 Hello, everyone. Welcome to our 2016 Student Invited Speaker Seminar from the IPIV department. It's my pleasure to introduce Professor Caroline Britosi from Stanford University and Howard Hughes Medical Institute. Professor Britosi studied chemistry at Harvard University as an undergrad, then moved to the West Coast to do a PhD at Berkeley with Mark Bynarski working on oligosaccharide synthesis. From there, she moved to UCSF to do postdoc work with Stephen Rosen looking at the interactions between selectins and saccharides and the role in inflammation. Caroline has far too many awards and accolades for us to name here, so I'll just highlight a few notable ones. She's membership in the National Academy of Sciences as well as the American Academy of Arts and Sciences and HHMI. She's received a number of awards, including the Lemelson MIT Award for Innovators, the MacArthur Genius Grant, and the Arthur Cope Award, which is presented by the American Chemical Society. It's the highest honor in organic chemistry. And in addition to these scientific achievements, Professor Britosi has mentored a number of PhD students and postdocs who have gone on to be very successful in their own research groups. And I've had the opportunity to meet a number of Britosi alumni now at conferences and meetings. And they'll just have such great things to say about Caroline and their faces kind of light up when they talk about their time in the Britosi group. And so I think it seems especially appropriate for a student-invited speaker to have such a great reputation among her own students as an advisor and mentor, in addition to scientific accomplishments. And we look forward to learning about science in the Britosi lab today in the area of glyco-calix engineering. Thank you. It's always fun to come to Wisconsin, even more fun when the students are hosting you, with all due respect to the faculty. And I've really had a wonderful day and had a chance to meet some of the students in postdocs and even a chance to meet undergraduates. I'm trying to convince to apply to our graduate program. So that was fun. And I thought what I would do is talk about a project area in the lab that it started about, I mean, in earnest, maybe about six years ago, and has kind of progressed through some stages of basic science questions to the point where we have some ideas for therapeutic interventions that map into the general space of cancer immune therapy. And so am I on? Can you hear me? OK. So I wasn't telling you to be quiet, actually. I realized that when this picture went up, everyone all of a sudden got quiet. And I was like, I think you thought I was telling you to be quiet. But I wasn't. This is just an image. It's Halloween month and so on. And what I'm going to talk about is the fact that we've known for some time. In fact, when I say we, I mean, there's 60 years of researchers have known for that duration of time that cancer cells are cloaked in sialic acid. And it hasn't been clear until more recently why that is and what that means and what we could potentially do about it. And so that's what this is. This is sort of like the bad guy is cloaked in sialic acid. You're going to remember that. But then the real title of this talk is Targeting the Cancer Glycocalyx for Immune Therapy. And so I think the notion of harnessing the power of the immune system to fight cancer is one that has now become kind of a household concept. There's been a lot of excitement in the last two years because of some very high profile clinical successes with drugs that are able to do just that. And it's been hard to avoid reading about this in the news outlets because, first of all, the successes in cancer immune therapy have certainly made it to the covers of all the journals that the scientists will recognize. Even was awarded the Breakthrough of the Year by Science Magazine. This was now about two years ago. People are writing books about a potential cure for cancer as coming from this immune therapy area. If you're in line at the supermarket, you might notice that Newsweek is writing about curing cancer. And Time Magazine, true to form, has kind of turned immune therapy into like a Game of Thrones episode or something inside the brutally selective, hugely expensive, life-saving trials of immunotherapy. Who doesn't want to read about that? And then I think new scientists kind of got there first with this very bold and provocative statement that immune therapy may well be cancer's penicillin moment. And that's a very strong statement when you think about what that means. And so what you might ask has prompted people to make such bold and optimistic statements about what immune therapy might do in the war against cancer. And I think there were a couple of very famous people who were literally kind of pulled from the grave by immune therapies. And the one who's most recognizable here in our country is former president Jimmy Carter. So you might recall that president Carter had a diagnosis of metastatic melanoma, which was in his brain. And before immune therapies were on the market, that was inevitably a death sentence. And his doctors gave him a few months to live. And this is a time when many people start saying they're goodbyes. But he had the fortune of, first of all, getting that diagnosis at a time when Merck had a recent approval for an immune therapy. And he was able to go on that drug. And then he had the second good fortune of being one of the roughly 20% of metastatic melanoma patients who actually respond to that drug. And he had what would be considered a durable response, which means that now, a few years later, he's essentially cancer-free. He's in remission and may even have been cured, which is quite a turnaround from the situation for people with that diagnosis just a few years ago. So that's what's got everybody really excited. Because even though the immune therapies that are out there are not curing even a majority of the patients that are being treated, what they are doing is taking a small group of patients and giving them life. So it's quite stunning. So in case you're not familiar with the backstory on these immune therapies, I thought it would just take a few minutes to summarize what's been happening. And so far, most of the focus has been around activating an immune cell called a T cell, which is one of many different types of immune cells in the body. So T cells, like all immune cells, have the job of going around through your circulation and looking for signs of cells that have gone bad. And the way that they test whether a cell has gone bad is they form a synapse with that cell. So let's say here it is, and it's a tumor cell. And when cells have gone bad, there are usually indicators of that on cell