My name is Laurie DeLave and I would like to welcome you to the Hans Popper Basic Science State-of-the-Art Lecture. Our presenter is Professor Neil Henderson, Chair of Tissue Repair and Regeneration at the Center for Inflammation Research at the University of Edinburgh in the United Kingdom. Professor Henderson received his medical training in Edinburgh and London before completing a Wellcome Trust-funded PhD at the MRC Center for Inflammation Research in Edinburgh. He then trained in hepatology and was awarded a Wellcome Trust Intermediate Fellowship, during which time he spent a three-year postdoctoral period based at the University of California in San Francisco in the US. He returned to Edinburgh where he has remained since. Among his numerous honors and awards is a Wellcome Trust Senior Research Fellowship in Clinical Science that he has held since 2014. The title of his talk today is using single-cell genomics approaches to decode human liver disease. This reflects the important body of work Professor Henderson and his colleagues contributed to literature on single-cell genomics to understand that the cellular and molecular mechanisms that drive organ fibrosis, most notably their highly cited paper in Nature in 2019, resolving the fibrotic niche of human liver cirrhosis at a single-cell level. With no further ado, I give you Professor Neil Henderson. Hello, everybody. And I'd just like to say a very big thanks to the Organising Committee for asking me to give this Hans Popper Basic Science State of the Art Lecture. It's a huge privilege. And so thank you very much to the Organising Committee and ASLD. So what I'd like to discuss in the next 25 minutes or so is the work we've been doing in Edinburgh and also talk about the work around the world that's been using this new technology, single-cell genomics, to help decode and understand human liver disease and really across human liver disease from fibrosis to regeneration. This technology has really been very disruptive and exciting to be a part of. Just a quick disclosure statement with my research funding and consultancy. So what I'd like to talk about initially is the fibrotic niche of the liver. And as we all know, it's very much a cellular, multicellular orchestra. There are many conserved cell types within this fibrotic niche in the liver compared to other fibrotic organs. And this is a deliberately oversimplified schematic of the fibrotic niche. And as many people in this audience will know, this is the space of DISA in the liver. And I'm biased, I'm a hepatologist, but we have beautiful repeating 3D architecture throughout the liver. And there's a lot of clockwork involved in driving fibrosis and some outstanding work over the last 30, 40 years in preclinical mouse models and cell culture has allowed us to really start to unpick and understand how liver fibrosis actually occurs. But what we were keen to do was to harness this new technology, single-cell genomics, to help us understand more about how human liver fibrosis actually occurs through this multicell lineage orchestra. So precision medicine is an increasingly used term in medicine in general. And I think we're getting to a stage where the technologies are allowing us in human liver fibrosis and other liver diseases to really get towards true precision medicine. What we set out to do a few years ago was to really increase our understanding of the fibrotic niche in human liver cirrhosis and fibrosis. And in some ways, the holy grail would be, could we block fibrosis with minimal perturbation of homeostasis and with minimal perturbation of homeostatic function? And on top of that, I guess the icing on the cake would be epithelial regeneration as well. And so what we felt in Edinburgh was that we needed higher resolution. So four or five years ago now, we started to bring in single-cell genomics as a technology, single-cell transcriptomics, and really try and harness this to give us higher resolution on this fibrotic niche in human liver fibrosis. Really excitingly, there's been a blossoming of these technologies across medicine. And just to give you an example of the excitement that this has brought in, the Chan Zuckerberg Initiative, CZI have funded a large number of consortia in different organs to really help understand how biology occurs at single-cell level in human tissues. And this is just an example of people who are part of this consortium for the Human Liver Cell Atlas funded by CZI, and you may recognize a number of people in some of the pictures here. This atlas is designed to understand human liver homeostasis, but there are plans afoot to allow us to start looking at disease as part of this atlas as well. And I just draw your eye to a review we wrote last year that no doubt, although it came out in August last year, is out of date already, I'm sure, in this rapidly evolving field. But really, there's been a rapid adoption of single-cell genomics across the breadth of hepatology, and it's a very exciting time to be part of this. So in Edinburgh, we moved out to this site about 15 years ago from our old teaching hospital, and this is the new one. And QMRI is a research institute, which is slap bang beside the National Liver Transplant Unit as you can see there. So this allowed us to get tissue from the surgical theaters at time of transplantation, and