Donald G Comb Lecture - Eva Nogales

See Dr. Eva Nogales present her work on the "Structure and Dynamics of Macromolecular Complexes using Cryo-EM."

Script

Tom Evans: Hello everyone.  Welcome to the inaugural Donald G. Comb Honorary Lectureship!  A scientific seminar series is a great way to recognize Don and his accomplishments and his passion for science. It’s very exciting to have Dr. Eva Nogales as the speaker to kick off this lectureship series.  After meeting with her, it is clear that she is the perfect person to start this series.  Before Bruno introduces Dr. Nogales, I want to talk a bit about research at NEB.  Research at NEB, and I am referring to all research performed here, is beautifully complex; encompassing basic, discovery driven projects that lead to publications first and foremost all the way to applied projects aimed at solving a specific commercial problem.
And the richness and breadth of our research is what makes us so unique.  We encourage basic research like an academic research institution, as exemplified by over 1300 peer reviewed publications in our history, AND we encourage very applied research like a single-mindedly commercial company, illustrated by our catalog of products and over 550 patents. 

All of this work is actively supported by NEB, giving us freedom from the whims of funding agencies or typical product-outcome funding.  What a fantastic idea.  It works so well.  The interplay between basic and applied research has allowed us to embrace changing technologies and customer needs.  We learn what our customers need by doing research like our customers. We also give back to the world through research; the obvious example is our work on filariasis, roundworms that cause blindness and elephantiasis in regions of the world that don’t have the resources to find solutions fast enough; interestingly, even with the advent of the Gates Foundation, NEB is one of the biggest funders of filariasis research in the world and our parasitology labs are leaders in this area.  The richness, complexity and freedom of our research program has been and continues to be an integral part of what NEB is as a company and contributes to our continued success.  Please join me in thanking Don for dedication to research and his vision that made NEB what it is today. 

Bruno Manta: There’s nothing common, standard, about her career path. Eva studied physic in Spain in the eighties, being one of the few women in class. When she was wrapping up her bachelor in physics, she met a scientist (Dr. Bordas) in a conference, was offered a PhD position abroad and immediately decide to move to UK, something not common for Spaniards at that time, much less for Spaniard women. She took risk and decided to study biological molecules using techniques that, at that time, were applied mostly to other types of materials. Her PhD thesis was so good that she ended up in moving to the US for postdoc. Based on her experience, she was given a simple problem: solve the structure of tubulin, a problem that was being under active research for at least forty years. 5 years and 4000 images later the problem was cracked, so she called her postdoc supervisor Ken Downing and said: “Ken, it’s done”. The rest is history.

This is how unique, lineal and outstanding will looks like Eva’s career if we focus only on the academic facts. But, indeed, Eva is a very common person, in the good sense. Indeed, Eva followed her fiancé to the US without having a plan for herself, and she end up being extremely lucky finding a postdoc mentor that changed her life. She got married and had kids at the same time she was assistant professor, something that is not recommended in the manual of “how to be a successful scientist”. She raised them, cook for them and worry for them. Even now, with one of the busiest agenda one can imagine, you may find Eva in a conference getting ready to deliver the keynote lecture or just shopping for shoes for her older kid or answering some school-related email. And, even being incredible busy as she is, Eva don’t miss a dance night ever, and I can assure that because I saw her in many conferences.
 
Before passing the mic to Eva, as a young scientist I also want to highlight Eva’s role on training the new generation. You may not be aware of that, but many of the current young leaders in the field of Cryo-EM were trained in her lab. Everywhere you go, from East to West coast, to Europe and beyond. …you’ll find former members of Eva’s lab. Her trainees went to academia, industry, consulting, etc.  The recipe for such a successful scientist factory is, in her words, very simple: some luck, a lot of freedom, good questions, hard work and a lot of room for fun. 

Eva is clearly a scientific leader of our time, so it’s a pleasure for me to have her here with us today.

Eva Nogales: Well, I have been here only for a few hours and I already love the place. I think you guys must feel very, very fortunate to work here. It really does feel ... I feel that happiness and the passion that Jim was mentioning earlier on in the morning. I'm also amazed, it has never happened to me, that people invite me to give a talk and it's not only that they know about my papers, but they know details about my life like some of the things that Jim or Bruno were saying. I'm very impressed.

Eva Nogales: In any case, I know that this is being recorded, but I think I've been told that it's okay if I encourage you to stop me at any time during the talk and ask me questions. I will start very general and then become more and more detailed. If you guys get lost at the beginning, it will be really bad, so I would not only not mind that you ask me questions, I would actually love it because it would mean you're actually paying attention.

