Hi. My name is Rachel Green and I’m at the Johns Hopkins University School of Medicine and in the Howard Hughes Medical Institute. What I’m going to share with you today is a story from my own lab focusing on how the ribosome, in eukaryotic cells, identifies bad messenger RNAs to trigger a series of events in the cell that lead to the degradation of the incomplete protein product, to messenger RNA decay, and to the recycling of the ribosome complex. So, which messenger RNA defects in the cell might you like to monitor? What could be the problem with a given messenger RNA in a eukaryotic cell? As we remember from learning about mRNA processing, we know that eukaryotic messenger RNAs come out of the nucleus with a cap at their 5′ end and with a polyA tail at the 3′ end, and these are key signatures on a messenger RNA in eukaryotic cells that tells a ribosome that this is a good message and that this one should be translated. Also, we know that the typical messenger RNAs have a start codon that signals the beginning of an open reading frame and a stop codon that signifies the end of an open reading frame, and what the cell would like to do is start at the beginning and go to the end and make a full-length protein. So, what can go wrong? Well, there can be lots of events that take place in RNA processing or with mutations that would result in a messenger RNA that might actually not be optimal. One example that many of you may have heard about is, for example, if a messenger RNA contains a premature termination codon. That is, instead of having just a stop codon at the end of the open reading frame, they, by mutation of some of other process – perhaps a mis-splicing process – they have a termination codon that comes up early in the open reading frame. Amazingly enough, eukaryotic cells have a mechanism in place to determine if a termination codon is the right termination codon or the wrong termination codon. And they trigger this sequence of events that ultimately leads to messenger RNA decay. Another example of a possible problem with the messenger RNA would be an endonucleolytically-cleaved messenger RNA. If the messenger RNA picked up a cleavage in the middle of, the ribosome would never reach a stop codon at the end, and the ribosome would be stuck in the absence of a stop codon and the ability to promote the normal termination process. So, this is a messenger RNA that the cell would like to get rid of, and not use again. It would also like to get rid of that incomplete polypeptide product. A final example that you can imagine might happen is that the messenger RNA might not actually have a stop codon, through some mis-splicing event or some mutation, and in that case the ribosome might actually get all the way through the 3′ untranslated region of that gene and encounter a polyA tail. PolyA encodes lysine, and what’s thought to happen is that the ribosome senses that it’s making lysine-lysine-lysine-lysine, and it triggers a process of decay and rescue in the cell that we refer to as non-stop decay. So, Nonsense-Mediated Decay, No-Go Decay, and Non-Stop Decay are probably three related processes that are all aimed at keeping under control in the cell u nproductive messenger RNAs, getting rid of the messenger RNAs, degrading the protein product, and rescuing the ribosomes for subsequent translation events, and that’s what the focus of today’s lecture will be on. So, how did we become interested in these processes of decay, of mRNA surveillance? We became interested because we noticed in the literature an interesting observation. For a number of years, we had studied the termination process in eukaryotic systems, whereby a stop codon is recognized by termination factors in eukaryotic cells to release the growing polypeptide chain. And we had studied these factors and how they work biochemically for a number of years. There was a publication that came out of Roy Parker’s lab, however, where Roy’s lab put into a messenger RNA in yeast a long stem-loop structure – it was 34 basepairs in length, this stem – and in fact it was such a large structure that no ribosome could possible push through it easily. And what Roy’s lab observed was that this messenger RNA was targeted for decay in the eukaryotic cell in a process he called No-Go-Decay, because the ribosome can’t get through it, but the key observation in this study was that this process of mRNA decay was dependent on two proteins, Dom34 and Hbs1, which turned out to be homologues of the eukaryotic termination factors eRF1 and eRF3 that we had been studying, and we had been studying them biochemically. So, this seemed like a key insight, because what it suggested was that recognition of a bad messenger RNA is in fact ribosome-driven, as you might suspect, because the ribosome must be the enzyme, or the catalyst, or the macromolecule, that decides whether a message is good or not. It seemed like a very logical idea that the ribosome is what determines whether a messenger RNA is good or bad. Should