surface molecules. So for example, a tumor cell might have a protein called an MHC molecule that recognizes and presents a peptide outside the cell. And that peptide might have mutations in it that reflect cancerous mutations. Or that peptide may be much more abundant than it would normally be on a healthy cell because it's a cancer cell. So those would be signatures that the cell had gone bad. And so the T cell has a receptor that can recognize that signal. And that's the T cell receptor. And when the T cell receptor is engaged at this synapse, and usually it's multiple copies of the T cell receptor, multiple ligands, that activates the T cell. And if it's a killer T cell, it'll kill that tumor target. Now, T cells also have receptors that they can bring to the synapse that would oppose the activating signal through the T cell receptor. And so the two that are now famous in the immune therapy world are proteins called PD1 and CTLA4. And these are proteins that can also engage ligands on target cells. And their corresponding ligands are called PDL1 and B7. And these two proteins are normally considered signatures of a healthy cell. So a healthy cell will have these kinds of proteins. And so when they engage these receptors, that delivers a negative signal to the T cell, which basically counteracts any activating signal through the T cell receptor. And that will inhibit the T cell. And that's important because we don't want T cells just killing good cells, only bad cells. So the T cell has this calculation at the synapse to determine whether the good is better than the bad, or the bad is better than the good. So now because of this system, it might not surprise you to learn that successful tumors that have been able to sustain themselves and grow and expand in your body have often gained the ability to express these inactivating ligands. Because that is a phenotype that would be selected for under the pressure of T cells to find bad cells and kill them. So many tumor cells, and particularly these melanomas, they're known often to be very high expressors of PDL1 or B7. And that's how they can basically turn off the T cell, avoid being killed, and continue to thrive. So when all of this immunology was worked out, people in the pharmaceutical industry started thinking about ways that they could intervene in this calculation to tip the scales more toward activation of the T cell. So there's been a lot of basic science focusing on this synapse, including a lot of work in understanding what is happening inside the T cell when these inhibitory and activating receptors are clustering together. And how does the calculation take place? And so here's a cartoon from a review article that I borrowed to illustrate this where now the tumor target cell is up top and the T cell is on the bottom. And so here's the PD1 receptor on the T cell. Here's the ligand on the tumor cell. And when PD1 is clustered at the synapse between these two cells, it is able to recruit in the cytosolic phase a family of phosphatases that are called SHP1, SHP2, and together they're the ship phosphatases. And they get recruited by virtue of a sequence of amino acids referred to as an item motif. And that's an acronym. And it stands for immunoreceptor tyrosine-based inhibitory motif. And the item motif and the binding of that motif to these phosphatases, that's what allows PD1 to turn off activation of the T cell. Because the activating signals coming from the T cell receptor are often mediated through kinases that are phosphorylating things in a signaling pathway. And these phosphatases undo that. They take off the phosphates that the kinases put on. So it's kind of like a battle cytosolically between the phosphatases and the kinase. And then the winner of that battle governs the response of the T cell. All right. So what the pharmaceutical industry has been doing is figuring out how to block interactions between PD1, PDL1, or CTLA4, and B7. And in the case of PD1, where there's this item domain, if you can prevent the recruitment of PD1 to the synapse, then you prevent recruitment of the phosphatases. So that's basically the biochemistry. And that has now been done in a variety of ways, reflecting a collection of now clinically available and approved drugs. And it started with Bristol-Meyers-Squibb, development of a monoclonal antibody now called urevoi, which binds CTLA4 in a way that prevents its interaction with B7. And then just around the time that urevoi was approved through the pipeline came Merck's drug Ketruda. And Ketruda is the drug that Jimmy Carter was on. And that's a monoclonal antibody that blocks PD1 and prevents its interaction with PDL1. And then BMS also developed their own anti-PD1, which is now called OPDIVO, which is basically indistinguishable from Ketruda in its properties. And then just about two months ago, there was another approval, this time from Genentech. And they developed a monoclonal called tessentric, which binds PDL1 and blocks the interaction from the opposite face of the synapse. And now folks are looking at combinations of these things, like this antibody with this one, or this one with this one, and so on and so forth. So these are the drugs that are now considered to be the immune checkpoint inhibitors. And this is what everybody's been talking about in the news outlet. But this is far from the only important synapse in the immunology of tumors. T cells are just one of many different kinds of immune cells that can all contribute to recognition of cells gone bad and to elimination of those cells. And so now the industry is kind of reaching out into these other immune synapses, looking for other targets. And that's a very hot area of research, both in academic labs as well as in companies. So against this backdrop, we found it interesting that there was this extensive body of literature correlating changes in self-surface glycosylation with cancer. And the literature in this area can be traced all the way back to the 1960s, as far as what I've been able to find. And these are papers that were published even before I was born, long before I was born. Not long enough. And so over the years, there's this general observation that if you inventory the collection of glycans that are on cell surface glycoproteins and glycolipids from healthy cells and compare that pattern to what you find on cancerous cells derived from that same tissue, there are some striking differences and often very stereotypical tissue-specific differences. And it's an extensive body of literature in which these differences have been noted. But myself, I think of it as also a rather frustrating body of literature because it's almost entirely correlative without a whole lot of mechanistic insight into why these