Scientist Stories: James Crowe, Viral Immunology & Antibodies

Okay, thank you everyone. Welcome to the influenza webinar series hosted by the Human Vaccines Project. The Human Vaccines Project is a nonprofit public-private partnership with the goal of decoding the human immune system to accelerate the development of vaccines. Vaccines and immunotherapies. In order to make real progress towards the development of universal influenza vaccine, collaboration among researchers around the world is essential. We hope this monthly webinar will foster the scientific foundation needed to advance influenza research. The series of webinars will be broadcast on the third Tuesday of each month and posted online. For a copy of today's webinar, please visit the Human Vaccines Project website at http://www.humanvaccinesproject.org/talks. T-A-L-K-S. Following the presentation, we will have time for questions from the audience. Please type your questions into the question box on the GoTo interface. I will read these questions following the presentation. And now it's my honor to introduce Dr. James Crowe. Dr. Crowe's laboratory has a broad portfolio of work in viral immunology and cell biology with an aim to discover mechanisms important to develop new vaccines. In addition to his laboratory, Dr. Crowe directs the Vanderbilt Vaccine Center and the Flow Cytometry and Cell Sorting Core at Vanderbilt. Dr. Crowe has been the recipient of numerous investigator awards and is an elected fellow of, among others, the American Academy of Microbiology, the American Association for the Advancement of Science, the American Society for Clinical Investigation, the American Academy of Pediatrics, the Infectious Disease Society of America, and the Association for Pediatric Society. He was elected the National Academy of Medicine— he was elected to the National Academy of Medicine in 2014. Today, Dr. Crowe will tell us about the complexity of human B cell responses to influenza vaccination. Thank you very much. Go ahead, Dr. Crowe.

Great. Well, thanks for joining us, and it's great to talk to you today about B-cell and antibody responses to influenza. The last month we had an awesome presentation from Jesse Bloom, uh, that touched on a lot of the new techniques and approaches around looking at viral diversity, especially in the HA protein on influenza. So you have this arms race in the immune system where the viral sequences are quasi-species and they're moving in time, and Jesse very eloquently talked us through how current researchers are studying that. Today we're going to look on the other side at the antibody B-cell response, which also generates swarms and quasi-species. So in effect, you've got to have interdisciplinary science to understand response because the virus is moving and the immune response has to chase it, and both of those systems are using polymerases that are error-prone and sustain a lot of variants and mutations. So that's, that's what we're going to be getting at today, is the complexity in the human B-cell response. This type of science has really exploded in the last 10 years because of new technologies for isolating human monoclonal antibodies. And when I was a postdoc in the 1990s, this was just a pipe dream. Phage display had just come online, and people were starting to make human antibodies, but they weren't paired with the, you know, the natural heavy chain and light chain pairings. And so there was a, there was a flush of human antibody activity at that time, but more recently, better techniques have evolved. And Jody, I'm not sure I can Advance slides. Somebody needs to give me control.

Can you try again?

Um, I'm still not getting, getting advance.

Okay, I'll do it for you. Apologies.

Okay, so let's go to the next slide. Looks like you have control.

You should have control now, Jim.

I'm still not able to advance my slides here. Okay, so as I mentioned, a lot of groups have developed techniques for making human monoclonal antibodies And we have our own particular approach, which uses EBV transformation to expand B-cell lines. We find that those EBV-transformed B-cell lines are somewhat unstable over time, so we fuse them to a myeloma cell, which is a non-secreting tumor cell. And by fusing the, the human B-cells with the myeloma cell, we end up with a hybridoma, which has been done with mouse cells for years, but it was very difficult to do with human cells. We developed an electrofusion—

