Hello, everyone. Thank you for joining the Liver Fibrosis Special Interest Group's program, Multidisciplinary Perspectives in Developing New Treatment for Nerve Fibrosis. I am Yujin Hoshida from University of Texas Southwestern Medical Center. I will co-chair with my colleague, Professor Hezer-Francis from Indiana University. We've seen promising progresses in nerve drug development, but it is still a challenging process with several obstacles. To further streamline the process and accelerate clinical translation of the therapeutic discoveries, we identified the four critical key steps summarized here, starting from understanding of disease pathogenesis, followed by target discovery, preclinical and clinical drug development, and then regulatory approval. We are fortunate to have these outstanding speakers to overview and outline the recent discovery developments, as well as remaining challenges for each of the four key steps. Good afternoon. My name is Jennifer Chen, and I am an assistant professor at UCSF. I am delighted to participate in this exciting session, and I thank the organizers for the kind invitation. The topic of my talk is New Molecular Mechanism of NASH Liver Fibrosis, and these are my disclosures. In this talk, I will discuss the design and implementation of a small molecule screen, the identification of acid ceramides as an antifibrotic target, and the validation of acid ceramides using multiple models of fibrogenesis. Cirrhosis affects approximately 600,000 Americans and accounts for 30,000 deaths per year. An estimated 16 million Americans have non-alcoholic standard hepatitis, NASH. Hepatic fibrosis is the final common pathway that leads to end-stage liver disease, and unfortunately, there are a lack of therapies and predictive biomarkers for patients with hepatic fibrosis. There have been several studies that have shown that fibrosis is the histologic feature associated with disease progression in NASH. This was a retrospective analysis of over 600 patients followed for 12 years, and this figure shows that patients with NASH and fibrosis, shown in red, have significantly decreased transplant-free survival compared to those with NASH and no fibrosis, which is shown in orange. These data highlight the importance of halting fibrosis progression to treat patients with NASH. Activation of hepatic stellate cells is the key step in hepatic fibrogenesis. In their quiescent state, stellate cells store lipid droplets. In response to chronic injury, these stellate cells undergo activation, leading to loss of lipid droplets, induction of contractile filaments such as alpha smooth muscle actin, and production of extracellular matrix proteins such as type 1 collagen. These activated hepatic stellate cells then produce the fibrotic scar. The goal of my research program is to identify novel mechanisms that inactivate hepatic stellate cells as a strategy to develop new therapies and biomarkers for my patients with hepatic fibrosis. During my gastroenterology and hepatology fellowship at Massachusetts General Hospital, I worked with Ray Chung and Alan Mullen to design a small molecule screen. We wanted to take an unbiased approach to identify compounds that inactivate hepatic stellate cells. As our positive control for stellate cell inactivation, we plated stellate cells on nitrogel. This leads to inactivation as evidenced by accumulation of lipid droplets, which can be visualized using staining, just shown here in green. So this was the primary readout for the screen, identifying compounds that promote lipid accumulation, visualized by BODEPI staining. This diagram shows the workflow of the screen. We plated hematopoietic stellate cells on plastic on day one, compounds were added on day three, and cells were fixed and stained with BODEPI on day five. In a pilot screen, we analyzed 1,600 compounds in duplicate and identified 21 hits, of which four were tricyclic antidepressants. We then validated that TCAs promote stellate cell inactivation. Nortriptyline, a representative TCA, promoted lipid droplet accumulation, shown here by BODEPI staining. Nortriptyline also led to a dose-dependent reduction of FTA2 in collagen expression and an increase in PPR gamma expression. Nortriptyline reduced alpha smooth muscle actin protein levels, and even in the presence of TGF-beta, a potent stimulus of stellate cell activation, nortriptyline suppressed expression of pro-fibrotic genes. We then validated that nortriptyline reduced fibrosis development using a carbon tetrachloride model. After four weeks of treatment, mice receiving nortriptyline had a reduction in fibrosis, as measured by a decrease in hydroxyproline and a decrease in serious red staining. This was an interesting finding, since TCAs have been used in clinical use for over decades. We asked whether TCA use was associated with a decreased risk of fibrosis progression in patients. We collaborated with Adil Butt, and using his cohort of 128,000 patients with chronic hepatitis C infection, we observed that TCA use was associated with a lower risk of developing cirrhosis, as measured by APRI scores. This association persisted even after controlling for other variables, including diabetes and alcohol use. TCA use was also associated with an increase in cirrhosis-free survival. We then wanted to understand how TCAs were inactivating stellate cells, and this is important because we observed the dose used in mice was 50 times higher than the dose used in patients. So to identify how TCAs were inactivating stellate cells, we performed RNA sequencing, which highlighted the sphingomyelinase pathway. Sphingomyelin is converted to ceramide by acid sphingomyelinase. Ceramide is hydrolyzed by acid ceramidase to sphingosine, and sphingosine is phosphorylated by sphingosine kinases 1 and 2 to form sphingosine 1 phosphate. Our data show that TCAs inhibit acid ceramidase and increase ceramide. We show that nortriptyline inhibited acid ceramidase enzymatic activity to a similar extent as shown with depletion of acid ceramidase using siRNA. Nortriptyline increased ceramide and reduced sphingosine levels. We then showed that inhibition of acid ceramidase promotes stellate cell inactivation. We used a tool compound, B13, and we observed a dose-dependent reduction in ActA2 on the collagen expression and an increase in PPR gamma expression. B13 also reduced alpha-smooth muscle actin protein levels. And we observed similar findings when we added exogenous ceramide, which led to the question, how does ceramide promote stellate cell inactivation? So to elucidate the mechanism, we performed RNA sequencing of human hepatic stellate cells treated with ceramide. Ingenuity pathway analysis identified that ceramide significantly downregulated genes involved in stellate cell activation and actin cytoskeleton signaling, consistent with our previous studies. The analysis also identified that the YAP signaling pathway was downregulated by ceramide. We performed gene set enrichment analysis using a publicly available dataset of siRNAs targeting YAP-TAS and hep G2 cells with our ceramide dataset. This showed enrichment of genes downregulated with YAP-TAS knockdown among the genes downregulated by ceramide, suggesting that ceramide mediates its transcriptional response by inhibiting YAP and TAS. We then confirmed that ceramide suppresses expression of multiple YAP-TAS transcriptional targets, including ANKRD1, CTGF, Imodal-2, and Ser-61. Originally described in Drosophila, the HIPPO pathway is a kinase cascade that culminates in the phosphorylation of the key effectors YAP and TAS. Specifically, the lab's kinases phosphorylate YAP-TAS at specific serine residues. Phosphorylation