changes have occurred and what that might mean for the biology of cancer. So we know, for example, that many different tumor types from different tissues of origin often arrive at glycophenotypes of these varieties. Hypercyalylation is the one that caught our attention first and foremost. So many tumors have been characterized as having kind of higher density of glycan structures terminating in this blue sugar sialic acid. And there are different structures on different tumors and sometimes combinations of structures on individual tumors. But from a distance taken collectively, there is this phenotype which recurs across really the world of oncology. And then there's another phenotype which is very common on solid tumors. And that is the overexpression of glycoproteins called musins. And the definition of a musin is a glycoprotein that has a long stretch of amino acid residues with lots of serine and threonines that are glycosylated in dense clusters. So they're very highly glycosylated proteins. And the density of those sugars forces the polypeptide backbone to extend and rigidify, which is why I draw the cartoon like a big hairbrush. And these molecules really do project off the surface of cells like they tower hundreds of nanometers above the cell surface, much taller than the garden variety proteins in the rest of their environment. And sometimes these two phenotypes go together because if these glycans on the musin are sialylated glycans, well, that just contributes to their overpopulation on the surface of these tumors. And again, the literature commenting on this is more than half a century old. So now, in the modern era, having an understanding of the role of the immune system in keeping tumors at bay and also the role of immunosuppression in allowing tumors to thrive, there is another lens by which one can now understand these glycol phenotypes. And the connection came together after the Human Genome Sequencing Project revealed that us humans have a large family of sialic acid binding receptors on our immune cells. So this family is known as the siglex, and that's an acronym that stands for sialic acid binding immunoglobulin superfamily lectin-like receptors. They all have a related domain architecture where there's a sialic acid binding module. It's called a viset domain, and it's the little crescent moon shape here. And then they have a number of Ig domain repeats, a single span transmembrane domain, and then of the 14 siglex in human, nine of them in the cytosol have that famous item motif. In other words, cytosolically, many of the siglex look very much like PD1, that T cell immunosuppressive receptor. And in fact, we know at least for some of the siglex that have been directly characterized that they also play a role in down-modulating the activity of immune cells. The best studied of these, by far, is siglex 2, which is also called CD22. And we know from the work, founding work, really, of Jim Paulson, that when CD22 is engaged at the immune synapse between B cells and their targets, it down-modulates reactivity because it opposes signaling through the B cell receptor. And many of you know here at University of Wisconsin that in Laura Kiesling's lab, work has shown that you can harness CD22 as a mediator of immune modulation using synthetic polymers that can bind and cluster that receptor. Well, CD22, cytosolically, is almost indistinguishable from many other members of this family, including siglex that are on immune cells that are known to be important in protection against cancer. And that includes CD, we'll call it siglex 5, on dendritic cells, siglex 7 on NK cells, siglex 8 on eosinophils and mast cells, and siglex 9 on macrophages and also on neutrophils. So virtually every immune cell has a receptor that functions analogously to PD1. But on the outside, these receptors bind sialic acid. So now from that perspective, I think it makes perfect sense that successful tumors that have been able to avoid recognition by the immune system may be hyper-sialylated. That phenotype might allow them to recruit siglex to immune synapses and to basically hypnotize the immune cell and put it to sleep and avoid activation. And so this was the question that we sought out to address experimentally when we started working on this. And it's a question whose answer I think in nowadays is sort of hiding in plain sight. But at the time when we started this, it wasn't quite so clear. Do the siglex function as checkpoint receptors in the context of tumor immune evasion? And if they do, one might be able to target them with drugs just as Merck and BMS and Genentech have done for the T cell checkpoint receptors PD1 and CTLA-4. So we started with a focus on the NK cell. And that's because it is known in the literature both in humans and in mouse that NK cells play an important role in anti-tumor immunity. If you deplete NK cells from animals, tumors tend to grow more vigorously. And if NK cells are compromised in human, same thing. We're more prone to malignancies. And just like T cells, NK cells will form a synapse with tumor targets. And they have a collection of activating receptors and also inhibitory receptors that if they're engaged at the synapse, participate in this calculation. And just like the T cell, if they have more activation than inhibition, they get stimulated. And what they do is they degranulate, spit out these little vesicles that have membrane-poor forming proteins like perforin, reactive oxygen species, and lytic enzymes, and just annihilate the cell that they're touching. They're quite vicious, actually, if you see videos of this in action. And so the first question we asked was, if the tumor target is hypercyalylated, does that phenotype offer protection against the direct killing of that tumor by the NK cell? And so in cartoon form, the question was, if there are cyalylated glycans on that tumor target at sufficient density, can they recruit the NK cell siglec, which is siglec 7, to that synapse, bringing along its ITIM domain, and then would this constellation lead to inhibition of the NK cell and protection of the tumor? So to test this experimentally, we took advantage of some chemistry that we had developed for, quite honestly, other projects in the lab. And it was the use of synthetic glycopolymers that emulate cell surface musins. And we had developed a method to make these glycopolymers where we could functionalize the two ends independently and install a membrane anchoring group on one side, an optical probe on the other. And this anchoring group was just a phospholipid tail. And we found that if we just add solutions of these to cultured cells in the media, just through mass action, those lipids would insert into the plasma membrane. And we could