Thank you all for coming. Thank you to the organizers. Thank you to my friend Leo for setting me up perfectly. It's almost as if it were scripted or well organized, I should say. I'm Bob Dixon, I'm a faculty member at the University of Michigan in Pulmonary and Critical Care Medicine, as well as Microbiology and Immunology. And I'm here to talk about the negative or positive impact of the microbiome in ARDS. And to segue from what Leo just described, in 2023, it's really not arguable that there are respiratory bacteria that are altered in ARDS, correlated with disease severity, predictive of outcomes. But like Leo said, the million dollar question is, is it causal? Are they actually involved in pathogenesis? That's really the focus of what I'm talking about. I have no financial disclosures. I'm gonna try to confuse you and then resolve that confusion because I'm gonna show you data suggesting both that the microbiome is participating in the pathogenesis of ARDS. It's the bad guy, it's the culprit. But also I'm gonna show you data that the microbiome is protective and on our side. And then I'm gonna explain why I think that can be resolved with a more nuanced understanding of what exactly we're talking about. So let me start with the bad guy, a force for evil. I'm gonna start with talking about oxygen, which of course everyone in the room knows is really our first line therapy for ARDS. I hope you also know it's a potential cause of ARDS. It surprises some, but if you take a healthy mammal of any species, you put them in pure oxygen, they will die within a week. And what they die of is respiratory failure. And under the microscope, their lungs are diffuse alveolar damage. They die of ARDS. This is data from our group, but this is epithelial necrosis. These are highland membranes that get alveolar edema. Oxygen causes ARDS. But as Leo said, oxygen also matters profoundly, not just to our lungs, but to the microbiota everywhere. It's a determinant of community structure. So a few years ago, a very talented postdoc working with me, Shauna Ashley and I, asked these questions. Does the administration of oxygen change respiratory microbiota? And does that potentially contribute to ARDS? And I'm gonna summarize in one slide about five years of Shauna's life, but it's all published. And it's a translational story where first we started with observational human data. We looked among our ICU patients at the University of Michigan, early exposure to FiO2, the higher the FiO2 in the first couple of days, the more staph you grew out of subsequent respiratory cultures. Staph seems privileged. It's uniquely tolerant of oxygen compared to Pseudomonas and other potential pathogens. But that's observational human data. It's highly confounded. Maybe they were brewing a staphylococcal pneumonia before. So we moved to animal experimentation. And lo and behold, we can recapitulate this in healthy mice that are not confounded. So this is taking healthy adult mice, putting them in hyperoxia, and then looking at their lung bacteria in subsequent days. And what you can see is within 24 hours, you have a drop in clostridia, bacteroidea, the common anaerobes that we normally find in the respiratory tract. And what fills that ecologic space is staph, which again is exquisitely tolerant of oxygen compared to the competition. So we can show that it's not just confounding. Oxygen does seem to change respiratory microbiota. We can make a story about temporality. We can show that this happens before this, namely the dysbiosis occurs within a day or so, and injury doesn't occur until day three or four. So like Leo said, you had to ask yourself, is the dysbiosis a consequence of the injury or rather than a cause? At least temporally, it's happening prior with these models. We can look at heterogeneity analysis. So we do the same model in a large number of mice that are microbiologically divergent. And we show that variation in lung bacteria correlates strongly with variation in lung inflammation and injury. So again, another clue towards pathogenesis. But the real kicker is this, is if you take the microbiome out, you actually pull it out of the system. So to do this, we use germ-free mice, which are black six, genetically identical mice to everything else we're using in other experiments. We give them the same exposure and we ask if the microbiome is a culprit, then it should be protective. And sure enough, it is. So this is showing you lung injury. So we quantify it with alveolar IgM. It's basically how leaky your lungs are. And these are all black six, genetically identical mice. Conventional mice, hyperoxia, causes severe lung injury, no surprise. Germ-free mice are protected, both biochemically, histopathologically. So at least an argument that