Scientist Stories: Roger Kornberg, Unexpected Role of Chromatin in Transcription

We're going to start the final session now. And so the final guest speaking is Roger Kornberg, who's a professor of structural biology at Stanford Medical School. And I think most of you know Roger comes from a distinguished scientific family, but has made many important contributions on his own. In fact, in back-to-back papers in 1974, he announced the discovery of the nucleosome, which you heard about earlier today. And that work was done with Francis Crick and Aaron Klug in the same laboratory where Sid began his work on tRNA. And in his own laboratory, Roger has done obviously very important work looking at the structure of the transcriptional apparatus and in particular solving that structure to the atomic resolution of RNA polymerase II transcribing DNA into messenger RNA, a fundamental process in biology. So it's terrific to welcome him. He won the Nobel Prize in 2006 for this work, and it's great to be— for him to be the final speaker at today's symposium.

So thank you, Dan, and let me add to what others have said what a privilege and pleasure it is to congratulate you, Sid, to participate in this celebration of your work. Unlike the others who were either a teacher or a fellow student or ultimately a student of yours, I represent the rest of the world. All of the many additional admirers of your work. I will, in the brief time that is available, speak primarily about two subjects. I'll begin with an introduction showing you just a view of the RNA polymerase II pre-initiation complex, and this is the work of Kenji Murakami Nir Kalisman, David Bushnell, and then a former postdoctoral fellow who's been at Scripps for many years, Francisco Asturias, and his postdoctoral fellow, Kwong Lee Tsai. I'll spend most of the time after that introduction speaking about the transcription of chromatin, and all of this work is due to Shigeaki Nagai. If time permits, I'll comment briefly on a point that Mark Patashny raised., and in this respect, work done by Jali Lorch and Barbara Meyer-Davis. And then I hope to leave time at the end for a second subject, and that is the folding of chromatin fibers beyond the level of the nucleosome, in this case the work of Kyle Egan and Tom Hartl. Now as I mentioned, and you heard from Dan, and you're all doubtless aware, all that I tell you relates to transcription by one of the three polymerases of eukaryotic cells, RNA polymerase II. And many of you are doubtless aware of what Mark diagrammed in one of his slides, namely the components of the polymerase II transcription machinery, which we and others identified by now some 30 years ago. The polymerase itself is capable alone of unwinding the DNA double helix and of RNA synthesis. But alone unable to recognize a start site and initiate the process of transcription. And it is for those essential functions that it requires the so-called general transcription factors that go by the letter names B, D, E, F, and H. Now it was a thought for some time that this constellation of proteins is not only sufficient for the initiation of transcription, for promoter-dependent transcription, but sufficient also for a response to regulatory influences. And indeed, as Mark would attest if he were here, there were many proofs of that point of direct interaction of transcriptional activator proteins with components of this machinery published during the 1980s. Nevertheless, in the early '90s, my colleagues Kelleher, Flanagan, Kim, and Bjorklund discovered an additional component required for an influence of a transcriptional activator protein, such as Mark mentioned, upon the transcription machinery that I have just described. This they isolated from yeast as a 21-protein, million-molecular-weight complex. We now know that it is conserved throughout eukaryotes. We refer to it as mediator for the obvious reason that it goes between regulatory molecules, both activators and repressors, and the transcription machinery. And we now know that some 60 proteins assemble in a large complex of about 3 million Daltons at every RNA polymerase II promoter prior to every round of transcription. Of transcription. To understand the mechanism of the machinery, we began long ago to investigate its structure. And we started, as you will have heard, with the polymerase itself, an assembly of a dozen subunits with a mass greater than half a million Daltons and far larger than could even be solved at the time when we started. But the process, when it finally culminated after about 15 years, led, as you've heard, to the structure of the polymerase in the act of transcription with both gene DNA and product RNA associated at near atomic resolution. Another 15 years later, we finally succeeded in assembling the entire complex of polymerase with all of the general factors, an assembly of, in the form that we studied, about 1.5 megadaltons., and just last year could report the structure determined by cryo-electron microscopy as shown