Founder Stories: Laura Walker, Generation of Neutralizing Antibodies for Pandemics

Okay, it's just 10 o'clock.

We'll get started.

Hello, and thank you to everyone around the world for joining us for this week's Global COVID Lab meeting. My name is Kristin Aboud. I am the science editor at the Human Vaccines Project, and I will be your moderator today. The number of global COVID-19 cases is surging, largely due to rising infection rates in South America, and India. Over the weekend, India recorded more than 400,000 new cases a day, with the death toll surging in parallel. Other countries are helping to try to address the vast shortage of critical supplies in the country, sending oxygen, medications, and vaccines. Meanwhile, scientists around the world are assessing just how effective the existing COVID-19 vaccines are against circulating viral variants, which are contributing to new waves of infections globally and are likely to only increase in number as the virus continues to spread so extensively. In addition to vaccines, monoclonal antibodies, which are the subject of today's lab meeting, are another strategy to both treat and prevent COVID-19. In the U.S., combination monoclonal antibody products from Regeneron and Eli Lilly have received emergency authorization for use in individuals infected with SARS-CoV-2 who are at high risk of progressing to severe COVID-19 disease. And both companies have announced results from studies showing that their two antibody cocktails are also effective at preventing COVID infections altogether. We're going to hear much more about exciting antibody development from today's speaker. But before we get started with today's presentation, I wanted to note that the application portal is now open for the Michelson Prizes: Next Generation Grants. Presented by the Human Vaccines Project and the Michelson Medical Research Foundation, the Michelson Prizes are given to support promising young researchers who are applying disruptive concepts and innovative methodologies to advance human immunology vaccine discovery, and immunotherapy research for major global diseases. We're posting the link to the Michelson Prizes in the chat, so please visit the site for more information or to apply. Just one more note before we begin. The information presented today includes some pre-published data that is currently under peer review. At the request of the author, we ask that you not reproduce or disseminate the data presented. This session will be recorded and made available on our website and our social media channels for you to review. With that, I am happy to introduce today's speaker. Dr. Laura Walker is the Senior Director of Antibody Sciences at Adimab and the Chief Scientific Officer at Adagio Therapeutics. Her team at Adimab is focused on understanding human B-cell responses to to viral infection, and discovering broadly cross-reactive neutralizing antibodies to a wide variety of emerging viral pathogens, including Ebola, Zika, yellow fever, and SARS-CoV-2. At Adagio Therapeutics, Laura oversees a broad portfolio of research in the area of viral immunology, including the identification of potent neutralizing antibodies to advance into clinical development. Prior to joining Adimab in 2012, Laura earned her PhD in immunology and microbiology at the Scripps Research Institute and completed a postdoctoral research fellowship in immunology at the University of California, San Francisco. I've personally been following Dr. Walker's work since 2009, when she published a seminal paper on the identification of broadly neutralizing antibodies Against HIV. During the presentation, please send me your questions using the Q&A function in Zoom. I will ask our speaker a broad selection of your questions after the presentation, and we will have about 25 minutes for discussion. It is my pleasure to now introduce Dr. Laura Walker.

Thanks very much. So I guess I should share my screen now.

Yes, that would be perfect.

There we go. Can you see it?

Yes.

