Scientist Stories: Joan Steitz, Noncoding RNAs with a viral twist

So it's with very great personal pleasure that I introduce to you our next speaker, Professor Joan Steitz. And I say it's a personal pleasure not only because she's a worldwide esteemed scientist, but also because we've been very close friends for a very long time. Neither of us will admit to this time. So Joan, it's great to have you here. At the Hall Institute again. Joan is Sterling Professor of Molecular Biophysics and Biochemistry and a Howard Hughes Medical Institute Investigator at Yale University. So a little bit about her early career. She did her PhD at Harvard with Jim Watson, and then she spent her postdoctoral period at the LMB Molecular Biology— Laboratory of Molecular Biology at Cambridge, where which is where we met. And basically Joan has lived her entire research career in the world of RNA, a world that I left some— at that time when I left, almost at that time, certainly left when I came to Australia. She's been a formidable pioneer in this world, and we'll be hearing a lot more about that from her today. She has tremendous list of honours She's a fellow of the American Academy of Arts and Sciences, the National Academy of Sciences of the US, the Institute of Medicine of the US, and also a foreign member of the Royal Society of London. I think one thing that stands out in her CV to me is that she's been awarded by the President of the US in person the National Medal of Science. I won't go on about her awards. She's an amazing scientist. Her research is marked by its incisiveness and clarity, as are her lectures. Her talk today is Non-Coding RNAs with a Viral Twist. Thank you, Joan. Please welcome Professor Joan Stein.

