Good morning. It's so wonderful to see so many of here today. To hear our Leon Schiff lecturer this morning, who is Dr. Jacqueline Marr. She is a professor of medicine and gastroenterology at the University of California, San Francisco, where she serves as chief of the gastroenterology division at the Zuckerberg San Francisco General Hospital, program director of the NIH-funded UCSF Liver Center, and co-director of the T32 postdoctoral training program in hepatology. Jackie, as she's affectionately known, obtained her medical degree from Duke University and remained at Duke for a residency in internal medicine. After that, she moved to UCSF, where they were smart enough to retain her, and she quickly rose through academic ranks to the leader that she is. She has devoted her career to basic studies of disease mechanisms underlying alcoholic and non-alcoholic steatohepatitis. She's an active member of this society, having served formerly as chair of the basic research committee and having completed a three-year term on the governing board as secretary. Dr. Marr has extensive editorial experience, having been on the NAE for American Journal of Physiology, Gastroenterology, and Hepatology. She has served on numerous grant and abstract review panels for the NIH, the AGA, and the AASLD. It's no surprise that she's further recognized for her fairness, for her expertise, and her stature as a member of the NIDDK Advisory Council. Welcome, and we look forward to your lecture, The Evolution of Disease Modeling in Novelty. So, thank you, Harmeet. And I also want to thank the AASLD leadership and the members of the program committee for giving me the opportunity to honor the legacy of Leon Schiff. Oops, I'm already not doing my slides correctly here. How do I do this? Left button? Back? I need to go back. I can't see that. Thank you. I'm sorry, my eyes are not good. I think I need to go back even further. The slide that I'm missing is Leon Schiff's photograph, so I'm so sorry about that. So before I begin, I wanted to take just a moment to honor the gentleman for whom this lectureship is named, Leon Schiff. Dr. Schiff was one of six founding members of the organization that we now know as AASLD, and was its first president in 1950. He was an immigrant from Latvia, moved with his family to the United States at the age of five, and settled in Cincinnati. He went through all of his education in Cincinnati, including undergraduate, graduate, medical school, and residency, and ultimately joined the faculty at the University of Cincinnati Medical Center in 1930. There he stayed until 1970, where he was chief of the service of gastroenterology, and actually has the distinction of having founded the GI division at that institution. In 1970, he moved to Miami to be with his son, Eugene, at the Schiff Center for Liver Diseases at the University of Miami. Many of you may know Dr. Schiff best for his textbook of hepatology, Schiff's Textbook of Hepatology, which had its first edition in 1956, and is now in its 12th edition. So now onto the topic at hand, the evolution of disease modeling in NAFL. Here are my disclosures, and I'd like to acknowledge financial support from the following institutions in current and years past. So what am I going to do today? I'd like to discuss with you one individual's perspective on my experience as an observer, and sometimes a participant in fatty liver disease research that's occurred over the last now 30 years. It's impossible for me to be able to cover all of that and do justice. And so what I've decided to do is to spend some time on the nuts and bolts of models of fatty liver disease, but spend additional time really telling you about some of the observations that I believe have been seminal in actually founding our knowledge about the pathophysiology of this disease. Those of you who are looking astutely can see that the buttons end in 2016. That's just to remind me that most of the research that occurred between 1994 and 2016 actually was conducted primarily in rodents, but on or around 2016, there was a shift towards the advancement of human models of fatty liver disease as well, and so I'm going to touch on that at the end of my talk. And I will apologize in advance to anyone who is a zebrafish biologist, because I am not going to discuss zebrafish models of fatty liver disease today. So one of the first models of fatty liver disease that actually attracted the attention of the hepatology community was the obese mouse. Now, obese mice have been around since 1949, but the gene that's responsible for the defect, leptin, wasn't cloned until 1994. And as you're aware, OBOB mice are deficient in leptin. These mice develop obesity and fatty liver because they're hyperphagic, they're normal at birth, but they quickly become obese, diabetic, insulin-resistant, and have marked hepatic steatosis, as you can see in this slide. Regrettably, though, it was recognized relatively early that despite the pronounced steatosis that these mice have, they're actually relatively weak fibrosers. And further experimentation with this model actually uncovered the reason for this defect. As