SCCM Resource Library-Putting Basic Science to Work: How State-of-the-Art Scientific Technology Can Inform Critical Care

Putting Basic Science to Work: How State-of-the-Art Scientific Technology Can Inform Critical Care

Video Transcription

Good morning, everybody. Welcome to this session here on bringing basic science to the bedside. It's good to see some people here. I guess everybody's worried about guidelines  and protocols, which you can just read about. But in any case, thank you for being here. These aren't things that you can just read about everything we're going to tell you about.  But I think what we can provide you are some nuances and some things to think about. So I'm glad you're here. My name is Hector Wong.  I'm from Cincinnati Children's and University of Cincinnati. I'll be your moderator today. And I guess I go first also. I think what we're going to do is about 15 to 20 minutes  or so. And then we'll take questions after each session rather than waiting to the end. And we'll try to stay on time. And hopefully, we can make it worth your while to be here.  All right, so my topic is on integrating genomics into clinical medicine. These are my disclosures, some grants from the NIH, some patents that I hold with my institution,  and then sitting on the scientific advisory board of the companies listed here. So as initially intended, the term genomics was coined back in the 80s.  I think it was at a Gordon conference and over a cup of coffee, I believe. I forget that legend. But in any case, strictly speaking,  it's supposed to be this interdisciplinary field. It's focused on structure, function, evolution, mapping, and editing of genomes. And here, the term genomes means the complete set  of DNA of an organism. It doesn't have to be human. It could be bacterial. It could be other species. But the idea is it's a complete set rather than focusing on individual genes.  Over time, it's kind of evolved, if you will, into what I call generic genomics in that people sort of think about when they think about high volume, high dimensional data that  involves nucleic acids, they refer to that as genomics. And that's OK. That's OK. But it's always kind of nice to know where things have come from.  And so with that in mind, I'm going to focus on transcriptomics, which is, in my opinion, I guess I'm biased because this is what I do.  I think it's some of the work that's closest to the bedside. And here, we're not looking at gene structure or editing of genomes, but we're looking at gene expression.  And the difference here is that you're looking at all genes at once, if you will. And the enabling technologies are microarray, which is now old school now.  Now, you're not cool anymore if you're doing microarray. Now, the thing to do is RNA-seq, which has many of many advantages. And then more and more, we're now  starting to get into single cell RNA-seq, which you can get transcriptomic profiles from individual cells. And we're starting to do that with neutrophils,  but that's proving to be a big challenge. So I'm not going to share that with you today. So what we've done, we've done a lot of transcriptomics  in the context of pediatric septic shock. And basically, what we've done is we've taken whole blood-derived RNA. And that's something that we can talk about, the source of the RNA.  And we take samples at day one and three of acute illness. We've done microarray analysis. And then we've undergone a number of iterations with bioinformatics and so forth,  always linked to very detailed clinical data. I think that's what a lot of young people forget about. You can have all the transcriptomic data in the world, but if you don't have  really rich and deep clinical annotations for phenotyping, it's not going to be that worthwhile. And so a lot has come up of this, right? We can't cover it all in an hour.  This is just one example that might interest you. This is from about five or six years ago, where we simply said, OK, among kids with septic shock,  if we compare those who get steroids and those who do not, is there any difference in their gene expression pattern? And one would think that the answer would be yes.  And actually, not that many genes are differentially expressed in this group. However, what's interesting is the majority of these genes  correspond to the adaptive immune system, which is pretty relevant to sepsis. And they're decreased in those who get steroids. So as we think about this whole controversy,  which still in 2020, we don't know about steroids, we have one of the experts here in the audience here on steroids in terms of clinical trials.  Food for thought, that if you give kids steroids at the genomic level, the genes corresponding to adaptive immunity are repressed. Then there's this story about olfactamide in four.  It's a little bit ironic that we're going from transcriptomics to one gene, but if you bear with me, I'll show you why. It's a glycoprotein that's found in neutrophil granules.  We don't really know its function. But in our data sets and in other data sets externally, it always seems to be the highest expressed gene when  you compare the patients who have poor outcomes, in our case, the non-survivors. And so what's really interesting to me is that this gene is only found in a subset of neutrophils.  So not all neutrophils express olfactamide in four. And so Matt Alder, who's one of our junior faculty guys, has worked this up some more.  This is flow cytometry to look at olfactamide in four positive neutrophils. The y-axis is staining for neutrophils. The x-axis is for olfactamide in four.  That box is all the neutrophils. And those are the olfactamide in four neutrophils. This happens to be a patient with a high percentage,  but it's remarkable that you have two really, really, really distinct populations. It's either there or it's not. And so Matt has worked this up.  In his first paper, he looked at kids with septic shock, and he basically quantified the percentage of olfactamide in four neutrophils. And what he found is that the percentage  of these neutrophils, the more of those that you have, it's independently associated with poor outcome. That's if you account for baseline illness severity, age,  comorbidity, et cetera. So he's pursuing this, thinking that perhaps this is a pathogenic subset of neutrophils. And imagine if that's true for ages now  in critical care, we've been thinking about anti-inflammatory strategies, but we realize that the neutrophils play a big role, but you still need them around.  Well, maybe there's one bad actor there, or one bad acting subpopulation that's olfactamide in four, and we have a way to identify them. Matt has chased this down further.  He's knocked the gene out in mice. And in your standard CLP model, you can see there that the mice that are null for olfactamide in four have a survival advantage.  For the pediatricians in the audience, this effect is actually even more dramatic in juvenile models. So Matt has a model where he can take little pups,  they're like the size of an eraser, and be able to do sepsis studies in them. And in that population, it's even more dramatic, the survival advantage that he can see.  So this is something that he's pursuing pretty aggressively, and you should see more of this in the future. This is something that we showed yesterday.  We've developed this biomarker-based model based on proteins that allows us to really nicely stratify kids with sepsis at baseline into mortality risk,  based on their mortality risk. And it's pretty robust. But the way that we found those genes, the way that we found those proteins, were through our original transcriptomic studies.  Related to that, this is another example. This is a good heads-up to the junior people out there, is that there are vast amounts of publicly available transcriptomic data.  So if you go to a database called GEO, which is the Gene Expression Omnibus, I think. And in Europe, it's called ArrayExpress. There are vast and vast amounts of publicly available  transcriptomic data. And so this is a really good example of leveraging that data. These were three groups across the country, the US, that is. I think, yeah, it was all US, who  downloaded all of the publicly available data for sepsis and septic shock across neonates, adults, kids, and so forth, and came up with parsimonious models for mortality prediction.  Great group effort, but it's also a good example of how you can leverage this publicly available data. The leader in this, in my opinion, has been Tim Sweeney.  He and Professor Kottry at Stanford have done a lot of the cataloging of the publicly available data. And they've had two papers in science translational medicine.  One where they had a diagnostic gene set to differentiate between sepsis and what we'd like to call sterile inflammation in critically ill patients.  And then they went on to find another gene set that distinguishes bacterial from viral infections. And so now what they're trying to do  is combine these two gene sets as a diagnostic  age-old question for us as intensivists is when our patients have fever and tachycardia and things like that, are they just generically inflamed or are they infected?  And so you may actually see this at the bedside sometime in the next four or five years. So that's sort of a whirlwind tour of what's possible.  What I want to focus on for the rest of this session is the use of transcriptomics to identify subgroups of patients. In this case, being patients with sepsis.  We always talk about how the diseases that we look after in the ICU are all very heterogeneous and how can we subclassify these patients.  And so four groups now have worked on this. You see a lot of different terminology out there.  We use the term endotype, but what you use doesn't matter. What's important here is understanding that what these groups are describing  are subclasses of disease, in this case sepsis, based on gene expression patterns. That's the fundamental exciting thing to me about that.  And so the approach has been pretty standard, microarray from whole blood derived RNA, using statistics to find genes that are differentially regulated  and those statistics vary from group to group. And then some sort of clustering algorithm. Again, those vary also, but the idea there is that you're grouping patients agnostic  to what they look like clinically, you're grouping them just based on similarity or dissimilarity of gene expression. So these algorithms know nothing about the patient.  They don't know if they're old or young, if they died, if they're sick or it's not so sick. It just clusters them based on similar patterns of gene expression.  And then an important thing is that all these groups have gone back in independent data sets or external data sets, they've validated their findings.  So this is the group from the UK, the UK Genomic Advances in Sepsis or Gain Studies. They looked at adults with sepsis secondary to community acquired pneumonia  and they reported two groups that they call sepsis response signatures one and two.  The SRS1 group has features of what they call endotoxin tolerance, T cell exhaustion and decreased expression of the HLA1 molecules. And this group also has higher mortality  and then they were able to turn around and replicate this in a subsequent paper.  This is a different group from the Netherlands, Tom Vanderpoel's group, the MARS project and again adults with sepsis. They reported four groups, okay,  that they call MARS1, 2, 3 and 4. Again, that common theme, the MARS1 group is characterized by decreased expression of genes corresponding to innate and adaptive immunity.  And that's the group that has greater 28-day mortality. Here's Tim Sweeney again. Again, downloading and cataloging all publicly available transcriptomic data  that's out there in both GEO and ArrayExpress and clustering to find subgroups of patients. And they reported what they're calling three groups  that they call inflammopathic, adaptive and coagulopathic. And those terms reflect what the genes correspond to, what sort of pathways they respond to.  And they found that the coagulopathic group had higher mortality. Again, a really, really good example of being able to use what's out there already publicly available.  And then back in 2015, we looked at kids with septic shock and we reported these two endotypes that we call A and B with those gene expression mosaics there.  The endotype A patients have decreased expression of genes corresponding to adaptive immunity in the glucocorticoid receptor signaling pathway,  which I would argue are probably pretty relevant to what we see in the ICU in this patient population. And the kids that are endotype A have higher mortality  and organ failure burden. And interestingly, among those kids, if you're endotype A and you get corticosteroids, your odds of mortality go up.  And we've been able to replicate this or validate this. And recently, I think I have a slide for that. No, I don't have a slide for that. But we were able to replicate that  in an external adult cohort.  And I showed that in a session yesterday. So those are the four groups that have reported these sort of molecular subclasses of sepsis.  They've all reported differences in outcome. But I would argue that that's not the goal. That's not the goal of these efforts, right? The goal here is to identify subgroups  that have biological commonalities. And then therefore, it could have the potential to inform therapy, this concept of predictive enrichment.  Perhaps for the journals and the reviewers and the audiences, the differences in outcome strengthen the association or the interest.  But again, I would argue that that's not the case, that that's not the important thing. Because imagine if you found two groups, their mortality was the same,  but you found that one group was more responsive to, I don't know, protein C, right? That would be intrinsically very, very important, even if at baseline, the two groups  have no difference in outcome.  But not everybody agrees with me on that one. So here are the four studies, right, all the way from 2015, and then after that, one every year. Is it a bunch of nonsense?  Are we all saying the same thing, but through different gene signatures? Or we're not sure yet? And I think we're somewhere between the latter two, and I'll show you why.  So there's commonalities, right? They've all been found. All of those groups have been found using this sort of unbiased subgroup discovery where you just allow the computer  or the algorithm to cluster the patients, agnostic to clinical phenotype. They've all described a subgroup that has patterns corresponding to adaptive immune system, right?  And if that's true, that's really important, right? The new rage now in sepsis is that we're gonna use these immunostimulatory strategies to improve outcomes in sepsis.  If what these different groups are reporting, it's not too hard to imagine that if you apply these immune stimulatory strategies to all patients, regardless of their subgroup,  that they're gonna behave very differently to those therapies, all right, if that's true.  Remember, though, we're all doing this based on gene expression, and immune dysfunction is inferred, right? None of us have actually measured immune function.  We're inferring changes in immune function based on gene expression patterns. All show association with membership and outcome, and again, I don't think that's really  the important point there. So here's some overlaps. Julian Knight's group said, all right, is SR1 similar to pediatric septic shock endotype A?  And there is overlap in the signaling pathway, but very minimal overlap in the exact genes that define the subgroups. Tom Vanderpoel's group said that there was overlap  between MARS3 and SRS2,  Tim Sweeney's group showed a lot of overlap between SRS1 and our endotypes and their inflammopathic and adaptive clusters. And that's not too surprising because they  used all of our data. They cataloged all that data and put it together. So I'm not surprised that they were able to find more overlap. So we've looked at this.  We downloaded Julian Knight's group, their data from the UK. And we assigned their patients to endotype A or B. And again, just to remind you of some commonalities,  the biological features there is basically repression of the adaptive immune system in some way, at least at the gene expression level. And they find increased mortality in SRS1.  And we find increased mortality in endotype A. So functionally, those two groups seem somewhat the same. But when we categorized Julian Knight's group,  this was a two-by-two table there. And the statistics say that there's overlap. But your eye tells you that there's not redundancy there. They're not really that closely aligned.  But remember, Julian Knight's, the GAIN study, the median age was 65.  