decorate the cells with these glycopolymers at densities that we control simply by the dose and that we can quantitate by just taking advantage of the optical probe. And so the chemistry here basically allowed us to make molecules that are every bit as long as natural cell surface musins, but with a structural modification that made them infinitely more accessible, synthetically. So I'm showing now the actual chemical structure that you would see if you zeroed in on the backbone of one of these musins, where there are serine and threonine residues glycosylated on their oxygens with a sugar and acetyl galactosamine, which then can be elaborated in a variety of ways. And we know from some NMR structural work and some synthetic model work from earlier decades that having these things in proximity with this kind of spatial arrangement is what basically forces the backbone into this extended conformation. So to emulate that with molecules that were more accessible synthetically, we designed what we call glycodomane memetics that rather than the polypeptide, they have a polyalkane backbone. And rather than these glycosidic linkages to amino acid side chains, we attached sugars to those backbones simply through oxime forming reactions, which are simple, operationally simple, and chemistries you can do in water on just free sugars that you've made in the lab. So this simplifies the synthesis to just making a polymer based on methylvinyl ketone and then separately building glycans that have amino oxy groups at the reducing end and then just mixing these together. So in the forward sense, we used a polymerization method called raft, which is a radical living polymerization method with very good control over polymer length. Or in other words, we can make these with very low poly dispersity indices. And we build the phospholipid tail directly into the initiator. So that's the canonical thiocarbonate raft reagent. And we build the poly-MVK directly off of this initiator at lengths that we select based on the ratio of the monomer in the initiator. And then after building the polymer, we simply combine these with the amino oxy sugar, which forms oxime linkages. And at the same time, nucleophilically cleaves this thiocarbonate to liberate the free self-hydro group. And that, of course, can be conjugated with your probe of interest through malamide chemistry to make the molecules that we use. And we characterize these using the usual tools, but it's fun to kind of show this TEM image because you can see these individual polymer molecules really do look like rods. They're quite rigidified because of the spacing of those sugars. So through that method, we put together a panel of glycopolomers in which the glycan structures reflected various different sialylated glycans that had been associated with cancers. Simple things like a monosaccharide of sialic acid, disaccharides like the prostate-associated antigen sialyl-TN, more complicated structures like this GD3 structure, which has been found in abundance on renal cell carcinomas and certain breast cancers, sialylated Lewis antigens, and then a couple of control structures that had no sialic acid. And with these glycopolomers, we did some in vitro cell culture assays to see to what extent NK cells could kill tumor cells that had been decorated with various densities of these molecules. So we took the target cells, incubated them with the glycopolomers to decorate the surface, washed them off, get rid of all the unbound polymers, and then add a suspension of NK cells from freshly drawn blood. And then you kind of rock these cells together for a few hours, and they form synapses, and then things happen. And either they kill the cells or they don't. You can quantitate this. And I borrowed this from Google Images because it's a reminder that this is not what the synapse actually looks like. It's not like two balls bouncing off of each other. This is the NK cell, and it's almost like reaching out with a tongue and tasting the cell to determine what to do next. I like to think of it that way. OK. So here was the outcome of those cell killing experiments. And what's plotted here is the percent inhibition of killing achieved by these various decorated glycopolomers, starting with the situation in which the tumor targets had no polymer. So that's kind of our baseline. And the targets for this experiment were the T-cell acute lymphoblastic leukemia cell line called the jercat cell. And so this is sort of where we start. But if those jercat cells were coated with a glycopolomer just with the monosaccharide cialic acid, there was substantial protection from killing. And that was also true for any glycopolomer, essentially, that had a terminal cialic acid group. And there were some subtle differences in the protection of different structures. But for the most part, all of them were protective. And by contrast, glycopolomers with unsyallylated sugars had no significant protection. So that was interesting to see. But does this have anything to do with engagement of siglex 7 on the N-K cell? And to address that, we did experiments with commercial antibodies, their mouse monoclonals, that bind siglex and block their carbohydrate binding. So these are function-blocking anti-siglex antibodies. And that's what these data are. So now we're plotting the percent cytotoxicity. So it's kind of the inverse of the top. So the jercat cells, when they're not coated with a cialylated polymer, there's a certain amount of killing. But when they are coated with that cialylated polymer, they're protected. But now if we add an antibody against siglex 7, the protection over here is kind of reversed. So now they're susceptible, again, to killing when siglex 7 is inactivated. By contrast, an antibody to the related siglex 9, which is not found on most N-K cells, but is prominent on macrophages, there was no effect. And likewise, an isotype control antibody had no effect. So together, we can conclude from these data that the cialicide polymers protect the jercat cells from N-K cell killing. And they do that in a manner that depends on siglex 7. Now our model for this is that the siglex 7 signaling would be, of course, occurring at the synapse between the two cells. So we wanted to test whether siglex 7 was actually being recruited. And we did this using immunofluorescence microscopy. So this is an actual synapse formed in cell culture between the jercat cell and that freshly isolated N-K cell. And this is the circumstance where the jercat cell is coated with a cialylated glycopolymer. So it's protected. And you'll note, when we stain that same doublet with an antibody against siglex 7, you can see that siglex 7 is concentrated. There's more intensity right at the synapse between the