 I'm Zhao from Zhao Tong University in Shanghai, China, and the Rutgers University in the US. So he's going to tell us about community ecology of the human gut microbiota in disease and health. OK. Thank you, Professor Zhou. And also, I would like to thank the organizing community to give me the opportunity to come to join this very interesting conference. This is my second computational biology conference. The first one was in China, 2004 in Kunming. And most of the speakers were Chinese or overseas Chinese professors. But at that time, Terry Hua was there. At that time, I was the only speaker talking about bacterial community. But now we have a whole meeting devoted to microbial community. So this is really wonderful. So I changed my title from a multi-omics approach dissecting the microbial community function to this understanding community ecology of gut microbiota just because I have been inspired by so many talks talking about the gut community. So I would like to share my thoughts with you. I also would like to thank my team, particularly in Shanghai, and also my collaborators and founding agencies. And I have been blessed with a wonderful team in Shanghai. And we know now gut microbiota can contribute to our phenotypes in disease or health just because when they grow in our gut, they produce various bioactive compounds which can get into our bloodstream, circulate, and regulate our genes, impact our immunity, and modulate our metabolism. And more than 100 years ago, Professor Matiniakoff already proposed that toxic compounds produced by gut microbiota may be driving aging and aging related diseases. Chinese medicine also believes that fecal toxicity is driving all kinds of diseases. Now we know neurotoxins, carcinogens, and immunotoxins. So all kinds of toxins have been identified and studied in various systems and have been shown that they can get into our bloodstream either through the enteroheptic circulation or more often through partially impaired gut barrier. But from a microbial ecological perspective, we know gut microbiota is just a microbial ecosystem. And the most important question as a microbial ecologist, I would like to answer is, who does what in the microbial? And if you look at, if you borrow concepts from macroecosystem equality, we know that ecosystems are providing benefits to human society. So each benefit can be considered as an ecosystem service. And so gut microbiota must also provide all kinds of ecosystem service to benefit human host. And many of such benefits can only be provided by gut bacteria, such as production of short-term fatty acids, which we do not encode in our own genome. And we must rely on our gut bacteria for that kind of function. And we know from macroecology, species are not equally important, like in a closed forest, tall trees are the so-called foundation species. They are the species, when they grow to a certain abundance level, they close the system, they create a very unique inside environment, which is structuring the whole ecosystem. So we would like to know whether in a healthy gut microbiota can we identify such foundation species. And we also know that in macroecosystems, species are not independent from each other. Different species form different functional groups, which can be called guild. And the members of each guild, they increase or decrease together, they thrive or decline together. So we would like also to know if in a gut microbiota ecosystem, different bacterial populations or species, they form different functional groups or different guilds. So these are the ecological questions that we would like to ask. But if you look at the human gut, it's just a walking bioreactor. It's a chemostat. And the nutrients we're constantly putting into the system are from two sources, diet and also host. The non-digestible and undigested dietary components that have escaped our digesting absorption will inevitably be used by some kind of bacteria. So this is one source of nutrients for sustaining our gut microbiota. The other source is from the host. From the mucins we secrete, or from the left cells of the column. And also some other secretions, like bioacids. So it's a combination of diet and host derived nutrients, which are sustaining the gut microbiota community. We know that the human body is already very complex. Adding to that complexity, now you have a microbiota, which is also very complex. So how can we tackle this super complexity? And in addition to this very complex system, we also needed to ask the causality question. So there are so many evidence that's published that disease people have a different gut microbiota or other parts of the microbiota from healthy people. But the microbiota changed because of the disease. All the disease happened because of the microbiota change. Taking our egg, this is always a very fundamental question when you study gut microbiota in health and disease. So we argue that even though it's a complex microbiota system, we still to need to follow caucasic populace for identifying the causative agent of a particular infectious disease. But we need to consider the polymicrobial and ecological nature of gut microbiota. So first, we need to do microbiome-wide association study. We should identify all the members which are positively or negatively associated with the disease or with a particular health phenotype. And then we should isolate those associated members either into pure cultural or into defined consortium. And then we should put this into a germ-free animal background and see if we can give the right environmental condition, see if we can reproduce the phenotype. If we can, then we should understand the molecular mechanism. So only after we complete all these cycles of research, maybe we can claim that gut microbiota plays a causative role and which members can be used as both biomarkers and also as drug targets. But before we do that, we should realize that strings are the functional units and the living units, functional units of bacteria. Because by definition, strings