Okay, great. Well, I just wanted to start by thanking the organizers for inviting me to speak at this virtual event today. I'm going to talk about some of the work that my group has been doing over the past few years identifying and also engineering broadly neutralizing antibodies to to various different emerging viral pathogens. So I thought I would start today by kind of introducing this concept of rapid response to emerging infectious diseases and just talk a little bit about what I think are some of the major limitations of that type of approach, and then present an alternative approach. And this is one that my group has worked on for many years, which really relies on the identification of broadly cross-reactive neutralizing antibodies to variable viruses. And then just show a few slides on some of the work that we've done historically on this in the context of Ebola viruses, essentially as a proof of concept. And then finally talk about how we've recently applied this concept to SARS-like viruses. So as many of you, I'm sure, are aware, infectious diseases are now emerging or re-emerging almost every year. And this trend started about 2 decades now, in 2002, 2003 with the original SARS virus. And that virus has been followed by many other viruses, only some of which are listed here, including 2 novel pathogenic coronaviruses. MERS, which emerged in the Middle East in 2012 and is still endemic in the Middle East. And of course, SARS-CoV-2, which emerged at the end of 2019, as you all know, and is still persisting as a pandemic. So the challenge with emerging infectious diseases, at least from a vaccine perspective and from a therapeutic development perspective, is that in many cases, not always, but in many cases, these epidemics rise and fall relatively quickly. And so on the left, I'm just showing you some of the timelines for some of the outbreaks that I listed on the previous slide. And these, these timelines are kind of capturing the bulk of the outbreak. Of course, there are long tails on both ends. But the point I'm trying to make here is that in many cases, on average, these epidemics last on the order of months, not on the order of years, which of course means that you need to act very quickly. You need to develop vaccines and therapies within just a couple of months to really have a significant impact on an ongoing epidemic or to stop an epidemic from happening in the first place. And so even— oh, sorry, I should also— oops, my pointer. Unfortunately, what we've now seen happen over and over again, and I think that Rino Rappuoli did a very good job depicting this in a perspective piece that he published a few years ago in PNAS, what we've often seen is that a new virus emerges, like what we saw with Zika and Ebola and the original SARS virus and many others. There's a lot of attention in the media after the first cases are reported, and then there's a lot of money pumped into drug development and academia and biotech and pharma. And everybody's working on whatever particular virus it is. And then as he's showing here, it's sort of over the hump here. The epidemic ends either because of herd immunity or because of, you know, public health measures. And then he's showing the end of the money and the end of the project. And so in the end, a lot of value is lost in terms of, you know, drug development when we go through this cycle. And of course, we don't develop vaccines and therapies generally in time to stop these epidemics from happening in the first place, which is of course what you would want to do in an ideal situation. And so even for viruses that stick around for a long time or become endemic in the population, like what we are very likely to see happen with SARS-CoV-2, the problem isn't solved, as we've all seen play out over the past year or so. Although we've developed vaccines and therapies at an unprecedented pace, that obviously hasn't happened before, where a million people have died and all of society has, has shut down, etc. And so here I'm just showing you the timeline for the development of benlanivimab, which is a monoclonal antibody. This was the first monoclonal antibody that was approved under an EUA, developed by Eli Lilly. That antibody was developed in 8 months. So again, an unprecedented speed. That antibody has now been abandoned due to the emergence of resistance, which I'll talk about later. And then here are the timelines for the development of the first two mRNA vaccines from Pfizer and from Moderna. So it took them about a year. And so just trying to highlight here that although these vaccines and therapies were developed extraordinarily quickly, they weren't developed fast enough to stop, you know, this pandemic from happening in the first place. And again, that's what you would want to do in an ideal situation. And so I'm going to present the argument that an alternative approach, and this is what I'm calling a preemptive approach, might be a better solution to this problem. And so for this approach, really, it relies, as I mentioned at the beginning, on the generation of broad and potent neutralizing antibodies against known viral pathogens. And so these are antibodies that are either cross-reactive among different strains of the same virus. You can think about HIV or influenza, or even different families of viruses like dengue and Zika, for example, or SARS-CoV-1 and SARS-CoV-2. And so if you could imagine, if we had developed an antibody after you know, the SARS outbreak in 2004, 2005, that was really pan-SARS, or even better, pan-SARS vaccine, we might not be in the same situation today. And so on the left, I'm just, I'm sort of highlighting this concept of broad neutralization using influenza hemagglutinin. So by definition, these broadly neutralizing antibodies are binding to conserved epitopes, epitopes that are shared across many different members of the family. And so this is just showing the structure of influenza HA. Obviously, this is a very variable virus. And here I'm coloring the structure based on conservation, and highlighting here is that there are regions even on very variable viruses like influenza or even HIV that are highly conserved. You can see the stem region here is very conserved. And there have, of course, been antibodies described that target this region that are broadly neutralizing, and these antibodies are now— some of them are being developed as therapeutics, and they're the basis for many universal flu vaccine efforts.. And so over the past 10 years or so, there have been many different broadly neutralizing antibodies described by us and many others to many different viruses. And that's largely been driven by advances in new technologies for the isolation of monoclonal antibodies from human B cells. And so there are really two leading technologies. One is single B cell cloning and sorting by flow cytometry. That's what we do at Adimab, and I'll talk a bit about that today.. And then the other technology I'm calling single B-cell activation and culture, where you essentially activate memory B-cells and then you screen those supernatants for binding or even for, for function. So these are the technologies that have enabled the discovery of these types of antibodies. So I'm just going to show a few slides now on some work that we've done historically identifying a pan-Ebola virus antibody cocktail, essentially as a proof of concept that you can actually develop antibodies that work not only against current viruses but also future versions. Of the virus as well. And so some of you probably know there are 6 different Ebola virus species that have been identified to date, and those are listed here. The ones that I'm showing in different colors are Ebola viruses that have caused outbreaks in humans, that are known to be pathogenic in humans. And then on the right is just showing a map of Africa depicting where these different outbreaks have occurred. And so this is, of course, the 2014 outbreak in West Africa. So when you hear about Ebola virus in the news, they're typically referring to Ebola Zaire, which is the species that has caused most of these outbreaks. So importantly, all anti-Ebola virus monoclonal antibodies, antibody cocktails, for example, mAb114, the Regeneron antibody cocktail, both of which have been approved now, as well as the Merck VSV vaccine, all of those vaccines and therapies are Ebola Zaire specific. And so what that means is if one of these other pathogenic Ebola viruses emerges next, or if we see the emergence of a novel Ebola virus, which I'll talk about at the end, we don't have any effective countermeasures. And so what you would want is a monoclonal antibody or antibody cocktail that works against all these different pathogenic Ebola viruses, as well as new emerging Ebola viruses and escape variants as well. And so that's what we set out to do in our study. And this was a few years ago. And I'm essentially summarizing about 5 years of work on a single slide here. But I think it makes the point. So essentially the workflow for the identification of a pan-Ebola virus antibody cocktail. So back in 2015, we were able to obtain a blood sample from a survivor of the 2014 Ebola Zaire outbreak. This was a single vial of PBMCs. So we sorted all of the B cells in that vial based on reactivity with Ebola GP. So GP is the glycoprotein that's expressed on the surface of the virus. And it's the only known target for neutralizing antibodies. And so we sorted out every B cell that showed reactivity with this protein. We screened those all for binding and for function and did a fair amount of work. I'm, of course, just— antibodies that showed binding to Ebola here. Actually cross-react and cross-neutralize these other Ebola virus species? And the answer to that question was 2. There were 2 antibodies out of the 349 that cross-neutralized Sudan virus and Bundibugyo virus, Sudan being only about 50% homologous to Ebola Zaire. So this really highlights the value of deep mining of antibody repertoires, because often— not always, but often— these broadly neutralizing antibodies are extraordinarily rare in antibody repertoires. And of course, that's because your immune system selects for affinity. For binding to the, you know, the virus that's infected you, not to future versions of the virus. Of course, there's no selection for breadth. So the other issue is that even if you can identify broadly neutralizing antibodies, it's often true that those antibodies have higher affinity and neutralization potency against the virus that the person was infected with and not the other related viruses. And so the way that we've solved this