Our median age is 2.5 years. So that's a vast, vast difference with respect to comorbidities and so forth and that kind of thing.  So we looked to see if there's interaction with age. And this is a little bit of a complicated slide. The y-axis is the observed risk of mortality. The x is deciles of age.  And then the patients are grouped into A1, A2,  in other words, the overlap with Julian Knight's group. And what's striking here is that if you're both A and SRS1,  what it means, but it's kind of interesting. This is a little bit of a different way to look at it. It's the predicted risk of mortality based on a model that tests the interaction  between endotype SRS and age. And again, you see that younger patients who are A1, in other words, are endotype A, and Julianites SRS1, those patients do really, really poorly if  you're younger. Again, we need to figure out what that means. So getting back to the original topic here, integrating translational genomics, this is  just one example, right, of the subgroups that we could potentially integrate. But I think we need to harmonize, we need to get together and find a consensus grouping, right?  We're all doing different things around the same area. And to me, it doesn't matter who's right or who's wrong.  There's great precedent for this, right? So this was people working in the colorectal cancer field, and they were at the same stage six or seven years ago.  They had four or five different investigative groups who they'd come up with their own molecular classification based on gene expression of colon cancer. They worked together.  They organized their.  So, in any case, just to take home messages here, four groups have reported these subgroups and validated. The validation part is really, really important.  All share this common theme of immune dysfunction. All report associations with members  Not sure if we're all describing the same thing with different gene signature, but the important thing is that we have an opportunity now to collaborate, work together.  There's a vast amount of data.  Hopefully, we'll get to that. So, thank you for listening. I hope that was useful.  By my watch, we have four minutes to stay on time.  I'd like to ask you a question that's been nagging me for 20 years. When we were doing our studies on endothelin activation of macrophages, when we made the  rats diabetic, or when we took human macrophages, boy, did the receptor biology change. Glycosylation and glycation of receptors, I mean, a transfusion reaction is really just  an abnormality of glycosylation, not proteins. So right now we're all fixated on genomics and we're fixated on proteins, but where's  glycosylation and all of these post-translational modification steps that may mean so much in this very dysregulated group of patients we all take care of?  Yeah, that's a great and sobering point because you're absolutely right. It's even one step further, right?  We're assuming that these changes in transcription are leading to changes in protein expression  and that may not necessarily be true, and then there's a whole other layer that you're referring to as post-translational modifications of proteins, whether it be glycosylation or  other things. Yeah, this is missing here, right, and I think that's another layer that needs to be disentangled.  Part of the challenge, part of the challenge, though, is those of us that are trying to get some of this work to the bedside is that we don't have a lot of time for some of these  diagnostics, right? I mean, if you have a tumor, you can have that tumor be characterized and you can take  two or three weeks and things like that and come back with something more comprehensive. In our world, as you know, you have two or three hours to make these decisions, so some  of those things are really challenging, but no doubt, great point, sobering point that makes me step back is that, yeah, there's a whole other layer of regulation that none  of this is even touching. It's not accounting for it at all, so very well put. My question, this is great work.  I'm curious as to how much timing matters and how stable these classes are and if it really, when we look at the differences across the studies, if that could explain some of it.  Yep, yeah, right, so our patients are very dynamic. Things change fast, so I'll speak for our own data in that if you look over a two-day  period, about a third of the kids shift endotype, for lack of a better term. They'll switch from A to B or B to A, things like that, and the other ones stay pretty  stable, and that's important because if you stay in A, you do really, really, really poorly. If you start out as an A and you change to a B, you do okay, so that temporal shift is  okay, but then that gets to the world of diagnostics and things like that, right, is that when are you capturing these patients, and that's a big challenge.  I mean, one of the things that I like about our data is that the way we enroll patients, it's pretty unselected, so I think it reflects what you and I do at the bedside, so the fact  that we've been able to validate some of this stuff even within that mess, it's been pretty  stable, but yes, without a doubt, that dynamic element is really, really critical, and everything  that I just showed you from the subclasses, whether it's Julian Knight's group or our group, is a single time point, and so that's really important, but it's a single time point  early, right, in the acute phase, which is when you want to make decisions. That was great, Hector.  Coming from someone who can't even spell sepsis, is there a, does pathogen lead to specific subclass, or do all pathogens just lead to A or B?  Yeah, so we've looked and looked, at least in our data set, we've looked really, really, really hard, because at the end of the day, sepsis is an infectious disease, right?  So you would think that the pathogen would lead to different signatures, and we have not been able to find a pathogen-specific signature. I mean, we just haven't.  And maybe it's a numbers thing. Our database is large enough now that we have a lot of kids, say, with pneumococcus or E. coli and things like that,  and we just haven't been able to find that. You would think that there is, but we just haven't. I know Julian Knight's group, they compared signatures from people  with community-acquired pneumonia and also intra-abdominal sepsis. They called it fecal peritonitis. Very little difference between those two.  But it is an infectious disease, so I think there is gonna be an element of that there.  Yeah, last question, and then we're gonna get moving for time. So you showed us that the mortality was different in those subgroups with age, that one of those last graphs.  Does the proportion of patients at a specific age that fall into each subgroup change?  So is the likelihood of being in one subgroup versus the other change with age? So in our data, the A patients, if you will, tend to be younger.  They tend to be younger. And younger kids are at more risk of mortality, but if you account for that, allocation to A needs that. That A1 business, it seems like younger people  tend to overlap to that A1, but we need to validate that.  Thank you for your questions. I'd love to talk more about it, but out of respect for our other speakers, we're gonna keep going. Rick, I think you go next.  So Richard Levy from now Columbia University is going to speak with us now about  a whole nother layer of complexity, the mitochondria, right? So none of this gene expression actually relates to that, so that'll be good. Thanks, Hector.  Thanks, Hector, and I want to thank the organizers for the opportunity to speak with you this morning.  Let's see if we can get this going. So, let's see, here we go. No conflicts of interest to disclose. So mitochondria are responsible for a number of important roles within the cell  to maintain cellular homeostasis as well as to mediate cell death. And we know from inborn errors of metabolism and genetic mutations that dysfunctional mitochondria  can lead to impairments of virtually every cell type,  tissue type, and organ system.  The recognition of the importance of mitochondria for health and disease is readily apparent when you look at the sheer number of publications  over the last 40 years regarding mitochondria. You see the steady rise here in the number of publications, and currently, mitochondria lead all organelles  with regard to study in the medical literature to date.  So the link between mitochondrial dysfunction, acquired mitochondrial impairments, and critical illness has been pretty well established for a number of critical illnesses,  such as sepsis, acute lung injury, ICU-acquired skeletal muscle dysfunction, acute renal failure, critical illness-related immune dysfunction, and even with regard to DAMPS,  danger-associated molecular patterns, it's now recognized that mitochondrial components can serve as DAMPS, including mitochondrial DNA that is released into circulation  and can mediate systemic toxicity. So oxidative phosphorylation is probably one of the more important processes that mitochondria carry out.  This is the process of generating energy, cellular ATP via aerobic respiration. The majority of mammalian cells utilize this process to generate cellular energy.  And we can assess this type of mitochondrial function in preclinical studies using both in vivo and in vitro techniques. In an in vivo technique,  we can use magnetic resonance spectroscopy to look at mitochondrial energy capacity, as well as steady-state ATP synthesis  To do this, we would use a P31 isotope, and we can quantify high-energy phosphates in various tissues, as well as phosphocreatine in organic phosphate.  If we use C13, we can quantify  Krebs cycle intermediates in the tissue to get a sense of TCA flux. The advantages of this approach is that it's non-invasive, it's physiologically relevant,  but the downsides, obviously, is that it requires very expensive equipment and, obviously, technical expertise. We can also measure mitochondrial function  and oxidative phosphorylation in vitro by measuring enzyme activity using spectrophotometric approaches, as well as respiration via polarographic approaches.  And we can measure ATP via bioluminescence.  This is one of the nicest reviews on the methodology  in steady-state enzyme kinetics in the electron transport chain. It appeared in Nature Protocols in 2012. And the concept here is that we measure  each individual electron transport chain complex activity in the presence of excessive electron donors and excessive electron acceptors. And we can monitor the change in absorbance  of either the electron donor or receptor on the basis of reduction or oxidation of those donors or receptors. And we can then measure the activity  in the presence of specific inhibitors, such as rotenone for complex one, antamycin A for complex three, or cyanide for complex four. So we can get inhibitor-specific activities  of each of these enzymes. We can then normalize them to a matrix enzyme within the Krebs cycle citrate synthase to get a further level of normalization  above and beyond milligrams of mitochondrial protein. Now, because these assays are performed in the presence of excessive substrate,  we're really measuring the steady-state kinetics that approach the maximum velocity, or Vmax. And we can augment this approach by doing these assays  with different levels of substrate. And that allows us to determine 1.5 Vmax, or Km. And we do this by generating a reciprocal plot to then determine Km by the x-intercept,  or Vmax by the y-intercept. And this allows us to compare different samples to determine if an enzyme is inhibited, and then the type of inhibition.  And we performed this type of approach a number of years ago to demonstrate that cytochrome oxidase was non-competitively inhibited in the murine septic heart,  48 hours after a double puncture. What you see here is the septic mice, the Vmax is significantly decreased compared to controls, whereas there was no change in Km.  Mitochondrial respiration, O2 consumption, is a very important variable that can be measured using the polarographic technique,  using amperometric O2 sensors, or phosphorescent probes. With the amperometric O2 sensors, we use a Clark-type electrode, which is a platinum-coated or gold-coated electrode  to measure oxygen content in the buffer. And in 1958, Britton, Chance, and Williams defined these different states of respiration in this manner.  And in 2002, these states were modified by Nichols and Ferguson. And this assay can be done in isolated mitochondria, permeabilized cells and tissue, or in cells in suspension.  In 2006, the Seahorse Flux Analyzer was introduced into the field. And this was a significant advance because it allowed measurement of mitochondrial respiration  in intact adherent cells, and also permitted us to measure respiration with much smaller volumes in a high-throughput manner.  Here's an example of the curves, the O2 consumption curves you get with the polarographic method and the Clark-type electrode. Here are the different states of respiration.  You have state two, following the addition of substrates, in this case pyruvate and malate. You have state three, respiration after ADP.  State four, induced oligomycin state respiration, which allows us to determine proton leak rates. And then you have maximum uncoupled state five respiration  after the addition of an uncoupler or an ionophore, such as FCCP.  Now here's an example of high-resolution respirometry as measured with the Ouroboros technology. This is, again, a polarographic method.  Similar concepts here with inhibitors and uncouplers. Here, the advantage here is that much more sensitive technique, sensitive detection of oxygen, and much smaller volumes.  So it allows for smaller sample sizes.  And here is an example of the Seahorse technology of how investigators report oxygen consumption rate. Again, the concept is the same.  It's oxygen consumption rate in the presence of inhibitors, such as oligomycin, rotenone, antamycin A, and uncouplers, such as FCCP or CCCP.  And this allows measurement of basal respiration, leak rates of respiration, and maximum respiration, as well as ATP turnover.  Now membrane potential is another important variable that should be considered and should be measured. And this is the electrical component  of the proton motive force that's generated by complex one, three, and four hydrogen ion pumping in the process of oxidative phosphorylation.  Now the membrane potential can be quantified using cationic fluorescent probes, as well as the lipophilic cation TPP. When measuring with the fluorescent probes,  we do this with fluorescence microscopy or with flow cytometry. And when we do it with the TPP probe, you can do this with an ion selective probe,  and you can actually do it while simultaneously  measuring oxygen consumption in that manner as well. The concept here with the TPP probe is that all of these cations readily permeate  and diffuse across the mitochondrial membranes. And as the membrane potential is generated within mitochondria, the more negative the matrix becomes.  This attracts the cations into the matrix of the mitochondria, and we can quantify the amount of TPP in solution, and determine what the membrane potential is  by the Nernst equation that will allow us to determine how much TPP entered into mitochondria.  On the left here is an example of quantifying the membrane potential using TMRM fluorescence. Here you see these are neurons in culture. And the investigators here quantify  the relative amount of fluorescence  during different states of respiration and in the presence of inhibitors and uncouplers. The limitation with this technique  is you oftentimes just get relative fluorescence and a relative membrane potential, not an absolute number. So your controls will be very important in this technique.  This is work from our lab, and we like to simultaneously measure the oxygen consumption with a polygraphic Clark-type electrode, along with the TPP ion-selective probe.  And what this allows us to do is to then correlate changes in respiration with changes in membrane potential to really get a better sense of the mitochondrial function  with regard to oxygen consumption and membrane potential, and to really determine if there is dysfunction and what type it is.  So we really believe that one should measure O2 consumption simultaneous with membrane potential to really have an understanding. And that's because everything with regard  to mitochondrial O2 consumption relates to the proton motor force or the membrane potential. There are three modules or three processes  that revolve around or relate to this proton motor force. Substrate oxidation, proton leak, and ATP turnover.  