I want to, first of all, say thank you to the organizers for inviting me today. Oh, sorry, I'm a little short. Is that better? Everybody can hear me better now? Okay. I want to thank the organizers, first of all, for inviting me today to talk about this topic. I was joking with my students that I was going to put a slide with the emojis for how I felt about coming to ASLD and talking about this topic. So I was very excited to receive the talk, and then I looked at the talk title and I panicked because this isn't my primary interest of thinking about how to predict efficacy of weight loss strategies. Then I was very stressed as I went through the literature and realized that we don't quite have an answer yet. So I'm going to say that first and foremost. But then I had a few aha moments, and then I realized what I wanted to tell you about, about how I think we can get to those answers for how we can use the microbiota to predict who and who will not respond to various interventions. So I don't have anything to disclose at this time. So this slide isn't going to be any news to all of you in the audience, but the way we think about the increased incidence of specific disorders in the lab or what we call new age disorders has really increased over the past half century. So these are just maps of the United States outlining the increasing prevalence and incidence of self-reported obesity defined as a BMI greater than or equal to 30 kgs per meter squared across the United States. You can see that there really has been a march as far as increased incidence. And this goes right along with other metabolic diseases that go hand-in-hand with obesity, including type 2 diabetes. But what we know, and when things get worse, is when CDC recolors these maps to show that the incidence has severely increased greater than 35 percent in various states across the U.S. And now this really does go hand-in-hand with non-alcoholic betty liver disease prevalence as well. So this is just showing across the world really the prevalence as a percent of the population of NAFLD within specific regions on top of the percentage of the population that can carry risk alleles, specifically for like phospholipase domain-containing 3 or PNPLA3 genotypes across the world. So you can see that incidence is growing and expanding as far as NAFLD is concerned across the world as well. So we know that the development of obesity is multifactorial and complex, and there's many components that come into play that work perhaps through genetic susceptibility to drive the onset of diseases like obesity. And really what we've been focusing on mainly in our lab is looking at gut microbiota, nutritional status, and my day job is actually looking at how circadian rhythms work through these. But today we're going to talk about how we can use the gut microbiota, perhaps mediated through diet, to predict how people will respond to treatment. So when we talk about things that can influence the gut microbiota and could be contributing to this onset of disease, we know that our diets have changed dramatically. So we really have gone from maybe what you might consider to be this hunter-gatherer type population to now we have readily available, high caloric density, highly processed foods that are high in fat content as well as simple sugars with low fiber intake and sedentary lifestyle on top of that. So I like this next slide just as a demonstration in a human population, different human populations, of how diet can really have an impact on the composition of the gut microbiota. So this is a little bit of an older study now at this point, and this is looking at 16S ribosomal RNA sequencing to look at the profile of the community members within two groups of children, those raised in Burkina Faso, Africa, as compared to children raised in the EU. And you can see just from the colors on the pie chart here that there's a dramatically different composition in the gut bacteria where the children from Burkina Faso, Africa, have a large proportion of bacteroides within their communities as compared to the children in the EU that have this expansion of firmicutes. And when they went back to look at factors that could be mediating or contributing and associating with these differences, what they really saw was that diet was driving these differences. And so when you just look at broad categories of the dietary intake, including K-cals that were taken in per day as well as macronutrient content, you can see that it differs fairly dramatically between these two groups of children where the children in the EU are taking in a more calorically dense diet, as you could imagine, based on the previous slide I showed you. And then the macronutrient content is quite different where children in the EU are taking in quite a bit more fat as well as carbohydrate. And I would argue that the carbohydrate intake between these two groups of children is dramatically different with far less complex carbohydrate being taken in by the children in the EU. Now Jeff Gordon and company have done a great job in demonstrating how the gut microbiota can be causal in the development of obesity. And others have now confirmed this across the world in that germ-free mice raised in the complete absence of gut microbiota, whether or not they're exposed to a low-fat or a high-fat diet, remain lean no matter what. However, when you compare that to their conventionally raised counterparts, as we all know, these animals that contain a full community of gut microbiota are exposed to high-fat diet. They develop severe obesity. And what many have shown is that if you take this gut microbiota from the fat, an obese mouse, and transplant it into germ-free animals, they also develop obesity and take on the phenotypic characteristics of the donor. So how do we assess the gut microbiota? And I think this is really key when we're starting