at serine 127 promotes cytoplasmic sequestration. Phosphorylation at serine 397 recruits a key adapter protein, beta-TRCP, which then leads to proteasomal degradation of YAP and TAS. When YAP and TAS are phosphorylated, they cannot translocate to the nucleus, where they've been shown to promote fibrosis in multiple organs, including the liver, the lung, the kidney, and the skin. I was first introduced to this pathway by the late Andy Taker, and much of this work has been inspired by him. We first asked, does ceramide regulate nuclear localization of YAP and TAS? Sarah Awesome, a very talented research assistant in my laboratory, led these studies and observed that ceramide inhibits nuclear localization of YAP and TAS, which is shown here in green. We then quantify the nucleotide-cytoplasmic ratio of YAP and TAS per cell, and we observed that ceramide significantly reduced this ratio. We then asked whether ceramide promotes phosphorylation to regulate localization, and indeed we observed that ceramide promotes phosphorylation at serine 397, leading to both YAP and TAS degradation. Since ceramide promotes YAP-TAS degradation, we then interrogated the degradation pathway in more detail. We observed that in the presence of the proteasome inhibitor MG132, ceramide no longer inhibits nuclear localization of YAP and TAS, and the nucleotide-cytoplasmic ratio is quantified below. And we observed similar findings when we depleted beta-TRCP, suggesting that ceramide promotes beta-TRCP-mediated proteasomal degradation of YAP and TAS. We next asked whether regulation of YAP-TAS is mediated by the last kinases. So here we nucleoaffected stellate cells with a YAP-397 mutant, which cannot be phosphorylated and is persistently nuclear. In the presence of the mutant, ceramide no longer inhibits nuclear localization of YAP and TAS, which is quantified to the right. Furthermore, in the presence of the mutant, ceramide no longer inactivates hepatic stellate cells. In the presence of the control YAP, ceramide decreases ActA2 and collagen expression. However, in the presence of the mutant, this effect is lost, showing that constitutive activation of YAP overcomes the effect of ceramide. And we observed similar findings when we depleted the last kinases. So mechanistically, our data suggests that acid ceramides inhibition increases ceramide to promote lax-mediated phosphorylation and degradation of YAP and TAS, leading to stellate cell inactivation. We then wanted to see whether inhibiting this pathway was relevant using models of fibrogenesis. Here, we induced fibrosis in mice using the carbon tetrachloride model, and we treated mice with B13, an acid ceramides inhibitor, for the last three weeks of carbon tetrachloride treatment. We observed that B13 ameliorates fibrosis as evidenced by a reduction in hydroxyproline and a decrease in serious red staining, which is visualized in the middle and quantified as collagen proportional area to the right. In collaboration with Joanna Schaub at Plant Therapeutics, we then turned to an ex vivo model of fibrogenesis, and we used precision-cut liver slices from fibrotic rafts. So specifically, we induced NASH fibrosis in rafts by feeding them the choline-deficient L-amino acid-defined high-fat diet, CDAHFD, for 18 weeks. We prepared precision-cut liver slices from these fibrotic livers and treated them with two different acid ceramides inhibitors, B13 and Serenib 1. We demonstrate that both inhibitors significantly reduced pro-fibrotic gene expression. Similarly, we prepared precision-cut liver slices from two patients undergoing liver transplantation at UCSF, one with alcoholic cirrhosis and one with primary biliary cholangitis, and observed that treatment with the acid ceramides inhibitors Serenib 1 significantly reduced pro-fibrotic gene expression. These data using rat and human liver tissue suggest that inhibiting this pathway may reduce fibrogenesis when fibroids may reduce fibrogenesis when fibrosis is already established. To further demonstrate cell type specificity, we deleted acid ceramides and stellate cells by crossing a flox acid ceramides mouse with the PDGF receptor beta-CRE. We observed that deletion of acid ceramides and stellate cells reduced fibrosis in the carbon tetrachloride model, as shown with a decrease in hydroxyproline and a reduction in serious red staining. And consistent with our in vitro data, we observed that deletion of acid ceramides inhibits YAP nuclear localization. In collaboration with Megan Mooring from Dean Yim Lamai's group, we demonstrated that control mice receiving carbon tetrachloride have nuclear YAP staining in the stellate cells, as depicted by the presence of pink cells, which shows the overlap between the red YAP staining and the blue DAPI staining. However, in the liver tissue from the knockout mice, we observed decreased nuclear YAP staining, which is quantified here to the right. YAP and TATs play a critical role in mechanotransduction and regulate extracellular matrix stiffness. Nadia Ayad, a graduate student in Val Weaver's lab, performed atomic force microscopy, and this revealed significant decreases in matrix stiffness in liver tissue from the knockout mice compared to control mice treated with carbon tetrachloride, consistent with ceramide regulation of YAP-TATs in vivo. We also analyzed the impact of acid ceramides deletion using the CDA-HFD model of NASH fibrosis, and observed that these knockout mice had significantly reduced fibrosis, as evidenced by a reduction in hydroxyproline and serious red staining. We also observed a reduction in the YAP-TATs transcriptional targets, ANKRD1 and CTGF, in the knockout mice, suggesting YAP-TATs inhibition in vivo. Importantly, we observed that acid ceramides depletion in stellate cells does not worsen steatosis or lobular inflammation. We also observed that acid ceramides depletion in stellate cells does not significantly impact liver-to-body weight ratio, ALT, or levels of hepatic triglycerides. We've recently analyzed these mice in another dietary model of NASH, the fructose palmitate cholesterol, FPC model, and we observed similar findings. We then wanted to demonstrate whether this pathway was relevant in patients, and we measured acid ceramides expression in patients with chronic hep C infection, with no fibrosis, and those with advanced fibrosis. And we observed that acid ceramides expression was increased in activated stellate cells in those with advanced fibrosis, compared to those with no fibrosis. To further demonstrate the clinical relevance in patients, my collaborator, Stephanie Christensen, generated a gene signature, the Ceramide Responsiveness Score, which consists of the top 100 genes down-regulated by ceramide from our stellate cell dataset. She then examined a dataset developed by Sydney Moylan and colleagues at Duke of 72 patients with NAFLD, 40 with mild fibrosis and 32 with advanced fibrosis. Using genes at variation analysis, we observed that the ceramide responsiveness score was significantly increased in those with advanced fibrosis suggesting that the biology inhibited by ceramide is enhanced with increasing fibrosis severity. She then generated enrichment score metrics for the stellate cell activation and the YAP pathways. This heat map shows the 72 patients with NAFLD as individual columns with yellow representing those with advanced fibrosis and purple representing those with mild fibrosis. The scores are shown in individual rows with the red representing a high score and blue representing a low score. This heat map shows that all three signature scores, including the ceramide responsiveness score, were coordinately upregulated