 And we've been watching the microbiome rush in. So let me give a little bit of a cartography view about the microbiome. A few years ago we had the first big looks at the composition of the microbiome, and the way that it felt just a few years ago is a bit like this. Now we've been hearing the last two days some splendid summaries of what happened from HMP, Metahit, and so much work that's been done in addition around the world but basically based in the NIH portfolio on these areas. And of course the large-scale compositional features of the microbiome are understood. We know something about the inter-individual variation of the microbiota. We know a little bit or beginning to understand what the questions are about how to think about diet and microbial composition. And we've learned a bit about how it changes during the lifespan, which is very important for understanding physiology and also for designing studies with regard to disease. Now what I'm going to do is summarize where I think we're going this morning in three ways. First of all, some highlights of disease association studies that have really influenced me and some of which we'll be hearing about this morning. And I think that they've fallen to a few categories of thinking about how the microbiome becomes a translational agent in disease biology. Then we'll also give a little bit of my taste about functional ecology, which I know has been brought up early in the meeting as well. But I think it's very important in thinking about how to frame what we should be studying about the microbiome in order to understand its contribution to disease and how to intervene. And then finally some conceptual questions about how one would actually carry out interventions. So first of all, they hit and run idea. Again, I don't believe this was covered in the previous discussions, but I think what's really important is we think about the microbiome as being a group of guys that make some very important products which could affect their physiology in some cases in the disease manner. But something which has really lingered with me, and we'll hear a bit more about this morning, is that there's a hit and run effect that the microbiome can play, meaning that at certain points in life, particularly in early life, the products of the microbiome can affect the physiology of us at a unique window of time. And so even if later on your microbiome totally changes, or if you at that point decide to therapeutically modify it, it might not matter because that unique impact that it had on our physiology or disease biology has already taken place. And so an example of that, which we'll be hearing about in a bit, I believe from Marty Blazer, is how your early encounters with microbiome really shapes your lifelong physiology. So I think that's going to be quite an important story. And a second, which is about windows of time and where the immune system is modified by encounter with microbiome. And one example which I'm highlighting here is how the formation and activity in variant NKT cells, which are quite important in asthma and perhaps in colitis, are their formation. It's really dependent on the presence and the type of microbiome in that window of life. And even if you wait in mice a few weeks later and now do your modification, the effect has already taken place. So window of time, receptive group of organisms in the body. And so of course that's going to be quite important to studying whether or not organisms have disease and what to do about it. We'll also be hearing about another thing about, I guess, good and bad guys. This is the second idea. The classic clinical microbiology idea is that there's particular organisms that are making particular products which are an issue for us in terms of disease state. And Stan Hasen's group and that group of collaborators have come up with a remarkable story that trimethylamine metabolizing organisms will release that metabolite and that metabolite actually affects physiology related to apple sclerosis. And so there was really a number of studies on this, but some elegant ones just this past year on the formation of this critical metabolite, which happens in most of us on a Western diet. But if you ingest antibiotics for a short period of time, then you're not making that metabolite. And a crossover study, you can restore that after you've done. So that metabolism is dependent on our organisms. People who are omnivores have those organisms. People who are vegans, at least some don't. And the formation of this metabolite has been revealed in a number of studies affects risk of atherosclerotic disease and perhaps other physiology related to bad cardiovascular outcome, presumably because of the effect of the oxidized form of trimethylamine, which then becomes a bioactive end product. So we'll be hearing more about that story. There's trimethylamine being oxidized. Second issue which we've heard a good deal about is C. difficile. And of course there's two ways to think about the physiology of that disease. One of them is we know that that organism is making exotoxins, which are important for causing the colitis. You could also think about it as an environmental disease where you've degraded the environment, which has allowed the bloom of those organisms. And so you could think of it either as a niche problem or as a bad guy problem. I've chosen in this case to highlight the one, but it's certainly important to share the other. And so the idea has been to restore the microbiome using simple things like fecal transplant. And this is, I think, important for proof of concept, but we're all looking forward to more sophisticated ways to approach this. And so if there's really a deficit, like some guys which are taking care of the habitat in a way that doesn't permit C. difficile to bloom or carry out its function, then what you'd like to do is restore those organisms either by putting the exact ones back in or something which is biologically equivalent. And I was very much of a fan of this study which was