Scientist Stories: Aaron Ring, Autoantibodies in COVID-19

started. Hello and thank you to everyone around the world for joining us for this week's lab meeting. My name is Kristin Aboud. I am the science editor at the Human Vaccines Project and I will be your moderator today. Like most weeks, there is no shortage of news on COVID vaccines. On Monday, Pfizer and BioNTech reported that their mRNA-based vaccine is both safe and highly effective in children ages 5 to 11. This finding was based on results of their trial involving more than 2,200 kids. The companies say they will submit their application for emergency use authorization by the end of this month. Experts say this could make it possible that the US FDA will greenlight vaccinations among this age group by the end of October or early November. This news will come as relief to many parents in the U.S. where infection rates are soaring among children. Internationally, good news came from India this week where the country's health minister said they would resume exports of domestically manufactured COVID vaccines starting next month, helping to fulfill their commitment to COVAX, the UN-backed effort to deliver COVID vaccines. This would bolster vaccine availability in many countries where supply remains extremely limited. After a slow start with vaccinations and a devastating wave of infections due to the Delta variant earlier this year, India has made tremendous progress with its vaccine distribution. 61% of adults in the country have now received their first dose, and the government says it will finish vaccinating all 944 million adults in the country by the end of this year. Access to COVID vaccines was also a prominent theme this week at the annual United Nations General Assembly meeting. Before we begin, begin today's presentation, just one note. The information presented today includes some pre-published data that is currently under peer review. 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 very happy to introduce today's speaker, Dr. Aaron Ring, Assistant Professor of Immunobiology at the Yale School of Medicine. Dr. Ring's research focuses on using directed evolution to create new pharmacologic tools and therapeutics against immune receptors, as well as to tune immune cytokines and growth factors for defined activities. His laboratory has developed new technologies to detect functional autoantibodies against the exoproteome. Dr. Ring is a recipient of the NIH Director's Early Independence Award, a Milstein Young Investigator Award from the International Cytokine and Interferon Society, And he was also named a Pew-Stewart Scholar and Robert T. McCluskey Yale Scholar. During his 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. With that, it is my pleasure to now introduce Dr. Aaron Ring.