I translate it or should I not? And that’s where our work started. We knew that Dom34 was a homologue of eRF1 and that’s highlighted by this structural view here. There were structures know of eRF1 and Dom34. eRF1, in recognizing stop codons, has a codon-recognition motif down here called the NIKS domain, and at the top, interacting with the large subunit of the ribosome, is a tri-amino acid motif, a GGQ motif, that’s responsible for the catalysis of peptide release. We see that Dom34 has similar domains. This domain is clearly related to this domain – you can see that in terms of the beta sheets and the alpha helices. However, it’s missing this domain with the GGQ motif that would be responsible for releasing the growing polypeptide chain. We see that there’s also no codon recognition motif, but the other feature they have in common is they have this domain, here, that’s known to interact with the GTPases that are essential to the termination step. So, this was our starting point. We had a protein that looked like it was related to a termination factor, and we suspected must engage the ribosome, and we had the biochemical tools that we wanted to use to ask questions about what Dom34 might do in order to bring about that messenger RNA decay that Roy Parker’s lab had observed. So, with these ideas in mind, what we did is we incorporated Dom34 and Hbs1, these homologues of the termination factors, into what was in place in our lab, which is an in vitro reconstituted translation system from the yeast Saccharomyces cerevisiae. In our lab, in the freezer, we had purified components of the ribosome, the small and large subunits of the ribosome. We had tRNAs, messenger RNAs, we had five key initiation factors that John Lorisher’s lab had shown to be essential for in vitro initiating on an AUG codon and getting that initiator tRNA into the ribosome, we had elongation factors, we had termination factors, and to that list we added Dom34 and Hbs1. What such a reconstituted system allows us to do is to form ribosome complexes with various messenger RNAs specifying various peptides and other steps in the translation system. We can place, for example, a stop codon into the aminoacyl site of the ribosome to be recognized by a factor, or we can place a sense codon, and we can follow the activity of these complexes. Shown here is a blue ball for the growing polypeptide chain, which is to signify that we can label the tRNA substrates, we can label the tRNAs, we can label the amino acids, we can label the messenger RNAs, we can label the ribosomes, and we can follow the fates of various components in the system to ask about biochemical reactions. And so that’s what we did, and I’m going to show you one example of a biochemical reaction that we performed to learn something about Dom34 and Hbs1 function, understanding that these factors really had just been implicated in mRNA decay and, really, it wasn’t known what they might do on the ribosome. So, what we did in this case is we used our biochemical system to reconstitute two different ribosome complexes. For the one on the left, what we’ve done is we’ve formed a complex, a dipeptidyl-tRNA, where we have two amino acids on a tRNA in the P site – we’ve gotten there by initiating and elongating for one round of protein synthesis – and in the A site we’ve placed a stop codon, but it ends up not mattering. And in this complex we see that the amino acid is labeled. So, on the methionyl-tRNA, the methionine carries a radioactive label. We formed a similar complex on the other side, here, but in this case the tRNA itself is radioactively labeled and we can follow the tRNA rather than the amino acids. In this case, what we did is we took these complexes, these large ribosomal labeled complexes with labels in different places, and we ran them on a native gel. A native gel allows them to run more or less intact with all their components there. And we asked where the label went. So, at the top of the gel, the intact ribosome complex runs here at the 80S position. And what we see is that in the absence of any factors it runs as a big 80S complex. However, when we add eRF1 and eRF3 – the known termination factors – exactly what we would expect happens, which is, in the case here where we’re following the peptide, we release the peptide, the dipeptide Met-Phe. Whereas on the right here, where we’ve got a tRNA that’s been labeled, what happens is the tRNA is generated here – this is a deacylated tRNA lacking the amino acids on it. So, this is the released product with termination factors. We next asked what happened when we add Dom34 and Hbs1, these factors that seem to play a role in mRNA surveillance, but we don’t know what they do on the ribosome, and we saw the appearance of a different band, here – not a peptide band – and a different band, here – not a deacylated tRNA band. And what we reasoned at the time when we saw a band that was common to these two different complexes, but different from either the result