muted—

that made the efficiency much higher. So once you have a hybridoma, these are not clonal in these multiwell plates. So we use flow cytometry or some other single-cell technique to get a single cell, make it a clone. Then you have a monoclonal secreting hybridoma, and those secrete human IgG of or human immunoglobulin. Sometimes they're IgM, IgG, IgA, and we express and purify those. We also can clone the genes out of these so that we can manipulate the antibodies and make them recombinantly in cell lines like 293F cells, or CHO cell lines would be the manufacturing cell line of choice. And over time, we've hardened this off to where we can do 100-plate screens. We can do this on robotic liquid handlers. At any one time in our lab, we're probably expanding 1,000 human monoclonal antibodies, and we can scale them to gram amounts for things like primate studies. So this is one approach. Others have used plasmablasts or single-cell sorting and that type of thing. And we have a— there's a large number of investigators in the field. Patrick Wilson, David A. Corti, and Tony Lanzavecchia in Switzerland, the Adimab group, and on and on, Florian Kramer, who have all developed their own methodologies. So now you're seeing thousands of human monoclonal antibodies to influenza. And I'm going to mostly talk about antibodies we made today, but it's not— the field is not limited to our work. So one thing you have to remember about influenza, there's serotypes or types. Type A is the one that most people have focused on for human infections. Of course, there's Type B, and the vaccines contain both Type A and B antigens. Within Type A, there are 18 subtypes that have been defined, and this is just a rough phylogeny that shows those subtypes also fall into 2 major groups, Group 1 and 2. The 3 human subtypes that have commonly circulated, H1 and H3 currently, and H2 from '57 to '68, are— they actually cross two different groups, which makes things complicated. If you look at avian viruses that have crossed over to humans more recently, H5, H7, and H9 also represent the two groups. Other viruses have been reported occasionally from birds to humans, like H6 and H10, but the ones that have caused outbreaks are H5, 7, and 9. And then of course there's pig viruses. A lot of flu viruses are in pig populations, and at least in the U.S. there's been a number of fatalities from H3 variant virus, which is in pig populations in the United States and crosses over, infects, and kills children. Some people call this state fair uh, flu virus, pig virus sometimes. So these are the types of, um, the subtypes that have crossover to humans that we know we need to pay attention to, uh, and the other ones might be of importance in the future, of course, because they're in birds, uh, typically. So I'm just going to spin through some snapshots of how antibodies recognize influenza, and I'm going to focus today on hemagglutinin or HA antibodies. There's also NA antibodies And that's a big new future direction in the field. So this is a 3-dimensional structure of the 1918 South Carolina H1HA. Of course, there's a globular head and then a stem region at the bottom. You would have the transmembrane domain, which is truncated in these soluble forms of HA. And here you have a complex of a human FAB that we isolated from an approximately 100-year-old person who had survived the 1918 influenza infection as a child. And this is a typical stem antibody. So Crucell and Wayne Marasco are two different groups that identified this type of antibody a number of years ago. And it typically— the canonical antibody interacts only with the heavy chain shown here and the FAB in yellow. Light chain's not really involved. And the reason is there is a canonical antibody variable gene, VH169, that encodes a hydrophobic loop which sticks into a shallow hydrophobic pocket in stem. So when you hear the whole movement of universal influenza vaccines was sort of kicked off by the observation that this type of antibody not only binds H1, but it also binds H5. That was very confusing at first, Now we know there's heterosubtypic antibodies, and the canonical one is the H169 to H1 and H5, but broader antibodies have been isolated over time. So this is— this antibody is actually not— this structure is not published yet, but from Ian Wilson's group at Scripps, and we isolated this antibody a number of years ago. The more typical antibody that people measure in neutralization or hemagglutination inhibition antibodies is a receptor binding domain specific antibody. So this is an FAB, 3 copies of an FAB, and ribbon representation on a space-filling form with glycans of the 1918 HA. And the receptor binding domain is right up here on the top of the head, and this binds sialic acid, which is the receptor for influenza. What was really interesting is you, you typically would think, because there's a lot of variability around the renom of the receptor binding domain, most of these antibodies are strain-specific. But when we obtained this antibody again from a nearly 100-year-old person in 2007, I believe, shortly thereafter an outbreak occurred of the 2009 pandemic virus. And when we looked at the footprint of this particular antibody, which was induced by 1918 and did not react with, uh, more recent H1 viruses, when the 2009— this California virus came out in the open box, we found that the 1918 antibody bound identically to the 2009 virus, and that's because the residues in the footprint are very, very similar. And it sort of revealed that the 2009 H1 was highly antigenically related to 1918, and very likely the 1918 influenza persisted in swine populations over the 20th century. The 2009 pandemic virus was actually a triple reassortment, probably with a human virus, a bird virus, and pig virus. But antigenically, the H1 is is derived from, uh, 1918. And the other interesting thing about, uh, this work was that it predicted that very elderly people would be relatively protected against the 2009 pandemic. And in fact, they were. The oldest ones who had this type of 1918 antibody around, sometimes in high titer, we saw subjects who were in their 90s or 100 who had HAI titers over 1,000 to 1918 flu. So it's amazing you not only have the B cells persisting, but you even have serum titers of this type of antibody for 60 or more years after virus has circulated. So sort of the world record for demonstration of B cell memory. Now, as we have done this type of work making human monoclonal antibodies to influenza HA, our work and that of others has demonstrated that there are canonical modes by which antibodies recognize these, these structures, especially receptor binding domain. And in the receptor binding domain, there's two principal ways that the tip of an antibody interrogates or recognizes the receptor binding domain. One is to put a carboxyl group mimicking sialic acid receptor you can do that with an antibody CDR that displays an aspartic acid residue at the tip. The other way is to have a hydrophobic residue at the tip, and these are typically aromatics. I'm going to just show you examples of that. So Jens Krause and our group a number of years ago isolated this H1-neutralizing antibody called 5J8, and then Steve Harrison's group published the structure. I believe the Duke investigators made this antibody. These are two different FABs and two different investigator