in patients with advanced fibrosis, suggesting that promoting ceramide may be a successful strategy to reduce fibrosis. These studies were recently published in Science Translational Medicine. In summary, a small molecule screen identifies tricyclic antidepressants as antifibrotic compounds. We show that TCAs inhibit the enzyme acid to ceramides to inactivate hepatic stellate cells. Targeting acid ceramides promotes proteasomal degradation of YAP and TAZ. Pharmacologic inhibition or genetic knockout of acid ceramides and stellate cells reduces fibrosis. The ceramide responsiveness score may serve as a potential biomarker to identify those with advanced fibrosis. We have several ongoing projects in the laboratory. They include deciphering how targeting acid ceramides regulates lat-mediated signaling. We are characterizing the conditional knockout mice in models of NASH, specifically analyzing insulin resistance, changes in the immune microenvironment, and analyzing ceramide subspecies. We are working to develop the ceramide responsiveness score as a predictive biomarker. We are analyzing other hits from the small molecule screen. My group is presenting a poster at this meeting, and I encourage everyone to join that session. I'd like to highlight the members of my laboratory who contributed to this work. Sarah Asimov and Amy Yu in particular. This has been an incredible collaboration which started in Boston with Ray Chong, Ellen Mullen, Brian Fuchs, and the late Amy Tager. These are my funding sources and my lab is hiring. If you're interested in joining, please contact us. Thank you very much for your attention. I look forward to your questions and listening to the other speakers in this session. Hello from me as well. Hello from Boston. First of all, I would like to thank the organizing committee for the kind of invitation. It's a spectacular venue and a great pleasure to be with you here today. My lab focuses on the integration and application of novel technologies and data-driven approaches for precision medicine applications. We're located, as I said, in Boston and our main footprint is Beth Israel Deaconess Medical Center, Harvard Medical School, and the Broad Institute of MIT and Harvard. In the past decade, an exciting realization has been that diverse liver diseases ranging from non-alcoholic cellular hepatitis, ALD, cirrhosis, to hepatocellular carcinoma follow along a spectrum. This realization helped us to drive our research program in a comprehensive way. But if you actually dive deeper in each independent liver pathology, you can see that there is a complex multicellular interplay. This is an example, for instance, from NASH. There is a complex multicellular interplay that might be driving each phenotype with significant differences and varying complexity. In this example here in NASH, we have a hepatocyte macrophage stellate cell axis with numerous connections and cell states, along with an interaction addition with the lymphocyte component. Then there are multiple ways to traverse this plot. But as an example, metabolic insults promote hepatocyte cell doses and injury activating this multicellular network comprising macrophages, NKT, NK, B cells that control cell activation and downstream fight process. So there is an extensive cell-cell communication with multiple signaling pathways involved. If we just select another one, I just select PSC as a second example, primary sclerosing cholangitis, since we're also working extensively on it. Recent studies have also identified interesting cellular characteristics. For instance, PSC patient livers have a significant increase of adrenobatic macrophage populations, CD68, CD206, CD68-206-CGR5, which is the major bioreceptor with effects varying on cell localization. PSC livers also have increased populations of non-classical intermediate CD16 monocytes resulting from recruitment of the blood. While naive CD4 T cells were also identified in PSC livers, which also exhibited a strong resident signature, it's easy to understand that the environment, the microbiome, which is different in PSC, the differences in the genetic background, but also the interplay and communication between the cell types and their localization, which are not present in healthy liver, they create this very complex landscape. It's very difficult to understand which cells and which cell states are in the driver's seat or passenger's seat. To this end, we invested extensively maximizing the amount of information we can extract from bulk RNA-seq. In this approach we call whole genome and beyond, we can get multiple layers of information from mutational landscape, TCRs, gene expression, non-coding genes. Now we have a new application we're developing for microbiome. We can get a lot of information in multiple layers in order to start capturing this complexity that I mentioned. However, despite this significant advances and this wealth of information we're getting from bulk, it still is bulk and it lacks a granularity. We can discuss about cell types, but we cannot discuss about cell states. We can see for instance CD8, but we cannot understand whether they're exhausted, what exactly is their signature, what the cells are doing. Our understanding is that this complexity needs to be tackled for liver pathologies. Now, single cell and then more recently, spatial technologies can give us this granularity. We can interrogate specific cells, each individual to understand exactly that, the cell states and the transcriptional programs. With in-situ technologies, we can keep the tissue intact and understand their localization and colocalization. Around this realization, we have established our research program, which currently produces from tissue sample, single cell and spatial concordant from the same sample, which enables us to capture the different cells, population, the states, specific events, cell-cell communication, architecture through the spatial localization, colocalization, how the cells affect each other and they interact with each other, and the effects of the microenvironment. I will go now and use an example on our work on COVID-19 and the application of this framework. Because this is an extensive project, it will give me the ability to go step by step and show the application of this process. We now know a lot about COVID-19 and the information where the community is gathering is still expanding in a super linear rate. But despite the incredible progress, there are still numerous often challenges. Unfortunately, the numbers are still growing. About a year and a half ago, this is exactly the date where we presented this on our pathology meeting about this project that I'm going to discuss. We actually restructured our research program and applied this framework which we established for liver and for our cancer work for COVID-19, which had just hit the Northeast and it was growing rapidly. We wanted to capture how COVID-19 affects the cellular phenotypes across tissues, the cellular interplay, and a question that was phrased very early on, we knew that COVID-19 has a multi-tissue phenotype. But it still is an open question, especially for liver. We started this effort with Yuri and Gordon Chang from medicine, and Jonathan Yack was the head of our autopsy program in our pathology department. But this effort quickly grew and formed a disciplinary team comprising colleagues and experts from pathology and medicine and support from technology developers. With the help of everyone, including our colleagues in the ICU, our postdocs in the lab, and most importantly, our patients and their families, we were able to collect multiple tissue samples from patients who succumbed to the disease. We subjected the samples into sequencing, which as we will discuss shortly, offers an excellent representation of parenthematic cell types across tissues. While concurrently using novel special technology, we're able to quantify four protein markers for antibodies in situ and 18,000 genes on the tissue section. This integrative effort has been transformative for this effort and for our research program down the road. Along the way, we joined the Boston-wide efforts to expand this cohort with exceptional colleagues from academic medical centers in the Northeast and abroad. Through this effort, we were able to collect more than 420 patient samples from 11 tissue types. 