in this slide. So what you see here is first of all a space-filling model, so an envelope of the entire structure of this 33-protein complex as determined by cryo-electron microscopy. And then superimposed on that envelope are the crystal structures of the polymerase in gray, of the general transcription factors B, which is red, the crystal structure of the TATA box binding subunit of factor D, which is green, factor E in pink, factor F in blue, and the portion of factor H for which crystal structures have been determined superimposed as well. And you see that there's really a very impressive, near perfect fit of the crystal structure of polymerase and of the other crystal structures to the map. In addition, one sees the DNA, and at this resolution, the major and minor grooves of the DNA even resolve. And what I'll show you in the next picture is a view down the axis of the DNA double helix. So we rotate this around, you view down the axis of the DNA double helix,. In this picture, also superimposed, is just one of the crystal structures, that of the ATP-dependent DNA helicase subunit of Factor H, which is responsible for unwinding the double helix to bring about the initiation of transcription. The notable feature of this picture is the lack of any interaction of polymerase with the DNA in a polymerase pre-initiation complex. Indeed, the DNA is associated only with the general factors. And if there were time, I would explain why that is and in more detail the significance. But I'd like to go on. If there were time, I would also show you a cryo-EM picture of not only polymerase and the general factors but also mediator. But rather what I'd like to do is explain in the time I have this afternoon why this picture, even that of the polymerase, general factors, and mediator, is incomplete. And the assembly of proteins responsible for the initiation of transcription of all genes transcribed by polymerase II is a good deal larger, more complex, and I think more interesting. Now the reason why the structure proves to be much more complicated than we realized is because of a mistaken assumption that we and others have made over the years. We always thought that the template for assembly of this complex and for transcription is a naked DNA molecule, and there was good reason to believe so. I mean, in the first place, I think it's obvious that the DNA can't be at the same time surrounded, as you see here, by all of the components of the pre-initiation complex, and at the same time wrapped around a set of 8 histone molecules in a nucleosome. Yalie Lorchenai had shown in 1987 that wrapping promoter DNA around histones in the nucleosome prevents the initiation of transcription in vitro. And Michael Grunstein and colleagues showed the very next year a similar inhibitory effect of histones upon the initiation of transcription in yeast cells in vivo, on which basis we, Mike Grunstein and others, came to believe what you heard Mark say earlier. The nucleosome serves as a general gene repressor. It prevents expression of all the many thousands of genes in eukaryotes except those whose expression is brought about by specific positive regulatory mechanisms. Again, it would seem apparent that any such mechanism must entail at some stage the removal of histones or the disruption of the nucleosome. In fact, the late Ira Herschkowitz, a great yeast geneticist who died unfortunately at a, at a at an unfortunate early age, Ira Hershowitz had intuited the existence of the proteins responsible for removing histones to allow transcription. From genetic studies in yeast, he concluded or he deduced the existence of a so-called switch-sniff complex. Marko also mentioned it in his lecture. He identified 5 genes which he believed would comprise that— would compose that complex and which would be responsible for relieving the inhibition of transcription by histones. When some years later my colleagues and I isolated the SWI/SNF complex from yeast, we could verify his hypothesis. We discovered it contained a number of additional proteins, one of which, so-called SWAP73, we knew had a homolog in the yeast genome, and then pursuit of that homolog revealed another even larger assembly of proteins which, like SWI/SNF, serves as a chromatin remodeling complex and plays a role in removing histones for transcription. The SWI/SNF complex is in fact relatively scarce, only about 100 molecules in a yeast cell, and the genes for SWI/SNF proteins can be deleted without effect upon vegetative growth. By contrast, the RISC chromatin remodeling complex is abundant on the order of 1,000 molecules per cell, and the genes that encode it are essential for cell growth. We discovered over the years that followed two important chromatin remodeling activities of the RISC complex. RISC in the presence of ATP will slide histone octamers and in that way expose expose a site for interaction with a protein molecule. RISC in the presence of ATP and a histone chaperone will actually dislodge the histones from DNA