Great. Thank you so much. And I echo that. Thank you for everyone for sticking it out till the end. And thank you so much for the invitation to speak about immunomodulation and sepsis. Here's my disclosures. So our updated definition of sepsis defines it as the life-threatening organ dysfunction caused by a dysregulated host response to infection. And this definition really posits that we are very good at identifying regulated and dysregulated responses. And so I think my objective today really is to go through with you how are we measuring and identifying regulated and dysregulated responses and looking at some of the trials that have tried to manipulate this and think about how we'll do this moving forward. So certainly we've learned a lot about the host response to infection. And of course, we need to simultaneously detect pathogen, recognize it as foreign, activate our leukocytes and white blood cells and immune system to start attacking that pathogen, activate our vasculature to enable cells to come to the site of infection and neutralize that pathogen. And then we need mechanisms to prevent tissue damage. So we need to be down regulating it. And certainly I'd refer you to this really outstanding review by Drs. Wiersinghe and Vanderpool recently. But as we think about manipulating the immune system, we could think about attacking each one of these places. We've certainly observed that patients with sepsis and certainly the more severe organ dysfunction and those likely to die seem to have excessive hyper inflammation. And therefore, maybe we should be attacking this aspect of the immune response. And whereas we used to believe that the compensatory anti-inflammatory response started later and was maybe a second phase, the more we investigate, the more we realize that these processes, these anti-inflammatory down regulation of adaptive immunity and T cell suppressor cells is happening as early as we can detect patients with sepsis. So if I was giving this talk 10 years ago, I'd be saying, you know, we've tried very hard to target the elevated cytokines that we see are in excess in the plasma of patients with sepsis, septic shock, or those who die. And yet every time we try to intervene, we have essentially negative trials. And certainly the 90s were a story of the ICU being referred to as the pharmaceutical graveyard where drugs come to die. But perhaps we're entering a new era. Perhaps there is kind of a new hope ahead if we can move towards a precision paradigm where we can identify the causal players, target the right patient at the right time in the right way. So how do we measure this? And if we start again, I'm highlighting just that we really believe both the hyper inflammation and the impaired adaptive immunity are happening concurrently. Starting first with hyper inflammation, should we be looking at inflammatory cytokines? Certainly COVID showed us that IL-6 and IL-8 were less dramatically elevated in the plasma of COVID patients than typical bacterial sepsis. And yet tocilizumab and anti-IL-6 receptor seems to be helpful. So maybe the extent of inflammation is not the whole story. Should we be targeting vascular activation? Should we be looking at specifically at cells, immature neutrophils, or kind of activated neutrophils? And I think one way that we really need to move this forward is trying to dissect which are the causal actors. We in critical care certainly have a history of focusing on potentially correlated plasma proteins. And most cytokines are moving together. But we have new techniques that we could perhaps try to dissect. Is it IL-6 or IL-8 or IL-10 or interleukin-1 beta? How can we best attack the immune system? Or should we stay broad and use something like steroids that probably have a broad effect on multiple causal actors? Again, we can use observational data and sometimes genetic techniques like Mendelian randomization or mediation analysis, thinking of patients essentially randomized to being high expressors or low expressors of cytokines based on their genetics. And that might give us insight into which of these inflammatory markers is having the strongest effect. And then we can feed that forward back to our causal models, whether those are animals or in vitro, trying to generate more data, which truly are the causal actors. Certainly, we need more quantitative traits in sepsis. So we need whether it's plasma proteins, metabolites, mRNA transcripts. Not only do we need to identify them, but then we have some work to do to really identify thresholds that can be useful at the bedside. How do we tell someone is high IL-6 versus intermediate IL-6? I think we've also learned to expect and anticipate heterogeneity. And so going forward, each clinical trial, we should be explicitly collecting the samples we will need to test for this heterogeneity retroactively. Our group focused on the interleukin 1 pathway when we identified a genetic variant in the interleukin 1 receptor antagonist gene, IL-1RN, that was consistently