other liver sections over to the institute to make single-cell suspensions very quickly. And this is a broad schematic of our pipeline. As you can see on the left, we take healthy or cirrhotic liver tissue in a variety of etiologies of cirrhosis that we transplant for, very similar to the demographic of liver disease in many other countries around the world. And this includes NAFLD, PSC, PBC, alcohol autoimmune hepatitis, and really any indication that we transplant for, we can make single-cell suspensions from to allow us to do single-cell RNA sequencing. We use plate-based technologies to begin with, but we quickly moved over to 10X genomics. And you can see that device on the right there, and this allows barcoding of single cells to then allow single-cell RNA sequencing of the individual cells. And I've listed some further modalities that we've adopted in the last couple of years, including single nuclei RNA sequencing, and I'll come back to that. And that's been a very useful technique to us in addition to single-cell RNA sequencing, and also spatial transcriptomics and multi-home sequencing, whereby you can combine multiple modality readouts from the same cell. So this was our initial work that we published in 2019, and this was a single-cell census of healthy and cirrhotic human liver, around 135,000 cells. And so what we did initially here was just take the deliberately basic approach of dividing the datasets, an NL5 in healthy and cirrhotic in these TSNEs, and we now use UMAPs for dimensionality reduction, which allows people like me, who I'm fundamentally a biologist, I'm not a bioinformatician by any means, but it allows you to just eyeball the data and see if there are any striking differences straight away between healthy and cirrhotic datasets. And so we then manually annotated all these clusters, as you can see on the right-hand side, into the various cell lineages. There are automated annotation approaches available now. We tend to still use signatures and manually annotate the different cell lineages, but you can see very quickly that there are multiple non-parenchymal cell lineages here. And in the next few slides, I'm going to use macrophages as an exemplar cell type of how we started to really dig into this data and try to make sense of it and try to understand new biology by it. And so this is initial clustering of mononuclear phagocytes. Clustering is in itself an art. If you overcluster, then it's very difficult to validate at the wet lab level what new clusters you think you found. If you undercluster, you're at risk of missing rare and interesting subpopulations of cells. So there's definitely an art to this. And what I would say is that we've learned an awful lot about clustering and where perhaps to set the threshold with this type of data, but it's an important consideration. And so what struck us straight away, and this is just informatics-based data on the bottom graph, is that there were two, I'm calling SAMs here, scar-associated macrophage populations. But in essence, this graph just says that at the informatics level, there were two macrophage subpopulations that were markedly expanded in cirrhosis compared to the healthy datasets. And so what John Wilson-Kanemori, the bioinformatician in our group, then did was he was able to identify differential gene markers for these macrophage populations that were expanded in the informatics datasets and disease. And two of these markers that John came up with, TREM2 and CD9, Prakash Ramakandran, who was leading on the wet lab side of the work in our lab, was then able to prospectively take cirrhotic livers, make single-cell suspensions and do flow cytometry to investigate whether these markers looked appropriate and relevant and basically validate whether these TREM2-positive, CD9-positive macrophages were truly expanding in cirrhosis. And you can see here with the graph on the right, which is quantitation, that sure enough, the TREM2-positive, CD9 macrophages were expanding in human liver cirrhosis. So then what we did was we wanted to ask the question of, well, we can see the expanding number by flow cytometry and quantitation, but topographically, where do these TREM2-positive, CD9-positive macrophages exist within the liver? Are they sprinkled throughout the nodular parenchyma of hepatocytes or are they actually topographically more restricted to areas of scar? And so this micrograph on the left, you can see TREM2 and CD9 are marked in red and white and collagen one is the green stain. And again, many people will be well aware of liver histopathology in cirrhosis, but if not, the central darker area in the middle is a hepatocyte nodule, and then you have broad swathes of scar fibrotic septae around that in green and collagen. And so where I was drawn to, we thought that the TREM2-positive, CD9-positive macrophages were indeed inhabiting areas of scar, but to quantitate that, Prakash in the lab then used an ImageJ approach to help quantitate whether the macrophages were randomly distributed throughout the liver or not. And in essence, the graphs along the bottom show that these macrophages were enriched in fibrotic septae and very much localized to areas of scarring. So just to switch over to one of the other cell lineages we looked at within the non-parenchymal data set, this is human liver