Scientist Stories: Roger Kornberg, The Nucleosome

Well, this is our lucky day. We have not only one but two keynote speakers today, and it gives me tremendous pleasure to introduce our second keynote speaker, Dr. Roger Kornberg, who traveled all the way from California, where he's a professor at Stanford Medical School, to be with us today. And as you guys all know, Roger was awarded the Nobel Prize for his amazing work on RNA polymerases, but he's also a pioneer, as you've heard from Gene, in nucleosomes, and he published his first nucleosome papers in 1978, before most of you were alive, although— '74, I'm sorry. Really a pioneer, 1974. And even in those days, he was really interested in the effect of DNA sequence on nucleosome structure and positioning, and this is a very timely topic as we've heard, and also in the, in what happens as he put it so beautifully, what happens when an irresistible force meets an immovable object, the immovable object being the nucleosome and the irresistible force being the RNA polymerase. And I think you'll be talking a little about this. So thank you very much for being with us and please welcome Roger.

Thank you very much. So I do owe Carolyn a particular thanks. I'm here because of your persuasive powers, and so Carolyn really gets the credit or the blame, depending. And I think everyone knows and you've heard and even some in, I should say, impressive detail about the contribution of histones in the nucleosome to chromosome structure, and I will therefore focus almost entirely upon histone and nucleosome function. In fact, as I was thinking about doing so today, it struck me how I should really have appreciated what is part of the point of my lecture long ago. I might, if I had thought more carefully about the subject, already realized at that time in 1973 what is one of the comments that I will make in the course of the lecture today. I didn't want to take much time to review the history of the subject, well known to many of you and certainly not of interest to those much younger for whom the nucleosome should be taken for granted, but I would like to point out what it was that was known at the time that might already have led me and perhaps others to realize the significance of nucleosomes for chromosome function. I'll mention at the outset the contributions from colleagues to some of the work I'll describe. So in the course of the lecture, I'll speak about the structure of an RNA polymerase II pre-initiation complex, and that is due to the work especially of Kenji Murakami and Nir Kalisman. With a contribution from a former postdoctoral fellow, Francisco Asturias, and his current postdoc, Kwong Lee Tsai, my longtime colleague Dave Bushnell. Probably there won't be time, what I had thought to do, to mention work of Phil Robinson. We'll see as we go along. I'll speak especially about the transcription of chromatin. This is the work of Shigeaki Nagai, of Yale Lorch, and of Barbara Meyer Davis. Now, the information from the past, which as I say might have already led me to one of the conclusions that I'll mention towards the end, shown here. So this is what we knew at the time in 1973, and especially I would call your attention to what has already been mentioned, in chromatin roughly equal weights of histone protein and DNA, and age-old observation. Not a lot of information to go on at that time, and to make matters worse, there was important misinformation. The ratios of the histones appeared to vary from one organism or even tissue to another, and so on and so forth, and in particular, we believed the unit size of DNA from the paper that Gene Thomas mentioned of Hewish and Burgoyne was 500 base pairs. Now I won't explain why we thought that, but that was the information available at the time. And the— what followed, again, I won't take time to explain. Jean Thomas, fortunately, has mentioned some of the background. She alluded to the discovery of the histone tetramer, and as I said, I wasn't going to go into this, but I think it might amuse you to know I seized upon the paper of Bender-Westhausen and von Holdt which represented a failed attempt to separate the histones. They, unlike all their predecessors, were unable to separate the several histone types from one another. I won't explain why I pursued that, only to say that I was sure that I would succeed in demonstrating that they had isolated a histone H3-H4 dimer, and so I decided for my purposes as a postdoctoral fellow at the time, I would isolate and crystallize the H3-H4 dimer.. And I mentioned this to Jean. I told her I was going to do equilibrium ultracentrifugation for the purpose, and she suggested, as she mentioned, doing chemical cross-linking, in which she had experience and expertise. And then you've heard from her the outcome. It proved not to be a dimer but a tetramer. And from that single fact, together with the few additional points that I've mentioned, and the misinformation notwithstanding, I could deduce the existence of a particle consisting of the 8 histone molecules about which were wrapped approximately 200 base pairs of DNA. Now for the purposes of understanding histone function and the comment that I made, I will