in this one model of ARDS, the microbiome's playing a role. That's largely mouse data. Do we have any human data? Leo's already showed you some human data to support this. This is our paper showing that, working with Louis Bos from the Netherlands, showing that if you look at mechanically ventilated patients, the quantity of bacterial DNA in the lungs is predictive of bad outcomes, which, as Leo showed you, has been shown in multiple studies now. It's a very consistent signal. Again, suggesting that the microbiome, even in these clinically uninfected patients, seems to be predictive of bad outcomes. Maybe it is the bad guy. So let me switch gears and show you how the microbiome might be our friend. To do that, I'm gonna walk you through a few findings from a recent paper of ours at the end of this past year by Rishi Chandraj and Jen Baker, looking at specifically the effects of anti-anaerobic antibiotics, which, of course, are clinically ubiquitous. If you've given your patient Piptazo or Unison or Flagyl, you have given them potent anti-anaerobic antibiotics that affect the gut microbiome. We've known from prior work that taking those protective gut anaerobes out of the gut makes animals more vulnerable to critical illness, lung injury, organ failure, and we recapitulated that. So this is work that Jen did with two different models of lung injury. This is a Klebsiella pneumonia model and a hyperoxia model that I already described. And in both cases, if you pre-treat the animals, these are non-infectious prior to the exposure, if you pre-treat them with Piptazo, which has potent anti-anaerobic activity, and you compare that to cefepime, which does not, or saline, they do worse. So in both models, they have worse pneumonia, they have worse lung injury, they have higher mortality. So I can turn a non-lethal hyperoxia model into a lethal model just by giving them Zosyn for three days ahead of time. That's mice, do we think it's true in humans? So Rishi looked at observational data from our center. So this is more than 3,000 mechanically ventilated patients in our ICUs. We looked at who got antibiotics, and unsurprisingly, all of them. It's basically 99% of them get antibiotics. It's hard to be on a vent, at least in our ICUs, and not get antibiotics. But there is variation. So 2 3rds of them got anti-anaerobic coverage like Zosyn, Piptazo, 1 3rd of them did not. And there are other talks on this topic, but most of those anti-anaerobic prescriptions are not necessary. They are over-prescribed. Most patients do not need anti-anaerobic coverage. So for legitimate clinical indication, very, very rare here. But this is enough variation. 2 3rds get their anaerobes blown up, and 1 3rd do not. Is there a difference in their microbiome? Absolutely. So this is looking at rectal swabs from these patients at the time of ICU admission. We could look back and tell you, based on their gut bacteria, whether they got Zosyn or cefepime in the ER. So within a day, you see a hundred-fold drop in bacterial density and the identity of anaerobes in the gut. So it has a very different effect on the gut microbiome. In terms of outcomes, this is very, very troubling. So if we look at overall survival, this is VAP-free survival, this is ventilator-free survival, this is overall survival, and the signal is the same. If you get your anaerobes depleted with Zosyn as compared to cefepime, 7% absolute increase in mortality at 30 days. So tremendous effect size. Now it's observational data. When we control for everything we can control for, still a 5% delta. And like I'm saying, we see it in the mice too. So it's not as if we, in our unconfounded mouse experiments, we don't see the same thing. So this troubles my sleep. I have given gallons and gallons of Zosyn to patients and have never given much thought to what I'm doing to essentially this organ, this organ that I am just inducing organ failure in with it. So how do we resolve this? How do we reconcile the fact that in some context, it looks like the microbiome is a force for good and some force for evil? There's a few things. One, as Leo said, the microbiome is not one thing. Like saying, is the microbiome good or bad in ARDS is like saying, is the immune system good or bad in ARDS? I can show you circumstances where neutrophils really do seem to be the bad guy. I would not give up my neutrophils. I think they're protective in most circumstances. So it varies across patients. It varies across anatomic compartments and across time. So across patients, I would argue it is the most dramatic source of heterogeneity across our patients. Any two patients are 99.9% identical in their host genome. They may have zero taxonomic overlap in their gut and respiratory microbiota. So huge variation across