There's only one process that generates the membrane potential, and that's the process of substrate oxidation. This is the process within the electron transport chain  of oxidizing either NADH or FADH2 to generate the proton motor force by pumping of hydrogen ions by complex one, three, and four. There are two processes that degrade  or consume the membrane potential. That's proton leak and ATP turnover. Now, proton leak is probably, I would say, fairly underappreciated and oftentimes overlooked.  And it's a very important process. It is a physiologic process, but can also be pathological.  There are leak channels that mediate this proton leak, and they're analogous to that little hole in the top of your bathroom sink to prevent water from overflowing.  If you consider the amount of water in the sink as the membrane potential, the mitochondria need to make that membrane potential to generate ATP.  The ATP synthase uses that membrane potential. However, you don't want it to overflow just like you don't want the membrane potential  to get too high because that will lead to oxidative stress. So mitochondria have these leak channels, which are analogous to these holes,  to prevent that overflowing of the membrane potential. And you can consider in the analogy here that the water flowing from the faucet is electron flux.  So these are necessary and important leak channels. However, you can also have leak that's pathological. For example, if you break a pipe or the drain has a hole in it,  you'll have a leak here of water, and in this case, you'll have a leak of protons. And in response, the mitochondria will need to rapidly generate the membrane potential  to keep up with the leak rate in order to make ATP. And here, the mitochondria would have a much higher electron flux and O2 consumption. So we can quantify the different processes  using a process of modular kinetics, an assay called modular kinetics. And the reason why it's important to do modular kinetics and break those three processes apart,  and most people overlook this, is because if you do standard O2 consumption without membrane potential, we miss the fact that in any given state,  many of these processes are happening simultaneously and overlapping, and they can confound each other or even be additive.  So you can have substrate oxidation, ATP turnover happening and changes happening in different directions. State three may look totally normal,  but in fact, substrate oxidation and ATP turnover can be totally abnormal, and you would miss that nuance. So we perform modular kinetics in a dynamic way.  And here, all three of these curves are overlapped just for simplicity. But we can measure ATP turnover, substrate oxidation, and proton leaks separately and independently.  And we do this by either inducing state four with oligomycin or state three with an uncoupler, and then titrate the respiration by using inhibitors or uncouplers.  And the purpose is to actually change the membrane potential over a range of membrane potentials to see how the mitochondria defend that membrane potential  with a change in respiration. And you can compare samples to determine how ATP turnover, substrate oxidation, and proton leak differ.  And here's an example of why this is so important. These are proton leak conductance curves. And if you look at the middle panel here, you have an X sample and a control sample.  State four respiration oligos, which is where you start the assay, are here in the black dots. And if we were only to measure state four oligomycin,  we would say, oh, they have the identical oxygen consumption rates. These samples are identical with regard to their state for oligomycin leak rate.  The reality is now, if you do it in the context of the membrane potential in a dynamic way over a range of membrane potentials, you see that these have markedly different  leak conductance curves. They have markedly different proton leak in each sample. And the way you interpret this data is you take the highest common membrane potential  and then compare the rates of respiration at that membrane potential. And you can see the X sample has a markedly higher rate  of respiration to defend that same exact membrane potential. That's massively inefficient and has a significantly higher amount of proton leak in that sample.  And that nuance would be completely missed if you simply did O2 consumption without knowing membrane potential. OK, I want to switch to calcium uptake, which  is a very important process that mitochondria mediate for calcium homeostasis. But we also know that if mitochondria take up too much calcium, it can induce the mitochondrial  permeability transition pore to open. And we can quantify the threshold for that pore opening to determine if the mitochondrial transition  pore is regulated normally or if it's pathologically open. We can do that with calcium retention capacity assays or with calcium loading capacity assays.  With the calcium retention capacity assay, we add a fluorescent calcium in pulses to the mitochondrial sample. And you can see mitochondria rapidly take up calcium  as soon as it's pulsed in the medium, as evidenced by the rapid downward deflection here. At some point, the calcium will trigger the pore to open.  And that's evidenced here by calcium being returned back into the buffer medium. And we can quantify the threshold for the pore opening because we know how much calcium we've added.  Here you see cyclosporine, which is a transition pore inhibitor, allows these samples to retain much more calcium because it's making the transition pore more resilient.  We can do a variation on a theme with a calcium loading capacity assay. Here we add all the calcium on the front end. And we calibrate the amount of calcium  that we've added to the sample. We then energize the mitochondria with a substrate, in this case, succinate. And you can track and follow the uptake of calcium  into the mitochondria. And then we can see where the transition pore is triggered here by the plateau of this curve and then when it's released back into the buffer medium.  And here you see cyclosporine allows these mitochondria to retain much more calcium. And we know exactly how much calcium is retained here  and taken up by the mitochondria because we've quantified it on the front end. OK, another really powerful technique is to do the calcium cobalt assay  to assess for mitochondrial transition pore opening. I like this assay a lot because it's very visual. And you can actually see, via fluorescence microscopy,  evidence of the pore opening. The concept here is that calcine is a fluorescent probe that gets into all compartments within the cell.  And it gets into the intramitochondrial compartment. We can then quench the fluorescence using cobalt chloride. When the pore is closed and intact,  cobalt cannot get into the mitochondria. And so you're left with these beautiful punctate green mitochondria fluorescence. However, if the pore opens, cobalt  can then get into the intramitochondrial space and can quench that intramitochondrial fluorescence. And that's evidence that the transition pore is open. This is work from our lab.  We demonstrated this assay in vivo using intravital microscopy. This is in the brain. This is a single neuron that we tracked over time in layer two of the neocortex of a P10 mouse.  We loaded and exposed the brain to calcine and TMRE to highlight the cells here. And you see that this neuron takes up calcine. And you see this very nice green nuclei.  And then you start to see the punctate green mitochondrial fluorescence. And you can see the perinuclear respiring mitochondria here in red.  We then quench with cobalt. And what you see is these beautiful green punctate mitochondria here. And over time, the mitochondrial fluorescence becomes quenched.  And it dissipates pretty rapidly. And that indicates that the pore is pathologically open in these mutant mice. OK, so just as with what Hector was talking about previously  is that omics has come to mitochondrial biology. We can do genomic approaches, metabolomics, proteomics to look at super complex formation,  interactomics, interaction between mitochondria and other organelles, such as the endoplasmic reticulum. We can do transcriptomics, epigenomics, and lipidomics,  which is not shown here, to look at lipid structure within mitochondria. And when we do this type of techniques, we get a very complementary approach  with omics and the classic functional assays to get structure and function information to tell us about mitochondrial integrity and whether they are functional or dysfunctional.  OK, in the last few minutes, I just wanted to talk about mitochondrial morphology and imaging, because this is really the state of the art of where the field is moving to.  Using a variety of fluorescence techniques, investigators are looking at shape, content, networking, dynamics, and fusion and fission events.  Here's an example of imaging of a single cell, an HEK293 cell. And what the investigators did was to use Mitograph Surface View, which is a publicly accessible, open platform,  to allow them to identify 3D anatomic structure of the individual mitochondria. And it also allows them to look at networking  by identifying the skeletons and the nodes of the mitochondria. Let's see if we can get to the next slide.  I don't think you can. Can we?  OK, good. So next, folks are really interested in looking at contact sites, looking at fission and fusion, as well as interaction with the endoplasmic reticulum.  This is done using combined imaging techniques. Here, the red is the ER-stained red. You see a single mitochondrion here in blue.  And the contact sites between the mitochondrion and the ER is in green here. And what they've identified here is a nice fission event. And you can see the mitochondrion clearly  undergoes fission in that image. Other folks have looked at stimulation emission depleted microscopy. And they've developed new and better fluorescent probes  to allow this beautiful imaging. Here, you see intercristae emergence. I want you to notice the dynamic activity of the mitochondrion here. They move around quite a bit.  And they are quite dynamic, just in steady state here. And you see intercristae emergence there. This image shows mitochondrial fusion. You're going to see the fusion event here  with the two mitochondrion here at the bottom. You're going to see they come together. And then they will fuse.  So this is a really powerful technique, which is obviously very complicated. But it's really important to be able to see these types of images. Now, this is an interesting image  showing mitochondrial rupture of this mitochondrion in the center here, which is probably a prelude to apoptosis if you have more than one  mitochondrion do this, as cytochrome c and other things are released into the cytosol. So in conclusion, mitochondria are certainly  important to health and disease and critical illness. There are various techniques to assess mitochondrial function that are available. And I would argue that we should be  looking at more than one technique when we want to characterize mitochondrial function. And the high-resolution imaging is a rapidly developing,  state-of-the-art technology that we're going to see much more of. So I want to thank you for your time. And thanks for the opportunity. Thank you.  Thanks for a really interesting talk. I wondered if you had thoughts on the importance and potential relevance of reactive oxygen species generation by mitochondria and how  that may relate to pathogenesis of disease versus sort of normal signaling and disentangling. Yeah, that's a great question. ROSs are important for homeostasis  as well as mediators of pathophysiology for sure. And I think to build on a little bit of what Hector was talking about this morning, why we are so simpatico is we're  both pediatric guys too. So I think with development, we know that the pore, for example, in cells that are developing from the embryological state to mature,  the transition pore is very important with triggering the switch from glycolysis to aerobic respiration. ROSs mediate that as well. These are really important questions.  And there are several techniques that I didn't have time to get into to quantify ROSs. There are obviously caveats with those techniques,  but very important and certainly round out the picture.  So if I shock a rat into a no-reflow phenomena so that it can't clear its lactate, even though I get its blood pressure back, and I look at its mitochondria, what am I going to see?  That's a great question. The field is certainly interested in the reperfusion phase of resuscitation. That's where the damage is believed to be mediated.  And it's largely a ROS event from complex I. And there are people who are working on complex I inhibitors during the no-flow to prevent the reflow, reperfusion ROS event.  There are others. I have a colleague at Wayne State who's working on a near-infrared light of a certain wavelength to inhibit cytochrome oxidase to modulate the membrane  to prevent that rebound effect of a hyperpolarization, which is also a ROS-mediated event. So the long answer to your short question is it's complicated.  But you're going to see a lot of different things depending upon organ system, tissue type, where you're looking within the cell, as well as timing.  And this gets to the question that was raised earlier of when you're looking is critically important. These are all variables that I think  are poorly controlled for in the literature and must be looked at carefully.  All right, our next speaker is Ken Remy from Washington University. He's going to talk to us about models, I believe. Ken did not get or read the instructions  about haircut or facial hair. But we will get him up here anyway. It's good to be the junior guy on stage. You get a little bit of teasing.  So actually, before I start, a quick question. How many people in the room are part of the research section? Raise your hand.  Not everybody, OK. So if you wouldn't mind, so one of the hats I wear is I'm the chair of the basic and translational science section.  Did you even know that we had one of those, too? So we actually, under the research section, have one of those as well. And this year, we're going to be doing  a lot of increased conversations about some of the things we're talking about today, including next year having a pre-course, hopefully.  So please join so we can get more faces into the group. That's my first plug. My second plug is I'm going to not talk about heavy science today.  I feel good not talking about my main area, which is basic science related to heme modulation and signaling pathways in the immune system.  We're going to have more of a philosophical conversation about how to build better animal models. But it'd be great to hear people's thoughts at the very end.  And I'll try to go rather fast. And there's some slides I'll breeze over in the interest of time so you can look at afterwards. But this is what I've been tasked to do,  is to discuss the strengths and weaknesses of our current animal models, including humanizing models, and focus on the relevance of these to human critical illness.  And so I ask these questions up front. Are our models flawed? Are we asking the right questions that are being appropriately answered with our current models?  Are we accurately modeling diseases in our animal studies? And then finally, are we accurately modeling our clinical studies to our basic science findings?  I think these questions probably can sit for many days. And we can talk for weeks on this. But I think these are the key questions that we need to ask  so that perhaps us as a society and as the scientific group that's doing critical illness trials really start characterizing what are the best ways to move forward  so that we can potentially use these wonderful omics and scientific approaches in humans. So I get to pick on my mentor for a moment, Richard Hotchkiss,  frankly, because he was supposed to give this talk. So I got asked to come in at the last second here. But every day, I ask about a new study that I'm going to do in my laboratory.  And he continually tells me the best model for human disease is the human. He's in part correct, although he's built his career initially on animal studies.  So I give him a hard time. And I continue to do animal work. But why do we need animals? We do need them because they serve as a good model  to help us understand how living tissues function and the biology of the underlying disease. It's that by understanding these whole model approaches,  we're able to develop treatments or cures, if you will. They allow researchers to understand the fundamental ways in which the diseases affect living tissues.  We can't model them with computer models because they can't replicate the complicated interactions of the whole system that we see in a whole animal model.  And furthermore, it's required by most funding agencies as well as regulatory agencies to move towards humans. So what have we learned? We've learned a lot.  