Scientist Stories: Herbert "Skip" Virgin, Animal Models and Microbiome Research

Good morning to everyone. Can everyone hear me? Okay, good. All the way in the back there was a thumbs up, so that's good. I very much appreciate the opportunity to speak here. Looking forward to the symposium and hearing all the other wonderful speakers. It's quite a group that's been assembled. Today I'm going to talk about the issue that's really on this slide, and that is that there are chromosomal genes, And we're all very excited about our own genomes. That's great. Genes are very important. But then there's this community that lives in and on us. And just to say, the viruses live both on us and in us, as we'll get to. So that's a little different than bacteria, fungi, protists, except during infection. But it's pretty clear that this community, including the genes, interacts with these genes in very important ways.. And that influences both biology, basic biology, disease biology, and how we interpret animal models for our studies. So that's the subject of the talk. In my lab, we study viral persistence, how interferons, autophagy genes, and inflammation regulate immunity, but also the viral, all the viruses that live in and on us. And today I'm going to be talking about these studies. And 3 faculty members who are now at Oregon Health Sciences University, Washington University, and UT Southwestern as new tenure-track assistant professors contributed substantially to the work that I'm going to review today, most of which has been published. And the way I'm going to do this is I'm going to talk about concepts and I'll show you a piece or maybe 2 pieces of data which are the central data for that particular concept. In an overview kind of a mode. So, I'm going to define some terms here. In my talk, the microbiome is all of the organisms that live in or on the host, including the virome, the bacterial microbiome, archaea, the meiaphona, fungi. And the metagenome is the sum of all those genes plus our own genes. Now, unfortunately, in the literature, microbiome has come to mean bacteria. But that's not the whole story. And so I think it's very important that when we talk to each other that we don't substitute this particular word. I'm going to go back here. We don't consider that this is only bacteria and that we be specific. If we talk about viruses interacting with bacteria, we should be specific. And fungi are important as are protists. So, that's just how I'm going to use the terms in this talk. So, by outline, I'm going to talk about the virome as a concept. I'm going to try to convince you that this set of viruses can have very substantial physiologic effects. That then, if we're thinking about one component of the microbiome, that it turns out that these components interact with other components in biologically meaningful ways. Such that the silos of expertise that we find ourselves in— virus people, protist people, bacterial people— actually limit how we understand biology because there are interactions between those groupings. And I'm calling those here transkingdom. I'm going to talk a little bit about those potentially occurring in humans and then close with a way of using some of this knowledge to change, alter, perhaps improve mouse models. So, that's the outline. So, we'll start with the virome. So, this is the original concept from 2009. I'm gonna come around here. Can people still hear me? So, I can see the screen a little bit better. This is the number of people on Earth at that time, and this is the number of those people who have all these different viruses. People in the room here, you have an average of 10 chronic virus infections. There are 8 human herpesviruses. All of you have 4 or 5. Some of you have all 8. If I were to swab your throats right now, you are shedding these viruses, but you are apparently healthy. They are part of your genome in a certain sense, and they are in active interplay with your immune system, such that if you immunocompromised a human, these viruses can cause disease. Ah, but that means that they're running all the time. The immune system's suppressing them. They're persisting. Which means that the immune system itself is in constant motion because these viruses are inside us and other mammals. That's the concept. Okay? This does not represent all of the viruses. The viruses have not been plumbed. We can only attribute 5 to 17% of sequences in any viral prep from human stool or mouse stool to a known virus. But we know when we assemble them that there are novel viruses there. So, there's a massive number of organisms that we haven't even begun to study. So, this is the concept. Over the ensuing years, this has evolved a bit. So, we believe each person and each mammal, each model organism has its own virome. These, for us, are in exchange with other mammals— rodents, primates, bats, etc. These diversify very rapidly, and broadly speaking, you can think of them as eukaryotic viruses, which are viruses that require eukaryotic cell for their replication, the endogenous viruses that live in our chromosomes and that emerge under various circumstances, and then prokaryotic or archaeal viruses, which persist in