And you have to forgive me, I'm just looking for the screen sharing privileges. I got it now. Perfect. Right. Well, it's a real pleasure to speak at this, you know, this venue today. Tremendously excited about this new research direction in my lab to look at autoantibodies in, you know, post-acute infectious setting. I have to tell you, you know, if you'd asked me 2 years ago, 3 years ago, if we'd be working in this area, of course I would have said no. I'm just a a Yeoman cancer immunotherapist, you know, trying to engineer, you know, new immunomodulatory drugs for cancer. But lately, we've had a shift in our research interest, or I should say rather, you know, a new direction of research interest, which is we've gotten interested in the natural pharmacology of the immune system. That's to say, what are the antibodies that patients are making and how do they influence disease outcomes, not just in cancer, but across many different diseases. But before I get started, I just want to acknowledge that, you know, because I engineer a lot of proteins and my lab has also developed technologies that I'll be discussing, I have been involved in the commercial efforts to bring those into the clinic, and so these are my disclosures. But as I was saying, we've gotten really— having been immunopharmacologists all these years, we've gotten really interested in the natural pharmacology of the immune system. That is to say, what are the antibodies that can influence disease outcomes? And certainly, when you think about antibodies, autoantibodies, that is, you think typically what comes to mind are autoantibodies that drive disease, like in rheumatologic illness, autoimmune diseases, where antibodies target different tissues and cause tissue damage and illness. But there's emerging evidence that not all antibodies are bad. In fact, across a wide range of disease indications, there are numerous examples where antibodies can actually be protective. So for example, in autoimmune disease, patients that make antibodies against cytokines, they have less severe illness in general than those that don't. In cancer, patients who make antibodies that opsonize their tumors, that target their tumors, they have better survival than patients that don't. This makes sense. Like for example, patients with breast cancer who make anti-HER2 antibodies, they're literally making their own trastuzumab. You know, no surprise that they would do better. But even in diseases that aren't typically, you know, thought to be immune-centric like neurodegeneration, autoantibodies have been shown, have been found to be disease-modifying. So for example, in Alzheimer's disease, patients who make anti-amyloid antibodies have more slowly progressing disease than those that don't. Anti-prion antibodies, anti-Tau, anti-synuclein, all examples of what's been found. But I think what you may have surmised from this list of examples here is that all of these examples were really hypothesis-driven. They were, you know, pretty straightforward targets, um, given what we already knew about the biology of these diseases. Um, it's sort of like looking under the lamppost. And so what we've gotten really interested in trying to address is how can we find these sorts of functional disease-modifying antibodies, or disease-driving, you know, finding new antibodies that that are causing disease in an unbiased fashion across the whole proteome. And so one of our premises here is, you know, we actually don't really care about the whole proteome, to be honest. Antibodies are large extracellular secreted proteins, so by and large they're going to come into contact with proteins that are in the same compartment, same extracellular compartment. They don't really cross through the membrane and access, you know, intracellular antigens. So our first, you know, thought was, well, maybe we should focus on the extracellular proteome, the exoproteome. And when antibodies can come into contact with extracellular antigens, they can exert a wide range of functions. So they can, for example, inhibit or activate cell signaling pathways by engaging with either the ligands or the receptors themselves, or they can target molecules for degradation or even whole cells for degradation via antibody effector functions like complement-directed killing, antibody-directed cell-mediated cytotoxicity of natural killer cells, neutrophils, or phagocytosis for macrophages. So that was really the premise we had, is that antibodies that are going to be exerting the biggest impact on our physiology are going to be those that recognize this fraction of the proteome. But it turns out that, that this extracellular proteome has been one of the most difficult to sample in terms of extracellular serological responses, antibody responses. And the reason why is really twofold. First is that most antibodies in general recognize proteins in their native three-dimensional conformation, not, you know, basically discontinuous epitopes that form when a protein is folded, as opposed to a linear epitope. That is, you know, independent of conformation. And that problem is compounded by the fact that extracellular proteins are among the more, most difficult to work with. They have unique folding requirements, you know, with disulfide bond formation. They can be modified with post-translational modifications like disulfide, or sorry, like glycosylation. They tend to be more fragile. You can't make them in, you know, your run-of-the-mill expression systems. And so for that reason, a lot of the conventional approaches for identifying new antibody responses like protein peptide arrays or really innovative approaches like phage immunoprecipitation, by and large, they're missing these epitopes that we think are among the most important. And so over the past few years, we've wondered, you know, can we, can we address this problem? Can we try to develop a system that allows us to detect these extracellular binding autoantibodies comprehensively. And so our first big bet we made was that yeast, the humble yeast, the baker's yeast, Saccharomyces cerevisiae, could capture a lot of the human extracellular proteome. And we've been working with yeast for many years to engineer proteins for directed evolution, and we knew that we could actively display a wide range of extracellular proteins tethered to the surface of yeast, for instance, the Agg1-Agg2 system. And so what we did is we said, well, why don't we try making a comprehensive library that contains thousands of extracellular human proteins tethered to the surface of yeast, displayed on the surface of yeast, such that, you know, each protein would have not just be represented on yeast, but have a unique genetic barcode contained within the plasmid that we could use to track it. And so that we, over the last few years, we've been painstakingly making this library. And with this library in hand, we worked on developing approaches, methods where we can query it for antibodies that recognize these proteins. And so what we've done is we've developed this technique called REAP, rapid extracellular antigen profiling, or exoproteome antigen profiling. And the concept is really straightforward. We just take a bit of patient sample, we apply it to to the yeast library in microtiter plates. And this allows us to sample hundreds of samples at a time in a multiplex fashion. And then we just isolate the yeast that are coated with patient antibody using magnetic separation and then sequence the barcodes. It's that simple. We turn an antigen antibody binding event into a sequencing readout. And what this allows us to do is to basically detect antibody responses against currently around 6,000 different extracellular antigens. In prior versions of the library that I'll present today, it's closer to like 2,800, but our current library has about 6,000. It just takes a little bit of sample, you know, less than 50 microliters of patient serum or plasma. And, you know, because we're sequencing, we can multiplex together, you know, several hundred samples at a time. So I just want to first, before diving into the COVID data, I just want to give you a little bit of background about the system and how it's working and its performance. And so we first developed the REAP system, we first wanted to benchmark it on an existing disease where we knew there should be well-characterized extracellular autoantibodies that were functional. And so we decided to look at this rare genetic autoimmune disease called APoS1. This is a very rare disease. It's caused by loss of the AIR genes. These patients lack central tolerance, and as a result, they get just devastating autoimmune disease, polyendocrinopathies, where they're developing antibodies against different, you know, diverse endocrine tissues. But they also paradoxically get immunosuppressed, and that's a result of the fact that these patients make ultra-high affinity, ultra-high titer antibodies against a wide range of cytokines. You know, notably type 1 interferon, but, you know, other cytokines like IL-17 are characteristic, IL-22, et cetera. So I wondered, you know, could we find these known knowns in APICET? And the answer was a resounding absolutely. So what you're looking at here in this heat map is a summary of all the autoantibody responses we found in these patients across the whole extracellular proteome that we queried. And so each patient here is represented as a column, and each row is an autoantigen that we detected in at least one patient. So first thing I'll say is the data matrix is pretty sparse if you think about it. Here we queried 3,000 proteins, but we only saw autoantibody responses to about 100, right? So that kind of makes sense. You wouldn't expect there to be an antibody against every single protein in every patient, right? But what we did see right off the bat is, you know, these, these pathognomonic anti-type 1 interferon autoantibodies were near universally present in these patients with APS1. We also found the known IL-22 antibodies, IL-17, other cytokines that were known. We found known tissue-targeted antibodies like anti-gastric intrinsic factor antibodies. These cause vitamin B12 deficiency in this disease, and of course many others like pernicious anemia, and antibodies against lipocalin-1 that had been described before. But then we found a bunch of completely novel autoantibodies even in this really well-characterized autoimmune patient cohort. So most of these— sorry, most of these responses were relatively private, meaning we only saw them in like 1 or 2 different patients. But there were some examples of antibodies that were relatively public, like antibodies against this glycoprotein hormone GPHB5, and that was associated with patients who had hypoparathyroidism in this particular disease. So this, this gave us a lot of confidence that the, the system was working as we anticipated. And in fact, not only could we find these, these known knowns, if you will, but we could find a bunch