groups and structures, Stage 65 on the right and 5J8 on the left. And you can see that on the left, the heavy chain, which is in darker pink here, inserts the heavy chain CDR3, or the hypervariable loop, which is called the complementarity-determining region 3, CDR3, into the receptor binding domain. Also shown in yellow is sialic acid bound in NH1 structure, and then HA is shown space-filling here. On the right, the same thing occurs in which you get a heavy chain CDR3 inserting in the receptor binding domain. But if you look closely, these antibodies are coming in from different, different directions and over different rims of the receptor binding domain. And if you look closely at the CDR3 representations, again, the sialic acid is in the same position, of course, on the same HA, but again, you're getting different angles of approach of these independently derived human monoclonal antibodies. But very interestingly, essentially not only is the antibody framework a scaffold, but on top of that, the CDR3 is essentially a scaffold that delivers the final effector molecule, which is aspartate. It's just super interesting that both antibodies put an aspartate in an identical position, which overlaps with the interaction site of sialic acid. So basically, an antibody that delivers aspartic acid in the right position with the right length CDR will bind this site and neutralize. So that's one of the canonical modes. Now, you think these— when you get a rule like this, you think, okay, well, now I can recognize a flu antibody. I just look at the CDR3 and see an aspartate. Especially if it has similar structure. But we found another antibody after experimental vaccination with H5 subunit vaccine. We were testing vaccines here at Vanderbilt in our vaccine testing evaluation unit, vaccine treatment evaluation unit, so NIH-sponsored vaccine trials of subunit vaccines. And we isolated this antibody, and in this case, in this sort of turquoise color the antibody puts a CDR3 in, but the aspartate points in the wrong direction. So HA's on the bottom and the aspartate points away. Katie Wenarski got this co-crystal structure with Vince Miller here at Vanderbilt. And if you look at all the other structures to which I've been referring that have aspartates, they interact with the HA, but the H5.3 antibody points the aspartate away. So it has it in the right position,, but it doesn't display it in the same manner. So I think it's interesting that you have these canonical modes of recognition, but they're not hard, fixed rules, and it's not just a sequence-based search. You really have to understand the structure of both the antibody and the antigen to apply these rules. So that's the aspartate, and then I'm going to switch to the other canonical mode, which is the aromatic hydrophobic residues., and for this I'm going to show you some antibodies that neutralize 1957 H2N2. So there's been a series of pandemics in human population. 1918 was the first. Once we got 1918 antibodies from 100-year-old people, we thought, well, wow, we ought to be able to get antibodies to every pandemic that has occurred because all the other ones were more recent. So we looked at middle-aged individuals who had lived between '57 and '68 when the H2N2 virus circulated in human populations. And we obtained these 3— we obtained a large panel, but these 3 monoclonal antibodies in Wilson's group at Scripps obtained co-crystal structures shown here for an antibody 8m2, 2g1, and 8f8. And they all bind at the top, they bind to the receptor binding domain, which is why they have hemagglutination inhibition activity, which is blocking sialic acid. So they all sort of look the same, you know, if you stand back and look at them. But if you, if you indicate the alignment of the axis of the heavy chain and the light chain, you see that they're actually coming in in 3 different orientations. And then if you look at the antibody footprint or epitope on the bottom of each of these, shown in yellow, you see that the general antigenic site is very similar. Of course, this is the receptor binding domain, but in fact, the shape of the perimeter of the footprint differs in each case. And in the very center, you see an aromatic residue, and it's not even always the same residue. It can be a phenylalanine here, phenylalanine, and this is a tyrosine. So this is a case in which Same thing, the antibody framework is a scaffold for CDRs, and that CDR is a scaffold to deliver an aromatic. And if you just look at a space-filling representation of the receptor binding domain on the left, on the right is sialic acid bound into the receptor binding domain. On the left are the aromatics. I'm just showing you the aromatics at the tip. Of the CDRs of these antibodies, and they come in from completely different angles, but essentially they deliver that aromatic residue in the correct position. And of course, that mimics one of the interactions of sialic acid. So these are the two principal molecular determinants of binding to the receptor binding domain of influenza using human antibodies. And in fact, at least one antibody has been identified that has both of those interactions in a, you know, in a single antibody. So that's H1 and H2 antibodies. I'm gonna just show you a snapshot of H5N1. So these are bird flu viruses that pass from bird to bird, and they're very high, they're high pathogenicity variants, high path. And so these become systemic because they have an easily cleavable HA, and they'll kill birds very easily, but they also, for humans that have a direct contact with these birds, can become infected and become systemically ill. So we were studying the human B-cell response to H5HA, and around the time that we were isolating antibodies, again from vaccinees at Vanderbilt who were being inoculated in our Vaccine Treatment Evaluation Unit with the investigators here, Kathy Edwards and Buddy Creech, were doing trials with subunit vaccines, and we were taking B cells from those individuals. Around the same time, Ron Fouchier in the Netherlands and Yoshi Kawoka in Wisconsin and Japan, they were trying to understand why does H5 not pass from person to person. They can go from bird-to-person infect, but you don't have transmissibility between humans. So they passaged viruses in ferrets serially, and these viruses adapted to ferrets, and ferrets have similar receptors allowing adaptation essentially to a human-type transmissibility. And so, of course, these were very controversial papers in which the the regulatory authorities recommended holding and not publishing this information. Eventually it was published, and the mutations that occurred in those HA that facilitated respiratory droplet transmission were very few. There were only about 4 mutations in these viruses, which number one gives you pause because these are definitely occurring out in, in bird populations in nature, but also raised a lot of discussion about dual use and that sort of thing. So I think the scientific community had decided that because these occur in nature, we need to understand how they work, and that's why they were published. So if you look at, in green, these transmissibility mutations there on the rim of the receptor binding domain, and there were previously identified major antigenic sites from urine antibodies at Udal and and others in Walter Gerhardt's lab had done way