of the reaction with eRF1 and eRF3 is we reasoned this might be a peptidyl-tRNA. That in fact what was happening was that Dom34 and Hbs1, when interacting with these ribosome complexes, was actually resulting in the release of the entire peptidyl-tRNA complex, and we were able to confirm that by adding a different reagent known as peptidyl hydrolase, which releases, then, free peptide or free tRNA. So this was a key biochemical reaction that we used to identify a novel activity for Dom34 and Hbs1 proteins, and this activity suggested that part of what these proteins were doing is they were recognizing ribosomes and they were facilitating some destabilizing event that led to the release of a peptidyl-tRNA from the ribosome. So, this was the beginning of a biochemical story and this is sort of the summary of that story. What we found over time was that a ribosome complex, when presented with Dom34 and Hbs1… what, actually, Dom34 and Hbs1 did was effectively separated the ribosomal subunits. The peptidyl-tRNA tended to partition with the large subunit, depending on the length of the polypeptide chain. What we found was that this reaction by Dom34 and Hbs1 was codon independent. That was consistent with the structure that I showed you several slides ago where there’s no codon recognition domain on this protein, Dom34. We saw that peptide was not released from the tRNA. That’s consistent with the fact that the Dom34 protein lacks the GGQ motif that’s responsible for that activity in termination factors. And a little aside is that these proteins prefer to split subunits when the messenger RNA template, here, tends to be short. So, if the messenger RNA is short from the 3′ end coming in to the wall of the ribosome, we found that in fact that was a preferred biochemical substrate for these proteins. So, how does all this make sense given what we knew about Dom34 and Hbs1 in the process of so-called No-Go-Decay? Well, over the same time period that we were performing biochemical reactions there were a number of labs studying this in vivo, and they had come up with a number of important insights, in particular, Toshi Inada’s lab. And it fit in with models he had presented, which is the idea is that the ribosome gets stuck at a particular place in a messenger RNA, maybe, for example, a big stem-loop structure. There’s an endonucleolytic cleavage that actually happens behind the ribosome, and that was documented in Toshi’s lab, leading to an endonucleolytically-cleaved messenger RNA. And given that there are many ribosomes on any given messenger RNA, what that means is all the ribosomes that come behind that endonucleolytic cleavage site encounter that endonucleolytically-cleaved end, and that these would be ideal targets for these proteins, Dom34 and Hbs1. In the absence of these factors, the ribosome is stuck on the message with no way of getting off. And what these factors might allow is for recycling of these ribosome complexes on defective messenger RNAs, leading to the recycling reaction and the reutilization of these ribosomes. So, our biochemical data fit in nicely with other results in the field that were taking place in intact cells. And so those were our thoughts, and so we wanted to ask next a more general question. Knowing the biochemical activity of these proteins, and knowing a little bit about how these proteins behaved on reporters in various cells, we wanted to ask questions about the general biological significance of a rescue response. Which messenger RNAs are typically targeted in a cell? Is is the same in wild type conditions or in stress conditions? And how can we go about thinking about that? We knew from the beginning that Dom34 is a gene that’s non-essential in yeast, but we know that it’s essential in higher eukaryotes. The gene is called Pelota in higher eukaryotes and is an embryonic lethal in mice. So, it’s an important gene in higher eukaryotes. We did, however, know that the Dom34 deletion in yeast was synthetic with a variety of different things including, for example, the deletion of a ribosomal protein gene. And it’s known that when you delete a ribosomal protein gene, for example, one copy of a ribosomal protein gene, that there are ribosome insufficiencies in the cell, because biogenesis isn’t really able to keep up with what it needs to do. And so the idea might be that if you delete a rescue factor for ribosomes, and put it together with a ribosome biogenesis defect, that that’s when cells really get sick, and so that would explain the synthetic lethality. And so the question that we really wanted to ask next was, “What might the cellular targets for Dom34’s function and Dom34 rescue?” We were lucky because at the time we were thinking about these ideas there was a new approach that had just come online from Jonathan Weissman’s lab and had been developed by Nick Ingolia, a postdoc in his lab. It was a method known as ribosome profiling that allows one to systematically look at the occupancy of all ribosomes in