bacteria. These are the raw material for transkingdom interactions between viruses, worms, viruses, bacteria. These then interact with host genetic variation and influence the genotype-phenotype relationship in manners such that if you wish to understand how genes function, you need to backtrack and think about these other elements that in sum that generates the level of systemic inflammation, the immunophenotype I'll define in a moment, disease susceptibility. Some of these are symbionts. They're good for you, or at least mice. There's a wide range of variation here such that since they are important, that may contribute to inter-individual variation. And of course, some of these are pathogens. And so there are classical pathogens. But this is perhaps the smallest part of what viruses really do when they interact with other organisms. So this is the structure of the talk. I'm going to talk about each of these elements in various parts as we go forward. So the first concept was this issue that if we're chronically infected all the time but apparently healthy, that may define our immunophenotype. And the immunophenotype is the state of your immune system when a new thing comes along, either a self-antigen or a foreign organism. Okay? And so these viruses may be setting up our immune system and then controlling the genotype-phenotype relationship. There are two observations that got us started. One is that a chronic herpes virus stimulated interferon secretion, changed gene expression, and led to resistance to infection in tumors. And then a second that a chronic norovirus, which we discovered these guys, which was a lot of fun, interacts with a host autophagy gene and changes the phenotype of a disease related to IBD. So I'm going to briefly show you some of the data that just got us started in this concept. So this is the first one. It's a pretty simple experiment. We took mice and we gave them a latent gamma herpes virus infection. This is related to your EBV or KSHV that infect humans. And we also had a virus because we'd done the genetics where we could engineer the virus so that it could undergo an acute infection but could not establish a latent or persistent infection. Of course, that's what herpes viruses do. They're forever. Once you've got it, you're never getting rid of it. Okay, but we can engineer viruses that have acute infection but don't have chronic infection. And then we gave these animals a very high dose of an intracellular bacteria, Listeria monocytogenes. Papers will be down the left-hand corner in the talk. The latently infected animals were protected compared to the controls. This lasted 3 months. It actually lasts 6 months. And the mechanism of this turns out to be that the virus is stimulating the set point of the innate immune system by inducing the expression of interferon gamma. So, the virus turning over is driving gamma, and the interferon gamma protects against the bacteria. Since we have at least 4, it seems conceivable that the stimulation all the time of those kinds of viruses that are living inside your body but are, for instance, at your mucosal surfaces, may alter how we then interact with pathogens or other organisms. So that's an example of herpes virus latency conferring a symbiotic advantage. On the host. Now, how important are these? Well, we hypothesized that phenotypic characteristics of mammalian hosts might be substantially altered. So we decided to do what we thought was a pretty tough test. If it changes the immunophenotype to have a chronic eukaryotic virus, could that viral infection alter a profound genetic phenotype in a host? Okay? So to answer that question, I'm going to tell you a little bit about this gene, HOIL1. So HOIL1 is a human— is mutated in humans. Children who are deficient can suffer from severe infections, but also hyperinflammation at baseline. The original study, which we participated in, 3 children out of— in 2 families died from this deficiency. It turns out that if you make a mouse, you get quite a striking phenotype. Knockout mice, if you infect them with Toxoplasma gondii, they die. They're very susceptible, so they're immunodeficient. Listeria monocytogenes, they die. Very immunodeficient. Tuberculosis, they are resistant. So it's the opposite. And this herpes virus that I mentioned to you, they are also resistant. Since these are severely immunodeficient, these two here, and this is controllable, this gave us the opportunity to give this virus to these mice to see whether we could reverse this phenotype. Now remember that immunodeficiency in humans, even for the same gene, you can have widely different phenotypes in the patients. So this is the experiment. We took the mutant animals, we gave them a latent infection, and then we challenged them with Listeria. And this is what happens. These are the mice challenged with Listeria. They die. They're very susceptible. And this is if you take the mutant mice and give them a chronic viral infection, you reverse the phenotype. Okay? So these are powerful enough effects that you can reverse a profound genetic immunodeficiency. The reviewers of this paper really didn't want to see it for just one gene. So they had us do it for Caspase