back in the day at Wistar. So mouse monoclonals have been used to identify, you know, major sites that recognized by monoclonal antibodies. So we started making human antibodies to H5, and we found the one that I've already told you about that has the aspartate that points in the wrong direction. This antibody was interesting because Andrew Ward's group was working with us to do single particle EM reconstructions, and we were mapping the epitope and paratope of this antibody, which was neutralizing and had HAI activity. And when we were able to make the variant or mutant HAs that had the transmissibility mutations on the receptor binding rim, we realized that the CDR CDRs of the antibody did contact the receptor binding domain shown in orange here. And by DXMS proteomic studies, we showed this peptide was contacted, and we actually made an escape mutant for the escape mutant virus using reassortant viruses. So all of the contact residues were sort of in this blue zone, and it avoided interacting with the mutations that mediated the transmissibility. So in fact, this antibody from a conventional subunit vaccination protocol is able to neutralize not only conventional H5s but also the ones that were made to be transmissible. So I think this is very reassuring that conventional vaccines would probably be able to cover such, such viruses if they occurred in nature or, if people who, uh, had bad intentions made them in a lab and released them on purpose, which, uh, hasn't occurred obviously. So, uh, now I'm going to talk about the H3 variant viruses. We've made, uh, human monoclonals again from vaccinees in our experimental unit here, uh, who were inoculated with H3 variant, uh, subunit vaccines. So not the conventional seasonal ones that we've all been getting but experimental vaccine that incorporated the H3 variant HA. And when we got— these are 3 representative monoclonals from those vaccinees. If we look at the reactivity of these 3 monoclonals shown in 3 colors, they do not recognize the more recent seasonal viruses that are the progenitors for the HAs that are used in the seasonal vaccines we're getting., and yet these antibodies recognize viruses from the late '80s through 2009. So there was a period of circulation of H3 viruses, um, that, um, that are antigenically distinct from the ones that we have now, which is why we may need an H3 variant, um, vaccine or therapeutics or antibodies. And it also is very similar to the story with 1918 H1, in that it looks like what actually happened is H3 viruses that were in humans, and during this period entered into pigs, and so those are the viruses that are re-entering the human population now, and we're calling them swine variant viruses, but, you know, very likely they could be human viruses that got into pigs, have sat there. Now the humans no longer have herd immunity, to the older H3s unless— if you're an older person, you do, but if you're a kid, the herd immunity in kids is to the more recent ones, not the older ones. That's why the kids have fatalities, whereas the older folks do not get severe disease with H3 variant. So these types of monoclonal studies elucidate not only just fine molecular details, but they also can be used to interpret the epidemiology of the viruses. The other use of antibodies like this, monoclonal antibodies, is to understand mechanisms of neutralization. So I've gone into detail about receptor binding domain antibodies that block receptor binding, and if you block attachment of a virus to a cell, that's probably the most efficient way. There's also stem antibodies that bind down the bottom, of course, and they inhibit fusion. So the HA is a metastatal protein and has to open and extend the fusion peptide and then fold back with heptad repeats. It's a very canonical way of doing membrane fusion in many viruses, and even in some host proteins use this. So the fusion peptide is tucked down at the bottom, so stem antibodies do mediate some fusion inhibition. They also, and that's probably the basis of their neutralizing activity, although most stem antibodies, you know, on average are not very potent antibodies. Some rare ones have high potency, but the typical stem antibody is low potency, like a microgram per mL IC50. And therefore, there are some limitations on what you can do with stem as a vaccine antigen. And I personally don't believe a stem-only vaccine would be sufficient to give high-level protection in humans over time. It probably needs to be a component of broad immunity, but the typical antibody you get up to the stem is not that potent. So there's neutralization activity from fusion inhibition, but there's also ADCC activity and other FC-mediated activities that are important for stem antibodies. So this is a third class of antibody that binds to the side of the HA head, which is relatively conserved actually. And this binds to a region that some have termed the vestigial esterase domain because there was a putative legacy enzyme site that's no longer used. So this particular antibody, which we got again from an H3 variant, vaccine trial binds to the side of the HA head to the vestigial esterase domain. We've recently published this in Nature Communications if you want to look at the structure. And again, Ian Wilson's group with Peng Zhang made the structure, and Sandeep Banguru and our group made the antibody. Interestingly, if you look at how the mechanism by which this antibody works, it has a lot of activities. It has ADCC activity, It does neutralize, but one of the interesting things is that it doesn't block attachment because it doesn't block the receptor binding domain. So virus attaches and goes in, but it inhibits virus on the way out. So these are transmission electron micrographs, and you can see influenza particles budding off of the apical surface of infected cells But in every case, by Immunogold labeling, we see the antibody present right at the scission point where normally the particle would butt off, but we see these particles sort of stuck and they're tethered. And the antibody, you know, we always see it in the junction between the cell surface and the particles. So we believe that this antibody is actually not blocking attachments, blocking egress of the virus from infected cells, and it's probably cross-linking HA on the cell surface and on the particle. It's sort of a tethering mechanism, which is a new mechanism of neutralization of flu. So that's— I've gone through H1, 2, 3, 5. What about H7 viruses? So again, these are avian viruses. We've used a number of techniques to map the epitopes, and I've already shown you single particle EM reconstructions from Andrew Ward and co-crystals from Ian Wilson's group. This is a— these are data from a different type of experiment called hydrogen-deuterium exchange, or deuterium exchange mass spectrometry, in which you You take HA, and in each case, this is a representation of one protomer of the HA trimer, and you label that with deuterium in solution, and then you do mass spec, and then you do a separate experiment where you add an antibody FAB, you make a complex, and then you label, and whatever peptides disappear in surface labeling in the presence of the antibody are interpreted to be the epitope. We just pseudo-colored on these representations the protomer in blue, the, the DXMS peptides that we identified. This was done with Sheng