endothelial cells and taking a very similar approach to the macrophage populations, we then asked the question of, first of all, how many clusters of endothelia are there within these data sets? And secondly, are any of the endothelial subpopulations enriched in cirrhosis? And I would like to draw your eye to the ACKR1 purple population or more pinky population, and also the lilac PLVAT positive population in the middle, as these were markedly expanded in cirrhosis. And so again, taking the approach of trying to investigate and validate these populations, you can see here along the top, a micrograph of human liver cirrhosis and staining with ACKR1 and PLVAT within that. And we use ERG as a nice nuclear endothelial stain because as people know, endothelia can be little skinny guys and quite difficult to identify at times, but ERG is a nice extra marker of endothelia. And in essence, two readouts from this experiment were important. It confirmed these ACKR1 PLVAT positive endothelial subpopulations, but also we found a very restricted topography of these populations to areas of scar. And so if you look to the left of the micrograph and also the top right, these are hepatocyte nodules, and we very rarely saw any of these ACKR1 or PLVAT positive endothelial subpopulations in the nodules, but they were very restricted to scar. And again, we have quantitation along the bottom whereby we confirmed what we could see by eye with increasing N and also quantitation of that. So we then reached a stage where we're able to identify scar-associated endothelia, scar-associated macrophages, and I've not shown you the data in the interest of time today but also scar-associated mesenchymal cells. But a key question was how do these newly identified multilineage scar-associated subpopulations actually come together and drive human liver fibrosis? So one approach we used to try and get at that and understand that in and amongst this high-dimensional, really rich dataset was to use a ligand-receptor interaction algorithm that Mariana Efremova and Sarah Teichman's lab in Cambridge had developed at that time called CellPhoneDB. And this is a really nice algorithm that allows you to take any subpopulation you have in either your homeostatic healthy liver datasets or your diseased cirrhotic datasets and ask the question of, let's, for example, see scar-associated endothelial certain subpopulation, how does that interact with the scar-associated mesenchyme or scar-associated macrophages? And similarly, you can do this in homeostasis with the various different subpopulations. And as you can imagine, this outputs a lot of data. And so what I would say is that, for me, there is no substitute for the human eye and lots of pairs of our eyes pouring over the data. But what we found with this analysis was that it both sprung up ligands receptor interactions one might expect from the prior literature, and also novel interactions. And so on this slide, I just want to show some of the ways we've tried to validate these putative ligand receptor interactions from the informatics data with wet lab approaches. So we've taken here PLVAC-positive endothelia prospectively isolated on flow from diseased human livers, cirrhotic livers. And you can see in the flow cytometry graph on top left, we've got JAGAD1, a known Notch ligand that is confirmed to be upregulated by flow and then quantitated. And then on the top right micrograph, you can then look topographically at the various ligands and receptors that you've identified in the cell phone algorithm. And you can see here, PGGFR-alpha is marking mesenchymal cells. Notch 3 is present on these mesenchymal cells as a receptor for DLL4, the Notch ligand. And then taking this in a co-culture setting, bottom left, we have human stellate cells co-incubated with SCAR-associated PLVAC-positive endothelia, which increases collagen 1 expression in the stellate cells. And then when we add a Notch inhibitor DBZ, this reduces collagen 1 expression in the stellate cells co-cultured with the SCAR-associated endothelial cells. And the bottom right graph shows primary human stellate cells incubated with a small inhibitory RNA for Notch 3. And you can see this reduces collagen 1 expression in the human stellate cells. So there are obviously various ways one can go about trying to validate the findings from the single cell data. You can use co-culture primary human disease cells as we've used here. You can look in tissue topographically with multiplex IF. You can use preclinical rodent models, although I would say that one of the limitations of the rodent models is they sometimes don't recapitulate the exact populations we're seeing in human disease. And that's for the obvious reasons. Rodent models tend to be short term, short duration, through no one's fault. It's just the way it is. Versus, you know, if you're looking at a Nash cirrhosis, that might have been burning away in that liver for maybe 20 years in a patient. So lots of different ways to try and validate the findings in informatics data. So what this allows is you can then build up interactome maps, if you like, of the various scar associated populations and the various ligands and receptors throughout these cells. And we've summarized here some of the ligands and receptors we've been interrogating over the last couple of years. But