Thanks, Ken, for the invitation. This has been an interesting session. So my topic that was assigned was Detox the Bug to Generate a Happy Host. I'm probably going to take a little bit of a tangential approach to addressing this particular topic. Let's see. Here we go. Let me start with my disclosures. I have a financial agreement with Aridus Pharmaceuticals related to patents owned by the University of Chicago. And I may receive royalty income based on technology developed in my lab that's currently owned by Washington University and subject to licensing by Forward Defense. This has all been reviewed by Washington University's Conflict of Interest Committee. So the objectives, as I've defined them, is to recognize the importance of the host-pathogen interaction in sepsis, to appreciate the molecular complexity by which pathogens initiate injury in the context of sepsis, consider the role of host genetic polymorphisms in susceptibility to pathogen-mediated injury, and to examine opportunities for host-pathogen-targeted interventions to prevent and treat sepsis. I feel like I should probably expand my disclosures a little bit because I took the words deep dive quite literally. So I'm a molecular microbiologist. And I really like to think about sepsis from the perspective of, how is sepsis initiated? And what's the initial molecular dance, if you will, that happens between a pathogen and a host that incites the injury? And I think as I look at the field, and certainly as I stand in the ICU and see a patient with sepsis, we are looking at the fourth and fifth act. So where my lab focuses and what I really think about is, how is sepsis catalyzed? What is the insult that's provided by the pathogen? So I will take a fairly deep dive to talk a little bit about a molecular mechanism that we've discovered to understand how it is that pathogens specifically intertwine with the host to initiate sepsis and the implications that this has for clinical therapies. So I like to think about the host-pathogen interaction really as this sort of seesaw. And on one side, I would put the pathogen. On the other side, I would put the host. When one thinks about the pathogen, there's a lot of richness of the biology that's become known in the past handful of years, including understanding microbes and their niche, which differ across the body and differ between you and I, competition within that niche. So why is one bug there and another one is not? Pathogen acquisition and transmission, thinking about the commensal, which in theory is not harmful to us, but the pathogen or the pathobiont, which may be. Virulence factors, which are the tools or the weaponry utilized by pathogens. And regulation of gene expression, which can be very different between two almost identical bacterial organisms. And then on the host side, we have just a burgeoning field of the microbiome and understanding how the constituents of the microbiome and, more importantly, their function impact the host. Immunologic competence, some of which is genomically encoded and some of which is acquired through experience. As I mentioned, the genetic constitution and how this may influence virulence factor susceptibility, which I'll get into. And then tissue-specific regulation. What about the site of infection and the site where the host and the pathogen interact? This is very different in the skin than it is from the gut. So I think many of you have probably seen this. What I really like about this graph of bacterial-associated mortality that was published by the Global Burden of Infectious Disease Study in 2019 is when you look at the list of pathogens across the bottom, and I won't sort of go into any of them in detail, what you recognize is that many of the pathogens on this list are the exact ones that we see in the ICU. And when you look across the bottom of that list, what you will also appreciate is that very few of those organisms have vaccine-mediated protection against them. And I think when I look at the field of sepsis and I look through the lens of a pediatrician at what has been the major victory in the past 100 years, it's the utilization of vaccines to actually prevent extreme consequences of infection. I think we see in the COVID-19 pandemic this works. So when I look at this graph, I really recognize it's incumbent upon us to think deeply about the pathogen, to understand how can we target the pathogen, because this has been the most tried and true mechanism by which we have protected the population against lethal disease. So when I think about sepsis, I conceptualize the host, the pathogen, the microbiome or the broader constituent of the microbial load that the patient has, and then the host immune response. And what I've schematized here is just four different exemplars of how these combinations may play out. And I think everyone knows as you stand at the bedside, you may have a given host who is infected with a certain pathogen. We certainly don't know much about