So I'd like to start by saying how absolutely wonderful it is to be here. I was fortunate in 2000 to spend a 3-month sabbatical here at the WeHi working in the labs of Jerry and Suzanne. And it was just such a refreshingly inspiring experience. And I'm so pleased to be able to come back and help you celebrate the 100th anniversary of this remarkable institute. So as Suzanne said, what I'm going to talk about is RNA, non-coding RNAs, and specifically viral non-coding RNAs. And this is simply because some mammalian viruses, certainly not all, do make non-coding RNAs. Just like host genomes make non-coding RNAs. And the question in both cases is, what are they doing? In the case of viruses, what one might think is that they're doing something either to enhance the viral life cycle or to help counteract the defenses that the host organism raises against viral infection. But there are a couple of other things that I think are important to sort of keep in mind when you're thinking about viral non-coding RNAs. One is since viral genomes are short, if they're going to expend precious space making a non-coding RNA, it ought to be doing something pretty important for the virus. Another thing is that viruses are constantly exchanging genetic information with their host cells over evolutionary time. But it turns out that just because something that you find in a virus looks like something you think you know about in the host, it isn't always true that they function in the same way because viruses are very clever at picking up these pieces of host cell genetic information and then changing it a lot to the ends of the virus. And finally, as with all studies of viruses, you eventually come back to the host cell and you see evolutionary relationships that are very enlightening. So I like to start— OK, I think I might have the wrong one here. Let's see if this works better. I always like to start talking about non-coding RNAs with this wonderful slide from the Australian bioinformaticist John Maddick. And what he's plotted here is the amount of non-coding DNA in various genomes, and then these are all organisms. And this is plotted as a function of something called developmental complexity, which is sort of a mystical term, but I think you all sort of get what's meant by it. So all the blue organisms down here are single-celled bacteria, eubacteria, archaeobacteria, et cetera. And notice that they only have up to about 20% non-coding DNA. Black organisms here include yeasts, single-celled eukaryotes. Up in this range, we have invertebrates and plants. And I think as you probably know, when we get up to vertebrates and mammals and humans, we have as much as approximately 98% non-coding DNA. But the really stunning realization about this huge amount of extra DNA that's come just within the last 10 years, I think, is that it appears that at least at some time in the life of the organism, all of this DNA is made into RNA. So what that suggests is that even if these RNA molecules are very low in abundance or perhaps very short-lived, they still have the potential for contributing to signaling and regulation in important ways. We have known for many years, or the information has accrued over many years, that there are many non-coding RNAs that play very important regulatory roles in gene expression. And some of them are illustrated here, the most abundant ones that we know the most about. For instance, they're the sRNAs in the spliceosome that are the RNA components of the SNRPs that identify the ends of introns. Set up the spliceosome to excise introns to make mature messenger RNAs. There are non-coding RNAs involved in transcription, such as the 7SK RNA that's essential for transcription elongation. In the nucleus, there are many small nuclear, or snoRNAs, that are involved in putting specific modifications into ribosomal RNAs. In germ cells, we have piRNAs that are, at least in part, responsible for keeping transposons from moving around and wreaking havoc with genomes in the germ cells. And now there's an ever-growing set of long non-coding RNAs that includes things like axist that clearly recruit things to the chromosomes that are then involved in sort of more extended control of larger regions of expression from chromatin. Why does this always happen? I didn't press anything. Oh, it's them. Oh, OK. Oh, they're just telling us how to get out, right? In case need arises. OK, let me go ahead then. OK. So when we go out to the cytoplasm, of course, we have the classical non-coding RNAs involved in translation. We have the experimentalist tool of using small interfering RNAs to knock down gene expression. And we have this fabulous huge class of so-called microRNAs, little 20 to 22 nucleotide long RNAs, of which we have probably up to 1,000 different species in our cells that are involved in regulating at the post-transcriptional level the output of cellular messages, mostly by, along with associated proteins, binding to the 3' untranslated regions and controlling both translation directly and deadenylation, so the level of the mRNA. So this is obviously just the tip of the iceberg. And the other thing I want to point out is that in all these cases, we know that these non-coding RNAs act very— in association with very tightly bound proteins that are essential to their function. So we're really talking about RNPs, ribonucleoproteins, rather than just RNAs. So the non-coding RNAs from viruses that my lab has been studying for quite a number of years now come from the gamma herpes— whoops. OK. Come from the gamma herpesviruses. And here are listed the names of the viruses. Here are the names of the RNA or the RNP. And over here is listed how we're coming on finding functions. So gamma herpesviruses, as you all probably know, infect lymphoid cells, B or T cells. They have large double-stranded DNA genomes encoding maybe 100 to 200 gene products. They are all oncogenic. And importantly, they all have both latent phases to their life cycle, where the genome disappears inside the cell and only expresses a few functions, and lytic phases, where they're induced to go into replicating the virus. And what you can notice from looking at the year of discovery of some of these non-coding RNAs versus all the question marks that are left over here is that finding functions is very challenging. It takes lots and lots of years. But recently, there's sort of been an upturn in what we've been able to figure out on the molecular level. And I think it's just that technology has finally caught up. And now we can do things to find out about associations of these non-coding RNAs that we previously couldn't do. So we're not just looking for a needle in a haystack. We have some direction. In which we're searching. So obviously, I'm not going to talk to you about all these different non-coding RNAs. I'm essentially going to tell you one story. But first of all, I just want to say the other places that we've made some progress recently. One surprise was with these RNAs from herpesvirus samurai, where they look just like splicing SNPs. It's absolutely obvious that where the virus got them was to adopt them from the host cell. But in the one case where we really know about the function, it turns out that the snR binds a host microRNA, causes it to be degraded. So what the virus has done is taken something that was used for splicing in the host and repurposed it into a degradation machine in the case of the virus. And then what this particular microRNA does is it's involved in downregulating all sorts of genes that are involved in T cell activation. And by getting rid of this microRNA, then the levels of these genes go up slightly. And that's beneficial for the viral latency. Keeps the cells percolating along so that the virus can survive over time. In another case that I'm not going to tell you more about, we don't know the actual function of this long non-coding RNA. But by studying it, we figured out a way that nuclear RNAs stabilize themselves, which is by grabbing their poly-A tails and forming a triple helical structure in the RNA. And that this seems to be a way that now many RNAs also in host cells stabilize themselves over long periods of time in the nasty environment of the nucleus. So the story I am going to tell you about has to do with Epstein-Barr virus and one of these RNAs. It's called EBER2. And the reason that I want to talk about it here is, first of all, I waited 33 years to be able to tell anybody what the function of this RNA was since its discovery back in 1981. And we have made real progress. And secondly, when I was here on sabbatical in 2000, it turns out I tried to do some work on this with Jerry and Suzanne's blessing and also with lots of help from David Wang. What I tried to do didn't really go anywhere, but it led to where we are today. So the next slide. OK, this is the take-home lesson of what I'm going to tell you about, is that this RNA does something sort of strange. It directs a host transcription factor to its binding sites on the viral genome. And it does this in an unusual way. So it's surprising what it does. It's surprising how it does it. And as I'll talk about at the end, it sort of presents a new paradigm for thinking about a different way that long non-coding RNAs might function in cells. And this is the work primarily of a very courageous postdoc in the lab, Nara Lee. Who's had help from Walter Moss, who's a bioinformatics structural RNA structure sort of person, and Tracy Ario, a technician. So let me start by giving you a little bit more introduction to EBV. I'm sure you all know that most of us are infected with it, whether or not we know we had mononucleosis as teenagers. It hangs out in our memory B cells. And you also probably know that cancers can develop from cells that have been infected with EBV, several different types of cancers, lymphomas, carcinomas, et cetera. What's shown down here is, I already mentioned, there's double-stranded linear DNA in the viral genome. But when the EBV infects cells, it goes into latency and the genome circularizes. And then if it's stimulated to go lytic, then it actually replicates by what's probably a rolling circle type