two cells. And this is just a false color overlay showing siglex 7 staining in red. And the green is just our glycopolymer. But by contrast, when the jercat cells were not coated with a glycopolymer, so this is the circumstance where they will get killed, those synapses form. But siglex 7 is uniformly distributed. It's not recruited specifically to the synapse. And here's the false color overlay. So again, this is consistent with the idea that the hyper-cylylated state allows recruitment of a siglex that then inhibits N-K cell activation. So this gave us confidence to kind of keep going and to think about how one might take advantage of this biology from the standpoint of therapy. And from that vantage point, it was interesting to us that N-K cells have a different mechanism for killing tumor cells that relies on antibodies. So in this way, N-K cells are often thought of as the bridge between innate immunity and adaptive immunity. So they can recognize cells gone bad just all by themselves. But if those cells are coated with an antibody, they really get excited. And that's because N-K cells have a very strongly activating receptor, which is called FC gamma R3. And it's an FC receptor that recognizes a portion of the conserved part of antibodies that's actually kind of up here near the hinge region. So the cartoon is not that accurate. That binding site is actually up here. But when FC gamma R3 is clustered at an immune synapse, N-K cells, they really fire off. And they will kill that cell quite aggressively. And that's not only thought to be important for our natural ability to prevent cancer from growing, but it's also an important mechanism of action for some very recognizable monoclonal antibody cancer drugs. So you might have heard of molecules like Rituxin or Rituximab and Erbitux and Herceptin. These are antibodies that bind targets that are known to be overexpressed on many cancers. And Herceptin is kind of the most famous. This is Genentech's big blockbuster cancer drug that recognizes a protein called Her2. And Her2 is a growth factor receptor that's overexpressed on a lot of breast cancers and also gastric cancers. And Herceptin will bind to that target, recruit N-K cells, and activate them to kill the Her2 positive target by virtue of FC gamma R3 signaling. So that's how these drugs work. Now Herceptin and the rest, they're famous because people think of them as landmarks of personalized medicine. So in the clinic, in practice, the way these drugs are prescribed is based on the outcome of a biopsy analysis. So some tissue is removed. It's analyzed for Her2 expression. And if it's Her2 positive, you get the drug. So the patients are selected based on a molecular signature of the tumor. But it's actually not that simple. So if you kind of look through the clinical literature on Herceptin, this is actually what the numbers look like. So 74% of breast cancer patient biopsies show Her2 positivity. So that sounds like a huge patient population. But the truth is, we know from the clinical trials and evaluation of Herceptin that only the patients with the highest levels of Her2, which is about the 20% of those patients, actually respond at a reasonable frequency. And so there are the patients that are scored as Her2 3 plus, whereas a patient that gets a Her2 2 plus or 1 plus score is generally not eligible for Herceptin because they have a very low likelihood of responding to Herceptin. So they don't even get the drug. But even the ones that do get the drug, not all of them respond. And it turns out only about 18% of them respond to Herceptin as a single agent. So when you add all these numbers together, the truth is that the vast majority of Her2 positive breast cancer patients are either unresponsive to or ineligible for Herceptin treatment, which actually is kind of the same numbers with the checkpoint inhibitors. So the PD1 antibody, same thing. About 20% of eligible patients respond. And the other 80% don't. And nobody knows why. And it's the same with Herceptin. 80% don't respond. We don't really know why. And there's been a lot of work trying to figure out why, as you could imagine. But throughout that vast body of literature, nowhere can you find anybody actually considering a role for the glycol calyx. So we thought this would be interesting to model experimentally as well. Or in other words, if you have a cancer cell that's Her2 positive able to bind Herceptin and to recruit N-case cells through FC gamma R3, would hypercyylylation and recruitment of siglex 7 undermine that activation pathway and actually afford resistance to Herceptin? Could hypercyylylation actually be a drug resistance mechanism? So we modeled this in virtually the same way as the previous experiments, this time using a Her2 positive gastric cancer cell line called NCIN87. And so if you combine these Her2 positive cells with N-case cells and Herceptin, you see a very potent killing response as a function of the N-case cell to target ratio. So Herceptin is very effective against these cells in vitro. But if first you coat the NCIN87 cells with a sialylated polymer, all of a sudden now they're not so susceptible to killing via Herceptin and the N-case cells. They're protected from that, whereas control polymers that don't bind siglex offer no such protection. So at least in this cell culture model, it is possible for sialylation and siglex 7 engagement to override the activation signal through FC gamma R3. So now, of course, there are some implications of this from the clinical standpoint. If this is actually happening in a clinical context, it's quite obvious how you might intervene. And you would do that just the same way that we are now intervening in the engagement of PD1 with PDL1, same exact problem. You could imagine blocking the siglec with a siglec blocking antibody quite analogous to ketruda or optivo, those two drugs. And from a therapeutic standpoint, you would want this to be a human antibody or a humanized antibody. And so that's a concept that is under way, and we're going to test that. But somewhat more interesting from my standpoint, for reasons I won't elaborate on now, I could talk about it later, it seemed to me that a more potent intervention would be to target the ligands for siglec 7. And it's a little less clear how to do that analogously to blocking PDL1 because the ligands for siglec 7 have actually not been characterized in molecular detail, at least in the in vivo context. But we do know that they are sialylated glycans. And maybe they're sialylated glycans of lots of different structures. Maybe they're sialylated glycans on one glycoprotein. Maybe they're on lots of different glycoproteins. We don't know. But they absolutely depend on sialic acid. And so we thought