100 samples were processed for the regional publication and we continued this effort. We performed single cell and single-nucleus sequencing per sample, as well as spatial transcriptomics and spatial proteomics for a subset of those samples. The process, this is for an example from lung, but this went across different tissues, and the process was as follows. H&E images were studied, and then IHC against the SARS-CoV-2 spike protein and other different proteins. ISH using RNA-scope against the SARS-CoV-2, transcriptase-2, temporase-2, in order to identify potential viral presence in the tissue sections. Then this information guided our region of interest selection, which was based on architectural features of our presence, inflammation, tissue damage, and the protein markers, which we had selected per tissue. Then viral scores were also calculated and correlated with additional probes which were added to the nanostring, whole transcriptome atlas, which we were using against SARS-CoV-2. In the original atlas paper, it comprised results from four tissues, including the first liver samples. It was recently published, and we focused extensively on lung, which is the organ with primary interest, of course, in COVID-19 research. We were not only able to provide the first detailed cellular map of the COVID-19 lung, but also prioritize cell targets, cell types and states explaining a substantial part of the ARDS lung phenotype, which could potentially be leveraged for targeted therapeutics. We extended this approach to liver, and this is, I'm going to show a very recent and published work by combining single-cell but also spatial investigations. Region of interest selection in liver was performed a lot based on architecture across the liver lobule, portals on 1, 2, and 3, as we see here across, and also the antibody markers that we used, and cell-specific selection. This combined team effort, and on the single-nucleus sequencing side, enabled us to generate 90,000 nuclei from 20 subjects with a very realistic representation of all cell populations, including the epithelial component, which is more often than not, lost using single cell. This was clearly a victory for single-nucleus sequencing. It's definitely advantageous for tissues such as liver, especially for liver. Since as you can see in these examples from leading publications which utilize single-cell sequencing, you have a higher granularity and over-representation of the non-parietomatic populations. This very high number of nuclei enabled us to characterize in high-resolution the function of hepatocytes, and identifying numerous affected cells exhibiting stress, apoptotic, or acute response phenotypes in our COVID-19 patients, which we also compared with healthy controls. We were able to characterize also the immune, endothelial, and mesenchymal cell compartments in very, very high detail and identify cell populations again only present in the patients. While the very large number of cells enabled us to identify six clusters in the cholangiocyte compartment, immature, small, large, reactive, ducts, intermediate cholangiocytes, and prophylogenic cells were identified thanks to this very high granularity. By extending a method we used in the original Atlas paper, but we adjusted it for liver, we were able to identify liver cells actively infected by SARS-CoV-2. As we can see here, this is the hepatocyte compartment, which is a small proportion of the cells can be infected because we know which nuclei have the viral presence. This can give us the ability to identify transcriptional programs enriched in those populations, and understand what is the phenotype and how they affect the other cells and cell communication. While having the spatial data annotated with anatomical locations such as zonation and cell markers. So this is from an actual patient slide, we can see the four markers in a selected region. This enabled us to register the single cell clusters in situ and then maximize granularity. For instance, here we can see that the abundance of a specific cholangiocyte cluster across the spatial regions and the integration of these two technologies enabled us, not only to identify the cell types of interest, but also identify them across the tissue and understand whether they're located in specific zones and their abundance. We can see the difference also in pathways or cell populations within the same patient in different regions of the liver sample, but also between patients. Jared from our group and Puriya Anandheri from Wyn Hyde's lab, just part of this effort, developed a method for pathway activity on spatial transcriptomics, which permitted the detection of regional pathway analysis and pathway level comparisons with healthy sample data. Information which we are now using not only to understand liver biology, but also the effects of COVID-19 on the liver. And we can see, for instance, the differences of transcriptional programs across zones in regions where there are infected cells across the liver lobules. It's quite empowering. And this, of course, is ongoing work. And I just cherry-picked some of the results. But I would like to close mentioning the most important take-home message for us through this effort is that the integration of these two technologies, of these two modalities, on the same patient samples has been truly transformative for us. Single cell, and especially for liver, single nucleus sequencing offered as the granularity, while spatial transcriptomics brought in CTL, in tissue architecture information. We learned a lot from our investigation of five tissues, but now we're learning so much more about the liver COVID-19 phenotype, but also basic and fundamental liver biology, which we're now bringing also to other liver pathologies. Closing, I would like to mention that this work, of course, was the result of a collaboration from many talented people. It really takes a village. And I would like to especially thank the members of our lab and the BIDMC team for the tremendous long hours and the breakneck pace to bring this project forward, as well as the technology developer, for the access to the WTA assay right after it was developed, before reaching the market and support throughout this effort. Closing, I would like to share, along with our excitement about these technologies, the news of the creation of the Spatial Technologies Unit in BIDMC and in the Harvard ecosystem with a mandate for rendering spatial and single cell technologies accessible to all, academia and industry, regardless of where you're located. And we're looking also, and this is quite important, we're looking for talented dry and wet lab scientists and postdocs to join this effort. These amazing technologies are setting up side by side to be integrated or tested or piloted. So please feel free to reach out if you would like to work with us or visit our website for more information. Would like to thank you all for your attention. We'd be more