Scientist Stories: Craig Pikkard, Gene Regulation from Plants to Humans

Rob sends his regrets and happy— and suggested that I give this talk and sort of in Rob's honor, I decided to add in the first half of the talk being something that would be on a topic that would be very similar to things that the Martinson lab cares about, which is small RNA biogenesis. And then, the second half will be the short talk I would have given tomorrow anyway. So, we'll sort of be in the transition between today's session and tomorrow's session anyhow with this plan. Plan. So 2 hours to get it put together, and so hopefully I had no chance to rehearse it, so we'll see how it goes. All right, so both the 2 topics I want to talk about fall under the common heading of mechanisms of selective gene silencing and epigenetic control. These are things we care about in my lab. So the first story I want to talk about is new work that we've been doing on trying to understand the RNAs of RNA-directed DNA methylation. So, to remind you of this pathway, or show it to you for the first time if you haven't seen it, this is a pathway that's important for the silencing of transposons. We just heard about transposons in the last talk as well. But many of the retrovirus-like transposons or retrotransposons are in fact controlled by this pathway, as are some genes. And key to the pathway are two RNA-dependent RNA polymerases that are unique to plants that my lab discovered almost 20 years ago, Pol IV and Pol V. These are evolutionary derivatives of RNA polymerase II that have 12 subunits just like Pol II but are dedicated to generating non-coding RNAs rather than messenger RNAs. And these non-coding RNAs are key to the silencing pathway. So Pol IV sought to kick off the pathway working in partnership with an RNA-dependent RNA polymerase called RDR2. And in Arabidopsis there are 6 of these RNA-dependent RNA polymerases, but only this one is involved. And the reason why it's involved is it actually physically interacts with Pol IV. Now together they generate double-stranded RNAs. One of the 4 different dicers in Arabidopsis is responsible for processing these into 24-nucleotide siRNAs. These are then loaded into an Argonaute protein, primarily Argonaute 4. And again, Arabidopsis has 10 Argonauts, but this is the major player. And then the small RNA actually guides the process of DNA methylation to homologous sequences, but does— seems to do so through the interaction with RNA polymerase V transcripts. This was shown by Andrzej Wierzbicki in my lab when he was a postdoc. It also interacts with the C-terminal domain of Pol V, as shown by Terry Lagrange working with Steve Jacobson in collaboration as well. Now, through other protein-protein interactions, ultimately the de novo DNA methyltransferase is recruited in ways we don't fully understand, and then we methylate the adjacent DNA. Now, if that happens to be DNA that consists of a promoter, that can be sufficient to block that promoter from firing. This is how we can achieve transcriptional silencing by this pathway. Now, this is a stripped-down, fairly simplified version of the pathway, mostly focused on Pol IV and Pol V. And mostly what I want to tell you about is going to be focused on the first half of this pathway that we've had recent success understanding in biochemical detail. That's what I want to tell you about here this evening. Okay, so support for the model. We put forward this model some years back, and genomic data has supported it reasonably well. This was one of our first attempts at doing genomic analyses in collaboration with Joe Ecker's lab. And what's shown here is ChIP-seq signals for where Pol V is sitting. And these are two peaks of Pol V occupancy. They are now missing in the Pol V mutant, for instance. They're gone. DNA methylation is color-coded for different patterns of methylation, but we can ignore that for now. You'll just notice that there's quite a lot of methylation. Sitting under these regions. And if you deep sequence the small RNAs, which are dependent on Pol IV, so they're gone in the mutant, you'll notice that the methylation occurs where the small RNAs overlap with the Pol V transcripts. And that's basically what the model's been. So the model has held up reasonably well over the years. Now, surprisingly, we haven't really known much about the RNAs made by Pol IV and Pol V in the now 12 years since they were shown to be involved in this pathway. So we kind of know what they do without knowing precisely how. There's been some recent progress now in the last couple of years understanding the nature of the Pol IV and RDR2 transcripts. And I'll show you some of those data here. So both our lab and Steve Jacobson's lab published at about the same time sort of our first analysis of what these short RNAs that are made by these guys look like. And both Steve's lab and my lab had the exact same rationale for looking from southern blots— or northern blots like the one I'm showing you here. So we're looking at a locus that produces very abundant 24-nucleotide siRNAs here. And then, if you look at what happens as you knock out the different DICERs— so DICER1 you can't completely knock out, it's lethal, it's required for microRNAs. But Dicers 2, 3, and 4 you can knock out. And you'll notice as soon as you knock out Dicer 3, things happen. These bands are gone and there begins to be this smear. Now you can still dice or