 And I went to thank the organizer of this meeting for having invited me and for what is turning out to be a very interesting meeting where the discussions are quite broad and I'm quite fascinated by being able to cover different topics. So I have always been interested in looking at things. I have to tell you that I don't see in three dimensions. My world is two-dimensional. So I always have this thing that I want to see things in 3D. So I started to do not yet, not yet. I started to do crystallography, cryo-EM, and now we're doing lifestyle imaging. And so what I'm going to do today is I want to show you two problems that we're addressing. One is very mechanistic, inside the cells. I'll show you how we can use fancy microscopy to track teeny objects and how they change composition. So this will be focused in lipids. So that will be the first part. And then I'm going to switch totally gears to a different level of organization. I'm going to go to the ability now that we have to image things in three dimensions close to single-molecular sensitivity with very, very high resolution. Now, this is a revolution. For me, it has changed completely my lifestyle. I must confess I have some problems sleeping. And the person that is responsible for doing that is Eric Betzig. So we're good friends and met him a few years ago. And you might recognize he was one of the people that got recognition with the Nobel Prize for single-molecular super-resolution imaging. And sort of when he was getting that recognition, already was developing another form of microscopy. It's an acronym which is very long. It's Lattice Lightsheet Microscope. The details don't matter too much right now. But it's a way in which you can see non-invasively inside cells with exquisite detail. So everything was fine. We built an instrument. We got it in the lab. And we've been doing this for several years. And then suddenly he came with another idea which I'll show you, which is to bring the technology that the astronomers use to sharpen the images when they're using the telescope. When you do that, now it turns out that you can go into animals, for example, or organisms. You will see also from David's talk later. And you can go now into the multicellular space and look at what's going on inside cells. So I think what I wanted to convey sort of a philosophical level is that at least I'm switching from a mode where I was doing hypothesis testing for those of us that are in the United States. That's what NIH always wants. What's your hypothesis, right? Papers also, the journals, tend to also ask you, what's your model, what's your hypothesis? To another mode which I'm calling hypothesis driving. So it's like renaissance to me. We just look, describe, and then let our brains go. So I'm going to give you these two stories. One of them, I need to turn off the light, sorry. So one of them has to do with the issue of phosphorinosidites. So these are specialized lipids that we have inside the cells. They have charge. And they tend to be platforms on which proteins can and they signal things. So I'm going to show you an experiment that was during my lab by Kang Min-Hae. He's a postdoc in the lab with help from Bobby Marsland. He was a PhD student, so he did the theoretical part, which I won't have a chance to present today. I'm sorry for that. And so this has to do with these lipid conversions going on inside the cells. So I'm going to focus on the endocytic machinery. I should add that through my career, I've always been interested in how cells eat and how cells spit. So when cells eat, there are many ways they do that. And the one I like is the classroom, what we call the classroom mediated route of entry. I'll come back to that in a minute. So I'm going to show you that he developed a special type of lipid sensors that with exquisite sensitivity can detect the lipid composition inside cells. And then the second part will be this problem of how you can now image in multicellular context with very high sensitivity. So this is a work that has been driven by two people. In Eric's lab is by Sung Lee, who developed these adaptive optics, as we call it, a lattice-like microscope. And Gokul is a fellow in my lab who is driving and spreading all the efforts that we're doing, both in terms of image analysis, building the systems, et cetera. OK, so I'm going to then start with dynamics of phosphonide conversion. Then I'll go to the imaging in tissues. And I would like to point out that this we can do with very high temporal resolution. We can go to 20 milliseconds per plane, actually per plane of imaging. We can even do faster than that. In some cases, we're going to do a few milliseconds. And the precision of our measurements can be about 30 nanometers. So that's the location where objects are. Of course, the resolution is not as high. It's more in the 300 to 400 nanometers range. So let me start with this movie. This is what I call a molecular animation. A molecular movie describes how clustering works. It summarized studies from a number of groups, including us, where we went through the crystal structures, binding constants, cloning, et cetera. And what is depicted here is how this machinery is forming. And it's what we understand with some level of certainty. So the time this movie is running is what it takes a cell to make what's called a clustering called a pit, which is a state where we are. The membrane is evaporating. And there are hundreds of proteins that are coming in place, a lot of very weak interactions, all in the high micro molar level. But then there's a neck from the vesicle. This is the protein dynamic. That's a source of energy that comes here. That's how we break the neck. So it's working to do the tightening of the neck until now you get efficient. And then the last part of the reaction is how you get rid of this glypholic coat. So that is done with an enzyme, which is another source of energy, it's an ATPase, which is a 70. So that's an ATPase. And there's a cofactor that brings the HSC 70 to the site where the reaction is. So you will see in a minute why I'm pointing this out. But what you see here is all the result from our cryo-en crystal structures that we have performed through the years. OK, so what you saw here takes about 35 to 50 seconds in average. We have hundreds of these guys doing this in every cell. So every eukaryotic cell has it. And we have a similar process that goes from yeast to mava. We will hear more from, I think, from David later on. OK, now there's a puzzle here. What happens is that this machinery that was encoding this does it only after this structure has pinched off from the membrane. So even though the components can recognize the code components, they are not recruited to the lattice that it's forming. It's only after they pinch that this gets recruited. That's a puzzle, what's the signal for this? So it turns out that the signal is lipids. So I'm going to show you that when you are making a coated pit, so it's this code that is forming here and is deforming. And that's the