problem at Adimab for many different viruses, including this one, is we can take those antibodies and we can improve their specificity and their affinity to these other heterologous viruses. And that's using yeast library technology. And I'll talk about it, how we do that later in the talk. So we essentially took these two lead antibodies. We improved their affinity to Sudan GP and Burgujio GP while maintaining their affinity to Ebola Zaire. And that led to about a 30-fold improvement in neutralization potency against these other viruses while maintaining potency against Zaire. We did this with two antibodies, and I'll show you the data on breadth of protection. And that antibody cocktail is currently in phase one trials with MapBiome. As I mentioned, I condensed a lot of work into this one slide. If you're interested in any of the details, you can look at any of these different papers. And so this is just getting straight to the point with the protection data. And so what we found was that our cocktail of two antibodies, both of which are broadly neutralizing —and they target non-overlapping epitopes within the fusion loop—that the cocktail was able to reverse advanced Ebola Zaire, Sudan, and Budubusho disease in non-human primates, which is the gold standard model for Ebola viruses. And this was true at just a single 25-mg/kg dose. And so on the left is showing you the Zaire data. We gave the antibody cocktail at day 4 post-infection. You can see we're seeing 100% survival. Similar data with Sudan, even at a lower dose at a 7.5 mcg/kg dose. And then on the right is the Bundy data. We gave the cocktail at a very late time point at day 7 post-infection, and we are still seeing 80% survival. And then below here, I'm just showing you the viral load data to make the point that we were administering the antibody cocktail at a time that these animals had very high levels of viremia. Most of them had fevers and were registering clinical scores. And you can see how quickly these viral loads are knocked down to undetectable levels after antibody administration. And so to our knowledge, this is the broadest and the most potent antibody cocktail that's been described to date for Ebola viruses. And this is likely the cocktail that will be used if another non-Zaire Ebola virus species emerges. All right, so this is the last slide on this. And it's going back to a point that I was trying to make at the beginning, which is what— What you would really like is an antibody that works not only against current viruses, but also future versions of the virus as well. And as it happened, while we were working on this project, a novel Ebola virus happened to emerge. It was identified in free-tailed bats in Sierra Leone. It's called Bombali virus. And this Ebola virus has about 64% amino acid identity to Ebola Zaire. And so of course we asked the question, does our antibody cocktail neutralize Bombali virus?? And the answer was yes, which is shown in purple here. It's just showing you neutralizing activity against Bombali and the other 5 Ebola virus species. You can see the cocktail is actually more potent against Bombali than the others. And so this, of course, didn't— it isn't good luck. This is— this was the case because the 2 antibodies target highly conserved epitopes. So it was likely that they would work against a novel emerging Ebola virus. So although we don't know yet whether or not Bombali is actually pathogenic in humans or whether or not it will spill over and cause an outbreak, This is just a proof of concept that our cocktail should work against future emerging Ebola viruses. All right, so now I'm going to talk about all the work that we've done over the past year or so applying this concept to SARS-related viruses. So as I alluded to at the very beginning, it's essentially a certainty that novel coronaviruses will emerge in the future. Just the same way that SARS-CoV-2 has followed SARS, MERS, and actually 4 seasonal endemic coronaviruses. And we know that not just because this is the 7th coronavirus to spill over into the human population, but also because there are many studies showing that there's massive genetic diversity among coronaviruses that are circulating in bat reservoirs. And at least for a subset of these coronaviruses, it's been shown as some of these SARS-like viruses are able to infect human cells via human ACE2 without— very high-risk pandemic pathogens, some of which could spill over directly from bats into humans without any adaptation. So on the left is showing you a phylogenetic tree from a paper that came out from a group at UC Irvine this year. This is showing you a tree of many of these different SARS-like viruses called sarbecoviruses. And the lines here depict where they circulate geographically. And this is an important point. Which is that many of these SARS-like viruses are circulating in the same geographic regions, essentially in the same cities, in the same provinces in China, which of course offers ample opportunity for recombination. And we know that coronaviruses readily recombine because their genomes are very modular. So essentially, if you have a bat that's infected with more than one coronavirus, a piece of one can recombine with a piece of another, and now you have a novel coronavirus. And that's how this diversity is generated. So we know that about 6.5% of bats in China harbor coronaviruses. And I think more importantly, there are at least a few studies showing that a fair proportion of residents in rural China, about 1% to 3% of residents, which seems like a small number, but that translates to, I think, 5 million people, have serological evidence of exposure to these SARS-related bat viruses. And what that tells you is that these viruses are actually spilling over from bats into humans relatively frequently. We don't hear about it because typically they only cause mild disease or asymptomatic disease, or one person gets very sick but the virus doesn't transmit very well. But the point is, because these spillover events are happening all of the time, the probability that we'll see the emergence of another coronavirus that has struck this ideal balance between pathogenicity and transmissibility is actually very high. And so what we need are not just, you know, vaccines and antibodies and other therapies, that work against SARS-CoV-2. We need these drugs that work against future emerging coronaviruses as well, as well as variants of SARS-CoV-2 that, as you all know, are emerging now in the human population and that could emerge in animal reservoirs in the future, similar to what was seen last year with the mink populations in Denmark. So just a couple slides here on the variants. So as I'm sure most of you know, there are thousands of SARS-CoV-2 variants that are now circulating globally, with some of these variants emerging to much higher frequencies than others. And that's, of course, because they have a fitness advantage. They're either more transmissible or they're able to evade neutralizing antibody responses. And so arguably, there are two variants that are more concerning than the others. And that would be the B.1.351 variant, which was first detected in South Africa, and the P.1 variant, which was first identified in Brazil. And I say that because these variants are much less susceptible, particularly B.1.351, much less susceptible to neutralization by convalescent serum and sera from vaccinees. And some of that data is shown below. This is from a lab at the Reagan Institute. It's just looking at full decrease in neutralizing activity relative to wild type. And you can see the B.1.351 variant in particular is significantly more resistant to neutralization. And in many cases where you can detect neutralizing activity against the D614G or wild type virus, many individuals use detect no neutralization against this variant from South Africa. And as many of you, I'm sure, are aware, this decrease in neutralizing activity is translating, at least to some extent, into reduced vaccine efficacy. And so it's been shown that the AstraZeneca vaccine did not work against the variant that first emerged in South Africa. The Novavax and J&J vaccines showed significantly reduced efficacy against this variant. And so I'm going to make the argument, which I think is an argument that many people agree with, that these variants are emerging, that the 351 and the P.1 variant are emerging to evade pressure from neutralizing antibodies. So this is showing the structure of the SARS-CoV-2 spike protein with one receptor binding domain in the upper formation. And I'm just highlighting here the mutations that are present in both the B.1.351 and the P.1 variants. And so they have the same mutations in the RBD with the exception that the 351 variant has a K417N and the P.1 variant has a K417T. So very similar RBD mutations. They have mutations in other places in the spike. I'm highlighting the RBD mutations because the RBD is the primary target for neutralizing antibodies. And so why are these mutations emerging? And so there's been a lot of work done now by us and many others looking at antibody responses to SARS-CoV-2. What these studies have shown is that the antibody response to this virus is surprisingly public. So basically across many different, many different people are making extremely similar types of antibodies that are not only binding the same epitopes, but are actually engaging the same residues. And so these antibody classes have been categorized as class 1, class 2, and class 3 antibodies.. And I'm just listing some of the mutations that are associated with escape from many antibodies in these classes. So I'm highlighting here is that K417N is a common escape mutation for Class I antibodies and E484K is a common escape mutation for Class II antibodies. And we know Class I and Class II antibodies are dominant over Class III when you do escape studies with convalescent sera. This is just from one particular paper from Jesse Bloom, but there are others. Indeed, the, you can you can identify escape mutations when you pressure with here, and the most common mutation is at position 484. And so it seems likely that these mutations are emerging due to pressure by these very common classes of SARS-CoV-2 antibodies that are sensitive to the same mutations. And then importantly, many clinical stage SARS-CoV-2 antibodies were of course isolated from COVID-19 patients. And so what you would expect is that many of these antibodies belong to these common classes, which they do, and that many of them would be sensitive to these same mutations, which I'll show you here. So this is some data looking