and expose a region of the DNA in that manner. We wanted to find out, to determine whether what I have told you is the whole story. So is it true, in fact, that RISC enables the initiation of transcription by the Pol II transcription machinery. We wanted to recapitulate the process starting from chromatin and culminating in transcription with purified proteins in a cell-free system. Now others in the years gone by have performed experiments along these lines. They produced chromatin by the addition of purified histones to template DNA, or with the use of extracts from cells that would serve as chromatin assembly systems. We thought it was important to pursue the problem with the use of chromatin assembled not in vitro in that manner, but as it is formed in cells in vivo. After all, we could not know it has not yet been determined what is the precise pattern of post-translational modifications of the histones of nucleosomes associated with a promoter. It is— it was not known what additional non-histone proteins might play a role. And so with the purpose of investigating as a template for transcription chromatin isolated from nature, we performed the procedure illustrated in the in this slide. So we began following a protocol invented by Mark Gartenberg here at Yale, inserting sites for recognition by a recombinase either side of a gene. And for these studies, we studied the FO5 gene of yeast, though as I'll tell you, the results are completely general. Then by expression of the recombinase in the yeast cells in vivo, one brings about the excision of the gene, in this case Faux5, in the form of a circle still associated with the natural set of chromosomal proteins. The circle could be isolated by affinity means. We purified it to some 200,000-fold the virtual homogeneity. When Shigeaki Nagai set about transcription with polymerase and the general factors, he found, as we expected, no detectable transcription from the gene isolated in the repressed form as chromatin. The same DNA extracted from the chromatin would, of course, give rise to measurable transcription. I should say that an experiment like this, which Nagai performed, was not straightforward because of the extraordinarily small quantity material. We're isolating chromatin from single copy genes, means we recover attomolar, that is 10 to the minus 18th molar quantities of chromatin, and detection of transcription was a challenge, but which he nevertheless overcame. Shigeaki then found that by the addition of 5 more proteins, the activator protein for the FO5 gene called FO4, TF2S, which was previously regarded as a transcription elongation factor, but we find is essential for initiation, the mediator that I've mentioned, another complex to which Mark alluded, the SAGA complex, and a remodeling protein complex which could be either SWI/SNF or RISC in these experiments. Now in the presence of that group of proteins, the chromatin template was efficiently transcribed. In fact, to our surprise, to a far greater extent than naked DNA extracted from the chromatin circle. We had thought if we were ultimately successful, we would achieve a level of transcription from chromatin comparable to that from a naked molecule. But to our astonishment, the level of transcription from chromatin was far greater. And in fact, as the concentration of polymerase and the general factors and other proteins was reduced,, the ratio of transcription from chromatin to that from the naked DNA molecule increased to the point where it was at least 20-fold greater at the lowest concentration that we were able to use. So this would indicate that chromatin is better able to recruit— so it shows a higher affinity for the proteins involved in transcription than a naked DNA molecule, which is a point I'll return to in a moment. I might mention in passing that Chromatin transcription requires acetyl-CoA. Of course naked DNA transcription does not. And also chromatin transcription requires a subunit of the SAGA complex called SGF29. So a form of SAGA deleted for that complex shows significantly diminished transcription from the chromatin template. So just to summarize what I've told you to this point. Transcription from chromatin is a good deal greater than that from a naked DNA molecule, and it increases relative to that from naked DNA as the concentration of the transcription proteins decreases, which points to an interaction of chromatin, and in particular, as I'll show you, the nucleosome, with the transcription machinery. And then I just mentioned that that interaction involves both histone acetylation and the histone 3-K4 trimethyl mark that is recognized by a subunit of the SAGA complex. Now we could identify which nucleosome of the chromatin template is involved in this interaction by virtue of studies done many years ago by Wolfram Hertz and colleagues. So Hertz had shown that the FO5 gene as you might imagine, and as Mark illustrated in his diagrams, is packaged in an array of nucleosomes in the repressed state. Hertz