I know animals get a bad rap these days. But in fact, a lot of the therapies that we currently have were derived initially in animals, if not all of them.  And those include and are not restricted to a number of different diagnostic tests as well as vaccines associated other therapies. So how have they helped?  This is just a quick list. Certainly Landsteiner and Popper, Salk and Sabin used animal models for polio, from diabetes to understanding how insulin worked in canines,  to cystic fibrosis. That went quicker. Sorry. To cystic fibrosis, respiratory distress syndrome, hypertension, obesity, and also the development of TPA and stroke.  These were instrumental, and we would not be where we are today without these models. In 2010, a European commission was convened to evaluate the utilization of murine research  and understand better ways to develop models. And they asked a couple of different questions. And these four, I think, were the most interesting.  The mouse model is the most common used model, but is it the best one? How useful and valid are these models for mimicking human disease? What characterizes these mouse models  have proven useful for basic science, drug discovery, clinical research, and what measures are needed to improve new and existing mouse models?  The same questions I started this conversation keep coming back. So can results from human studies really be applied to humans? There's a lot of similarities.  We know from Richard and others' work that immune function in mice is pretty similar. Cardiovascular function from a lot of the work  at the NIH in the 1980s, demonstrated in a canine model, certainly has been helpful. But they provide us an index of safety. They allow us to evaluate something  that ordinarily we can't do in humans. And in fact, almost all regulatory agencies, as I mentioned, require this. So if you think about the mouse, about over 95% of all the genes  are homologous in mice to humans. The difference really lies in really the redundancies and gene expression differences. So let's look at a couple early studies.  So Richard's group had looked at both human and mouse differences in apoptosis and sepsis when they looked at both the thymus and splenocytes  and found pretty high similarities between both. This was done at autopsy. Furthermore, they looked at BCL2 and CASP3 pathways in both mice and in humans and demonstrated  pretty significant similarities. And although there were differences in global gene expression, there were key similarities in the key pathways.  Furthermore, in 2002, Richard wrote a paper that demonstrated that there was significance in looking at, at least in his hands, his mouse models to human sepsis,  looking at loss of DTH response, apoptosis, the Th1 to Th2 class shift, increase in Tregs, T cell exhaustion, PD-1 expression, increased MDSCs, and defective neutrophil  functions, oxidative burst. And so in 2020, we've got a number of different other resources that could help further refine understanding these similarities and or even differences.  So we've got CreLOX mice and CRISPR, which Dan Chertow is going to talk about today. They both confer different benefits. CreLOX have great temporal and tissue-specific control  of modification. And CRISPR don't really have that, but are able to be used in both different organisms and cell culture and can be paired with CreLOX.  And they allow us to further refine and modulate specific sites of existing genomes to be able to further evaluate some of the findings  that some of the folks that preceded me today discussed. Furthermore, we've got the ability to humanize our mice. We can take immunodeficient mice and reconstitute them  with a human immune system by injecting RAG2 or L2RGs with a specific human gene of interest, with the immune cells in vivo. And then furthermore, we can further  evaluate basically a human immune system within a mouse and screen and identify highly effective combinations of different therapeutics.  There are some drawbacks still, some both humoral and complement mismatches that still need to be further evaluated. There is a virome and microbiome differences  that need to further be elucidated. And lymphoid architecture, that certainly is an area that needs to be further evaluated. But why do we need better animal models?  I mean, seemingly, I just told you that they've been great. We've been able to cure polio. We've been able to use TPA and stroke. Well, they don't sufficiently with certainty  actually model humans. In fact, one third of all successful animal studies translated to human trials. But one tenth of those only led to an approved intervention.  So a significant difference actually made it to patients for benefit. Why? Well, there's a whole host of reasons. There's internal and external validity issues,  including selection, performance detection, and attrition bias. There's variation in single center animal studies that leads to irreproducible data.  There's genetic drift, strain differences. There's environmental issues. We don't think about this. And it was great to hear Hector talk  about using 14-day-old mice, because that does accurately reflect the human population of interest. But most of us look at animals between the ages  of six weeks and 10 weeks, which really only reflect a 20-year-old. If we did all of our sepsis trials in 20-year-olds, it'd be very difficult to really differentiate a benefit,  because 20-year-olds usually don't die. We never really get to understanding senescence if we're studying six-week-old mice.  And so we really need to start studying the respective ages that we're interested in from humans in the animals at that age that we are reflecting them. So some other issues.  So this paper caused a nice stir in 2013 and led towards a number of different editorials in The New York Times and other journals. Siak et al. looked at model validity  and looked at correlation between human genes and mouse genetic studies and tried to see, really, where the correlation lie. Do mouse studies really inform human studies?  And so in this figure here, what you notice in the upper left-hand corner is they looked at about a little under 5,000 genes in humans and their murine orthologs and looked  at both human burn, trauma, and endotoxin and looked if there was a correlation between the mouse models. And what they found is excellent genomic responses,  at least in human trauma and in burns, from human studies. But what they found was that the mouse studies actually poorly correlated with one another  and were random when it looked at human studies. You could flip a coin from the mouse studies and then figure out whether or not it was informative.  She further evaluated a number of things over time. So they looked at genes for human burns, trauma, and endotoxin in the murine models. And what was interesting is that over time,  the genes changed significantly over time, but very minimally in mice, not really reflective and similar between the two species. And then furthermore, they looked  at endotoxemia at activated pathways and suppressive pathways against three murine model pathways and really found that the human studies correlated quite nicely with one another.  But when looked at with comparison to mice, very little correlation. So how do we use this information to build better models?  Well, first and foremost, we've got to avoid over-interpreting our animal results. We actually have to evaluate sex, age, and strain differences.  We've got to allow for better internal and external validity of these animal studies. We need to study the animals out to the same age-related endpoints.  Now, this is something that doesn't commonly occur, but we probably need to start instituting this. Not only for our IACUC protocols, we have to have sample size calculations,  but we should re-reporting them into our manuscripts. We should have an eligibility criteria and treatment allocation plans, understanding how we do our blinding,  what's the flow of animals, how do we control for the different physiological variables, control of our study subjects, and then describe early what our initial statistical analysis  approach was supposed to be. We also should consider having co-clinical trials. So co-clinical trials are important because it allows us to evaluate in mice,  in the mouse clinical space, the same way that we take care of patients in the hospital. It allows for having treatment protocols and the same outcome measures.  So I'll give an example, at least from some of my background. So Chuck Natanson at the NIH, I worked in his lab and we looked at the storage age of red blood  cells between the continuum of seven versus 42 days because the FDA allows for 42-day-old blood to be transfused in humans. That's the shelf life of the Twinkie, if you will.  And commonly, patients will get the oldest blood first and then will get the younger blood if they're out of the old blood.  Usually more in adults than children, but commonly happens. But studies really haven't evaluated this. And so an animal model is helpful  because it allows us to look at those youngest versus the end of their shelf life. It could also further define where the risk occurs, whether it's the age of the product,  the volume of transfusion, or the clinical populations at risk. So in Chuck's model, we induced sepsis via staph aureus via bronchoscopy.  The animals had mechanical ventilation, sedation, antibiotics, cardiovascular support, as we would do in humans. And over 96 hours at each time point,  with Swangan's catheter in place, a central venous catheter, an arterial line, a tracheostomy in the animal, you're able to mimic exactly what we do at the bedside with humans.  They had stress ulcer prophylaxis. They had DVT prophylaxis. They had an aggressive support regimen that included fluid resuscitation.  And over time, they had a number of different parameters that were measured. And at 96 hours, you're euthanized. And we evaluated this with an exchange transfusion  for either seven versus 42-day-old blood that was dog-matched that met the FDA gold standard. And these are the results. And it was in this study that they demonstrated  that 42-day-old blood had worsened survival in comparison to seven-day-old blood. Great study. But the problem was is that if you  look at all of the storage age or chronologic stage studies that have been done from neonates all the way through adults, across the board, the randomized control trials really  have demonstrated no difference. The problem is they didn't do this. They didn't explore the early-aged blood and old blood. All of these studies uniformly evaluated  young and younger blood. They actually never looked at past, really, 21 days. So in fact, they didn't actually mimic the animal study that  was informative to demonstrate that there may, in fact, be a risk that lies here. And so we still don't have the answer to this question. And in this case, I don't believe  it's the animal study that failed us. I think it's the clinical trial that didn't take into consideration the model that  was demonstrative of a risk and then evaluated that in humans. I'm going to move past this, but this was just a nice chart that I was able to comprise from this paper that  looked at some of the problems using humanized mice and ways to overcome some of these limitations. So I come to this real question. Is it the intervention studies and human trials?  Do they fail because the animal study is flawed? Or does the model itself just not translate? Or are we evaluating findings in animal studies in flawed clinical trials?  And should we consider, and this is what I'm hoping this year we could accomplish as a white paper, evaluating consensus guidelines  and critical care that delineate from bench to bedside how we should translate our studies? So Richard Hotchkiss and I are embarking on a number of different things.  And one of them is to, in our animal models, we're mimicking exactly the same thing in humans, which is we're using a real-time ELISPOT platform  to look at, simplistically, the innate and adaptive immune system by basically counting the number of cells that produce the cytokine of interest after stimulation  and then give those animals in vitro, their whole blood, after separation for PBMCs and lymphocytes, a specific intervention. And what we see in vitro response,  we then give back to the animal. And we've been able to actually evaluate this also in humans and have demonstrated that we can find responders versus non-responders  and can use a precise way to perhaps use some therapies that have failed previously in human trials. They may have failed human trials  because they were misapplied to the wrong populations. That's, in part, part of the issues, I think, with some of the modeling from animals to humans  is that we're not applying the data that we have before us to the appropriate human populations. We're trying to hit a power, a number.  But we're actually not actually using the right patients at the right time for the right therapies. So certainly, mice have a genome that's similar.  They're great for at least developing and obtaining specific early functional information. We need to understand better our genetically modified mice  and our transhumanized mice to see how these are going to be useful for preclinical screening and safety and then moving them into specific human studies.  But I don't think that animal models in today's realm are flawed. I think that there's opportunity for many of us to consider also moving from the bedside to the bench.  We don't conduct multi-center animal studies. But perhaps that would actually help us better understand environmental differences, husbandry differences,  understand differences with a lot of the different environmental cues so that we could actually have reproducible data across different animal centers that then will help inform us  as we move forward. So how do we get there? So I say design animal studies with a clinical trial in mind. We should be partnering with our friends that  are performing clinical trials. With sample size calculations, a co-clinical trial design that mimics the bedside. Have similar primary and secondary outcome variables,  which most of us don't do with our animal studies, is think prospectively to those human trials. What are those secondary outcomes going to look like?  And then multi-center animal studies. We should reflect these correct populations in the animal study to precisely define the human population  to an intervention and go back to the bench when the clinical trial fails after the animal model had success to understand the why. And then finally, develop consensus statements  on recommitted translational study designs. And with that, I'll just take questions.  Anyone feeling philosophical? It was a really great talk. I have one question. As Dr. Wong presented earlier, we're really focused on heterogeneity and the human populations.  And I'm curious if you have ideas of how we can better model heterogeneity in our mouse models. So I think some of that heterogeneity  would actually come out if we did multi-center animal studies. I think that some of the importance of some of the transgenic mouse strains, they're really helpful for honing  in on mechanistic or causal pathways. We then should take those areas that we've defined and attempted to modulate in the actual transgenic model and then put it  into an inbred strain that doesn't have that same mutation, and then conduct it over maybe a couple different sites, and then mimic some of the real-time changes  that we do at the bedside that we don't reflect in our animal studies. That may provide, at least in some ways, although the care may be more homogenous,  we may be able to minimize some of the heterogeneity that we're seeing and then be able to translate that. I think that's the only way to do that.  The other thing is design better human studies. I know that sounds crazy, but I mean, if you look at the last 10 years, I don't mean to be, we seemingly have  in our important journals, the majority of them have a primary outcome variable that's never significant, but it always seems on subgroup analysis where we're making our findings.  Why? There's a breakdown for some reason, whereas our animal studies, our primary outcome variables seemingly have a difference. Either we're not using prospectively  a good statistical design for our animal studies and then just having to do the same thing we do in humans, or there's something with the design flaw  that we're having an issue with.  Hi, Sam Poliak, School of Pharmacy, University of Pittsburgh. So I'm not feeling philosophical, I'm feeling pharmacological. My question is really one of drug concentration  and effect in biomarker interaction. I think one of the reasons a lot of the trials fail is because we didn't pick the right dose.  