explain at the end how I could perhaps have had an idea all along. The important points from that time were first, one could calculate from equal weights of histone and DNA and the idea of a histone octamer, a unit length not of 500 base pairs but of 200, and secondly, one could calculate from the longest X-ray spacing measured, 110 angstroms, a packing ratio of DNA in the chromatin fiber due to coiling in the nucleosome of about 7. Now I will return at the end of the lecture to the significance of these two observations, for the function of the nucleosome, which will be the subject of most of the rest of what I have to say. I think the first clear evidence regarding the function came many years later in 1987 when Yalie Lorch and I showed that wrapping promoter DNA in a nucleosome would prevent the initiation of transcription by RNA polymerase II. And then in the very next year, Min Han and Michael Grunstein showed that depletion of nucleosomes in yeast cells in vivo would allow the initiation of transcription. And from those observations, we—Yale and I and Michael and Min Han—could conclude that the nucleosome serves as a general gene repressor. The nucleosome prevents expression of all the many thousands of genes in eukaryotes except those whose transcription is brought about by specific positive regulatory mechanisms. Now the experiment that Yali Lurch and I performed to demonstrate inhibition of transcription in vitro was done with the purified proteins that were available at the time, and these were the polymerase and the general transcription factors, BDE. E, F, and H. The polymerase we already knew from the previous work of Bob Rader and others is alone capable of unwinding DNA and of transcribing DNA, but the general— but in the absence of general factors, the polymerase alone would be unable to recognize a promoter or initiate transcription. And I'll explain to you now how structural studies that we've done over the past couple of decades have shown, first of all, why the general transcription factors are required and what is their role, and more particularly why polymerase and the general factors are subject to repression by RNA— by the nucleosome. Why it is, how it is that the nucleosome can serve as it does. As a general gene repressor. Now the structural studies began with RNA polymerase II, and I'll just show you one picture of the structure pertinent to the rest of what I have to say. So here you see the structure of a transcribing complex which was solved about 16 years ago, and it shows the DNA double helix as it enters the complex unwinding before the active center. There's a sharp bend in the template strand in consequence of which the axis of the entering DNA double helix is at nearly right angles to that of the exiting RNA-DNA hybrid helix. And this is the one point about the structure that is relevant to the rest of what I have to say. The binding site for nucleic acid in a transcribing polymerase requires a right angle bend of the DNA. Now in the years after we pursued this, structure and related ones, learned much about the process of transcription, but we were unable for the most part to understand the roles of the general factors, unable for the most part to determine the reason for their complexity and their contribution to the initiation of transcription. It was only 3 or 4 years ago when Kenji Murakami succeeded finally in assembling a complete pre-initiation complex. Consisting of the polymerase and 5 of the general factors with actually some additional components, we could arrive at what is shown in the next slide, which is a cryo-EM structure of the complete PIC at, depending on the part of the structure, from 6 to 11 Å resolution. Superimposed on the electron density map from cryo-EM are the crystal structures of the polymerase in gray,— the crystal structures of TFIIb here in red, the TATA box binding protein, portions of the structures available from crystallography of some of the other components, A, E, F, and a part of H. Now, in addition, one could see at this resolution— whoops— the DNA and even the grooves in the double helix. And if we rotate around, and so I will turn, as it were, around to the left so that one views down the axis of the double helix, this is the notable finding. Here's the DNA in green. It is associated with a component of factor H, which is a DNA-dependent ATPase helicase that is responsible for unwinding the DNA to create the transcription bubble for initiation of transcription. But what is significant about this picture, there is no contact between DNA and RNA polymerase II and a Pol II pre-initiation complex. And in retrospect, we can understand why and might even anticipate it and can deduce what is the reason for the general factors. Because duplex DNA is comparatively rigid, certainly on the scale of couple of turns of the double helix. It must follow a straight path, whereas I told you the binding site for DNA in the transcribing complex requires a 90-degree bend. So the general factors recognize the promoter DNA in its straight, unbent form, and they bring it to the surface of the polymerase situated directly above the active site. The helicase melts the DNA after