it happens, hepatic stellate cells have receptors for leptin, and they respond to leptin by proliferating and activating. So in the absence of leptin, there is a reduced fibrogenic response in the liver. And so for this unfortunate defect, the OBOB mice did not develop a strong foothold as models of liver disease in the community. That said, there are a whole host of either normal or rather spontaneously mutant or genetically engineered mutant mice that exhibit either spontaneous hepatic steatosis or a predisposition to hepatic steatosis with challenge. I'm not going to talk about these individually, but I thought that it would be useful for you to see them divided into groups based upon their defining features. On the left are mice that actually develop steatosis because they overeat. In the middle are animals that develop hepatic steatosis because they have impaired hepatic lipid secretion. And on the right are models that developed hepatic steatosis because of altered lipid metabolism, whether it be defects in hepatic lipogenesis, fatty acid uptake, or fatty acid oxidation. So as valuable as these models have been in defining the contribution of these specific molecules to a hepatic steatotic phenotype, what has still been missing from this equation and is an unmet need is for a model that can be reproducible in a wild-type mouse. And for this, research has turned to diets. So what about diets? One of the earliest diets that came to the attention of hepatologists was the methionine choline-deficient diet, or MCD diet. First publications with this diet appeared in about 1996. And as you can see on this slide, MCD-fed mice developed all of the trappings of human NASH, including steatosis, hepatic injury, inflammation, and with enough time, fibrosis. Another advantage of these animals was that the disease was relatively rapid in onset. The histology that you see here on the H&E actually occurred within three weeks of beginning the diet. However, like the Oboe mice, this mouse had a major disadvantage, which was despite its histologic replication of fatty liver disease, the mice didn't display any of the metabolic features of human NAFLD. In fact, they didn't have insulin resistance, and nor did they gain weight. In fact, they lost weight, as you can see in the blue line on this slide. So what to do? This diet, actually, our group was a strong proponent of this diet in the beginning, and I was quite reluctant at the time to just drop it without trying to make some kind of sort of lemonade out of lemons before letting it go. And so we actually did experiments capitalizing on the pathophysiology of the MCD mouse in order to learn something about nutrition. I'm not going to talk about those studies today, but I'd be happy to discuss them with anyone offline. What I will do instead is to talk again about, and these are not the right slides. What I'd like to do now is to actually show you a table of diets that have been the evolution, if you will, following the abandonment of the MCD diet. First, there are vestiges of the MCD diet, one being above the red line. So I'm going to draw your attention to the CDA-HFD diet, which is a choline-deficient diet still, but the methionine's been added back. And so this is a kinder, gentler, if you will, methionine-choline-deficient diet that can allow for longer periods of feeding. It allows for fibrosis and inflammation without the weight loss that occurred in the MCD animal. Unfortunately, though, these animals don't develop strong obesity or insulin-resistant signals. They do gain weight, but not as much weight as you would expect in other high-energy diet models. On the bottom, what you see is a group of diets that are more what I'll call conventional diets. These are all high-energy that have been used to model fatty liver disease. And I'm going to draw your attention to the four at the bottom, because the Western diet is sort of the granddaddy of them all. The four at the bottom, as you can see, are almost regressing to the mean, if you will, and they all have features in common that lead to the development of the pathophysiology that's desired. And I'm going to, in the next few seconds, tell you about what these defining features are. First, they all have about equal amounts of carbohydrate and fat. They're all about 40-40. This is good because it's what humans eat, but it's also good because you need a good amount of carbohydrate in order to stoke de novo lipogenesis, which is an important driver of the disease. Second, they're all supplemented or incorporated with additional fructose, either in the diet or in the drinking water. And this is yet a second stimulus to de novo lipogenesis, which will enhance fatty liver and enhance the potential injury response. Third, they all include cholesterol. And this is the additive that I'm not particularly fond of because 2% cholesterol is much higher than a human would ever eat, but it does seem to be the secret sauce that's quite important for the development of really strong fibrotic response in this animal model. And there may be an explanation for this demonstrated by the laboratory out of Columbia