at neutralizing activity of the Eli Lilly antibody cocktail, and the next slide has the Regeneron cocktail against some of these emerging variants of concern. And so on the left here showing you on the y-axis neutralization IC50, so you can see both of the antibodies in the Eli Lilly cocktail, LYCO555 and LYCOV16, are quite potent against the original virus, the Victoria strain is very similar to the original virus that emerged out of Wuhan, China. This is showing you data for the B.1.1.7 variant. You can see LIKOV555 neutralizes this one very well, whereas LIKOV16 loses significant activity. And neither of the antibodies show any detectable neutralizing activity against the B.1.351 or the P.1 variant. And so if you're infected with one of these viruses, this cocktail is not going to be effective. Here's the data for the Regeneron cocktail. So again, you can see both antibodies are very potent against the original virus as well as the variant that first emerged in the UK, the B.1.1.7 variant. One of the antibodies, Regeneron 10987, which is a class III antibody, maintains its activity against B.1.351 and P.1, whereas Regeneron 10933, as you can see with the curves here, loses very significant neutralizing activity. I should have mentioned these are all authentic virus neutralization assays. And so this cocktail is essentially reduced to Regeneron-10987 monotherapy against these particular variants that have the E44K and the K417N or T mutations. And we don't know how well this cocktail actually works as a monotherapy. For example, Regeneron-10987 may rapidly select for escape mutations upon treatment. And so this is a problem. We're now seeing the emergence of variants that are leading to reduced vaccine efficacy that are escaping antibodies or even antibody cocktails. And I'll make the argument that at least in principle, inherent features of broadly neutralizing antibodies should offer a higher barrier to resistance. And there's two reasons for that. One is because broadly neutralizing antibodies often, not always, but they often target immunorecessive epitopes. So epitopes that are not targeted or not readily targeted by the endogenous immune response, which of course reduces immune pressure on those particular sites. And the second reason, which is perhaps the more important reason, is that by definition broadly neutralizing antibodies engage highly conserved residues, which is of course what allows them to be broadly neutralizing. And these highly conserved residues are often conserved, not always, but they're often conserved because they're important for viral fitness. And so although you might see escape in vitro from broadly neutralizing antibodies, the number of pathways to escape is typically lower, at least based on our experience. All right, so similar to what I just showed earlier with the Ebola project, I'll just walk through sort of the workflow for discovering these types of antibodies and also for optimizing them. And so last year, last March, we obtained a blood sample from the NIH, not from a COVID-19 patient, but actually from a SARS survivor. So this was someone who was infected with SARS in 2003. The blood was drawn 13 years later and then shipped to us a few years after that. And so to identify broadly neutralizing antibodies, we sorted B cells from the sample, not with a SARS S protein, but with a SARS-CoV-2 S protein. Of course, this person has never been infected with SARS-CoV-2. And so by definition, we would be pulling out broad binding antibodies. And so from that effort, we identified 200 cross-binding antibodies, so antibodies that cross-bound SARS-CoV-2 1 and SARS-2. And of those 200, 7 of them showed broadly neutralizing activity. And we published this work last summer. I'm not going to talk about it today. I'm going to talk about the second half of the project. So although these, these 7 antibodies were broadly neutralizing, as I alluded to when I was talking about the Ebola work, these antibodies were much more potent against SARS-CoV-1 than SARS-CoV-2, exactly what you would expect. And so we selected 3 antibodies, which I'll walk through, We affinity engineered them using yeast library technology. And then from that effort, we identified a lead clone called Adagio2. That was all described in a paper in January, so you can find any information that I don't talk about today there. And we've half-life extended this antibody with a 2-amino acid substitution in the Fc region to create a new molecule, which we call Adagio20. So on some slides you'll see Adagio20 and some Adagio2, but they're essentially the same thing. All right, so this is showing you the data for the 7 broadly neutralizing antibodies that we identified directly from the SARS survivor B cells. So this is neutralization IC50 on the y-axis. I get SARS-CoV-2, which is shown in blue, SARS-1, which is in red here, and then WIV-1, which is a SARS-related bat virus that I was talking about earlier. This is essentially a representative pre-emerging coronavirus that could spill over into the human population. It's been identified in bats. And so what you can see is that these 7 antibodies show varying degrees of neutralization breadth and potency. All of the antibodies are more potent against SARS-CoV-1 than SARS-CoV-2. Again, it's expected. And so what we wanted to do was affinity mature these antibodies to improve their neutralization potency against SARS-CoV-2 while also maintaining or even ideally improving their potency against other SARS-related viruses. So these were the 3 antibodies that we selected for affinity optimization. So at Addimmune, we have a lot of different methods that we use to generate affinity maturation libraries. For antibodies that we isolate from human B cells, we usually use an oligo-based approach where we introduce diversity into the CDR regions. And then we build libraries with DNA shuffling or by stitching all of those together. And so for these 3 clones, that's the method we used. We made yeast display libraries that were on the order of 10^7, 10^8 diversity. And then we did selections on those libraries by flow cytometry to identify clones that were improved in binding over the parent clones. Those yeast cells are then plated. They're sequenced directly from the yeast. The IgGs are produced directly from the yeast. And then if needed, you can go into another round of affinity optimization. And so that's what we did. And this is just sort of showing you an example of one of the clones. So we made a heavy chain diversity library and a light chain diversity library, and we selected them separately. So the blue is showing the library, and the light blue here is the parent. And so you can see, you draw gates around clones that show improvement in binding over the parent clone. This is using an S1 monomer. And then we sort out all of the improved heavy chains from the library from the second round of selection, all of the improved light chains from the light chain library, and then we make a new library again to select for clones with further improvements in affinity. And so we did that across 3 different clones. This is just showing you the terminal round of selection. You can see the libraries are binding. Seems to be that there are clones in the library that are significantly improved over the parent based on these flow cytometry plots. And so of course, as I mentioned, these yeast cells are sorted and the antibodies are produced.. And so this is showing you essentially the binding affinity evolution for those 3 different lineages. So the parental clones are shown in light blue. And then all of the progeny coming out of this affinity maturation effort are shown in purple. And so you can see we're getting somewhere between 20 and about 1,000-fold improvement in monovalent bindings. This is FAB binding to the SARS-CoV-2 spike protein. The improvement in binding translated into an improvement in neutralization, although it wasn't linear. So for a 1,000-fold improvement in binding, we're not seeing a 1,000-fold improvement in neutralization. More like 70-fold for Adagio-1 and 2, which you can see here. We also ran a number of clinical stage antibodies as comparators to make sure we were in the same range, which we were. You can see Adagio-1 and 2 have neutralization potencies very similar to Regeneron 10933. It's worth noting here that although VR7831 doesn't neutralize in this particular assay, it does neutralize in other assays when you use different types of targets though. This is a pseudovirus assay, an MLV pseudovirus assay, which we know correlates with authentic SARS-CoV-2. But of course, I'll show you that data. So based on the neutralization data and developability data, we ended up selecting Adagio-2 as our lead antibody. So that's the one I'll talk about for the rest of the presentation. So as I mentioned, we had looked at neutralization in the pseudovirus assay. So the next question was, does this antibody actually neutralize authentic SARS-CoV-2 just as well? How does it look relative to other clinical stage antibodies? And does it maintain its breadth of activity?. And so this is a phylogenetic tree of some of these SARS-related viruses, and I just want to highlight here that it's really the clade 1 viruses that are of greatest interest because these are the ones that I was mentioning earlier that are able to infect human cells via human ACE2. Of these viruses, only some of which are listed here, there are 4 that are available for neutralization testing. And that would be SARS-CoV-1, SARS-CoV-2, which isn't listed here, WIV-1, which I mentioned earlier, and SHCO14, which is one of the most divergent viruses from SARS-CoV-2 within this clade 1 sarbecovirus subgenus. And so if your antibody neutralizes SARS-CoV-2 and SHCO14, it should neutralize most of the viruses in between just based on a sequence homology. So this is showing you that data. So the SARS-CoV-2, right? So we ran different clinical stage antibodies. So one from VR, the two from Regeneron, two from Lilly, and one of the antibodies from AstraZeneca and Adagio2, and two different authentic SARS-CoV-2 neutralization assays, one using HeLa ACE2 cells and one using Vero cells. And so you can see Adagio-2 shows extremely potent activity with an IC50 around 1 nanogram per mL in both of these different assays. So again, very similar, as we saw in the pseudovirus assay, to Regeneron 10933, and a bit more potent than some of these other clinical stage antibodies, although you can see all of the antibodies are very potent. But importantly, Adagio-2 showed very potent activity against the other SARS-related viruses as well. Whereas