which in single-stranded form it's free to bend, can descend into the cleft allowing the initiation of transcription. Now coming back to the picture from the previous slide, you can immediately appreciate, it's perfectly obvious why assembly in a nucleosome would inhibit the initiation of transcription. Clearly impossible for both the DNA both to be incorporated in such a complex and also to be wrapped around histones in the nucleosome at the same time. Well, this is a point which one surmised even before the structure was available, and determination of the structure really validated the idea. We have since gone on to add mediator to the picture, mediator which we discovered long ago is responsible for communication between regulatory proteins proteins in the machinery and is the remaining component of the core machinery for transcription, assembles with polymerase and the general factors in a 3 million molecular weight complex, 2.5 million of which are shown here in a cryo-EM structure done by Phil Robinson, and it simply serves to reinforce the conclusion from the previous slide. Now, the question then, of course, is, has long been, how is repression by the nucleosome relieved for transcription? And as you're all doubtless aware, the pioneering work of the late Ira Herskowitz gave the first insight, the first step in the answer to that question. So Ira identified by genetic means in yeast the SWI/SNF switch and SNF genes and correctly deduced that they form a complex which in some way relieves the inhibition of transcription by histones in the nucleosome. Some years later, my colleague Brad Cairns isolated another chromatin remodeling complex closely similar to the SWI/SNF originally identified by Ira Herskowitz we refer to it as RISC, for remodels the structure of chromatin. It's larger than SWI/SNF. It is essential for cell growth, unlike SWI/SNF, and much more abundant, and it really is the principal remodeling complex for the purpose it serves also in higher cells. Over the years, Brad and Yali Lorch and I and others showed RISC, like Switch SNF, is able both to slide and also to remove histones from DNA in an ATP-dependent manner. And on this basis, one might imagine that the problem was solved. So at this point, there was a straightforward way of understanding gene activation and the initiation of transcription. It was only required for a remodeling complex to remove a nucleosome from a promoter would then become available for assembly with polymerase and the general factors, and transcription would ensue. Even though it would seem like the problem was solved in what I have just described as straightforward, it would nevertheless obviously be necessary to prove the point, and to do so by isolating a gene in its native state and bringing about its transcription in the presence of remodeling complexes and all other essential factors. Now we— of course, there were previous studies done with chromatin assembled with purified histones and promoter DNA in vitro along these lines, but we thought it important to repeat the experiment with a gene isolated from nature, a gene assembled in chromatin in the form that occurs in cells, and bring about transcription with the proteins at hand and whatever else might be required. Important because even to this day we don't actually know the histone modification state of a single copy gene in a eukaryotic cell, the promoter or in the vicinity thereof. We don't know always the precise location of nucleosomes, and that can be difficult to reproduce in an experiment performed by reconstitution of DNA with with purified histone proteins. We are, at the present, unsure whether or not there may be additional proteins associated with promoters in cells that are important, even required. So therefore, Shigeaki Nagai, following a procedure that we and others had developed earlier, pursued the isolation, in this case of the FO5 gene from yeast in the form of chromatin, and it was done by insertion of sites for recognition by a site-specific recombinase both upstream and downstream of the gene, then induction of recombinase expression in the yeast cells in vivo, so the gene was excised in the form of a chromatin circle free from the rest of the chromosomal material, and it could be isolated, purified a couple of hundred thousand-fold by affinity chromatography for the for the purpose of transcription. Now when Shigeaki then added polymerase and the general factors, he found, as expected, no detectable transcription from the purified chromatin circle, whereas when DNA was extracted, the histones removed, then of course it served perfectly well as a template for transcription by, again, polymerase. Race and the general factors. This only recapitulates the observation that we had made many years before and, as I say, supported by the studies of Han and Grunstein also performed not in vitro but in vivo. When Shigeaki then pursued additional proteins that might elicit transcription from the repressed gene in the form of a chromatin circle, he began with the FO4 gene activator protein for FO5. He found a requirement for what had believed to be an elongation factor, but actually genetic studies of Fred Winston and