of novel responses that hadn't been described before. The other thing we learned from this data is, is REAP isn't just a yes/no answer. But it actually, it's a quantitative readout. So as you titrate the amount of patient sample on the library, if you look at the enrichment of individual antigens, you can see that you see sigmoidal dose-response curve. So you can essentially calculate virtual titers that, you know, this is really a quantitative readout of the autoantibodies that are present. And it compares favorably in terms of sensitivity for some antigens than, you know, as compared to like ELISA, which is the gold standard. So we thought this was a good, you know, first foray in, you know, into using this technology and gave us confidence that we could detect autoantibody responses against the extracellular proteome. So we wanted to apply it to a bit tougher nut to crack. We wanted to apply it to a much more heterogeneous sort of real-world disease, if you were, and decide to look at lupus. As I'm sure you all know, this is probably one of the most common autoimmune conditions, you know, polygenetic, multifactorial, and best known for, you know, it's characterized by loss of tolerance to nucleic acid and nuclear antigens. So those antibodies against like ANA, antinuclear antigens, very well characterized in this disease. And, you know, of course lupus is extremely heterogeneous in terms of its clinical presentations. But what is less known is the impact of disease-modifying antibodies that target the extracellular proteome in these patients. There are some reports that they— that some of these patients have anti-interferon autoantibodies that might be associated with less severe disease outcomes, but certainly nothing like we, you know, know for other diseases like APOCET. And so we profiled 107 different patients with lupus, you know, compared to healthy donor. And here on this heat map, we have arranged the patients according to their disease severity, the SLEDAI score, or actually disease activity at that particular point in time. And one thing I'll just say, if you look globally, this is a night and day difference with APoSIT. Instead of seeing large patterns, block patterns of reactivity in these patients, you see a starry sky, a constellation of fairly private autoreactivities. Each patient is kind of like a snowflake here. It has their own unique complement of antibodies. And to be honest, this is really the much more realistic picture of what we're seeing across many different diseases, that there aren't really a lot of public antigens. The most common autoantigens we'll find usually be in like the 10% to 20% range, but that many autoantibodies we're seeing are only present in 1% of patients with a particular disease or less. And so this was very instructive for us. But, you know, we still were able to glean some really interesting observations even from this fairly heterogeneous disparate collection of autoantibody responses. So first thing we noticed is that there were some autoantibodies that we only saw in patients who had severe disease, that had, you know, high SLIDI scores. So, for example, patients with antibodies against this chemokine CCL8 or this podocyte-expressed protein endocyanin, also called CD248, were uniquely present in patients who had active disease. And it actually correlated with particular features of lupus. These patients had lupus nephritis, which was, which was pretty interesting. But then we found examples of antibodies that were only present in patients with less severe disease, make us think, you know, hey, these antibodies might be protective. So for example, patients who had antibodies against immunoreceptors any one of these immunoreceptors, they had lower SLEDAI scores than those that didn't. And that suggests, you know, perhaps those antibodies were serving a disease, you know, modifying form, even if the exact function wasn't immediately obvious. But then there were some responses that just were screaming out to us that they were functional and could be impacting the disease. So here's an example of a patient with a very low disease activity score who had ultra-high titer antibodies against the cytokine IL-33. And not only were they present at greater than 1:10,000 full titers, but they were extremely potently neutralizing. So that made us think, yeah, this patient may have been protected by virtue of these antibodies. And in fact, IL-33 is elevated in lupus. It does contribute to autoantibody production. And there even is a preclinical report that IL-33 blockade is protective in a lupus pro mouse model. So by learning from these sort of clinical trials of nature, we're hopeful that we can potentially learn about new drug targets, but also we can learn about new mechanisms, pathology in these diseases by studying these types of responses. Okay, so I realize that's a long setup, and this is a meeting about COVID So of course, I want to tell you what we found in COVID. Now, as the pandemic was hitting, you know, in early 2020 in the United States, we had just gotten this technology off the ground. And so we were kind of primed to think about the influence of autoantibodies in any disease. But, but it became apparent early on that there were some really strong clues that, that antibody autoreactivity could be contributing to the pathoideology of COVID-19. So certainly the role of the innate immune system is unquestionable, that you get this exuberant maladaptive inflammatory response. But, you know, in the original SARS-1, there are reports establishing that patients who had severe disease oftentimes developed antibodies against the lung parenchyma, perhaps due to molecular mimicry. That was the hypothesis put out there. But then, you know, as groups started profiling COVID patients themselves, it became apparent that the humoral immune system might be a cause of autoreactivity. So Ignacio Sanz's group found that COVID-19 patients had B cells in their blood, these extrafollicular B cells, these double-negative B cells that are basically found really only in autoimmune conditions like lupus, that were present in high levels in COVID patients. And in fact, the greater numbers of these B cells in a patient's blood, the worse their disease. And then classic rheumatologic panels, ANA, phospholipid antibodies, other conventional rheumatologic autoantigens were found to be elevated in COVID patients, further influencing— implicating autoantibodies in the disease. And then there was a real just absolute mind-blowing paper of extraordinary significance. And that was this report by Jean-Laurent Casanova's group, you bastard, et al., where they found that patients who had anti-interferon autoantibodies had much more severe disease outcomes than patients who didn't. And this, you know, they hypothesized that these antibodies predated the infection. These patients were walking around with them. And certainly they'd been seen in other conditions like lupus, like cancer, seen not uncommonly. And they were enriched in patients with really bad outcomes. And that immediately really implicated antibodies as having a functional impact in the course of COVID disease. There were other reports of similar mechanisms, like Max Krummel's group at UCSF, where they found that antibodies could also be inhibiting interferon signaling through interactions with the inhibitory Fc receptors. So we were, like I said, a lot of clues out there that not that COVID could drive autoimmunity, but that COVID outcomes could also be influenced by preexisting autoimmunity or immunity that was affecting immune function. And so we're really fortunate at Yale to have phenomenal leadership by people like my colleague Akiko Iwasaki, Albert Ko, Saad Omar, who realized early on that we were going to need to be able to study this disease in patients. And before we were even getting the first patients in at the Yale New Haven Hospital, they were already working to get a protocol up and running to acquire samples from these patients. That's called the Yale IMPACT cohort. And they collected samples from acutely infected patients in the Yale New Haven Health System for several hundred patients. And so we were able to collaborate with them, and we collaborated very closely with the Kiko's lab in particular, to profile a few hundred different patients infected with COVID-19 at different, varying disease severity. Most, I have to be honest, mostly pretty acutely ill because these were patients who were admitted to the hospital by and large. Some of them were not admitted, they were sort of found surreptitiously in healthcare workers that we were monitoring who got infected. They had more mild illness, but, but most of our patients were hospitalized, and many were even in, you know, the ICU. And what we found just globally, just at a high level, is that COVID patients had remarkable levels of autoantibodies. In fact, the average COVID patient, even with mild or asymptomatic disease, had higher levels of antibody, autoantibody reactivities than lupus, severe lupus patients. And some of these patients with severe COVID had numbers of autoactivities that approached what we see in APICET. So that was really striking. This is a lot more auto-reactivity than we've seen in other conditions. And that really piqued our interest. The other immediately obvious thing was that the targets of these autoantibodies, the immune system itself was a huge target. We saw tons of antibodies against cytokines. So we confirmed what the Kastanova group found with the interferon autoantibodies. But we also found many different cytokines targeted, chemokines, cell surface receptors of immune cells as being targets of these autoantibodies. And I should just say, on this heat map, the patients— it looks like it got clipped off here at the bottom. My apologies. But these patients are arranged by their disease severity. So these are the patients on the far left who had the most severe disease. These patients had moderate disease, mild, asymptomatic, and then controls here. And so just going back to the interferon autoantibodies, again, the concept here, the hypothesis is that the patients had severe disease because these antibodies were blocking interferon. And interferon being such a crucial antiviral cytokine, if you didn't have sufficient interferon activity early on in the infection, the virus would just, you know, have, you know, open season to replicate, and that would cause more severe disease. And so, you know, we confirmed that. We, like Casanova's group, found that these interferon antibodies were enriched in severely affected patients. And we're also able to correlate this or associate it with specific clinical features. So we found that patients who had these interferon autoantibodies, they had higher levels of viral titers at initial presentation when they came into the hospital. The other thing we found is if they had these antibodies, they had a much slower clearance of the virus over time