Li at UCSD, who's a proteomics expert. We also made point mutants, escape mutant viruses, which are shown in pink. And very gratifying that the critical residues, which are the escape mutations, They also reside in or near the dXMS peptides. So these are the epitopes. And so for 1, 2, 3, 4, 5 different human antibodies from H7 vaccinees, we see that there are various places on the HA head that you can bind and neutralize. So these are all binding to various rims of the receptor binding domain. And this, this antibody H75 is actually quite interesting because if I showed you the side view, you couldn't even see the epitope. It comes in from the top and it's extremely interesting antibody with some very unusual properties that we're trying to figure out right now with our Scripps collaborators. Most of the potent neutralizing antibodies that we obtained from H7 vaccinees bind to the receptor binding domain, just like for the other subtypes. So this is FAB of a human monoclonal called H7.137 in complex with H7HA of the trimer. And again, you have a heavy chain on the right and a light chain on the left, and we can see exactly how the the antibody loops interrogate the receptor binding site in a pretty canonical way. So this is again the most potent and effective way to neutralize flu, and this is where you get your most of the protection probably from subunit vaccines is this type of antibody. The problem being that these loops around the receptor binding domain quite variable. So for H7, there's been various what are called waves of virus because you get antigenic drift. And there was a very severe fifth wave last year, raised a lot of concern. There's even a sixth wave ongoing now. Jesse talked about that last month. So finally, what I've shown you, I, you know, I would say are relatively mid-throughput techniques in which we're taking blood from subjects and making B-cell lines and screening through those with liquid handlers and robotics. And typically we end up with dozens, maybe 50 or 100, uh, human monoclonal antibodies when we do that. And other people do very similar scale things. But I think in the future we're going to get a comprehensive or near complete mapping of influenza repertoires because of the techniques that have come online, uh, and maybe the last 3 years or so. So we're doing a project in the Human Vaccines Project, which is sponsoring this webinar actually, and our, our work has been termed the Human Immunome Project. And what we're doing is taking large amounts of blood cells from people, we leukapheresis them, so we put a large needle in one arm and a large needle in the other. We basically filter all their blood 5 or 6 times and remove all of the circulating white blood cells. And we can obtain about 40 billion white blood cells by doing that. It's actually a very safe procedure. And the people recover their counts very, very rapidly, because not all your blood cell— not all your immune cells are in your blood. Of course, most of them are in your bone marrow and other tissues. But we remove about 40 billion, and we sequence those samples for heavy chain and light chain sequences, and also for TCR alpha and beta sequences. Until we don't get any new sequences back, essentially. We started representing these VDJ combinations so we can understand how complex is a person's repertoire. And there are tens of millions of clonotypes, and for some of those clonotypes, we've seen as many as 60,000 somatic variants for a single lineage. So the human immune system, even at one time point, is incredibly complex and very diversified. And this is great because the complexity we see in human immunomes is far greater than you would see in a circulating influenza phylogeny like Jesse described. So we're always worried because flu is so hypervariable, it's always drifting and shifting, but the complexity we see in the human immune system far exceeds that. So if we can understand the rules of how this stuff works, we ought to be able to make vaccines that make very, very broad or universal responses. So that's kind of the basis of what's going on. If we make a monoclonal by the conventional means that I showed you before, and then we do this sort of repertoire analysis from the same individual or even other individuals, we can take our monoclonal recombined antibody variable gene sequences and blast them essentially against these large repertoires. And when we do that, We can identify lineages. This is— we're not the only people doing this. Of course, it's been done in the HIV field to find broadly neutralizing antibodies and study the evolution of the lineages over time. So this is an experiment in which we, we studied a person who got a seasonal vaccine that was either 3 or 4 subunit vaccines, the typical vaccine that you would get as an annual shot, and bled that person on multiple time points basically every day for a month, and each color here in this lineage is a different time point, and each size of the nodes is the number of sequences we got back in the next-gen sequencing. And this little triangle is a single monoclonal antibody we got by conventional means that's basically pan-influenza A. It's a very broad antibody. It's not stem, it's a head antibody, and we see somatic variants of that antibody in the repertoire at multiple time points, What's amazing, now that we have the next-gen sequencing technologies, there's also been a revolution in synthetic genomics. We're currently synthesizing thousands of antibodies that we get from the sequencing, and we can make these as recombinant proteins. We've been doing experiments. The current experiments are several thousand antibodies from a single sample. And by doing that, we can look at the effect of somatic mutation within a lineage and then look at lots of lineages. Another technology that's really going to blow this field up is single-cell technologies. So that's important to lots of biology people doing single-cell transcriptomics and so on. So you can use a flow cytometer to identify B cells with markers like CD19 and CD27 you can identify even plasmablasts. So if you have a subject who was infected in the last several days, approximately 7 days, they will have circulating plasmablasts for a short period of time. Or you can identify B cells by CD19 and add a fluorescently tagged or detected antigen. So In this case, this is influenza or some type of glycoprotein that's been— influenza HA or NA, or you could use any, you know, viral protein. So the plasmablasts declare themselves as specific to whatever vaccine or infection you have recently received. The memory cells are pulled out by antigen labeling. Either way, you can take these cells and put them into a single-cell device, of which there are many coming along. Some are sort of DIY that you build yourself, but the 10x Genomics device is the one we're currently using. And you can drop in 10,000 cells per chip essentially and get individual sequences of heavy chain and light chain linked. 