a cell on their various mRNA templates. And we reasoned that if Dom34 was a protein that played a function, in yeast cells for example, if we had a wild type strain and a knockout strain, we might be able to ask what the role of Dom34 is in yeast cells by asking where ribosomes accumulate in strains lacking that protein. So, the basic idea behind ribosome profiling is you can take all the ribosomes that are on messenger RNAs in a cell, and you can isolate those messenger RNAs with ribosomes bound to them. By using a nuclease, you can digest away all the free messenger RNA, leaving just the little bit of messenger RNA that’s protected by a ribosome, and that protected mRNA bit is about 30 nucleotides in length. You can then isolate that 30-or-so nucleotide long messenger RNA from all the ribosomes in the cell, you can submit it for high-throughout sequencing. You get hundreds of millions of reads of ribosome footprints from the cell and you can ask what the occupancy – the global occupancy of ribosomes on messages – is in a cell. You typically couple this experiment with an mRNA seq experiment, where you randomly fragment all of the messenger RNAs in a cell just to make sure you can account for cloning biases and other potential artifacts from a cell, and then you can also ask how efficiently a ribosome will occupy a given messenger RNA because you know how much of each messenger RNA is present and you know how many ribosome footprints you have. So, that was the experiment that we envisioned using to ask about the in vivo role for rescue factors. We began like everybody else, submitting some samples and asking how well the system works, and so this gives me an opportunity to show you what this sort of data look like. What we did is we subjected, from wild type yeast and from Dom34 mutant yeast – missing Dom34 – ribosomes footprints and mRNA-seq data. The first thing you can do with those data is you can ask how they align to reading frames. And you can see, very simply, in this panel at the top, that if take the ribosome footprints they predominantly map to one single reading frame within all the open reading frames, suggesting that ribosomes are moving three nucleotides along at a time down a messenger RNA template, and that this methodology is able to capture that three nucleotide movement. There’s a periodicity to the footprint signature. By contrast, you can look at the mRNA-seq data and see that the reads distribute to all three frames, because the ribosome isn’t imposing any periodicity onto those frames and so they distribute equally over all reading frames. So that suggested that we were capturing something relevant to translation and three nucleotide movement along a messenger RNA template. You can see that’s true along the open reading frames as well; if you align all of the start codons, here, to the beginning of the gene, and you align all the stop codons and you do what we call an average gene or metagene analysis… first of all, what you can see is that the mRNA-seq reads, in green, distribute all along the open reading frame, as well as in the untranslated regions of the gene, out here, and at the 3′ end of the gene. So, these regions distribute all along the message. By contrast, the ribosome footprints distribute just along the open reading frame, beginning at ‘start’ and ending at ‘stop’. And in fact in these data here, you can see that three nucleotide periodicity emerging. So, we felt very confident that this was a method that might give us some signature of ribosome activity in the cell that might tell us something new about the function of these rescue factors. I’m going to tell you that we added another little touch to these studies to extend the sort of analysis that the full-length ribosome footprints might allow us to think about. What we reasoned is that, if there were big impediments in the cell that prevented the ribosomes from getting through, there might be not just one ribosome stacked on a messenger RNA, but two. So, we call these disomes, and you can actually… after RNAse treatment, we see this black trace here… we can actually see a little bit of a residual disome peak and so we isolated that disome peak as well, reasoning that big stalls in the cell that were a problem and might be subject to rescue might be enriched in that peak. We also were generous in cutting our mRNA fragments from a gel because we knew that we might… based on the biochemical activity of Dom34 and Hbs1, we reasoned that short mRNA footprints might actually be also biased in their preference by Dom34 enzyme, because we know that Dom34 likes short messenger RNA templates. So, we isolated shorter and longer footprints as well, so we were generous in slicing our fragments from a gel. So, what do those data look like? If we take, from a wild type strain or a Dom34 delta strain… if we take those footprints, we send them out for high-throughput sequencing, and ask what the average length of the fragments that you obtain, for example, from