patients. It matters if we're talking about the gut or the lung, as Leo was discussing, and it varies temporally. Unlike the host genome, it changes based on your time in the ICU. What you have on day three is not what you came in with. So that's one source of confusion here. Two, in terms of about anatomic compartments, so I'll come back to this in a moment. The microbiome is biologically potent. So like I said, we talk about it in a cute way, but it is functionally an organ. It can promote or disrupt homeostasis. It does this via more than one mechanism. So first of all, you have to ask yourself, are we talking about bacteria in the lower gut or in the lung? As Leo just argued, and I think has spent the last decade showing us, even though as exciting as the gut-lung axis is, we don't always need to invoke it. Sometimes it's lung-lung. Sometimes it's local interactions between respiratory microbiota, the respiratory host response and lung injury that matters. So that's one source of complexity here. But also in terms of that gut-lung axis, why might gut bacteria be contributing to or protecting from lung injury? There's multiple mechanisms. The one everyone already knows is colonization resistance against potential pathogens. This is why our colleagues in infectious disease and antimicrobial stewardship have been for ever telling us that we should be more reasonable with our use of antibiotics. Once you deplete those gut anaerobes, C. diff takes over other nosocomial pathogens. That's old news, right? But we also have to take into account is direct translocation of gut bacteria. Now, this is a very old idea. It was studied for decades. Went away in the 90s. Culture-based studies did not find convincing evidence of it. I think it's back on the table. So a few years ago, we published a series of studies, of experiments looking in animal models of sepsis. When you induce sepsis in mice, either via cecal ligation and puncture or endotoxin, it disrupts their lung microbiota. I can tell, based on their lung communities, which group they were in. And it does so in a way that looks like it's enriched with gut bacteria. So this is Bacteroidales. This is the dominant bacterial taxonomic group in the gut of these mice. It surges in the lungs after sepsis and then normalizes it by two weeks out. And because we can sample the upper respiratory tract, we can argue that it's not just aspiration. It's not just top-down. That's mice. In humans, we also find a third of patients with ARDS have this similar bacteroides, these gut anaerobe in the lung, and it's positively correlated with their serum TNF-alpha. So the more systemically inflamed and shocky you are, the more of these gut bacteria we find. And then in that same paper that Leo mentioned from the UCSF group, looking at lung bacteria predicting outcomes in trauma patients, they found that the lung bacteria among patients who did develop ARDS compared to patients who didn't, the driving difference, the most potent explanation was Enterobacteriaceae. So a very common, like a definitional gut gram-negative rod. And in that same study that I did with Louis Boss from the Netherlands, we actually find the exact same bug discriminating patients who do and do not have ARDS. So again, it's mounting evidence that in ARDS, we find more gut bacteria in the lungs than belong there. That's observational human data. It's confounded. We use animal models. So I'm very lucky to work at the Max Harry Weill Institute at the University of Michigan. We have a large animal lab. This is us modeling sepsis in pigs. They get an E. coli pyelonephritis. And a lovely thing you can do with large animal models you can't do with mice or humans is you can design an experiment to sample what goes into the lung and comes out of the lung to determine what's getting filtered and what's getting produced. So by sampling pulmonary artery blood and time-matched carotid artery blood, we can actually triangulate what's going on in the lungs. And when we measure bacterial DNA, we see a consistent step down. So there's more bacterial DNA going into the lung than coming out of the lung, suggesting that at least in the septic shock state, the lungs filtering it, right? It's getting there. So I think that direct translocation of gut bacteria is again on the table as a possible mechanism. But the two biggest plausible mechanisms that we look to for the gut microbiome in foreign lung disease haven't even been mentioned. One is calibration of systemic and alveolar immunity. This is all the rage in other fields. So if you're on oncology and you want a paper in science, what you do is you show that your patient's with melanoma. Whether you do or do not respond to immunotherapy depends on your gut bacteria, and they're actually