within the same name of the bacterial species can share up to 30% of genomic difference. We know that the human and mice only share 10% of genomic consequence difference. So that's why we need to go down to a string level to understand the causality and the ecology. So how can we tackle this complexity? If you look at my post address, it's exactly the same information but written in English and Chinese. In Chinese, you start with the country, the city, the university, the building, the room, the family, and then the person. It's top down. But in English, it's the opposite. You start with the person, the family, the room, all the way up to the country. So English speaking people are bottom up people. Chinese people are top down people. Well, it's not just the language, the culture. The way we do science is also different. So this top down thinking and bottom up thinking, if you look at all the parts, you can understand all the parts of an engine, but you still do not know how the engine works and whether the engine has any problem or not. But when you put all the parts together, you have some new functions emerge. It's so-called emergent functions, like noise, exhaust gas, and the vibration. So to understand a whole, one must study the whole. You should look at the emergent functions, try to understand the complex system. And so in Chinese, the medical doctors, traditional medical doctors, they do exactly the top down approach. So they never divide a human body into different parts, but they measure emergent functions at a whole body level. Like look at your tongue, they'll touch your pulse. And then they have their own principles to do pattern recognition and diagnosis. And then they give you a decoction, they give you acupuncture, they ask you to change your diet. They do everything they can. And then they read the changes of the whole body level emergent functions to decide what to do next. And eventually to help you recover your health. But we know this is all empirical. But in modern day engineering, you can mount thousands of sensors to a running engine. And you can do online monitoring of the health and problem diagnostics of your running engine by focusing on vibration noise and exhaust gas. But can we do this to human body? Yeah. And if you look at urine, fecal matter, and blood, these are the three windows only a live person can have. But so these are the emergent functions. But they contain all kinds of biological information, which can be profiled and quantified and analyzed captured by using all kinds of omics technology. So if you do molecular profiling of molecular level variations of these three windows over time of a cohort of people, either natural progression or disease development or responding to any type of treatment or intervention, then you may be able to identify patterns, signatures in these three windows, which are associated with a particular phenotype you are interested at the whole body level. So how can we do that? One way is to correlate changes of gut microbiota with metabolites in their environment. So this is a proof of principle study we did with Professor Jeremy Nicholson from Imperial College and also several other groups in China. So we collected a urine and fecal sample over a monthly interval from a seven member of four generation Chinese family. And then we analyzed the microbiome variations intra-individually and the intra-individually. We also analyzed metabolites by using NMR-based metabolomics in the urine samples. And then we do OPSDA modeling to do two-way correlation analysis between these two data metrics. So in this proof of principle study, we identified that 10 members of the gut microbiota each showed association with at least one urine metabolite. This particular member, fecalobacterium prosonitia, showed eight associations, six positive and two negative. So this is the indication that if you were a postdoc, I would ask you to pick a bacterium to study. You may want to study this one, because this may have the widest impact directly to human metabolism. So this was highlighted when it was published in early 2008. It was highlighted by Nature Reviews microbiology as a platform technology for finding out who does what in the microbiome. And yesterday, Dr. Nilsen gave a beautiful example of using this approach to identify potentially some k-bacteria impacting human insulin. So I would like to, with all these concepts and methodology, I would like to focus on obesity and type 2 diabetes to give you two examples for understanding the community ecology. We know that from many publications from Professor Jeff Gordon's group and many other groups, gut bacteria may play a very important role in development of metabolic diseases, obesity, or type 2 diabetes. But we still need to identify the key members which are causatively contributing to the disease or to recover of disease. The approach we use is to change the diet of a disease group of people and see if we can change the phenotype. And over time, we collect urine blood and fecal samples at different time points before, during, and after. And then we do molecular profiling of changes in these three windows and correlation analysis. See if we can identify any potentially important members of the gut microbiota which are potentially contributing to the disease or to recovery of health. How can you change the gut microbiota to a healthier structure? Well, we learn from traditional Chinese medicine. Because Chinese medicine has a very long tradition using food as medicine or medicine as food. So in China, you have an officially published list of plants by Ministry of Health of China. And plants in that list can be used as common food by everybody. But they also have been used as medicine by Chinese doctors for thousands of years. So after more than a decade of study on these plants, we realize that there are two kinds of ingredients which are very important in those plants. One kind of are non-digestible but fermentable carbohydrates. Mainly various plant polysaccharides. And