compared to patients who didn't. We also, as others have shown, we found ex vivo these antibodies are very potently neutralizing, but being able to show that clinical correlation I think was important that this really did correlate with with an impaired clearance of virus. But, you know, we found, you know, interferon very important, you know, found in about 5 to 10% of patients with severe disease. By no means was the only target of immune-directed autoantibodies. We found that, you know, over 30 cytokines, 2 dozen chemokines were targeted by patient autoantibodies in COVID. So of course, you have the type 1 interferon. We also found a lot of antibodies against type 3 interferon. IL-1 family members, including IL-1 beta, the interleukin-18 receptor, different colony stimulating factors like GM-CSF, type 2 cytokines, TNF superfamily members. A lot of diverse antibodies that you're seeing. And they're present in fairly decent titers ranging from 1:100 to 1:10,000. And they were functional. If you look in signaling assays, we found that these antibodies blocked the function of the receptors. So antibodies that hit GM-CSF blocked GM-CSF. Antibodies that hit chemokines blocked the signaling of those chemokines in GPCR assays. So it helped us, you know, build some conviction that these antibodies could be having an impact. But the problem was this: unlike interferon, which, you know, if you're thinking about antibodies as alleles, like genetic alleles, you'd say the allele frequency of the interferon antibodies was a bit higher, right? Even in— and Casanova's group has a more recent paper where they've actually profiled the prevalence of these antibodies across the general population. And essentially, you know, what they find is that they increase with age, and then, you know, as patients get, get very elderly, you know, they can, they can go as high as like 5% of elderly patients over, I think, 75 or 80 years of age. So they're, you know, they kind of started off at a higher frequency, and because of that, you you can form associations just by the distribution of the antibodies. But for these other targets, oftentimes we only saw them in a patient or two. So even though they were enriched in patients with severe disease or found basically exclusively in patients with severe disease, we weren't able to statistically really establish that they were associated the same way that we could with the interferon. And so what we tried to do was to model the effects of these antibodies preclinically. You know, Akiko's lab has, you know, nicely established murine models of coronavirus infection. These are the standard K18 ACE2 human-expressing ACE2 mouse model. And what we did is, you know, we— they infected the mice with the coronavirus, but then prior to infecting them, we would pre-dose them with murine surrogates of these autoantibodies and then looked at the impact, you know, on the outcomes. And you can see if you block the interferon pathway, the mice do much worse. But we also found that, you know, if you block IL-18, we found IL-18 receptor was a target. If you block IL-1 beta, IL-1 beta came up several times in patients with severe disease, IL-21, GM-CSF, that the mice did worse. So by impairing different features of immune responses in these mice, these antibodies exacerbated COVID. So again, you know, we're trying to, you know, build more evidence for these rare responses we're seeing, maybe in like 1 in 200, 1 in 250 patients. And certainly it would be very valuable to profile an even broader set of patients across more diverse cohorts. But I think this type of data does help us bootstrap our way to being able to say more confidently that these antibodies were impacting function. But, you know, it wasn't just immune signaling molecules that were targeted by these antibodies. We also found that immune surface molecules were highly prevalent in COVID patients. So, for example, we found that many patients with severe disease had antibodies that targeted their B cells. So here's an example of a patient who had anti-CD38 antibodies. And my first question to my team was, check that this is not a cancer patient, a myeloma patient. This— the titers are so high It made me think this was a drug like daratumumab. But this patient had no history of cancer, no history of a drug infusion. Instead, had extremely high titers of these CD38 antibodies. And in fact, this patient's serum potently drove phagocytosis of B cells ex vivo. And when patients had these types of antibodies, they had lower levels of B cells in their blood. We had matched flow cytometry data from this cohort.. And as a result, these patients also had lower antibody responses to the virus. So basically, these patients were endogenously depleting their B cells and therefore had impaired humoral responses. And this is what's seen in iatrogenic B cell depletion in patients who have autoimmune disease who are getting antibodies like Rituxan. They have similar bad outcomes with COVID and that's as a result of impaired humoral immunity. But it wasn't just B cells. We found similar responses to monocytes. Similar responses to patients with antibodies against T cells, where CD3 epsilon, where they had specifically lower T lymphocytes, but not like CD4, CD8, NKT, but not NK cells were normal, B cells were normal. And these antibodies again were functional ex vivo, they drove phagocytosis. So basically antibodies that targeted the immune system either by interfering with signaling or depleting immune cell populations were associated with worse outcomes in patients. But it wasn't just the immune system. We found a wide range of antibodies against many different tissue autoantigens, against antibodies, antigens expressed in the central nervous system, vascular and blood system, platelets, coagulation factors, connective tissues, very, very diverse set of antibodies. And I should say, you know, because we were studying these antibodies in the context of an acute infection, these patients had a lot of reasons to be sick. It wasn't typically possible to try to associate particular antibodies with acute disease outcomes in these patients, but we were able to draw some inferences. So for example, we found that patients who made antibodies against the hypocretin-2 receptor— this is also called the orexin-2 receptor— that was associated with having worse neurological outcomes, very crudely here measured by Glasgow Coma Scale. But we also found these antibodies, we validated that they truly were binding the GPCR and they were inhibiting the function of that receptor. I should say parenthetically, these antibodies had previously been seen in narcolepsy and also had been associated with narcolepsy induced by the Pandemrix vaccine. Some years ago. So, you know, really intense interest from some of the folks who previously worked on these antibodies before. But, you know, of course, when you see antibodies against these different diverse tissue antigens, it makes you think, could these types of antibodies, could be, could drive, you know, long-term sequelae of COVID you know, the long-hauler syndrome, post-acute COVID syndrome Long COVID, etc. So we're working on that now. I don't have data ready to present on that, but I will say, you know, we have been profiling patients with Long COVID. But I want to say we do have longitudinal data on a lot of these antibody responses, and we're able to classify antibodies depending on their temporal dynamics. So some antibodies, we saw them very early on in the patient's disease course. They predated the appearance of the spike antibody. So for example, these type 1 interferon autoantibodies, high titers at the very initial presentation, we would surmise then that they're likely pre-existing. We can't prove it for most of these patients, we don't have a pre-infection baseline, but it stands to reason that they're likely pre-existing. In other cases, we clearly saw antibodies come up while the patient was in the hospital. They weren't present initially, but then they came up.. And in some cases we saw antibodies that came up or that waned or that were present, you know, at higher levels early on and then went away as the infection carried on. So we saw all sorts of diverse antibody temporal dynamics. I will say now, and again, not 100% ready to present this, but I'll just summarize it. We've now looked at these exact same patients that we saw over a year ago, you know, in the initial phases of the pandemic. Some of them now over 300, 365 days post-hospitalization, you know, remeasured their autoantibody levels. And, you know, some good news is that many of the antibodies do wane to background, but it's not all good news. And there's, there are many, many examples of antibodies that are, that are newly acquired from the infection but remained at high levels, you know, even now about a year out from their initial infection. And so we're really intensely interested in understanding are those antibodies pathological? Are they driving some of these long-term sequelae of the disease? So just to summarize, you know, what I would have— in wrap-up, hopefully I've been able to relay that this REAP technology is pretty robust to detect these interesting extracellular autoantibodies, that we were able to benchmark it in autoimmune diseases like APOCED and lupus, but that in COVID what we found was that there is a huge array of autoantibodies directed against the extracellular proteome that they impair with immune function in virological control. And they also can impact tissue function, potentially impact tissue function, and provide some clues into longer-term sequelae of COVID I will say some ongoing areas for us, we've been looking at You know, are autoantibodies, are they related to vaccine-related adverse events? Do autoantibody levels change with vaccination? We're working on a report now there. It's by and large pretty good news. We don't really see new autoantibodies come up post-vaccination with mRNA or Johnson Johnson. And we even don't see new autoantibodies form in patients who have pre-existing autoimmune disease who get vaccinated. So that's been good news. We also haven't really seen autoantibodies in patients who get myocarditis and other vaccine-related adverse events. And, you know, as I said, you know, a few times now, we're intensely interested in actively profiling autoantibody responses in, you know, in patients who have long COVID. So before I, you know, answer any questions, I just want to acknowledge that the folks that, you know, who led this work, I had a trio of really talented students in my lab who developed the REAP technology and applied it to COVID and other diseases— Eric Wong, Ila Dai, Jill Jaycox. We had just tremendous collaborators— Kiko Iwasaki and her two students, John Klein and Tianying Mao, the Yale IMPACT team, Albert Ko, Nathan Grubaugh, Sadomar. You can see their names on the left there. A really phenomenal team. And of course, you know, our, our support.