than happy to take any questions. Good afternoon. My name is Rob Myers, and I lead clinical development in liver fibrosis and gastroenterology at Celiad. I'd like to thank the organizers for inviting me to present today regarding drug development challenges in NASH cirrhosis from a biotech industry perspective. These are my disclosures. It's no surprise to this audience that the unmet medical need in NASH cirrhosis is high. Cirrhosis is associated with increased risk of liver-related and all-cause mortality, and the prevalence of NASH cirrhosis is growing exponentially due to rising rates of obesity and diabetes. Importantly, the burden of NASH cirrhosis is substantial from both the patient and societal levels. And in light of this high unmet need, multiple randomized trials have been conducted of new therapies in this population, but unfortunately, none have been successful, and therefore, no therapies have been improved for NASH cirrhosis. Now, one of the key challenges, in my view, for drug development in NASH cirrhosis is that current regulatory guidance by both the FDA and the EMA advocate event-based clinical trials for approval of therapies for NASH cirrhosis. The challenge is that these event-based trials are long and large and have limited feasibility. What I'm showing in this table are estimates of required sample size under a variety of assumptions in this patient population. And I've highlighted in red what I think are reasonable assumptions. First, that the hazard ratio for events for an active therapy versus placebo would be 0.8. In other words, a 20% relative reduction in risk of events defined as decompensation, transplantation, or death during follow-up. We've also assumed a 30% incidence of these events over five years of follow-up, and this is extrapolated from pooled data from Gilead's Cituzumab and CilantroTip Stellar IV trials. And what you can see, that for five, six, or seven-year trials, under these assumptions, you'll require at least 5,000 patients to have an adequately powered clinical trial. Now, based on estimates like this, it's clear that alternative surrogate endpoints are needed for this patient population. Indeed, in 2015, the FDA and the ASLD put out this guidance document where a few potential surrogate markers of the risk of mortality were proposed for patients with NASH cirrhosis. You can see changes in the MELD score and Child-Pugh score, as well as HVPG, were considered. I'm going to talk a bit about HVPG today, but would also like to discuss additional options, including liver histology, non-invasive tests of fibrosis, and the development of new varices. Well, let's start with liver histology. The challenge currently is that regulatory agencies have considered the evidence base for a link between changes in fibrosis and clinical outcomes to be unclear in the cirrhosis population at this time. Now, this is, of course, in contradiction to guidance regarding non-cirrhotic NASH drug development, where the endpoints of NASH resolution or fibrosis improvement have been considered as acceptable surrogate endpoints. The data on this slide address the link between changes in fibrosis and clinical events in the cirrhosis population, and it comes from our simtuzumab and solancer tip stellar IV trials. What I'm showing on the left is the incidence of liver-related events, defined as decompensation, transplantation, or death in these trials, according to whether or not these cirrhotic patients had cirrhosis regression on their 48-week liver biopsy. And as it turns out, 16% of these patients had cirrhosis regression over 48 weeks. And what you can see quite clearly is the patients who had cirrhosis regression had a markedly reduced incidence of liver-related events during follow-up. In fact, a roughly 85% relative risk reduction versus those patients who did not have cirrhosis regression. Now, it's important to note that the patients who had cirrhosis regression had evidence of milder fibrosis at baseline, including on liver histology and non-invasive tests of fibrosis. And therefore, the skeptic might argue that these patients did not truly have cirrhosis regression and that this was simply sampling variability. However, I think the data on the right at least partially refutes this. What I'm showing here are changes in a variety of histologic and non-invasive parameters of fibrosis, according to whether or not patients had cirrhosis regression. And what's clear is that patients who had cirrhosis regression, shown in blue, had greater reductions in hepatic collagen content and alpha-smooth muscle acting expression on biopsy, as well as significantly different changes in ELF score, FIB4, and liver stiffness by transient elastography, and in the subset of patients who had HVPG measurement, a greater reduction in HVPG over time. So these data suggest to me that cirrhosis regression determined histologically is a validated surrogate endpoint and should be considered acceptable for clinical trials in NASH cirrhosis. Of course, one of the challenges of conventional histologic assessment is limited reproducibility between readers, and as well, conventional ordinal staging systems have limited sensitivity to detection of treatment effects. And therefore, we've spent quite a bit of effort in collaboration with PATH-AI to look at alternative approaches to evaluation of fibrosis, particularly use of machine learning models and images of liver biopsies. We've come up with a couple of scores which allow us to quantify fibrosis. One is referred to as DELTA, or the Deep Learning Treatment Assessment score, and this is closely related to the machine learning NASH3 and fibrosis score, which identify patterns associated with fibrosis using machine learning. And what I'm showing you here on the left are data from our ATLAS trial, where we looked at silofexorin for sulcostat versus placebo treatment in patients with F3 and F4 fibrosis, and I'm showing changes in the DELTA liver fibrosis score in these two treatment groups according to whether or not patients had fibrosis regression over 48 weeks, as deemed by the central pathologist. What you can see is that among placebo-treated patients, there's no difference in the DELTA fibrosis score, according to whether or not the patient had fibrosis reduction or not, according to the central pathologist. However, with silofexorin for sulcostat treatment, you see a significant reduction in DELTA in those fibrosis responders versus non-responders, and this to me suggests that this score may be more sensitive to change than that of conventional histologic reading. We've also generated machine learning models that can predict the hepatic venous pressure gradient, and here I'm showing an overlay of this model, which is focused on patterns of fibrosis that correlate with HVPG. What you can see in the table on the bottom is that patients who had a hemodynamic response, in other words a 20% reduction in HVPG during follow-up, had a significant reduction in their machine learning HVPG score compared with non-responders, and a similar observation was made with respect to cirrhosis regression. Those patients who regressed their cirrhosis had a greater reduction in this machine learning HVPG score. So, these data, if validated, may be one or two approaches for using liver histology to more accurately or more sensitively detect changes in fibrosis, as well as HVPG in clinical trials. Now, we'd all like to move away from liver histology due to its, you know, clear disadvantages, and in this regard, a wealth of literature has supported the prognostic utility of non-invasive tests of fibrosis at baseline in patients with NASH cirrhosis, and this