their microbiome when we have inferred things about the host immune system, and this patient will end up with sepsis. The second patient is genetically distinct from the first, may be subjected to or exposed to a different pathogen, has a different microbiome, different immune response. They do fine. This third patient I've depicted, you get the gist. Different pathogen, different microbiome may have a similar immune response, but will end up with a ventilator-associated pneumonia. And then in the fourth condition, a surgical site infection. Our challenge when we stand at the bedside is we have no idea which patient has exposure to which pathogen in the context of a microbiome that may render them susceptible. So this grid is not interpretable by us at the bedside. So I'm gonna extend this sort of a little bit further to capture some of the nuance of this problem as I see it. So if you, again, have this constellation of host pathogen microbiome and immune system, let's say that this patient is set up for survival. If you make a subtle change to the immune response, even in the context of the same pathogen, similar microbiome, this patient may die. Perhaps you make two changes, change the context of the microbiome, yet the immune response may be functionally similar and the patient may survive. And then in the fourth condition, change the pathogen. The microbiome could be similar. The immune response could be similar and that patient will die. So I think to me, this illustrates the real complexity that we face when we think about sepsis in the context of the host-microbe interaction at the bedside. Because we have hosts that are genetically undefined to us. We have pathogens that are often not discriminated or not known early in the course of disease, even with modern day molecular microbiologic tools to discern them. We really have not interrogated the microbiome in any way that enables traction to determine predictive factors within the microbiome that may lend to one outcome or another. And again, as has been mentioned very nicely in this session, we have a pretty poor handle on what is the protective or the non-protective immune response in the context of this host. And I think that where the field has likely gotten sideways, so to speak, with mice, is schematized here. We've utilized mice extensively and I'm gonna share some mouse data with you because I'm probably a mouse biologist more than anything. We've used LPS, we've used CLP. And what these result in is inflammation. And we have assumed that this inflammation, as evident in the mouse in response to these stimuli, is the molecular readout of sepsis. And I think that lacks a level of discrimination that we really need within the field. And I would urge that we should maybe think about this just a touch differently so that we don't throw away the mouse. So what I'll try to convince you with a series of studies that I'll share is that what we really need to understand is how pathogen-specific factors actually drive the nature of the inflammatory response, which is intended to be protective. The host immune response is never driven by an intent to harm the host. It's maladaptive. So what is it about the pathogen that actually interfaces with the host to drive the immune response? And how on a molecular basis does that yield the outcome of sepsis? So what I'll tell you is a story of what I would call molecular specification. What makes one pathogen different than another on a molecular basis? And how does that drive the phenotype within sepsis? So I'm gonna start with a little bit of background on Staph aureus. This is a bug that I'm certain is very common in all of your clinical experience. It's an interesting organism. It's been studied for well over 100 years, and there's incredible molecular genetic detail known about Staph aureus. It causes a really wide array of infections, which makes it fairly unique amongst bacterial pathogens because it can be at home in any tissue in the body. And what's really been vexing in the field about Staph aureus is that this organism causes frequent recurrent disease. So what it suggests is that this organism can actually subvert the host immune response. We do not develop protective immunity, as we know, against Staph aureus. So this is what I call the anatomy of the pathogen. So I've just depicted the organism, or Staph, as these golden circles. And what I've depicted around this cluster of Staph aureus are a number of virulence factors, or so-called weaponry, if you will, that the organism uses to insult the human, cause injury, and also to defend the bacteria against host-mediated immune responses. This is a really impressive armamentarium that this organism has. I won't go into any of the details of these, save to say I'll talk a little bit about the cytolytic toxins, because that's the window through which we've viewed this host-pathogen interaction. So I'm gonna start with a story many, many years ago, back in the day when