So yeah, thank you for your attention. Thank you so much, Dr. Ring, for that fascinating presentation. We'll kick off with some questions. So I guess one question is, how unique is this, and how do— what do we know about autoantibodies that are stimulated by other viral infections?

Yeah, so what I'll say is there is no silver lining to COVID, right? Absolutely none. It's like one of the worst things that's ever happened. However, it has presented an opportunity to study the immune consequences of acute infection, just because there's been so many events. And a lot of groups had the foresight to collect samples prospectively. And so we now, I'd say we have a better ability to study COVID than any other viral infection to date. So the answer is, you know, we have not, we have not yet applied REAP to other viral illnesses. We are looking to do that now. We're linking up with some groups who do have repositories from people infected with flu, infected, you know, with other respiratory viruses, other pathogens like bacterial pathogens, etc. And certainly there's been a longstanding relationship between infection and autoimmunity that's long been hypothesized and many examples found. But so the question is, is this unique to COVID? No, this general phenomenon is not unique to COVID. I do think it is striking though. I would, without hard data, having not compared quantitatively other infections the same way we have COVID, it does seem striking just how prevalent autoantibody formation is. In COVID patients. And so it does make you wonder, is it more common in COVID than other infections? There are some clues, I think, that it might be. You know, Ignacio Sanchez's group has shown that a lot of the antibodies that are used to recognize the viral antigens like spike, a lot of them are using frameworks like the VH434 framework, which has an innate propensity toward, you know, poly-reactivity, self-reactivity This isn't really molecular mimicry per se. It's just saying that the antibody repertoires that are being used may be somewhat predisposed toward autoreactivity. That could be a feature of the virus. But yeah, it's too soon to say. And I just want to be clear, we're not at all saying and haven't said that this is COVID-specific. We're just saying we're seeing it in pretty alarming level in COVID.