10x actually has proprietary bioinformatics tools, so it just kicks out the sequences of your heavy and light chains. We then are using these to synthesize all of those sequences, put them on our liquid handlers for DNA preparation of plasmids, and then we transfect in 96-well plates. We actually do 96-well Protein G purifications. So this has allowed us to scale these types of experiments to where we're interrogating thousands of antibodies at a time. So I think the next-gen sequencing, synthetic genomics, and single-cell technology are going to allow people to, to make tens of thousands of antibodies at a time. I mean, that's, that's not a fantasy, that's happening right now. I think one of the limitations of these things is you get a lot of antibodies and you have to deal with them. In the plasma blast technique, we've done a lot of this after vaccination or or infection, and this was a natural infection instance where our group recently expressed, I think this is just 100 antibodies, and you can see it's great. Well, you get HA antibodies, 11%, 25% of them are NA. We looked at NP and M1 as a minority, but a lot of them are probably to other influenza proteins. If only 10% of your antibodies are HA, that's a relatively inefficient yield of HA antibodies versus sorting with HA directly. Plasmablast is an interesting technique, but it has some limitations. You have to just think through, are you trying to catalog the whole response and understand the complexity to HA, NA, M1, MP, et cetera, or are you trying to develop a therapeutic antibody? What is your purpose? I think that guides the selection of the tools that you would use. So I'm gonna end the talk just with some summary points. And first, what I've gone over today, at the beginning I talked about canonical modes of interacting with the HAA or the stem, either driven by particular amino acids like aspartate or aromatics, or the stem, BH169, there's also BH330. The most potent neutralizing antibodies are the head domain. The broad antibodies you see typically are the stem, but now that we're doing very high-throughput screening, occasionally we can see universal or broadly protective antibodies that are to the head domain. And certainly, I talked about 100-year-old people, healthy adults, children— your prior exposure greatly determines the, the types of B cells that you have available to respond in the memory spots, and you may or may not see particular antigens depending on your prior exposure. The other thing— I just give you two more summaries. One is, you know, what's the purpose of doing this sort of academic work? Well, we're trying to understand not just the sequence diversity, the antibody response, or or viruses, so you can sequence viruses, but the viruses encode proteins, and those proteins serve not only as structures but antigens, and how to understand the magic of what an antigen is and how the immune system recognizes that. So the antigenic landscape is basically what we're trying to do with most of these. Once we have monoclonals, we can determine various mechanisms of immunity in vitro or in animals, even in humans. Some monoclonal antibodies have been administered to humans followed by challenge, so you can understand, do antibodies even mediate in vivo protection. And the mechanisms of immunity work is meant to connect to new correlates of immunity. So the only correlative immunity that's currently accepted by the FDA for vaccine licensure is the hemagglutination inhibition assay, which only measures antibodies to the receptor binding domain. It does not measure antibodies to stem or any of these other alternate mechanisms like the egress inhibition that I showed you. So if we can fully delineate the mechanisms of immunity, develop laboratory tests that measure those things, and then validate those tests using human samples, and studies, we might be able to develop new correlates beyond HAI that would be very helpful to the universal vaccine effort in the world. The other exciting future direction, I think, is once we have structures of an antibody on an epitope, you can essentially throw the antibody away and now use the epitope for computational design, a rational process in which you do structure-based design of an epitope and display that epitope on some kind of particle, like a nanoparticle where you make it multivalent. And this has been termed reverse vaccinology. So you start with the end in mind. You, you think, okay, I know the type of antibody that mediates protection, I want that antibody, so what are the epitopes that induce those antibodies? And so you make the antibody, make the structure, computationally designed, and then express that epitope and display it in a multivalent platform. And reverse vaccinology, I think, has great promise, but it, it's not a technology that's hardened off. We're not using that for commercial vaccine development right now, but I would say in 10 years that might become the main way that we do vaccine development. And then finally, as we are making these human monoclonals, In a way, they're reagents for all the stuff I just said above, but in fact, I believe antibodies could become really the principal way that we deal with all infectious diseases, especially emerging things, more so than vaccines. And the reason is vaccines typically, if you look at the data, they take about 50 years of development before licensure. So the, for instance, the flu mist vaccine that was a live attenuated cocktail vaccine that was intranasally delivered was essentially invented and generated when I was born at the beginning of the 1960s, and it took nearly 50 years to get that vaccine licensed, and then it's been on and off recommended for use since licensure, which it's only been out for a few years. So vaccines are actually quite difficult to develop and get over the bar, but antibodies have a pretty straight path. In fact, more than half of the drugs in the pharmaceutical pipeline right now are antibodies for cancer, immunotherapy, etc. There's a very high level of experience with the safety of antibodies they can be used in all populations from a newborn to a 100-year-old person. They can be used in immunocompromised, which is not really true of many vaccines. And you could say, well, antibodies are very short half-life, but MedImmune and Roche and others have developed Fc mutations in antibodies that extend the half-life. And MedImmune has actually shown in humans they can achieve approximately 90-day half-life. So if you have a 90-day half-life of your molecule, and you, you find molecules that are, have a very high level of potency, effectively the higher you get, the longer your functional protection would last in vivo. And if you can extend that molecule to 90 days instead of 21-day half-life with mutations, then you can envision having 6 to 12 months of passive protection from an antibody. And this exceeds what we currently achieve with seasonal subunit vaccines, which very often only have a duration of protection for 3 or 4 months. So I think it's entirely realistic that antibodies could replace vaccines for influenza prophylaxis. So I've mentioned a lot of projects, a lot of molecules, structures, techniques today. We've done a lot of work in collaboration with especially the Ward and Wilson labs at Scripps. Mount Sinai group we did a lot of the avian work with. Work I did not talk about today at ECC with a group in Melbourne, Stephen Kent's group. Our VTU collaborators and computational collaborator Jens Smiler, Ben Spiller, crystallographer, former members of the lab who made antibodies, and then our group now. This is a picture of our group We've also gotten funding from the Human Vaccines Project for the human immunome work, NIH, DITRA, and DARPA.