also showed that upon transcriptional activation, so upon induction of transcription of the FO5 gene in yeast in vivo, 2 nucleosomes are removed and 1 is caused to slide. It shifts in position exposing the TATA box of the promoter but still covering the transcription start site. Now we could insert the sites for recognition by the recombinase to make chromatin circles upstream and downstream of the promoter, and in that way isolate circles from the gene in the repressed state that would bear 3 nucleosomes, circles from the gene in the activated state bearing only a single nucleosome located over the transcription start site but exposing the TATA box. We found still there was enhanced transcription from the chromatin template compared with the naked DNA molecule. On this basis then, we believe contrary to what we had imagined that the presence of a nucleosome covering the transcription start site would be inhibitory to transcription, rather that the nucleosome covering the start site, commonly referred to as the +1 nucleosome, is actually a component of the transcription machinery. We imagine that the transcription machinery we previously thought comprised polymerase, the general factors, and mediator now entails assembly of a much larger complex. Beyond those 57 proteins, a total of 99, a complex of over 5 million daltons, again, which must be formed at every polymerase II promoter prior to the initiation of transcription. Now I said what we have observed for the FO5 gene of yeast we believe to be general. We think it's applicable to all yeast genes, in fact to all genes of eukaryotes, and I'll explain now briefly why. So this picture which I have just shown you of a single nucleosome in the activated state covering the transcription start site flanking an exposed region is observed throughout the yeast genome. So you see a picture here of a type of analysis that has been reported by many research groups. This is our own version of the same thing. It's obtained by mapping the locations of nucleosomes with the use of cleavage between them by micrococcal nuclease and then deep sequencing. What you see plotted in red is the likelihood of occurrence of a nucleosome. As a function of position along a 100-kb region of one of the yeast chromosomes. Also indicated below are the locations of transcribed regions with the start sites by the plus signs. We zoom in on one portion. What you see, what is true throughout the genome, namely an array of nucleosomes but with the apparent absence of one immediately upstream of a transcription start site, and then a very well-positioned nucleosome, the so-called +1 nucleosome, covering that site. The so-called free or nucleosome-free region includes the TATA box and regulatory elements, and those are available for interaction with activator or repressor proteins and for the beginning of assembly of the transcription machinery. In the interest of time, I won't comment on the question that is posed here of what is the basis of the NFR, but I'll skip directly to a further and last topic of my presentation. I'll simply summarize what I've told you up to this point in the following way. We think what emerges is what I might call a two-step authentication mechanism. For the turn-on of transcription. A first step in which a region that includes regulatory elements in the TATA box is exposed, but still leaving a nucleosome covering the start site, and then a second step that requires that nucleosome and culminates in the initiation of transcription. It's rather like the immune system that requires two steps to assure fidelity of the response. In this case, two steps and an intermediate in which a nucleosome still covering a start site prevents adventitious initiation of transcription. Now everything I've told you to this point relates to chromatin at the only level we presently understand, namely an array of nucleosomes which in, as I've said, in the repressed state and in as Mark emphasized as well, more or less uniformly packages the genomic DNA. The next slide will, in a way, summarize our ignorance of the problem. So this is a picture many of you will have seen, an electron microscope thin section of a eukaryotic nucleus with heterochromatin, the condensed region around the periphery, so-called euchromatin in the interior. I showed this picture for 20, 25 years to medical students at Stanford when I taught histology to medical students. And then finally the Stanford administration in its wisdom eliminated basic science from the medical curriculum. And so I don't have many opportunities of showing it again. I do so now for a particular reason, which is to call attention to the question of what is the organization of chromosomal material, and in particular, what is the structure of a gene in interphase in a eukaryotic cell nucleus. Now I will tell you then only briefly about experiments which I think shed light on this question, and they come from our use of a method introduced by others that you heard mentioned, chromosome