dynamic in this location. And then you pinch, and then the encoding machinery comes and destroys. There's the enzyme HSS70, which is being recruited. And then coding, and this oxaline, which is the cofactor. So it turns out that this cofactor, which is structurally depicted here in the bottom, has in the C-terminus, what we call is the J domain, which is going to recognize the HSS70, has a binding site for clathrin. And then the amino terminus has this thing that we call the P10-like domain. P10 is a fault that is a tumor suppressor, and it's a phosphoenositite enzyme. So it's actually removing phosphoenositites. Now here, this has sequence similarity. The structure is quite similar to the P10. But the catalytic pocket doesn't exist. It has been mutated. So what I'm going to show you is that this portion here actually binds lipids. It likes to recognize certain classes of phosphoenositites. I'm going to show you that what is being recognized is PI3P and PI4P. It also recognizes 34P2. It's a special head group phosphorylation we get. The plasma membrane is very rich in 45P2. And I'm going to try to show you how we think it happens that we can make the conversion of lipid that we have the PI45P2 here into getting the 3 and the 4 later on as we pinch. So this is the biology. So what we have here is a cell that's been placed in the microscope. And we're looking at the bottom surface of the cell. So what is in red is the clathrin, which was this trisculium that has a fluorophore in red. And each one of those spots corresponds to each one of those structures that are forming in the bottom of the cell. Later, I'm going to show you an image throughout the cell as well. The image to the right is this oxylin. We added a GFP. And you can see that it's blinking. So it turns out that those blinks are happening at the end of the code when it's falling apart. So for example, here you have fluorescence intensity. And I've just clustered a number of these traces by cohorts. So just focus, let's say, in the very last one, which I'm going to trace with my pen. So the fluorescence intensity in red, which is the clathrin, is going up all the way to the end. So this is the assembly process. And then it catastrophically comes down. And that coincides with bringing up this cofactor on the cell as a pulse. This happens after the vesicle has pinched. So that's sort of the biology of the process. So what we have found on taking this protein and following how it gets recruited. And then with a trick I'm going to show in a minute where we can replace with the natural P10 with other domains of other proteins known to bind specifically to lipids. What's going on is that the plasma membrane has PI4P2, which is sort of depicted here. There's also a little bit of PI4P trace amounts. But then when we pinch, which is at this point here, there's a pulse of PI4P while we're consuming for 5P2. We have a pulse also of PI3P. And then PI34 appears and stays and then goes away. Now, for those of you that know the endocytic pathway, this vesicle here, after it uncode, is going to traffic to the endosomal compartment. Unfortunately, I won't have time to show you now. But we have done the same class of experiments as I show you with this oxylin following RAP5. And it turns out that RAP5 is only recruited after you encoded. And it's actually part of the recognition of this 34. So I'm going to show you that not only you get these waves of lipids, but these are also the platforms at which things come in. So oxylin likes PI4 and PI3. And that's the signal that brings the encoding machinery when it uncodes. Then we're going to get a wave of 34. And that is what's bringing the RAP5. I'll show you also that we have the enzymes that are responsible for modifying these lipids. So I won't have time to go through the whole math, but we have done simulated all this. So we had all the results and the timings and changing and putting mutations, et cetera. And then what we have done with Bobby is put all the binding constants, all the rates of all the enzymes. And it turns out that not only you can model the process, but you can also do predictions in the system, which I think solidifies what we had. OK, so then the idea is that we have our clustering code. This is actually this is one of the first. I think this was the largest cryo-EM. The largest structure solved by cryo-EM that was done about, I think it's now 14 years ago. This structure is 22 megadaltons. At the time, it took us about three weeks of CPU time to solve the structure. Today, you can do it in a few hours, just a little bit of context. This model here in blue is the clustering. In red is the component of oxyline that is binding to the code. And in green is the HSC 70. And the resolution we have here is roughly somewhere between 8 and 10 angstroms. OK, so the idea for our sensor is that we can replace now the p10 domain with any lipid binding domain that we can get hold from the literature that people know, their specificities. We put the EGF here at the minoterminus. And this guy is recruited exactly like the natural protein. It will give you a spike if we get the correct lipid binder or it won't recruit, et cetera. I should add that these changes don't interfere with the biology. So then the sitting machinery is fine. And the movies that I'm showing you is all single-molecal sensitivity. So we don't need to overflow or overload the system anymore. So these are examples of some of the domains that we have replaced. So this is the p10-like for which we solved the crystal structure a few years ago. And we have other domains from the literature that we took, pH domain, PX, ENTH SSR, with known lipid binding specificities. So just simply you swap. And then you end up having this sort of coincidence detector where it gets recruited now in different fashion. So these are now chymographs. So we're just looking at the bottom traces. So this is time in this axis. And then we have the signal for clustering in red and in green will be this auxiliary of the replacements. So these are the spikes that we have. It will put PI3P, the same with PI4P. And with PI4P, you get a spike that then lives longer. If I were giving you the signal for PI4P2, the signal would have come together with the red and then disappear. So that's a consumption. So the summary of this sort of results that we have is that we have a process where we get the signal of the PI4P2 until we pinch and then it gets consumed. And then again, we get a spike on PI4P, the same for 3. And the 3, 4, 5, the 3, 4, P2 goes up and then decays much slower. So I won't be able today to go through every single experiment we have done to figure out which enzymes were doing that. So I've just listed them here in the next part here. So there's a program, for example, called Synaptogenin. Synaptogenin binds to the clustering code. This is work that was done in Pietro de Camille's lab primarily. So Synaptogenin is recruited. The moment you have a clustering code, it comes on place. Synaptogenin is consuming the 4, 5, P2. It's