toxin-antitoxin sera were utilized to protect against diphtheria. There were a number of children that received a diphtheria-antitoxin-toxin mixture in order to be protected against diphtheria. This is in the early days of immunotherapy. And many of these children died, and they succumbed to what appeared to be the vaccination. An inquiry was established to understand why this happened. And very early in his career, McFarland Burnett, who won a Nobel Prize, discerned that this Bundenberg disaster was the consequence of vials of this antitoxin that were contaminated with a different bacterial toxin. He did some really elegant work, and I'm just gonna share this really briefly. He took filtrates from organisms that he cultured, and demonstrated that these filtrates from this antitoxin that was contaminated could kill rabbits in less than 15 minutes, could elicit a dermonecrotic response in the skin of rabbits, and could also cause hemolysis. And in the day, these were, I think, relatively primitive molecular biologic tools, but the observations were really incredible. And what he commented on was that these three very distinct outcomes were manifestations of a single antigenic substance. So killing rabbits, causing hemolysis, and causing dermonecrosis from a single antigenic substance. Many years later, scrolling forward, nearly 100 years, this was defined to be the alpha toxin of Staph aureus. This is a really, I think, beautiful molecular structure that was solved in the mid-1990s. This toxin is a canonical pore-forming toxin secreted by the bacteria. It binds to the host cell membrane, generates a heptameric pore, and pokes a hole in the cell membrane. So you would imagine if you poke enough holes in any eukaryotic cell membrane, the cell viability will decline. So that's the presumed mechanism of action of this toxin, which has been studied now for almost 100 years. When I first started studying this toxin, the model that was appreciated was that as the toxin is secreted as a monomer, it binds to the host cell membrane, assembles into the heptamer, and then forms a pore that perforates the membrane. What we asked the question of was, how is it that this toxin binds to the membrane? And we discerned that it binds to a receptor called ADAM10. ADAM10 is a zinc-dependent metalloprotease that is ubiquitously expressed on every cell in our bodies, interestingly, with the exception of the human red cell. When the toxin binds to ADAM10, two things happen. First, the toxin is able to form a pore. And secondly, ADAM10 becomes activated. So by virtue of this interaction, the toxin is enabled to perform its primary function of perforating the cell, but the toxin also catalyzes the molecular activity of ADAM10. And ADAM10, as a metalloprotease, cleaves a number of different proteins on the surface of our cells. And what we're able to demonstrate in this context is that when the toxin elicits this cleavage event driven by ADAM10, this underlies the pathogenesis of disease. So we're able to generate Staph aureus that are either wild-type, which I've shown here, that makes the toxin, a knockout variant that does not have the toxin, but is otherwise genetically identical, and then we recover or restore the expression of the toxin. So we can use this toolbox to study disease. So these are just three models of Staph aureus disease in mice. I'm just gonna draw your attention to the right, which is a sepsis model. So what you see is survival on the Y-axis and time post-infection on the X-axis. These are animals that are injected IV with live Staphylococci. You can see following wild-type Staph infection, these animals all succumb. Following infection with a strain that's identical, except it does not have the toxin, they're protected against disease. So when we look on a molecular level at how the toxin ADAM10 interaction facilitates injury, what we know is that ADAM10 mediates cleavage of VE catherin on the surface of the endothelium, which then leads to an injury to the endothelial barrier. So we wanted to explore this in more detail and really ask the question on a mechanistic basis of early in infection, what's happening to the endothelium? We all know at the bedside, the endothelium is a really significant problem in the context of sepsis. So we first demonstrated, these are just a flow-based assay study of endothelial cells. You can see VE catherin is stained green in this picture. If we put the toxin with these cells, you start to see the VE catherin disappear, indicating that intoxication of the cells leads to cleavage or loss of the catherin. So this barrier is broken. So then what we did was we leveraged host genetics to knock out ADAM10 on the endothelium. There's a very convoluted sort of genetics behind this because this had to be both a cell type specific and a temporally specific knockout, because obviously ADAM10 is important for how endothelial cells function. But in the end, without going into the details, we detected a