cut with other ribonucleases, so there are still products made. But in any event, as you start making double mutants with Dicer 3 and other dicers, this smear gets even stronger and stronger so that in the triple mutant it's quite strong up here. Now you'll notice these are not terribly long. They only average about 32 nucleotides in size. And we are very interested in what are these things, and might they be the precursors? If you also— if you're triple mutant for the Dicers and you're also mutant for either RDR2 or Pol IV, they are wiped out. So these things are completely dependent on Pol IV and on RDR2. And you'll notice that because you really can't tell these two apart in terms of their molecular phenotype, we haven't known for a fact which one really goes first. And that's something I'm going to be showing you here as well. But, in any event, we were curious, what are these things, these Pol IV and RDR2-dependent RNAs? Former postdoc Todd Blevins, who now has his own lab in Strasbourg in France, led the charge to clone those things. And here is from just one example from deep sequencing back to 5S locus. These things in blue are precursors, are those RNAs that accumulate in a Pol IV/RDR2 double mutant. Again, they average about 30-something nucleotides long. They don't start at any specific place or end at any specific place. They like to initiate with a purine, and they just occur sort of in all possible registers. Now, if you look at where the small RNAs in wild type come from, you have a very nice overlap, and there is a fairly good tendency of the very ends, either the 5' end or, well, the left side I should say, or right side of these RNAs, matching these precursors. So these precursors are only big enough to be diced once, and they appear to be— you can cut them from either end to generate the siRNAs. So we finally knew what we were looking for after all of these years, after 10 years of searching for these transcripts. The problem was we were doing deep sequencing, and we were running all the stuff off the bottom of the gel because they're only 30-something nucleotides long. But that's what they make are these relatively short transcripts. And I'm going to tell you why they're short as well, too. Okay, so this has still been a question in all these years. It's a model that Pol IV really goes before RDR2. I'm going to show you that the model is right. It actually is true, but we haven't had biochemical proof until here very recently. All right, so here I'm going to show you some unpublished biochemical data done by a very talented graduate student, Jasleen Singh, in the lab. So back in 2012, Jeremy Haig, a former postdoc who has a position at Monsanto these days, he devised the first in vitro assay for Pol IV and RDR2 and Pol V. And to do it requires that we use an oligonucleotide template to which we hybridize an RNA that could base pair for about 9 base pairs or so. So this mimics the base pairing that occurs in an elongating RNA polymerase. So there's always a little bit of RNA-DNA hybrid. Now, we can end-label the RNA, and we showed— Jeremy showed years back that if you add the Pol IV-RDR2 complex, which has been affinity purified, you could in fact elongate that RNA primer in a templated fashion to the end. But what we generate is Pol IV and RDR transcripts that are that fall off of the end of this thing and are only single-stranded. So here's first evidence that it's Pol IV that's making that first strand. So we can do this reaction where we have the wild-type complex, which Pol IV and RDR2 are associated, or we can isolate Pol IV from an RDR2-null mutant. So P4 means that it's isolated in a mutant that doesn't have RDR2 at all, a null. This case. So here's what the transcripts look like. Here's our hot primer. We make these stuttering transcripts, including full-length stuff that runs off the end. Now, for years we were generating these transcripts, and we were never generating double-stranded RNA in these reactions. We did not understand why. We have both Pol IV and RDR2 in the reactions. So the new breakthrough, which Jasleen found, came from revisiting something in Jeremy Haig's paper from 2012 in which we noticed that if the template were double-stranded, we hardly made anything go through a second strand. So I had had Jasleen test out different templates in which we included non-template DNA of different length to try to give the Pol IV a head start before it encountered it. And we were wondering whether it would just stop at the edge or not. In the course of doing those experiments, we stumbled on the mechanism by which Pol IV and RDR2 are coupled. So I want to tell you that story here. Okay, so if you give it a template like that, what happens is that the Pol IV complex, when it encounters the edge of the non-template strand, it doesn't terminate right there. It goes about 10 to 12 nucleotides more in a sequence-independent manner, and then it terminates. And that termination event somehow is key to coupling RDR2 to make the second strand. So here's what that looks like. So the only difference between these first lanes again is that now we're adding the non-template strand the way I'm drawing it here. And we get this premature termination that's very strong. And again, it's Pol IV that's doing the termination because we get the same result in the RDR2 null mutant background as well. Now, with Jasleen's major contribution to the