 Good afternoon everyone. So it's almost four. Let's get started and welcome to the Provost lecture series And this is our fifth one. We started last November and it has been going very strong and Here is just some background information for those of you who are here the first time and So today we're going to celebrate Professor of Scotland Before I do that again, I just want to thank everyone from the provost office they have been very helpful dealing with all the logistics and people from CPR and also The president the president's office and everybody else who helped to make this series possible and today's speaker is Dear colleague of Scotland and this is The abstract he shared with us to post online as you can see of is from Sweden and He moved to OIST In 2010 so he is one of the originals at OIST and Since then he has built a very successful Unit and he also has been the dean for the OIST graduate school since 2018 and OIST research expertise It's a cryo EM and electron tomography And so we're going to learn quite a bit today during his talk and so what I Learned about OIST was actually before I came to OIST that was April 2014 and The topic was related to space allocation and I'm sure many of you have personal experience how to fight for space or how to acquire space and I I decided to join OIST in August 2014 since I already had experience from other universities I decided I'm going to start the process ahead of schedule and making sure I get what was promised and so I I contacted at the time VP John Dickerson who is here and some of you probably still remember him and He told me yes So your lab space is not ready. It's lab 3. It won't be completed in 2015 But we set aside some space for you so you can go to the Basement level level A in lab 2 and wait for your space for a year Fine So I had some email back and forth and and talking to John Dickerson April 15 So many many emails and he said Amy nice try but not so fast We still need to accommodate other incoming faculty blah blah blah and and so the space you see between Elliott's an old space is not going begging and so so From that point on I know I realize of is a very important person and The spaces are they set aside for all and so this picture is actually shared with me By all fun and you are you a now I understand why You and John Dickerson you guys are very good friends And so what I have learned from all things often I so we're being our office Next to each other for the past eight years So I learned from all that yeah, just don't take everything too seriously be reasonable be a good colleague and do whatever you can to achieve what you want to achieve and So since then we actually have some really good collaborations with oaths unit. Here are some papers. So we learn certain techniques from his group how to culture bacteria and his former post now also helped us to perform cryo em and other Very important imaging techniques what else I have learned from off You need a reliable car Especially it needs to be good and fast and furious, but safety first so you can see throughout the years So off upgraded his cars And I also learned from of how to write the proposals in Japan So I remember after I submitted to my first proposal. I was very hopeful I would get my proposal funded and of course not so I I think that was June 2015 so I complained to off to next door and he told me this is all wrong I showed him my proposal. It's like too many details too scientific So you have to write the proposal using manga style and that was super helpful. I I adopted of proposal writing techniques and and so far so good I learned from off as well be a good mentor and Food especially cakes are very important. So oaths first PhD student phasal has been Very successful. So he's now an associate professor at Harvard Medical School and So oaths unit they they regularly have social events and making sure everybody's happy and And so this so these photos are Shared with me from your unit members So you can kind of see the the progression of your unit from 2010 2013 15 and 2017 So I want everybody pay attention to the pictures of oath Okay, so what I want to illustrate is that Oaf has never changed from so that was many years ago was Habu or somewhere and when when you're When your kids are still a lot younger and this is from 2015 This is from November 2022 and so I don't know. What's your anti-aging secret? maybe always wear hats and or Habu that or Okinawa suits you and Most importantly of told me you need to be excited about your research. Maybe that's what has kept you forever young and so oaths expertise its electron tomography and So it involves sample preparation. You have to use the microscopy and At the end so you have to perform reconstruction To create this a 3d visualization and So oath is going to go through all of these techniques in more details and But he shared this slide with me. I think you use the same wine when you introduced Ichiro So I'm sure this is holds very dearly to you it involves dance And so hopefully we're going to to see some demo at the end of your lecture So with that I'm going to hand this over to oath after your lecture, so we're going to present you some gifts and For everybody else. So after the lecture Outside so we'll serve snacks and coffee and also Please Be mindful to prevent COVID So with that I will hand the floor to off All right, do you all hear me Interesting question. Is there somebody in here who doesn't hear me? Thank you for this very nice introduction. I I think that It's a bit of a one of my nerves when it comes to my professional activities doing all these years and that is that I Actually think science is fun so You know, I think that's true for really a majority in here science is fun and so When I now took my vacation Which happened to be from some time ago Then I realized that I actually still want to be at the department So I'm in inside my office fixing up things and trying to finalize things except today So I realized that my calendar was soon filling up with all kinds of strange stuff But just today so already tomorrow. I'm free again sort of I Before I really start into this I would like to say that I I've as you know have had two functions here, so I've been the regular professor sort of and now I Transform from During the night of March 