Okay, great. So again, we'll be talking today about our work on long non-coding RNAs in liver fibrosis. And just in terms of disclosures, we have a number of projects going on with our pharma collaborators, although those are separate from what I'll be talking about today. Okay. And in terms of liver fibrosis, I usually have more of an introduction, but I think at this point we've had that introduced very well already, and I'll just say that from our point of view, we're really focusing on the hepatic stellate cells because we know those are going to transform into the myofibroblasts, or at least the majority of myofibroblasts, that are going to be responsible for the chronic injury and are the primary source of fibrosis there that's going to lead to end-stage liver disease and liver failure. And if we think about the normal liver, we know there are hepatic stellate cells, which are marked in green up there in the top diagram. Those are in their quiescent state in a normal, healthy liver. Again, as we've had nicely described already, that during chronic injury, those cells are going to differentiate into the HSC myofibroblasts, those kind of elongated green ones towards the bottom. They expand, they proliferate, they become more contractile, and they produce lots of extracellular matrix proteins, majorities of which are collagen proteins, which really contribute to that fibrotic scar and chronic liver injury. And so what I'd like to do today is really talk about, first, what I mean by what are long non-coding RNAs, because that's, again, the point of most of our talk, and what those are by definition and why we think this might be relevant to liver fibrosis, both in terms of understanding the pathogenesis and, potentially, therapeutically, why they could be relevant. Then I'll discuss the identification of long non-coding RNAs that are expressed both in human and mouse hepatic stellate cells, and then, finally, talk about the details of one of those long non-coding RNAs, which we think is actually potentially a target to inhibit collagen and other extracellular matrix protein production. And so when we talk about long non-coding RNAs, we're talking about RNA transcripts that are five prime capped, and they're three prime polyadenylated, and they're single-stranded, so essentially the same features as messenger RNAs. The real difference is they don't encode proteins, or at least as far as we can detect, they do not encode proteins. These are separated from, when we talk about microRNAs, you know, they're small RNAs by definition that there are over 200 nucleotides in them. And then what has really attracted us to study them in terms of liver fibrosis is that they're actually much more abundant than protein-coding genes. So people can argue over the exact numbers, but they're on the order of 20,000, 25,000 annotated protein-coding genes in humans. But if we look at the long non-coding RNA data, there are over 160,000 of these genes annotated in human at this point, and we expect those to continue to expand as more cell types are investigated. And finally, this diversity allows there to be more cell type-specific patterns of long non-coding RNAs. So there's going to be more diversity of a long non-coding RNA population in individual cell types compared to what we're going to see with protein-coding genes, just because there's a greater diversity. As people have begun to study these long non-coding RNAs, we've begun to understand some of their functions. This is not meant to be an exhaustive list, but really just to give an idea of some of the types of activities that the long non-coding RNAs can take on in the cell. We understand that they can fold into secondary structure, and it's probably that secondary structure that allows them to interact with both RNA and proteins, and then they can also interact with DNA in other cases. So on the upper left, we're seeing an example of how a long non-coding RNA can act as a protein scaffold. So they can interact with multiple proteins and help bring those together to nucleate some activity in the cell. In the upper right, they can act as well as a competitor. So there are examples where a protein may normally bind to DNA. If the long non-coding RNA is present, that protein may now bind to that long non-coding RNA due to similar structure, and now that protein is no longer binding to the DNA target. In the lower left, you can see examples of protein-DNA scaffolding, where the proteins and DNA can be brought together by these long non-coding RNAs. And finally, there have been recent examples as well of the term of microRNA sponges, and the idea that if a microRNA is targeting a messenger RNA in the cell, if there's a link RNA that's happened to be in that cell type that can also be targeted, it can suck away or sponge away those microRNAs so that messenger RNA is no longer affected or not affected to the same degree. So, again, these are some examples. I think the way I would think about it even more is that if you were to take the first 50 proteins and try to extrapolate what proteins do from those 50 proteins, there are going to be a lot of other functions we just don't understand at this point, but this is really a starting point as we begin to think about function. And so, as I mentioned, that these long non-coding RNA transcripts tend to be more diverse in cells than the protein-coding genes, and so what we really wanted to do is to identify long non-coding RNAs that were normally expressed in activated stellate cells that were involved with extracellular matrix production, collagen production, and understand those that, if we could identify those that were there that were unique to stellate cells, those could become potential therapeutic targets to inhibit fibrosis. And the idea is that if those are expressed in a stellate cell, but they're not expressed in other cell types in the liver, then we'd worry less about the off-target effects as long as we can get those targeting constructs to the stellate cells. And as I mentioned before, there tends to be a lot of diversity among long non-coding RNAs, so if you want to move into a new cell type, we really can't just take what's annotated and look to see what's expressed. We really have to know what's expressed in our cell type of interest. So to do that, we went back with human stellate cells, we performed high-depth of RNA sequencing so we could rebuild the entire transcriptome in those stellate cells. We can then throw away all the genes encoding proteins, we can throw away the genes encoding microRNAs, we can throw away those that have higher coding potential by