Eva Nogales: All right. I'm going to talk about a number of things with the theme being what is in the talk, which is how we've been using cryo-electron microscopy, a technique that has really become a major methodology in structural biology, but that it was not so when I started working on it, to study about the structure and dynamics of very large macromolecular complexes that otherwise would have been impossible to characterize. I just want you to get an idea, this not working, of what it's all going to be about.

Eva Nogales: I'm interested in molecular mechanisms of what I consider to be essential biological processes for the life of the eukaryotic cell. My way to go about it, because I'm very primary is to see them, is to visualize them. This is all that my talk is going to be about. It's going to be about visualizing molecules in action. When I talk about molecules, I'm going to be talking about things that are typically larger than 100 kilodaltons, can be over a megadalton in size, or can be an infinite size like polymers. When I talk about seeing them in action, I mean that I'm going to see motions that can be as large as hundreds of angstroms and that are also going to be seen in the presence of interaction with other partners. I'll cover this with a number of examples.

Eva Nogales: Now, when I talk about visualizing, I'm talking about transmission electron microscopy. This is a technique that is very old. It's been used for many years in many different modalities. Top electric microscopes, if you're using them on inorganic materials that are not radiation-sensitive, these days, can give you from a single image, resolutions that are of the order of 0.8 angstrom. This is not the case for biological molecules because we have to deal with the problem of radiation damage. Just like when you shoot X-rays into proteins and nucleic acids, when you shoot electrons of the energies that we use, we damage them. We pass energy from the electrons into the sample. We ionize them. The ions move around, generate more radicals, breaking bonds. Obviously, we're able to do something, and that is because we take care of things in a number of ways.

Eva Nogales: The first one has to do with cryo-fixation. The cryogenic parts of our studies plays a double role. One is, as you well know, biological molecules work and are happy in an aqueous environment, but water is incompatible with the ultra-high vacuum inside an electron microscope. On the other hand, if you freeze the sample and you keep it a liquid nitrogen temperature, the phase diagram of water will tell you that you can put it in the vacuum of the electron microscope without the water subliming.

Eva Nogales: The second is that when the molecules are now at liquid nitrogen temperatures, the sample is going to get damaged by the primary electrons just like before, radicals are slowed down in their diffusion, so we get a little window in which we can bombard the sample with electrons and still get enough signal to be able to go down the pipeline. We still use low dose. We're not like the material scientists that put thousands of electrons per angstrom square in the sample. We put 10 to 20 electrons per angstrom square, so the images are very noisy. I'll show you some. But then, we overcome that. They have enough signal that through computational analysis, we are going to be able to visualize them in detail. I'll show you how.

Eva Nogales: Something that is very important for some of the molecules that we study is that we use incredibly little amount of sample. For NMR, for crystallography, it is required you have an over-expression system or otherwise rely on samples that are very abundant in the cell, like ribosomes, say. But, there are many cases where, I'm going to show you examples, the complexes are very large and there is no actual practical reconstitution system. You have to rely on endogen material that is very rare. That is not going to be a limitation for cryo-electron microscope.

Eva Nogales: In Cryo-EM, you start with a purified sample that is in the conditions in which it's happy, biochemically speaking. We thin that sample to a layer that is about the thickness approximately of the molecule of interest. Then, we freeze it very quickly, hundreds of thousands of degrees per second if you want. What that does is stops the sample so quickly that water molecules don't even have the time to reorganize into a crystal, so that water that is frozen in here is in anamorphosis states that actually preserve the aqueous layer of water around the molecules. It's in this frozen-hydrated states that this samples going to go into the vacuum of the electron microscope.

Eva Nogales: Once in the microscope, we're going to illuminate them with electrons, and we're going to generate images, that is very important for you to realized, are projection images. They're not a surface, they're not a section. It's the integral, it's the summation of the density in the direction of the electron path. Very much like an X-ray when you go to the doctors, and the image has your nose as well as the back of your head just added up on the image.

Eva Nogales: All right. This is an example. This is GroEL. Some of you may have recognized it, an example image. The first thing that we have to do, we take hundreds of thousands of images like that, that have tens or hundreds of representations of the object in different orientations. The first thing that we do is we call this picking particles, something that used to be done by graduate students, then undergraduates, and now we have artificial intelligent creatures doing it for us. In any case, with this, we generate galleries that, to start with, can be classified in what we call 2D classification. These selects from the gallery, all the images that are in one defined orientation. These images are related by rotations in planes and translations, so we can do two-dimensional alignment, and then put them together. By summing these noisy images, boost the signal, and we get now class images that looks a lot crisper. But, this is still 2D. We have to do this for all the different views of the object.