knockouts and IL-6 knockouts. Same thing happens. So multiple genetic immunodeficiencies can be reversed by chronic infection with one virus. So that's an example of an interaction here in the genotype-phenotype relationship where a virus is altering the effect of a host gene with a significant effect on biology. So the second example. So it turns out that if you take barrier-raised mice that lack this autophagy gene, ATG1601, they're actually hypomorphs. And you look at them, they have normal panacells. Those are some cells in your intestine which make antimicrobial peptides. If you give them chemical agent DSS, you can induce colitis. But in fact, the small intestine is normal when you give DSS even though the colon is abnormal. And so these wild-type mice or the mutant mice, this is what it looks like. If you take the mutant mouse but not the wild-type mouse and give it a chronic viral infection, now you get abnormal paneth cells. You get a new kind of colitis from DSS. And now you get a completely new phenotype, which is villous atrophy or blunting. The virus— the virus in the wild-type mouse doesn't do any of this. The virus in the mutant mouse does this. So this is an example of— sorry about the little green guy saying I'm sharing my screen. I'm pleased to be sharing the screen. This is a virus plus gene interaction. The gene doesn't do it, the virus doesn't do it, only the virus plus the gene does it. How important are these phenotypes? Well, our collaborator Thad Stappenbeck has looked at this gene and taken it to humans and looked at the role of different genes including ATG16 in making abnormal paneth cells. And it's seen in humans. If you then RNA-seq patients' intestinal tissues, you find that there are two subtypes by gene expression within Crohn's disease. This correlates with this and, in fact, predicts recurrence after surgery. So patients who have abnormal Paneth cells and a particular signature recur much more rapidly, allowing Crohn's disease to be subdivided into two separate diseases based on these pathologies. So now think about that. Mouse, virus, no phenotype. Mouse gene, no phenotype. Mouse virus plus gene, phenotype. Phenotype not seen in humans before. Show it in humans, then see that it has a disease correlate. Okay? So these interactions matter. It turns out that you can prevent the virus-triggered pathology by giving antibiotics, indicating that the virus is interacting with bacteria. Okay, so that is an example of another interaction with host genetics which gives a phenotype. So now to this concept of transkingdom that I just introduced. So, you know, we all talk about our various pieces of the microbiome, myself included, being viruses, sometimes bacteria. And then we say that that's explaining a lot of biology, but at the end of the day It's really complicated. One, it's not my fault that it's complicated. Okay? All right? Just saying. But importantly, we can't explain all of any phenotype. So I've been thinking about this and I think this is in parallel to this. This is what the solar system looked like before Copernicus. Now this was the idea that the Earth was the center of everything and to map the planets This is what you have to say the orbits of the other planets look like, okay? Now, I think that we tend to think about the pieces of the microbiome interacting with the human as if the human is the center. But the solar system only made sense when all of the components were put relative to each other, and then you get this, which seems more accurate as we currently understand it. There's a certain parallel to the way that we look at the microbiome. We tend to think that the, the human is the center and we look at just the individual component. It's really complicated. One wonders whether we could simplify it with a more inclusive view. So I would suggest that this concept is outlined in these two different reviews. We may come back to some of the issues, but the idea is that the piece that we're missing is that there are interactions between the parts of the microbiome. And then that explains how the microbiome interacts with us. So, in this paradigm, it turns out that helminthic worms control the virus which controls the gene expression which controls bacterial resistance. And that bacteria actually control a persistent virus which interacts with the host gene to control susceptibility to disease. So, I'll briefly take you through those.