others have shown that TF2S is also important in the initiation of transcription in yeast in vivo. He added mediator and then also SAGA for this TPP— this TATA box-containing gene, a remodeling complex, and for most studies SWIT/SNF. And then he found what is shown here, now significant transcription from the gene isolated in the repressed state, shown to be, as it were, promoter-dependent in part by changing a base in the TATA box which abolished transcription, a reaction that required not only polymerase and the GTFs but all the additional proteins., but what was perhaps most striking, we thought we would be successful if we could succeed in bringing about transcription from the repressed chromatin template at a level comparable to that obtained from naked DNA. We didn't anticipate that we would find a level of transcription from chromatin much higher than that that is obtained from naked DNA, and that's a property which which proved to be dependent on the concentration of all the proteins in the reaction as Shigeaki progressively diluted the entire set of transcription proteins. The ratio of transcription from chromatin to that from naked DNA increased to the furthest point that he could reach upon dilution where it was nearly 20-fold greater from chromatin than from naked DNA, and I strongly suspect that at the concentrations and under the conditions obtain in vivo, in fact the nucleosome or chromatin is not only permissive but required for the initiation of RNA polymerase II transcription. Shigeaki went on to show that a requirement for acetyl-CoA in this reaction required for transcription from chromatin but not from naked DNA, and also an important contribution from the H3K4 trimethyl modification well known to be a signature of a gene engaged in Pol II transcription, and by deleting one of the subunits of the SAGA complex known to interact with H3K4, he saw the relative transcription greatly diminished. Now, at this point, one might summarize as follows. As I've told you, the level of transcription from chromatin is not only greater than that from naked DNA, but it increases relative to that from naked DNA as the concentrations of transcription factors decrease, and that would imply an interaction between those factors and chromatin. And then the question becomes which component of chromatin, or which nucleosome if you will. And to address that question, let me mention or remind you of the classic studies of Wolfram Hertz and his colleagues. So the late Wolfram Hertz and his group showed long ago that the FO5 gene in the repressed state is associated with an array of nucleosomes that cover not only the TATA box and transcription start site, but also one of the regulatory elements that binds the FO4 transcriptional activator protein. Hertz and colleagues showed that in the activated state, Following induction of FO5 transcription in yeast in vivo, two nucleosomes, one of which covers one of those regulatory elements, are removed creating a nucleosome-depleted or free region, and the nucleosome that covered the TATA box and transcription start site is not removed but shifts downstream and now exposes the TATA box but still covers the transcription start site. So we could now, following the procedure that I mentioned, insert sites for the recombinase both upstream and downstream of the promoter and the so-called +1 nucleosome, isolate a promoter circle in the repressed state with 3 nucleosomes from yeast and perform the same experiment. So Shigeaki has done that and he has seen that once again the isolation relative transcription from chromatin compared with DNA is greatly enhanced. In this case, the experiment that is shown, a result performed on the activated circle, there is only a single nucleosome present, and it covers the transcription start site, and it is responsible for a significant enhancement of transcription over that obtained from naked DNA. On this basis, we suspect that we have not proved that a proper pre-initiation complex goes beyond polymerase, the general factors, and mediator, and may well include SAGA, a remodeling complex, and the nucleosome. Now, I wanted to mention that the conclusions from this work on Pol V, we believe, are general, they apply to all genes in yeast and indeed we think most if not all genes transcribed by polymerase II in eukaryotes. And the reason is because of what is shown here. So this is a picture which will be familiar to many of you. It's a map obtained by micrococcal nuclease digestion and then DNA sequencing of the ends of the DNA produced by micrococcal cleavage. Showing the distribution of nucleosomes, in this case across about 100 kb of yeast chromosome 6. And what it reveals here, the likelihood of a nucleosome plotted as a function of position in the genome, the more or less monotonous array of nucleosomes punctuated at particular sites by these notable absences, and if we zoom in on one of the regions, of what you see, those absences consist of a region of DNA in the order of 150 to 200 base pairs immediately upstream of start sites of transcription, which lie just inside the boundary of what I explained a moment ago, maybe referred to as the +1 nucleosome. We've studied