most of the other clinical stage antibodies failed to neutralize these other SARS-related viruses, with the exception of BIR7831, which is also broadly neutralizing, although you can see it's about 52— I mean, 10 to 50-fold less potent than Adagio-2 depending on the assay. This antibody was also identified from a SARS survivor, which is interesting.. And so as I mentioned, there are only 4 of these sarbecoviruses really available for neutralizing, neutralization testing. And so essentially as a surrogate for neutralization, so let's just look at binding to representative sarbecoviruses. And so the ones that are highlighted here are the ones we chose. We just tried to span the whole range. So I'm showing you the data for the Adagio-123 antibodies and these other clinical stage antibodies, as well as ACE2. So you can see Adagio-02 has very broad binding activity. It bound to 12 out of these 13 clade 1 viruses that we tested. Interestingly, the one The one that it didn't bind, RATG13, is ironically closest in sequence to SARS-CoV-2. This virus didn't bind to human ACE2, at least in our assay, or it's extremely low affinity. And the mutation that's actually required to bring on ACE2 binding also brings on Adagio-2 binding. And so if this one spills over, maybe it will select for that mutation and bring on antibody binding. You can also see the VEER antibody, as expected, is extremely broad. It bound to all 13 of these sericoviruses, whereas the other clinical stage antibodies showed much more limited binding breadth, again, as you would expect based on where, where their epitopes are. And so we've done a lot of work mapping the epitope recognized by Adagio-2 using fine epitope mapping and structural studies. This is showing you the fine epitope mapping data, and we looked at this by making an RBD display library, or essentially mutagenizing the receptor binding domain by sampling every amino acid at every position across the RBD. And then we select for loss of binding by Adagio-2. And then we select for binding to other antibodies that bind to non-overlapping epitopes to make sure that we're not just selecting for global knockout. So we were using VIR7831 and CR3022 because they don't compete with Adagio-2. So on the right is showing you a map of the different mutations that knock out Adagio-2 binding. ACE2 binding is shown on the bottom. So you can see many of the mutations that knock out Adagio-2 binding also knock out ACE2 binding, which potentially explains why those are very conserved. And on the bottom is showing you the conservation at these 4 positions: position 405, 502, 504, and 505. So as expected, you can see that the residues at these positions are highly conserved, at least across the clade 1 viruses, which is what you would expect given the breadth of neutralization via DODGY02. And so this is showing you the structure of the antibody, cryo-EM structure, which is solved by Jason McCollin's lab.. And this is just to go back to the point I was making earlier, which is that often broadly neutralizing antibodies target sort of these immunokaisin epitopes. And so this is the surface targeted by Adagio-2. You can see it's targeting a surface that's distinct from these commonly induced classes of antibodies, these class I, II, and III antibodies. And the mutations that are associated with lack of binding by Adagio-2 don't overlap with these other classes. And when you look at the map of mutations from Jesse Bloom's lab, where he's pressuring with convalescence serum, you don't see any mutations at the Adagio-2 site, which again speaks to the fact that this is likely an epitope that's very rarely targeted by endogenous antibodies. And we've now isolated over 1,000 antibodies from COVID-19 patients, and we've only identified one that looks like the parent of Adagio-2, which again suggests it's a very rare epitope. And perhaps for this reason, as well as the fact that I mentioned earlier, In fact, the point I mentioned earlier, which is that these antibodies target conserved residues, when we look at binding of Adagio-2, and again here I'm showing you Adagio-10, 20, 30, which are half-life extended versions of Adagio-1, 2, and 3. But you can see Adagio-20 and these other broadly neutralizing antibodies, these two antibodies target a site called class IV, which has been shown to be very conserved. VR7831, as I mentioned earlier, is also broadly neutralizing. These antibodies bind to all of the— all or most of the commonly circulating SARS-CoV-2 variants that we've tested to date. So we essentially went into the GISAID database, downloaded all of the sequences, filtered them to look at the ones that have mutations in the RBD, and then made the most common variants. And so you can see most of these antibodies bind to all of the variants. Adagio-10 only has one knockout mutation. In contrast, when you look at the SARS-CoV-2-only neutralizers, the ones that are not broadly neutralizing, you can see that there are many resistant variants variants that are circulating in nature, including the variants that I mentioned earlier, the B.1.351 and the P.1 variants, which aren't recognized by 3 out of these 4 clinical-stage antibodies. And again, this is likely due to the fact that these antibodies are binding to variable epitopes. So that's just binding. What about neutralization? So we've looked at neutralizing activity of Adagio-10, 20, VIR-7831. Again, these 3 broadly neutralizing antibodies, as well as many of these SARS-2-specific antibodies, against the Victoria isolate, which I mentioned earlier, the B.1.1.7, 351, and P.1 variants. And again, what you can see is that Adagio and TEN-020 maintain very potent activity across all these different variants. The VIRE antibody maintains its breadth, whereas when you look at the SARS-2 specific antibodies, you can see many of them, not all, but many of them are losing some activity or all of their activity against these new emerging SARS-CoV-2 variants. On the right is showing you some more recent data that was collected by a lab at the NIH looking at Adagio-10 and Adagio-20 across many of these different common single point variants, some of which escape recognition by clinical stage antibodies, as well as many of these newer emerging variants of concern, one from Japan, Uganda, Southern California, and New York, which are shown here at the end. And you can see Adagio-10 and 20 again are maintaining their activity within 10-fold of the original D614G virus. So this is some in vivo data. So we've done in vivo studies now in mice, hamsters, and NHPs. For the sake of time, I cut out the NHP and hamster data, and I'll just focus on the mouse data. And that's because this is the data that supports the fact that breadth of neutralization translates into breadth of protection. So this is a study that we did in collaboration with Ralph Baric's lab using mouse-adapted SARS-CoV-2. So here we're administering 200 microgram Adagio-2 IP, and then the mice were challenged with SARS-CoV-2 intranasally on day 0, and we monitored the mice for weight and respiratory changes every day, and the mice were sacked at day 4 to look at lung viral loads and histopathology. So you can see that the mice that were treated with Adagio-2 showed very minimal weight loss and were completely protected against respiratory burden, hemorrhaging in the lung, and lung viral replication. And when we repeated the studies with SARS-CoV,. You can see the results look almost identical, minimal weight loss and complete protection against clinical measures of SARS-CoV-2, SARS-CoV-1 disease. And so again, breadth of neutralization is translating into breadth of protection. This is just showing you the antibody development timeline. So we received the sample, as I mentioned, from the NIH at the end of March in 2020. We identified the cross-binding antibodies the beginning of April and had neutralization data just a few days after that. We immediately started on the affinity maturation effort. That was complete at the end of May. We spun out a company called Adagio Therapeutics, the company that's developing the antibody and doing all the clinical trials. That was in the middle of June. The antibodies went into cell line development at the end of June. We filed the IND in December. And we started our first in-human trial at the beginning of February. And we expect a readout, I believe, in fall of this year. And I just want to end with one slide on a vaccine design, because I think it's an important point, which is that these broadly neutralizing antibodies not only have value for prophylaxis and therapy, but also in guiding the design of vaccines that induce similar types of antibodies. This is showing you some of the targets of these broadly neutralizing antibodies that have been described to date that target distinct epitopes. You've got the Adagio 10, 20, and 30 site, VIR7831, which targets a distinct site, and then these so-called class IV antibodies defined first, I believe, by COVAY116, which target yet another broadly neutralizing site, There are other antibodies in the S2 region that have also been described that are broadly neutralizing. And so you can imagine designing vaccines that really focus the antibody response on these conserved sites, the idea being to induce antibodies that work not only against the original form of SARS-CoV-2, but all these different new emerging variants and future sarbecoviruses that spill over into the human population. And so this is, of course, something that we and many others are are working on. So most importantly, the acknowledgments. Acknowledgments for the Ebola virus work. So most of the work at Adimab was done by Eileen Goodman. The neutralization assays were done in Kartik Chandran's lab at Albert Einstein. The animal studies, the NHP studies, were done at US Amarin and John Dye's lab at UTMB and the Public Health Agency of Canada and the Guy Spritlag and the Key Lab. And MAPBio is running the Phase 1 clinical trial. Acknowledgments for the SARS-2 work shown here. So most, a lot of the work I showed was done by Garrett Raposo and Kenzie Kaku at Adimab. I want to thank everybody at Adagio, who they're all working very hard to develop the antibody and get it through clinical trials. Gavin Screeton's lab ran the variant neutralization assays. There were other neutralization assays done at Scripps, which I don't think I showed here today, as well as the FDA and Carol Weiss's lab. The mouse model was from Ralph Baric. John Dye's lab ran a monkey study, which I didn't have time to show today. And the structure that I showed was done in Jason McCollum's lab at UT Austin.