a single ribosome, from a monosome peak, what do those fragments look like? We can see the distribution, here, where we look at read length on the x axis and we look at the fraction of reads having that read length on the y axis, what we see is that a majority of our reads are in this range from 28-30 nucleotides in length. That was the original fragment length studied by the Weissman lab, and that’s what we think is a full ribosome with a full-length messenger RNA bound to it. We actually see a little peak, here, at 21 nucleotides in length. This is a fragment that’s been characterized by Liana Lareau, and what she’s identified this as is probably representing a rotated or ratcheted state of the ribosome in the process of translocating along a messenger RNA, and that won’t be the focus of any further discussion here. Finally, the fragments that we were most interested in are these truncated fragments down here that peak around 16 nucleotides in length, and we believe, and I’ll show you evidence to support that, that these are short messenger RNA fragments bound to ribosomes that have run to the end of a messenger RNA template, and therefore there’s only 16 nucleotides there, but these are principal targets of Dom34 action in the cell. So, what do the data look like? You get 100 million sequencing reads from a wild type and mutant strain and you ask, how do they distribute over your transcriptome? How do they distribute along a messenger RNA? So, we can look at an abundant gene, here, PGK1. It’s just a garden variety gene. We can see that the ribosome reads distribute all along its length, from the beginning of the open reading frame to the end, and what we see is that in a Dom34 knockout (KO) strain the pattern is really very equivalent. This is what typical ribosome profiling data looks like. It’s bouncy. Included in this bounce is probably some natural pausing by ribosomes, as well as sequencing biases, and cloning biases, and amplification biases. But what you can see is in terms of what we were interested in knowing is in this particular gene there was no obvious disruption… there were no obvious pauses in this gene, or problems, that led to Dom34 action. There were no piles of ribosomes in the absence of Dom34 that accumulated. As we saw these data, we began to worry about whether we actually had the tools in place, however, to observe significant pausing, and so we did a nice trick. We put in a histidine analog into yeast known as 3-AT – 3-aminotriazole – that actually blocks biosynthesis of histidine and allows you to actually… you’re deficient in histidine in the cell and so we anticipated that there would be pauses, ribosomes piled up at histidine codons throughout the genome. And in fact that’s exactly what we saw. In a wild type or in a Dom34 knockout strain, all of a sudden you see piles of ribosomes exactly where there are histidine codons throughout every gene in the genome. So, this was a really nice indicator for us that we understood something about ribosome functions – ribosomes should pause at histidine codons in the absence of histidine – and really, though, as we anticipated, Dom34 doesn’t respond to general amino acid starvation, and we didn’t anticipate that it would, but it gave us a feeling for the type of result that these data could yield – that we could understand something about pausing and we might be able to decipher something about Dom34 function. So, with those tools in hand, we began to look more globally at our transcriptome and ask, were there any genes on which there was an increased number of reads in the Dom34 deletion strain relative to a wild type strain. The way that you can do this experiment is you can take all the reads on a given gene and you can plot it – the number of reads by mRNA-seq in the knockout versus the wildtype strain, and put it on the x axis, or the difference in the number of ribosome footprints on a given gene in the knockout strain versus the wild type strain – and what you can see is a very uninteresting pattern from the perspective of a high-throughput experiment, where almost all of our genes have no interesting pattern differences. They all come out around one, which means in a wild type and a knockout strain we have approximately the same number of reads. That’s both by mRNA-seq, so Dom34 maybe isn’t a decay factor, and it’s also in the occupancy by ribosomes. There is one exception, I’m gonna get back to it. It’s a gene known as Hac1. It’s a transcription factor that’s really very interesting and ended up being our big positive that told us, we think, how Dom34 functions in the cell. I’ll show you another similar plot, however, that showed us something new and exciting about Dom34 function in the cell, and it had to do with those short reads. So, in fact, in the previous slide what I was focusing on was full-length reads that are generated in the Dom34 deletion mutant, the 30 nucleotide long standard ribosome reads. And in fact that’s what’s shown here on the x axis, which