Eva Nogales: Then, we have to do something that I'm going to tell you is like magic, but it's not, it's mathematically sound, which is to find the relative orientation of these views with respect to one another. With that information, we can then combine them to get a three-dimensional structure. This will be the density map for GroEL. In this case, the resolution is about seven angstroms or so. This is a resolution in which alpha helices are clearly separated, so if you have crystalline structures of pieces, you can dock them in very accurately like a three-dimensional puzzle. In some cases, you can generate homology models or even from some prediction, you can use the map to interpret it in terms of a structure.

Eva Nogales: Of course, we can get to higher resolution, again, as this allowed, but it is how high can resolution go? It depends on the samples. Samples that are very well behaved, so samples that you can buy out of a catalog, that has been crystallized many times, and that are very well behaved; those can go ... Right now, I think the record is 1.6-angstrom resolution. Now when you are studying things that are very difficult, that did not crystallize, that are very large, that are very flexible; then it may be harder to get that resolution. I think typically is more like three or four.

Eva Nogales: At that kind of resolution, with maps from Cryo-EM, which have very, very clean, just think if there's any crystallographers in the room, we get amplitudes and phases from the images, so the density doesn't have to get refined. This is the final density. The only thing that we refine is the atomic model. At three, even four-angstrom resolution, it's actually possible to trace ab initio. This is it. This is a snapshot. These images are taken of frozen, hydrated samples that were frozen in time and the conditions where they're perfectly happy doing their biochemical reactions. The images are noisy, but we can average them, and then we can combine different views to get three-dimensional structures. In this process, we never had to crystallize. Because we're not limited to crystallizable units, we don't have to break things down. We can use the study, fully assemble complexes that are units of function and the conditions where biochemically active.

Eva Nogales: As I said, it does require a very small amount of samples, so I'm going to give you an example today, where we just use a few microliters that are at nanomolar concentrations. This is unthinkable. I always love to give this example. I collaborate a lot with my colleague Jennifer Doudna, and I remember the first time that they were working on a CRISPR system. It was not a Cas9, it was one of these larger complex cascade from e coli. They had been trying to crystallize it and were not successful. At a certain point, I said, "Okay, give me the Eppendorf tube after you suck the protein to put in the mosquito. We're going to just rinse it off and make some grids with that, and we got the structure, so you really need very little.

Eva Nogales: Now, this technique has been around for a while, but it has only been in the last, I would say, five or six years, that really help make it into the big scene of structural biology. That has resulted from new detector technology that has really revolutionized the field and new software that really has taken full advantage of those improved images. One of the things that this has allowed us to do is to deal with different functional states that co-exist in the sample, either compositional or conformational. What before used to be a challenge because it meant that having this variety made your averages very blurry. Now that we can separate them, it's a source of richness, of information because it means that we can get to the equilibrium of the states that exist for those samples, so we have extra information.

Eva Nogales: This has also allowed us to go better in resolution. I think this is the most common regimen, but as I said, very well behaved samples, better than two angstroms; very poorly behaved samples, maybe just down to 9 or 10. But, these two things actually go together because, as I say, typically, the reason why you cannot get too high a resolution, now that we typically not limited by the amount of data, has to do with having variability in the sample that leads to mixing and blurriness, but if we can separate, we can get to higher resolutions, so those two things have come together.

Eva Nogales: All right. I'm going to show you today two biological examples. One I'm going to go through fairly quickly, just concentrate on one example at the very end. Then, I'm going to go into great detail into another. As Bruno was saying, I started my scientific career moving from physics to biophysics in the study of tubulin and microtubules. With the new technology, we've been able to obtain atomic details of microtubules now. This is something that we do in a fairly routine basis.

Eva Nogales: This is the alpha/beta-tubulin heterodimer, as it assembles into protofilaments, that then associate into a tube. We're able to study these microtubules at resolutions that are around three, three in a half angstrom. We studied how they change with nucleotide, for example, but also, we study how they interact with many factors. This here is kinesin, the motor domain of kinesin bound to a tubulin dimer. Kinesin is important to establish the organization of the cell by moving organelles around. But, we study, especially, many proteins that are involved in mitosis. EB proteins are proteins that are very important for the dynamics of ends and for localizing different components of the cell with respect to chromosomes and microtubule ends. Brc-1, for example, is involved in setting up the organizational, the arrangement of microtubules. TPX2 nucleates microtubules around chromosomes.