And I'm happy to take any questions. Thank you so much. That was wonderful. Really appreciate it. We have a bunch of questions coming in for you. So maybe we'll start with whether you know the RBD-directed neutralizing antibody, whether it binds to the initial SARS-CoV-2 virus as well as the escape mutant virus. Sorry, I don't think I understand the question. Okay, so the question again was whether the RBD-directed neutralizing antibody against the initial SARS-CoV-2 virus also binds to the escape mutant virus.

I think they're asking whether or not a DOTGO2 binds to some of these knockout mutants that we described, or if those are associated with escape. Yes, exactly. Um, when we do in vitro escape assays, we're able to select for escape variants with all of the broadly neutralizing antibodies and all of the SARS-2-specific antibodies that we've looked at to date. The difference is that for the broad antibodies, we select for fewer escape mutations. So for example, for Adagio-2, and I think Vera has described an antibody that's similar to Adagio-2, Adagio-3, and they've seen a very similar result with a very similar mutation. Only a single escape, we only identified a single escape mutation at position 504. We don't actually know yet whether or not that that's a fit virus. So whether or not you would select for that in a patient, but we do select for one. Regeneron's published data on their cocktail, and we've looked at their antibodies as well, and we select for multiples. So that's really the distinguishing features, I think, when you look at broad antibodies versus SARS-2 specific is the number of escape mutations, number of pathways to escape.