a 4, 5, P2 kindness. But then what happens is that even though you have the enzyme here and you're consuming the substrate, it's exchanging with the adjacent lipids. So they're diffusing very fast within milliseconds. So even though the enzyme is consuming, you replace very quickly. And the same is true for some of the other enzymes that we have here. So there's another one here called OCRL, another enzyme that also the Camille has been working. Also, Luigi in Italy. So this is also a 4, 5, P2 kindness. It's also recruited to the clustering code, but it's recruited actually at this stage, a little later, when the code is already falling apart. So those two are required to consume the 4, 5, P2. If you get rid of one or the other, nothing happens. You have to get rid of both. Even worse, if you get rid of one of them, the other enzyme shoots up in concentration. For those of you that think about signaling inside the cells, I found that extremely fascinating and weird at the same time because the regulation of the OCRL or the synaptogen, depending which order you do the experiments at the transcriptional level. So the somehow is sensing lipid composition in the coded pits. And that somehow gets transferred all the way to transcription to regulate. I must confess I have no idea how that works. So I have a few other enzymes. And it really doesn't matter. I won't go into that detail right now, but we have other enzymes that are also modulating the other lipids. And here at the bottom, I just mentioned two oxalins. There's one called oxalin 1, and the other one is oxalin 2. So oxalin 1 is the prevalent form of this cofactor in the brain, although you have it in other tissues as well. And oxalin 2 is present in basically every tissue in your body. So what we found is that oxalin 1 likes to recognize PI3P. And oxalin 1 works in the plasma membrane. Oxalin 2 recognizes preferentially PI4P and also 3.4P2. It also goes to the plasma membrane, so that part is good. But then when you look in 3D, where else do you have this oxalin working? It turns out that there are coated vesicles in the secretory pathway, and those vesicles contain PI4P. They don't contain PI3P. So sure enough, this oxalin 2 goes to those coated vesicles that are inside the cell. So the signal that is sending the enzyme to the right place is dependent on the lipid composition of the vesicles. And finally, I also mentioned that RAP5 is being recruited to the codes after you encode it. And that requires having the PI3P4P2 lipid. So there's also a protein that engages and then brings it in, and the GIF actually required for the RAP. So with this, what I'm trying to point out is that we have a system which is very complex. Things are happening very, very fast. It's all in a matter of seconds. There's a vast amount of traffic. There are all these signals going in. All the interactions here are all in the micromolar level. So if you go with one protein at a time, given the cytosolic concentrations, nothing should work. And it's working, I think, because you tend to have, usually, in this protein, there are two binding sites. Let's say one for clathrin, one for a lipid. Each one is micromolar, and you end up having these coincidence detectors. OK, and so this is just a summary of the lipid situation that we have. So the plasma membrane, I just pointed out, we have a 45P2 and 4P. The endosomes like to have PI3P. And then in this mult vesicular bodies, which are sort of deeper forms of endosominal structures of all these lipids depicted here, and also in the secretary pathway, as I just pointed out, we have PI4P. And so what we found is that there's this transient modifications that happen along the cardiovascular. So imagine this is the cardiovascular pit, and you're making the cardiovascular. I'm just highlighting here the various switches that we're having in lipids and the wraps that are being recruited. So I think that something similar can happen somewhere else in the cell. We haven't studied this in detail. But I imagine that the same will happen, or something similar could happen in the secretary pathway, where you're also having lots of signals and sort of diverted traffic going on. OK, so I now want to change gears to the second part of the talk, which is imaging things inside the cell. So it's an expansion of this. But what I'm going to now change is the way I'm going to present you the data. So I won't show you enzymes anymore, and there are going to be lots of movies. You will see some movies where we can follow the endocytic system. I'll show you that you can follow, for example, the endocytic traffic, not only at the level of a single cell, but also multi-cellar organisms. And I'll also show you other type of biologies that, as I said, at least for me, is keeping me awake. OK, so we're used to use microscopes where if we have an objective, you shoot a beam throughout the cell, and the beam is like the pointer. So it's basically a Gaussian-shaped beam. It illuminates the whole cell. And then what we're doing is imaging in a focal plane. So that's how a standard microscope with its white field or a confocal scope is operating. There is another way of imaging, and it started with experiment some years ago, which involved using sheets of light. So you can have still a Gaussian sheet coming from an objective, and it's coming through an angle, through the sample. Then you can sweep the beam, and then orthogonal to the plane that you're forming. You can image with a second objective. So this form of imaging is light sheet microscopy. It's very powerful to look at the organization of cells, and for example, developing embryos. The problem of this, however, is that this beam here is quite white. It can be cell microns in diameter. So it kills your ability to have high resolution and also similar to the standard form of microscopy, because this beam is basically illuminating the whole cell. You end up getting lots of bleaching. And it's very invasive to the cells. So Eric Betzig then invented another way of doing the microscopy, where he can have a beam that becomes much thinner, and that's an instrument that he calls the lattice light sheet microscope. The trick is very simple. So under normal conditions, you will have an objective. And if you illuminate the back aperture and objective, you get your Gaussian beam. You're basically taking a Fourier transform of the beam. If you illuminate with a ring, what you get is a vessel coming out, a vessel function. So now it's a very complicated beam with lots of arches. Now if you create a lattice, an array of those beams, and you can organize them so that they're all coplanar, it's still a very complicated beam structure. But if you bring all those beams together, since you're using a laser to do this illumination, you'll get interference. And there's a sweet spot where everything becomes very tight. And you effectively get now this sheet of light. You still have to deal a little bit the beam with a Galois mirror very fast. But you end up having now a beam where this