maybe we could intervene by targeting enzymes to the surface of the cancer cells that are able to cut those sialic acids off locally. And those enzymes are called sialidases. And sialidases are actually found in many different organisms, including humans. Lots of microbes make them too. And so that became a more recent direction of the project was can we make chimeric molecules that have antibody tumor targeting capacity, like Herceptin, able to activate FC gamma R3, but also adorned with a sialidase. They can strip the sialicides from the synapse. OK. So here are some examples of sialidases that we've been working with. And some of these are actually commercially available. Others you still have to make yourself. But for example, here's one that's off the shelf from Vibrio cholera. And it's really a gorgeous molecule, a rather large molecule. It's got a catalytic domain in the core. And then it's got these two ancillary domains that are thought to serve kind of a lectin-like function. And it's a very efficient enzyme for stripping sialicides off of cells, because it can bind cell surface sialicides, actually, through these lectin domains while it chops off sialic acids with its catalytic domain. And then another bacterium, Salmonella, has a sialidase, which is kind of like the catalytic core missing these two other domains. So it's a much smaller protein that simply has that sialic acid cleavage function. And then we humans have four putative sialidases, the best characterized being a protein nu2, which is quite related structurally to the Salmonella enzyme. And then Clostridia has one, which has one small ancillary domain. And we, for various reasons, took a look at all of these. And where we started was with the Vibrio sialidase simply because that was the first one for which we had a very productive expression system up and running. But as you'll see later, it's probably not the best one. But it is where we started. So we set out to make a chimeric molecule, let's call it Herceptin sialidase version 1.0, which was a gigantic beast of a molecule, 313 kilodaltons, where we situated two sialidase units right near the C terminus of the heavy chain of the monoclonal antibody. And we chose this very end of the heavy chain because, as I mentioned earlier, our ambition was that this molecule would also be able to engage FC gamma R3 at the synapse and initiate the activating signal. And that epitope has been mapped to a region that's up here near what's called CH1. That's the number one conserved region of the heavy chain, as well as a kind of part of the hinge region. So this was about as far away as we could conceive of positioning this kind of mutation compared to where the biology would happen. Now to make these fusions, we took advantage of some chemistry that we had developed in the context of making antibody drug conjugates, which are chemical fusions between an antibody and a small molecule. And now our hope was to use the same approach to make chemical fusions between the antibody and the enzyme. So the key technology is what we refer to as the aldehyde tag method for site-specific protein modification. And there's a whole story here as well that I won't tell. But in a nutshell, if you clone this short consensus motif into a protein of interest, it can be recognized by an enzyme called FGE, which stands for Formyl Glycine Generating Enzyme. And that enzyme will oxidize the cysteine side chain, convert that thiol to an aldehyde. And all of this can happen inside the cell that is expressing the monoclonal antibody, like the cho cell. So we are able to isolate proteins that have aldehydes at a site encoded by this motif. And then we do chemistry on those aldehydes. And it can be really straight up old school bioconjugation chemistry, like condensing aldehydes with amino oxy groups to make oxymes. And so we just adapted that process to the synthesis of these conjugates. And I have a shameless plug here because I'm in Wisconsin. And there's a company out here now that has commercialized this. They call it the smart tag technology. And Catalan has a big site just outside of Madison. And they're hiring, by the way, if you're looking for jobs. And one of you works there already. I think someone from Ron Raines' group has a job there now. OK. All right, so we clone that aldehyde tag near the C-terminus of Herceptin's heavy chain. And then we condensed that aldehyde tagged protein with a linker through the oxyme chemistry. And that allowed us to kind of translate the aldehyde into an azide. And then in parallel, we took that vibrio cholera sialidase and through nonspecific lysine modification chemistry, so kind of heterogeneous modification, we introduced a complementary group, a short linker with a cycle octine. And azides and cycle octines will react selectively with each other. And that we call copper-free click chemistry to make this conjugated product where, because of the symmetry, there are now two sialidases, one on each heavy chain through that chemical linker. OK, so we made that. And so the first question we asked was, will this conjugate take the sialic acids off of her two cells selectively, her two positive cells? So to judge that selectivity, we did a cell culture experiment combining two different cell types, one her two negative and the other her two positive, mixed together. And then we treated them as a cult culture to determine whether we could take the sialic acids off selectively from the her two positive cells and to identify a therapeutic window or a dose that would allow that to happen. So this just kind of sets the stage. Here's a field of cells. And you can't tell which is which just by looking at the DAPI staining in the phase image. But when you stain them with her two antibody, you can see that these two are her two positive, which is these two. This is the same field. And SNA is a lectin that binds sialic acid. So you can see all of them are sialylated. But only these two are her two positive. And here's the merge. And we can analyze a whole population of these cells by flow cytometry. So this is the her two axis. And this is the sialic acid axis. And you can see these are her two negative, her two positive. They both have roughly the same sialic acid density. So now, what happens when you treat that co-culture with, let's call it, a low dose of her septin sialidase conjugate? Here's a field of cells. We don't know which is which. But you can see that these are the her two positives. That's these two. And now, look at the sialic acid channel. They've been stripped of their sialicides, whereas the her two negatives are still sialylated. And here's an overlay. And here's the flow data. So again, this was the her two positive cell population. Their