is, if we look at a wild type or a Dom34 knockout versus a wild type strain, that the long full-length reads are pretty much the same throughout all the genes in the genome. There’s no differences in their distribution. However, if we look specifically at the short reads, the 15-18 nucleotide reads in the Dom34 versus the wild type strain, this distribution skews quite dramatically, suggesting that in a Dom34 knockout strain ribosome occupancy on those short-mer fragments is significantly enriched, consistent with the idea that Dom34 is specifically responsible for clearing those ribosomes under normal circumstances, and that when Dom34 is missing you’re actually enriched in those short-mer reads in the Dom34 knockout strain. So, that was an exciting idea and it was consistent with our biochemistry and the ideas we had going into this project. So, we next actually… what I want to show you next is what we learned about the HAC1 gene and our increased ribosome occupancy in the Dom34 knockout strain, because it ends up being a very clear, positive effect of Dom34 in yeast, and it tells us something about Dom34 function. But first, I need to tell you a little bit about the HAC1 gene. The HAC1 gene… it encodes a transcription factor that’s involved in the unfolded protein response that’s been extensively studied by Peter Walter’s lab, and in yeast it turns out it’s the single gene that in yeast cells is exported from the nucleus with an intact intron – it’s not spliced by the spliceosome in the nucleus. In fact, it’s spliced in the cytoplasm by an endonuclease, a non-canonical pathway, that’s known as IRE1, specifically when IRE1 endonuclease is activated by unfolded proteins in the endoplasmic reticulum. So, what that means is that in the cytoplasm of yeast cells sits an unspliced gene that looks like this, the HAC1 gene, with a start codon and a stop codon. And it’s waiting for a signature to be cleaved by this unusual splicing pathway. But what we know is in fact that there’s some constitutive splicing that happens at the 5′ splice site, and we know it’s unproductive because this has been documented, and we know that there’s some amount of this sort of truncated messenger RNA sitting around in the cell even in normal cells. And to us this looked like it might be a No-Go substrate. It’s an open reading frame that ends without a stop codon and might need to be rescued by Dom34. And in fact that’s true. We can look at our ribosome profiling data aligned specifically along the HAC1 gene, and we see at the bottom of each set of two, here, the wild type reads and the wild type ribosome footprints, and above them the Dom34 knockout footprints, and what you see is that both for the disome reads – so those peaks that we isolated for two ribosomes in a row – as well as the short reads – the 16-mer reads of a single monosome peak, so just one ribosome – we see we are significantly enriched in the Dom34 deletion strain relative to the wild type strain, consistent with the idea that ribosomes indeed are piled up at this endonucleolytically cleaved junction, so we’re missing the intron, now, because this is endonucleolytically cleaved. The ribosomes are stuck. Under normal circumstances, they’re cleared by Dom34, and in a delta-Dom34 strain you accumulate ribosomes at these positions. So, this was a beautiful signature of the function of this protein in the cell. Moreover, when we looked, actually, upstream of that paused set of ribosomes, we saw another signature of Dom34 activity in the cell that looked very similar: a set of short reads in the Dom34 delta and a set of disome peak reads in the Dom34 delta, consistent with the idea that in fact when the ribosome is paused at this first position, through a normal endonucleolytic cleavage by IRE1, there’s another cleavage, presumably by the endonuclease that’s typically involved in this process called No-Go-Decay, it’s a second endonucleolytic cleavage that leads to another set of ribosomes piled up that get cleared by this system. So, this was, for us, a beautiful example of what Dom34 does in the cell on this single gene, and in retrospect we think that this might be consistent with the idea that in normal cells there are very few Dom34 targets that are crisis, that are big, big targets that Dom34 needs to act on. But it provided evidence for the model, which is that when ribosomes pause, endonucleolytic cleavages do happen, and when those endonucleolytic cleavages happen this system comes in and rescues ribosomes, and in the absence of this system you accumulate piles of ribosomes. So, it was really a very consistent set of data for the model for No-Go-Decay, and for what ribosomes do in a crisis. It just didn’t provide us with very many targets. Alright, I’m going to tell you another story, now, that’s related and has to do with the role of Dom34. Dom34 had also been implicated in this Non-Stop-Decay pathway that I described in the first two slides. Non-Stop-Decay, as I mentioned, is what happens