thickness here is about 400 nanometers in thickness. And so that's close to doing turf microscopy, total internal reflection microscopy, but in 3D. Because now we have this objective that is looking orthogonal. OK, so that's how the instrument works. So basically you have the lattice. You dig there, and then if you have a cell that is expressing fluorescent markers, in this case, a marker just at the cell surface, you can just move the sample against the. It's still optical illusion, yeah? Excuse me? It's still optical illusion. Yeah, it's always, yeah, yeah, yeah. So although we can do better than that, I mean, if you're interested, we can discuss that later. We can structure also the light to go break the barrier of the fraction limit. So you can scan your sample against the beam, and you can do that either by moving the sample with respect to the beam, or you can move the beam and the objective and have a stationary sample. It does details really that doesn't matter. But you can collect data, and the fastest we have done is about 3, 5 milliseconds per plane. Normally we do 20 milliseconds per plane. And that's so that we don't bleach too fast our fluorophores, usually GFP or similar molecules, right? And then the amount of light that you're putting in the cells is low enough that it's not damaging. And you get, for example, cell division, things like this. It's very hard to do an experiment when a cell divides using a normal microscope because immediately you get DNA damage, and the cells just freeze. All this is gone now with this form of microscopy. So all this was done in Eric's lab. OK, so I just brought one example just to give you a little bit of context. So this is now from our lab. This is a cell. And we have a surface marker in this cell. And this particular cell has the spikes that you see here. So these are philipodia. So I must say that before this movie, I always have seen the philipodia, these projections, sort of parallel to the glass. So with the normal microscopes, it's very hard to see these guys sticking out. With this now form of microscopy, now you do not get a good sense of the 3D. So this particular movie was such that we're taking a spacing of 0.3 microns per plane. There are 150 planes in each stack. And for this particular experiment, it took us three seconds to collect the stack. There are 15,000 images. And this is a seven-minute movie. No bleaching that we pick up. No clue. Is that good? I should know. I don't know. It doesn't really matter this whole time. Who cares? No, no. The reason we were looking at this is because Emma, when I was looking at the experiment, he was interested in exosomes. So he chose a cell that was making exosomes. I just don't remember the name. And this part in here is supposedly enriched in endosomes CD9. It's a tetraspanin. Malachosate is a transmembrane protein. Well, it turns out that that's how we found that it was full of polyporeal. We're not expecting that. It was just an accident. So this is another cell. In this case, this is a cell that's being gene-edited with this protein called AP2. That's a protein that links clathrin to the receptors. And so this has been gene-edited. This follows principles that David brought in some years ago to convince the community that we should be using gene-editing instead of transfections to then follow fluorescent proteins, ideas that you can express this at endogenous levels. So this cell is expressing this. And this AP2. So each spot is one of those clathrin coats. And this happened to be 50 planes, I think, every two seconds. Now, the reason I'm showing this movie is because this is a cell that we just drop on glass. And you can see that this cell dropped, and now it's spreading out. So what you see here is a volumetric imaging of the whole cell. And you see that there's material that is flowing backwards. This is the dorsal surface of the cell. So it's a bit contrary to it. But what is depicted here, you have the cell that is expanding. So this is a lamellopodia and lamella. So the membrane is expanding to the left. And then we have a backflow. It's a centripetal flow that we're having in the top. So that's counterintuitive. We can discuss later the science of this. But this ended up being the motivation for this poster thing that we have for this. There's now a markup. That mark took mark actually made this. This is actually a mistake. So here what is encoded is time in colors. So what is blue is things that are quite old in the image. And these are different frames, each frame with different colors. And you see the fresh guys are from outside and going in. So that's the poster. Yeah, Mark, thank you. OK. So let me compare you now to spinning disk with a lattice like she might have. Some of us use spinning disk. I was very proud of the spinning disk because I think we got the third spinning disk in the United States. I think that was in 1999. So we used it heavily. And I thought it was a cool instrument until I'm going to show you the movie What Happened. And now I put my spinning disk in eBay. So this is, again, the gene-edited cell. It's going to be for AP2. And I'm going to show you what happens when you collect these 50 planes as a function of time. So we're taking stacks as a function of time. And the cell in the bottom is a similar cell. These are some cells, actually. This is a form of cancer cells. And the bottom will be the same thing. And the way we organize the experiment is that the amount of light that is coming to the two cells is equivalent so that we can get the same signal to noise. So you're going to see the two movies running together. And here in the right, you're going to see a plot of the signal to noise for the top and the bottom. So this is what happens. So you see, after 20 stacks, we don't see anything with the spinning disk. And the other one, we weren't, I think, 100 of those stacks. Maybe we lost 10% of the signal and we stopped because we got bored. So the point here is that this form of microscopy is very non-invasive and allows you to now do volumetric imaging. And you can then extract information. You can get positions. You can get the amount of signal. You can get composition. You can get rates. You can get kinetics. So an example is shown here. So in this case, we're now tracking every single clustering code here. And the traces that you see here correspond to the, in this case, we're just following the position of the clustering code. So it's color coded on time. But we can also follow content. So I won't have a chance to show you, but we can now put cargo. We can put the preferred cargo for these clustering codes could be transferring or LDLs or viruses. And we can now track for every single endocytic event how many molecules are being captured, how fast being recruited, and what's happening. And then we can, of course, do this with several molecules at the same time. So in the lab, we have now, I think, generated at least 30, 40 lines, different genes for the endocytic traffic and trying to get correlations sometimes to understand