reported just about a year ago, which actually took the effort in a mouse model to work through in a systematic way organisms which could fill that hole in the ecosystem and hence prevent the C. difficile bloom. And from this actually a relatively small number of organisms were identified that could do that. And I presume this area will refine and we'll see this as a way to, first of all, diagnose people that have the ecosystem problem and then some ideas of what to do about it. Now, one of the highlights of the microbiome really since the beginning has been its role in energy metabolism and there's been a good deal of discussion about that. And with regard to obesity, the effects are direct and indirect or cause and effect because organisms eat what we eat and so if we have a diet which promotes obesity then there'll be different organisms because of that. But there's been very elegant work which has demonstrated that the organisms actually they and their energy metabolism drives the obesity state or for that matter the malnourished state. And so we'll be hearing a bit more about that and how to intervene on it. And I wanted to call out this example because I think it points out an interesting issue about how to target the microbiome with regard to obesity. So if you say that you have organisms which are say if your issue is obesity and you want to target organisms which are promoting that state, then one idea which was tested here was to use complex carbohydrates to shift the community of organisms to make it less favorable for those ones which are known to be contributing to the obesity state in the first world model. And so here's an example of how this goes where you started this weight state and with a regular high-fat diet obesity or body weight or for that matter non-lean weight and so on goes up. But if you and if you have an lean body weight then a lean fat diet you actually go down in weight but you could even on a high-fat diet using complex carbohydrates you could avert at least the increase which is very cool. The thing though is you're not actually making things quite the same because if excuse me if you look at compositional analysis this is sort of a PCA type of analysis that the normal microbial community which would be sort of this cluster doesn't get restored when you put back complex carbohydrates but it's just functionally it's different and so what we need to think about as well is that satisfactory biologically or have we created or are we looking at other sorts of risks that could be a very interesting thing to challenge to monitor exactly how we're changing the microbiome as we carry out an intervention and to determine whether or not it's beneficial or not in the larger sense. Now I think a really exciting area quite surprising to many of us has been the relationship between the microbiome and behavior and that's been studied in some elegant ways in mice and to a certain extent in people we'll be hearing the first data focused fully data focused talk is going to be taking an overview of that and one thing which I want to call out partly because of our pride in Los Angeles and I see Emma and Mary's here is studies of emotional tone in individuals which are simply getting the products of microbial products and you change the intake of that and basically crossover design and you could watch and then you challenge those individuals with an adverse visual emotional visual stimulus and you look at their response to that using fMRI and actually the microbial products will change the emotional tone of those individuals so I think this is a very exciting area obviously it's a very complex area to study but one that points out the breadth of the physiology and behavioral disease being such an important issue in the world that the microbiome might be affecting that. Now we think about microbial effects being local but they also can be systemic in interesting ways and one of them which I was just calling out is the relationship of microbiome to behavior but another one surprisingly it's relationship to things like lymphoma risk and this is one example using looking at individuals at risk for lymphoma because they have genetic problems with DNA damage and so they go on to things like lymphoma at an early age people with a taxicelastic tasia and a number of those other genetic problems with DNA repair most of those individuals will die from lymphoma or in other cases typically it's leukemia in their early second to third decade of life. Now the surprising thing is that when we study this actually collaboratively at UCLA we're surprised to find out that in this genetic disease where there's 100% penetrance and death with lymphoma we found that if you change the mouse house that those mice did not get lymphoma for a very long time instead of acquiring the disease at an early age and then going on to death it was actually the lymphoma uptake occurred more like a year of age and actually most of the mice died of old age rather than dying of lymphoma. This is very exciting because no one had imagined that these sorts of diseases were any less than 100% penetrant cancer phenotype. And what's interesting is that when we track this down it was related to the microbial composition of those mice and the microbial composition changed just the physiologic burden of genotoxicity DNA damage that was occurring in lymphocytes or any immune cell types throughout the mouse not just in the gut but throughout the mouse day by day. So this is the level of damage of DNA damage in a regular lymphocyte having basically sort of a Rockville or Maryland diet but with a Malawi or Ameridian diet equivalent those microbiota are driving a lower level of DNA damage. And we've actually studied a good deal what are the products of the microorganisms how they're eliciting certain cytokines which drive that systemic constitutive level of DNA damage. It's not a problem to most of us because we repair our DNA damage but for individuals who have difficulties with that then that puts you at