HomeArticlesDavid Bartel (Whitehead Institute/MIT/HHMI) Part 2: MicroRNAs: Regulation by Mammalian microRNAs
David Bartel (Whitehead Institute/MIT/HHMI) Part 2: MicroRNAs: Regulation by Mammalian microRNAs
September 10, 2019
Hello, and welcome to the second part of this series on microRNAs. In the first part, I gave you a general introduction about microRNAs, and in the second two parts I want to focus more on the types of experiments that are being done and in this part, I’d like to describe some of the experiments that we’ve been doing to measure the regulatory effects of microRNAs. So, up until about 2005, the prevailing idea was that microRNAs predominantly decrease the amount of protein that’s coming from the targets. somehow doing some sort of translational repression, without really affecting the amount of the mRNA. And then in 2005, Lee Lim published this experiment where he showed that when he added a microRNA, for example, mir-124, this is a microRNA that’s preferentially expressed in the brain, very highly expressed in the brain, and when added this microRNA to these HeLa cells, this human cell line, that many of the messages did not change, those in blue, but there were some that went up, and many more that went down. And those mRNAs that went down were preferentially mRNAs that had sites to mir-124 in their 3′ UTR, these seed match sites that I described in the introduction. And what’s more, these messages that went down were preferentially mRNAs that were actually lower in the brain than in other tissues in the animal. And so what this meant is by adding this microRNA that’s very highly expressed in the brain, to these cells, that it would make their transcription profile shift a little bit towards that of the brain. And he got a similar result when he added mir-1, which is highly expressed in the muscle, and that would shift the transcription profile a little bit more towards that of the muscle. So, this is very interesting and one of the major conclusions here was that microRNAs, in addition to doing this translational repression were actually having a quite detectable effect on the levels of the mRNA. And subsequent experiments from the Schier lab, the Belasco lab, the Izaurralde lab, showed that this change in the mRNA was mediated by shortening of the polyA tail and then decapping and degradation of the targeted mRNAs. So, we started using this data set and others that Lee had generated to begin to look at, what are the features of these mRNAs that could be, determine which mRNAs are targeted and which are not. You know, this is a lot of really useful data for looking at microRNA targeting. But of course what this data is missing is the additional repression that’s happening at the protein level, the additional translation repression, and at the time, it was actually thought that most of the repression was happening at the protein level, and that there would be some targets where you really would only be able to see the effects of the microRNAs by looking at the changes in the protein. So, we thought it’d be really important to also look at the changes in the protein. And so to do this, we started collaborating with Steven Gygi, specifically Judit Villen, in Steven’s lab, and, to look at what’s happening to the proteins. So, the way this experiment is set up is that we started with, we did some experiments like what Lee did where we’d add a microRNA to cells that normally don’t have that microRNA, but we also looked at what was happening to endogenous targets by doing these loss of function experiments where we’d start with either wild-type mice, which have the microRNA, and mutant mice, which lack the microRNA, and we’d isolate cells from those mice, and we were dealing here with this microRNA mir-223. This is a microRNA that’s very highly expressed in the mouse neutrophils. And so we could isolate the progenitors of these neutrophils, have them differentiate into neutrophils in these two different medias that would allow us to later be able to distinguish between the proteins that came from the WT mouse and the mutant mouse. And in that way, we could see the differences in protein that were caused by the presence of the microRNA. And so the, once these neutrophils had differentiated, we would mix them together and Judit did the proteomics experiment, and then we also looked simultaneously at what was happening to the messenger RNA, with and without the messenger RNA. Alright, so in this way we could look at the, what was happening to both the protein, and the messenger RNA of the endogenous targets. And so for context, we looked at what was happening at messages that did not have a site to the microRNA, things that would not be predicted to be targeted by that microRNA. So, that’s just the data that I’m showing you here, so on the vertical axis, we have the changes in the protein, and then on the horizontal axis, changes in the mRNA. And you can see that, as you’d expect, for these proteins and their corresponding mRNAs, that don’t have sites to the microRNA, things are clustered around zero. This is on a log scale, so 1 is a two-fold change, 2 is a four-fold change, et cetera. What about the proteins and their corresponding messages that had sites to the microRNA? Well, you can see here that there’s a shift in the distribution. There’s still plenty of them that really aren’t changing, that really don’t seem to be responding to the microRNA, but of those that do change, you see that they’re coming up here, to the upper right. And this difference in distribution can be explained by about 60 of these mRNAs that have sites to the microRNA being directly targeted by that marked RNA and being shifted both in the protein level, as well as their mRNA level. And so we could use these types of data, I should say that we’re looking at about a