what's going on. OK. And those details in the right really doesn't matter for now. OK. So we've done lots of studies with cells in glass, and that's OK. And I was very excited. In fact, we have one machine in the lab that we have built together with Eric, and then we had to transport it to my lab. Nobody wanted to get the insurance. It was like a third of a million dollars, something like that instrument. And I wanted to move it from Janelia research campus in Virginia to our lab. And I said, OK, it's easy, right? I'll just hire a piano-moving company. And they're used to move expensive pianos. They refuse to do that. I tried FedEx. I said, well, as long as you have an invoice and it shows exactly what it is, well, no way I could do that except the cameras and the objectives, right? So we send that by FedEx. The rest, we put it in an SUV, and we drove it for 12 hours to Boston. OK, we built it. Everything was fine. And everything was working all right for a couple of years. And then I decided to get a second instrument so we can do virology. So we actually got it. Everything was up and going. But then I ended up interacting now with Eric in his new microscope, the one I'm going to describe now. And that was a disaster, right? Because now I got jealous. I need to have that instrument as well. So I sold my lattice number two to a colleague. And now we're building a new microscope, which I'm going to show you the results. What I'm going to show you was done with a prototype, which was a very large instrument. I was like, I think the length of the instrument was maybe 2 and 1 half of these tables, very big. So in the last eight or nine months, a team together, from Eric and our lab, we've been rethinking how to shrink the instrument to make it sort of a more compact instrument that can then be spread to use in other labs. So I think by July, we should have the instrument built. And then after a few months of tests, I think one can release for other people if they're interested to build it. And so I want to show you now what you can do with that instrument. And that's why I'm writing here, can we do anything in this cellular setting? And I would like to convince you the answer is yes. OK. So the trick that Eric invented or adopted is taking the lattice slide sheet microscope and having these adaptive optics. So adaptive optics is the trick that the astronomers use when they're using their ground-based telescopes to look at the stars. And as you know, when you're looking at the stars, they're all blinking. That's because the refractive index is changing all the time because of thermals. The astronomers, what they do is they have a mirror that is deformable. So they project the image back into that mirror. And they shoot a beam, a laser beam, to about 90 kilometers up. And they excite sodium ions. They get a spot. They know it's a spot. It comes all blurry back. The mirror changes shape dynamically. You get back your focus spot. Now the optics are corrected. And that's how you can get good pictures. So the same logic Eric introduced in the instrument, but it's a little more complicated because we have to deal not just with what's going on in the detection site, but we also have to deal with what goes on in the excitation. Because the sample that we're going to put here, which can be now an embryo, or can be an organoid, or embedded in collagen, has lots of perturbations in the optics. And you really need to fine tune it. So we need to do now corrections in the excitation, in the detection, in the focus side. So this is in practical terms depicted here in the left. So you will see in a minute the biological example. This is what happened in the telescope. So it's just flicking between correction and non-correction. Everything gets fine and sharp. So just here in the right is how the optics is done in the telescope. We're going to do something similar in our microscope. So we also need to change the way we are illuminating. We told you that we're using a lattice light sheet microscope. What you see here is the cross section of these beams that were coming to create the sheet. So we create this pattern, and then we're having a mirror that is going to deler this, and that's creating a plane that if you would have an objective that is in the ceiling looking down, that's how you now get your plane of imaging and your focal plane. So that's when everything is OK. But if you put a sample that is not OK, like this sample here where we just put, let's say, matricle or collagen, everything gets aberrated. So we then need to correct. And what we do in this case, instead of another flexible mirror, we go back to the optics that generated this. And this is a reflecting surface that has about 1 million optical components. And very quickly, you're flickering so that you can retune the image that you create in the back aperture than to create this lattice. So if you're interested, we can discuss this physics later. But I would like to go now to the biology. And then the other thing we need to do is be sure that the focal distance between the objective and the plane is perfect. So it turns out that when you have changes in temperature in the room, everything changes a little bit. In fact, this is the most expensive thermometer you can think of. I mean, nothing of a change in temperature change completely changes your focal plane. So we have to also very actively correct the focal plane. And we do that by just retuning constantly the sharpness of the image. OK, so you do all these things. And now comes David's sample. So this is a sample that he provided. This is basically a cluster of cells that he will discuss assuming more detail later. The only reason I'm showing you this is the difference with and without the correction, right? So you can see that after you correct, now you can track every single event. So Gokul and Francois and my lab brought code that allows you to track in 3D every event. And in this particular case, what is color coded here is the time it takes a code pit to form and then But this, what you see here in the right, is now tracking every single endocytic event that happened in one cell, but now a whole cluster of cells. So we're looking through the whole volume of this. So that's the difference that we can get. OK, so now I'm going to show you examples. I'm going to show you examples now with animals. So we ended up tuning on zero fish because it's a transparent organism, so it's useful. And so I'm going to show you images that are made not just of a single stack, like when we're doing the cells, but now it's going to be dozens or hundreds of those stacks taking all as a concatenation. Then we glue them together. So now you get a full movie of a very, very large movie and that then will be cycling. So this particular example I'm going to show you first comes from this region of the animal. This is a zebrafish. It's about 30 hours post fertilization, so we have it immobilized in the microscope. So we're going to be looking at this region of the animal. So that's where