risk for going on to lymphoma. And so we did actually careful analysis to identify the organisms that were different in those two consorts of microbial communities to identify the ones that were promoting DNA damage or the ones which were actually protecting against it and from that we were able to identify a number of candidates for each which are the physiologic ones. And the nice thing is if you put the ones which are protective back into a regular mouse then you reduce the amount of DNA damage and you reduce the accumulation, the onset of lymphoma. So it's quite exciting. One could imagine that basically eating in this case it was a lactobacillus strain was a single organism which could make a difference and one could imagine that that sort of intervention might be beneficial for people who are at risk for this and that could be the equivalent of a yogurt diet. So I think this is something which we'll have to look at in terms of bad and good guy physiology for microbial composition but also ways in which one could target particular organisms to intervene. Now there's another more complex way to think about the microbial community and how it affects disease susceptibility and that's at the ecosystem level and worked by many and I know that David Roman was listed to highlight that is to think about the ecosystem issues in the microbiome and inflammatory bowel disease is a good example of that where we know that it's a complex microbial composition which is associated with disease penetrance. And here I've called out an example of work by a group of individuals and this one was led by Curtis Huttenhower which identified how the microbial composition indicated by this individual, this aggregate matrix of a PicoA type of data representation that the composition in normal individuals might be down here but there's sort of a microbial succession which occurs as you go from a non-disease state into increasing disease activity. And those sort of ecosystem differences are also associated with, and this is a study of humans so it's actually a quite impressive study that if you look at epidemiologically at lifestyle traits which are known to be disease susceptibility factors in IBD that you actually find that some of those environmental factors are also related to changing the microbial composition. So this is quite exciting in terms of seeing a coherence between susceptibility traits and the disease phenotype but then we have to ask how much of this is related to driving the disease and how much of it is related to a response to disease and those both might be relevant in terms of intervention but of course some of them which are secondary might not be then a useful therapeutic target for intervention. So this is a very big challenge right now I think in any sort of study of microbial composition and disease this is a bit of a summary of that showing some of the microbial taxa which are associated with disease states but I think what this comes to is the issue about correlation and causation it's a famous issue in philosophy for the last few hundred years and so I called out this worker and Bishop Barkley was the one who had written the one that I had to study in college and actually I've been fond of it ever since but it does point out some ways that we could approach it by being very careful about core design understanding habitat but probably most importantly finding ways to intervene to validate that particular organisms or ecosystems are important and there was a great wonderful study by George Tsuchihara from the Scripps Oceanographic Institute which studied bioinformatically this classic problem people I guess when you look at coastal fisheries there's the succession relationship back and forth between sardines and anchovies and you see this all around the world and so in marine biology the question is is this a causal relationship or is this just a correlation and some very new computational biology was presented in this paper to argue whether or not it was I'd like a vote from everybody how many of you would say that this is a causal relationship and how many would say it's correlation everybody has to vote so who says that it's causal okay some bold individuals causal and who says that it's correlation okay that seems to be the vote we're a democratic country and actually the computational biology says that it's correlation so congratulations this shows the wisdom of the crowd but actually we don't know because it's computational biology I didn't really mean that I can't admire it more than anybody could so I guess in summary this is how I feel where we are right now that we've gone from understanding there's a new hemisphere to getting some idea about what it looks like and a little bit about what's inside but of course there's a lot more to be done now one guide for me about where we should be going in this is to thinking about the ecosystem with regard to function as opposed to just composition and the way that I this sort of data I'm sure has been showed during this last day or so this is just the composition of about a hundred different individuals these are healthy individuals and this just demonstrates the remarkable variation that each of us has because these are all normal but you can see how different it is now that may be biologically important may not certainly in things like in the vaginal microenvironment we know that composition can be very tightly related to function but it's not necessarily the case and the way when I look at composition I sort of think about this it's like looking at the UPC codes imagine if we went into a grocery store and all we could see is cans with UPC codes but we didn't know what was inside the cans would be sort of tough to know what to buy to make a good meal and I think also we know that one of the difficulties with phylogeny is that at the phylogenetic level there could be such diversity probably that's related to acquired traits and we've heard a bit about phage and other sorts of things strain differences but even