breakthrough here was he started toying with different nucleases to find out what's the nature of these transcripts. And this turned out to be a very important observation. So, in this case, he incubated these products with RNase I, which would take care of any single-stranded RNA, and RNase H, which would degrade any RNA-DNA hybrids. And he noticed something very interesting, which is that in only this one lane, as you'll notice, do we have a product. So, to get this product that's resistant to these nucleases requires that we have RDR2,, and it also requires that non-template strand. So single-strand should have been wiped out, hybrid should have been wiped out. We have something protected, and so we thought, "Aha! Could it be the double-stranded RNA at last that wouldn't be degraded?" So as we thought more about that, we're starting these reactions with an unlabeled RNA. So we have a single phosphate on the first strand of the primer that we're making. But if we make a new strand de novo by RDR2, we're going to have a triphosphate on the end of that. So we thought if we have made the second strand, it should have a triphosphate, and if so, it should be a substrate for capping enzyme. So mRNAs, as you know, the 5' end has a 7-methylguanosine cap. You can do that reaction in vitro with labeled GTP. So we ran these reactions again, this time with no end label on the RNA to start with. So we ran completely cold reactions. And the only label in the next reactions I'll show you come from doing this GTP reaction at the very end. And it works like a charm. And it's a little bit longer because we've capped it. We've added a GTP onto the end of it. So that all looked good. Like, we really are making the second strand. We did another experiment to convince ourselves, and that was to design a template where we could control the nucleotide content. So we designed a template that doesn't have any Ts in the, in the, the first strand, and we set it up again with this primer and the non-template strand that's going to induce this coupled reaction. So, again, in the first strand we're going to make the new RNA. It's going to include uracils across from these A positions in the template. But because we're going to label in this reaction with ATP, we can't get any labeling in that first strand, only if the second strand is made. So that's the trick, is to do ATP where we can only get the second strand labeled. And that works like a charm as well. And again, it's dependent on the non-template strand, Pol IV, and RDR2 to do it. Better yet, we've made recombinant Dicer III in the lab. We've made recombinant RDR2 that works. We've recently made recombinant Pol IV, in fact, that works in the last couple of weeks by reassembling all 12 subunits. And I'll show you this nice reaction that Jasleen did where he just added in the Dicer III to see what would happen. And we can actually do the biogenesis reaction in the test tube. So a bit of a complicated slide here in that we've got Pol III-RDR2 in every reaction. We've got the non-template strand, which induces that early termination. We've got RNase I and H, which cleans up all the single strand and the hybrid stuff. Two other lanes to point out that are important. RNase III, which is a double-strand-specific nuclease, cleaves these things. That's that clean lane here again, so more evidence that this is double-stranded. And then we also tried just titrating in increasing amounts of Dicer III in the reaction. And as you'll see, what we make are 24-nucleotide siRNAs made almost quantitatively in this reaction. So, a little bit of update to the model. The models, often the way they are drawn, have Pol IV and RDR2 making this long RNA that's sliced like a salami. That's not true. They make very short transcripts, as a matter of fact, only big enough to be diced once. And these things are attached, and this reaction we think is a coupled reaction where they probably, as the Pol IV transcripts are being generated, they're being captured by RDR2, and then as soon as they're released, they then can make the second strand to make these short double-stranded RNAs. So we've got this first part of the pathway working in vitro. We've also made recombinant Argonaut4. We have some evidence that, in fact, we can load it with these small RNAs, leaving this to be the next black box that we're trying to get after biochemically. So we're investing a lot in the lab in producing all these complexes, including Pol V and the Argonauts and these various helper proteins that aren't even on this slide. And we're looking for good postdocs who want to do some biochemistry. So if you're interested, we should talk. Okay, so that was story number 1 in honor of Rob, which I told him in the bar the other day and he was excited about. So anyway, and some of the students were there too, so now you've seen the story. So we hope to get that written up before too long. All right, so what I was going to tell you about tomorrow is story number 2, which is something completely different. Which is the study that we've been doing for a long, long time now, trying to understand the selective silencing of ribosomal RNA genes. So I want to remind you what these things are. So the ribosome, of course, are the machines that make proteins. And a ribosome is 2/3 RNA. So everything in gold here is an RNA molecule. So there are actually 4 ribosomal RNAs. 3 of which are generated from