the spine and the muscle is located. And the signal is coming from a marker that only goes to the plasma membrane. So I just want to show you, pay attention to the pre and post correction so you can get a sense of how it looks. So muscle is here and these have other cells. And it's a huge volume that we're collecting. It's about 170 by 180 by 180 microns. And this is about 154 stacks that we have collected. And this is sort of the volume that we're imaging. Now what you saw here is a static point. It's not movie yet. So I'm now going to show you what happens when you do movies. So here we go. So I'm going to show you a few examples. So this is clathrin. So this is a transgene animal where green are the clathrin that has been labeled with mnion. And the red happens to be the plasma membrane. We're going to be looking now at muscle tissue. And this is like 2 and 1 half days post fertilization. So the animal is already a real animal. So this was very hard to look at this. So Gokul developed this trick where he identifies every single cell and then that's an expansion, a virtual expansion. So each one of this is now a muscle, a fiber. But this is a single time point. There's no dynamics yet. So if you hold on for a few seconds, you will see that now the movie will start. Now we have time. So you see every single clathrin codifesticle is forming both endocytic from the plasma membrane. And we also have the guys in the transgolgen network. These are the secretory guys. So it is a very complex image at this point. Just for you, we just separated each cell so you can get a visual on this. But we can actually analyze this mathematically. And again, using the same trick I showed you before, we can track now all the endocytic events. And they're now here. So we can separate what's endocytic, what's plasma membrane. Basically, you can now follow endocytosis in the context of the animal. And you can ask a lot of questions. In this particular case, we're comparing the dynamics of brain and muscle. And if you're interested, we can go into those details. I mean, it's not so important right now. Just to indicate that we can do the analysis of things. OK, I've been interested for a long time and what happens when a cell when it divides. Some years ago, we found that when a cell divides and goes through mitosis, the volume is roughly constant. But the cell loses surface area in mitosis. And then the two daughter cells massively recover, massively recover surface area. When they do cytokinesis. So this particular movie, where now you see the single isolated cells, we're tracking the plasma membrane, we're also tracking organelles. So I can also tell you where the location of the organelles are as the cells are dividing. So we have hundreds of cells that we're tracking here. And this happens to be now the brain, which is a rich tissue where you're getting a lot of what's called asymmetric division going on. So cells are taking decisions to decide whether you're making a neuron or other cell types. So again, the trick of separating the cells from each other, nothing manual, my friend. It's all automatic. So we're dealing with this and so it's all automatic, right? So you can go to Paris and have fun. So this will show you for one cell, just for one cell, what happens on time. So here you are going to have the volume and the surface area for this cell. And what is being plotted here is the relative position of the organelles as you divide. So there's some jiggering around and then there's a reset of the position as you go through mitosis. So for example, the mitochondria go to the periphery of the cells when you divide in metaphors. Not surprising. You have all this at the skeleton that's reorganized. You have the microtubules, the spindles pushed away. The ER is the same. And then you can see here, for example, that you have the surface area of the cell interface. There's a dip. So the dip is because we found that endocytosis is kept. But recycling shuts down. So you keep consuming your membrane through the classroom pathway. You remove membrane as you round up. And as soon as you do cytokinesis, you reactivate the exocytic pathway. And all the membrane comes back. And the surface area of the two daughter cells added is the same as the mother or the father. And that recycling depends on calcium and uses the same scenarios as regulated secretion. And for some of you that have done tissue culture and you see bleving cells, sometimes one thinks that's apoptotic cells. Actually, the cells are dividing and bleving all the time. And that's this particular phenomenon. OK, I'm going to show you technically how we collect these large volumes. So what you have here is now just a slab. This is a three micron slab. And this is a tile of 7 by 7 by 3 of those stacks. So I'm going to show you different regions, how it shows with no correction. I'm sorry for the light. I don't think you can see that very well. But this is with no correction. This is when we do the adaptive optics. And then finally, the complete image. And then what we do is we first do the adaptive optics in the whole volume. And then as we keep cycling stochastically, we go through different locations and we correct. So the correction takes a fraction of a second. So we don't lose time in the actual data acquisition. OK, so I'm almost done. This is one more example here. In this case, what you're going to see now is these blocks. So each one of those was one of those stacks. We had to write code to stitch them back together so that the images are all nice. And what you're going to see now is the eye of the fish. And in a minute, you will see that you can see all the texture of the eye. And the reason we were doing this is because the eye is very active in cell division. And there's interesting decisions, again, that are happening in what's called a neuroimpetilium, that links where the retina is going to appear from the rest of the tissue. So here Gokul is showing you different ways of presenting the data. That's a little bit easy. You can see now where the eye will form. This is one day after fertilization. And then this is actually a movie. And the reason we were doing that is because we were looking, we were interested in just following what was going on in cell division here. OK, this is a more canonical way in which you can present. This is an ortho projection just to show you the quality of the data. I mean, there's no bleaching and there's no loss of intensity. And how long does this take this kind of? This particular one takes, well, each stack must be, each stack is maybe three seconds. And there were 120 of those. So how do you see the mitosis? Well, here because? Not here, when you looked at the mitosis. Yeah, because the cycle that we're doing is every minute or something, right? OK. So you can just go for hours. Yeah. OK, some of you like cancer. So here we go cancer. So this is an experiment that comes from Martin and Mathis Group. So they generated, they provided a cancer, human cancer cell that is in green. And then the