computational approaches. And by doing that, we can filter down to a list of long non-coding RNAs that we think are expressed specifically, or at least expressed in these human stellate cells. And by that analysis, we identified about 2,800 long non-coding RNAs in these fibroblasts. On the lower left is just an example of what we're doing with the data. So this is stranded data, so we know which strand the transcripts are coming from. The red marks areas where we're mapping three different exons of a novel long non-coding RNA. The blue marks that there's no reads underneath. We also do H3K4 trimethyl analysis as well. This is a mark that tells us where transcription start sites are in the cell. And so this helps us feel more confident that we actually have identified a long non-coding RNA, and we've identified that five prime end because we had that K4 mark there. And so again, if we do this genome-wide, we can identify about 2,800 long non-coding RNAs. We wanted to then try to break those down into smaller numbers that we could actually start thinking about what the individual function of some of these might be. And so we looked at different criteria, some of which have been discussed today already, in terms of characteristics that might allow us to identify those that might be more relevant to fibrosis. I won't go through the analysis, but we looked at those that were regulated by TGF beta signaling. And by analyzing expression of these transcripts over 40 different tissues, we could identify those that were also highly restricted to hepatic stellate cells, both compared to other tissues and compared to liver. We also wanted to think a little bit about function. And while we don't know the function of most of these transcripts, most of them have not been described before and haven't been worked out in much detail, we do know that proteins are annotated by function, and at least gives us a way to try to begin to predict what the link RNAs might do. So we took advantage of co-expression analysis. And again, this is the idea that the genes, both coding or non-coding, that are co-expressed across many different tissues in the same patterns are often part of the same networks. And so if we could identify the networks of proteins and long non-coding RNAs that were present specifically in the stellate cells and know what those proteins are doing, that gives us an idea of which link RNAs might be a part of those similar processes. And here, what we're showing is one of the networks, and it's not important to really take home anything other than that red means it's a long non-coding RNA, blue circles means it's a messenger RNA. And this is one of the dominant clusters that we pulled out of the network analysis. And what it's showing is that there are both a set of protein coding genes and long non-coding RNAs that are part of a co-expression network. The gray lines represent that the individual transcripts are co-expressed. And what we're seeing here is that there are a large number of both long non-coding RNAs and protein coding genes as part of this network. We don't know what the long non-coding RNAs are doing. They're not been annotated in much detail. But we do know these proteins are doing, and these proteins are actually all part of the extracellular matrix network, which suggests that these long non-coding RNAs may also be part of that same process. We can then filter that down a little bit further by comparing that network with all the other qualities we were looking at in terms of being regulated by TGF-beta signaling, being restricted to stellate cells, and generate a much smaller network that we can actually start thinking about in terms of individual long non-coding RNAs. And so for the rest of my talk, I'm going to focus on the long non-coding RNA you can see up there on the mid-right, the arrow pointing next to it, that's co-expressed with both COL1A1 and COL1A2 in hepatic stellate cells. So we went back, we cloned this product by doing race just to allow us to define the five prime end and the three prime end in the stellate cells and made sure we could clone the full-length transcript, and then began to look at the expression of that transcript. On the left here, we're looking at activation of stellate cells, so we isolated human stellate cells from a patient sample, isolated those and looked at RNA expression at two hours, and then after one passage and two passages as those cells are activating, and can see that the telic expression is now increasing. And again, if we look at the expression both in stellate cells versus whole liver, we can see those are again highly enriched in the stellate cell population, much lower in the full liver population, and what we're seeing there may also just represent the small fraction of hepatic stellate cells as well. We also want to understand functionally what they might be doing, and so in this experiment, we're taking lochnucleic acids, so these are single-stranded DNA that combine to RNA and degrade those, and we're transfecting human hepatic stellate cells with those two different LNAs. And when we do that, we can see a depletion of the telic, which is the name for this lung encoding RNA, in both cases by about 80%, and in those same cells, what we're seeing as well is that we're seeing a depletion of collagen expression, so about a 50% depletion of collagen expression when we deplete this lung encoding RNA as well, again suggesting that it may be a target that can regulate collagen expression. And then I'm going to summarize some data that I'm not going to have time to present. I kind of touched on some of those pieces of data just in the last couple slides, but what we see with this lung encoding RNA, which again we're calling TILAC for TGF beta-induced link RNA activating collagen, is that it's increasing expression as we go from a quiescent stellate cell to an activated stellate cell. If we look in the liver, we see that the primary source of that link RNA is coming from the hepatic stellate cell. Again, if we deplete that to the stellate cell, we can see a decrease in collagen expression from those human stellate cells as well. We'd also like to be able to look at this in more of an in vivo model, and that requires us to move into mouse systems, but we needed to figure out if there was an equivalent lung encoding RNA in the mouse as well. And so to do this, we took advantage of Robert Schwabe's help, and his lab was kind enough to sort out GFP-positive labeled stellate cells, both from quiescent mice and mice having been treated with CCL4 for four weeks. We could then perform the same type of RNA sequencing