University that cholesterol has the ability to stabilize TAS and hepatocytes, which can promote hepatocyte injury and hepatic fibrosis. So what I'd like to do now is to switch gears for a moment and talk to you about some of these experiments, manipulations of models, if you will, that in my view have really yielded seminal observations that have changed the paradigm of how we think about the pathophysiology of fatty liver disease. And the first one has to do with mouse models that shed light on adipose tissue liver interactions. So in 2003, it was discovered that mice with obesity have macrophage infiltration into their adipose tissue, as you can see in brown on the slides on the right. The macrophage infiltration is associated with adipocyte enlargement, which is in turn associated with a suppression of adiponectin secretion. So in response to this, there was a theory proposed that obesity would lead to adipocyte enlargement and suppression of adiponectin, which would lead to macrophage infiltration, which could lead to lipolysis, which could in turn lead to fatty liver. But the question is how to demonstrate that cause and effect. And so here's where a targeted manipulation of a mouse model has proven a point. So the strategy that was adopted was to try to inhibit this pathway by overexpressing adiponectin in the adipose tissue of a mouse and seeing if you could inhibit the fatty liver response. And the results of that experiment represent one of the most striking phenotypes that I have ever seen. Here you see an Obi-Obi mouse on the left and an Obi-Obi mouse on the right that has a transgene for adiponectin in its adipose tissue. The mice on the right become super obese. As a matter of fact, as you can see in the graph on the left, they reach 100 grams. Despite that remarkable expansion of adipose tissue, the adipocytes in the histology below are actually small. And furthermore, they have no inflammatory response. So what's associated with small, non-inflammed adipocytes? There is increased, or I'm sorry, reduced lipolysis and reduced or absent fatty liver along with preserved insulin sensitivity. And so these two experiments have demonstrated cause and effect that have really formed the foundation of our concept of how adipocyte liver interactions in the context of fatty liver disease. So what about a second seminal observation? This is a graph from the classical experiments that were performed by Peter Turnbaugh when he was in Jeff Gordon's lab showing that the microbiome of obese mice is different from the microbiome of lean mice. This in itself is a provocative finding, but what's remarkable is the mechanistic studies that followed and proved an important point. So what they were able to do was demonstrate that microbiotic transfer of stool from either wild type or obese mice into germ-free mice led to either a normal or an obese phenotype without any change in food consumption in the recipient, demonstrating that the obese phenotype was transmissible from the stool. And they went on further to demonstrate that this characteristic was not unique to the OBOB phenotype, but it was reproducible also in wild type mice that were simply fed a high energy diet and had diet-induced obesity. So these investigators actually didn't look at liver outcomes, but liver outcomes were subsequently determined by investigators who performed FMT from, in this case, patients who were either healthy or had NASH. And what you can see in this slide is that the mice that received the NASH microbiome had steatosis at baseline, which was exacerbated with a high-fat diet in the presence of this abnormal microbiome. And so again, together, these experiments demonstrated by direct manipulation of a mouse model, a foundational new paradigm that has led to what we now understand as the connection between microbiome and metabolism leading to abnormal liver phenotypes. So what about another kind of model? So in this instance, you're not changing the mouse, you're not changing the diet, you're changing the way the diet is given. So in this experiment, which was based from UC San Diego in the laboratory of Satchin Ponda, what they showed was they could feed mice a high-energy diet. And if left to their own devices, the mice would eat this diet ad lib and become obese and develop fatty liver, as you can see here in the histology. On the other hand, if you fed the same diet to mice but only restricted the time that they could eat it to 12 hours out of every 24, they did not become obese, they did not develop fatty liver, and they ate the same amount of food. So this is a phenomenal observation. The mechanism underlying it is yet incompletely understood, but it's known to involve the microbiome. And so the reason that I'm demonstrating it here is, as you probably are aware, these experiments have spawned an enormous amount of clinical investigation in time-restricted feeding as a way to either prevent or treat fatty liver disease in humans. So another thing that you can do to a model that has nothing to do with the mouse or the diet is to change the experimental environment. In this study, mice were housed in