conformation capture, or in its most elaborate form, so-called Hi-C. And the method is very straightforward. It involves fixation with formaldehyde to cross-link proteins, cleavage of the DNA, and then finally ligation of the cut ends. If the proteins have brought together two regions located at distant sites in the genome, sequencing across the junction will reveal their proximity in the chromosome. And this has been done, as I say, by many others over the years, and this is just an example from work by Decker and colleagues of the results that are obtained. What you see here, each red dot that is plotted here corresponds to a point of cross-linking, so of approximation between a location in the DNA sequence of the chromosome plotted here, and another location in the same chromosome plotted on the other axis. And what is revealed by this type of analysis done in many laboratories are the— this is such an alternation of regions of intrachromosomal interaction and less such intrachromosomal contact. It struck me such an alternation of interaction and absence of such was reminiscent of polytene chromosomes of Drosophila. And so we asked the question whether the application of Hi-C to dipteran larval tissue containing polytene chromosomes would produce a similar result to what you saw in the previous slide. Kyle Egan found Indeed, the so-called topologically associating domains, or TADs, could be detected when the procedure was applied to salivary gland tissue of the appropriate larval stage of Drosophila. And then what you see in the next slide is in the number of cases where DNA sequence coordinates of bands are known. And in these cases, the DNA of the bands is lined up at the left side. The amount of DNA in a band varies in length. And what you see in red is the overlap with a TAD revealed by the Hi-C approach with the DNA sequence coordinates of the band. In black are regions of non-overlap. Well, it's obvious that there's almost a perfect correspondence. Of TADs with bands. The same thing could be applied, so bands equal TADs, and I might add what this immediately reveals is something significant about the interpretation of a TAD. People had wondered, are these transient, or do they represent stable interactions? Do they relate to regulation, or do they relate to structure? What we learn from this simple observation TADs actually correspond to regions of persistent chromosomal condensation along the length of the chromosome. We could apply the same procedure, and it's plotted in a similar manner, to the so-called Hi-C, or chromosome conformation capture analysis of diploid cells. And what emerges is that not only are bands equivalent to polyteentads,, but they're nearly the same as diploid TADs, from which it emerges that the picture of a chromosome that one observes in the light microscope from examination of polytene tissue actually reveals the organization of the chromosomal material in the interphase nucleus of, well, in this case, cultured cells of Drosophila, but virtually the same result with embryonic tissue of Drosophila. Sopla-containing? Squash it when you're looking? Absolutely. You see exactly the same result. But no two squashes look the same. Sorry? No two squashes look the same. On the contrary, the band pattern which is characteristic is— I thought you meant the formaldehyde thicker. No, the formaldehyde is revealed— crossing reveals a local result, nothing related to anything macroscopic on the scale that you mentioned. Only at the level of individual protein-protein contact. So those experiments are of course done without squashing, but it wouldn't matter. How short range is that with the polytrombos? The range of the interaction of polytrombos is exactly the same as the range in diploid cells or in embryonic tissue. So I showed you that the TADs are typically in the order of a megabase in size, and that will be true throughout. We talked about this earlier, but can one be sure that there's no fat body contaminating your polyethylen preparation? Well, in the first place, of course, I absolutely don't know the answer to that question. All I can tell you is that when salivary glands are dissected, they give this result, and it is equivalent to what is found with KC cells of Drosophila. And equivalent to what is found with embryonic tissue of Drosophila. All I can tell you. Now one can, of course, put a ruler to the picture in the light microscope of a polytene chromosome, and with knowledge of the DNA coordinates from the deep sequencing, from the Hi-C analysis, state rather precisely what is the degree of compaction of the DNA along the length of the chromosome. And for this purpose, we actually took measurements that were made long ago by John Sedat, and they reveal what is shown here, that in between the bands, the region referred to as interbands, the packing ratio defined as the length of the DNA contained in that region of the fiber, packing ratio is between 5 and 10. And it just happens that by now, what, 40-some-odd years