third, let’s say, of the proteins that are expressed in these cells, so thousands of proteins, but still not most of them, but the fact that sixty of them are changing implies that at least 200 of the, there are at least 200 direct targets of this microRNA in this cell type. Ok, so, that’s useful information and it was also useful to be able to look at these data to distinguish between those that are responding to the microRNA and those that are not. And so, we looked at various features that would explain why that same site in the same 8-nucleotide site, or 7-nucleotide site, in one mRNA would cause a response, whereas in another mRNA, would not. And what we found was that the differences that we saw between different mRNAs that had sites to the microRNA could partly be explained by the strength of the site. So, if the site was just a seven nucleotide site that had the seed match to the microRNA, and either the A here, across from nucleotide 1, or another match here, across from the 8th nucleotide, but not both, ok, so that’s two different seven nucleotide sites, those 7 nucleotide sites, about a sixth of them responded. More than 18 percent of them responded to the microRNA loss. Whereas, if you had the full 8-mer, that had both matched eight and the A here across from 1, a third of them, more than a third of them responded to the microRNA. And this is looking at all sites, not just the sites that are conserved, but any sites in the 3′ UTRs. When we focus only on those then, that were conserved, what we saw is that a higher fraction of them responded, as you would have expected. So, conserved 8-mer sites, over two-thirds of those responded to microRNA loss. It also depends on where the microRNA is, I mean where the site is, ok, so the sites that are in UTRs responded much more frequently than those that were in open reading frames, although there was some effect of some sites in the open reading frame. So, there is some effective targeting there, but it’s not as frequent as what you see in the 3′ UTR. And where the site falls in the 3′ UTR, whether it’s in a favorable context or not, also mattered, and we already had a model for predicting which sites would be in most favorable contexts, and that model worked very well. For instance, sites that are within high local A/U content, lots of A’s and U’s, which don’t pair to other RNA as well as G’s and C’s, those sites are presumably more accessible, and in fact, those would respond better in this experiment. So, these types of differences in sites could help explain why some responded and some didn’t. And then there are also differences between the mRNAs, some mRNAs had multiple sites to that microRNA and of course, that gave additional repression. What we see in general is that when you have multiple sites to the same microRNA, or to two different microRNAs that are coexpressed in the cell, both of those sites are functional, and the more sites you have, the more repression you have and the degree of repression that we see as more and more sites are added is what you would expect if those sites are acting independently from each other. You can see some cooperativity if the sites are within about 40 nucleotides of each other, at least, as long as they’re not overlapping, but beyond that they seem to be acting independently. But with additional sites, more and more repression. We also see that some microRNAs are different than others. In this experiment, we’re just looking at one microRNA, mir-223, but as we start to get data from many, many different small RNAs, what we see is that the RNAs with the stronger predicted seed pairing seem to be more effective and likewise, the microRNAs that have fewer mRNAs with sites in the cell are more effective for the mRNAs for which they do have sites. In other words, if the microRNA isn’t distributing himself among many, many mRNAs, if it’s just going to fewer mRNAs, it’s more effective at targeting those mRNAs. That’s kind of what you would expect. And then we can actually build a model now, we have one, and it’s present here if you just google the TargetScan and get to the TargetScan site, you can see that the predicted targets here are ranked based on the model that has incorporated all of these features and we continue to improve that model. It’s not perfect, but it’s still the most predictive model that has been developed. And so other types of information that we can get from this experiment. Well, one thing that was quite interesting is that when you look at the degree of upregulation, so, remember, what’s happening here is that we’ve deleted the microRNA and so the repressed targets, now their protein is going up, and their mRNA is going up in the absence of the microRNA. And so the degree of increase, even when we’re looking at the protein is not huge here, ok, we have a log scale, so these that are around log2, so, we’re talking about a 4-fold, 5-fold increase here, at most. But many of them less than a two-fold change. So, you find less than a 50 percent repression by the microRNA. Yet, even though there’s, these sites are imparting rather modest repression, they’re very frequently conserved in evolution. Now, remember that most mRNAs are conserved targets of microRNAs, and so I think a very interesting implication of this is that because these sites give rather modest changes in protein output, yet they’re conserved in evolution over quite long evolutionary times, that means that the precise levels of many different proteins appears to be important and the loss of these sites, which would give you a bit of an increase in that protein, seems to actually have a fitness effect for a surprisingly large number of proteins over surprisingly large evolutionary time. So, pretty interesting implication from these studies. But of course, that’s just looking at individual