the other are various groups of phytochemicals. So they can be used as nutrients of beneficial bacteria and they can also work as a protectant for beneficial bacteria. So we developed about seven or eight years ago, we developed a dietary scheme which can satisfy the, provide a balanced nutrition to human but also provide enough nutrients for the microbiome. So this was featured in this article, science news article. So if you are interested, you could read that. I give you this case study. So N equals one clinical trial. So we focus on this young man, 175 kilograms. BMI nearly 60 and he was on this dietary program for 23 weeks and he lost more than 50 kilograms without exercise. So just this diet for about six months and he lost more than 50 kilograms and he recovered from type two diabetes, hypertension, hyperlipidemia. He was almost hyper everything before but then he recovered after 23 weeks and losing 51.4 kilograms. And he also had a reduced inflammation and reduced lipopolysaccharide biting protein indicating that toxin, indotoxin which can induce inflammation produced by gram-netic obtuse pathogens are decreasing in their load. And so we did a very simple PCR DGGE fingerprinting and over time. So you'll see three bands representing three major populations at the baseline disappeared very quickly and remained almost undetectable through after the clinical trial. We cut out the DNA and identified the DNA sequence to them and identified as members of enterobecta, genus. And these genus contain 10 species. They all can induce sepsis. So they are opportunistic pathogens gram-negative can induce inflammation. And we also did a metatonomic sequencing at the baseline nine week and 23 week after. And we found genes involved in genes involved in synthesis of lipopolysaccharide. Quite a few of them reduce their abundance. So then we did, we isolated several hundred putative colonies of enterobecta genus and we run co-migration against the baseline DGGE fingerprinting. So any colony which can migrate to identical position to at least one of these bands will be kept and characterized. So eventually we got strings, a predominant string enterobecta colloquial B29. So we focus on this particular pathogen. And then we inoculate this pathogen into germ-free mice and give germ-free mice either high-fat diet or normal to a diet. Professor Jeff Gordon group and many other groups already showed that germ-free mice cannot become obese. Even you give them high-fat diet. But after you're colonized with a whole gut or a normal microbiota, and then they can become obese and insular resistant. And when we give this a single pathogen isolated from this obese donor, it can also become obese. And it can develop a lot of visual fat and insular resistance, fatty liver, inflammation and Professor Jeff Gordon's group identified a gene in the gut which is necessary, the expression of this gene is necessary for burning stored fat. But it was shut down by whole gut microbiota and only can induced by hunger. But if the gut microbiota was disbiotic and this gene was turned off very tightly and even you feel hunger, you cannot turn this gene on, you cannot burn stored fat. And they also found that the whole gut microbiota actually can up-regulate genes in the liver for transforming glucose into new fat, ac1, phas and the PIPA gamma. So all these two groups of genes regulated by the gut microbiota as identified by Professor Jeff Gordon's group can be regulated by the single pathogen. This is probably why, because the colonization of this pathogen in the gut prevent the host from burning stored fat but encourage the host synthesizing new fat from glucose. So that's why they can accumulate such excessive amount of fat in the same time period. This is an example, this was published in 2012. And in the past five years, we have been working with Professor Philip Gerhard tried to identify the molecular mechanism. So we already, we did a mutation in LPS synthetic pathway and removed most of the pro-inflammatory activity of the LPS and then the mutant lost all the capacity for reducing obesity, fat liver and everything. So there's no obesity related development if you mutate LPS. We also did, we mutated the ptolecular receptor four and derived the germ free mice. If you give one type, no obesity. So only when you have LPS endotoxin and ptolecular receptor four intact and cross talk with each other, induce inflammation, you have all the phenotypes downstream. So the molecular interaction between the LPS and also ptolecular receptor four, maybe the first molecular event leading to all the downstream phenotypes. So this is an equals one study. We only focus on one single case. But we also did a clinical trial with a genetic form of childhood obesity. It's called the predivary syndrome, PWS. Children with this disease, they had a defect region in a number 15 chromosome from the father's side. They were born with very low muscle tone, couldn't even suck enough milk. So they were very small and undernourished before winning. But after winning, when they start to take solid food, they quickly develop a hypophesia. They are hungry all the time and you cannot satisfy them with any amount of food. And it's very difficult to control their body weight growth. However, when we work with Gondong Women and the Children's Hospital to do clinical trial with mobilo-obese children, we accidentally found that one-third of the children in our program, they were actually PWS patients. They were genetically obese, but they responded very well to the dietary intervention of their gut microbiota. Like this particular boy, 14 years old, 140 kilograms heavy. And he stayed on the program in a hospital for 285 days, reduced to 83.6 kilograms. He continued the intervention at home after 430 days. He was 73 kilograms. So he lost about half of his body weight, only undieted, no exercise, and recovered from other metabolic problems. And so eventually we recruited... It's a genetic problem. They don't grow much. It's a... He didn't help that. Only