a single precursor transcript generated by RNA polymerase I, and the 4th is made by RNA polymerase III. It makes the 5S. So these are the ones I'm going to be talking about. It's the genes that encode the 3 largest catalytic RNAs. Now, ribosomal RNA genes are arranged in long tandem arrays, hundreds of copies. And those loci are known as nucleolus organizer regions, thanks to Barbara McClintock. There's a good Cold Spring Harbor connection to a lot of this work as well. Now, here's an electron micrograph of 4 ribosomal RNA genes in action. So you've probably seen these in the textbooks, these so-called Miller spreads or Christmas tree-like structures where the central axis is the DNA. Coming off the central axis of RNA as they get longer and longer as you go from the 5' end of the promoter towards the 3' end of the genes in this case. And when these are active, that leads to all of the processing machinery coming to generate ribosomes and doing a lot of other things in the cell as well. And that's the driving force for the formation of the nucleolus. So just to remind you that here's a Nemarski image where nucleoli, of course, are the most prominent structures in the nucleus that we've known about ever since we could see nuclei with a microscope. Now, the fact is that not all ribosomal RNA genes are on at any one time. And one of the first clues that that was true came from a classic epigenetic phenomenon known as nucleolar dominance. So nucleolar dominance was described based on the study of hybrids. So what happens in many hybrids is that very often only the ribosomal RNA genes from one parent are expressed. So only those chromosomes from one parent form a nucleolus. That's how it gets its name. So I'll just show you one example of this from our work on Arabidopsis suessica, which is an allotetraploid hybrid. It has basically a diploid genome from both thaliana and aranosa combined into a single nucleus. Now, for many traits, the hybrid is intermediate between the two parents, but turns out not intermediate when it comes to the ribosomal RNA genes. But during development, the thaliana-derived genes are selectively silenced, whereas the ones from Arabidopsis are on. And I'll just show you one example down here, which is an RT-PCR experiment in which we can make primers that are specific for the thaliana-type genes or the Arabidopsis-type genes, and then we can use those primers to ask what goes on in the hybrid, which is all the other lanes. And what's labeled here is the time after the radical first emerges from an imbibed seed. So this is very early germination. And at 2 days, for instance, you detect transcripts from both sets of parents. But already by 4 days, already the thaliana ones are starting to be expressed at lower levels. By the time you get out to about 2 weeks or so, all the way through mature plants, only trace levels of thaliana genes are expressed, whereas the aranosa ones are dominant., and they just keep cranking. So the phenomenon in part was discovered in hybrids because hybrids have the advantage of you can tell the two parental types of ribosomal genes apart. Because the real problem with ribosomal RNA genes is they're darn near identical within a species and very difficult to study. But we ended up having a way around that, which I'm going to show you in a bit. So it's not actually a hybrid-specific phenomenon. And in both hybrids and non-hybrids, Part of the basis for this is selective silencing. So back in the 1980s when I first learned about this as a postdoc, the idea was that they were— the dominant genes were just better at competing for some limiting transcription factor. So they were on and the other guys were off by default. But we overturned that 20 years ago when Jeff Chen was a postdoc in the lab with relatively simple experiments that showed that, no, it's not differential activation, it's differential silencing. So Jeff did these simple experiments like this where we took those sowessica seeds and we germinated them on Petri dishes that either just had, you know, control media like MS salts and the like. And as I showed you, Arabidopsis genes are on, thalianin genes are off. But if you put in an inhibitor of DNA methylation or histone deacetylation or both, this is sufficient to prevent that silencing, that developmentally regulated silencing. Doesn't occur so that everybody is expressed in this case. So that argued that now there's plenty of transcription factors around but that one set of genes is being denied access to those genes. Now, the hybrids are difficult to work with for a number of reasons, and very difficult to do genetics on a polyploid as well. So we were fortunate to stumble upon the fact that we don't have to study this in the hybrid. We can look at this in good old non-hybrid Arabidopsis thaliana diploids. But that required that we find variation. So again, the genes are almost identical to one another, but Keith Early, a former PhD student in the lab, he noticed that despite the fact that each of these genes are virtually identical, at the very tail end of the gene, still within the region that's transcribed and part of the RNA that gets processed away, there is some variation in there just based on short invels or little repeats that differ in number. Such that you can put primers that flank that region and get different products. So here's a DNA PCR reaction where we can see at least 4 different