apart from that just at the genus level there's remarkable difference in function all the vegetables that you see in that beautiful image of whole foods or something like that is those are all from the same genus and so it gives you an idea of the remarkable diversity of function that can be captured at the level of a single genus and here's just an example of a single species the remarkable difference in function and proficiency that one could have so what that really begs us to do is to figure out methodologies to study ecosystems or groups of organisms with regard to function rather than their phylogeny and I've just pulled this in a number of these I've pulled as examples from Curtis Huttenhow I want to acknowledge some of his presentations of this and what I wanted to point out is that the composition at the phylogenetic level is actually a bit more simple to analyze when you look at it at the metagenomic level there's more homogeneity at a particular site in healthy individuals at the metagenomic level and at the transcripts it also is somewhat less noisy and transcripts are would be the expressed part and metaproteomic level is still being done and I know that Janet Jansons here who's done some important work to try and capture for the actual produced products and for that matter metabolites I'll comment on in a moment so the point then is that there are ways to study the proficiencies at the genetic level but actually the products of the microbiome and that would seem to be a very important place to look since it's really the products of the organisms rather than names which are driving our physiology or disease state an example of how members of a consortium of people that I've been working with have been studying this is to do endoscopy on humans take colonic samples and we've done about 300 individuals where we've actually just washed different parts of the mucosal surface endoscopically collected the washes from those different sites and then we've looked at a bunch of things we've looked at the OTUs look at microbial composition and then using imputed metagenome determine what the proficiencies are for the production of production and activities and then also take those very same samples and look to see what was actually being made locally with regard to metabolites and proteins and when you do that one thing that's very cool and this is where computational biology can be quite lovely is that if you look at the 2000 or so OTUs and you look at about 2,000 different metabolites and you just do a correlation analysis you find that a few hundred organisms and a few hundred metabolites are actually closely correlated and so a lot of them aren't but the ones that are you can see by this heat map that a lot of them really go together and when we analyze this in terms of the microbial composition so again what we're doing is we're taking a few hundred samples and each one we know the microbial composition phylogenetically and then we know the proteins that are made and we know the metabolites that are being made and we ask what are the ones that really go together if we looked at the bacteria and the human proteins and the bacterial composition and the human proteins there's a little bit of relatedness but not that much this is sort of a network schematic so there's a certain amount of guys that tend to cluster together but not so much at the level of human proteins and metabolites really very little but look at this if you look at the interactome of all the small molecules in the gut and the bacterial composition you find a very rich interactome and so that would indicate that the two things one is a great deal of the small molecules that are present in the limit of the gut are actually not what you ate or products that are being released by your intestinal mucosal lining even though that's a massive site for secretion but it's actually products that are formed by that stuff from the microbiota and since a lot of those are bioactive it would be very important to analyze them and so here's an example of a data set of a heat map of metabolites which are associated with the Crohn's disease disease state or ulcerative colitis or healthies and I think you can appreciate that at this sort of like 10,000 foot view that there are big differences in the metabolite features of individuals in a healthy state or in Crohn's or UC state and it's quite interesting that these individuals were all taken from individuals who had minimal inflammatory disease so now that they felt good and they were endoscopically normal so this is just sort of the underlying difference in metabolic or actually this is metagenomic I'm sorry this is at the metagenomic level but there is a difference in the production the capabilities of the organisms in these different disease states this is from again from Curtis Hutpower's consortium and it's taking a similar sort of data and pointing out that if you look at that data set it actually is fairly coherent that some of the products of microorganisms or capabilities of proficiency of the microorganisms in the disease state or in the healthy state are fairly different so you're really selecting for organisms which are capable of different things making different things and some of those products are reasonably likely to drive those disease states or to be interfering with things which might be healthy for example the decline of organisms that make free fatty acids which are known to be an important source of energy metabolism that are required for the healthy epithelium and those organisms are making that are depleted in IBD now let's think for a moment about inflammatory bowel disease or most diseases that we all concerned about have a genetic feature to them IBD is one of the ones that have been very well studied for that there's 163 genes which are known to be players probably each individual maybe has a dozen or two that are the players for that fall into different categories of function but what's interesting is that if you look at those 163 loci