So I wanted to comment on, you get a comment from you about really determining dose concentration relationships, target engagement with biomarkers,  so that when we actually do go humans, we know what we're targeting, we actually hit the target when we translate. Yeah, no, I think that's a great question.  I'm not sure necessarily that the mouse model is probably the best to do that. I think Dan could probably give you some great information  about a monkey model that he currently has at the NIH. I think that probably would be a relevant model to really hone in on some of these,  or a large animal model that may be more relevant. I don't have an answer for you in relationship to mouse, unfortunately.  Add that with repurposing, it becomes particularly difficult because we think we know the safety and we skip all of those, and then we go right from the rat to the rat.  That's right.  Just one somewhat philosophical question. I wondered if you had thoughts or comments to junior investigators embarking on a career in veering research now that the reaction  of some of this data by some funding agencies like NIGMS is to actually deprioritize applications for funding that involve veering models. I think that I'll use this as a moment  to just touch upon the development of a white paper, let's say, from SCM. I think that this is a unique opportunity where we can demonstrate the benefit of using an animal model,  but we need to partner specifically with our clinical trialists. And I think that even though that's attractive to, say, a lot of the different funders,  I know you're talking about the same funder that I have currently who's not going to fund animal studies,  I think it's a time for us to say that we're gonna actually improve that translational effect of design. This is where I think we've,  maybe the area that we can say this is vitally important because with all due respect, if you look at sepsis, and I know they're talking about guidelines,  uniformly, one in five in current practice, if I give antibiotics as soon as I know sepsis has occurred, I may still have a 20% mortality. There needs to be better than that.  That's not going to be in my purview based off of just ongoing supportive care. We're not going to be able to refine these unless we actually pattern our clinical trials off  of some of our successful animal studies. And then furthermore, do the opposite and develop better approaches. I think the work with omics that Hector  and some of the other speakers have mentioned is exactly what we need to integrate. We need to understand how to use all this really exciting  science, but it has to be with better study designs. And the only way it's going to happen is the translational aspects. In the 90s, we were running a septic pig  at Michael Reese in Chicago for 12 days at a time. At the end of the 12 days, the fellow running the pig smelled worse than the pig did and appreciated the nurses  more than anybody else in the program. But it was a good model in terms of being an 80 kilo animal on a ventilator, on peep. The Animal Use Committee loved it  because it was on opiates, and opiates are immunosuppressive. And it was on antibiotics, and the cytokine profile was dramatically moderated from what  we saw in the rat studies and the in vitro work. And I still think the idea of a couple of places in the United States running septic pigs that other people could participate  in, it cost about $6,000 a pig. It's not a cheap model, but if I get to heaven and God's a mouse, I'm in big trouble. But that's why I designed, why we had, we wrote it up.  It's in critical care medicine a long time ago. But I do think there's value to that model that we should consider doing multi-institutional. Thank you.  Good food for thought for everybody. And by the way, Andrew, Dr. Sarah Densmore from NIGMS is here, and she would more than welcome to speak with you about that.  Our next speaker is Daniel Chertow from the Clinical Center at the NIH, and I'm really looking forward to this. This should be good. Please. I'm going to reset my clock to make sure  that I'm respectful of my time here. So we're going to shift gears. We're going to talk about next generation diagnostics with CRISPR.  I'm going to serve as a translator and a cheerleader today a little bit, outside of my typical role. So it's already been alluded to.  I do basic and translational studies on severe emerging pathogens using large animal models. But that's not what I'm going to talk about today. I'm going to talk about CRISPR.  So let's see here. All right, I have no disclosures because I work for the federal government. Learning objectives. So I'm going to summarize the diversity of CRISPR-Cas  systems in prokaryotes, bacteria, and archaea, and highlight applications of CRISPR-Cas12 and 13 systems  that folks may have never heard of for their applications for rapid and acute diagnostics. So context. Why does it matter? Why do we need rapid diagnostics?  Well, I didn't plan this. But as of early December 2019, there was a cluster of pneumonia cases that were reported to the WHO. They had a common exposure to local seafood and animal  market. And a novel coronavirus was isolated and sequenced within weeks, which has got genetic similarity mostly to bats. And that's an electron micrograph  that was published in the New England Journal of this virus. Within a relatively short interval, there was widespread local, then regional, and now international spread.  The case count was over 46,000 on February 14. We're closer to 60-some thousand now. And there's concern that there's going  to be widespread disease in countries throughout the world. So right now, there's a case definition, which is a syndromic case definition, which  includes a respiratory illness and an epidemiologic risk factor. And that risk factor includes travel to an affected region  within 14 days and or contact with a suspected or confirmed case. If that person walked into your health care facility, these are the things that you would be doing with a focus  on infection control. You'd be isolating cases. You'd be wearing the appropriate protective equipment. And then you'd be contacting your local and state public  officials, not just for the public health investigation, but to facilitate you getting a confirmatory diagnosis. And right now, those samples are being  sent to the Centers for Disease Control so you can get a confirmatory diagnosis. And ultimately, the goal is to push out that testing to state health departments  throughout the country. You need a respiratory specimen from the upper or lower airway, and they use an RT-PCR-based assay to make a diagnosis. But the delay may be days or longer.  So it brings to mind, wouldn't it be nice in both our health care facilities, but also in remote settings where many of these infectious diseases ultimately emerge,  to be able to take an unknown and use various clinical specimen types, whether it be urine, blood, or stool, and implement a rapid and accurate diagnostic, ideally  that's affordable at the point of care, so that you can implement effective infection control and optimal clinical care. So today, I'm going to talk about,  this is my cheerleading and sort of translator role. This is not my research, but I am interested in it because of its applications for emerging infections.  So I made a commentary on this that was published in Science. And I'm going to introduce some concepts about these emerging techniques that  include these acronyms, HUDSON, SHERLOCK, and DETECTOR, which I'll talk to in a minute. But before we go there, unfortunately,  for better or for worse, we need to talk about some basic cell biology and molecular biology as it relates to bacteria. So CRISPR-Cas systems, we're going  to talk about their biological function. These provide bacteria-acquired immunity. I'm going to say that again. CRISPR-Cas systems provide bacteria-acquired immunity.  So all that we know about CRISPR are borrowed from our understanding of their biologic functions in bacteria and archaea. So what do they do?  They protect bacteria against invading genetic elements. And what are those? Those are phages. Phages are viruses that infect bacteria.  And plasmids, bacteria's shared genetic components, whether it's resistant elements or others, by these mobile genetic units. So CRISPR stands for clustered regular interspersed  short palindromic repeats. What are they? These are repeating sequences that are present in the bacterial genome, which are shown by the little blue bars there.  Those are repeating sequences, same genetic sequence. And interspersed with them are these spacer sequences. And so those spacer sequences I'm  going to talk to them about a little bit in a bit. But together, the repeating sequences and those spacer sequences make up the CRISPR array within a bacteria.  Adjacent to that is this Cas operon. So the Cas operon is a genetic component of the bacteria that codes for these Cas proteins, or CRISPR-associated proteins.  That's what Cas is, CRISPR-associated, that ultimately make up the CRISPR-Cas systems of a bacteria. So I said that this is a required immunity in bacteria.  So how does a bacteria gain immunity against an invading genetic element? The answer is it occurs in three stages. And those stages are adaptation, maturation, and interference.  We're going to go through those quickly. So this is a picture of a bacteria that is being invaded by a bacteriophage, or a virus. And so on the surface of the bacteria,  there's a virus that's injecting its nucleic acid or its genetic material into a bacteria. The bacteria recognizes that genetic sequence  as foreign by something called the protospacer or PAM, protospacer-adjacent motif, which is present on the invading virus's genetic element there. That PAM is recognized.  And then proteins of the bacteria, these Cas proteins, go ahead and cleave a little bit of that genetic sequence out of the invader and integrate it  into its own chromosomal DNA, into its own bacterial DNA. And that little piece of genetic sequence is now a memory element.  In case that bacteria were to see that invading virus again, it's got a little tag. It remembers it. It's got a piece of it that can then say, oh, I've gotcha.  So that gotcha part is the maturation and interference. So again, here we are. This is now a virus that's coming. It's infecting a bacteria again.  And it's injecting its nucleic acids. Well, the nucleic acids are coming in. But the bacteria's CRISPR array is now transcribed. And it transcribes these pre-CRISPR components here,  which includes the nucleic acids and the associated proteins. And those are then cleaved into the individual parts. And there's one of those parts that  has sequence specificity for that prior invading element that it recognized before. That sequence specificity recognize the invading elements and the associated Cas proteins  then go to work and cleave it and destroy that invading element. And so that's what takes place during maturation and interference process.  And then you get chopped up nucleic acids. So what I want to emphasize, which is really difficult to understand as a clinician and even somebody that hears all about CRISPR,  is that there's an enormous diversity in CRISPR-Cas systems that we need to increasingly begin to appreciate so that we can use their molecular tools  and apply them to things that are relevant for us at the bedside. And so that diversity, the way that the nomenclature works right now, there are two classes of CRISPR-Cas systems.  Each class has multiple types. And we're going to start out by talking about these class 1 systems. Hang with me for a second here. So this is the class 1 system.  So the class 1 systems, what they do is that they have, again, a piece of multiple Cas proteins that come together in this complex. And that's sort of shown there.  And that complex of proteins is associated with the CRISPR RNA. And that makes up, in class 1 systems, something called the cascade.  So cascade goes and recognizes a genetic element. It binds down to the DNA of an invading genetic element, which you can see there. And then once it's bound down, it  calls in this Cas 3 protein. And the Cas 3 protein, which is part of the bacteria, goes ahead and cleaves that little piece of DNA there. So you can see the scissor.  That's where there's a cut in the single strand of DNA there. Now, type 3 systems, they're similar to type 1 systems, although there's a nuance here.  Again, they have this complex of Cas proteins that are associated with the CRISPR RNA. So it's a CRISPR-like, excuse me, a cascade-like complex of proteins.  Now, that interference complex of these type 3 systems, instead of recognizing DNA, they recognize transcribed RNA. So it binds down to transcribed RNA.  And then if you look at the little scissors there, there's a cut, not just in the transcribed RNA, but in the double-stranded component of the DNA.  So there's cuts in the RNA and the DNA. The byproducts of that cleavage result in these little cyclic adenylates that bind to CSM6, which is also a protein that's  present within bacteria. And that complex of these cyclic adenylates and CSM6 result in nonspecific cleavage of RNA. So you have a specific recognition  of a genetic element that results in RNA and DNA cleavage. When that specific recognition takes place, other RNA that's in that milieu gets nonspecifically cleaved and degraded.  That's a really important piece that I'm going to come back to that's ultimately, you'll see why it's relevant in the upcoming parts of my talk. OK, so we're going to switch quickly  to talking about class 2 CRISPR-Cas6 systems. Again, these include types 2, 5, and 6. Again, going into the biology of this, the one that we all are most familiar with  are the interference complex of CRISPR-Cas9. And we know about this because of its role in gene editing. That's going to sort of taking off.  But the interference complex in CRISPR-Cas9, the way that that works is that it recognizes DNA and that it results in clean cuts where there's blunt ends in the double-stranded DNA.  So you can see where those scissors are, where there's clean cuts in the blunt ends, which is very amenable to then going in and making  very specific cuts where you can then edit the genome by having insertions of a target gene of interest. So that's a type 2 CRISPR-Cas9 system. And you can see those cuts there.  Cas12 is similar in that the interference complex also recognizes DNA. But instead of making those really nice, clean, blunt-ended cuts, you can see where the scissors are.  They're staggered cuts. So there's staggered cuts in the DNA that take place with Cas12 systems. And there's another feature of Cas12 systems  that are really important that I don't highlight on this slide, but I will later, when there's sequence-specific cleavage  of the DNA, that non-target DNA in that milieu is also degraded. So there's nonspecific cleavage that's taking place. And that's one of the pieces that's  being leveraged for these diagnostics that I'll discuss. So Cas13, these are example of type 6 systems. Cas13, again, where instead of DNA being the target, RNA is now the target.  So the interference system recognizes RNA. RNA is cleaved in a sequence-specific fashion. But when it is, when there's recognition  and sequence-specific cleavage, other RNA that doesn't have sequence homology with the CRISPR sequence, there's nonspecific degradation of RNA in the reaction mix.  And that's another feature, that nonspecific cleavage of RNA that's leveraged for these diagnostics that we're going to discuss. So now you know all that you need to know.  That's all of CRISPR-Cas biology in five minutes or less. And how is it being leveraged for use in rapid diagnostics? So here we are. We're back at the bedside, OK?  So this is the concept here that has been reflected. So bottom line is there's three steps to this. So you have your sample from your person that presents to you.  And then the first step is to prepare your sample by releasing and protecting nucleic acids within the sample. So you want to get those nucleic acids available for your assay.  Then you amplify those nucleic acids. And those can be DNA or RNA so that you can have increased amplification for detection of them. And the method that's being used there  is RPA, or recombinase polymerase amplification. RPA is a unique thing. It's called, it allows for isothermal amplification. When we think of PCR-based assays,  that requires this thermocycling on this machine that it's at this temperature and that temperature for annealing and breakdown. RPA, all that amplification can be  done at a single temperature. So that's an isothermal step that doesn't require that thermocycling to amplify your nucleic acid. And then the final part of this is  you need to accurately detect the target of the amplified signal. And the ways that that's being done that we're going to go through this are Hudson and Sherlock and detector.  