of mechanism. But then the RNA— or sorry, the double-stranded DNA is again linearized to go back into the virus. The action with respect to the circularization and the relinearization takes place at what are called the terminal repeats. And what these are are 5 to 20 copies on both ends of the linear double-stranded DNA of a tandem repeat of about 500 base pairs. So that's important because that's where we're going to end up focusing during what I'm about to say. OK, so the RNA I'm going to be telling you a little bit about the function of is ebra2. This one here. It and its partner, EBER1, which we still don't know very much about what it really does. They're about 170 long. They're RNA pol III transcripts, remarkably high in abundance, about a million copies per cell. So this is like the same number of SNPs that there are in the nucleus. It's 1/10 as many as there are ribosomes in the cytoplasm. Very, very abundant. Work in other labs has suggested that if you put these into certain types of B cells, they have a tumorogenic phenotype. And one thing that's perfectly clear is that they're strictly nuclear. That's been clear for many years. Obviously, very early on, people deleted these from the EBV genome. They sit over here and looked for phenotypes, and they weren't able to find any obvious phenotype. Which is why we've been looking for a function for so many years. OK, let me go on then. OK, so the method that made a difference, the technology in the case of EBER2, is called ChART, which stands for Capture Hybridization Analysis of RNA Targets. And it was figured out by Matt Simon, who's now an assistant professor in our department at Yale, but he did this work when he was in Bob King's Joshua Elkington's lab at Harvard as a postdoc. And ChIP is basically just like ChIP. You all know what ChIP is. You identify where a protein is sitting on chromatin. But here we're going to identify where an RNA is sitting on chromatin. And the idea is you start out the same. You formaldehyde cross-link the chromatin. You shear it down to pieces. But then instead of coming in with an antibody against the protein you're interested in where it's sitting, You come in with an oligo selection oligonucleotide hooked to a bead that then hybridizes specifically to your RNA of interest, and you pull out the DNA fragments that are associated. And then you work them up for sequencing. You do deep sequencing. And from that, you can figure out where this particular RNA was sitting on your chromatin. And so when our— OK, so the prerequisite to doing this is that you have to have some sequence on your RNA that's available to hybridize to your selection oligo. So this is the standard experiment that one would do to figure that out. You simply take a latently infected EBV extract that has EBRA2 RNPs in it, and you design DNA oligos that scan along the genome. You pop this into the extract. There's endogenous RNase H there that cleaves only the RNA strand of a DNA-RNA hybrid. And you look for the ones that cleave. And that tells you that this region and this region are available for interaction with an exogenous oligonucleotide, in this case DNA. So this then was the selection oligo that Nara chose. In order to do the pullout, which is actually fortunate because we're going to see that a lot of the biological activity is in this other exposed region of the RNA. So let me go on. So when Nara did the experiment, I really expected he was going to see Ebertu sitting all over the host chromatin in various interesting places. And that was what we would be analyzing. That's not here. But basically, what he found was it wasn't sitting on the host chromatin. Here is the background signal when the experiment was done on uninfected, non-EBV-infected B cells. But in cells that are latently infected with EBV, you see a fair amount of low signal all along, but then a huge peak at the end in that terminal repeat region. And if you blow that up, then what it looks like is you get these peaks over each one of these half-kB terminal repeat sequences. And I have to say right here is, of course, we don't really know that EBR2 is sitting on each one of these terminal repeats. It could be all sitting on one. But this is traditionally the way— repeated units are looked at in the sequence. And we'll talk a little bit more. It's just adjacent to this gene, which is an important oncogene of the virus, and sort of in the middle of this gene, in the first intron of this gene, which is important for maintaining latency. And this got even more interesting when Nara looked in the literature. And discovered that, in fact, a couple years earlier in 2012, Paul Lieberman's lab at UPenn had published a very similar profile, this time for a host transcription factor called PAX5 and where it sat on the EBV genome, exactly in the same place. So if they're both there, are they actually physically associating? Oh, first of all, what is PAX5? And a lot of you in the audience probably know a lot more about it than I do. It's a very, very important transcription factor in B cells. It's important for the commitment to the developmental pathway. It acts at various places along the pathway, both in a positive and negative way. And then it's also important for the functioning of B cells when you finally get there. And when the Lieberman lab figured out that Pax5 was sitting on these terminal repeats on the genome, of course, one of the things that they did was to look and see what would happen if you knocked down Pax5 in the cells. And what they found was that some of the latent EBV genes went up, some went down. And we'll come back to this later. So if these two things are in the same place, are they actually physically interacting? And we had the tools to do that, obviously. We could take immunoprecipitate with anti-PAX5 and ask, does EBER2 come down, which it does. Or we could use our selection oligo to pull out EBER2 and ask, does PAX5 come down, which it does. This doesn't say that the interaction is direct, and we in fact think it's indirect for several reasons that I could tell anybody who's curious about it and wants to ask. But we think it's an indirect interaction because of negative results. But in any case, they're clearly associated. So if they're associated physically, are they also functionally associated? And here it turned out, starting with work from the Lieberman lab, that they had shown that if you knock down PAX5, some of these genes right in the region of where PAX5 is sitting on the terminal repeats, in fact, go up. I pointed out the LMP1 and 2A genes. And actually, if I go on down here, here they are. So the terminal repeats are in the first intron of this gene and this gene. And these are adjacent. And what they had found was that if you knock down PAX5, that the amount of expression of these genes go up, and the same thing happened here. So the other thing that this predicted was that— let me see what this is. Yes, okay. So if you do PAX5 ChIP, then what you actually see if you knock down EBER2 is that there's less PAX5, sitting at the terminal repeats, whereas the amount of PAX5 sitting at the promoters doesn't change. So that's knocking down EBER2 and asking what happens to PAX5. If you knock down PAX5 and ask what happens to EBER2, we got a surprise. So here is just knocking down PAX5, showing again that we can get this gene in the vicinity to go up. Were knocking down Pax5 sitting on the terminal repeats. But what was surprising was the amount of EBER2 sitting on the terminal repeats did not go down. It was unaffected. So what that meant was that one of our two models for what might be going on, and what seemed to us the most likely, was in fact not happening. What we thought was most likely, if we had RNPs that that had in it PAX5 and EBER2, and it confronted the terminal repeats, which in fact have two very good PAX5 consensus sequences in them, that PAX5 would be bringing EBER2 to the terminal repeats. But when we knock down PAX5, we in fact do not see an effect on EBER2 in the terminal repeats. It's the opposite. So the model that we then came up with, a sort of speculative model, was that maybe what's happening is that EBR2 is somehow interacting, here we thought, with nascent transcripts coming from the region that would include the terminal repeat sequence. And that's bringing PAX5 to bind to its consensus binding sites. And we reasoned that if that were true, we should be able to predict this other RNA-RNA interaction between EBER2 and the terminal repeats. We should be able to block it and get the same phenotype as if we knock down EBER2. We should be able to show that transcription through the region has to occur in order to get the interaction. And we should be able to get physical evidence for the RNA-RNA interaction. And I want to just very quickly go through with you the data for these. We brought in Walter Moss, our bioinformatics person, and he used a program called RNA-UP to look for potential complementarities between EBER2 and this 544-nucleotide sequence of the terminal repeat and came up with this. And notice that this sequence is very close to one of the PAX5 consensus sequences. But it's on a different molecule. This is the RNA-RNA interaction, whereas this would be a DNA-RNA interaction. Nara then used an antisense morpholino oligo to block this sequence and ask what happened. And it had the same effects as actually knocking down EBER2. So you can just block the interacting sequence and get the same phenotype. He also used the CRISPR system, which it turns out— you all know it can be used to make mutations. But Jonathan Weissman's lab figured out that it can also be used to block RNA polymerase II transcription. And you do that by making mutations in the active site of Cas9 so that it doesn't cut the DNA, but the complex directed by the guide RNA you put in then just sits there and serves as a block to transcription. So he made guide RNAs to both of the upstream promoters for LMP2A and 2B, and showed that they, in fact, very effectively knocked down transcription. And when that happened, he, in fact, knocked down the amount of eBER2 that was associated with the terminal repeats. What?