analysis, rebuild the entire transcriptome in the mouse, throw away all the protein coding genes, all the microRNA genes, and identify those lung encoding RNAs in the mouse as well. And again, the reason we need to do this is because if those lung encoding RNAs haven't been annotated, which they hadn't been in mouse stellate cells, we're probably going to miss many of those if we just look at the genomic annotation for mouse. And that is basically what we found. I don't have time to go through the approaches that we need to use to try to find orthologs. I just want to say that the sequence conservation is very, very poor, and we usually have to look more at genomic location to find what we think are the orthologs. And when we go through this analysis, what we found is this transcript here that has not been previously annotated in stellate cells in a similar genomic position, we look at the RNA-seq reads up on the top, we see there's really no evidence of expression of this link RNA in a quiescent stellate cell. But as these cells are activated with four weeks of CCL4 treatment, we can see this gene is turned on, and we see a significant expression there. This again is based purely on genomic position, and if we go back and want to look at function, we can knock these down with the same type of approaches using antisense DNA. We're here now depleting with two different antisense oligos, targeting two different regions of that long non-coding RNA mouse stellate cells in culture, and we can see again that a depletion of those link RNAs are also associated with a decrease in collagen expression. We went on then to develop a mouse model so that we could look at the loss of this link RNA in in vivo models of liver fibrosis. So what we're doing here is we've used the CRISPR-Cas system to insert, it's up in the upper left here, it says GFP-PA, it's a GFP cDNA followed by a polyadenylation signal. And this allows us effectively to convert the link RNA expressing gene into a gene expressing GFP that no longer expressed that link RNA. So what would happen is the GFP is going to be expressed, you'd hit the polyadenylation signal that's going to terminate transcription, so we just don't express much of the downstream long non-coding RNA. And so on the bottom is just schematically what we'd expect. If you have a normal wild-type cell, you add CCL4, those cells would express the link RNA. If you have the Tylek GFP mouse, you add CCL4. Now those cells normally expressing Tylek should become GFP positive, which would allow us to track them, but they would no longer express the Tylek long non-coding RNA. And just my last couple slides, there's some preliminary data, but I thought it would at least be helpful to see kind of what we're starting to see in the mouse model. I've got a larger view and then a more zoomed-in view to try to illustrate this because I wasn't sure how they'd project. So these are two samples. One is the knockout mouse or the GFP knock-in mouse with an olive oil control. The second one is treated with CCL4 for eight weeks, and we're doing immunohistochemistry here for GFP. I think you can see on the right side with the CCL4 that you can sort of see some of that GFP staining with the immunohistochemistry that's present in the mouse treated to induce activation of the stellate cells, but no longer not present in the other untreated mice. I have just to zoom in here to try to illustrate that a little bit better. I think you can see on the right side with CCL4 treatment, there are a lot of cells marked with brown. They're smaller nuclei. The larger hepatocytes are excluded from the GFP, and this is very recent data. We're in the process of counter-staining with a number of other different markers for other fibroblasts and cellular populations in the liver, but the distribution certainly would be what we'd expect if these are marking activated stellate cells. The question is, what happens if we deplete the ENTILAC and try to induce fibrosis in mouse models? Again, I want to say this is a preliminary experiment. We're still analyzing the histology and the hydroxyproline levels, but we could get to the collagen expression a little bit more quickly. In this experiment, we're looking at mice, either wild-type mice or mice that were treated with a high-fat diet for 12 weeks, the CDA high-fat diet, so the choline-deficient L-amino acid-defined high-fat diet, and then doing the same thing with mice in which we've depleted the link RNA that we're talking about. I think you can see on the two sets of mice on the left that, if normal diet, we don't see evidence of significant collagen expression. When we treat those mice with high-fat diet for 12 weeks, we can see collagen expression is induced, but that induction is significantly attenuated when we have the loss of that long-line coding RNA. So again, our interpretation, and we're in the process, again, of doing more analysis and moving these into other fibrosis models, is that loss of this link RNA in vivo is also inhibiting full expression of collagen as well. And then just to summarize, I put this little diagram up on the top mainly to reiterate the points that we're trying to start these experiments in human stellate cells because there's poor conservation between human and mouse link RNAs in many cases, so we want to be making sure we're studying something that's relevant to human. We can then move into mouse to identify those orthologs, which allows us to set up in vivo models and eventually, with those, be able to test in vivo approaches to try to inhibit these long-line coding RNAs, which we hope eventually would allow us to address therapeutics approaches in humans. Again, what I've shown today is that long-line coding RNAs are abundant. They're uniquely enriched. There's certainly a population of these that's uniquely enriched in myofibroblasts. The TILAC is one of these long-line coding RNAs, regulates collagen expression both in hepatic stellate cell myofibroblasts in mouse and human, and that if we look at that ortholog of that link RNA in mouse models, it does also seem to have a similar effect in in vivo mouse models as well. I'd also like to just to thank the people who've done the work, so the work was done in large part by four postdocs over the last few years in the lab, Kaveh, Chen, Cheng, and Amin. We've also benefited greatly from help from Yuri in setting up some of these in vivo fibrosis models, and Robert with his assistance with the LREC pre-mice, and again, I appreciate your attention. Thank you.

long non-coding RNAs

liver fibrosis

hepatic stellate cells

TILAC

therapeutic intervention