mouse that lacks ADAM10 on the endothelium. So we can infect these animals and ask the question, when infected with Staph aureus, if ADAM10 is not present on the endothelium, are those mice protected against lethal infection? And the answer to that question is they are. So these are mice, again, looking at percent survival in the context of sepsis. These are mice infected, wild-type mice infected with Staph aureus. You see they succumb. This is about an LD50. And you can see that animals that lack ADAM10 expression only on the endothelium are protected against disease. And I'll show you what these animals actually look like. So I'm going to try to play two videos concurrently. On the left, what you'll see is a wild-type animal. And on the right, you'll see an animal that's lacking ADAM10 only on the endothelium. These mice are 16 hours post-infection. So they receive the same inoculum of Staph aureus. You can see these animals are probably within five or six hours of succumbing to infection. These animals are pretty spry, and the clear majority of them will go on to survive. So a single genetic determinant on the host endothelium yields this outcome in the context of disease. So we took a deeper look at what's the molecular phenotype or the cellular and tissue phenotype of infection. You can see in control animals, these are the livers. You can see large areas of necrosis, both in the gross specimens and in histopathologic specimens. These areas are much, much smaller in the ADAM10 knockout animals. You can see this reflected in a transaminase reduction on the right. We then wanted to look at what's happening within the vasculature. So we hypothesized that an injury to the vasculature should be met with platelet deposition on the surface of the endothelium as a normal host protective mechanism. I mean, that's indeed what we saw. So in control animals, when we look at the hepatic vasculature, so the vasculature is delimited by red in these fluorescence microscopy studies. And you can see platelets in green. You can see following wild-type infection. You can see platelets that are coalescing within the vasculature. And in contrast, you see very few platelets in the VECAD here in knockout. So protection of the endothelium stifles the deposition of these platelet thrombi within the liver vasculature. And that's just quantified here on the right. So we then stepped back and asked the question, is this a common mechanism? So ADAM10 is expressed on every cell. The endothelium is very often targeted. And we're often scratching our heads at the bedside of patients with a wide array of different infections asking the question, why is this endothelial insult so terrible? So we wanted to ask the question, do other pathogens use ADAM10 to elicit endovascular injury? And the rationale for doing this is that many of these pathogens actually have poor-forming bacterial toxins. So this is really a question of, is there a molecular similarity at the level of certain pathogens that allows them to tie into a uniform host response mechanism? So we started with a few pathogens, Pseudomonas, the pneumococcus, GBS, and we actually looked at Canada albicans. And we looked at sepsis outcomes in either wild-type animals or animals in which ADAM10 has been deleted on the endothelium. You can see that there's a survival difference when animals are infected with Pseudomonas or the pneumococcus. However, knockout of ADAM10 on the endothelium doesn't matter for GBS or Canada. So arguing there's molecular specificity of the host-pathogen interaction on the endothelium and that we can dichotomize or discriminate pathogens based on this molecular phenotype. We then went on to ask the question, does this phenotype relate to microvascular thrombosis? And indeed, it does. So what you'll see here, again, these intravascular or intravital microscopy images of the liver, looking at Pseudomonas, the pneumococcus, GBS, and Canada. These are control mice, and then on the bottom are the ADAM10 endothelial knockout mice. And it's a little easier to see with the quantification below. In Pseudomonas and the pneumococcus, you see a reduction in the platelet thrombi in the absence of ADAM10. GBS actually elicits a lot of platelet thrombi, but it's independent of ADAM10 expression on the endothelium. And then in Canada, there's actually very limited platelet thrombi within the endovasculature. So the ADAM10 phenotype links to microvascular thrombosis in a pathogen-specific manner. So this then allows us to redefine this initial schematic and really think more in more of a structured manner about the fact that certain pathogens have the molecular machinery to tie into ADAM10 on the vasculature as part of the molecular pathogenesis of sepsis. And in contrast, what we know is that other pathogens do not rely on ADAM10 or ADAM10 expression specifically on the endothelium, even though they're able to elicit a lethal outcome in sepsis. So this is really, to me, one of the first