to think about how we want to use it as a predictor for how individuals will respond to an intervention. And first of all, I want to say that context is key. So a lot of what we read in the literature these days is there was increased levels of diversity. There was increased richness. There was differences in taxonomic composition. But we really have to think about that within a specific disease state. So what I'm going to tell you about as far as maybe metrics that we can use in the context of obesity are distinct to obesity and shouldn't necessarily be translated across all disease phenotypes. So just because we see an increase in diversity doesn't always mean that that's good. And so I just want everybody to kind of keep that in mind because I think that can oftentimes get lost in the details. So context is key here. So diversity. What are we talking about when we're talking about diversity? Well, it's richness. So the observed features within a specific sample. So how many unique sequences do we have within a specific sample? And then the evenness and distribution of the features that we observe. So the equitability and the proportion that we see. Then we can also talk about the actual taxonomic distribution between two environments. So here we have people compared with a normal BMI to those with an obese BMI, and now we can look at the distribution of the different microbiota in each case. But then we can move on from just a fingerprint to looking at the genetic potential. So this would be more metagenomic-type sequencing analyses where we can sort of get a better understanding of the functional context that's going on in regards to the community members that we observe within a specific environment. And then we can consider the functional output. So this is moving beyond just the genomic potential and looking from, let's say, metabolomics perspective or proteomics or lipidomics perspective. And all of these are very important, and I'm going to argue that I don't think we should use a single metric. We should try to incorporate many different ones when we're trying to think about prediction. So with these qualities in mind, what do we observe between lean versus obese gut microbiota qualities? We know that in lean individuals who are typically taking in a healthier type of diet, perhaps high-fiber, low-fat, high-protein diet, that they oftentimes have what we would call a eubiosis. And in this case, we're going to say that it's actually high diversity. And this is compared to obese individuals who are perhaps taking in a high-fat diet, simple carbohydrate, low-fiber diet. And in this case, they have a dysbiosis within the microbiota, and this is considered to be low diversity. We can look at specific members, and those with eubiosis in lean individuals have been shown to have increased capacity for fermentation. So many of these organisms and bacteria listed here have increased fermentative capacity, so Prevotella, Christian, Sonella, Bifidobacteria, and then something that we're going to talk a little bit about later, Acromancea mucinophila. Whereas, as we just heard before, in a dysbiotic scenario, we can oftentimes see an expansion of bile-tolerant organisms, such as Bilophila allostipes and Bacteroides. We see an increase in short-chain fatty acids in eubiotic states and decreased levels of trimethylamine, and others have shown that TMA can actually contribute to atherosclerosis. And then in a dysbiotic state, we have decreased levels of short-chain fatty acids and, as I just mentioned, increased levels of TMA. So with this knowledge, we can then start to think about how the gut microbiota may actually begin to serve as a mediator as a response to treatment options for obesity. So this is probably something many of you as clinicians work through on a daily basis when you're talking with your obese subjects that are trying to lose weight. And depending upon the amount of weight that they need to lose, you will assign either lifestyle changes for those who fail that, perhaps pharmacological agents, and then for those that fail that or have morbid obesity, you would prescribe, perhaps, surgical interventions. And we know that the gut microbiota are mediated by almost all of these things listed within the lists in either one of these specific treatment options. But we know that people oftentimes fail all of these therapies. So they'll fail lifestyle changes, they'll fail pharmacological interventions, and some will even fail surgical interventions. And the question becomes, why? What's so different about these individuals? We know that genetic susceptibility is very important, but could it be that they have a dysbiosis to begin with that is unable to literally respond to the specific intervention that's prescribed? And then on top of this, we can think about the microbiome-based therapies that have very mixed outcomes in the literature, specifically probiotics, prebiotics, symbiotics, and now fecal microbiota transplant is now being used in specific studies to determine whether or not it can be used as a therapy for obesity. But again, many individuals will fail these microbiome-based interventions, and the question is, is it because of the microbiome that's already in existence and we can't penetrate this microbiome to change it in any important way for the host? And I think this is important to think about as to how the gut microbiota could be sitting in the middle, really, of these interventions that we're prescribing. So is the gut microbiota itself serving as the mediator? So when we're talking about delivering prebiotics, we're really talking about how do we change the gut microbiota community to translate into a specific outcome for the host, such as weight loss. But on the other side of that, is, in fact, the microbiota serving as an effect modifier of our given intervention? So we know that the gut microbiota