and if you just act in healthy individuals whether those disease polymorphisms are related to your microbial composition it turns out that a lot of them are so here's the Manhattan plot of the basically you might say these are the 163 loci and what we've done is done basically a QTL analysis to ask what is are there organisms or concepts of organisms whose abundance are different depending on your disease allele polymorphism and so I think you can appreciate that many of these loci are associated with significant hits where they're changing microbial composition in normal individuals so this raises the idea that some of these disease polymorphisms they might actually be playing a role because they're changing your ability to garden your microorganisms and hence the microbial composition or their production are putting you at risk and so we've actually looked at an example of that and here I'll just show you one we know that there's a set of genes which are associated with inflammatory bowel disease which are genes which are involved in the formation of glycans of the mucin glycans and mucin of course is the protective layer throughout the gut and other mucosa that have a big effect in host microbial interaction and so we asked was one of these disease polymorphisms which is associated with Crohn's disease it turns out that there's a no polymorphism that 16% of healthy individuals so 16% of you people actually have a no polymorphism of foot too so you don't have terminal feucose on your O-glycans and the reason for that is that some organisms that are bad like Helicobacter like feucose and they bind to it and that allows them to be a pathogen and so if you have a no polymorphism for feucose you don't have a problem with Helicobacter but you do have a problem with organisms which like a terminal galactose which is what's left when you don't have the terminal feucose so you have to sort of choose your poison and 16% and so we have different populations around the world that probably because of infectious disease burdens have that difference so this is a so it's secretor, non-secretor are foot too positive or foot too negative so you could make mice that are the same way and what we've done is we've analyzed the microbial composition and function and it turns out that there's actually quite a concordance between mice and humans and how you garden your microorganisms your whole microbiome depending on foot too and they're largely concordant and what we've done is we've analyzed it also at the level of metagenome and metabolites and we find out that again between mouse and humans the microbial composition is fairly different we find that their composition is actually dictated by your foot too state and in a significant manner in terms of ones which go up functions which go up and functions which go down depending on foot too and the majority of these differences are the same ones that you see in IBD so this is a correlation and that's actually quite cool even with people that don't have the foot too state the way you garden your organisms with regard to IBD the way you garden them with regard to this one disease polymorphism are concordant which suggests that this might be a genetic trait which changes out organisms in a way that are bringing in functions and products which might be deleterious and put you at a closer threshold for disease so way to think about that that the genetics of the host are changing your microbial composition your what you eat and the intake of products affect the organisms as well and the end result and maybe this is the business end of thinking about microbial community or what their functions and products are which then are affecting the host state and whether it's pushing it closer or away from a threshold for disease so what does this mean in terms of analyzing the microbiome with regard to disease and ideas for intervention well first of all this certainly emphasizes the idea of ecology in the microbiome which I know has been an important theme for this meeting and so it points out that we should be trying to embrace the logic analysis as opposed to individual organisms as a way to think about the microbiome and particularly think about it with regard to disease states we also know that the functionality of the microbiome is going to be as informative or perhaps in some way particularly illustrative or explanatory for how microbial composition is driving disease states and it makes sense mechanistically because microbial products organisms I mean we our problem with diseases aren't their names but our problems with organisms or the benefit that we get from organisms are what they're doing so what does that mean for intervention well at a simple level it's that well you find out what are the pathobioins the deleterious organisms find a way to deplete them or find a way to add back beneficial organisms but I think a more elegant way to think about it is how we could consider targeting functions so not worrying about changing in or out organisms but changing the function of them so if we know that pathway number 27 is a particular biochemical pathway which is beneficial or deleterious then you can imagine that we could eat something that would drive the pathways to change them in a way which is desirable or maybe the way that we take drugs like a coxtoe inhibitor to block that pathway to block the formation of an undesirable product we can imagine treating the microbiome sort of in the way we treat our own physiology which is to block a particular pathway using a drug which is which is desirable or undesirable so this gets me to the radio metaphor this I think is a pretty famous article that was a plomical article that came out in cancer cell about a decade ago which is talking about what's the problem with biologists trying to study systems in disease states and so the metaphor that he used was what would happen if a biologist was presented with a radio and needs to determine what was it why was it wrong and how to fix it and so the way that a