31 to April Fool's Day Into a professor emeritus. I don't still don't know if I change color or something like that But I will be a new kind of person on that date All right, so But I was also a dean for a long time now. It's approaching five years And with the transition time it's actually about five years already and I would say that because I'm not going to talk about the Dean ship here. It would be about my research The Graduate School here is composed of quite a lot of Sections five of them and also a lot of very very good people So I would like to say that Although I'm not talking about that here in my time. I would like to Say to you that I really appreciated my time as a dean for the grad school It's been very rewarding to me and I've learned a lot about human nature Which is really quite a lot of fun All right. Now. Let's me get into this so I Call this a quest to enable viewing individual proteins at 10 angstrom resolution now Why do I say it like that? Well, because I was trained as a protein crystallographer And that was my PhD and so I did Some structures there and then as postdoc. Well it had to be Protein crystallography again. So I did virus crystals virus structures And then I realized that what really intrigued me with the viruses was the inside of the virus I'm not saying that any of these things is Better than anything else. It's just what I was intrigued by and it turned out that there was no method to look at individual not repeating parts So I decided to Actually develop tomography. I had just spent a couple of weeks with Aaron Kluge's lab in MRC in Cambridge and they said well, it doesn't work You need to have repeating things But if you can really solve the equations about alignment and everything, it's okay. Maybe you can try So of course I did it Because as soon as somebody said to me that you can't do it I will I'm always wondering if that's based on, you know, some mathematical proof Or they just think you can't do it In this case, I was pretty sure that they just hadn't put their mind to it in Cambridge They wanted to reach high-resolution very quickly and went the way of this one was symmetrizing molecules and Henderson and his colleagues they went with instead using crystals 2d crystals So I wanted to start and in the beginning I couldn't get better than about 8 8 and a half nanometer in 3d, but it was a very large particle So I wanted to reach a level where you can uniquely Define structural elements and that's about one nanometer give or take a little bit So everybody there in Sweden in particular said no you can't do it So step-by-step I've been increasing the resolution all the time and as you will see towards the end here We are actually there So what is tomography? It's the same technique you use in older times of radar and also in human reconstructions with CT scanners in hospitals and Basically, and let's see if this works. Can you see this red dot? Somewhere in the middle Okay, so you have an object and you project with the beam through it and you record pictures like this And if you just record one slice Like a line on each photograph You can fold that back with back projection mathematics and you have one slice in there so if you proceed through with all the Slices you can have here you reconstruct the object like a sliced up piece of bread, right? And then of course after that you can turn it around in all kinds of directions. You have a 3d Volume now this is done in all of these it's very common with Tomographic principles today, so you have it in as I said radar older types the newer radar is They have introduced a bit more steps, but basically theoretically backwards. It's it's tomography Etc and tomography can be in several dimensions so you can go from lines to 2d or you can go from pictures to 3d, etc all right, so tomography in in In a way when you look at it and you go over to the four-year space instead There are people here who don't know what that is, but it's like diffraction space And or the focal plane in a lens set up. That's where you have a four-year space. Well In in that term Crystallography is a special case of tomography Because tomography doesn't require that you have any symmetry or whatever. It's like for crystallographers P1 Symmetry isn't in there unless you're imposing it somehow So if you have a bit of a molecule in there, you see the molecule and everything else in there in your in your object the problem is in When you record the data so in electron microscopes you want to record all these different directions and like super stereo somehow But limitations with the specimen holder, etc Makes it very very difficult to record 360 degrees There are lots of lamps who have tried this out and it's so far not successful to high resolution and One of them told me that When they make a holder so that they can start rotating the specimen the energy dissipation in the materials makes it start wobbling and That's very difficult, but maybe in the future People will be able to create specimen holders so that you can actually Rotate the specimen all the way around it's Missing some data Means that you have a deficiency in the 3d reconstruction so you have lower resolution in certain directions than others But it's not as bad as it sounds and you will see later on that in Our 3d reconstructions we hardly even see it nowadays the effect of it So what do you do? well Let's take a normal morning You go to the lab and you prepare the specimen You put the specimen in the electron microscope and you automatically collect all the data and And then you transfer the data over to Your computer in our cases usually I max but it works even on machines like this one and I back Pro and After some calculations later the same day You can then view your structure in 3d. This is an antibody That you see over here So the process in this particular case zoom over and it's like a few hours But the specimen if you really want to not miss any details One of the best ways is to use frozen frozen hydrate it so it means that the specimen is The electron microscope grid with the specimen on here you don't see it here But it's a small small drop a couple of microliters or so Containing your specimen and use have it so much smeared out as possible over the surface and then You shoot it down into liquid ethane that