So I'll end there, and I'm happy to take questions. Thank you, Dr. Crowe, for that wonderful presentation. So if everyone would please type their questions into the question box, we'll give everyone a minute to do that, and then I can read them to Dr. Crowe, and he will answer them.

Thank you. We stunned everybody with an hour of influenza.

So while we're waiting, Dr. Croce, do you think that a universal flu vaccine should be actually more likely to be a universal flu antibody? I know antibodies are very specific, but if you could create like a cocktail of them or something, do you have thoughts on that?

Yeah, I think, honestly, I think it's more realistic in the short term to develop a cocktail that would be a universal antibody cocktail because we have a lot of these molecules in hand. So if you, if you combine a very broad head antibody and a very broad stem and maybe, uh, one or two by alternate mechanisms, you could rationally, uh, cover all the strains. I think part of the difficulty when people start talking about universal is what does that mean? So broadly protective, are you trying to get all H1s and H3s and Bs that are circulating now, or are you trying to cover also the pre-pandemic threats like H5 and 7 and 9, or are you trying to get even H17 and H18, which are only in bats, and we don't even know if they even ever cross over to humans. So you have to define your goal, but within H1, 3, and B, it's entirely plausible that we could put together a cocktail of antibodies that would have a universal coverage. Making an antibody in Making a vaccine to induce those is plausible, but, you know, again, it's sort of years down the line. I think a lot of people are going to be trying to do that.

Great, thanks. We have our first question. The question is, is the repertoire of the mucosal antibodies different from that in the serum?