Eva Nogales: But, we also study other type of microtubule-associated proteins, and I'm going to tell you a little story about that, and that's the case of tau. Tau is a neuronal, microtubule-binding protein that is important for stabilizing microtubules in axons and for bundling them, but also has a bad name because when it fold in microtubules in certain states, it can form fibrils that give rise to tauopathies, Alzheimer's being the best known.

Eva Nogales: I just want to show you the structures as such that we can generate ab initio models from them. In many cases, some of these structures of things that bind to the microtubule have been obtained by themselves, and is very curious. They tend to change the structure of the microtubule, but also they change conformation when they bind to the microtubule, so even in cases where the structure was known, it was actually not the physiological one when they are on the microtubule surface. But, also there were many that actually only take shape when they bind to the microtubule, and they bind across tubulin subunits, so you can not even bind them to a single dimer and try to crystallize them. That is the case for TPX2 and for tau.


Eva Nogales:
I'm going to give you the example of tau because it was the most difficult that we did, that we have to do, and I'll show you why. As I was saying, microtubules are involved in organizing the cell, in separating chromosomes during cell division. They are very important also for the morphology of cells, so in the case of neurons, they are very important to establish axons, make them grow, and go to where they need to make connections. Tau is very important in the axons. It is actually an unstructured microtubule-associated protein, so in solution, it doesn't have a structure. As I said, when it comes off microtubules, when it gets hyperphosphorylated, or in case of some hereditary mutations, it can form tangles that give rise to Alzheimer's disease or other diseases.

Eva Nogales: Tau has been studied for a long time, almost as long as microtubules. It's a fairly large protein, and the main region interacting with microtubules have been defined, is right here, is made up by these four pseudo-repeats that are depicted right here, that have some similarity, but they're not absolutely identical. It had been proposed that each one of them could bind to tubulin and the four of them will bind, adding avidity and affinity to the process. But, the structure of tau had resisted characterization.

Eva Nogales: Liz, Nisreen, and Simon got together and first looked at the structure of the full-length tau bound to microtubules. I hope you can see, and I'm just going to blow it up. It has this footprint of an extended polypeptide chain that goes over a tubulin dimer, so that would be a monomer-monomer interface within the dimer, but also across a dimer-dimer interface of the next dimer will be here. This is a polymerization contact, and it's really a staple those two tubulins together. The tricky here, the thing that made it very difficult is that tau has four repeats that likely have a very similar footprint, and what we're getting here is a mixture. They're so similar that we can not separate them and distinguishing them, so what you see here is a density that is the average of repeated one, two, three and four. If we wanted to do chain tracing, we couldn't do it with this mixture, so this is what we did.

Eva Nogales: First of all, we chopped everything but the four repeats and we saw that still, we got the same density, so this really corresponds to these repeats' footprint on the microtubule. But then in order to be able to get to an atomic structure, we generated a synthetic tau, that instead of having this four, distinct repeat, has four copies of repeat1, so we still have the affinity and the avidity that those four binding sites give you, but now whatever footprint we see has to come from the R1. When we did that, this is the repeat1, and it's colored by conservation, so some regions are more conserved than other. We could trace 12 C-alphas, but we were not sure what the register of sequences was. I'm showing you this because it was the most challenging case that we have to deal with.  

Eva Nogales: We got together with Frank DiMaio, who's is a specialist in Rosetta modeling, and what he did was to calculate the energies for 12 amino acids within this repeat, but in different positions, and then found one that gave us the minimal energy. This was okay, but we wanted to be sure that this was right, so we did ... By the way, the model was actually very appealing because, I didn't mention, but there is a critical serine residue that is hyperphosphorylated in Alzheimer's disease. These four part of the best defined region of this footprint, and it's involved in interactions that contribute to the affinity, so the idea was when the serine gets phosphorylated, this interaction with this glutamate will go away and this will contribute to it's coming off the microtubule and forming other kind of structures that are deleterious for the cell.

Eva Nogales: But, so what we did to make sure is we repeated this experiment, but instead of R1, which is a little weird, it's most distinctive of the repeats, we looked at four repeats of R2. In this case, we could trace 27 C-alphas, but they could go in ... Again, the shift with respect to these four repeats could be different. We found the same result, the same, this matched. The difference in energy is larger because now we're modeling a larger part, so we've seen more of that binding energy, and that gives us great confidence that the model was correct.