ago when I put forward the idea of the nucleosome, I could calculate from what I surmised to be the DNA content of a nucleosome and the approximate diameter of a particle, calculated a ratio of 6.8. So without any doubt, what we have in interbands are open chains, extended chromatin fibers. We know from the sequence analysis they came in about 5% of the total DNA, and within the interbands reside, for the most part, the regulatory sequences and promoters of Drosophila genes, those that undergo transcription. Drosophila cytologists and geneticists identify two types of bands they call gray and black. And again, by such measurements, we know that the gray bands have an intermediate packing ratio between the interbands and black bands. They represent such fibers condensed some 2 to 10-fold. They contain about a quarter of all the DNA and, for the most part, the bodies of transcriptionally active genes. And finally, the black bands contain such fibers now condensed 20 to 30-fold, the majority of the DNA and all, and in every case, genes that are not transcribed. So this gives some color to the picture that I showed you before. Now we have a better idea of what is the organization of the chromosomal material within the interphase nucleus, and in particular, what is the structure of a gene undergoing transcription. So I— that is the conclusion of what I have to say, and I know time is short and everybody wants to hear what Sid will comment at the end. I only wanted to reiterate what I said at the beginning and to emphasize something which I think is often forgotten, namely the insights that Sid derived and everything that I've told you comes not from studies of human cells, but what we call model organisms. And I think that there's still much to be learned from work on these lines, that the best is yet to come. Congratulations, Sid, and thank you all. Roger, a question that is coming up all the time is a very differentiated cell versus a non-differentiated itself, and if you can actually detect any differences there. In these experiments, we have not attempted to detect such a difference. But let me add that it may be helpful to mention polytene chromosomes, of course, lack one feature that Mark mentioned and that is common in very many differentiated tissues that would be an important property of what Tom Manieris alluded to before the looping of chromosomal material to bring enhancers at some distance in proximity to promoters. Of course, that's not possible in a polytene chromosome because you have 1,000 copies lined up and they are comparatively rigid. And indeed, one can show that, for example, in the— some of the important loci that exhibit this, I don't know, antennapedia or others where this happens. Those features are detectable by this Hi-C analysis in cells of an appropriate origin, the differentiated cells where it occurs, and not in the polytene chromosomes. So what we look at in the polytene case is the— if you were— the basic level of chromosomal organization of interphase nuclei in general. There are additional features that may be superimposed in differentiated tissues. You left out a section of chromosomes having to do with IAT content. Indeed. I'd be glad to mention that. I just was worried about the time. But I'll tell you in a word. In fact, I could probably go back to it very quickly if I can recover this. But if not, I'll just say it in a word. I think it's best not to try. So it had been suggested by many that the mechanism of formation of a nucleosome-free region depended on the abundance of AT-rich DNA. Mark alluded to that point, and it was thought that that it is a passive mechanism, that it is because of what had long been known that AT-rich DNA in particular, poly-D-A, poly-D-T, is actually refractory to the formation of a nucleosome. It is sufficiently rigid it won't bend around the histone octamer to form a nucleosome. We tested that idea in a straightforward way. We discovered we could form nucleosomes. They had— the histone octamer had a comparable affinity for nucleosome-free regions from yeast as for DNA derived from the bodies of genes. We went on to show that if we assemble a nucleosome on DNA from a nucleosome-free region, and side by side DNA coming from, for example, an open reading frame, that the RISC remodeling complex would remove the nucleosome at least an order of magnitude more rapidly from the NFR. And then finally, we could show that if we start with a nucleosome assembled on DNA that is of typical origin and has no significant oligo-D(A)-oligo-D(T) stretches, for example, RISC would remove a nucleosome only very slowly. If we insert simply a 7-residue stretch of oligo-D(A)-oligo-D(T), then the nucleosome is rapidly removed. So indeed, as people have suggested, the presence of AT-rich sequences plays a role, but not through a passive, rather an active mechanism. It's something which is controlled, it's ATP-dependent, and it involves the function of the RISC chromatin remodeling complex. Any last questions?