sites, the effects of the microRNAs are compounded when you have multiple sites to multiple microRNAs, and so you can actually get quite substantial repression with many sites to all the different microRNAs that are in the cell. But the evolution, of course, is acting just on single sites. So, the other thing to say is that, and this is the major take-home from this experiment, is that what we’re seeing here is the degree of protein change that you see in the vertical axis is actually not that different from the degree of mRNA change that you see in the horizontal axis. In other words, most of the protein change that we saw here could be explained by a change in the mRNA. So, this is very interesting and it implied that the translational repression component of this regulation wasn’t as great as had previously been thought. So, we were interested in looking more carefully at what’s happening here. And the basic question is what are the molecular consequences of microRNA targeting. We know that some of what’s happening is a decrease in the mRNA, so here I’m just showing you one of the targets of a microRNA for mRNA molecules, and in the presence of the microRNA, it goes to two mRNA molecules. That’s clearly happening, and then the question is, how much can that explain the decrease in the protein, or, on top of that decrease in mRNA, is there also a decrease in ribosomes per mRNA? Which is this second possibility here. So, how much of it is a decrease in the mRNA, and is there an additional component, and to what degree is there additional decrease in the number of ribosomes in the remaining mRNAs. So, the way that we decided to look at this, of course we got quite a bit of information looking at the protein, but in the meantime, there was another technique that had come on, that had been developed that we were very excited about using because it would allow us to look at differences in the ribosomes, and allow us to look in a more quantitative way at more mRNAs than what we could do at that time looking at proteins. Remember, in the proteins, we’re only able to look at about a third of the proteins that are expressed in the cell. And this method that we implemented was developed by Jonathan Weissman’s lab and it’s called ribosome footprint profiling and the way this works is that the ribosomes are arrested, cells are lysed, and then the mRNA between the ribosomes is digested with an RNase, and that leaves these protected fragments, what we called the RPFs, or the ribosome protected fragments. And those then can be sequenced with high-throughput sequencing and each of these sequencing reads then tells us where one of the ribosomes is on the message and we get millions of these reads, so we can look at the location of millions of ribosomes and look at how that changes with and without the microRNA. And so we did that with the miR-223 system in neutrophils and also adding microRNAs to HeLa cells, and that we published, and I’m also just showing you another experiment that’s not yet published but a similar type of data where we’re looking at the ribosome protected fragments and how they change when B-cells, activated B-cells, which normally express a lot of miR-155, when they no longer have that miR-155, when that 155 has been deleted from the genome. And you can see that again, there are a lot of these that don’t really change, but those that do change, because the 155 was repressing the amount of protein, and the amount of ribosomes that were on these messages, when you lose the 155, the number of ribosomes increases for many of these targets. And when you, at the same time, you see an increase in the mRNA, so things are shifting up here to the upper right rather than to the other quadrants. And basically, pretty much a corresponding increase in the amount of mRNA change compared to protein change, for these endogenous targets of miR-155. So, overall, and we’ve done this experiment, and these types of experiments, now in many different cell types, what we’re seeing is that overall what’s happening is that the ribosome changes, and the protein changes, are mostly reflected in mRNA changes, so the mRNA changes can mostly explain what’s happening to the ribosome, and what’s happening to the protein. In other words, in this example, the reason, the major reason that there are fewer ribosomes here in the presence of the microRNA is because there are fewer mRNA molecules. They’ve gone from 4 to 2, so from 16 ribosomes to about 7, because there are fewer mRNAs. And we really don’t see any evidence for some targets repressed only at the protein level. In these types of experiments, you can see some outliers, but you see them also among the messages that are not targeted by the microRNA. So, we really don’t have any evidence for specific messages repressed only at the protein level. Nonetheless, with these very quantitative methods, we can detect some change in ribosome occupancy, so here we’ve gone at the top here we’ve gone from an average of let’s say of about 4 ribosomes per mRNA and now here we’re about 3.5. So, there’s some influence by, on top of, the change in the mRNA, but it’s rather modest. And we’ve looked in lots of different cell types and overall what we see is that between 70 to 90 percent of the effects can be explained by the changes in the mRNA. Doesn’t matter if the cells are dividing or arrested, or if they’re stressed or not stressed. And all these experiments that I’ve described so far have been looking at steady state, where the mRNAs and the microRNAs have been sort of allowed to reach equilibrium, and that’s of course the way that you want to look at it to understand the relative contributions of these processes. But still, it’s also very interesting to look at the dynamics of this repression, and so we’ve also been doing those types of