associated with reduced risk for ARDS, also with better survival in sepsis. And the minor allele was also associated with higher levels of plasma IL-1RA in volunteers who were given LPS or endotoxin as a stimulus. So we believe this is a functional SNP. We did some RNA sequencing to identify this locus as a site of allelic imbalance, meaning that you transcribe the C allele more efficiently, just a slight bias, more efficiently into the full transcript. So perhaps these patients are protected from ARDS and from death and sepsis because they are more efficient producers of IL-1RA. And then we see in the plasma in the VAST trial that these patients also had lower expression of plasma IL-1 beta when measured. So of course, this brings up a paradigm, well, if these patients are genetically protected from ARDS or sepsis death due to more efficient IL-1RA production, then could recombinant IL-1RA be a helpful drug? And of course, that's been well tested in sepsis. Two phase three trials failed to show an overall mortality benefit. That was statistically significant. I'm showing you the Fisher trial back here in 1994. But because our genetic data suggested that there is heterogeneity in how we express IL-1RA, we asked whether plasma IL-1RA or IL-1 beta might interact with the treatment. And I will say we hypothesized that it would be either the high IL-1 beta patients who would benefit from getting recombinant IL-1RA, or it would be the low plasma IL-1RA patients. And we were wrong. We did see a statistical interaction between the treatment effect and your baseline plasma IL-1RA levels, but in the opposite direction than what we predicted. Now, IL-1RA and IL-1 beta are highly correlated in the plasma. So perhaps we are just, IL-1RA is a better marker for dysregulated IL-1 beta signaling. But as I show you here, it was actually the high plasma IL-1RA patients here in red that had a statistically significant mortality benefit from being given exogenous, being randomized to exogenous recombinant IL-1RA. This, of course, is retrospective work. It's a subgroup analysis and really warrants testing again, but highlights again that heterogeneity. And we have other examples of heterogeneous treatment effect. So more recently with COVID, we seem to have heterogeneous effect of steroids that perhaps can be predicted by your clinical status, whether you're on oxygen or not on oxygen. Pratik Sinha did a nice study in the New York population very early during the first COVID wave and applied a latent class analysis to ask how many classes of COVID patients are there. These were all hospitalized, critically ill patients, but their latent class analysis detected two subphenotypes that differed by a lot of the severity of illness markers that you might anticipate, such as lactate, troponin, creatinine, and other markers of organ dysfunction. And they observed a statistical interaction between corticosteroid use and that subphenotype where it was the more hyperinflammatory or class two patients who seemed to benefit, whereas the less inflamed patients did not benefit. Perhaps similarly, we investigated a hospitalized cohort of patients with COVID and did a fairly high throughput proteomic analysis with a platform called O-Link. So we had about 700 plasma proteins, and we could identify a plasma signature of severity. And certainly, the more severe your protein severity score, the higher your mortality. We investigated in our cohort where approximately 40% of patients received steroids. This was early in the pandemic before the publication of the recovery trial. And we also observed statistical interaction between your protein severity score and the apparent mortality effect of steroids, where patients with a low protein severity score seemed to do worse when they were randomized to steroids. Caveats to both those last two studies, those were non-randomized application of steroids, but suggest that, again, our subjects are heterogeneous. We should anticipate that their response to treatment might be heterogeneous. We've seen heterogeneity of steroid effect that could be predictable by gene expression. Hector Wong, of course, did groundbreaking work to identify different classes of gene expression in children with sepsis, and really innovatively had a technique to display the gene expression array with what he termed mosaics, where very quickly, fairly untrained investigators could detect these different patterns where subclass A had a much higher mortality in children compared to the either two subclasses. And in subsequent work, he showed that, again, in a non-randomized fashion, subclass A septic children who received corticosteroids seemed to have a higher mortality than those who did not receive steroids. And he has designed a clinical trial to test this prospectively. The Vanish investigators, led by Drs. Ancliffe and Gordon, also investigated a subgroup analysis of the patients who had gene expression. Their group had shown two gene expression classes in adult sepsis that they