Right. So I guess then maybe this is a related question, which may have the same answer, but do you think there would be any overlap between autoantibodies found in severe COVID and those associated with sepsis?

Yeah, we haven't, we haven't looked in sepsis yet. So I don't know. There probably is going to be some overlap. You know, one of the hypotheses for why we see so many immune-directed hits is that those are the antigens that are present in the inflamed milieu. And so some of these extrafollicular B cells, you know, that's— those are the antigens that are present, and that's why they're expanding. It's a very attractive hypothesis that makes you think, yeah, you know, they're probably in some of these inflammatory conditions, you could see a similar immune milieu. But, you know, like I said, we haven't profiled them yet, so we don't know. But it seems reasonable that there could be some overlap.

And are any of the autoantibodies cross-reactive with specific SARS-CoV-2 proteins?

We haven't looked at that ourselves. Other groups have looked at individual cloned antibodies that, you know, that are shown to recognize spike or nucleocapsid or other antigens and then profiled their polyreactivity, usually against things like you know, just sort of like gold standard poly-reactivity assays, insulin, you know, phospholipids, etc. And some of those antibodies have been poly-reactive, and it's really, it's actually kind of crazy how frequently the anti-coronavirus antibodies are very close to germline. They're kind of almost like, you know, there seems to be some innate repertoire of antibodies that can recognize those proteins. And the fact that they're not somatic hypermutated, oftentimes that hypermutation affinity maturation selects out some of that poly-reactivity. So, I mean, it's important, I think, to distinguish this because it's not really molecular mimicry per se, what's going on there. And I should say the fact that we don't see a lot of, or really any, induced autoreactivity in the setting of vaccination, I think, does argue against molecular mimicry being a major mechanism of autoimmunity, particularly if you're focused on spike. Of course, we don't look at nucleocapsid, et cetera. So I'm kind of more favoring that it's less so antibodies against the virus, but rather the virus is, number one, creating an inflammatory milieu that could be driving some of these antibodies, could be leading to losses in peripheral tolerance, not just due to the inflammatory milieu, but the fact that those antigens are present. And then again, like I said, the virus might be innately tickling some of these more innately polyreactive antibody frameworks. Those different hypotheses haven't really been dissected. I think all of them could be happening.