you can very quickly build up these interactome maps. But the key, of course, is functional validation of these. So what I'd like to do in the coming slides is just move on from the initial single cell work we've done and really talk about some of the new multimodal approaches that people are using in tissue fibrosis, and more specifically in liver fibrosis. So as I mentioned earlier, there's been this incredible rapid explosion of multimodal single cell approaches. And nowadays you can measure multiple readouts from the same single cell, which I think we'd all agree 10, 15 years ago would have seemed almost impossible. But now we can measure, for example, the epigenome along with the transcriptome. We have spatial transcriptome that's coming in, and I'll talk a little bit in a couple of minutes about that too. But in essence, I think it really is quite an exciting or a very exciting time in hepatology, but also in biomedicine, because we're building up these single cell maps of organ fibrosis and in the liver that's going to allow us to identify new subpopulations, new cell states that are disease specific, disease associated, and one would very much hope that that can only drive forward precision medicine and targeting of liver fibrosis. And if anyone's interested, myself and two colleagues wrote a review in Nature last year summarizing some of these single cell genomics approaches in the context of fibrosis. So a couple of the techniques that we've been working hard on in the last couple of years is a technique called single nuclei RNA sequencing and also spatial transcriptomics. And what we feel one of the beauties of single nuclei RNA sequencing is that this allows us in Edinburgh to open up a lot of the epithelial data. So you may have noticed in the earlier data sets we generated, it was very non-parenchymal cell dominated. And this is because the two to three centimetre cubed wedge biopsies we get from X plants don't allow perfusion to allow hepatocyte isolation. And we tried various ways to do that, but it was very challenging. So single nuclei RNA sequencing has allowed us to open up hepatocyte data and more cholangiocyte data. And the other big advantage of this approach is that it allows you to access biobanks of frozen human liver tissue. And so single nuclei RNA sequencing has been transformative for us and other groups around the world in that way. It also allows sharing of frozen tissue in various centres and has been really, really helpful. Spatial transcriptomics is a very rapidly evolving nascent technology. We use the Visium system from 10X Genomics as many other people do. And it's a very clever system whereby you can barcode on an array. So it's almost like a next gen microarray and you have barcoding of these 5,000 spots across a matrix, which you lay down your frozen tissue onto. And this allows you to start doing what's called in-situ on-slide sequencing, which is a very exciting new advance, which should combine, you know, genome wide sequencing data with topography within the tissue. And you can look back at your original H&E and track exactly where the gene expression is. And there's new kits going to be coming available that will allow this approach in formalin fixed paraffin embedded tissue, as well as the frozen tissue, which the kit's been around for a year, a year and a half or so. So it's exciting times in this area as well. So we've been using single nuclei RNA sequencing to try to understand the changes that occur in the cellular and molecular mechanisms as human NASH progresses. And so we've been building up large data sets in healthy liver, but also in increasing fibrosis and steatosis grade, as well as Nafldex plants. And I would very much like to mention Ross and Bas on the wet lab side in my group who've been driving this work forward. And also John Wilson-Kanemori, the bioinformatician in my group. So we're now beginning to build up large data sets across NASH progression, which will hopefully allow us to identify the key targets as this disease evolves. In terms of spatial transcriptomics, and I acknowledge Andreas and Ross in my group who've been pushing this on, spatial transcriptomics of human NASH as shown here, and what this immediately, which was, which was nice to see because we'd never done this technique before to really see how the data looks, is very quickly you can pick up the scarred versus non-scarred areas, which you would hope, but it's nice to see what actually in front of you. And the top image here is end features or numbers of genes on a heat map. So more pink, orange is more genes. And you can see in the bottom left that you can then take this data and start to really interrogate it. And cluster zero is hepatocyte rich spots, whereas cluster one is non-parenchymal cells in the scar areas. And very quickly, in a reassuring way, you can see that it does pick out even at the top level areas of hepatocytes versus areas of non-parenchymal cells in scar and the heat maps on the right very much concur with that. So we are currently increasing our experience with spatial transcriptomics. As you can imagine, the informatics pipelines are similar but different from the previous single cell and single nuclei RNA sequencing. But again, you can very quickly pick out what looks like appropriate topology. So this is hepatocyte