is data from our combined Symptusamab and Stellar programs in patients with cirrhosis, which confirm that. Effectively, you can see significant associations between baseline ELF, TMP1, a component of ELF, the NAFLD fibrosis score, as well as liver stiffness by transient elastography on clinical events. But what's interesting here is that we also see significant association between changes from baseline in these parameters, particularly for TMP1, NAFLD fibrosis score, and liver stiffness by transient elastography, suggesting that changes in these parameters have prognostic utility. Now, I would caution, however, that the optimal cutoffs for these changes that are predictive of fibrosis response and the risk of events are currently unclear, and therefore, additional work validating these endpoints needs to be done before they can be considered as validated surrogate endpoints in clinical trials. Let's move on to HVPG. I talked a little bit about HVPG before, and it's clear that clinically significant portal hypertension, or an HVPG above 10, as well as an HVPG response, are well-validated prognostic factors. But there are a few challenges with using HVPG in clinical trials, and what I'm showing on the left are data from our Symptusamab trial, where we did indeed use HVPG as an endpoint. When we were designing this trial, we reached out to 341 sites to see whether or not they had the capability to perform HVPG measurements, and you can see that only about one-third or 117 sites had this capability. Ultimately, we were only able to include 78 sites in the trial, and these are sites that we thought would provide high-quality HVPG measurements. So not only is HVPG availability limited, but it's vital that adequate training and standardization of measurements is incorporated into the design of these trials if it's going to be used. Another important consideration, particularly in the NASH cirrhosis population, is that body weight and changes in body weight can confound changes in HVPG. So something to think about if you're considering using this as an endpoint in a trial. Then lastly, a point that I think is very important is that the placebo response for HVPG is very high. This too is data from our Symptusamab trial, where we looked at the proportion of patients who had at least a 20% reduction in HVPG or a reduction to below 10, and you can see that roughly one-third of placebo-treated patients had a response. And obviously, this has important implications for sample size estimation in clinical trials if you're using HVPG as an endpoint. Alternative to HVPG that's also related to portal hypertension is the development of new varices. On the left, I'm showing…or on the right, I'm showing some data from a recently reported phase 2 trial of belipectin, the galactin-3 inhibitor, in patients with NASH cirrhosis. And while HVPG was measured in this trial, unfortunately, significant differences were not observed between the groups. However, in a post-hoc analysis, the investigators did see a significant reduction in the incidence of new varices among patients who did not have varices at baseline in patients treated with belipectin 2 mg per kilogram versus placebo. And based on these data, the sponsor has initiated an adaptive phase 2b3 design in patients with NASH cirrhosis using the incidence of new varices as the primary endpoint in the trial. I think there are a few advantages and disadvantages of this endpoint, which I'd like to highlight here. Now, first, you know, it's clear that the development of new varices heralds clinically significant portal hypertension, and this is associated with an increased risk of complication, so I think this is indeed prognostic. Also, the incidence of new varices in this patient population is not low. It's on the order of 5 to 10 percent per year. And thirdly, endoscopic surveillance is standard of care in patients with cirrhosis, and therefore, using this endpoint in a clinical trial of this patient population, I think, would be deemed acceptable to investigators and patients. But I do see a couple of disadvantages that are highlighted here. One is that, by definition, this endpoint excludes patients who have varices at baseline, which we know accounts for about 40 percent of compensated cirrhotics in clinical practice. And then much…or very similar to the issue regarding liver histology is that intra- and inter-observer agreement for the grading of varices by gastroenterologists and hepatologists is not perfect. This is some data reported by Fatin and colleagues last year, which showed kappa values of 0.59 to 0.84. These are not bad, but of course not perfect and would need to be considered when designing a clinical trial with this endpoint as the primary endpoint. I want to switch gears a little bit before I close, you know, moving away from surrogate endpoints and talk a little bit about this patient population. And I want to highlight that patients with NASH cirrhosis are extremely heterogeneous. Many patients are more F3-like than they are F4-like, for example. And this is data, I think, which illustrates the point quite nicely from our STELLAR-3 and STELLAR-4 trials of solancertin. So this includes patients with F3 as well as those with F4. And what we've looked at is the incidence of disease progression in these trials, defined as progression to cirrhosis in F3 patients or decompensation, transplantation, or death in F4 patients, shown on the left. And then the proportion of patients who had at least a one-stage improvement in fibrosis without worsening of NASH on the right, with categorized patients according to baseline health score as either less than 9.8, greater than 11.3, or in between. And what you can see is that disease progression and fibrosis response are clearly dependent on baseline fibrosis severity. So I think this is a very important consideration when you're designing a trial in patients with NASH cirrhosis as to the spectrum of disease you'd like to include in the population based on your endpoints. It also emphasizes the importance of appropriate stratification of randomization due to this heterogeneity. In fact, we've typically stratified our randomization based on the baseline ELF score. I think another potential stratification factor could be, for example, the ISHAC stage, which is more granular than the NASH-CRN fibrosis stage. Lastly, this brings up an unanswered question to me, in fact, is whether or not there's a point of no return. In other words, is there a severity of cirrhosis at which fibrosis won't regress and or we can't reduce the incidence of clinical events? I think the word is still out on this, but additional research is necessary in this regard. In summary, NASH cirrhosis is a major unmet medical need and drug development in this population is vital. Histologic cirrhosis regression is associated with improved clinical outcomes and should be considered an acceptable surrogate endpoint in clinical trials of NASH cirrhosis. Finally, additional validation of non-invasive endpoints likely to predict clinical benefit, including non-invasive tests of fibrosis, is a key priority. I'd like to thank you for your attention. Dr. Ishida, Dr. Francis, good afternoon, everyone. I want to thank the organizers for inviting me to give this talk. Okay, as many of you know, I'm at the FDA. I started there in early 2018. The Division of Hepatology and Nutrition, or DHN, formally launched in March of 2020 before the lockdown. Our portfolio involves largely NASH, which we'll talk about mainly today, as well as drug-induced liver injury. This is my conflict statement, and here are my key takeaways. So, I want to talk about, one, the