It's not like necessarily either or.

Yeah, yeah, absolutely.

We do have a question on whether you've been able to drill down on at least some of these antibodies to determine what proportion are high titer IGG with high avidity as opposed to mostly IgM with low avidity?

And there— Yeah, great question. I just want to be 100% clear, we only look at IgG. We don't look at IgM. IgM, unfortunately, we're really interested in IgM. The problem is it's just so sort of innately, I want to say a little sticky, that it just sticks to like all yeast all the time. So we kind of remove the IgM from the process. We can look at IgA. We haven't done that yet with COVID It'd be really interesting to do that. But yeah, we haven't looked at IgM. So everything we're looking at is IgG, and we have drilled down pretty significantly into many of these antigens. Now I just want to, you know, iterate that, you know, we found, you know, quite some number of antigens, you know, several hundred, so we can't validate them all, right? But we did characterize many, you know, for example, we titered them via ELISA Luciferase immunoprecipitation assays. We did functional assays showing their ability to block pathways. So many of them are very high titer responses, you know, 1:1,000. These are greater than 1:100. This particular response is greater than 1:10,000. So it really just depended on the antigen. Some of them were low titer as well.

So yeah, we have done some profiling there.

Right, right.

Do you— have you noticed any age-related differences among the people you've studied, or do you suspect there would be?

Absolutely. We know that antibodies go up with age, and autoantibodies go up with age sort of innately. And so we haven't— in all of our analysis we did, we presented in our paper, of course we did regression models to try to correct for things like age and sex, et cetera. But yeah, that, you know, that does kind of lead to an idea that, you know, one of the features that could be potentially driving more severe outcomes is that older people may be more prone to develop these antibodies or have existing antibodies like the interferon autoantibodies that predispose them to severe disease, perhaps other immune targeting antibodies. Our cohort is not really big enough and diverse enough, I think, to really tease that out. Because it's heavily skewed toward patients who are elderly, just given that these were by and large hospitalized patients. But yeah, that's a great question. And I think as we, you know, expand our studies on more patients and more diverse cohorts, that we could start to tease out those, those influences.

Right. So we have a question about whether it would be possible to anticipate the appearance of some autoantibodies after certain types of virus infection through bioinformatics.

For example? Yeah, well, I think that's one of the great hopes that we can do with approaches like REAP, perhaps with like FIP-seq, or just high-throughput unbiased characterization of autoantibody responses. And if we can develop a large enough atlas, a serological atlas of many different patients with different infections, different past medical history, different coexisting comorbidities, we might be able to predict some of those antibodies in addition to saying which antibodies are also influencing the disease as well. I don't think we have enough data points for that yet. I think you would need a massive atlas of tens of thousands of patients to do that. But yeah, it's— I think that's sort of the next frontier for the types of questions we want to address. And the types of questions we can address with the data matrix we're creating here is enormous, right? We have many thousands of patients across thousands of antigens, and in many cases we have a longitudinality as well. And so it really does lend itself to those types of advanced data science approaches.

Right. And we just got a question: have you, have you looked at changes in REAP pre- and post-SARS-CoV-2 or other vaccines in the absence of prior infection?

We have. We're writing up a report on that now. Hopefully we can release that as a preprint in the next few weeks. Just to summarize, it's a very prosaic result. So we have patients— not patients, just volunteers— who we have pre and post-vaccine at day 7, day 14, post-dose 2. Maybe we'll be able to expand this after we get the boosters coming in. And we have not just like healthcare worker volunteers who did this, but also we have in collaboration with the Benaroya Institute, we've been looking at patients who have preexisting autoimmune disease, pre-post vaccination. And I just want to reassure everyone, we didn't know what we were to expect. And I was kind of worried, man, what if we see a lot of autoantibodies? That's going to be a bummer to be like the guy shading on the vaccines, right? But you guys can all rest easy. It's a very prosaic result. The vaccines do not increase autoantibody reactivities like COVID, full stop. And patients with existing autoimmune disease don't either. So yeah, that's— I don't quite have that ready to present. I don't have the figures made, but I have the result. I'm happy to share it with you guys that you can rest easy.

So we have a question about whether you've detected antibodies that are against the pancreas or insulin, because it's been described that there's been more type 1 insulin dependent diabetes cases after COVID?