Eva Nogales: This is the two superimposed, the R1 and the R2, just to show you that the R2 has a larger footprint on the microtubule. This part that is common is almost identical. It's also the part that is conserved. Again, it includes the serine that I was telling you about that is right there. For the case of R2, which is more similar to R3 and R4, we can see this full footprint going across the dimer and then across a dimer interface making these key interactions that serve as a staple between tubulins of units. We could come out actually of our model because the only part that is hard to see is the connection between the repeats, but that is fitted very well by the remaining part of the sequence, which is this PGGG that tells us the four repeats come perfectly cross four tubulin dimers, working as a tandem, and therefore, really providing great levels of stability for the microtubule.

Eva Nogales: Just to let you know that, of course, at the resolutions that we have right now, we can look at the small molecules binding to microtubules. This is taxol. This was done a couple of years ago. We can get even better resolutions now, but this is taxol, which is very broadly used in the treatment of solid tumors. Other alternative to taxols that are being considered, zampanolide is different, but binds in the same binding pocket; or peloruside, which binds at a completely different site. There is enough resolution in there that we can pinpoint where the molecule is, but also start position it. This is accurate enough that now computational chemists can go into the structure and really tell you what are the most likely hydrogen bonds network that is keeping that molecule in place.

Eva Nogales: But, I want to use the rest of the time to tell you about really challenging study. Microtubules, of course, are ideal for Cryo-EM because they have polymers that cannot be crystallized, but that are really, really good sample for image analysis and Cryo-EM studies. But, this is another example of something that could not be studied by any other means. This is a large protein assembly, is the pol II transcription pre-initiation complex that has to assemble at the co-promoter of every protein gene in our inner genome, in the nucleus of our cells. It includes large factors like the polymerase its self, half a megadalton; or TFIID, which I'm going to tell you a lot about, which is 1.3; or TFIIH, that is also about a half a megadalton.

Eva Nogales: We started working on transcription when I became an assistant professor at Berkeley through collaboration with Robert Tjian, which is one of the major biochemists in the eukaryotic transcription. TFIID was very easy for him to convince me that this was an important complex to study. It's involved in the recognition of co-promoters for all pol II genes, no matter what combination of sequences. If it's a pol II gene, TFIID is going to recognize the core promoter and is going to be involved in the recruitment of the rest of the transcription pre-initiation complex, so bringing the polymerase to the transcription. It's that site that sits among those core promoter sequences. But also, it's actually important for regulation because it itself interacts with gene-specific activators and repressors as well as with post-translational modification in histone that marks areas of the genome that are active for transcription. But, it is very challenging, so it's one of these cases, it's 1.3 megadaltons, is made up of the TATA-binding protein that bind TATA box co-promotants, but also 13 TBP associated factors or TAFs, some of which exist in two copies, so this adds up to 1.3 megadaltons.

Eva Nogales: There's no effective reconstitution system. I believe now there's a bio archive with the first full reconstitution. I don't know how efficient it is, but when relying on purification from endogenous sources, we purify this from huge growths of HeLa cells through immunopurification, we get very limited amounts. This complex has very poor stability, so biochemists get around it by having it in crystal or sugars. We can't use any such thing because that will match the contrast of the protein. In the end, we'll have the same density the buffer as the protein, but the most challenging thing we didn't know when we started.

Eva Nogales: Some of the initial work was done by Patricia work. This was our first Cryo-EM structure of TFIID, and at that point, we just described it as three main lobes that were A, B, and C. We were very frustrated by the fact that the resolution was about 40, 45 angstrom. At the time when the methodology that we have today didn't exist, we teamed up with Pawel Pencek, who has been one of the early developer of software, and developed for our use a 3D variance method that, although will not give us the different states, it will pinpoint regions where there was a lot of change. TFIID really lighted up very dramatically, so we know that it was very flexible.

Eva Nogales: But, we did not really find out until years later when Mike, when he was a graduate student, now he has his own lab at University of Michigan, showed that of the three lobes that we very imaginatively names A, B, and C, the A lobe moves very dramatically with respect with a more stable BC core. Just to give you an idea, so this is a movie just where we put different classes together, superimposed them on this to give you an idea of the range of motion, the range of the states that we can see from one particle to the next. This about 400 kilodalton model moves by 150 angstrom, so this very dramatic. He quantitative that. He was a biophysics graduate student, so he quantitated it. This is just protein position of lobe A with respect to this BC core. This shows that this is a continuum, but bimodal distribution between what he called a canonical state, where lobe A is touching lobe C, and a rearranged state, where lobe A is touching lobe B.