itself has the capacity to perform xenobiotic metabolism, and as we heard just previously, can even modify bile acids within the gut. So based off of the intervention that we're giving, for example, a drug, is, in fact, the gut microbiota acting on this drug to change its capacity to then influence the host and the desired outcome that we're looking for, and in this case, weight loss? So with this knowledge, how can we then go on to identify and predict who will respond, and can we identify responders versus non-responders? So I really liked this specific review from Dr. Patrice Connie's group, who's done a lot of work looking at prebiotic interventions and examining acromancia muciniphila in the context of obesity and metabolic diseases. So when we're in the clinic and we're thinking about the individuals that we're treating, we really do want to identify responders versus non-responders, and this cross-cuts many disease states, from obesity to NAFLD to cancer drugs as well. And what we really need to have is information in regards to their baseline. And so when we're talking about baseline, we're not only talking about the microbiome, but also all of the environmental, clinical, and even omics technologies that we can apply to these individuals. And oftentimes what we'll see is that we can stratify individuals just from the get-go based off of baseline. So in this case, we have individuals with a status X versus status Y based on all of the metadata that's collected for the individuals, and when we look at the gut microbiota that's present within these individuals, we note that one group has low diversity while the other has a high diversity. So now we can think about when we put in an intervention or give an intervention, how are these two different groups of individuals going to respond? So let's induce a weight loss intervention. Perhaps it's caloric restriction. Say it's a high-protein diet, for example. And now on the backside of this, we have a group of non-responders and a group of responders. And again, those with the low diversity microbiota from the beginning did not and were unable to respond to the given intervention, whereas responders had improved outcomes. So all of the environmental, microbiome, clinical, all of the omics stay relatively the same for the non-responder because nothing really has changed. However, the responder now has a different set of metadata associated with it that can hopefully then be used to feedback to examine what was going on at baseline and what shifted in order to cause response in that individual. Then hopefully from this information we can identify mechanisms, targets, and strategies that we can use to then target the non-responder profile, whether it's any one of these factors, particularly the microbiome, to sort of force that individual then to be a responder. And then we can also up front then identify these features for response optimization to make a non-responder become a responder. So this is kind of the framework I'm thinking about this in. So using the microbiome as a predictive tool, the questions and the challenges. So can the host responses to an intervention be predicted by coupling pre-intervention health status and microbial composition data? What information and data is necessary in order to do so? What types of studies and technologies are needed? And if we can predict in a training population, say in a clinical trial, and these settings can be achieved, then how do we scale up to a broader population so we know that gut microbiota, for instance, is so unique per individual? So how can we take our training set and really expand it out to a broader population? So next I'm gonna walk you through a few studies that have been done, where I think they have incorporated the baseline data and identified responders and non-responders and tried to identify features of the microbiome that could aid in prediction. So all of these, and I wasn't really sure for one of them whether or not it was actually the same data set used by the other two studies, but ultimately this was the study design for the first study. So it was a non-randomized design. It was somewhat small subjects, 49, eight males, 41 females, that were considered obese or overweight. The intervention that was given was a six-week energy-restricted high-protein diet, and then they were placed on a six-week weight-maintenance diet. And in this case, the microbiome measurement that they were using was actually fairly simple and not very in-depth, but it was quantitative PCR for seven dominant bacterial groups listed here. And then after they had this data, they tried to perform graphical Bayesian network framework to really identify predictive factors that would either drive weight loss or regain after individuals were placed back on the weight-maintenance diet. And so this is the outcome that they observed as far as weight loss and regain in the individuals from baseline through the 12-week period. And so you can see that what they were able to do in their data set is stratify it into three categories of high weight loss as compared to relatively low to no weight loss within these individuals, and start to look back at the data to see if anything could predict this specific outcome. And what they observed, that obviously cluster C had the worst outcome and worst weight loss, and at the onset at baseline exhibited the highest levels of this specific group of gut bacteria that includes lactobacillus. From what we just heard about earlier, maybe this could in fact be playing a causative role in this situation. And what they noted was that the abundance of these specific microbes at baseline correlated with weight regain and consumption of a starch-containing diet. However, when they performed the Bayesian network analysis, unfortunately this difference that was observed at baseline was kicked out, and it