how these nucleosome-free regions are formed, we and others, especially Madhani and his group and others have, I think, good reason to believe that the remodeling complex and especially RISC performs that important role. But what I would like to just emphasize for now is the occurrence of the transcription start site almost invariably about 10 to 15 base pairs inside the boundary of the +1 nucleosome. And recent work of others, and I can't recall the details, has shown that the kind of movement of a nucleosome associated with activation of the FO5 gene is quite general, and that this picture is one that pertains very widely amongst eukaryote organisms. So with that, let me return to where I began, namely the idea of the nucleosome as a general gene repressor. Well, what I think emerges from the studies that I have just described, the nucleosome, we believe, performs a dual role. It is on the one hand important for silencing transcription, but it is on the other hand required for the initiation of transcription. So I think it may be viewed in a way as imparting to the regulation of transcription an important level of control. It allows what you might call a two-step authentication process. So in an initial step when a gene is prepared for transcription in the appropriate cell type, we know that nucleus hypersensitive sites will appear. Clearly a structure that contains probably a nucleosome-free region or whatever may be the counterpart in a higher cell is created, but the nucleosome remains on the transcription start site, and that prevents the initiation of transcription except and until the appropriate regulatory factors interacting with enhancers and transmitting a signal through a mediator for the assembly of a pre-initiation complex at the promoter finally takes place. Now the nucleosome then serves as a general gene repressor, and finally I might emphasize that it does more than that. It serves as a general transcription repressor., and there are many examples of evidence from the literature, and I'll cite one in particular, that of Craig Kaplan and Fred Winston, who showed years ago that if one cripples an important histone chaperone, SBT6, then one finds depletion of nucleosomes in the open reading frames of yeast genes and extensive spurious transcription initiation in the open reading frames. Of yeast genes. Clearly then, the nucleosome performs an important role that is more general than the regulation of transcription at promoters, as I've mentioned, and is more general and extends to cryptic transcription start sites throughout the genome. So with that, I return to where I began, and I'll explain why I said what I did at the outset that I wanted ought long ago to have realized what is an important point regarding the functional role of the nucleosome. So the calculation that I could make long ago that led to the idea of 200 base pair unit size of DNA as the repeat length in chromatin was based on dividing, or as it were, the calculation of the amount of DNA per histone octamer, that in turn assumes that all of the DNA is associated with octamers. Well, clearly all must be bound in some way, shape, or form by the nucleosome if it is properly to extinguish transcription, including preventing all cryptic initiation sites from participation in the process. But beyond that, what is striking, histones are the major chromosomal protein. I don't know the exact percentage, but they constitute the vast majority of chromosomal protein. But I could calculate at the time, and we have since verified for interbands in Drosophila, the packing ratio resulting from packaging in nucleosomes is about 7, a far cry from the majority of chromosome condensation that occurs in the interphase nucleus, which we have shown ranges from 30 to several hundred, and which as everybody knows must be several hundred thousandfold in the mitotic state. So the majority of chromosomal protein is used to achieve only a very small extent of condensation. As I say, from this information I should have realized at the time that an important role, perhaps the evolutionary purpose of the histones, goes far beyond structure and that it really pertains to function. And then finally, a concluding remark, if you were Mother Nature wanting to design a mechanism for extinguishing all transcription except that required in a particular cell type and brought about by specific positive mechanisms, well, you might have thought of coating the DNA with protein, but that would be a highly inefficient process. We know that typical DNA-binding proteins range from 25 to 100 100,000, and they will protect 10 or 20 base pairs of DNA. Far more efficient would be, instead of coating the DNA with protein, to coat protein with DNA, where we can wrap a single molecule or assembly of proteins about 100 KD with not 10 but 150 base pairs of DNA. So with that, as I say, And in the context of this gathering, which is partly retrospective, I can only fault myself for not having realized the conclusion I have drawn in this lecture long ago. Thank you for your attention. Thanks, Roger, for an excellent talk and very exciting talk. I bet there are questions, and I see here already the first one.