is cooled by liquid nitrogen Okay, and this freezing Process here is so fast That you don't form any ice here, you know ice is very bad because it is bigger in volume So if you have cells or whatever in there they burst if you do it You might know that if you have strawberries and you accidentally freeze them in the fridge Right, you take them out and saw them and they are just saggy blobs, right? That's what's happening the ice destroys the cells, but with this procedure you can actually Preserve the specimen because it's so fast and it's even faster than this looks like because It's a transition temperature at minus 44 to 45 So the very very small temperature interval There is a phase transition and you get the solid material There are some tricks one of the tricks is to When you do the plunging into liquid ethane it has to be controlled because if you do things in Here you see Evaporation velocity in nanometers per second. Well your specimen is hopefully something like Below 50 nanometers Right and if you see here that You evaporate like 50 Nanometer per second. Yep Then you don't want to do it in dry climate So that's why people use special humidity chambers When they prepare the grid and then they shoot it down in a fraction of a second in so that they don't lose so much Water if you lose water in your Very thin layer the concentration changes tremendously of different components, right? So it's very important to have a Control on this one. We call it molecular electron tomography and it removes The step that comes after this removes shot noise. I'll talk about that but what it is is you have your Specimen on a grid that you photograph and there are some black dots here Which are gold markers that you can align everything on down to two and a half angstrom or something like that and Then you can you can take the tilt series by tilting and Then you can reconstruct and then this is where your headache starts because when you look at this one If you still have a specimen to look at it means you haven't irradiated much because it's sensitive to electron dose so The what you have down here is a molecule which is reasonably well to look at if you choose a low resolution like Five nanometer worse normally you can in some cases go to higher But you get a lot of noise in there and that's because the You use such a low radiation that you see the Fluttering of your detector in terms of Intensity, I mean you know how it is to go in a room like this If we lower the temp the light here more and more and more and more finally You will just see gray things in here. You won't see color and it will look grainy But grainy stuff is what you have in here along with your reconstruction. However You can add a step and that's what we have done This added step means that you start with this number four here and then you you actually refine First of all, you check up whether the 3d reconstruction fits your data or not If it doesn't fit the data because it could be scaling error or kinds of things numerically Then you can fix it but a very important thing is that Because you know the variance, you know, it's a detector. It was on this beauty there, etc. So you can actually say Whether you have too much information in the system or not So what you do is you mix these two states Mathematically means you have cross derivatives and all kinds of nasty stuff in there And you can remove everything that has too much information. It sounds kind of crazy, right? It's just that The maximum information in any picture you have is when two pixels never correlate That means you need to describe every pixel. That's a lot of information so that means Because noise is of that order, it will always have more information In the noise than in the actual picture So you remove noise and then you affect this and you cycle this and Voila, you get down and I will show you later. We get down to one nanometer now So here's an example a long time ago of an antibody Rotating and going through 10 cycles of refinement and you can see how efficient the noise is removed Through the cycles so this antibody is actually Here and up in that corner. You can see it's actually binding To Here it binds to this antigen Alright Here's another example. I did this is also a long time ago or the look of corticoid receptor and I realized these are two Supposed to be identical molecules in the same purified sample Now when you do the 3d reconstruction You can realize that the different domains is the structure Like you see particularly in this one in Relationship to the bigger bulk at the end they point in different directions Which means that there is internal dynamics in in the structure, which makes It very difficult for anybody trying to for instance crystallize this or something So nobody has been able to crystallize the entire group of corticoid receptor But they have crystallized this piece and the middle piece but no longer not yet the back part here Because it seems to vary also But tomography where you don't average naturally It doesn't matter You see what you see in there. So we pushed all this better and better and better as we were here And this is an example of catalase when I really started to doubt myself Because as you can see here is a crystallographic structure Of catalase and you can count one two three four five different helices Here is an example of one of our our reconstructions and you can count the same helices one two three four five This one looks kind of funny in this area and this is another one too and To my horror I realized that Even if you get to high resolution on these individual ones The dose shoots away randomly in in these particles. So the Curation of all this is what single particle people had already figured out a long time ago. You average All right, if you average many of them the error where The electrons have shot away some part of the structure Gets averaged out So I was thinking how do we get to that because the point the point of tomography is like you can see in a solution All the different molecules and what do you do with them? so We went went to a different Way to do it. So we started this another real project with synaptic tagmin and Synaptic tagmin is a molecule that Here depicted in two more or less identical images here so the synaptic tagmin has One peptide