doing fecal microbiota transplant and turning non-responders into responders. So thinking of the gut as an immune organ and you can calibrate the host response via the gut. Finally, dissemination of microbially derived metabolites. So the gut microbiome is a metabolic organ. It makes at least 400 detectable small molecules that spread throughout the body and have effects on end organs. And these two matter, right? So calibration of immune tone and your metabolic activity is critically important in critical illness. And we've been looking at this recently using a specific index that tells you something about both, which is fever and temperature. So if you think about what is fever telling you, it's telling you the patient's inflammatory and metabolic state. We were inspired by a paper from a few years ago by Siva Bhavani and Matt Chirpek that looked at septic patients in multiple cohorts and found that they could cluster them, they could sub-phenotype them and group them by their temperature trajectories. So there's patients that, we all see this clinically, come in hot and they stay hot, people that come in hot and they defer vest, people that are normothermic and people that are hypothermic. And these are consistent trajectory phenotypes that pop up in multiple cohorts. They differ demographically. They also differ prognostically. The hypothermic patients do worst. We think that this is fascinating. We wondered why, what's driving that heterogeneity? Of course, I think the answer is the microbiome. So we tested that hypothesis. In a paper we just published in Blue, Cale Bongers, one of my trainees, tackled this question. The first thing we did was recapitulate what they found. So in our own ICU patients, we find the same four trajectories of patients based on their temperature curves. When we compare those patients to each other, their gut bacteria are distinct. And specifically, it's formicides, it's this gram-positive phylum, and specifically lacnosporiceae, which is a potent producer of short-chain fatty acids that seems to discriminate them. The more of those you have, the more like you are to be in the febrile side. But that's observational. Human data, it's confounded. Can we find it in mice too? We definitely can. So this is all on genetically identical black six mice. We're modeling sepsis with endotoxin. And the first thing you notice, the thing to know is that mice don't get fevers. They get hypothermic when you make them septic or you give them an infection. This is what that shows. If you use germ-free mice without a microbiome, their temperature barely moves. All of the heterogeneity is present. And again, these genetically identical mouse, mice, same age, same gender, we've controlled for all the variation that we can think to. And there's tremendous variation in their temperature response, whereas there's none in germ-free mice. And that variation here, whether you maintain your temperature or you drop three degrees, is explained by the microbiome. And it's, again, lacnosporiceae, the same bacterial taxonomic group that we found driving it in the humans as well. So in this case, 36% of the variation in temperature is explained just by this one bacterial family in the gut. So we think that this is compelling evidence that if you want to explain inflammation and metabolism in the ICU, you have to consider the gut microbiome as a potential driver of that. So my last point is we're always talking about modulating the microbiome as a potential therapeutic strategy for sepsis for ARDS. My message is we're already doing so. This is not over the horizon. Maybe one day we'll have precision therapeutics and we can tweak the microbiome for good. We're already doing it. And we are just beginning to understand the biological and clinical consequences for this. And it's coming back to what I described with anti-anaerobic antibiotics. So I'll bring you back. This will be my last figure. But giving anti-anaerobic antibiotics to patients coming into the ICU requiring a ventilator, at least in our observational study and our murine modeling, increases their risk of dying. Like I said, this troubles my sleep. And I would ask this. For all of the debate and deliberation about time to antibiotics, let's open it up to which antibiotic, right? Like a short delay to make sure that patient doesn't get unnecessary anaerobic depletion inducing a type of organ failure may be worthwhile. So with that, hopefully I've confused you and then resolved at least some of that confusion. And I again thank the organizers. I acknowledge everyone I've worked with on this data and happy to continue the discussion. So thank you all. Thank you.

microbiome

acute respiratory distress syndrome

respiratory bacteria

dysbiosis

gut microbiome