Right. Okay. Thank you. And have you used the epitopes targeted by your antibodies to generate immunogens for vaccine use?

Yeah, so that's something that we're working on now, but we don't have results yet, so I can't comment on it. But yes, we are doing it using all the strategies you would imagine using that people have used in the HIV field and in the influenza field.

Yeah, and I guess the follow-up question to that is, with an exceptionally rare antibody like Adagio-2, do you think its epitope would allow you to induce Adagio-like antibodies, the Adagio-2-like antibodies?

So that's a good question. When you look at the substance, so the parents of Adagio-2, which are broad but they're not very potent. So Adagio, well, the parents of Adagio 1, 2, and 3, I should say, those parental antibodies are broad, but their potency is relatively modest. Their potency runs 100 nanograms per mL. We have found antibodies that are similar to those antibodies in SARS-CoV-2 patients. And so they do exist indicating that you can elicit them. The germlines aren't particularly rare. They don't have massive numbers of somatic mutations. They don't look like HIV antibodies, for example, where you look at them and you say, oh my God, they'll never be able to induce that specificity. They look relatively vanilla. So I think by masking these common sites, these very immunogenic, immunodominant epitopes like these class I, II, III sites, you might redirect the response to these more conserved sites. I think that remains to be seen. But I don't think there's a problem with precursor frequency just based on the germlines that are used and the fact that they don't have unusual sequence features like super long H3s and they're not polyreactive and things like that.

Right, right. And can you talk about the antibody-dependent enhancement of disease and if your approach addresses that in any way?

Um, so when you think about ADE, I think you have to consider the fact that although, you know, for many viruses, um, you know, you can, you can, you know, set up these in vitro ADE assays and detect ADE in a cell line, generally ADE doesn't translate into animals with most viruses. And that's been shown with SARS-CoV-2, that antibodies that enhance in vitro have actually been shown to either do nothing something in animals or actually protect. And even if you do see it in animals, whether or not it'll translate to humans. So there are only 2, maybe 3 examples that I'm aware of, of actual clinical ADE of disease in humans. And that's in the case of dengue, of course, which is sort of the classical Fc receptor-mediated ADE. And RSV, that early RSV trial in the 1960s with the formalin-activated RSV. And in both of those cases, when you think about the mechanism, so the dengue mechanism relies on FC receptors, FC receptor-mediated uptake of the virus, which requires the target cells to be expressing FC receptors. That's not really the case here. And then the RSV case, that was associated with the induction of non-neutralizing antibodies. So with these very potent SARS-CoV-2-specific antibodies, it's hard to imagine seeing AD in humans, especially because we haven't seen any evidence for that. With convalescent sera, with vaccines. And in some ways, I think the best evidence would be the Novavax data from their trial in South Africa, where what they saw was that people that were previously infected with the wild-type virus and then reinfected with the B.1.351, those, those people didn't show any enhanced disease. In their first midterm readout, they were saying there was no protection from the previous infection, but in their later readout, they actually showed 50%. Protection. And so infection with the wild-type virus, although we know that those neutralizing titers are probably extremely low or non-existent against the B.1.351 variant, those individuals were still, they still had some degree of protection or no protection. It's not like we were seeing all those people, you know, were hospitalized or something. And so to date, I don't think there's any evidence of clinical ADE.

Okay, okay, thank you. And have you attempted to put pressure on the virus with lower concentration of antibody to see if you can elicit natural escape?