So we're going to discuss those briefly.  Double checking my time. OK, I'm still OK. Hudson, heating unextracted diagnostic samples to obliterate. This is just your sample prep. So you have your sample.  And this is using heat and chemical inactivation to release and protect your nucleic acids. And this takes about a total 25 minutes or so to complete this step.  This is a basic heat and chemical inactivation. Now you have your nucleic acids, which are on the left side of that slide. You do a reverse transcript, with or without  a reverse transcriptase step, depending upon if you've got RNA or DNA. You do your RPA for isothermal amplification. You've got amplification of your nucleic acids.  And then those nucleic acids are reverse transcribed back to RNA. So you want your target here to be RNA. So they use something called a T7 transcription  to give you your RNA nucleic acid. And then this is now using a Cas13 detection system in your diagnostic sample. Remember, I told you Cas13 recognizes RNA  in a sequence-specific fashion. And so the other part of this reaction mix, which are shown there, is basically RNA, which is compounded to a fluorescent reporter.  So now when Cas13 recognizes the sequence of interest, a virus or whatever you're looking for, there's cleavage of that RNA. And as I mentioned, there's nonspecific cleavage  that also takes place. And that nonspecific cleavage of that RNA in your reaction mix allows for the fluorescent signal, which is the cool part.  And then you get that cleavage signal, which is right there. And so these authors, Merival, that all use this as a proof of concept to say, hey,  can we detect Zika virus in clinical specimens? And they compared it to a validated PCR assay, which is this Altona assay. And they found that they had good concordance detecting  all the positives and all the negatives using their CRISPR-based assay. And then they said, well, we can do it with extracted nucleic acids.  But can we do it with a relevant sample type? And the sample type in this case is saliva. So Zika is present in multiple body fluids, including saliva.  And there are two groups here. And the x-axis there is a dilution curve at different concentrations. And they're comparing presence of nucleic acid in PBS versus saliva.  And you can see at 0.9 atom molar concentration, they were able to detect Zika nucleic acid. So for the audience, atom molar, for the unoriented group,  is 10 to the minus 18th molar. So you're starting to get down there and obviously has a sensitivity of a PCR-based assay. So you think about our rapid diagnostics  where we're looking for antigens and we're looking for antibodies using these that are not sensitive. This is a molecular test that has a sensitivity of like a PCR.  So detector, what's detector? DNA endonuclease targeted CRISPR trans reporter. So what this does is, in this case here, your target of interest is DNA. And this uses a Cas12 system.  So I mentioned Cas12 recognizes DNA. So you have your DNA all the way to the left there. You have an RPA step, which amplifies it in a very short interval.  And then you use a Cas12 detection system. And again, there's this nonspecific cleavage when you have targeted cleavage in a sequence-specific fashion.  And those nonspecific DNA components that are combined with the fluorescent reporter, they ultimately are cleaved when the reaction takes place. And you have a fluorescent reporter  that gives you your signal for a positive test. So these authors used detector. And they, again, compared a PCR for detection of two for HPV16 and 18.  And what they found is that detector performed, was concordant with the PCR, validated PCR test in all but two of their sample types.  So not perfect, but awfully close getting there. Whoops. An advance on that was this idea of, well, can we multiplex? So Gutenberg et al used a combination  of Cas12 and Cas13 systems to multiplex and be able to detect four different targets in one reaction mix. The complexity of this slide, long story short,  is that they were quite successful. And that's sort of shown by the heat map on the right there. And then the next step was, OK, well, you have this new cool molecular technology.  But how do we make it rapid and use it at the bedside? And the idea was, can we get this onto a lateral flow readout? Think a pregnancy test strip.  Think a urinalysis test strip to get an answer. And the long story short is they used Cas13 to detect Zika. And they used the same concept as reporter cleavage.  And on a test strip, were able to successfully detect the presence of Zika using this lateral flow readout. Now, this particular reaction that I showed you,  which was published on, took about one hour. So ideally, it would take shorter. And so I had mentioned in one of the introductory slides the CSM6.  So they leveraged the biology of CSM6 to further amplify not the nucleic acid, but the detection signal. So CSM6 also was a protein that allows  for nonspecific cleavage of RNA in the reaction mix. And so in this assay, they combined Cas13 and CSM6. And they implemented on a lateral flow device  with the idea that if they can get a better amplification signal, it would take less long for the reaction to read out. And they accomplished this now in 20 minutes  with a readout on a paper test strip. So in summary, this is not quite ready for prime time. But this is happening in the background.  And you can imagine how this might be clinically useful, both for us in the ICU in a modern setting, but also in clinics and health care settings across the world.  I hope I've emphasized that CRISPR-Cas systems are quite diverse, that Cas12 and Cas13 applications have shared some promise for rapid, accurate, and simple ID diagnostics.  And of course, like other emerging technologies, the ultimate impact will depend upon careful validation in the clinical setting. And ideally, in a global setting,  these will have to be low-cost tests. Now, the reagents, I should say, are quite low cost. But how they get, and they don't require high-tech  machinery, et cetera, et cetera, but how that's scaled up and ultimately provided in low-resource settings is yet to be determined. Thank you for your attention.  So to me, an obvious application is detecting bacterial DNA in blood specimens. I'm not that smart, so I'm sure somebody's thinking about that. What would be the barriers to that?  Yeah, so that's a good question. I mean, right now, the greatest barriers for detection, because this is the idea of, can we  use existing molecular tests to diagnose bacterial sepsis? And right now, for using unbiased next-generation sequencing, the primary limitation of that  is there's a lot more host than pathogen. So if you have an unbiased approach, then they're having a hard time finding the pathogen. There are new emerging technologies  that are becoming available, some of them commercially, others. There's one out of Columbia that you may have heard of called BackCapSeq, which is a targeted biasing approach which  has amplification as primers for the whole host of known bacterial pathogens. That takes time and money, et cetera, et cetera. So can CRISPR be applied?  CRISPR theoretically could be applied. The issue is that you have to know what you're going after. And so these assays, you can see there's one pathogen,  or they've got multiplexed up to four pathogens for these point-of-care rapid diagnostics. Now, I happen to know that there are also  efforts to doing more of that sort of BackCapSeq approach. But instead of using PCR or sequencing, they're attempting to use CRISPR,  where they'll go after the whole range of human pathogens. Conceptually, it's totally feasible. But that would not be done at the point-of-care.  That would be done with the machine in the laboratory. It would have an interval to take some time. And then the issue there is you're  going to have high sensitivity, but how specific are you? So if you look at those assays like BackCapSeq, you'll pick up all sorts of things. And it's remarkable.  And actually, the truth is that some of them are real. Some of them are what we have circulating in our bloodstream as a function of when you floss your teeth,  you get low-level bacteremia. Well, which of them are causing problems? And then some of them are environmental contaminants from the host, from how you collected the sample.  And some of them are environmental contaminants from the lab where the assay was performed. So that gets back to, we're going to need to sit around and think a little bit about how  we interpret these assays before we act on them clinically. So redefine what it means to do that. We're going to have to understand.  We're going to have to get better understanding of that.  I mean, you can see how this could be appealing in so many different ways. And as long as, you know, and we know that there are resistance elements, and if you  can identify the one, the two, or the five resistant elements that confer pan-resistance, you can create an assay that could presumably pick that up in a relatively quick way. Okay.  Moving along, our last speaker is Dr. Bala Venkatesh, who's from Princess Alexandra Hospital in Australia. We were talking about clinical trials.  For those of you who don't know, the adrenal trial, which I believe is the largest trial to date in corticosteroids in septic shock, was led by our next speaker. So welcome.  Thank you, Hector. And it's a great honor to be here. I'm a clinician, and I'm a clinician researcher.  So I've been sort of spellbound by the fantastic level of detail in the, you know, on basic science in the previous presentations.  So my plan today is to give you a clinical overview to all the clinical trials we've had in trying to modulate the immune response in critically ill patients with sepsis.  And so give you a brief overview of the immune response, the concepts of SIRS and CARS, the  systemic inflammatory response on the counter-regulatory response, and the early trials of immunomodulation  in sepsis, what we learned from that, the shift in thinking now from trying to, in fact, going away from the blocking the early phase to now to immunosuppression, which is the  main problem in sepsis, and we need to do immunostimulation. That's where the focus is now.  And what the challenges are in doing clinical trials in the ICU, particularly in this cohort.  So these are my disclosures, and I have no conflict of interest in relation to this talk. Way back in the 50s was the first, I guess, insight, the discovery that the manifestations  of an infection do not necessarily relate to the pathogen per se, but in fact it's the body's response to it.  And that's the discovery of the endogenous pathogen, which we now know is interleukin-1. And then a very elegant study in the 80s where they demonstrated that if you gave endotoxin  to genetically resistant mice, the mice did not die. But if you then transfused those mice with bone marrow cells from genetically sensitive  mice and then gave them endotoxin, the mice died, suggesting that the lethality of sepsis and endotoxemia is in fact the host response primarily and not the endotoxin per se.  So the concept of sepsis as being an immunological problem, the multi-organ failure, all the  manifestations were driven by the immune response of the body, led to a clinical trial with steroids as being the first anti-inflammatory agent.  And so we had Bill Schumer's work, famous paper, back in 1976, randomized controlled trial with placebo, methylprednisolone, and dexamethasone.  And he was able to demonstrate that patients who were assigned to the steroid arms had far better clinical outcomes as compared to those who were randomized to placebo.  Over the years, we learned about the immunological response, the cytokines, the proximal cytokines,  the involvement of the coagulation cascade and the immunological processes, the recognition  of cellular mechanisms, the endogenous mediators, and more importantly, and in fact more recently,  a significant genetic contribution to the immune response, which in fact contributes to the heterogeneity of sepsis to a large extent.  A recent publication in Nature where they identified there were more than 3,000 genes involved in response to a single endotoxin injection, just to outline the complexity  of the genetic response and its contribution to the inflammatory process. So as we understood the immunological factors involved in sepsis, trials began to appear.  These are the two cardinal trials back in the 80s by Elizabeth Ziegler, where they looked  at an E. coli human antiserum and the anti-endotoxin antibodies, and in both those trials there  were improvements in patient's outcome, but the major confounding factor was the patient selection where they were not standardized.  This was before the SERS criteria for sepsis were defined. And we then had anti-TNF studies, which came through, and I've just shown you three here.  We've had many TNF, anti-TNF studies, and again they were not associated with any significant improvement in patient outcomes.  We then had anti-interleukin-1 receptor antagonist trials, a randomized control trial, once again with no improvement in patient benefits.  And then we had trials, the activated protein C came in, thinking that if we can somehow link the immunological and the coagulation cascade together and try and identify a mediator  which would target that. And so the initial PROVA study was reported in early 2000s, and although there was an  improvement in patient outcomes, the improvement was confined only to the sicker cohort of critically ill patients.  So the ADDRA study followed that, and they examined the role of activated protein C in a cohort of patients who had a lower risk of death and could not identify any beneficial  improvements. And subsequently, the PROVA shock study was done, they took a heterogeneous cohort of  critically ill patients with septic shock and a high risk of death, and again, the trial was negative.  There were no differences in outcomes between placebo and those assigned to activated protein C.  We've had more than 100 trials over the years, phase II and phase III trials, targeting a whole range of mediators, and I'm not going to go through the entire list over there,  but as you can see, there has not been a single one which has stood out in terms of improving  patient-centered outcomes. And again, this is another very nice graph which I've taken from John Marshall's paper.  If you look at this, there are about 50, 60 trials listed there, and in the aggregate,  they all seem to suggest a trend towards improved patient outcomes, but they're all confounded by selection criteria, they're confounded by baseline imbalances between the placebo  and the treatment groups, and when you put them all together, the statistical analysis is also complicated, and that's one of the reasons why there's not been any further work  in these particular areas. It just needs a large, massive trial if you want to go back and look at these in more detail.  Then come the 2000s, then we had a new drug. People started to look at statins. It's just  not an anti-cholesterol agent. In fact, it's got many pleiotropic effects against complement and endothelial activation. It blocks the recognition of the pathogen by the receptors  and blocks the cytokine responses, and people felt that it might, in fact, be beneficial in sepsis.  And the interest in that was driven by this study which was reported by my colleague who I work with, Peter Kruger, and this was a retrospective cohort study where they looked  at patients with bacteremia, and then they divided them into three groups, patients who were on statins, and when they were hospitalized, statins were continued, then patients who  were on statins, when they were hospitalized, statins were discontinued, and patients who were never on statins.  The best outcome was in patients who were hospitalized, who were on statins, hospitalized, and statins were continued. And the worst outcome was patients who were on statins,  but in whom statins were discontinued at the time of hospitalization. So based on this observational study, we went on to do a randomized controlled trial, which  was in people coming into hospital with presumed infection, we randomized them to either continue  the statins or to stop the statins. And in that randomized trial, interestingly, we did not find any differences between the two groups with regards to progression to severe sepsis  or in terms of organ dysfunction or mortality. So that was the first RCT that we did. Then we, the previous cohort was patients coming  into the emergency department with infection, but who did not, a large proportion did not come to intensive care.  