wasn't actually identified as a predictor in this model. And it could have been due to the small sample size, or as I mentioned, this isn't a very in-depth analysis of the gut microbiota. Had they used something like 16S or metagenomics, they may have been able to identify and use this as a predictor for outcomes. So in a very similar study design, where we had this non-randomized clinical trial, again, the intervention was a six-week energy-restricted, high-protein diet, low glycemic index, followed by the six-week maintenance diet. And the microbiome measurement here was actually metagenomics, so what they call quantitative metagenomics, where we're really trying to get after the genetic potential of the community. When they looked at the individuals based off of gene content within the community, they noted that there were two, there was a bimodal distribution of the gene number observed within the individuals. So they noted that there were 18 individuals with low gene content, and 27 individuals with high gene content. And this is essentially how they stratified the population moving forward. So when they looked at outcomes from the intervention, between the low-GC and HGC, high-GC individuals, you could see that, as far as disease index was concerned, there was a difference at baseline at week zero between low-GC and high-GC-containing individuals within the microbiota. And there was a reduction in the disease index across all the groups, although there was no significance between the two at the end of the study. And when we're talking about the HOMA-IR, there was really only an improvement observed within the high-GC-content individuals, and not observed in the low-GC-content individuals. And this was also the case when they considered triglycerides. And ultimately, as far as inflammation is concerned, really only the high-GC-content individuals responded and had reduced inflammation. So what this group did was they were looking at the gene richness, so all of the unique genes that were observed in the high-GC versus low-GC individuals. And you can see that the microbial richness was fairly low, as expected, in the low-GC individuals. However, following the intervention, there was an increase, although these individuals were never able to recover completely to what was observed in the high-GC individuals following the intervention. And then there was some loss following the maintenance diet. So when they did a gene clustering analysis, they revealed that 26 out of 39 clustered varied over time, so they really focused in on these 39 clusters that varied across the timing of the interventions. And so again, based off of just these non-discriminant titles for clusters, you can see that the low-GC individuals started with lower levels of these genes as compared to the high-GC individuals. There was some regain in certain clusters following the intervention. And in certain cases, it was actually maintained as well. So perhaps we were, or they were actually able to change the gut microbiota profile and that contributed to the outcome. However, as I mentioned, it was really only the high-GC group that exhibited a marked reduction in adipose tissue and systemic inflammation. So perhaps the baseline microbiota and having low-GC actually prevented these individuals from responding. And had there been some change initiated at baseline in the microbial community, we perhaps could have changed a non-responder to a responder. So using the same study design, a follow-up was performed to examine whether the presence of a single species could impact response to the diet intervention. And so again, same study design as I mentioned before. And again, the intervention was the calorie-restricted diet, six-week maintenance diet. The microbiome measurement was again quantitative metagenomics. Genetic potential was examined. And then quantitative PCR for acromantia muciniphila, which as I mentioned is correlated with a lean individual gut microbiota over the intervention time. And so this is what they noticed was that they could stratify once again their populations at baseline by level of acromantia observed within the community at time point zero. So here in the black line, we have the acromantia high individuals and here in the gray line, the acromantia low individuals. And what was interesting to me about this was that actually following the intervention, there was actually a decrease in the relative abundance or actually the quantitative abundance of acromantia within the community following intervention. So this begs the question is, is this important? Is acromantia actually important in the community in order for an individual to respond, which I'll show you in the next slide. But you could see that those with low acromantia to begin with following intervention didn't necessarily increase and always stayed lower than the high acromantia containing group. So this is what the disease outcomes looked like for this specific analysis. So those individuals that had high acromantia from baseline actually had improved disease index from baseline. And those also that had low acromantia also exhibited improvement as well, although not nearly as great as what was observed in the high acromantia containing individuals. Waist circumference was decreased in both groups. Again, perhaps higher in the high acromantia containing group. Total cholesterol was reduced following the six week intervention, but then it rebounded back to levels observed at baseline. And similar features were observed with LDL. So what I did really like about this study was that they actually performed some metabolomic type measurements within the serum of the individual. So they looked at serum levels of short-chain fatty acids and they noted that the serum levels of the short-chain fatty acid acetate actually correlated with acromantium acinophila at baseline. So here in