So in our library, insulin— insulin's an interesting protein. It's heavily processed post-production. And so unfortunately, the yeast don't do a very good job making insulin. So insulin's not in our library. Sorry about that. But we do have a lot of pancreatic enzymes, pancreatic lipases that are in the library. We do see them in other conditions. I don't know if maybe the eagle-eyed among you there saw In the APOSET patients, we found patients who had anti-pancreatic lipase antibodies, colipase as well. In fact, the one patient who had a colipase antibody had exocrine pancreatic insufficiency. So they're in the library, we detect them. As far as I know, we haven't— we have not profiled any samples of patients who developed diabetes post-COVID-2. We would love to do that. If you guys had samples, you want to send it to us, we could definitely take a look. But in our cohort, I don't have to— I'll have to go back and look if we saw those. I don't know that we've followed up to see if they subsequently developed diabetes if they, if they have those antibodies. Right.

So I know you presented a little bit about the waning of some of the autoantibodies in people maybe up to a year after infection. Do you think that there will be a continual waning, or do you think the people who still have the antibodies are going to experience some autoimmunity.

Yeah, so I just want to be clear, like, even in the same patient, if we look at the same patient and still have, you know, a range of autoreactivities, some of them wane down. And, you know, basically it's like predicted decay given the half-life of an antibody. Like, it really is tracking well. In fact, in many of these patients, we have a spike in control. The patient's got tocilizumab, they got an antibody against IL-6 receptor, So we can actually look at the clearance of that antibody over several months and compare that to their autoantibodies. Yep, so yep, these antibodies are waning. But in those same patients, sometimes there's antibody responses that came up and stayed constant. And right now it's too early for me to speculate why that is and if other patterns to some responses that wane versus others that remain elevated. If I had to speculate, I would say probably there's multiple mechanisms of humoral autoreactivity here. In some cases, it probably, it definitely is this phenomenon of the extrafolicular double-negative B cells, these GC-independent B cell responses. And those seem to wane over time. They don't seem to form, you know, long-lasting memory. You know, Noxios Fonts has got a preprint on this. I think it's pretty nice that shows that those types of responses wane. I suspect a lot of the waning antibodies we see may be derived from those cells. But in the cases where they don't wane, that, that does make you think this might be a more conventional, you know, break in tolerance, and you get, you know, formation of a long-lived response. And that's what worries me because, you know, now you can, you know, very straightforwardly see how that type of response could cause a long-lasting autoimmunity, and potentially that could lead to conditions like long COVID. I want to be clear, our hypothesis is that autoantibodies could contribute to long COVID. We have not— I would say, you know, we have certainly not established, nor has anyone else, that this is, you know, an autoimmune disease. And certainly I think it would be premature to say, you know, we should be targeting autoantibodies in those patients. I just think right now we don't have— you know, there's several different hypotheses. Is it persistent viral reservoirs, maybe. Some of these patients do better with vaccination. Is it autoimmunity? It might be, or at least in part. You know, also, you know, long COVID is such a catch-all term, but it's a very heterogeneous disease. You know, likely this could be a constellation of different illnesses, very much like other, you know, post-infectious conditions like ME/CFS as well, which we're also studying now with REAP. In fact, we're comparing ME/CFS patients to long COVID patients to see, you know, other features in common between them. So stay tuned. It's a little too early for me to speak definitively about that, but potentially some clues of differences.

Yeah, and a lot of interest in that topic, obviously. Maybe we have time for one more question. Have you looked at any samples from persons living in Southeast Asia that may have been exposed to different coronaviruses over a lifetime?

And what their autoantibody profiles look like? I would love to. We haven't profiled those yet, but if anybody has those type of samples, please contact me because I would be delighted to take a look. You know, we actually put in a lot of the other coronavirus antigens like, you know, HKU, you know, 229E, NL63. So we actually, you know, can monitor antibody responses to other coronaviruses in addition to CoV-2.

That's great. Well, we have a lot of interest on using REAP for other infectious diseases among our participants.

So, I'll say, you know, to those interested, we are actually now expanding our library. We're going to add in about another 150 to 200 antigens from a wide range of viruses. So there's already groups doing this, really great groups. They have powerful technologies, the VirScan technology with phage display, etc. So, you know, by no means are we the only ones doing it. There's many people much further ahead than us. But, you know, what hasn't been looked at are some of the, you know, conformational autoantibodies against those viruses and then correlating that at the same time to autoantibody responses. And so we are very interested in this area, and we're going to have antibodies against the other herpesviruses. Other endogenous retroviruses, a very wide range of viruses, other coronaviruses, etc. Stay tuned, and if you're interested in autoimmunity and infection, like I said, I never would have thought 2, 3 years ago I'd be going to this area, but we're very interested.

If you have samples and you're interested, please contact me. That sounds very exciting and a very important and expanding field. So thank you so much, Dr. Ring, for your excellent presentation and this really important work. We really appreciate you taking the time to share this work with our audience and for answering questions. I would also like to just thank all of our attendees for once again participating in today's webinar and for submitting such great questions for our speaker. We're always fortunate to have such an engaged audience for these lab meetings. And with that, I invite you to join us 2 weeks from today on October 7th Our speaker that day will be Dr. David Fuhrman from the Buck Institute for Research on Aging, and his presentation will focus on systems biology of the aging immune system. And if you're interested in yet more research on COVID-19, please do 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. And finally, please visit our website and follow us on LinkedIn, where we will upload a recording of today's webinar. Thank you again so much, everyone, for participating today. Stay safe, and we hope that you'll join us again on the next Global COVID Lab meeting.

Thank you so much.