experiments, but we still haven’t been able to find a setting in mammalian cells in which the dynamics of microRNA repression impart some, really a substantial effect, that can be observed only at the protein or ribosome level. And so this is actually very good news for those of you who are looking at the effects of microRNAs in mammalian cells, because what it means is that you can get a pretty quantitative read-out of the effect of the microRNA by adding it or subtracting it, and looking at what happens to the mRNA without having to look at the protein or the ribosomes, which is more difficult than looking at mRNA changes. Again, in general, what we’re seeing ia that the changes in the ribosome, changes in the proteins, can mostly be explained by changes in mRNA. But there is one really interesting exception that has now been reported in another system, in the developing zebrafish embryo. And this was shown by Antonio Giraldez’s lab. What he found when they were studying miR-430, remember, miR-430 is this microRNA that comes on very early in zebrafish development, and it’s really important for the development of the brain, and for other things at this early stage, as I described in the first talk, the introduction to microRNAs. And so, what they found was that when this microRNA comes up here, in red, between the two hours post-fertilization and four hours post-fertilization, that the ribosome protected fragments goes down quite substantially here, and at four hours post-fertilization it’s down, and it continues down here at six hours post-fertilization. But there’s a delay in the mRNA going down, so you can see at four hours, the mRNA really hasn’t changed, and, but then later, at six hours, the mRNA changes. And what they see at six hours essentially is what we see in these other mammalian systems that we’ve looked at. Where most of the changes in ribosomes are explained by changes in mRNA. But at four hours post-fertilization, it’s very different. If you were just looking at the mRNA changes, four hours post-fertilization we would miss pretty much everything that the microRNA is doing. At this stage, you have to be looking at the ribosomes, and so the interpretation here, or one idea, is that maybe what’s happening here is that the microRNA is initially repressing the translation and then there’s this lag, and due to the dynamics of microRNA regulation, there’s this lag and only later comes the mRNA destabilization. So, we’ve been very interested in this, and also been looking in the early zebrafish embryo, at microRNA regulation, and one of the experiments that we’ve done is to inject a microRNA rather than looking at this, embryos with and without this miR-430, we’ve injected a microRNA, miR-155, that normally isn’t in these embryos, so we inject it at the one cell stage, so it’s there constant, throughout these stages of development, and what we see is that really from two hours post-fertilization, we see a very strong change in the ribosome protected fragments, as well as at four hours and at six hours, and again, here, it’s between four hours post-fertilization and six hours post-fertilization, that you see a change in the mRNA. So, in some respects, very similar to what was seen for miR-430. But remember, in this case, the microRNA was there from the very beginning. So, this is the first clue that it’s not so much a dynamics of microRNA repression that is causing this difference between four and six, it’s something intrinsic to these embryos that’s different. And so to consider that, I just want to remind you about the mechanism of microRNA action. So, remember that the microRNA recruits Argonaut to the 3′ UTR, that recruits this GW182 protein which in turn recruits the deadenylase complex, and that causes the poly-A tail to get shorter. And so, the way that we explain what’s happening here, and we’ve developed ways now to measure the lengths of millions of poly-A tails for millions of mRNAs, and I’ll just summarize what we’re seeing in the context of this experiment. What we’re seeing is that the reason that you have this change in ribosome protected fragments at this early stages is because there are very different consequences of shortening the poly-A tail. In all three of these stages, the microRNAs are causing the shortening of the poly-A tail, but at six hours that shortening of the poly-A tail causes the mRNA to get degraded, but these embryos earlier on are very different. They’re different in two respects. One is that those with shorter tails are not destabilized, and the second difference is that at this two hour and four hour embryos, what we see is that there’s a very strong coupling between the length of the poly-A tail, and the efficiency of translation, and that disappears at six hours. And so, here, the mRNAs that are targeted by the microRNA have shorter tails and that causes less translation, and here they’re targeted by the microRNA, have shorter tails, and that causes the mRNA to be destabilized. So, now we have a much better understanding of what’s happening in this very interesting context where you have to look at the ribosomes to understand what the microRNAs are doing. So, again, what we’re seeing here is that the changes in the ribosome, changes in the protein, are mostly reflected in changes in the mRNA. And in all contexts outside of the early embryo that we’ve looked at so far, that’s changes in the amount of the mRNA. Whereas in the early embryo, it’s changes in the length of the poly-A tail of that mRNA. So, I’d like to thank all these really excellent students and post-docs, who are listed here in purple for the really great work that they did on these projects that I described, as well as our really wonderful collaborators, who are listed here in blue. And I want to thank you for your attention, and have a great day.