when the ribosome encounters a polyA tail and translates poly-lysine. And the idea is that poly-lysine talks to the ribosome, through the exit tunnel, and says, “This is a problem. We need to stop here and rescue these ribosomes, degrade the message, and degrade the protein product, because poly-lysine doesn’t make sense.” And so the model is quite similar, which is instead of running into a stem-loop here, as in Roy Parker’s study, the ribosomes run into a polyA tail, endonucleolytic cleavage happens, and that’s been documented, so again the trailing ribosomes will be cleared by these proteins Dom34 and Hbs1, and we’re not quite sure what might happen to these ribosomes. But what we wanted to know was: was there any evidence of this sort of activity in our yeast cells that we could decipher from the Dom34 data that existed. What I’ll tell you first is that we can actually get some feeling for what iterated lysines in the cell do to ribosomes that encounter them, and we can do that by looking at all of the ribosomes in the yeast genome, or in the yeast transcriptome, and we can actually do sort of averaged gene or metagene analyses where we take single lysines and line them up together, or genes that have two lysines in a row and line them up together, or three, or, or five, and so on. And you can immediately see what’s been described at the level of a genome analysis, which is that multiple lysines in a row in fact do cause the ribosome to stutter a bit – you’ve got bigger peaks here, the ribosomes are stuck – and that kind of propagates along the length of the poly-lysine tract. So, that’s using ribosomes profiling data to ask questions about ribosome occupancy, which are consistent with the idea that poly-lysine is a problem. I should say, though, that these poly-lysines are in the middle of genes – they’re just normal coding lysines. So, we had another way that we thought we could look for what happens when ribosomes encounter poly-lysine, and that was by manipulating the yeast cells in a special way, and I’m going to tell you about that in this experiment, here. So, what you can see, for example, is if we take an individual gene – this gene is known as ENT5, it’s just an average yeast gene – and we can see in a wild type and a Dom34 knockout strain… these are average length reads, these are the 30-mer reads that Jonathan Weissman’s lab originally characterized, and what we see is what we expect, which is ribosome reads in the open reading frame, but not in the 3’UTR and not at the polyA tail, and that makes sense because this is a normal gene with a normal stop codon and ribosomes recognize the stop codon. But the experiment we did is we actually inserted into a yeast strain a suppressor tRNA, and what suppressor tRNAs do is they misrecognize stop codons and they read through stop codons, allowing stop codons to be bypassed at the some level, and therefore translation into the 3’UTR and, in genes that lack another stop codon in the 3’UTR, into the polyA tail. And so what we see here… these are full-length reads we’re looking at now… if we put a suppressor tRNA into a yeast strain and look on this same gene, we read through the stop codon, here, we accumulate ribosome reads in the 3’UTR, but we have enhanced ribosome reads in the Dom34 mutant. So, this is evidence that when you read through a stop codon and read into the 3’UTR and into the polyA tail, these ribosomes, under normal circumstances… this is sitting right on top of polyA… would be cleared by Dom34. And in its absence we have a large pile, so this is something Dom34 definitely does in cells. We were even more delighted, however, when we looked at the short-mer to see signs of this endonucleolytic cleavage that had been proposed for reading into poly-lysine. We see that here when we look, again, in the same strain with the nonsense-suppressor tRNA and the full-length ribosome reads sitting right here at the junction of the polyA tail. We see, trailing behind that signature… we see yellow reads – these are short-mer reads. These are messenger RNAs that are truncated and the ribosome has run to the end of that messenger RNA, and they trail, in fact, this grey peak because the endonucleolytic cleavage, we think, and it’s been shown, happens behind the ribosome that’s in the front. So, this ribosome encounters polyA tail, it struggles with it in some way, the ribosome reads that as a problem, endonucleolytic cleavage happens behind that, and you get an accumulation of reads. Again, identifying a clear role for Dom34 when you read into polyA tail. Does this happen in sort of… what if we look at this in a more global way? Well, we can take this sort of signature and we can do it in a more global or metagene analysis sort of way, where we align, here, at 0, the junction with the polyA tail. And there they are, those are the grey 30-nucleotide long reads. We’ve aligned all the polyA tails together on all the genes that have nonsense codons that are suppressed by the suppressor