opportunities to identify common themes in the early host-pathogen interaction that could be targetable based on pathogen identification. So I want to show you just a few more slides around this concept of toxin specificity for the endothelium and ADAM10. So in these three pathogens that are ADAM10 specific, we generated knockouts of the proposed toxins that cause this injury. And you can see that in the wild-type pathogens on the top, endovascular injury causes platelet microthrombus formation. However, with knockout of the bacterial toxin, you see that this injury is eliminated. And this is just quantified in the total thrombus area on the right. We then ask the question, can we target ADAM10 with a small molecule inhibitor of ADAM10 in order to protect against disease caused by these pathogens? GI254023X is a hydroxamate-based inhibitor of the metalloprotease itself that binds in the active site. We had previously shown for Staph aureus that GI would protect these animals against lethal sepsis. And we now show that in the context of Pseudomonas and the pneumococcus, treatment with this inhibitor of ADAM10 precludes the formation of platelet thrombi within the liver microvasculature, which is shown in the images on the left and then quantified on the right. So this really gives us a little bit different picture of what's happening in this so-called pre-sepsis state, if you will, when these early molecular interactions between the pathogen and the host are causing injury that is probably not immediately evident standing at the bedside. So I've dichotomized this into ADAM10-dependent and ADAM10-independent endovascular injury. There's a few predictions from this model which get at therapeutic or interventional opportunities. The first is that specific endothelial ADAM10 substrates likely mediate this injury. The second is that every pathogen could be classified based on ADAM10 dependence or the lack thereof. The third is that infection with ADAM10-dependent pathogens may benefit from antithrombotic therapy. And the fourth is that ADAM10 expression or expression level may govern the severity of infection. So I'm going to tackle a few of these in the next handful of slides to sort of bring this to a broader conclusion. So I'm going to talk a little bit about host polymorphisms in sepsis outcomes, looking specifically at or thinking about the ADAM10 RS653765 polymorphism. So this was a really very nice paper that was published a number of years ago, noting that an ADAM10 promoter polymorphism is a functional variant in severe sepsis. This group was a Chinese group that studied nearly 500 patients and classified them based on severity of sepsis and identified that this specific promoter polymorphism seems to govern severity of sepsis. So this paper immediately came to mind as we started to appreciate that ADAM10 expression on the endothelium is likely relevant for disease outcomes. So what we did was we looked at the mouse ADAM10 locus. So this promoter polymorphism is about 250 base pairs upstream from the start codon of the ADAM10 locus. And what we did was we generated what I'll call humanized mice. So I'll just show you down below. There's a mouse that I'll call the GG mouse, which is the G SNP type in homozygous form. And then the AA mouse, which is the A SNP type in homozygous form. So the GG mouse would be the one predicted based on the human studies to have a worsened outcome of disease because of a higher level of ADAM10 expression. And the AA mouse would be predicted to have a lower level of ADAM10 expression on the endothelium. And presumably, if the hypothesis is correct, a better outcome. So we subjected these mice to lethal Staph aureus infection. And indeed, what we saw is that animals with the GG phenotype have a worsened outcome compared to the animals with an AA genotype. So these studies are quite early. But I think they allow us to see that if we can map specific molecular interactions between a pathogen and the host, we can then start to ask the question, how do host genotypes influence the outcome of disease at a very granular level, including at the level of a single promoter polymorphism? So I want to finish by then asking the question, how can this type of knowledge inform an understanding of targeted interventions for sepsis? And again, I'm going to look through a very narrow lens of the best studied interaction of the host-pathogen interaction, which is this alpha-toxin-ADAM10 interaction. So I'll offer that there are probably three broad categories of interventional opportunities that we may be able to leverage. The first is targeting pathogen virulence factors. And again, I think this has been the mainstay of protection from lethal infection for decades. Vaccination, and then certainly there are new monoclonal therapies that I think probably within the next five years or so we'll see present in our ICUs. The second is targeting host factors that are required for molecular pathogenesis, such as