Tim. Thank you, great, great talk, Roger. Remember way, way back in the days of bacteria and Don Carruthers, where he claimed that the lac repressor was a transient activator? Is it— do you know in these pulled-out systems of natural chromatin if the enhancement in transcription you see is due to initiation and/or elongation.

So these experiments look only at initiation, but further to your question, we do not yet know if other proteins are involved, and the reason is because the amount that we can isolate of a single copy gene in chromatin is at the atom molar level. You know what that number is, not everybody would. And so we can't actually detect the proteins. So the only way to really pursue what I have described and tell if it's true is finally to succeed in reconstitution, but in the full extent that is required, namely with histones appropriately modified and placed at the right place and then show once again in such a system one sees an enhancement of the initiation of transcription. We're struggling to do that now. We've so far failed. I think the failure is telling, because if we place a nucleosome, we begin with unmodified histones, we see no success and no enhancement, indeed no transcription at all. Now we're trying, of course, to institute the relevant modifications and so forth.

Now the real point of the question is, not that you need to answer it, is it simply the loading of the pre-initiation complex and having it ready to go which gives you a rate increase, or is it that some aspect of the histone proteins become part of the machine itself, like the wheels or something?

Right. So I'll tell you what I— would speculate. I mean, what I think— what we actually believe is going on, and we're fairly sure from circumstantial evidence but have no proof. We know that two components in particular interact with the nucleosome: SAGA— well, I mentioned already the contact that it makes— and also mediator, and that we showed long ago. And there's whole other story relating to the mechanism of transcriptional activation through mediator, which I also recently, perhaps in my advancing senility, believe that I understand. But nevertheless, I can tell you privately how I think that works. But coming back to your question, we think both SAGA and mediator do interact with the nucleosome, and what's interesting is what one finds conserved across or common across genes in yeast and elsewhere. It isn't the location of the nucleosome relative to transcription start sites but relative to the position of the TATA box or TATA-like sequence that interacts with either SAGA or the TAFs. So it is the spacing from that position to the nucleosome seems to be the conserved feature. And that would make perfect sense if indeed there is a larger complex assembled that is, as it were, enhanced in energy or occupancy by the presence of the nucleosome.

And I think that addresses your question. The question also is whether some part of the histones are in the complex that runs down the—

No, I don't think so. Then I'm fairly sure what happens next. We know that the SSL2, which is the DNA-dependent ATP helicase of TFIIH, not only is responsible for creating a transcription bubble, we know that we've shown that remarkably the initial bubble is 80 base pairs, not 15 which occurs in the transcribing complex. Those 80 base pairs are a result of that molecule functioning, as you might imagine, as that helicase functioning as a translocase. Now I can comment to you privately how that melts the DNA, which I think we understand contrary to what you might imagine, not really functioning as a helicase. But more to the point, that bubble will extend well into the +1 nucleosome. So in the course of expanding the bubble, it unavoidably displaces the histones from in the DNA. But none of that's been proven.

That's just idle speculation. Last one, Roger. In this very nice study, you showed us when you isolate the little transcription mini chromosomes, did you try to mass spec what you actually purify on the— is it only only histones or what else is coming down? This is my first question. And then the second, when you mentioned GTFs, is this done with TF2D or with TBP?

So first of all, answer to your first question, at a molar quantity, or even if we really kill ourselves and make a lot from hundreds of liters of yeast, uh, maybe femtomolar quantities, we still didn't have enough material to see additional proteins in fact, at first blush we don't even see histones, and the only place that's been done really very effectively is by my former colleague Joachim Griesenbeck, and you heard that mentioned in one of the previous talks, where they succeeded in seeing histones when they isolated not a single copy gene but one that occurs in hundreds of copies, ribosomal RNA genes. In answer to your second question, we've done this experiment so far only with with a TATA box-containing gene representing about 20% of those in the yeast genome, and that only requires TBP, not TFIIID, but SAGA replaces the TAF complex in that context, and I'm fairly sure they're all the same. I think whether you have a TATA box and use only TBP but then you require SAGA, or whether you don't have a TATA box but a TATA-like sequence such as Frank Pugh has described, and then you bring in TF2D and that in effect recruits SAGA.

It's all the same. So this dogma is over. We just published now two papers where we show that this is not true. Which is what? Tell me. That TF2D is acting on all active genes and SAGA is also acting on all active genes. Oh, I didn't mean to suggest otherwise, of course. Until now it was said that 90% of the genes are regulated in yeast by TF2D. No, 20%.

If I understand correctly, some 20% of genes have a well-defined TATA box. I'm not questioning that. And I think the remainder do not. The important point, which if I understood you correctly, you have shown, they're all essentially the same in regard to mechanisms. Initiation of transcription. Is that correct? So what— They all use the same components, they all work in essentially the same way.

Yeah, so they both—

every active promoter both uses TF2D and SAGA.

Ah, that I didn't know.

Yeah, this is coming out in a week. Well, it adds up to the same thing in the end. Okay, thank you. Any other questions? Okay, so let's thank Roger again for this insightful talk and the great discussions.