signatures applied within the Visiam data and you can see nodules of hepatocytes with areas of scar that are not lighting up for the hepatocyte signatures as one would hope to see. And then when you overlay and apply signatures for scar associated mesenchymal cells, these are picking out the areas of fibrotic septae and not the hepatocyte nodules. So a lot of the early data we've accrued with this approach has been reassuring from a prior biology standpoint. What we're currently doing is pairing the single nuclei RNA sequencing datasets with the spatial transcriptomics data and you can use this using various informatics approaches. We've been using an algorithm called Harmony. But what this allows you to do is start to assign what the most enriched lineages are within each of those 5000 matrix spots to try and get a handle on. We know because each of these spots in the arrays is 55 micrometers in diameter, so multiple different cell lineages will contribute to those spots. But to really start to try and unpick and understand which spot may be enriched most for say hepatocytes or mesenchymal cells. And you can see here we've two samples on the left hand side, two NASH explant samples and there's quite reassuring overlay of the data from the two spatial transcriptomics datasets. And then on the right using cell assign, we can start to, as I say, assign cell lineages within these spots and really start to try and understand the in situ interactions that occur. Because whenever we make single nuclei or single cell suspensions, we lose all that spatial structure and relationships within the tissue, whereas with spatial transcriptomics that preserves that and so we're working hard trying to really bring together these multimodal approaches to leverage and get the richest possible information about how the fibrotic niche operates in human liver disease. So just to summarize, single cell genomics approaches and techniques really are driving a resolution revolution. And I think the biggest beauty for me is that it allows this powerful unbiased exploration of cell states and types at single cell level. And so we've seen multiple groups generating data where we have unexpected insights into liver biology and disease. And as I mentioned earlier, of course, this is being rolled out across medicine. It's a very rapidly evolving field. There are new algorithms informatically coming out almost every month. The new technologies are evolving and iterating and getting better continually. So it's very quick moving. It's been a lot of fun trying to keep up with it all, but it is a very, very rapidly moving field. As I mentioned a few minutes ago, the spatially resolved molecular profiling is also blossoming as an approach, and I think it will only get better and better in the next two to three years. And this should help us understand how these cells interact in situ and really the convergence and integration of all these multimodal technologies alongside very large scale initiatives such as the Human Cell Atlas really are an incredible opportunity to try and decode the cellular molecular mechanisms that regulate human liver disease. And so what we and many others hope is that these technologies will really help drive and push a new era of precision medicine. It's almost like, you know, we had a microscope before, whereas this is an incredible lens on all the cells and molecules that are playing a role in driving human liver disease. So I think, you know, this new era of precision medicine and the treatment of liver disease is upon us. And I think these technologies will be part of that as we drive it all forward as a community. So I'd just like to acknowledge many people in my lab group listed on the left there who've done an incredible amount of work over the last few years, pushing this all onwards and out in our group. I'd also like to acknowledge the Scottish Liver Transplant Unit and the patients who very kindly give us their bits of liver when we do the transplants. Also the transplant coordinators and the surgeons who help with our tissue procurement pipeline. There's a number of key collaborators I'd like to acknowledge, Sarah Teichman and John Marioni in Cambridge, who really helped me get single cell going in Edinburgh and were incredibly collaborative in terms of training. John Wilson Kanemori, who's the first bioinformatician in my group. So a big thanks to them. Quentin Anstey in Newcastle for the exciting ongoing work we're working with Quentin on, on NASH progression. Jackie Mayer at UCSF and Robert Schwabe at Columbia in New York for ongoing collaborations in this space as well. I'd also like to just highlight these two web browsers we have for our liver single cell datasets. One of the things we're very keen on is making all this data very accessible for other groups. It costs a lot of money to generate these datasets. So we and others are very keen that this data is used to the maximum. And so those two browsers may be of interest to people. And finally to acknowledge all the funding bodies that have helped us do this work over the last number of years. And thank you very much for your attention. I'd be very happy to take any questions. Thank you.

single-cell genomics

human liver disease

liver fibrosis

single-cell RNA sequencing

spatial transcriptomics

precision medicine