— So, this is the first story. Helminthic worms in the mouse and also humans induce an immune response encompassing interleukin-4 and interleukin-13. So, very simple experiment. If you take a mouse and you infect with a virus which is an indicator virus, when it emerges from latency, it makes the mouse light up because it expresses luciferase.. And then you can just image. Then when we give the worm, we can ask, did the worm reactivate the virus in the animal? Okay? And that's what happens. If you give the worm and you measure light coming from the animal, which is reflective of the reactivation, the worm reactivated the virus. I'm not going to explain all the data. It's all in the paper. But this is a very important evolutionary concept, this next slide. So, we think of cytokines acting in the immune system to control infection. This is not that. This is different. This is that the virus has evolved its own promoters to sense the cytokines that are made by the animal to manipulate and allow it to live the way it wants to live. So, it turns out that this is not about the host controlling the virus, this is about the virus evolving promoters such that it can leverage what the host is doing. —and so if you take the reactivation gene called Gene 50 of the virus and you give interferon gamma, you can turn it off. So gamma blocks reactivation. But if you give interleukin 4, which is induced by the worm, you can actually reverse it. And it turns out that this promoter for this gene actually has 5 different promoters and that one of them is responsive to interleukin 4, same one's responsive to interferon gamma. So the virus has evolved to sense these two cytokines and to sense the competition between them, and that's how it's controlling reactivation. Actually makes sense. The ancestral herpes virus is shared between birds, reptiles, and mammals. They have been studying you far, far, far, far longer than you have been studying them, okay? And so this gene actually has five different promoters, and it just arranged one of them to be responsive to the host immune system. So if that's all true, you should be able to take animals and give them anti-interferon gamma and get very little reactivation of the virus. IL-4, get very little. But if you give IL-4 to turn the virus on and you give anti-gamma to remove the brake, you should reactivate the virus. And that's what happens. This is a log scale. So the virus is actually a 2,000-fold part sensor, has a brake and a turn-on. And that's how latency is regulated by these viruses. The worm works through stimulating IL-4 as shown by the dependence on the transcription factor STAT6 for the phenotype that I just showed you. So this is another example of a transkingdom— or a first example of a transkingdom interaction in which a helminthic worm is turning on the eukaryotic virus, which has already been shown to have substantial effects on the host. Okay? So we really can't think of the virus independent of the worm. It is very unlikely that you should think about the worm independent of the virus. At the same time that we published this, David Artis, we collaborated with, and it turns out that the same worm changes the CD8 T cell response to another virus, the norovirus that I mentioned before. So that the same worm both inhibits the response to one virus and turns another virus on. Second transkingdom interaction example. Turns out that bacteria control the chronic infection with noroviruses. It's a very simple experiment I'll describe in a moment. So, these are the cruise ship viruses. They are responsible for the worst 48 hours of your life. And everybody has had them. We had the good fortune to be able to discover them in mice, which gave us a nice mouse model. Some of these viruses persist forever in the intestine of mice. That's this orange-colored one here. Other viruses do not. And this virus isn't actually just a bad virus. It just is a systemic virus. It's not a gut virus. So it goes to the spleen and lymph node and stuff like that. You can map the genes between the different viruses. But we're studying this persistence and we did the following experiment. We gave antibiotics to an animal and then gave it infection. And what we got was that antibiotics prevented persistent infection. So that the virus requires bacteria in order to establish a chronic infection, remembering that this is the virus that interacts with the autophagy gene to change the paneth cells. But it requires the bacteria to do that. Is it really the bacteria? Well, it turns out that if you treat with antibiotics, you can prevent the chronic infection. But if you do a fecal transplant back, you can restore it. This allowed us to do a pretty simple immunologic experiment. So we're immunologists. We'd like to know what's responsible in the host. Now we're going to map the host genes required for the ability of an antibiotic to block a virus infection. Antibiotics are pretty commonly used, I think. All right? And so we just took a bunch of different knockout mice and we looked at whether the antibiotics worked. And the antibiotics work without interferons, without adaptive immunity, without IgA, without toll signaling, without the sensors for inducing interferon. Unfortunately, they work without autophagy genes. It would have been a lot easier for us. But there were 3 genes which are required for the antibiotics to work. And that's STAT1, IRF3 involved in interferon signaling, and a receptor for a poorly studied interferon called interferon lambda or type 3 interferon. So that led to this experiment where we just take interferon lambda, which was already available. It's also been tested in humans. And interferon lambda cures the infection. All right? So bacteria required, works somehow through interferon lambda, interferon lambda can prevent the infection. And this is one of my favorite experiments. This shows that if you pay attention to these trans-kingdom interactions, you can discover new pieces