and another one C to a C to B B is larger than a and then there's a long tail that goes to a vesicle and here it's Drawn in a different way into the vesicle and the point is that These two Synaptic tagmin domains, they are dependent on calcium so when there is calcium present somehow they get to be very close to each other and They actually bind to a membrane down here Plasma membrane and it's thought that that helps to pull This vesicle down here. So first of all this vesicle is a synaptic vesicle. So that's where Your brain is keeping your nerve signals. It's pretty good if you can make use of it so you can think Okay. Well down here is the synapse border. So this action is thought to With calcium to bring these two together in some way so that they can bring the vesicles down and then there can be fusion With this membrane here and the fusion down to this one and the contents can be released It's a complicated procedure. So down here. I don't think one knows exactly what's going on but it's been an endless amount of Hypothesis around what synaptic tagmin does here All right, so we've been interested in finding out What does synaptic tagmin look like actually in there? It's it's not very big. So We were not sure first whether we could see it To make a short story even shorter. This is a correct caricature of it. So you have the synaptic vesicle the two Calcium containing or could contain calcium peptides of the synaptic tagmin, which is the sum of these two and then Binding to the synaptic cleft This One has to understand the dynamics in this process to be able to figure out How to moderate this this is a nerve signaling thing. So if you make a drug in here, for instance It's probably not good if it's on or off, but maybe 35% off or 67% off something like that to moderate it a bit. Otherwise, maybe there are very big effects on your thinking So it has been a lot of speculations Circulating in the press About these two domains how they are and how they can be related to each other in different geometric configurations all trying to explain how When there is a lot of calcium, for instance, they can bind and create Possibility for the synaptic vesicle to release the signal all right, so it turns out that the Crystallographic procedures have been used to solve the two when there is a lot of calcium and Another version where there is little calcium, but these are only two two different conformations and The question is which work and which doesn't work or how much in between can it be to work That you can't see with with crystallography. They've been able to pinpoint two versions here Not yet though with SPA because it's not big enough So here is an example of how big it is. So the two parts in it Are such that this the smaller protein is just 15 kilodolton and the other one is 17 and a half kilodolton so all together not a very large molecule and at least of This date there. I don't think there has been any attempt with cryo-em and single particle On this for the reason of the size of course The molecule is also rather small You can see 47 angstrom Roughly in size etc. So it it's a rather small thing, but the major issue with Single particle is the alignment of different parts and when it's small molecules almost impossible to align Now we are not dependent on that in tomography because we have our goal markers So we can actually continue and seeing all those small things so here is From the PDB database the 5 cch Showing the open configuration. So no calcium and here is another version that I Actually put up, but it's a small picture looks almost like a butterfly when they are close together So this is with calcium now We then designed a new way to do this procedure so We have our initial Reconstruction and then we do something called mining. So what just one tilt series and you reconstruct it and this Constrained refinement Comet is super efficient to remove noise So you can actually go into the reconstruction area over the entire specimen that you did and and ask for a molecule With the size of 35 kilonewton plus minus a few percent And then in this particular case it fished out a few hundred of them But they are all in different conformations However, when you analyze all of them You find out that they form clusters So there are a number of ways that the molecules like to be in in there And of course if it's a cluster Then even I could say well it make me make sense to average within the cluster So we proceeded that way so We had all these different structures that we clustered and then you can make Averages and fit them to the x-ray structure and then you can discuss the model So here's an example So in this case We had 771 particles of about 35 kilodolton in one experiment one one tilt series and then in this case Cluster seven contain 105 particles quite a number of them and it looks like this and with Fouichel correlation you can see that it's with the zero one Four three criteria and it's ten point zero angst from resolution on that one Another one Had slightly worse this one is eleven point eight angster resolution and the two Peptides are much further apart from each other in this case so one can actually go on with this and In this particular Picture you see X-ray experimental research that's the sort of greenish yellow here and the other one is computation of prediction because People have come up with this alpha fold version where you can predict the structure and here one can see that Alpha fold Is kind of Good but not in all aspects so alpha fold can be used yes But you have to know how to reinterpret it to make sense out of it So here are more Alpha fold predictions this one is for The monomer domain C2a and C2b and then you start to make a dimer and a tetramer and This is with calcium. So they are close together and on this one you talk about the same Prediction again by having them further apart. Well, we have looked at those and compared with with What we see from tomography and then one can see that maybe The prediction is not so good when the structure is complicated So it needs a lot more work Simply for for that to be trustworthy yet And if you take the full length Synaptic tag mean, you know, it has a Tail that goes all the way up to be anchored in the synaptic vesicle and then it gets even worse I think in terms of how