So the Well, let me clarify. That's a great question. So let me clarify some vocabulary. So repertoire, what do we mean by that? When I use repertoire, I'm typically talking about nucleotide sequences that are derived from circulating B cells or plasma cells, or we've actually looked at body tissues. If you take airway or gut and things like that, you can extract antibody RNAs and sequence those. Typically, I'm using repertoire to mean the gene sequences that encode antibodies. Some people also use the word repertoire to mean what are the various specificities or peptides. So what they really mean is diversity of antigenic recognition. So I'm not sure exactly how you're thinking about that, but— and then most of the sequencing techniques that we're using are taking blood cells are making RNA and then sequences. You can also take antibodies and sequence the proteins, and this is a new technique in the field. So Cell Signaling Company researchers, George Georgiou and others, have reported techniques, and we've actually been doing these where you, you take a plasma sample or a serum sample and and you run that on a column, say an HA column, you capture the antibodies that are specific to that protein, and then you get rid of all the other antibodies, and you elute that, you cut it to an FAB, and then you mass spec sequence it. And you can get the sequences of the antibody proteins that are in serum or plasma. And these are thought to reflect the, the antibodies being secreted not in the blood, so the The circulating cells in the blood are memory B cells. They're not secreting antibody at all. They're memory B cells. So the antibody that's in the serum plasma is being secreted in the bone marrow by long-lived plasma cells. So you can either go into the bone marrow and obtain plasma cells there, and Rafi Ahmed's group at Emory's been doing this to sequence or isolate or do antibody secreting assays. Or people do believe that you can use the serum or plasma antibody sequencing of the proteins to reflect the same thing. And it's actually easier to get serum or plasma, it's just that the techniques are more difficult. So that's the setup, the vocabulary I wanted to go over. So if you compare antibody protein sequences with the B-cell repertoire nucleotide sequences in a single individual, We've only done that not in influenza so much as in Ebola. We have a whole series of studies like this with Ebola, and what we've seen is there's a partial overlap. So you can see some of the antibody protein sequences also in the memory B cells, and so we don't know if there's different repertoires in the plasma bone— long-lived plasma cells in the bone marrow. And circulating memory B cells. Or, you know, alternatively, these proteomics mass spec sequencing techniques, they're not that efficient and you don't get a lot of depth. So we think, you know, the partial overlap might be in part because we're not seeing all the antibody protein sequences. But certainly there's at least some overlap in the Bola studies that we've done. When you look at what others have done, George Georgiou and others, You know, one of the interesting things about the proteomic sequencing is that there's maybe 10 to 100 dominant antibody sequences in those sets. So it may be that some antibodies are overrepresented, and when you do that, we look at that, they're quite limited number of clones and specificity. So it's a great question. Do the serologies that we do, do they reflect the memory B cells and the the other assays that we're doing experimentally or in clinic. I think, sort of rambled a long time there, but I think that remains to be seen. I think there's some correspondence between memory B cells and serum or plasma antibodies, but it's currently not known to be a complete overlap.

There may be some differences. Thank you, Dr. Crowe. In the last minute, we have one more question. The question is, Would monoclonal antibody-resistant viruses, for example, escape mutants, render prophylactic approaches obsolete within a single flu season?

Yeah, that's a great question too. Essentially, this is the problem with current seasonal vaccines. So when people talk about a match, you get in the population, we pick a strain, you know, something like 9 months before we want to be giving it to people, and we rush, rush, rush, and those strains are picked by experts who are gazing at sequences that occurred in the past year or in other continents, and by the time we manufacture those vaccines, often circulating virus has already moved on. So this is not just a monoclonal issue, it's a polyclonal issue in which the circulating viruses can drift away from the entire polyclonal immune response of humans that's induced by prior vaccines. Scott Hensley's done a lot of work at this, and, you know, the question is, if you have a bunch of antigenic sites, how can one mutation escape the polyclonal response? And Scott's group has shown that over time there are immunodominant antigens, and you tend to focus on, say, one particular site, and at that point, then the virus can escape from that. So viruses escape polyclonal responses already. That's what antigenic shift, or antigenic drift, excuse me, is all about minor changes in an immunodominant site. So if you then say, well, what about monoclonals? A single monoclonal antibody would be very susceptible to that type of escape. And I already showed you in H7 work that I showed you in the lab, we can take an antibody and a virus, and typically we can make an escape mutation which only takes a single point mutation. Now, having said all that, there are places on the virus that are critical for the virus, you know, attachment, conformational changes, fusion, etc. And so there are places on the virus where an antibody can recognize, and escape mutation may occur, but that virus may be debilitated so that it does not grow like a wild-type virus. It's actually attenuated in growth, and. And if it were completely attenuating, you might find, you know, this is sort of a fantasy of people like us who make monoclonals, you might be able to find an antibody whose escape mutants don't grow. Like the escape mutation would disrupt the HA protein in such a way that the virus could no longer replicate. And so you definitely see attenuated mutations in escape variants. That's pretty typical. Most monoclonal antibody-resistant mutants are attenuated for growth in vitro and animals. But, you know, probably some would completely inactivate the virus. It just has to be looked at when we make escape mutations, put that virus back in as a single mutation by reverse genetics, and then do single-cycle growth curves to determine is the escape mutant attenuated or not.

So that's how we approach that question. Thank you, Dr. Crowe. And I wanted to thank everyone for participating in this webinar. As a reminder, on September 18th, we'll be having our next webinar with Dr. Paul Thomas from St. Jude. Thank you very much, Dr. Crowe, and I hope everyone has a wonderful rest of their day.

Bye-bye. You guys.