                          
Eva Nogales: The thing that was really exciting, because up till here it was just a real nightmare, is nature just making things difficult for us, but there had to be a meaning to it. When he gave DNA and TFIIA, which helps TFIID bind the DNA, what he saw was a shift in that population towards the rearranged state. When he did the three-dimensional reconstruction, it was only the rearranged state that was binding to DNA, so it was very obvious that the different conformational states of TFIID have different capabilities to bind the DNA.

Eva Nogales: This is very important because this is a complex that serves as a hub or that collects information about histone states, activators, repressors, co-promoters; and then has to give an output, which is how much polymerase is loaded and how much mRNA is being produced. Having this flexible landscape that can be tuned by interacting with different factors to increase or decrease binding to DNA was really very exciting. But, this was still the blobology years of Cryo-EM, where we at a very low resolution because this is not state. This is many different states that we're averaging together. We have to do better.

Eva Nogales: At that time, Robert, another biophysics graduate student, that now is doing a postdoc at John Hopkins, decided to have another go, and what he first did was to purify TFIID that was stably bound to DNA. Because we have such small amounts, we cannot purify it by running a column or anything like that, so what he did was to biotinylate the DNA, then use streptavidin-coated magnetic beads, and in very low volumes add excess TFIID and TFIIA, then let it bind/incubate for a while, wash off whatever was not bound, and then release TFIID using a restriction enzyme that cut it off the beads. That will go in EM grids.

Eva Nogales: These are the two structures that Mike got before, and this was the structure of the TFIID purified that was stably bound to DNA, and I hope you can that the DNAs much more clearly seen, and you start to almost see like a shape in it. I just want to draw to your attention to the fact that this big blob in yellow, only one part is nicely seen and is the part that is binding this end of the DNA. The rest has become detached and is flopping around so much that have to play with the threshold, that's why there's two isosurfaces, to be able to see it.

Eva Nogales: What we do when something like that happens is we ignore the part that is smaller and flexible with respect to the rest. When we do our alignment, we concentrate on the rest because otherwise this flexible part is kind of blurring in the accuracy of the alignment. When he did that, which we fancifully called masking and focused refinement, he was able to improve the resolution. This shows a local resolution. It's a local resolution map, where you can see that there are regions that are more flexible and they're, again, seen at lower resolution. But, obviously, there are more features.

Eva Nogales: The resolution was enough that we could take the crystalline structure of TBP bound to TATA box DNA and IIA and fit it very nicely at this end of the DNA. That positioned the TATA, and then just moving along the length, we could position the other core promoter sequences that were in this promoter, and we could see that they were being touched by protein. But, what protein? This was still limited and this resolution is lower because, here and here, there is a flexing point. What Robert did was to mask one more time, and that improved lobe C to a resolution in the region of eight angstrom, where these alpha-helices can be very clearly separated.

Eva Nogales: Then, we were lucky because just a few months before, they had been a crystal structure of this complex of TAF1 and TAF7, where the presence of a winged-helix has lead the crystallographers to think that this could be involved in binding DNA. Here it is right there binding DNA. There was a region that was missing that was flexible in the crystal, but it turns out that it becomes stabilized when it binds to the DNA, the initiator motif in particular, and we could model this, this is small globular domain. Some bioinformaticists had identified TAF2 as having a structural homology to aminopeptidases, of all things, and we could take one of those structures of P10 there.

Eva Nogales: Then, someone had obtained the structure of a HEAT repeat within TAF6 and from a fungus. We saw that there were two copies in this end. Then, we proposed that was an alpha helix from TAF8, which turned out to be correct, but it was a stretch at the time. I still show it because we were right. But basically, we could see TBP binding on one side and TAF1 and TAF2 binding the downstream core promoter element. We have defined almost everything that bound the DNA, but we were still missing lobe B, we didn't know what was there, or all of lobe A, that we have even ignored.

Eva Nogales: That was the task for a third biophysics graduate student that published a full model of TFIID about just a little less than a year ago. He was using APO-TFIID, so by itself, big motions. He collected a huge data set and did massive sorting of different states. What I'm showing you here are three maps. The white transparent map is a real map that exists of one the states, but because there's a continuum of states, when you split the data to define all the pieces, you have so little data that you don't go to high resolution. But, you can combine more of the data and refine doing this streak of unmask the part the move and concentrate on that part. That allowed him to see these BC lobe at about four-angstrom resolution and lobe A, the one that is really, super mobile at about nine. TAF1/TAF7, when it's not bound to DNA, is just flopping crazily, so that is impossible to see when not bound to DNA. Ultimately, you get something like this, where it was possible for him to identify every component.