fish, the blood vessels of the fish are purple, right? And so just to show you the exquisite level of detail you can see right now. You can see the cells. You can see different forms of migration. You have tails and things protruding. I mean, you can measure the volumes and speeds and you can do tons of things, right? And so we have done some of that analysis as well. So I'm going to end with one more fish, one more movie. One more fish. OK, I'm going to show you. I'm going to show you in this particular movie. It's a movie. It's very interesting because it's the head of the animal. This is with Sean Megason. It's a colleague in Amsterdam, his group. So they are very interested in the development of the ear. The fish have ear. I didn't know that. They have ear. But they don't listen. They use ear, the inner ear just for 3D orientation. So we were focused in this region here just to see the early stages of how the ear formed. So just to orient yourself, what it's shown here is membrane in color. All the membranes are labeled. This is a transgene animal for a cat. I mean, it's a JFP going into the membrane. This is the outside of the animal. This is the skin. There's going to be a capillary here. This is brain. And the point of this movie, as you will see in a minute, is that you will see now neutrophils, part of the inflammation recognition system. They're going to be crawling here. And I just want to appreciate the level of detail that we have. And in blue, what you're going to see is dextran. We injected fluorescent dextran into the heart of the animal. So they also did endocytosis and phagocytosis. So you see the blue spots. This is the endocytosis that went on in the endothelial cells. And there you have this guy. And there's another one here. They're just exploring and go out. Now, this is a 3D movie. Unfortunately, I don't have it. I cannot have glasses. I don't have VR right now to show it to you. This is a 30 gigabyte movie. It's a bit large. But the level of detail is exquisite. You really can see all the details of the membrane. But you can see here it's slightly enlarged. And then what is interesting also is that there are lots of different biologists. That's why I said this is like renaissance. You just look at things and you get biologists. So this is the inner ear that is developing here. And there's cell division and you have migration, et cetera. Another thing you might notice is that the whole tissue is vibrating. This has nothing to do with the pulsing of the heart. I have no idea what that is. But something in the tissue is moving. Honestly, I don't know what it is. And another thing that we could also see was there's some Tom Conberg has coined the term he calls cytonym. It's like maybe like this philipodia, connections between cells. Now we can see it quite vividly. I'll show you in a second a movie you're going to watch in your iPhones and smartphones. OK, and then we can track. In this case, we're tracking the speed of the guys. So that's color coded for speed. So I normally ask people to not use their telephones or smartphones. I'm going to do exactly the opposite. There are two URLs here. These are movies. We can start with this one here in the bottom. I don't know whether the connection, the internet connection here is a bit slow, unfortunately. But this one here will show you that movie where you had a neutrophil moving. And when you bring it up and you activate this YouTube, you can then move your phone around and you can see the cell. You're like being inside of the animal and you can be in. And then you can see the neutrophil crawling. The upper, is it working? No, I think it's really a period here in the YouTube. A what? To you see YouTube. This is correct. Don't challenge me. I know, I shouldn't pick. And this one here is the, you saw the muscle, which I show you the classroom spots. So that's a static view. That's not a movie, but you can be inside the muscle and then you can just walk around 360 degrees, go up and down just so you can get a feel of the 3D of this, OK? So at least try to, so can I go to the next slide? Have you written down the, yes? You can take a picture. Yes, can I continue? Yes. OK. Yes, OK. I'm done. So basically, I just finished, I just show you where are we with imaging at the level of organisms. I just gave you a teeny, teeny summary of the experiments that we're doing. In my opinion, it's an explosion. It's hard to keep up a lot of issues, how to track, how to image, how to interpret the experiment, how to do experiments, right? And what should you do? What topic you should focus? But I think it's a revolution. I think it's a dream. I mean, at least when I started looking, I think it was very hard. And now I think it's possible. Now what we need is people that can help us do the analysis of things. So I just want to finish acknowledging what the people have done to work. So I started with the PIP sensors. I mentioned Kangmin, he's a postdoc in my lab. And Bobby, he just finished his PhD in physics. And now he's continuing postdoc training in theoretical computational biology. The work on the light sheet microscope started with a team that I have here in the right, where Eric Marino was the guy that built the instrument. And then Gokul has taken over. And he commands now my whole group to deal with imaging. Wes is the person that also helped us a lot from Eric Betsy and from Bichang Chang on the first design of the instrument. And then we moved to the adaptive optics. And Sung Lee, as I mentioned, was the one that built the second device. All this, without Eric, would be impossible. And it's an amazing interaction that we've had in the years. The work on zebrafish, it's some of the key experiments with the ear come from Sean Megason's lab, and Ian and Kishore were instrumental in helping setting up the experiments. And finally, I get funding support from NIH and two companies, Biogen and INS. And with this, I want to thank you so much for your attention. Feeling for the sensitivity of the lab's light sheet. So for example, compared to doing a turf movie, looking at class and media events, what about in the zebrafish? So we run normally our experiments at the 3GFP level of sensitivity. We can go better than that. But we chose that because then our photon dose is low enough that we don't bleach too fast. And all the things I've just shown to you are done at that level of sensitivity. So with that, you can go, as you can see, but extensively. But we have also done single molecule experiments. I'm just going to pull up on that and just ask you. So you're showing that you can see fluorescent proteins. But these are fluorescent proteins are massively localized to sub-cellular surfaces, structures. You mentioned single molecule, TFP imaging. How does that work with the ductile optics? So the trick is that the astronomers used a laser that shoot up to the sky to create an artificial star. It's called a guide star. We do the same trick. We have an extra laser. I didn't tell you. We use a two-photon microscope. I have