much you can believe of the structure But you we can use comment to predict protein structures and here comes a twist on this So comment can be used to Predict actual conformations of complex structures and conformational changes of the proteins while interacting with other proteins so with The artificial intelligence at least at this point it can't really predict complex structures and conformational changes but Geometrically constrained molecular dynamics can be used to fit such structures to our structures And there is a researcher from Toronto who works in Riken right now. Adnan Shloka who has developed this so If we compare here is a probe number one we use one of our structures in one of our Complexes that we we find a cluster and we search among all the generated PDB maps that Adnan Shloka can predict through his constrained molecular dynamics and we find a fit and It corresponds to if you remove the density corresponds to exactly this kind of distribution and You can actually look at another probe It fished out another of their predictions and it's this version And finally a third one Here is our structure and here is the probe it here's the probe and here is the Molecular dynamics generated structure and this corresponds to the atomic distribution so It means Yes at this point one can have in this case we had only About 1200 structures that Adnan had predicted but it can be a hundred thousand or more and you can use Any of these cluster structures to as probes and search through it's pretty quick actually and then the interesting part is that you fish out all those possible variations of the molecular dynamics that actually works and Then you can discard all the other ones and and predict which Molecular dynamics trees you should pursue or further more even more accurate fitting So it's an interesting Twist to this I think where tomography suddenly becomes possible to in some ways to interpret to high-resolution So what we're aiming after is to understand how to go about and find out how the molecules dance in there because when drug Companies generate a new drug They might not want to completely stop the action or Twist somebody on to a hundred percent. They might want, you know a small dance More reduced answer. That was the dance. You were after right? so I think that When I first did this thing Analystically on antibodies we modeled the whole procedure. It was so complicated. I realized We won't be able to do it on on the very complicated Molecular changes, but in this case We can figure out Directly from a lecture dynamics What has happened to the molecule and which ones fit fit to what cluster? now The clusters we've been working with is from one till series, but right now we're expanding it to work on Pool data so you can take 25 till series. Let's say with one chemistry upset subset and You reconstruct and you see what clusters you have in there Right, and then you can change a little bit on the pH or the salt concentration or whatever it is and So on and if you want to know what happens if you put it in the body, which is a very interesting thing, right? You can take a biopsy and then you can add your molecules to this biopsy and Reconstruct what kind of clusters you actually have in the biopsy without doing the animal experiment All right, so Yeah, I think we actually do reach One nanometer. What's stopping us actually? Well, it turns out that Rina who has been taking most of our photographs Is limited by you know, how accurate can you align the height adjustment in the microscope and you record the data? It's very critical and she's become a master of that. So that's why we reach down So far our record for your shell correlation structure is 9.3 angstrom, but If we were able to figure out the focus we would be able to go Way further down and there is this new technology that's coming up Which is called integrated Something focus IDPC stem is a new stem technique Which is not fully developed yet, but it works for Single particle analysis at least to 3.6 angstrom, but in that case when they make it work for tomography There is no focus issue So then we can use the same technique to get to much higher resolution that we have it today I would like to Say that there are many contributors to this work. So Gunnar Wilken. He is sitting here there He has been important for us to sober up our Way of using some of the mathematics in comet to a very important step where it's very accurate And can be used to higher resolution I did a test on on A nano particle silver particle actually a nano particle And there you see the lattice. So you have the silver atom distance and see every silver atom in some area So it's rather accurate on this kind of stuff Well, last year and and I we have been camping together since 1992 And you might remember him. He was here for many years ago to retire two years ago two and a half years ago We continue to work with him on software development and he and I will continue doing software development as I'm back as a emeritus but curiously enough so in 1992 not 1982, sorry, I Arranged the course together with Richard Henderson at the MRC in Cambridge and the course was running in Uppsala and the best student at that course was Los Yaran So that's why he ended up finally in my group Ritesh who is also here there He has been doing tomography earlier with me, but The late thing is a single particle Structure that we're struggling with trying to get published right now And then we have our master for data collection Rina Who is now? Continuing here at OIST in the support section imaging section and we also have this ongoing collaboration with Adnan Schlokka who is Developing this very exciting new version of molecular dynamics Where you don't have to calculate things that don't have anything to do with the real world So it's not completely random as the way it's been developed so far. I Did develop a lot of this stuff as I was at Karolinska Institute But lately It's been here at OIST So I have got a lot of numerous generous grants in Sweden and also some very very large grants here in Japan So I'm very happy I've been privileged. I think and I'm also very happy that my collaborators over the years have new positions Continue torment everybody else Well, that's about it I'm happy to have questions if you have any