The next study we did was in critically ill patients with sepsis. We did a randomized controlled trial of statins versus placebo. And in these patients, the primary endpoint  was interleukin 6 concentrations, and we had a whole range of secondary outcomes, which were all important patient-centered outcomes.  So the primary endpoint, again, there were no significant differences between placebo and statins. And when you look at the secondary outcomes with respect to mortality, length  of stay, again, there were no differences between the two cohorts. And other trials on statins from the, this was from the Irish group, and there's one  from the ARDS clinical trial network. Again, two negative trials using statins in patients  with ARDS and acute lung injury. So once again, the statins have sort of, people have lost  interest in statins as an anti-sepsis agent. Well, then we've had a number of trials looking at cytokine removal in septic shock. Again,  none of them have borne fruit, disappointingly. And before I move to the more recent drugs,  I just want to revisit the steroid story again. I told you briefly about Schumer's paper. The Schumer's paper was followed then by two large random mass control trials in the 80s,  one by Charlie Sprung, and the other one was by Roger Bone. And they both used 30 milligrams per kilo per day of methylprednisolone, which is equivalent to about 10,000 milligrams a  day of hydrocortisone. And in both those trials, the patients were randomized to receive methylpred or placebo,  and in both those trials, there was evidence of either no benefit or harm to patients assigned  to the steroid group. So high-dose steroids were clearly dead in the water. Interestingly, of course, in those trials, they did not measure any mediator concentrations. So we do not  know if there was any immunological effect from the high-dose steroids, but from a clinical perspective, there was certainly no benefit.  Then in the early 2000s, Didier Kaye did a very elegant study because there was an interest  in low-dose hydrocortisone. And so he performed a randomized crossover trial in patients of  septic shock. Patients were randomized to either placebo or low-dose hydrocortisone, and halfway through the trial, they crossed over. And basically, he was able to show that  in patients assigned to low-dose corticosteroids, the IL-6 and IL-8, the pro-inflammatory cytokines  and markers of neutrophil activation were far lower in the steroid group as compared to placebo.  So as you all know, there have been a number of low-dose steroid trials, and the most recent  ones were the adrenal trial, which I had the privilege to lead, and the other one was the APROX trial, which was published from France.  In both these trials, well, in our trial, we could not show mortality difference between the two arms, which the French group was able to demonstrate, but there were consistent  improvements and important patient-centered secondary outcomes in both trials. Stroke reversal, mechanical ventilation, early discharge from intensive care, and  in our trial, we showed a reduction in blood transfusion. Again, what the mechanism is, we do not know. We did not measure biomarker concentrations  in this trial as part of the original report, and certainly did not measure any cytokine concentrations in either of these trials.  We do know that there are established anti-inflammatory effects for hydrocortisone. The APROX trial  used fluterocortisone in addition. Interestingly, fluterocortisone has got some pro-inflammatory  effects, so where that sort of links in with the whole immunological story is something that needs to be teased out.  And of course, the recent entrant is vitamin C. There's been a lot of interest in vitamin C in sepsis, initially driven by Paul Maric's paper back in 2015, and on the grounds that  it's an antioxidant and anti-inflammatory agent. We've had two trials, two prospective randomized controlled trials. One of the citrus ALI trial,  which compared vitamin C versus placebo in patients with sepsis, and the primary outcome was progression to multi-organ failure or worsening organ dysfunction.  And again, I've shown you the graphs of the C-reactive protein and thrombomodulin concentrations  between the two arms. There were no significant differences between the two arms with respect  to that. And importantly, there was no differences in the primary outcome between the two groups.  And the second trial was the Vitamins trial, which was reported in Belfast last month. Again, this was a trial of patients with septic shock, randomized to receive vitamin C versus  placebo, and the primary outcome was shock reversal. Again, there were no differences between the two groups. They did not measure any biomarker concentrations, but it sort  of begs the question, if the inflammatory process is driving a lot of this, then these agents should be doing the work, and we clearly are not seeing that.  And again, we went back and looked at our own data from adrenal. We measured endogenous ascorbic acid concentrations, and we divided them into tertiles. And again, there were  no differences between the highest and the lowest tertile with respect to outcome or primary outcome, which is day 90 mortality. So the thinking, which was initially to block  the early phase of the immunological response, so the pro-inflammatory state, that's now  been largely abandoned. And the thinking now is that it's the late phase of immunosuppression,  which is responsible for the poor outcomes in sepsis. And the evidence for immunosuppression comes from a number of different reports. One is  the finding that there's lymphopenia in patients with sepsis, and that's work from Richard Hotchkiss. And this is from apoptosis of the immune effector cells. There's reduction in  cytokine production from T lymphocytes. And interestingly, there is a very interesting paper in patients who have enhanced IL-10 production, which is an anti-inflammatory  cytokine. If they have a higher ratio of IL-10 to TNF-alpha, they have a higher mortality,  again suggesting that the immune suppression could be a driver. And of course, the clinical  observation that patients who come into hospital with sepsis go on to develop a higher incidence  of hospital-acquired infections, suggesting that they are prone to immunosuppression. And again, from our own genomic data sets from adrenal, we've identified SNPs with that  particular gene, NINJ2, which is associated with day 90 mortality. And the clinical annotation of that gene is associated with myeloid and lymphocyte count production.  So there are trials of immunostimulatory agents, which have been reported. We've got interferon  gamma, GCSF, GMCSF, and IL-7. And of interest, IL-7 is the one which is now the, I guess there's a lot of interest in this particular drug, because it improves lymphocyte counts,  because persistent lymphopenia in sepsis is associated with mortality and adverse outcomes. So there are a number of IL-7 trials, which it's been shown to be of benefit in oncology.  In particular, in septic shock, the most recent one is the one by Bruno Francois, where they  compared two doses of IL-7 and placebo. It's a phase II trial. And they were able to demonstrate  an improvement in CD4 and CD8 T cell counts as compared to placebo. At both doses, the lymphocyte counts are much higher as compared to the placebo arm.  So based on that, we now have an international collaboration. We are working with Richard  Hotchkiss. We are pulling three large databases together to understand the impact of lymphopenia  and outcomes in critically ill patients. Lymphopenia is, in fact, if you go back to your ICU population  and look at their differential white cell count, lymphopenia is very common. And the question really is, does lymphopenia in sepsis, is it any different from lymphopenia in trauma?  So we want to answer that question first. So we're going to look at this large database  to see if there's any differences in outcomes between the septic and the non-septic population in preparation for a randomized controlled trial in the future.  This is bone marrow stem cells, immunotherapy for septic shock, which is being trialed in Canada. This is a phase one trial. And we had a, Hector mentioned the SRS1 signatures  before in his presentation. So the inpatients with septic shock, the Vanish, the British group, when they did the Vanish randomized controlled trial, from that they were able  to identify a cohort of patients who had the SRS1 and 2 signatures. So SRS1 is an immunosuppressive  signature, and SRS2 is an immunocompetent signature. And they were able to, the SRS2 patients who were immunocompetent, when they were assigned to corticosteroids, they had  a worse outcome. Now this is based on a post hoc analysis, and therefore they are planning an RCT based on a rapid detection of your SRS endotype, and then going on to randomizing  them to steroids versus placebo. So that's being planned, and I guess it's probably the first personalized immunomodulatory treatment in the critical care scene.  There are a number of challenges when we do these trials, and I guess the first thing, and a number of questions today have been on this, on heterogeneity. The first thing  is the study population. How do you pick them? We have traditionally used the SRS criteria, and now we now have the sepsis 3 criteria, and do these physiological criteria, do they  actually reflect what's going on in the plasma? And if you look at this, in fact they don't always do that. This is a very interesting paper from a subgroup analysis of the TNF  trial, where they looked at patients who met the SRS criteria physiologically, but if you look at their median interleukin 6 and interleukin 8 concentrations between the  two groups, they were vastly different. So a physiological similarity does not imply a biochemical homogeneity, and that's one of the problems we had.  The second thing is the timing of intervention, whether you get in on day 1 of septic shock  or you wait for day 3 when the immune suppression sets in. These are questions for which we  don't have reliable answers. We don't have good quality data on dose response relationships and how long we should give these treatments. And again, a question was asked before, the  pathogen heterogeneity, and again, this is a subgroup analysis of all the anti-TNF trials, and it's interesting. So in mouse models, which was LPS model, if you gave it anti-TNF,  the mice survived. But if it's a CLP model, there was worse outcomes in the mice receiving anti-TNF. And again, there were beneficial effects in E. coli and Staph aureus infections,  but deleterious effects in strep pneumo and cannabinoid sepsis. So the pathogen is also a significant driver of the heterogeneity and response to immunomodulatory agents.  And I told you about the genomic complexity, and I'll leave Hector to tell us about this  graph. And there are a number of other factors, the timing of initial antibiotics, the patient  co-morbidities, and the timing of intervention and the treatment practices. Do you give antibiotics  for three days, five days, seven days? We haven't been consistent with that. The duration of treatment, the dosing, and I think these are important considerations which contribute  to patient heterogeneity. So to conclude, I think we have certainly, we understand the immune process much better  compared to what we knew, say, 10, 20 years ago. And trials targeting the pro-inflammatory  pathways have not yielded fruit with respect to patient-centered outcomes. Low-dose corticosteroids,  a known anti-inflammatory drug, has been consistently shown to improve patient outcomes, although  the precise mechanism by which they do that has not been elucidated. There's now a focus on the delayed phase and attempts to target immunostimulant treatments, and one of which  is IL-7, and patient heterogeneity still poses a significant challenge. Thank you for your attention.  Thanks, Bala. That's a great overview, and I guess this is more of a comment or opinion rather than a question, but obviously you can respond to it. I'm really, I think that  the movement towards immunostimulatory therapy makes biological sense. I'm worried, though, that the movement seems to be using lymphopenia as a way to identify the patients who are  going to benefit. And I understand that that's pragmatic, right? We can get a CBC and measure lymphopenia easily anywhere in the world, and you can enroll a patient based on that.  But immune suppression in our patients is so much more complicated than what's reflected in lymphopenia, and I worry that what we're going to call a precision trial is going to  be more of an approximation trial, and we're going to be back to the same place. I really, really worry about, I can see that's gaining momentum, that, well, we're going to find  patients with lymphopenia, and those are the ones we're going to select for IL-7, and I really worry about that.  A great comment, and I, it may well end up being like the other trials that we've done in the past. The, I think the, it's, I mean, it is for pragmatic, it's not just for pragmatic  reasons. There is a lot of data showing association between persistent lymphopenia and poor outcomes.  If we fix that, whether it'll change things, we don't know that. And I think to a large extent, the large retrospective database will inform us in planning for a definitive  trial, and I guess that's as much as I can tell you at the moment. And there is now a phase two trial, which is, which will provide some pilot data that's being run in France  and that'll inform us as well.  I, this is a great presentation. I think we get into trouble because we limit our ideas to the discipline in which things were originally studied. If endothelin had been described  by an immunologist, it would be interleukin-34. It has that much effect on a macrophage, but it wasn't. It was discovered by Yanagisawa in 88, and we started studying it in 89. And  as we bounced back and forth between the cardiovascular literature and the immunology literature,  the thing that seemed to really matter was the endothelial cell at the terminal resistance arteriole and down in the capillary. That's where the adhesion molecules, the leukocyte  adherence had occurred and things got turned on. There was a balance between endothelin and nitric oxide, that the nitric oxide not only vasodilated, but it also turned off the  expression of the adhesion molecules. It also kept the platelets from being activated, kept  the leukocytes from being activated. So everything I've looked at embarrassingly now for 40 years  says that keeping the endothelial cell happy is a key aspect of avoiding sepsis and treating  sepsis. Statins work because they make endothelial cells happy. Steroids don't because they turn  off nitric oxide production, which is an important part of the healthy endothelium. And I'm stuck with, I described interleukin-3, so I've been in this a long time. I still have to  ask myself the question, am I going to do a better job draining the abscess, resuscitating the patient, supporting them as they heal, or are we going to find the magic bullet we've  been looking for for the entire 50 years of my career? Thank you to all of the presenters. This was a great session, by the way.  Thank you. I think it's a terrific question, and there's a lot of philosophy in that, and I think it's a very important one. There's no substitute for rapid recognition, source  control, early antibiotics, and rapid resuscitation. I don't think anyone argues with that. But we know that with all that, the mortality in sepsis is still about 25 to 30% in the  best centers in the world. So what can we do to fix that delayed mortality? And mortality in sepsis continues to rise even after the acute phase. So most people survive that,  and they go on to die after day 14. And I think that's what we are trying to address, because I think we are now getting in pretty early and resuscitating people extremely well,  but we do not understand what happens in the late phase. I think that, well, whether these – well, I – endothelin is a target, or whether we try to look at some other cytokine,  I do not know. I'm not an expert on that. But we can only go by what's there in the literature.  Hi, Diego Ray, Endpoint Health. Another approach for addressing immune suppression that's now out there is checkpoint inhibitors,  a specific example being BMS with their anti-PD-L1, which is especially interesting. It seems to be a nowadays rare case of a pharma company showing interest in this space.  But I was just curious if there's any comments on that approach in managing a suppression of the adaptive immune system. I think there's a lot of work in oncology on those,  and obviously they've led the field there. The data in sepsis are, again, very preliminary, and I'm not aware of any large trials in those.  And so I guess, yes, that's probably as much as I can say.