termed steroid response signatures one and two. And very surprisingly, steroid response signature two, which is usually a very low mortality subphenotype of sepsis, in red, these patients were randomized to steroids in Vanish and had a very high mortality of about 50% when typical SRS2 mortality is about 75%. This was a randomized trial, but a subgroup analysis. So these data are based on only 60 patients. But again, suggesting that perhaps some patients are harmed by our therapies. How about the other side of the immune dysregulation? How do we identify it? Perhaps we can just use lymphopenia. So the overwhelming majority of our septic patients are absolutely lymphopenic, with lymphocyte counts less than 1,000 on admission to the ICU. And the persistence of that lymphopenia, Richard Hotchkiss has shown as a strong predictor of mortality. Drs. Monterey and colleagues have shown that low expression of HLA-DR on monocytes, on CD14 monocytes, is also strongly associated with mortality and associated with tolerance to endotoxins. So these monocytes, when removed from septic patients and stimulated in a Petri dish, produce less inflammatory cytokines than normal monocytes. So could we use either of those as markers of patients who need immune enhancement? The IRIS-7 group, Drs. Francois and Hotchkiss among them, did a really compelling trial, a phase two trial, just as proof of principle that giving recombinant interleukin-7, this is classically considered a T-cell maturation factor, that when they gave this treatment for four weeks after sepsis, they could observe an increase in both CD4 and CD8 cells. Here in red and blue, those were the two different doses of interleukin-7. This trial was not powered for mortality. I think they had fewer than 50 subjects. And the same group have now a really compelling, almost point-of-care assay to kind of test precision response, again, by removing the patient's whole blood and then testing in a Petri dish when stimulated, do these patients' cells respond by producing more interferon gamma? And so you can see a readout of patients who seem to be responders here. Each dot is kind of like a colony of interferon gamma compared to the final well, two patients who are non-responders. These patients' cells do not respond to IL-7. So perhaps we could be doing these assays to ask ourselves whether a drug would be helpful for our patient. The PROVIDE trial was just published, I think, in late December, and really innovative trial where they tried to identify both hyperinflammation or immune paralysis early during sepsis. And the trial started as an observational trial, excuse me, an observational study of about 200 patients to better understand the immune expression. And then a small phase 2, 36 subject portion where, based on their assays, patients could be, if they were either hyperinflammatory or immunoparalyzed by these two definitions here, could be randomized to therapy specific to their immune activation status. So again, in the 240, they were able to show that the hyperinflammatory group had a very high mortality, over 70 percent, immune paralyzed in blue, just slightly lower mortality, and their intermediate group, who had neither hyperinflation nor immune paralysis, had the best survival. Then in the randomized fashion, 34 of the 36 subjects were in the hyperinflammatory group, and you can see that they did not observe a signal for better survival, and they observed a very high mortality of 80 percent. So where do we go from here? You know, I started the talk saying that we know that these processes are happening in the same patients concurrently, and so maybe it's time for us as critical care physicians to embrace combination therapy, like our oncology colleagues. Can we come up with the right markers and the right causal pathways to be precisely both suppressing certain arms of the immune system while enhancing others? And I think that will be our real challenge. So our way ahead, I think we need to start getting used to thinking of all of our treatments as potentially having both benefits and harms, and start looking for ways that we can predict what might be the individualized treatment effect for our subjects, realizing that all of our trial data is giving us kind of a summary or average that really is not any one true patient. All of our patients are these blue, green, and yellows who have maybe a mixture of predicted response to any therapy, and we need to think about all we're seeing from trials is that average effect. So we will need more observational study and more clinical trials that collect molecular data so that we can test for this heterogeneity. We need more rapid point of care type testing so that we can assay these and start to understand our thresholds of positive and negative. And importantly, we need more therapies that we can start testing in this way. Thank you so much for your attention.

immune modulation

sepsis

dysregulated response

hyper-inflammation

adaptive immunity

precision medicine