two-photon laser system, right? I'm surprised you actually need an internal reference. Can't you just use structured imaging to produce an internal? In theory, it's possible. In practice, it's a nightmare. So it's much simpler to just create the artificial spots. Part of the problem is because the biologist is moving, right? I need to have a point that is not moving to do the adjustment. So when I come with my two-photon laser, that's not moving, right? Just a practicality. I have a question on the first part of your talk about the sensors of phosphoresicides. So I may have missed it. But then what you're suggesting is that there are bursts and waves of different phosphorylositides after the vesicle has been building. But then what in your view is the consequence of those bursts? So how are they changing some properties of them? Oh, yeah, yeah, yeah, yeah, yeah. Have they been detected by? Yes, I went too fast. Let me, OK. So let's start. Let's start, OK? The puzzle was the way the story started was with a puzzle. I knew the proteins that were required to encode the clathrin code. It was HSC70 and Oxylin. We did the, in the field, we did the biochemistry. You know, there were structures. I mean, we did the cryo, we did crystallography. Everything was fine, OK? The problem is that these guys, ah, one more thing. I can even make a synthetic clathrin code. I can add the components and, poof, everything comes back, right? It comes out. Everything is fine, logical, right? The problem is that in vivo, when I'm building my clathrin-coded pit, it doesn't fall apart. It's actually building all the way from beginning to end. It builds, builds, builds, builds, builds. And then by magic, it falls apart. The magic occurs after you pinch. So there's a signal that says, now you can encode. So we then did the cofactor with GFP. And sure enough, it came as a spike. That was the Oxylin. It was coming as a spike at the end of the, when a coded vesicle was forming. Green also got the same things in it. So we knew it was coming as a spike. So that's a signal. So what was the signal, right? It could be a change in conformation in the code. That was one possibility. But then I said, well, wait a second. There's this domain that looks like p10. So that's how it started, right? So then it's showing up. If you get rid of that p10, the Oxylin is not recruited anymore. So clearly that was the signal that the, what was reading and recombined. So at the end, turns out that you have to have the right lipid, Oxylin likes PI3P. And when you have the right lipid, bingo, it comes. And we have now repeated everything in vitro. We can make artificial coded vesicles and it works exactly the same. Can I just say that a half moment for me, you didn't say it to her, is why does it happen after the fission? It's because it stops getting replacement of the phospholipids from the plasma membrane. This is when all these enzymes can come along and who cares what they are? Well, the enzymes, some of the enzymes come, they're even all the time, right? But the point is that if the membrane is open, you're getting a change, there's the fission. And then when you close, now that's it. So basically we're connecting a change in topology. We're connecting a change in topology with a change of composition, therefore a change in recruiting of enzymes or of factors or whatever, right? So I think the same principle occurs when you are trafficking endosomes and lysosomes and stuff like that. Yeah, my live question, I was very impressed by your movie to follow the development of the organ, like zebrafish and retina. Have you ever tried to compare the development between normal animal and the sort of saturation with the genesis animal, which have the program of the retina development? Have you done this sort of experiment? My answer has two parts. You're welcome to join the lab. No, it's a very good question, right? I just need to give you a sense of timing, right? This microscope started to become operational, I would say a year and a half ago. We've been collecting data that comes from other labs or from our lab for one year. The paper was then, we had to spend time writing the paper and the paper just got accepted in science. Yeah. And now we're in this, so we have dismantled the machine, doesn't exist because we're building the new one, right? So it's simply a practicality that we haven't been able to do more experiments. So, you know, suddenly in my lab, I'm desperately now looking for people that would like to come and join. And if you know somebody who would be interested in working in zebrafish that would like to come, I would be more than happy. Yeah, yeah. David. I'm wondering your model about the scission and the lipids continuously replenishing the lipids and exchanging between the vesicle. I wonder, because I know that proteins like bar proteins can slow the diffusion of lipids. And so I wonder, have you ever actually measured the diffusion? Yeah. So the experiments that we have done, I'll just describe what we have done, right? So we have targeted a PI4 kinase, a PI, sorry, I'm gonna start all over again. We have incubated the cells with a PI4 kinase inhibitor, right? And that immediately shuts down the signal in the coated pit, right? So within seconds, the signal is gone. So you're consuming, you're inhibiting the kinase is being consumed in the plasma membrane, but it's gone, right? We have done the same both with optogenetics and also with rapamycin to send a phosphatase for PI4, 5P2 to the plasma membrane and just disappears very quickly, right? I mean, also within, you know, we're talking of seconds. So if you're asking me at the millisecond level, I cannot answer that, right? But at the second level, at the second level, it is exchanging very fast, yeah? I was wondering if you could offer any more details of the magical breaking cells apart. How is this done, how do you separate? I imagine it's quite challenging. Well, okay. So basically, you map the position of the membrane, you do tensor analysis, you do water shedding and then you also check the position of the membrane. You check, you have the boundary, right? And then you have the thickness of the boundary and at the same time, we also have other markers in the cell. You have cytosolic markers, so that gives... So you have a way to tell which membrane is an outer membrane, but... Well, you get... So if you have a cell that has nothing, nothing beyond the cell, that's trivial, right? You have two cells that are next to each other that's non-trivial. So we just take this signal in between, right? We don't have the resolution of the optical microscope to see this type of thing, right? Because the optical resolution that we have in the z-axis is like, let's say, 700 nanometers, right? And x and y is a little bit better, so we don't even bother to do that, right? So it's like an average that you get. And then after you did that mapping, just then there's a transformation, right? Leave me a transformation. Yeah. Okay, so... Okay, well, thank you so much. Yeah.