types, which we call variants 1 through 4, one of which is only present in trace amounts. Now, because it's still in the transcribed region, we can also ask, "Well, are they all made into RNA?" And that's next door. So here we're looking at RT-PCRs of the RNA, and you'll notice that These guys are expressed more or less at the abundance of the DNA, but this guy that's about 50% of the total is not. It's off. But in chromatin modification mutants, which we know are important for preventing the silencing, now all the genes that you see at the DNA level are also expressed into RNA. So again, these are being selectively silenced by the exact same pathway that happens in hybrids. So what's going on? This is what nuclear dominance is in hybrid. It's just a manifestation of something that goes on in all eukaryotes, including you and I. Right now, we only have about half of our ribosomal RNA genes active, and the other half are off. And this is true in every eukaryote that's been looked at so far. So it's dosage control. It's all about dosage control. Okay, well, so we have these different variants. At least one of them gets selectively silenced, and we had a question of, well, how do they do that? Are these things intermingled with one another, or might they be segregated? So, a talented graduate student, Chinu Chandrasekharan, now a postdoc at Johns Hopkins, she took on this project of asking, "Well, can we find natural variation for these subtypes that would allow us to map them?" And you can. So, that same primer pair that we used in Columbia, which is our favorite strain here, used in other strains like Shaw or Lur or Berg, gives you this pattern. So, the same primer pair works, but you'll notice that they vary in the content of these subtypes. Chinoo also teamed up with our bioinformatics collaborators and went looking for, through deep sequence data, genomic data, all that stuff that people normally throw away because it's ribosomal genes. Send it to us. So anyway, so we went sorting through all those raw reads that normally get pitched to go looking for SNPs, and we found some that were present at high enough copies, and then she went in laboriously by PCR, verified that a lot of them were true. So she found, in total, more than a dozen either SNPs or tiny indels that define additional subtypes in addition to these things we call VAR1, 2, 3. And it turns out those SNPs also differ among the ecotypes, and they also differ in expression. So it turned out that of the— we ended up with like 14 or 15 subtypes. Half of them are constitutively active; the other half are silenced during development. But are not silenced if you're in a chromatin modification mutant. So they're subject to epigenetic silencing. Anyway, long story short, by taking advantage of this variation, if you have variation of the trait, you can cross these two things, you can let them segregate, we can try to map how these different variants segregate relative to markers to chromosomes 2 and 4. And what Chinoo stumbled upon was that all of the silent subtypes went to one of the two NORs. There are two in thaliana, one on chromosome 2, one on chromosome 4. All the silent ones went here, all the active ones went there. So the importance of this is that, again, since from the mid-1980s, we've been thinking that it's every gene for itself, either competing for transcription factors, and we sort of ruled that out, and then the idea was that, okay, now they're somehow competing for silencing. There's something different in their sequence that allows some to be silenced and not others. But this now suggested, wait, maybe that's not true at all. Maybe it has to do entirely with chromosomal position and not ribosomal gene sequence. So our best evidence for that came in a little study, and I'm just going to show you the bottom line of this paper published by Grisha Mohanath, who's now back in India, has his own lab. And Grisha identified in a chromatin modification mutant line in which genome instability was known to occur, We noticed that something was going on with the content of these different subtypes that we knew about. And what we were able to deduce with a lot of genetics is that most likely there was a chromosome break that occurred, a double-strand break happened in the active NOR, NOR4 in this case. And in the course of repairing that break, normally you would use the other chromosome 4 to fix it. But since it's sitting in ribosomal RNA genes, which are virtually identical in sequence except for SNPs out of, you know, repeated a single SNP out of 10,000 base pairs. Then it reached over and used chromosome 2 to fix it. And it ended up repairing it all the way to the end, including the telomere. So we lost this telomere and duplicated this guy on this end. So several million base pairs. These things—I didn't tell you the size yet, but each of these NORs is about 4 million base pairs. So they have about 375 genes that are each 10 kb in length. So this is a big chunk. And we can deduce that several million base pairs—we don't know exactly how much was transferred over here. But the bottom line is that what this ended up doing was it took the silent genes of NOR2 and moved them to the location of NOR4. And by golly, they're on in that location, really showing that it's not your ribosomal gene sequence. It has something to do with your chromosomal position. So from this we were deducing, ah, there's