tRNA. What we see is the pile of grey reads right here – these are the ribosomes reading into polyA tail – and here’s the pile of ribosomes trailing behind on the truncated messenger RNAs that are the result of the ribosome reading this polyA tail. It’s a 29 nucleotide lag and that’s consistent with the spacing that we would expect of a ribosome stuck in front and an endonuclease coming behind. So, this is a global view of what so-called Non-Stop-Decay looks like. The ribosomes are a signature for us identifying Non-Stop-Decay in this Dom34-delta strain. So, you might argue, though, that this is not particularly broad and far-reaching because we’ve put in a nonsense suppressor tRNA into this yeast strain, and that’s maybe not a general phenomenon. But what I’m going to argue is that in fact premature polyadenylation is an abundant and regular event in eukaryotic cells, and the signature of this event is largely erased by a variety of quality control pathways that are always taking place, making the signature of premature polyadenylation invisible by normal experiments. And these are two examples of that. RNA14 is a gene in yeast, and YAP1, that are known to be prematurely polyadenylated. Previous results from Pelechano et al in 2013, in fact using RNA sequencing methods, identified premature polyadenylation sites in these two genes. And what we see is, using our Dom34-delta strain and looking for grey and yellow reads – the long reads and the short reads – we see a clear signature that ribosomes are stuck in this gene on endonucleolytically-cleaved messenger RNAs that must be the result of the ribosome reading into poly-lysine. What I can tell you is that if we look broadly at yeast, in these strains – the Dom34-delta strain – and use the appearance of such reads in the middle of open reading frames as a signature of this process of Non-Stop-Decay happening, what I can tell you is it is incredibly broad. Many, many, many genes that we look at, as many as 50% of the genes that we look at, have piles of ribosomes in the middle of them if Dom34 is knocked out of the cell, suggesting that this is in fact a very regular event that happens in yeast cells and I suspect that it will happen in higher eukaryotes. So, I’ll go back to this slide, where I talked about the distribution of fragment sizes, as a way to end and tell you how important we think Dom34 might be in the cell. So, I showed you in the beginning that if we just, in an unbiased way, cut a big slice out from the gel and ask how many ribosomes are on full-length ribosome reads versus short reads, what you can see here was there was about 5% of the reads, maybe, that were in this area down here, if we sum them up over several lengths. So, you might say, those are typical Dom34 targets. They’re short reads and they might be typically targeted by the Dom34 system. And what I can tell you is if we knock out Dom34… this is actually a wild type strain that we’re looking at here… if we knock out Dom34, this peak size doubles. So, at least 5% of reads in that case, we can kind of confidently say, might be the result of Dom34 action in the cell, suggesting that in fact there’s plenty of short-mer reads in the cell and you could imagine there’s short-mers that result from the intermediate processes of exosomal decay of messenger RNAs, and so we would argue that this helps us to understand why this gene is essential in higher eukaryotes, and I suspect we’re going to learn more about that quite soon, and non-essential in yeast, but that it becomes essential in the case where ribosomes are limiting. So, I think it helps us to understand the in vivo role of Dom34. It plays an important role, that’s why it’s embryonic lethal in higher eukaryotes, and we’ve uncovered that role by working in a genetically-manipulable system where we can actually delete the gene and study the process in real-time. So with that I’m going to stop and conclude by telling you that I hope you’ve come to understand something about mRNA surveillance in general, and I think the take-home message is really that at the center of any mRNA surveillance process is the ribosome. The ribosome reads the messenger RNAs and it’s properties of the ribosome and the ribosomal machinery that are responsible for recognizing and recruiting factors to understand when messenger RNAs are defective in some way and need to be targeted for decay by normal pathways, that the incomplete protein product needs to be targeted for degradation, and that the ribosomes themselves need to be rescued by a pathway independent of normal termination process, in order to be put back into the pool for ribosome homeostasis. And with that I’ll mention the main players in my lab who did the work. The earliest biochemical work in the lab was done by Chris Shoemaker, and all the ribosome profiling work was really done by Nick Guydosh, a recent postdoc in the lab. And I’d like to also thank my funding sources from the NIH and the Howard Hughes Medical Institute. Thank you.