of biology. So I teach immunology. They still let me do that, okay? To the graduate students, the medical students, and I teach them basically, or I used to teach them, that the adaptive immune system is really smart. It sees the antigens, it recognizes them, and wipes out the infection, and that's how you sterilize tissues. And the innate immune system is kind of stupid. It's short-term. It just keeps everything under control until the strong cavalry of the adaptive immune system comes up and comes in and cleans up the mess. So we did the experiment of trying to cure the infection in normal mice. And it— this now we're allowing the persistence to be established and then the interferon lambda is curing. So we're clearing persistent infection. Now, of course, in the absence of the adaptive immune system, since adaptive immunity is required for sterilization of tissues, and that's what I've been teaching for the last 20 years, if you do this experiment in a Rag knockout mouse, the interferon lambda is not going to work. Works perfectly well. The innate immune system is capable of sterilizing tissues from infection. So there is sterilizing innate immunity. A really pretty paper came out at the same time that this paper came out from Gewirtz's lab showing a very similar phenotype for rotavirus infection. So the two major causes of gastroenteritis in humans both can be controlled by the innate immune system. The mechanism of this, how it sterilizes tissue, is not known and is under study. But it came from studying a transkingdom interaction. So that's an example of a transkingdom interaction in which commensal microbes and interferon lambda control the persistence of a virus that then interacts with the host. And additionally, that there is a form of innate immunity that can sterilize a tissue. So, does this happen in humans? Are there transkingdom interactions in humans? And of course the answer is that we do not know. I'll just show you some suggestive data from the enteric virome in inflammatory bowel disease. So this is a study that we published in 3 cohorts: Boston, United Kingdom with Miles Parks, Boston with Ramnik Xavier, and Chicago. And we took patients with or without IBD and we did their bacterial microbiome and their virome. Their bacterial microbiome looked exactly as expected. The sick patients had lower complexity of the bacteria. And then we did the virome. And we got a surprise. This is the richness of the bacteriophages, the number of species of bacteriophages in Crohn's disease, ulcerative colitis, and household controls. This is a very strong cohort because this is household controls and the virome is determined inside the house. So that, that's a very important control. Same things happened in two other cohorts. So the bacteria are becoming less complex taxonomically, but the viruses are becoming more complex taxonomically. Is there a relationship between the bacteria and the viruses. These are the discriminate bacteria in the United Kingdom cohort based on 16S. So these bacteria were associated with disease. These are the most common viruses. And if there's a red dot, this viral sequence is anti-correlated with Prevotella. Okay, if there's a blue dot, there's a positive correlation. This is basically a picture of the statistical relationship between bacteria and bacteriophage, Caudovirales, one group of the bacteriophages. And that's what the household controls look like. This is Crohn's disease. It's a different signature. And this is ulcerative colitis. It's a different signature. One explanation, unproven, is that this is a part of a predator-prey relationship. That actually the disease is related to the viruses killing the good bacteria, not just the absence of good bacteria or the presence of bad bacteria. And that's something that we're interested in studying. So that's a hypothesis, okay? So that's an example of a disease-specific alteration in the entire enteric virome. We're currently studying this issue in IBD, HIV/AIDS, vaccine responses, susceptibility to type 1 diabetes, trying to map the host genes that control the relationship and to develop diagnostics. We have a number of collaborators, multiple countries, and this is a very exciting effort for us. Part of what we want to do is develop animal models so that the things that we see here can be studied in mechanistic detail. This paper is exemplary. So this is from Doug Kwan's group in collaboration with us. It turns out that the bacterial microbiome in the vaginal vault predicts risk of acquisition of AIDS in a South African cohort of young women at high risk for HIV. And that that correlates with the induction of inflammation by those bacteria. That'll be coming out soon. So there are human studies going on, not just by us, but these included, which may shed some light on this. And the way it's going to shed light on this is that we're not just doing 16S. We're doing 16S, and now we're doing whole genome sequencing, or shotgun sequencing to do the bacteria. But at the same time, we're doing the viruses, and we're working on algorithms for worms. Fungi and the apicomplexans so that we can— and others— so that we can actually get a whole picture of what's there