something magic about chromosome 4 to be on, something magic about chromosome 2 to be off. Well, new data now that we haven't published yet suggests that That's not quite true either. It turns out that the NORs are kind of autonomous. And this is where some of the chromosome engineering we've been getting involved in comes in. So, for what I'll show you next, just in a few slides here, is making use of the Wisconsin DSLox T-DNA collection. So these are a very useful set of lines in which T-DNAs were placed around the genome some of which happen to be close to our NORs, at least close enough that we can use them. Now, they have within them what we call here a lock site launch pad. So the tDNA starts having two lock sites in it, but one of them is inside a DS element, a transposable element that can be mobilized upon heat shock when we express the transposase. Again, it's McClintock's thing, AC transposase to move these DS elements. So you can jump them. So from the launch pad, you can jump the Ds elements and they can go to other chromosomes as well. Now, but one of the loci stays behind. So now we've got loci that are different. So if they're on the same chromosome, if two loci pointed in the same direction, we can bring in Cre recombinase and delete everything in between them. Sometimes you can't get away with that because it's lethal if there's things in between that matter. But if the loci lands in the opposite orientation, you bring in Cre and you can invert it. So you don't necessarily die as a plant, but we might actually be able to remove some controlling element from near and move it to another distance. So we're exploring these things. But what I'm going to show you now, the most complete story that we have, comes from cases where we jumped these loci to other chromosomes. And now that allowed us to actually do reciprocal translocations between the tips of the chromosomes where we could move the NORs to chromosomes where they hadn't been before. And I'm just going to summarize, rather than show you the data, I'm going to show you what's happened. And this has been the work of a very fearless postdoc, Dingxiao Feng, in the lab. So here's the starting condition. NOR2 is off, so red, stop; green, go, being on. Using the reciprocal translocations where we have— the loci are not immediately at the NOR, but they're close. So on NOR2, the locus site that we're using for recombination is about 110 kb from the edge of the NOR. For NOR4, it's about 300. But again, we've got 4 million base pairs of NOR and then another 100 or another 300 kb in addition to them. So that's what we're moving in these experiments. Anyway, she can translocate NOR2 to chromosome 1 where it doesn't normally happen, and by golly, it's still off. NOR4 is dominant, this one is off. She's moved NOR4 to chromosome 1 or to chromosome 3, and it's still on, and this guy's still off. So the dominance is still okay even when we move the NORs to new locations. And very recently, she's gotten the one that we wanted all along, which is the actual exchange. Let's just move the two NORs to the other chromosome, and do we swap the dominance? The answer is no. Actually, it turns out that they're still autonomous. This guy in red is still off, and this guy is still on. Is still on. So we— there's something about the NOR and what's adjacent to them is sufficient, we now know. So whatever the information is, is within that 100 or 300 kb and then the 4 million base pairs of the NOR. Now, that's an interesting result that we just got because remember this one I just told you about where in a case where we had a break somewhere in NOR4, we're not exactly sure where, that was replaced by these sequences all the way to the end, in this case we did change what happened with the expression of the NOR. So I just showed you that these two NORs, you can move them anywhere and they're okay. But in this case where you made a hybrid NOR, the one that was normally off is now on, suggesting that there's something down here that matters. So we're kind of trying to find ways to zero in on these junctions. Again, we're having to deal with millions of base pairs at a time. To try to figure this out. So, one of the ways we're doing that, actually with the help of Dan Voigtis, actually his lab, Ding Xia, has been working to generate talons. It's been slow going, but we're close to being able to get in there and actually take out a region next door to NOR2 that we think might be important. There happens to be a 60-kb transposon junkyard sitting next to NOR2, which might be critical. So we want to get rid of that, so we're working at that. And, in addition, we're also trying to think about what it is about chromosome 4 that maybe might be important, too, and what are the— and what really are the signals, in addition. So we know that there's something about this— we think these junctions here that matters, and again, when we move them, our starting point is about 100 kb out here or about 300 kb on NOR4. Something in that interval is sufficient. So hopefully we ought to be able to go in and try to zero in on the region. So that's what we're working on. And with that, I'm going to end early. And there's the group, happy group in the lab. We've had a few changes since, but anyway, and there's room for more on both projects, in fact. So anyway, all right, so thank you. Thank you very much, Craig. So questions? Yes, please.