sialylation level dropped. But the her two negatives are largely the same. So we conclude that that is a dose, six nanomolar, where we can selectively desialylate the her two positives in the presence of the her two negatives. But as you might expect, when you crank up the dose 10 fold, you reach a point where there's off-target desialylation. And remember, this vibrio-color sialidase has those lectin domains, so it has another way to bind to the cell surface, independent of her two. And that's playing out here. So at this dose, we can just go right to the flow data. You can see that the her two positives, which started here, have dropped their sialic acids. But now the her two negatives are starting to drop down as well. So we consider this to be a dose where we can achieve selectivity, but this is the limit where there's off-target. So knowing that, this allowed us to kind of pick our dosing range around which to test the efficacy of these antibodies in mediating NK cell killing of the her two positive cells. So what we're doing now is basically comparing the cytotoxicity as a function of dose for naked herceptin and herceptin sialidase. And this is a cell line, ZR751, which would be considered low her two. These are her two one plus. Same with BT20. They're very similar. They have roughly the same levels of her two. They're breast cancer cell lines that are her two one plus. And in this circumstance, you can see quite clearly that while at high doses, herceptin can basically deliver some killing by the NK cell, but it's significantly potentiated when the herceptin is dragging a sialidase. You get a big increase in cytotoxicity, which is exactly what we were hoping to achieve. Interestingly, and also somewhat predictably, when you do the same experiment with high her two expressors, like the SKBR3 breast cancer cell line, which is her two three plus, now there really is no difference in the killing activity of herceptin alone or with the sialidase. They're equal potent and both quite potent. And that makes sense when you think about what's happening at the synapse. Because cells that are high in her two can bind a lot of herceptin, cluster a lot of FC gamma R3, and deliver a very strong activating signal. And whether or not Siglex 7 is engaged, it might not matter, because the activation is so strong. Who cares if you eliminate the Siglex from the synapse? And this is another control, her two negatives, where there's no toxicity, just to show that it is a her two dependent phenomenon. So these data are consistent with a model that you can kind of think of as this seesaw or a balance. So the conditions where her two is high, those are conditions where there'll be a very strong activating signal through the FC receptor. And there might be an inhibitory signal through the Siglex, but it's not overpowering this activating signal. So the seesaw tilts towards an activated, angry NK cell that will kill. And you can bring a sialidase to the synapse and eliminate signaling through the Siglex, but who cares? Because you're already leaning towards activation. It doesn't matter. It's in that situation where her two is low, that you could have a huge effect. Because when her two is low, and there's a weak activating signal, an inhibitory signal through Siglex 7 could dominate the synapse, and the NK cells don't kill. And if you take that away, now the seesaw tilts the other way, and you can get the NK cell to react. So we are most excited about the capability of this as a potential intervention for her two low patients, the kind that don't respond to her septin. And that's why they don't even get the drug. But maybe they would respond if you could strip off the sialic acids or block the Siglex. So that's kind of where we're going with this. But there's a lot of room for improvements. And the two main liabilities of that version 1.0 were, first of all, the massive size of this protein, which is 84 kilodaltons, put two of them on 150 kilodalton monoclonal, and you're above 300 KD. That's gigantic. Very difficult to think of manufacturing that thing. Also, these lectin domains are surely contributing to the off-target reactivity of this enzyme on sial-related cells. So I like to think of it as kind of like having a Mr. Pac-Man. Don't you see that? Look at that. It's like two hands in the mouth, right in the middle. So that's not good. But like I said, this is just one of many sialidases you could choose from. And once those biological data came through the pike, we were ever more incentivized to get a good expression system for this little one here, the salmonella enzyme. So that was the next thing that we cracked. We figured out how to make bucket loads of that protein. So now we basically went through a very similar process to generate what we call the Herceptin sialidase 2.0, which it's still pretty big, but at least it lost about 100 kilodaltons of weight. So that, I think, is going to be quite helpful. The other thing we did differently in this version 2.0 is we cleaned up the chemistry on the sialidase so that it is no longer kind of random modification of lysines, but now a site-specific modification of an engineered cysteine. So but otherwise, the chemistry was quite the same. We took our Herceptin, conjugated it to the linker with the azide. We put in a cysteine. We did some oleamide chemistry to introduce a cycle octine, clicked these two together, and made version 2.0. And last night, I got these data, so I didn't have time to make a pretty animated slide out of it. It's really beautiful. So these are all flow data, obviously. And here's 1.0. Here's 2.0. And same experiment by flow. So these are Her2 positive, Her2 negative. This is at the start before any treatment with anything. And now what you're looking at is different doses of the conjugate of 1.0, different doses of the conjugate of 2.0. And you can just jump to the highest dose here, because that's what matters. So version 1.0, a lot of off-target reactivity. The Her2 negatives are dropping. But version 2.0, even at 120 nanomolar, is essentially untouched. So I think we got rid of a lot of this off-target behavior, which is much better. And so 2.0 is now the rendition that is going into some animal models to see if we can't get more clinically relevant data sets. OK. Anyways, none of which relieves us of the burden of making probably the most important type of molecule, which is the 3.0. And that would be a genetic fusion between antibody and ideally a human cyanolidase. That's the kind of molecule where now you can talk about progressing this as a preclinical candidate. A human cyanolidase is less likely to be immunogenic than a human versus these bacterial cyanolidases. And a genetic fusion, of course, is a much more