Eva Nogales: I'm running out of time, so I'm going to skip through how he went about getting this structure and instead, if you don't trust me, I can do it during the question time. But, one of the things that helped is we had enough resolution in this part to be able to model everything, and then we knew that there were two copies of some TAFs, that happened to be almost all the ones that are here, that have to be there. This lobe looked similar in shape, so we could duplicate it and then add whatever was left and it fitted the density very well.

Eva Nogales: In any case, those two lobes, B and A, do very different things, but they have common scaffold. Lobe B is made up of TAF5 which has an enzyme in a helical domain in one of these beautiful WD40 beta-propeller motifs. This binds to three pairs of histone fold containing TAFs. That is 6 and 9, 4 and 12, and 8 and 10. Those are the three pairs. They form something that is like a hexamer of histones out of a histone core in a nucleosome, but then they're bound to this. Out of these, we know everything is in two copies, except for TAF8. This is now lobe A. It's in a different orientation, but then you can see it in all its glory. TAF8 is in a single copy. That's what makes this lobe unique. It's tethers it to this HEAT repeats and to tap TAF2, so this forms this ... TAF8 allows lobe B to be more rigidly stable with respect to the complex.

Eva Nogales: In this one, TAF8 has been substituted by TAF3, partnering with TAF10, and then now, this is a surface that serves as a binding site for another pair of unique TAFs that also have histone folds, 13 and 12. These ones bind TBP, which in the complex, is completely inhibited and is bound, this has been shown previously also by crystallography, with end terminal segments that extend from this white subunit, the TAF1, and going into there. This is held basically through flexible linkers. One is to TAF1 from those end terminal regions that bind here to this part. The other one is through TAF6, which has this HEAT repeat and then a linker that goes into a histone fold, so there's one going into here and one going into there.

Eva Nogales: We can take that and now also we improve any structure of the DNA-bound TFIID, and we could put the model of TAF4 in here, so we also now have this structure of the engage state. I just want to point out that none of these histone folds come close to the DNA, and this was surprising. People had predicted that they were going to exist. They were even crystalline structures of some of these pairs. Because it binds DNA, was the obvious thing to think. It turns out it is not the case.

Eva Nogales: These are the two lobes, aligned so that you can see the similarity. This is a histone core colored, so that you can see also the similarity. There really is something that looks like a octamer here and like a hexamer here. But, this is the thing. In a nucleosome, these histones have a very characteristic pattern of electrostatics with this positive surfaces interacting with the phosphate backbones. Those in the case of TAFs within TFIID, have been mutated away. They don't exist. The histone fold has been used as a scaffold that is no longer involved in DNA binding, just like the aminopeptidase fold has been mutated in the active setting, it's just used as a scaffold for TAF2.

Eva Nogales: This is a very complicated structure. We propose a model of how it comes together because it's kind of pseudo-symmetric but not quite. It is known that some of the subunits ... It all starts in the cytosome, but we think it's completed in the nucleus. In the cytoplasm, we know, for example, this is not our work, that the end terminal helical domain of TAF5 has to fall within the cytosolic chaperoning. Then, it's only released from the chaperoning, APO-binding to this TAF6/TAF9 histone pair, which also has the HEAT repeat. That liberates at three TAF, competes with three TAFs. This is able to interact with the TAF4/12 histone fold dimer. Within, there are other, three complexes. All of them have in common that they have at least one pair of histone fold containing subunits. This is TAF8/10, this unique pair interacting with the TAF2. This is the 11/13, via that unique pair, interacting with TBP, which interacts with TAF1, which is to interact with 7. Then, we think this exists as a unique element also, that TAF3/10 unique pair.

Eva Nogales: What these things have in common is that each one of these complexes have one nuclear localization sequence. This allows all of the component, including the ones that do not, to get into the nucleus. In the nucleus, we propose, this is based on our structure, that there are two complexes that come before and that form, one in which this combines with this, one in which this part combines with this. That is going to give rise to lobe B, and this is going to give rise ultimately to lobe A. This cannot interact with itself because of static hindrance of this, and this cannot interact with itself because the HEAT repeats by themselves have very poor affinity. It's only this combination that can form a stable dimer. That now generate this pseudo-symmetry, where these elements have this unique pair that can now interact with this and add the TBP module to the complex. This is our proposal, which just came out yesterday electronically in a Current Opinion. That's where you can go crazy with ideas.