Craig, if you delete NOR2— first off, have you deleted NOR2, and what are the fitness effects for deleting it?

Yeah, we have not deleted NOR2 yet, so we're not sure.

There's no ecotype that—

There is one, yeah. Bur-O ecotype actually has a dinky little NOR2. Yeah, we're not exactly sure how many genes, but small enough that we can see it on a CHEF gel. So I mean, it has very few. When you map to it, it looks like there's very few. And Fernando Rabanal in Magnus Norberg's lab actually did FISH to that. It has a tiny little FISH signal as well. Now, we don't know the size of NOR4, however, in that line to know whether it's just a bunch on the other and not on this one. But it's not always NOR2 that's silenced. I would have had time, but I cut out the slides that show that we've done some nice genetics of that, too, where there are ecotypes like Shaw that we worked with where NOR2 is clearly on. It's not off. And then we've made the Columbia-Shaw recombinant— look at the recombinant inbreds there to ask, well, what does it take now in order to silence? Because we start with Columbia where NOR2 is off. We have Shaw where NOR2 is on. So we differ by this trait, cross them together, do the genetics. It maps right back to the NORs. There's no evidence for any other loci. It's the NORs that matter. And it turns out that for the columbia NOR2 to be silenced requires that the recombinant hybrids have columbia NOR4. If it's shaw NOR4, you no longer silence NOR2. So the NORs are talking to each other as well. So whatever this signal is is coming from the NORs, we're quite sure.

[Speaker:MALE_SPEAKER] I mean, is there ever an environment that would get them both on, or I mean, if you— or a developmental stage that best guess that they need both sets of ribosomes?

Yeah, I think that they are, they're probably all needed once per life cycle. I think there's a good argument for that in a lot of eukaryotes. But that during most vegetative growth, you don't need 1,500. 400 is plenty, probably. So I think that the same way early on in development and in embryos and in cotyledons, you see both on, for instance. So it's only as the true leaves are coming out, and then when they transition to flowering, you can also get some being turned on again. So I think that the rapidly dividing cells use them. Plus they're not losing one whole NOR. They do use them once. Any other questions?

Yes. So you showed that hybrid manipulated plants have— so in earlier stages they produce both and then become— it looks like a mitotic silence. And what about in Columbia? There are different variants and then they have a look at an early stage, early developmental stage if they—

Yeah, in thaliana it's the same thing we see in the hybrid. So I showed the example of silencing in the hybrid where you can you know, tell the thaliana and the arenosa genes apart, and the thalia1 silence over a time course of about 10 days. In the non-hybrid thaliana, the same thing happens with the subtypes. The subtypes that get silenced follow that same time course, and they respond to the same chromatin modification mutants, prevents it from happening. So the hybrid— there's nothing magic about hybrids, basically, this phenomenon, even though that's how it was discovered, but it's something that goes on in all species. Not a system that was waiting, oh, what if I make a hybrid? Let's be able to silence it. It doesn't make sense anyway. So it's something that every organism does to control the dosage. So it's the same pathway.

Thank you. Yeah, V? Craig, so story number 1. Uh-huh. Okay, so RDR2 and Pol IV get together early. Uh-huh. In your experimental system, you're introducing the non-template DNA, so it's— penetrating through there and then stopping at 30-ish. In nature, how does it know to just stop at a 30-ish?

Yeah, thanks, actually. Thanks for, I didn't even have her plant that, but that's a good one. Yeah, this is why they're short. I think, we can't prove it yet, but we're working on it. I think what happens is you open a transcription bubble, which you always need to do to get a polymerase to initiate, so it needs to have single strands. So that's what transcription factors do normally. We don't know exactly how these guys do it. But once you make a bubble and you initiate and you reach the edge of the bubble and you go about 10 more and you terminate, that's why you're 30. That's my model. [Speaker:STUDENT] Is it because of the size of the bubble? [Speaker:PROFESSOR] Yeah, that's my guess. You open the bubble, you start in the bubble, you hit the edge of the double-stranded, you go about 10 more.