targeting a molecule like ADAM10. And then the third is understanding host susceptibility, which really gets at this concept of personalized precision medicine. If you know that you are a GG genotype, so to speak, at the ADAM10 polymorphism locus, you may be more susceptible than your neighbor in the next ICU room who may have an AA genotype from a conceptual background. So the data that would argue that this first approach that I propose is vaccination and or monoclonal antibody targeting is useful. There are multiple publications now that have demonstrated that antibody levels to multiple Staph aureus toxins, including the alpha toxin, actually relate to the clinical outcome of disease. These data have formed the basis for what several companies have now pursued in terms of monoclonal antibody generation to target the alpha toxin. And I suspect that we'll probably see those studies roll out at a greater level within our ICUs. We've taken an approach of asking a similar question of, what do we know about how these responses develop? And how do we identify a patient who's at risk? We looked at the serum antibody titer to the alpha toxin from Staph aureus in about 500 children ranging in age from 0 to 18 years. And you can see that the serologic titer increases over time. We then mirrored this with a study of the neutralizing antibody response to target this toxin, which is really the functional moiety of the antibody response. And you can see that this response also increases over time. But what you appreciate is that there's a population that one may be able to define as at risk. So a patient who is unable to have or does not have a neutralizing response to a given toxin could indeed be a population in which you could know ahead of time, is this patient at an increased risk? And this is really how I see the substrate for thinking about utilization of antitoxin monoclonal antibodies or vaccination is in a risk stratified population. The second goal that I had mentioned is thinking about, how could we leverage knowledge of molecular pathogenesis to consider sort of dichotomization of patient populations? So I've just schematized here a probability of survival in a clinical trial in sepsis where you really see no significant difference. And what I pose on the right is a question of, what if we stratify these patients based on whether they were infected with an ADAM10-dependent pathogen or an ADAM10-independent pathogen? Can you actually then see a clinical difference in the outcome because you've generated a relevant stratification for a therapy that is directed at the thrombosis response? I don't know the answer to this. I simply frame this as a conceptual framework for further discussion. A little bit of data that would suggest that this may be the case was published by Victor Nuzze's group a couple of years ago. This is a Staph aureus infection model looking at percent survival in the context of sepsis. With antithrombotic therapy, you can see that these animals had a better outcome than animals that received the vehicle alone. This was nicely mirrored in our studies where we were able to look at the liver vasculature and show that ticagrelor treatment yielded a smaller number of these platelet microthrombi within the vessel, which is quantified on the right. So again, an ability to reflect on prior studies now that there's more clinical phenotyping from a mechanistic basis and really understand, can we dichotomize outcomes based on this molecular knowledge? Last, I'll just pose a question of, can we utilize genotyping or SNP typing based on molecular identification of host-pathogen interactions to identify individuals at risk? So again, taking the example of the ADAM10 polymorphism, let's say we have an individual who is the AA genotype who's predicted to have a low level of ADAM10 on their endothelium and may be less susceptible, compared to a GA individual or a heterozygous, compared to the so-called high-risk GG genotype. One may view someone with a low risk as someone that you may observe clinically. In contrast, someone who has an intermediate risk, you may think differently about how to target an intervention. And for someone who has a high risk, you may be willing to consider higher-risk interventions such as prophylaxis in a way that we frankly don't now in ICUs with the advent of concern about antimicrobial therapy. So I think this type of molecular typing based on mechanistic insight of the host-pathogen interaction will afford us much more of an opportunity to consider truly personalized mechanisms or approaches to therapy. So with that, I will stop. I've just highlighted on this slide a really terrific group of folks who did all of the basic science work that I had presented, and then my sources of funding on the right. So thank you, and I think we'll have questions at the end.

molecular microbiologist

host-pathogen interaction

sepsis

ADAM10 metalloprotease

Staph aureus

virulence factors

precision medicine