Yes, so that's, that's what I was talking about earlier, that we can, you can, you can select for escape variants in vitro, a single escape variant, right? But we don't know whether or not that variant will be selected in animals or even in people. And I think, you know, it's worth mentioning, like, when you think about escape in resistance, you have to consider two sort of types of resistance. There's circulating resistance, which is sort of what I was talking about with the B.1.351 variant. These are isolates that are circulating, might be resistant to your antibody. And then there's treatment emergent resistance, which is of course when you treat with your antibody, you might select for these different variants. And there isn't a ton of data yet, you know, from humans on treatment emergent resistance. But if you look at the Eli Lilly trial, They had a monotherapy arm with LICO-555, and they had a cocktail arm with LICO-555 and LICO-16. And when— if you look in their EUA document, they have lots of data on resistance. And what you can see was that although they saw some resistance in the monotherapy arm— I think was 8%— in the LICO-555 group, it wasn't statistically, you know, significantly at least different than what they saw in the placebo group. In the placebo group, they were seeing something like 7%. And most of that resistance was driven by viruses containing E484K mutations. So why would there be resistance in the placebo group? This goes back to what I was mentioning at the beginning about these sort of canonical classes of antibodies. It's probably due to the fact that some proportion of patients, maybe that 7%, are inducing LYCO555 dominant serum neutralizing antibody responses. So again, a lot of pressure on that particular residue. And so the escape that they were seeing in the monotherapy arm was very similar to the placebo arm. And probably more importantly, there was no difference in clinical outcomes in patients that were treated with monotherapy versus combination therapy. And so to me, this is worst-case scenario. LYCO555 is the worst-case scenario because it's targeting this very variable epitope that E484K mutation confers escape. So I'm not that worried about treatment-emergent resistance. And also, if you look at the VEAR data, that's monotherapy. They're seeing 85% efficacy in the clinic. They didn't report on escape yet, but it's obviously not so much escape that they're seeing significantly reduced efficacy relative to what other companies are seeing. In fact, arguably they're seeing better efficacy, 85% versus, you know, Regeneron supporting 70%. Circulating resistance I see as a bigger concern given, you know, the emergence of all these variants of concern. But for the reasons I was mentioning earlier, the fact that DODGY-02 and some of these other antibodies target these subdominant epitopes, and they're targeting these conserved residues, the likelihood that we're going to see these mutations, generally single mutations that escape these antibodies, arise in nature relatively low. Although to me, the ideal cocktail would be— ideal therapeutic would be a cocktail of two broadly neutralizing antibodies, which is what we're moving towards now.

And I guess to that point, can you talk about the delivery for an antibody cocktail that you foresee? Because obviously there's a lot of constraint around intravenous, infusions versus the subcutaneous.

So what, what are you thinking about for delivery? Right, so I think some of the best news, you know, for, for, you know, in this field and in the therapeutic SARS-CoV-2 antibody field has been that the antibodies are showing efficacy at relatively low doses. So Lilly didn't see any difference between their, I think it was 8-gram or 7-gram dose and their 700-mig dose. They're with an antibody that's 10 times less potent than many of the others, including ours. They're seeing just as good efficacy at 500 mg. And so now you're getting down into these doses, which suggests you might even be able to go to 300 mg. And so now you're getting down into these low doses you can easily give IM or subq. And so everybody is now moving toward IM and subq. The only downside of IM and subq is that there's a delay to get to sort of the peak. Antibody concentration. You give an antibody IV, you're there immediately. So the IM sub-q, there's some delay, maybe a day or something, to get to higher concentrations. Whether or not that will matter, I don't know, but I'm sure data will be coming very soon. I think Regeneron already reported some viral load data with their sub-q arm, and it looked similar to IV, so that's very promising. I think that's, that's going to be the future. You go to a pharmacy, you get a test, and you get an antibody right there.

That's the ideal. Situation. Well, to that point, I mean, have you been, do you think, or maybe disheartened, or do you think there are any lessons for the uptake of the antibodies that have been approved under the emergency use authorizations in, you know, in developed countries where they haven't been used as widely or as broadly as maybe one would have anticipated?

Yeah, I mean, I think there are multiple reasons for that. One is the IV, you know, having to go in and get an IV infusion and have somebody wait with you for an hour and then another hour when everyone is busy with actual COVID patients. That obviously was a barrier. Another barrier is, you know, just getting patients in early enough. So you basically have to get these antibodies at a time that you don't feel very sick. And I think most people are resistant to going into a doctor's office when you don't feel that bad. Most people think, oh, I'll wait to see to see. And by the time you start feeling really bad, it's kind of too late. So you need to catch it very early. And then the third thing is, I think it was just very— and I think that this is still the case. I think there's a lot of confusion among, you know, doctors and patients and, you know, when it comes to monoclonals and how to actually get an antibody if you want one. Like, I even personally had a friend who was infected and like, we couldn't even figure out how to do it, you know? And I'm like, if I can't figure this out, like, you know, I don't have a PhD in this, definitely are not going to be able to figure out, it's so complicated. So hopefully all of that will be addressed in the future with IM delivery and, you know, education with doctors and with patients, hopefully.

Right, absolutely. Okay, well, thank you so much. I think that's all the time we have for the Q&A, so I'm just going to take over the screen share here. And just thank you again for the fascinating presentation. We really appreciate your sharing your work with our audience and for taking the time to answer questions. I'd also just like to thank the attendees for participating in today's webinar. We are fortunate to have such an engaged audience for these lab meetings. And I'd like to thank you for spending your time with us and submitting questions for our speaker. I also would like to invite you to join us 2 weeks from today on May 20th for the next Global COVID Lab meeting. Our speaker that day will be Dr. Bob Cedar from the Vaccine Research Center at the U.S. National Institute for Allergy and Infectious Diseases. He will speak about immune correlates and mechanisms of protection against the Washington and B.1.351 SARS-CoV-2 variants following mRNA vaccination in non-human primates. And if you're interested in more research on COVID-19, please sign up for the HVP COVID Report, a biweekly newsletter that provides insights from experts around the globe and highlights the latest scientific articles and data. Finally, please visit our website and follow us on LinkedIn, where we will upload a recording of today's webinar. And with that, I'll say thank you again for participating today. Thank you, Dr. Walker. Stay safe, and we hope you'll all join us again on the next Global COVID Lab meeting. Thank you so much.