I think that's why they're short. [Speaker:STUDENT] Would you expect the bubble to be subjected to thermodynamic breathing?

[Speaker:PROFESSOR] Well, we only know what happens like in E. coli and other things like that where sigma factor and things like that work and Pol 1, 2, and 3. I mean, they make bubbles of about that size, and you usually don't see them spread. I mean, you can make an R-loop. That's one way where you could constrain it. But we don't know because we don't know yet. We're working on it in vitro to see what we have. There's a couple of Proteins that have been found genetically, people that are aficionados of this pathway will know that both Pol IV and Pol V have associated proteins that matter that are members of the ATP-dependent DNA translocase family. But in the reviews everybody says they're chromatin remodelers. But there's no evidence for that. What there's evidence for is that they're homologous to ATP-dependent DNA translocases. So these are things that bind DNA and they burn ATP and they move. If they move on double-stranded DNA, yes, they might push a nucleosome out of the way. And then you're a chromatin remodeler. But the other thing you can do is you can go on a single strand, and now you're a helicase. And so we think that they might be helicases and not chromatin remodelers, and so we're trying to test that in vitro. But my guess— [Speaker:WOMAN] You could image the bubble. [Speaker:KEITH] Yeah, we're trying to image it all. We've actually got— I didn't show you that either. We've got a cryo-EM structure of RDR2. We're working on it for Paul Flohr. We're going to get the complex. We're going to try to actually see that reaction, what it looks like. But we have to move to the chromatin. Yeah, chromatin is next. That's the project we need postdocs. Anyway, so the next step is to actually get this working on a chromatin template. But my guess is that that might be it. If we can figure out how the bubble gets opened, that might actually be key. Once we get started, we just go to the end and we drop off. And there's models out there that maybe, you know, it encounters a DNA methylation mark and makes a mistake and stops. We have no evidence for that. That. In fact, we've tested it and it's not true in vitro. Yeah? That was very nice. That T cluster next to NOR2 is really interesting.

Did you find a good correlation between the sessions, our readoptive sessions?

Yeah. Not so far. So, in fact, so this other ecotype we talked about, like SHA, where NOR2 is on, as best we can tell from the 1001 Genomes Project, It still has it. And I'm a little bit nervous about that, and I should probably talk to some of the more genome experts. I wonder how much de novo assembly there has been, because my other worry is that it's a repetitive region, and just because you go back to the reference genome, do we really know that it's there just because we find it? But anyway, so with that caveat in mind, it appears that the transposon junkyard is still there in cases where NOR2 is not silenced, as if it doesn't matter. But we want to verify that by optical mapping or some other secondary— Ed's nodding knowingly. Yeah, right, yeah.

Okay. What happens during stress?

Do they get turned on? For the ribosomal genes? Uh-huh. Sorry. Yeah, we haven't done a lot of physiology. Yeah, it's kind of funny. For all the years I've studied this, I've never worried about if they ever make it into a ribosome. We just want to know whether they're on or off and how does that happen. And so being transcription jocks, we've sort of done that. So I don't know. And there is— we should look because I don't know if you guys know about it. Some of the coolest work that I know about happening right now is these specialized ribosomes. A woman, Maria Barna, at Stanford does beautiful work. It's not in place. Plants, but there are ribosomes out there with— that differ in their ribosomal protein content that are transcribing specific mRNAs in specific locations of the cell. So look up Maria Barna, the 3- or 4-cell paper. Beautiful stuff. It's really cool. There's a whole new code out there with— of specialized ribosomes. So maybe even these little SNP variants we've been describing in our types that are on or off, they might change the way different ribosomal proteins bind. And in plants, there's a lot of variation in ribosomal proteins. So it's possible you have different variants of the ribosomal proteins and it might matter for what they do. So, Julia, there we go, one more thing we have to add to your laundry list of analyses to try to figure out whether there's subsets of mRNAs on specialized ribosomes.

So with that, no more questions.

So please join me in— Great. Thank you.