manufacturable molecule, because you can make it in one fermentation rather than making things and trying to chemically click them together. So that is also kind of on the docket for future work. And that's the story. And there are many people to thank. All of this started, I mean, really, you could kind of date it back to David Rabuka, who was instrumental in this aldehyde tag technology as a student. And now he works for Catalan, although Catalan biologics west out in the Bay Area, although he's here in Madison quite often. Camille Godula developed that raft polymerization method for making those glycopolymers. Jason Hudak and Steve Canem did all of that early proof of concept cell culture killing assay stuff with NK cells and Ciglex. And then Elliot Woods and Han Shao have generated these antibody cyanolidase conjugates and did all of this preliminary biological work. Han is looking for faculty jobs. And I think he submitted his application to Wisconsin here. So take a look. It's in the pile or in the ether or wherever they are right now. But he's absolutely fantastic. And there aren't that many postdocs where you can say, hey, why don't we make a 313 kilodalton protein through chemical conjugation? He's like, yeah, whatever. And he's passing that torch to new student, Melissa Gray. And she made this version 2.0. So she's got a really exciting road ahead with the animal work. And these are the people that funded the work. And thank you very much for the invitation. For a few questions before we move out to the atrium for our reception. And I'll let you moderate. Any takers? Yeah. Can Ciglex, the titan, and the devil in the details? What's the devil in the details? Yeah, so let's start with talking about the implications of all of those Ciglex, OK? There's a lot of them. And so in this regard, it's quite a bit more complicated than the PD1, PDL1 binary relationship, right? So yeah, there's nine of the 14 human Ciglex have the ITIM domain and could potentially be involved in immune suppression, although not all of them have been characterized specifically that way. But we've looked at a variety of different immune cells to figure out who's expressed where. And what I'm showing here is a gross oversimplification. So for example, NK cells, they all have Ciglex 7. But some of them have Ciglex 9. And a lot of them have Ciglex 3. We've looked at neutrophils, macrophages, eosinophils. And all of them have at least two, sometimes three, Ciglex. Even T cells, which nobody finds Ciglex on T cells when they're quiescent, but when they're activated and exhausted, the Ciglex 8 comes up. So the problem is, how many Ciglex do you need to block in order to see an effect? And it might be more than one. And that makes the development, the therapeutic process, much more difficult. By contrast, though, if you just target the sialic acid with these sialidases, you inhibit all of them. That's kind of why we decided to go this route, rather than just plugging in antibodies to Ciglex. And the sialidases we're using are very broad spectrum. So they are able to strip pretty much any type of sialic acid, irrespective of the details of the linkage and what's underneath it. That could be good or bad. The potential bad consequence would be autoimmune side effects. It's a very obnoxious thing to do to a cell to strip off its sialic acids. Not only does it open it up to these immune pathways without Ciglex engagement, but complement is inhibited by sialicides. So that's how our blood cells protect themselves from complement-mediated lysis. And if you strip off sialic acids from blood cells, they get laced. Well, that's not good, unless it's in lymphoma. And then it's great. So I think it is a complicated axis of immune modulation to be tinkering with. But we're an academic lab, so why not? Sam, yeah, towards the end there. This plateau? Yeah, that's a good question. This is kind of always what you see. And it's the same with the checkpoint inhibitors. They reach a point where you just can't get beyond a certain level. It might be because this is going on for, depending on the exact experiment, but in these experiments, this is like a four-hour killing study. And there's turnover during that time. So it's not like you're not really at equilibrium. And HER2 is an internalizing antibody. So you're actually consuming the HER2. And presumably, you're consuming this as well. So probably the concentrations are not static. And it's a pretty high density of cells that we use to get good error bars on these. So it might relate to that. But that's a hand-waving explanation. I don't really know. Yeah, so we're trying to stay within a dose range where we're not going to get too much off-target with this thing. But yeah, you can get more killing with higher doses of either end. You can have different NK cells to target ratios. And you can boost the killing with more NK cells. You can boost killing with higher doses of antibody. But you do saturate after a while. And you can then have more NK cells and alleviate the saturation. That's possible. I mean, this is a clonal cell line. But having said that, there is heterogeneity in the HER2. We can see that in the histogram and the flocetometer. And the NK cells, they're not like clonal. These are isolated from blood. So that's a heterogeneous population as well. And some of them might be much more active than others. And when the active ones exhaust themselves, you might have NK cells that are just not doing anything in there. It's just a good question. There's a lot of variables that would hesitate to speculate. Are there glycoforms of HER2 that can form a resistance mechanism to her septum? Anything goes. And that's certainly possible. It's precedent, it's certainly, like you said, in more highly mutagenic organisms, like viruses. I can't think that anyone has even looked. But I know that there is some thought about the checkpoint inhibitors potentially being sensitive to glycoforms. I know BMS at least has at least articulated that in seminars that they've given. So yeah, I mean, it's entirely possible. And once again, when I think of the literature on HER2 resistance, it's absolutely hyper-focused on cytosolic biochemistry. Everyone's looking at, oh, is it loss of PTAN? Is there some signaling defect in the EGFR kinase? What's all downstream of EGF signaling? Is there now HER1? Has HER1 come up? It's all really focused obsessively, disproportionately focused on cytosolic signaling pathways. That it's an extracellular drug. You see what you're looking for, right? If people shine a spotlight over there, that's what they're looking for. And then the answer's over there, and no one even looks. That's biology. It's so big. It's just so big. And people don't know where to look.