the Spearman correlation plot you can see that when acromantia is at higher levels, the levels of serum acetate are also at higher levels. But again, interestingly enough, from what they observed in regards to decrease in abundance of acromantia from baseline, they observed a similar feature in regards to serum acetate in that time point zero levels were relatively high in the high acromantia group and they diminished over time following the intervention. So the outcomes in regards to acromantia mucinophila abundance in dietary intervention, higher baseline of acromantia mucinophila actually associated with better health outcomes from the specific dietary intervention. Higher abundance of acromantia mucinophila associated with more microbial genomic richness and baseline levels of the serum metabolites such as acetate correlated with increased levels of acromantia mucinophila. So perhaps we could use acromantia at least as a biomarker of who might better respond to specific outcomes and interventions. So conclusions from this, can we use the microbiota to predict outcomes to interventions? I don't think we're quite there yet based off of what I saw in the literature, but there's a lot of people trying and I think we have to keep trying. And I really think that the research is improving as technology improves and the cost of sequencing decreases. So 16S rRNA might not and probably isn't the best marker to be using to identify predictors, but metagenomics could give us some further information about the gene content and richness examined or observed within a specific individual. And if we can identify responders versus non-responders at baseline, can we use that information to modulate their microbiome prior to our intervention to promote a positive outcome? So for these individuals that have a dysbiosis, low diversity and richness of the microbiota, can we expose to a probiotic or prebiotic to change their microbiota at the beginning, which could promote better outcomes on the backside? So how do we move the needle in this direction? Well I think we need more longitudinal and interventional crossover study designs, which can help us to identify these responders versus non-responders. And we really need deeply phenotyped individuals. So I know many of you who have and run fatty liver clinics, you're getting a lot of phenotypic data on each individual. So how can you incorporate microbiome analyses meaningfully in your clinical practice, really? I think we need studies that have diet interventions with a baseline, because we need individuals really to serve as their own controls. And that's gonna allow us to broaden out to a bigger population. Even with drug trials, I think that we have to include microbiome as a measurement. Because as I said, microbiome can serve as an effect modifier. And so we need to be able to better predict how microbes are gonna interact with certain drugs. And especially for FMT studies for obesity. So far there's just been a lack of consistency across centers for outcomes. We're definitely trying, and I think that we're learning more. But we need to have a better understanding at baseline and increase the numbers of people receiving FMT. Because perhaps we need to do something at the front end for individuals, like treat with specific antibiotics to allow for that FMT to take hold. Which then could promote improvement of the intervention in the long run. And then I think we need to increase the microbiome-based endpoints in each study. So metagenomics coupled with metabolomics and coupling specific technologies. So there's been some nice work done where they've coupled MALDI-TOF even with 16S or whole genome sequencing to identify specific mediators, microbial mediators with a community for necrotizing enterocolitis, for example. So could we use those same tools and techniques in the context of obesity to promote better outcomes on the back end? And I think really in order to increase this predictive capacity, we have to go from the bedside to the bench and back to the bedside. And so this was a figure that I put together for a review we put into cell host and microbe. Where the gut microbiota kind of, as I said, sits in the middle of many of these aspects of cause. But they can be association and we need to tease apart this contribution. And we need to start probably with the human subject, right? What's the important question to ask? But use these tools that we have, these preclinical models, germ-free mice. Can we conventionalize germ-free mice, humanize them with responder versus non-responder microbiota, perform interventions in the animals, determine who responds, who doesn't, and then take that information back to the clinic. And then again, we can examine specific community members by using in vitro tools like intestinal organoids or culture systems. And again, these can inform each other so that ultimately we can use them as predictive tools to determine who and who will not respond. So thank you for listening to my stump speech there. I have to acknowledge a number of people. I've been doing some non-alcoholic fatty liver disease work in a mouse model with my collaborators at U of C, Eugene Chang and Dr. Charlton. I have a great team that I work with. They've just done some tremendous work in our mouse model. I'm gonna, of course, thank my funding sources. I'm supported by a K Award through NIDDK. And I'm going to give a shameless plug for you all to go visit a poster on Monday. This is a medical school student who worked with me two summers ago on our mouse model and he's presenting a poster. So please go hackle him for me because I won't be here on Monday to do it. And with that, I want to say thank you and it sounds like we'll take questions at the end. Thank you.

gratitude

weight loss strategies

microbiota

obesity

dietary changes

microbiome analysis

personalized interventions