thermoneutral conditions, and a difference in phenotype was observed. So just as background, normal vivaria operate at 22 degrees, which is good for humans but not good for mice. Normal body temperature for a mouse is 30 degrees, and so if you house a mouse at 30 degrees, it will be more comfortable, will expend less energy. So when mice are fed a high-energy diet under thermoneutral conditions, they readily develop signs of fatty liver disease, as you can see in the histology at the bottom. This is true despite the fact that the same didn't occur when the same animals or same group of animals was housed at 22 degrees. Another thing that was found in this particular experiment was that the pathology was evident in both male and female mice. And this is interesting because typically female mice had been resistant to the high-fat diet in this instance. So another aspect that I'd like to call out, which doesn't really involve specific manipulation of a mouse model, but nevertheless represents a tour de force of experimental manipulation, is the leveraging mouse models of fatty liver disease to unravel immune drivers. And this work has been done by several investigators, probably who are in the room today, who have, through the process of cell isolation and analysis, both single-cell analyses and flow cytometric analyses, really unraveled all of the innate and immune processes that seem to be operative in the context of fatty liver. I'll just call out two of them. One is the demonstration of tremendous macrophage heterogeneity in the liver, and the second is the activation of autoaggressive CD8 T cells with features of exhaustion that can lead to hepatocyte injury on their own, as well as hepatocellular carcinoma. So at this point, I wanted to stop for a minute and say, what is the best diet? I have to admit that I'm somewhat of a purist, and so I favor the diets that are the simplest. Wild-type mouse, high-energy diet, no gene manipulation, no dual sort of hit. Others have tried combinations, such as putting an animal on a diet and then treating it with carbon tetrachloride and LPS, or putting Obi-Obi mice on the MCD diet so the weight loss isn't such a problem, or using a STAM model in which mice receive streptozotacin and are then given a high-fat diet in the context of what turns out to be type 1 diabetes. I've never considered those to be particularly favorable, and so prefer instead to keep it simple. Right now, I think we have enough high-energy diets that we can reasonably reproduce what looks like fatty liver disease in a mouse model. But I may be a little naive, because there was a study published in 2016 by Teufel and colleagues demonstrating that despite some of the best mouse models, shown in red on the right, the transcriptomics of the livers of those mice are still quite distinct from any human liver, and these are human livers, whether they be normal, obese, or fatty liver disease. So there's almost no overlap. There's only one promise, which is that this study was published before the thermoneutrality experiments were performed, and there was actually a more recent study published by a group in McMaster University that said if you put a diet together in thermoneutrality and then look at this kind of investigation again, you can see that there's some more convergence between humans and mice. But I still look at this with some pause and recognize that this may underlie some of the differences that we've seen from preclinical outcomes to clinical outcomes when it comes to therapeutics for fatty liver disease. So that leads me to experiments that may address this problem, experiments that in fact model human NAFLD using human cells. One flavor is to use primary human cells, and they can be used in two forms, either as multicellular microtissues or organoids with heterotypic cells, hepatocytes, endothelial cells, Kupfer cells, and stellate cells, or they can employ microfluidics, organ on a chip, where cells are embedded in the center and fluids run out the side. These kinds of models do have some advantages in that the organoids and the microtissues have extended survival compared to monolayer cultures. And using these kinds of heterotypic models, you can interrogate the spectrum of NAFLD. In addition, you can use these in preclinical models to test drug responses. They do have some disadvantages. The cells can be hard to come by, and particularly in a system such as this, there's potentially no guarantee that all of the cell types might be coming from the same donor. So what's an alternative to use induced pluripotent stem cells? iPSCs do have advantage in that they're a renewable resource. They're obviously human. The ones I'm talking about are from human. And so there is an additional opportunity to perform controlled manipulation, particularly as it relates to genetic variants that might be present in the patients themselves. And so for this, I call your attention to a study that was published recently in the Journal of Hepatology looking at iPSCs that had an engineered mutation in the PNPLA3 product. So what you see here is a control cell, an isogenic cell line that