Video Summary

The video discusses the strengths and weaknesses of animal models in understanding human critical illness. Animal models are useful in studying disease biology and developing treatments, but their accuracy in mimicking human diseases is questioned. Researchers should improve animal models by considering factors such as gene expression, age, and strain differences. The speaker also recommends implementing co-clinical trials to align animal studies with human clinical practice.<br /><br />The video also focuses on immunomodulation in sepsis, where the body's immune response to infection often causes illness. Numerous clinical trials targeting cytokines and coagulation cascades have not shown significant improvements in patient outcomes. Trials exploring immunostimulatory therapies, such as low-dose corticosteroids and IL-7, have shown promise. However, patient heterogeneity poses a challenge in conducting clinical trials for sepsis, as factors like patient selection and timing of intervention can influence treatment effectiveness. Despite the challenges, researchers are striving to find better therapeutic strategies for sepsis patients.

Asset Subtitle

Immunology, Research, 2020

Asset Caption

The Society of Critical Care Medicine's Critical Care Congress features internationally renowned faculty and content sessions highlighting the most up-to-date, evidence-based developments in critical care medicine. This is a presentation from the 2020 Critical Care Congress held February 16-19, 2020, in Orlando, Florida, USA.

Learning Objectives

Review bench techniques in clinical research, including relevance in study design as a primary readout for clinical outcome variables
Analyze strengths and weaknesses and demystify commonly used state-of-the-art basic science technologies
Discuss how state-of-the-art technology driving discovery in basic scientific research is applicable to critical care
Appraise basic scientific discoveries that can translate to clinical trials and to the bedside in critical care

Integrating Translational Genomics Into Clinical Medicine

Hector R. Wong

Mitochondrial Origin of Multiorgan System Disease

Richard Levy

Building Better Models

Kenneth E. Remy

Next-Generation Diagnostics With CRISPR

Daniel Chertow

Immunomodulation in Translational Science

Balasubramanian Venkatesh

Meta Tag

Content Type Presentation

Knowledge Area Immunology

Knowledge Area Research

Knowledge Level Foundational

Knowledge Level Intermediate

Knowledge Level Advanced

Membership Level Select

Membership Level Professional

Membership Level Associate

Tag Genetics

Tag Infectious Diseases

Year 2020

Keywords

animal models

strengths

weaknesses

human critical illness

disease biology

treatments

accuracy

gene expression

age

strain differences

co-clinical trials

immunomodulation

sepsis

clinical trials

patient outcomes

	Society of Critical Care Medicine 500 Midway Drive Mount Prospect, IL 60056 USA Phone: +1 847 827-6888 Fax: +1 847 439-7226 Email: support@sccm.org	Contact Us About SCCM Newsroom Advertising & Sponsorship DONATE	MySCCM LearnICU Patients & Families Surviving Sepsis Campaign Critical Care Societies Collaborative	GET OUR NEWSLETTER
© Society of Critical Care Medicine. All rights reserved. \| Privacy Statement \| Terms & Conditions The Society of Critical Care Medicine, SCCM, and Critical Care Congress are registered trademarks of the Society of Critical Care Medicine.