helped with... No, he's this type of children. They don't grow much. By the age of 14, they already mature. So they don't have the potential to grow that much. But... So they... We recruited 17 children with PWS, and they stayed in the hospital for three months. And then same mobilo-obese children without genetic reason, simple-obese children stayed in the hospital for one month. So at the baseline and the end of each month, thorough medical checkup for medical phenotypes and also collection of urine blood and fecal samples. For one month in the program, they lost about 10% of the unusual body weight. And three months lost about 20%. And they recovered from... So they have improved glucose homeostasis, lipid profile, and liver function. And they also have significantly elevated inflammation. And then lipopolysaccharide, binding protein also significantly reduced. And then we transplanted the baseline microbiota and three months after microbiota from the same person to germ free mice. The baseline microbiota can induce inflammation in the first weeks. And then after that, they start to accumulate more fat. So the baseline microbiota can induce inflammation and can encourage more excessive accumulation of fat. But after intervention microbiota, they do not have that capacity. So this is the indication that even diet itself can improve directly host health, but it also changes the microbiota. And the changes the microbiota may have also partially contributed to their metabolic health improvement. We then did a metabolomics analysis of the fecal water sample. And they showed a significant shift before and after intervention. And mainly you see a lot of carbohydrates increase in the fecal water and many bacterial metabolites decreased. And then we analyzed the short-term fatty acids. They are the products from fermenting protein or fermenting carbohydrates. So we see a relative increase of acetate, but relative decrease of isobutriate and isovalerate. This is the indication that after we changed the diet, bacteria shifted from fermenting protein to get energy to fermenting carbohydrates to get energy. And the side products are changing from detrimental and becoming beneficial. And now we would like to understand the string-level changes of the gut community. So we did a metagenomic sequencing of all the samples, inter-individual samples and the intra-individual samples, because we dramatically changed the diet. So we induce a lot of changes, variations of the members of the gut microbiota. And so we would like to see the detailed structural changes. And then we used the canopy-based algorithm introduced by Professor Dr. Nielsen and yesterday. So we identified a little bit over two million non-redundant microbial genes. And after co-abundance analysis, we got 20,000 co-abundance gene groups. And 376 of these gene groups containing more than 700 genes, so they are each is potentially a bacterial genome. 161 of these are shared by more than 20% of the samples. So they are potentially the chromosomes of prevalent genomes. And then we did assembly of the genome based on each keg and we get 118 high-quality draft genomes. So they meet at least five of the six criteria for the American Human Microbome Project reference genomes. So the quality is almost as high as when you do pure cultural sequencing. So now we would like to see the co-abundance relationship between these genomes to understand their ecological relationship. So out of 161 co-abundance gene groups, genomes, we can delineate them into 18 potentially guilds. Each colored group is potentially a guild because they increase or decrease together. But if they are connected with red line, they co-occur. If they are connected with blue line, they could exclude. Our dietary scheme significantly promoted this group of bacteria. They increase their abundance dramatically after the new diet. And the three members of this group are from Bifidobacterium genus. And particularly Bifidobacterium pseudocantinolatum, these particular species had the highest number of negative correlation with the members of many other guilds. And those negatively associated are potentially pathogens or producing some detrimental compounds by looking at their genome. And so we hypothesize that our dietary scheme probably promoted this group of bacteria as the foundation species. And when they grow to a high abundance level, they produce a lot of things like they acidify the gut and they also produce many antimicrobials so that the gut environment becomes favorable to beneficial bacteria, but unfavorable to many detrimental bacteria. So this actually restructured the whole gut microbiota. If we do guild level abundance analysis, you'll start to see correlation of different guilds with individual human phenotypes. So three guilds showed a negative correlation with disease phenotypes. And nine guilds showed a positive correlation with disease phenotypes. And six guilds didn't respond to this dietary intervention. So now the major, all these 161 high quality draft genomes, they can, they amount to more than half of the total sequence we get. So that means they are prevalent and also dominant members of the gut ecosystem. So now they are organized into this skilled ecological structure. And blue-arrowed ones are potentially beneficial and the red-arrowed ones are potentially detrimental and the others may be neutral. So I would like to comment to that the currently many data analysis in a microbiome field is the so-called taxon-based analysis. So you do a genus level, family level, all the way up to final level analysis between disease and health. But this is a problem. Why? Because bacteria, they form functional group, they work together not based on their taxonomy. So if you look at the members of the guild, some guild they have members from four different phyla. Some have members from one phyla. So they work together based on the function, not based on their taxonomy. So if you do taxon-based analysis, you introduce a lot of noise. And also, even in the same