expresses abnormal PNPLA3 product, and then followed by a knockout for PNPLA3 altogether. And what you can see is kind of expected that the mutant PNPLA3 does have more fat, both spontaneously and with the addition of oleate, than the wild type. But surprisingly, the knockout has even more than the mutant. And the reason that this is unusual is that it contradicts studies that were recently or previously performed in mice. In mice, it's been shown, as you can see in the histology on the left, this is a PNPLA3 MM knock-in mouse that has some hepatic steatosis on chow and substantial hepatic steatosis on a high-energy diet. But if you suppress or silence PNPLA3 with an antisense oligonucleotide, in this instance, the fat goes away. So now we have a discrepant result. In the human cells, knocking out PNPLA3 makes the steatosis worse. In mice, it makes it better. And so this is actually not insignificant, because there are currently drugs in development that are intended to silence PNPLA3 as a treatment for fatty liver disease. So this underscores the need to validate what's found in animal models in human cells. So I'll just speak for a second about additional models of NAFLD. One is a multicellular IPC model that can actually try to reproduce what's been done in the cell culture. And the second is an organoid model, which Dr. Takebe, who's speaking right after me, will discuss at great lengths. So I'm not going to discuss these further. But instead, what I'd like to do is to talk to you about some experiments that we've been doing with IPSCs as models of human NAFLD. So in the study that I showed you just a moment ago, there was one cell line that was manipulated to either express abnormal PNPLA3 or knock it out. What our objective was in the study that we just performed was to see if we could identify differences among populations of patients, some who have NAFLD and some who don't. So what we did was to generate IPSCs from a population of patients, many of whom have NAFLD and some of whom are controls. We now have over 60 lines. But right now, we have data that we've published for 37 of them. We differentiated them to IPSCs, and the information that we obtained was rather provocative. So what we could see from the outset was that the IPSC-derived hepatocytes from patients with NAFLD had evidence of hepatic steatosis at baseline, where the controls didn't. And I want to underscore that these cells were not treated with any lipotoxic medium. This was strictly out of the box. So this was a spontaneous phenotype, and it was present in much of the population. Secondarily, what we found when we were looking at transcriptomics was that there is a transcriptomic distinction between the patients and the controls as well. So where this becomes interesting is that what this means is that there is a genetic difference between the IPSCs from patients and the IPSCs from controls that is responsible for this, because by definition, in the cell culture environment, we've taken the environmental issues out of the picture. So I'm just going to come back for one second to PNPLA-3 to tell you where this is driving us next. So if you see in the heat map on the left, underneath, I have the genotypes for PNPLA-3 for each of these patients. What you can see in the black is that a substantial number of the diseased patients have PNPLA-3 GG, but you also see that three of the control patients were PNPLA-3 GG as well. And if we look at the triglyceride content of those GG control patients, it's low in comparison to high. So that's at odds with what you just saw a minute ago, that a cell that was engineered to express abnormal PNPLA-3 GG became steatotic. So what's the difference? The difference is that each of the dots in this graph are different patients, and these patients do not have only PNPLA-3 GG to worry about. They have all the other risk genes for fatty liver disease that are probably coming into play. So what we are going to need to be doing in the future is to unravel the polygenic nature of the phenotype that we see in these cells in these experiments. So in the end, where do I see the future of fatty liver disease modeling? Honestly, I think that there's going to be room for both mouse models and human models. They're growing together in their capabilities, but there's still some distinctions between the two. So mice have the advantage of the ability to ramp up numbers, and they also have the ability to give you large numbers of populations of particularly immune cells that can be studied readily in vitro. Human models have the advantage of being able to manipulate human genetic variants that mice don't. But honestly, what I think is going to be the train coming down the road is that both of these models may wind up being reverse translational to the human multi-omics data that are being generated every day. So I think in the end, this will be a continuum with human data potentially being the driver of experiments that are subsequently performed in mice and humans. And with that, I will conclude. Thank you so much.

NAFLD

mouse models

metabolic features

hepatic steatosis

microbiome

diet

iPSCs