biologist would do it is that they would rip it up and take a look at what's inside of it and they'd first see that there's all these components and they'd write all these papers in nature saying that there's red components and blue components and things like that and so you've sort of discovered what the radio is you know it has all these components and then people would say well you know really not that the functions are doing the components are doing something and so then you start building pathways and the way that you do pathways is that you find out that certain components like the MICA like it's called a microphone but it's a MICA and I forget what the acronym was that he used but you find a few components that if one of those is absent then the radio doesn't work and so you start building up a pathway of the radio where you have sound expression coming out so this is again how you do it but you could keep doing this for years and years and then you find out there's other elements of the pathway and there's more science paper and then cell papers and reviews and all of this stuff but you still end up with understanding the radio as a pathway and you still don't know really how the radio is or how to fix it and so the comparison to that is how an engineer would look at a radio and instead consider it as a system where how the things are wired and how they relate one to another and so the way to define an electrical system and what it's doing is by an integrated decided as a system as opposed to components or pathways and so that metaphor made me think of the ecosystem analysis where an ecosystem is a way to think about the microbiome not as a set of components like microbial taxa or even as pathways as was describing while there's these metabolite number 27 is up or down but instead to think about this as an integrated system each one of these is a stable state and you can have different states of the microbiome which are robust and you could bang them around with diet and what have you but they'll stay in that state but what this ecologic state is doing and its products are going to be fairly different from this one and one or the other might be favorable unfavorable depending on your genetics or what have you so the challenges are to define what is that state that probably the best way to think of it is the ecosystem and then the other big challenge is to understand how you can navigate between states what's the interventions that you could do so it's more challenging than just to define the states because the process of getting from this state to this state is going to be tough what are the things to make this state unstable and how do you make sure you come over to this rather than to maybe a third state over here which is even worse than the one that you were in so we need robust knowledge of the ecosystem networks to define what are these stable ecologic states and how to navigate between them we also need to conceive what are the ecosystem problems that we're trying to fix is that a hit and run problem where the way that you need to study is different because you have to look at a particular window of time is it a broken or altered component so the ecosystem is okay but maybe a particular guy is a bad behavior in that system and or is it going to be a wrong ecosystem and so I've given examples of each of those now defining an ecosystem this gets into network biology where we really need to work closely with our bonphematic colleagues to try and understand whether we're seeing and by our clinical colleagues to carefully design studies which allow us to determine if it's cause and effect and if the targets that are best used to define the ecosystem state or to modify is to target the hubs the hubs of a network or the causal nodes those particular organisms, their products one or the other might turn out to be more beneficial more desirable or effective so that would be a big challenge for us going ahead and then we'll of course need to write analytics robust sampling and the pre-analytic platforms and we had a wonderful few of us a bunch of us were together last night talking about what we need to do to really distill reliable ways to make these data so that as everybody's doing them they'll be consistent enough that we'll be able to compare our data one to the next quantitative ecosystems and then components and so what I hope cartographically is we'll end up with something that looks like this rather than the ancient things and I'd finally like to acknowledge the people that have been that have contributed to some of the work that I was associated with and a lot we'll be hearing about a bit more for the rest of the morning so thanks very much Time for one question So today the theme is translation in the microbiome and I think it's a great pleasure for me to introduce Jess Goodman who's the chief scientist of VFDA so it's a very appropriate speaker for the this morning session Jess, Dr. Goodman has been the chief scientist of VFDA since 2009 he's in his prior role from 2003 to 2009 he was actually the director of the VFDA Center for Biologics Evaluation and Research known as CBER so again very appropriate for anybody who works on probiotic for example that's one of your contact there at FDA So Dr. Goodman received his MD from Albert Einstein College of Medicine in New York and he did his residency in fellowship training at the University of Pennsylvania and at UCLA and he's still an active infectious disease physician or practicing active infectious disease physician and he's worked from bed to bed site to really now in his role as a big picture public health science and he's been and he still is a big proponent of scientific excellence I think regulatory research and science based regulation and public health at VFDA so he has pretty strong ties to innovator innovation people and obviously to NIH and kind of cater that relationship between FDA and NIH so Dr. Goodman is going to talk about the microbiome getting to product that benefit people Dr. Goodman