basic drug approval process that goes on at the agency, and then talk about how the NASH with liver fibrosis indication differs because it uses the accelerated approval pathway and a surrogate endpoint, and the rationale for that basis, point number three. Then the last half of the talk, I'll describe for you some of the challenges that applicants face in the drug trials for NASH with fibrosis. Finally, to identify some of the more finer points that I think you as academicians, largely basic and clinical scientists, may be able to help in these challenges. Okay, so the basics of drug approval. Bottom line is that once an applicant is done with their preliminary work, they will submit usually two large-scale, adequately-powered, phase three, randomized, double-blind, conceivable control trials. If they're effective, they will likely take a population in which they show that the patients in that group in the study drug do better in terms of how they feel, function, or survive. If that's the case, we do a safety assessment. If the benefit risk assessment is in their favor, they get approval. If there's a problem, there's no benefit, for example, or the risks outweigh the benefits, we send a CRL, a complete response letter. For those of you who write NIH grants, it would be the same, really, as a triage summary statement in which we tell the applicant what we think was wrong, what the weak points are, where the substantial problems are, and what they would need to do to resubmit the application for potential approval. Hypertension is a good example, and I would just point out that hypertension is a validated clinical endpoint, and I have a whole list of these types of examples of endpoints and markers at the end of the talk that are in a PDF format for you to look at on your own. Now, accelerated approval is a component of one of the expedited programs published in the guidance that I have hyperlinked for you here. There are three criteria for this. One, that it meets a serious condition. Two, that it's a meaningful advantage over existing therapy. And three, there's an unmet medical need. I think it's safe to say that NASH meets all of those criteria. So, in standard approvals I described before, there's a clinical benefit that's demonstrated on a clinical endpoint, and that you get full market approval, and that's the end. In the accelerated approval pathway, the difference is that there is a surrogate endpoint, which is easily likely to predict clinical benefit. And in this case, NASH meets these criteria, but the applicant is responsible for doing a clinical trial, a confirmatory trial to verify clinical benefit, with the notion that under the law, if the clinical trial does not confirm clinical benefit, then the agency has the authority to pull the drug because they didn't show clinical benefit. So, you have to show clinical benefit. Okay. Why is liver biopsy a surrogate endpoint likely to predict clinical benefit? It comes from these two papers, at least, that we all know, I think, that fibrosis stage is the strongest predictor of mortality in NASH, and I think that's well-known to all of you. And by extension, not going into details, we also make the extension that necroinflammation and Maluni degeneration are associated with the cause along the causal pathway of fibrosis. In the staging criteria that we have available, which we use the NAS scoring system, we do not include steatosis for efficacy analysis by histopathology when the trials come in. We don't include steatosis. We can talk about that later if you like, but everything else is fair game. In this talk, we're not talking about cirrhosis, and we also do not include…we recommend that applicants do not include stage 0 and stage 1 fibrosis in their enrichment in the population because those two groups of patients with those kinds of fibrosis scores are not indicated for treatment. Again, we can discuss these later if you like. So, these are the three endpoints, efficacy endpoints, of the histology trial. There's resolution of acetyl-tiazet as an improvement of the NAS score with no worsening of the fibrosis score, or there is an improvement of at least one stage of the fibrosis score and no worsening of the NAS score, and or the last one is that there is improvement of both the NAS score and the fibrosis score. Now, the applicant only needs to get one of these. They don't need all three of them. Although there are other aspects of liver biopsies of NASH patients that are considered, they do not meet the adequacy of these established pre-specified endpoints that we use the agency to establish the provisional approval of the drug. Now, in the phase 4 trial, there are clinical outcomes that we look at in these phase 4 confirmatory trials. Progression of cirrhosis, any hepatic decompensates, compensatory event like variceal bleeding, refractory ascites, a substantial change in MELD score, liver transplantation, all-cause mortality, which includes a major adverse cardiovascular event, etc. So, the bottom line with these is that if the applicant pre-specifies a number of patients, the number of events over time, then once they've achieved the number of events over a period of time, then the trial stops and they're analyzed. So, pictorially, this is what happens with the applicant. They submit, they do the phase 3 double-blind placebo-controlled histology trial, they submit the data in the survey. The data are analyzed. If the data meet an effectiveness or benefit risk, they get conditional approval, the drug goes to market, people begin to use the drug. At that time, they can roll the patients over and maybe add more patients to the phase 4 clinical outcomes trial, which is a time-to-event study, and then at some point later on, three to five years later, perhaps they submit these data for clinical benefit verification. If, in fact, the study drug succeeds in resulting in fewer clinical events, longer time to duration to these events, the applicant meets the full marketing approval certification, and they get full market approval, the drug is then fully clinically verified, and the drug meets all the approval process for this pathway. So, note, there are still two randomized trials that are done, but they're different, they're separated. And stated another way, these are the accelerated approval endpoints, and these are the clinical outcomes approvals. Again, two clinical trials, randomized, double-blind, placebo-controlled, but done in a sequence as opposed to being done together, as I stated, with additional full market approval. Now, one of the things I think that if you've read the literature at all about NASH trials that have failed, there are challenges that applicants meet, and I'm just going to mention three of these. These were mentioned in an article that the division published in the May issue of hepatology. I'm just going to mention three here. One is standard of care, the other is DILI, and the third is the continued use of liver biopsy as the surrogate endpoint. So, standard of care is important because there is a high placebo response rate in some of these phase three and phase two trials, and this is a very frustrating issue to clinical site investigators and the applicants. There are many reasons for this. One of them is that weight loss seems to result in substantial reduction in fibrosis scores, for example. The other is that these patients that enter these trials, it's very difficult to recruit and maintain them, and I think when patients enter these trials, they're highly motivated. So, it's not clear whether the placebo patients are doing the same things that the subjects on the study drug are doing, and