2 minutes.

2 minutes. OK. So soroling cross-linking is what's used to show that you actually have an interaction. And I won't go into the details, but Nara was able to show that, in fact, there is an RNA-RNA interaction. So what I've shown you then is that these predictions that we had are, in fact, And this seems to be the mode of action. But why is the EBRA2 RNP sitting on the terminal repeats? What's that doing for the virus? And here, because the terminal repeats are so important to the DNA replication, Nara checked by knocking down EBRA2 what happened to DNA replication. This is a Southern blot. This is, in fact, lytic replication. Replication is down, and the amount of DNA that goes into the virus is much more— oh, gosh— is much more potently down, mimicking again what happens if you knock down Pax5. And this is important because the clinicians tell us that prior to the onset of an EBV-linked tumor, they almost invariably see a round of EBV lytic replication. So this could be a tie-in to the oncogenic potential of the EBVs. And finally, I just wanted to say that if this is true, you expect it to be in related EBVs that infect other primates. This one— whoops. Let me go back. OK. So this is a rhesus monkey that has an EBV-2 that's only 65% homologous to EBV-EBV-2. And here, the terminal repeat is 900 nucleotides long. And there is, in fact, a complementarity that maps to approximately the same place in the RNA. So finally, let me just close by saying, in addition to the models that people had previously for what long non-coding RNAs were doing— being a scaffold, bringing in proteins, holding proteins together, as in the case— I always like to use telomerase as an example of this. The tethering model exists. We think that there might be additional RNAs that would— oh, sorry. Additional RNAs in host cells, non-coding RNAs, that might use this mechanism of RNA-RNA interactions to bring protein factors to specific places on chromatin. And I think I'll end by just leaving these questions up here. Maybe people will— ask about this during the question period, but let me first of all thank Nara Lee and Walter Moss and people who've worked over many, many years on this problem, our funding sources. And let me go back to the open questions and hope that maybe we can discuss these.