because we've convinced ourselves that studying any one in isolation may miss important things. Does any of this mean anything to mouse models? So, you know, we spent a long time cleaning up mice and making them reproducible so the same thing happens in a given facility. I'm sorry, you can't hear? Okay. It's hard to see the screen from here. Am I on now? Okay. All right. So, can we use this information in any fashion? So, we did a very complicated and long experiment. Tiffany Reese, who's a new assistant professor at UT Southwestern, deserves a lot of credit for this. She did the worm study that I showed. That was a good enough study to get her a tenure-track job. That's good. But then she took on this experiment, which is one of these experiments which you do and you have no idea whether it's worth actually doing, but it takes an enormous amount of work. So, we're going to analyze the immunophenotype testing this hypothesis. Does a history of infection mimicking the early life human exposure to chronic infections alter the normal immune system or change the response to vaccination? And so, here's the structure of the experiment. These are 4 different experiments. They're done pretty much the same way. They're repeats of each other. We take mice and either mock infect them or just sequentially give them infections, a gamma herpes virus like EBV, a beta herpes virus, this mouse CMV. We gave them influenza and then we gave them a worm. Lots of people in the world would be exposed to this and, of course, many other things. Then after we allowed them to recover, we vaccinated them with yellow fever virus vaccine. And then we took blood to do gene expression on blood at day 0 before the vaccination and then after the vaccination. And then pooled all the data together and that's what I want to present to you. Okay? These are normal laboratory mice in our clean facility. It's not a germ-free facility, but it's a high containment barrier. And now we've created sequential infections in them, so what happens? So this is what you would see in standard laboratory mice. This is days after vaccination, before vaccination here, and after. And so this is the progression of the vaccine response for all of the genes which are altered. I'll tell you a little bit about the genes in a second. Some things go up and come down. Some things come up later. Some things are not terribly changed. Okay? Now we're going to look at what happens in the co-infected mice. So that's this column, this column, this column, and the last column. So the things that go up in normal mice— I don't know which ones are normal— in the standard mice, okay, don't change. Okay? Other things are up and change very little. So there's almost no statistical overlap between the gene expression patterns in the blood of the good laboratory mice compared to the serially infected laboratory mice. What do these genes do? So this is the interferon response and memory CD8 T cell response associated genes. This is platelet formation, coagulation, wound healing, and regulatory T cells. This is ERK activation, inflammation, macrophage activation. This is erythrocyte differentiation and genes that are expressed in RSV-infected compared to healthy infant humans, Staph aureus versus healthy infected humans. And this is neutrophil responses and memory CD4s. Basically what you're looking at is that what we go through as mammals early in life, if you mimic it in humans, mimic the human exposure in mice, you get a completely different phenotype. Does this have anything to do with human immune responses? I'm going to use a term here advisedly perhaps to stimulate discussion. Did we humanize the mice? I mean, everybody says that mice— mice don't predict models, they don't predict vaccine responses and everything. Well, maybe mice actually are very representative, but we've cleaned them up to the point which we've made them non-responsive. And I would point out that if mice are not predictive, it is still true that all the major medicines, for instance, in cancer immunotherapy, are targeting molecules which were originally discovered in mice. So there's some relationship between these mammalian organisms here, okay? So what we did is we took these signatures, okay, some of the T-cell signatures, and we asked in a study which compared maternal peripheral blood gene signatures to cord blood, okay, thus mimicking environmental exposure versus no environmental exposure, is there any overlap? And so there is significant statistical overlap between the co-infected mice and adult human signature compared to cord blood. And this one is an interferon signature. And there is an inverse signature, which is a naïve lymphocyte signature, which is changed by co-infection and is different in mothers compared to their cord blood. So, there are definitely sets of genes that we altered which are distinguished between those two groups in humans. So, I'm not proposing that we made the mice humanized, but we certainly changed genes which are different between— after exposure of humans to the environment. So, those are the two signatures. So, this is an example of sequential infection with common pathogens promoting a human-like gene expression signature and altered vaccine responses in animals. So, that's what I've spoken to you about today. I've tried to introduce you to the idea that one component of the microbiome, being the virome, is important.