Eva Nogales: There it is. What we have studied, and this is now based on data, are the dynamics of TFIID and how this is important for the loading of TBP. What we did to study this was to go back to square one and just mix the TFIID with the DNA and TFIIA, and see what's in there, what's in the mixture. We see these states, where you remember ... I mean, these are just the two ends, but this thing is bouncing from here to there. This is the same that we see when we study just APO-free TFIID. I'm always marking TBP in red. In this state, is where TBP is completely inhibited. It has the end terminal segments of TAF1 binding the concave and the lateral surface and it's also engaged with this TAF11/13.

Eva Nogales: Then, we see this structure. It's not very abundant, but it is there. We think it's an intermediate, in which the downstream DNA has already been bound tightly by the TAF1/TAF2 module. We know that because there's a lot of density for the DNA. While as you move away, there's less density, mean the DNA is flopping around because it's not very engaged very tightly. What we are proposing is that upon binding these sequences, these now positions the lobe A right here, where the DNA's being held so that TBP can start scanning looking a TATA or a pseudo-TATA, something that it could ultimately bind to. This gives the DNA a chance of competing away this otherwise inhibitory peptide. This is just due to local concentration and the geometry of the arrangement that TFIID's providing, upon binding these sequences. In here, we propose that TBP's kind of touching and going with very fast on and off rates, not bending the DNA, not engaging it very strongly.

Eva Nogales: We also see this, which is very similar, but it has TFIIA, so what we propose is that is this case, TFIIA, which binds to lobe B, is again being placed so that it has the right geometry, the right orientation, high local concentration to be able to compete away this inhibitory peptide. Under these conditions with the help of 2A, TBP really has a good chance and eventually, does this high-affinity binding, where it bends the DNA, and it clicks into position. This is incompatible with this interaction, so TBP is freed from the lobe and deployed into the DNA. This is important because now this frees the site that binds to TBP to TFIIB, which is the one that binds to the polymerase and gives rise to the assembly of the PIC. This is our proposal based on our data.

Eva Nogales: I really prepared a super-long presentation, so forgive me. This is the idea now so that you believe it, we made a movie. Let me go back one second. Let me give me a little bit of time. I describe these regions, which some of them is very mobile. These we put here. These are chromatin modification binding domains that are present at the end of long tethers, so we don't see the. We know that they're there. The structure of just that domain has been solved, so this is just a model just showing that where there are tethered with respect to the rest of the complex and the range of motion that you could expect. This is a core promoter, like the one that I show you. This is what is called the +1 nucleosome. It's a nucleosome that has several modifications that are recognized by these elements, and that is placed there with precision with chromatin remodeling complexes, that I can tell you about if you come to the second talk that I'm making today.

Eva Nogales: In any case, the idea is that thoseodification are recognized and they contribute to bringing TFIID to side where it can search for these sequences. If it recognizes, it binds them. It binds now lobe A to be in the position where TBP can start attaching the DNA, and with the help of TFIIA, which actually relieve this inhibitory interactions and ultimately be able to bind the DNA, coming off the lobe, recruiting TFIIB, which then recruit the polymerase, which comes right here, and it will be placed with respect to the transcription start site. It's the other story.

Eva Nogales: These are the people that work on what I show you today. Jie is my lab manager. He's also the person that grows huge amounts of HeLa cells and purify these complexes. Dan runs our cryos electron microscope. Patricia is a senior scientist in my lab that started the 2D story, that then was taken over by three sequential biophysics students, Mike, Robert, and Avi. I didn't tell you about how Yuan studied the whole assembly of the pre-initiation complex, the polymerase, the TFIIH. Basil and Kelly did the structure of TFIIH at high resolution.

Eva Nogales: I showed you before people working on microtubules. Greg and Gabe have now their own labs at Rockefeller and at Scripps, developed the methodology to be able to do microtubules at high resolution. I didn't tell you but microtubules are super tricky because they are pseudo-helical. Helical is great. Pseudo-helical is real difficult. Their methods were really improved by Rui, who now has his own lab at Washington University. The work on tau that I showed you were by these two postdocs that now have their labs at Cornell and the University of Cologne. Sorry it's just University of somewhere, Cologne in Germany. Nisreen, that was a graduate student in Ken Downing's lab and worked with them during this time. Stuart together with Liz and Nisreen, worked the drugs, on the anti-mitotic drugs. Thank you very much for your attention.


Loading Spinner