Thank you, Joan, for that really elegant story. I'm sure there are questions here, especially from immunologists.

Yes. Ellen. Where's Ellen? This is actually a molecular biology question. I'm down here, Joan, very close to the— There you are. There you are. I'm very interested in these long non-coding RNAs that also come from super enhancers. Oh, yes. And so one of the questions is whether you think that there's a way that these may be helping to recruit the transcription factor clusters that seem to be preferentially associated They could certainly be doing that. I mean, I have no idea. I have no evidence, but that would be a very attractive thing that they might be doing indeed. And actually, I wanted to talk to you about the fact, you know, why does a transcription factor need an RNP to help get it to its site? Is that because of the distant enhancer effects that you were talking about the other day, and you need to open up the the DNA. And in this particular case, you need to do it in a special way with an RNP. Thank you.

Yes, Steve is next. Steve Knott.

Follow up on a similar type of issue. PAX5, the ChIP-seq for PAX5 shows that it binds essentially every enhancer in a B cell. So it doesn't, you know, B cell have any trouble getting access to everywhere. Right, right. So you'd have to almost invoke a situation where you need preferential binding or like for a reason, I think. Accessibility doesn't work. Right, right. And hidden in my slides were the fact that when we knock down the EBER2 RNP, we don't affect the binding of PAX5 to a bona fide PAX5 binding promoter. I don't know, but one of the interesting things that— another interesting thing is that knocking down PAX5 or EBER2, in fact, increases the expression of the genes in the neighborhood, whereas you might expect it to go the opposite direction. And the one parallel with that is the work of Thomas Genuine on repeated sequences in the mouse genome signs where he finds that isolated transcription factor binding, which this would be, and in his case even of other PAX5 family members, in fact, is repressive to the chromatin. So lots of things are going on that we don't understand very well. Lots more to learn.

I'm afraid we're going to have to move on because time is running out. And I could suggest that over drinks later, all of those people who want to continue this conversation do it then. Can you join with me in thanking Joan again? Thank you.