we have to be mindful that these are international studies in all cases, so that may change the way patients behave. The other issue is drug-induced surgery. I think we're all very clear that NASH, like many other chronic diseases, is a lifelong drug. It will be lifelong, and therefore, this will be a problem, I think, and we have to…is there a drug-induced liver injury in these patients? And then, finally, the issue of liver biopsy, which I think doesn't need an explanation. So, weight loss. This is a prospective clinical study in which several hundred Cuban patients were followed at biopsy at the beginning and biopsy at the end. They were coached about eating and exercising. The main issue I want to follow is not the CO-leptitis for the NAS score, but fibrosis. How do these patients do? Bottom line is that most of the patients lost less than 10 percent of ideal body weight, and most of the patients had F0 and F1. About 70 percent had F0 and F1 fibrosis, which we would not consider treating with drug trial. The other thing is that about 30 percent of patients had F2 and F3 fibrosis of all these patients. Generally speaking, you have to have more than 10 percent ideal body weight loss to have a substantial reduction or regression by one stage here. So, I think this is a real issue, whether or not this is a durable fibrosis response to whether this can be sustained. These are trials that are year-long, and it's not clear to me whether or not this can be sustained beyond a clinical trial. This study had issues because the patients in the trial did not necessarily have a period of time. There were patients that a lot of data was imputed. We don't know how long it took patients to lose the 10 percent or more of their ideal body weight. So, I think these are, with any study, they're imperfect, but it does shed light on the fact that there is a conundrum about weight loss and fibrosis not with C. adipocytes. So, my question to the scientists in the audience is not, why does one of the biologic mechanisms that result in the improvement of hepatic fibrosis due to weight loss, and there seems to be a gradation here. The more weight loss you lose, of course, the more the fibrosis score is likely to improve. I don't think we know the answer to that, but that's a question that should be deciphered in the laboratory. Dilly, so it would seem, based upon the current hypothesis about drug-induced liver injury and NASH, that NASH would be a risk factor for patients who are going to be on a lifelong therapy. But if you look at the data, they're really quite scant with NASH. There really is no conclusive evidence that NASH patients have a higher likelihood of having injury from these drugs. Now, certainly, there's some finer issues about anti-fibrotics, but in general, with NASH, that's not the case. And I think it probably has to do with, you really can't do hypothesis testing and do rigorous studies in these patients. There was one thing I came across with NAFLD patients from a group of Italians in policy research many years ago that showed that NAFLD patients was in support for increasing the risk of Dilly in an obese, middle-aged Italian population of men. But again, these data are very loose, and we don't treat NAFLD patients with drugs. So it's not really clear whether there's a risk anyway. But I do think, and I think the growing consensus at work is that people are going to be likely taking these drugs the rest of their life. And finally, I want to talk about liver biopsy, because we hear a lot about this from our applicants, academics, and others. And this is really an area that's ripe for clinical and translational investigators. So, you know, I learned a lot at FDA, and one of the things I've learned about is biomarkers and the difficulties to establish them in various aspects of clinical trial design. And the FDA and NIH have a consortium called BEST, and there is a great syllabus that I've highlighted for you if you want to learn about it. It's a much more complicated field than I would ever imagine. There are all these different types of biomarkers, and I think that's one of the issues when biomarkers come in for analysis by experts, you know, what type of analysis needs to be conducted. But I will point out that biomarkers can be used for a variety of clinical tools, and they have to be validated in various and sundry ways. Analytical validation, clinical validation, there's two aspects of validation after the biomarker is studied. And I just point out this article from CGH that came out this year about the ELF score, and I think this title says a lot about where we need to go with biomarkers for clinical trials for NASH. Can be…this title says it all, predicting liver-related events in patients, in this case, with NASH and compensated cirrhosis. So, I think one of the problems with a lot of endpoints, with a lot of biomarkers that come in, is do you have a validated…excuse me, an endpoint biomarker that is based on information that predicts clinical outcomes? That is, can it predict mortality rates in patients with liver disease, with NASH? Can it predict that patients over time who develop compensated cirrhosis will be compensated at a higher frequency than those patients in a placebo cohort? Can it predict that patients who have NASH with liver fibrosis that go on to develop cirrhosis have a higher frequency of liver transplantation? If it can do that, then that particular biomarker is ripe to become a useful tool as a surrogate likely to predict clinical benefit that could supplant liver biopsy. What I think happens is that a lot of folks come in with surrogate endpoints in which they compare their data with liver biopsy, with liver histology, and that, I think, is a flawed comparison, as I've learned, because liver histology, as we've learned today, is a surrogate endpoint and you really cannot get information about a new surrogate endpoint from an existing one. You have to look at clinical outcomes in order to establish a new endpoint, and I think that's why there's a lot of frustration about these surrogate endpoints likely to bring clinical benefit coming online. I think that's really the issue. When you come to the FDA, one of the things that I remember someone telling me is you have to remember that expression, feels, functions, and survive. That is what we are about. So with that, I want to stop. I want to thank you for your attention. I'll be happy to take questions. Thank you. On behalf of Dr. Hashida, myself, and the liver fibrosis sick members, I want to thank all of our speakers for providing timely and important updates regarding new discoveries in liver fibrosis related to NASH. We've heard from expert leaders in the field regarding basic science discoveries, as well as new drug discovery possibilities from FDA and pharma. This work also demonstrates and highlights that there is significant crosstalk between hepatocytes and other liver cells that contribute to hepatic fibrosis and NASH progression. Identification of new signaling mechanisms may lead to targetable deliveries to patients suffering from NASH, and the use of computational biology is critical in assisting these discoveries. Considering that NASH has become a global health crisis, these discussions are both timely and warranted. Again, we thank you for attending the AASLD 2021 liver fibrosis SIG and look forward to your feedback. And we hope to see all of you again in person in 2022.

Liver Fibrosis

Special Interest Group

NASH

Dr. Yujin Hoshida

Professor Hezer-Francis

Nerve drug development

Disease pathogenesis

Acid ceramides

Antifibrotic target

Tricyclic antidepressants

Hepatic stellate cells