— that the virome itself, individual parts of it, a herpes virus, a chronic norovirus, can have very substantial physiologic effects, reversing immunodeficiency, creating new phenotypes that ultimately turn out to be related to human disease. That, in fact, these agents don't act in isolation. They're what we call transkingdom interactions between worms and viruses, between bacteria and viruses. That in turn have substantial effects, okay? That there is some data that you can find statistical relationships between different parts of the microbiome, being bacteriophages and bacteria, in specific human diseases. And that you can certainly leverage that to take what's a standard laboratory mouse and change it substantially, whether you've made it better or or worse, I'm not commenting on. But we made it different for sure. Okay? Now, if you think about this, just going to put this up here. This is a— now I've stopped. I'm not talking about data. I'm talking about opinion. What do you need as a field in order to really leverage this set of information that is very complicated, in fact, right? Because we've got individual things having profound effects and then interactions between already complicated things having substantial impact. So, first of all, I think we need careful clinical and animal studies. And I don't mean to say that people aren't careful, but there's a lot in the literature that we know is not right. And I suspect it's because the experiments have not been done with full acknowledgement of the controls which are needed for these kind of effects. We need consensus methods, proper controls, databases, and most importantly, full reporting. I don't believe there's one solution to animal experimentation. I don't think that we all have to do germ-free experiments or we all have to use antibiotics. I think that these models, including in model organisms, I think these models should be temp— arranged to answer a question. And then each model should be used for its benefits or deficiencies. If we do an experiment and you try to repeat it and you don't know exactly what we did and you can't repeat it, what we're going to do is we're going to have an argument in the literature for the next 5 years and we're not going to cure any diseases. So full disclosure is really important. The bioinformatics is very important. The viral bioinformatics you cannot do using bacterial bioinformatics. Viruses are not bacteria. They vary far more between things that are, in fact, the same agent. So there's a whole separate bioinformatics, and there's only a few labs in the world that really do this. That's important. Everybody needs to be able to do it. The databases are totally inadequate to studying anything other than the human and mouse and a few others, the model organisms, and bacteria. But to viruses, fungi, protists, the databases are completely inadequate to that task. Okay? Clearly biological hypotheses should be integrated between model organisms, and by model organism I'm including mouse, Drosophila, worms, squid. I mean multiple different models. Each have their advantages, but we don't integrate these things across so that we can ask the right questions. Question in a given system, taking advantage of what's great about a Drosophila system or a mouse system. Okay? Eventually, we're going to have to change these things. So we are in associational science now. I get to publish papers that are basically characterizing fecal material. Okay? And I just characterize them. I try to do a good job. I find associations and I publish papers. But there's not a single mechanistic element in most of these studies in our field. We are going to have to manipulate the ohms and we— and find out what happens, do an actual test of mechanism, and we have to find the mechanisms. I'm going to just throw out that in the end this is really important to disease, and we're currently doing all of our sequencing using methods which are not standard for human diagnostic testing. Okay? I run a big pathology department. I can tell you the research sequencing that is done in this field is not adequate for immediate translation to humans. Okay? It will all have to be done over again in order to bring it to humans. Another thing we could do is actually do the sequencing right in the first place. Okay? And that might actually be useful. You will need clinical-grade databases. And frankly, this is a super exciting field. It is clear that there's major pieces of biology that are as yet unexplored. And I don't think we're training enough people or putting the young people who actually know how to do this in powerful faculty positions early enough in order to move this forward. So that is how I'll end. That's an editorial statement at the end about if we could do all that, that'd be great. And I hope this conference will shed some light on these various things. There'll be lots of counter-opinions to discuss. So with that, I'll close, and thank you for your attention.