species, like in fecalibacteria prosnitia, we assemble the nine genomes, but they are in four different guild. So that means members of the same species, they don't behave the same. And so we need to go down to a string level. We look at this particular boy. In the first 105 days, we had multiple time points sampling. If you look at the genus level, you'll see a lot of variations. But there is a steady increase of bithidogenes. But if you look at the species level, nine species identified by the only pseudocannulatum actually showed a substantial increase. And then we isolated five strings from these species, from one single sample from this person and five strings. And we did the finished sequencing. You can see the five strings, they share 1,520 genes in their core genome. And they don't have much difference gene number wise. However, during the first 105 days, they showed a different response to the same dietary intervention. So this is a really a string level function. And one of the strings can actually be used as probiotics to alleviate high-fat diet induced obesity in mice. And how much time I have? Okay, so I would probably just quickly go over the... So we also analyzed the u-ray metabolites and we also found significant shift before and after. And then we identified the significantly changed the metabolites. So 14 metabolites significantly changed. So five of them actually promoted and the others reduced. Among the nine reduced metabolites by the dietary intervention, four actually are co-mortabolites by the host and also by the gut bacteria. So what happened was the gut bacteria ferment lipid or ferment protein into potentially toxic metabolites. And then these will get into the liver and the host liver will further modify this and adding sulfate or other groups and they become water soluble and they coming out of the u-ray. And TMAO has been known to induce asroscrosis and significantly reduced. But we don't know which bacteria actually convert calling in a diet to TMA. And then TMA get into the liver and become oxidizing to TMAO. And then TMAO can promote asroscrosis. So we don't know which bacteria are potentially doing this conversion from calling dietary calling to TMA. And then if you take the 161 high quality draft genome and this 13 metabolites changed, you do your correlation. You found among 118 high quality draft genomes, 31 show the positive correlation with the urine concentration of TMAO. Among these 13 actually have the two genes required for converting calling into TMA. So these 13 bacterial genomes are potentially the potential key bacteria for promoting or contributing to asroscrosis. And we are now isolating the bacteria and try to demonstrate the mechanism in germ-free mice. So it looks like that when you change the available nutrients to the gut microbiota, so you change the resource and there is a guild level response. So it's not a random individual response. You'll see a guild level response to the resource change. And you can actually dissect and establish that structure. And the most important thing is the foundation guild and also guild which are potentially beneficial or potentially detrimental. It's possible to identify them and for forming new hypothesis and to do further molecular study. So I would like to finish by emphasizing that foundation species is different from case stone species. So case stone species is this piece of stone which is not necessarily the major member but it's very critical for holding the whole structure. But the foundation stone is the piece of stone underneath the whole structure. So I think the potential species we identified they are not important for one particular guild. They are actually important for holding the whole healthy structure. So I think the most important impact of a microbound study to human health is that we have a new window to assess and monitor human health. Anything you do to human body to either to a particular individual or to a large number of people, you can just assess the response of the host by looking at how the gut microbiota respond and how they are metabolized and also other antigens which can interact with the host also changed and connected that with eventual organism level phenotype changes of the host. So this is the most important way a new methodology for assessing and monitoring human health. And so this can be applied to a large scale of people and also can be used to understand either a complex intervention like Chinese medicine or a complex intervention like a new diet. You can understand the molecular level response. So I would like to finish with this theoretical model. Genetically, we may live up to 150 years. That's our genetic potential. And so if you analyze the molecular level changes in urine, blood and the fecal and also other medical phenotypes and then you use all the data to quantify the health at any given time point. And if you do that throughout the lifespan, you may get a trajectory like this. So this is the ideal trajectory. You live healthy throughout your 150 years and you die in the last week. Right? And but unfortunately individuals born with genetic defect and die prematurely. But most people have a trajectory like this. Roughly 50 years old is a turning point. Before that we are okay. After that we are going down the hill and eventually end up in a hospital with one or several chronic disease. And modern day technology can keep a person alive for many years even though he was paralyzed. But if this decline of health after mid-age was due to our genetic defect, we still don't know what we can do. But if genetic defect only increase the risk but the actual manifestation of the disease requires environmental triggers. And the toxins produced by gut bacteria may be the most important triggers. And then we have hope. Because we know gut microbial community is plastic. And we can change the community back to a healthy status and monitor that to make sure that you remain in this region until the last day. So eat right, keep fit, live long, die quick. Okay.