Scientist Stories: Markus Sauer, Super Resolution Expansion Microscopy

Welcome everyone to this edition of Imaging One World, and today we have Markus Sauer from Würzburg. Who will be talking on super-resolution expansion microscopy. I guess Marcus doesn't need any introduction. Anyone in the field just knows him. But I have— I was watching his Microscopist interview with Peter O'Toole, and there he mentioned a very interesting bit, and I encourage everyone to watch that interview, which is basically he was initially not interested in physics and chemistry, and he deselected them during his grad courses, but, and he was interested in medicine, and then he went back on recommendation of his medicine supervisor, and then he came back to physics and chemistry. So I, I want to listen to Marcus, so I won't waste a lot of time detailing what he did. I guess everyone knows D-STORM and, and, and the impact it has made in microscopy and super-resolution microscopy. So, Markus, I'm all eager to listen to you, and thanks for coming to our series.

Thank you very much, Kirti, and thanks for the invitation to talk today here on this platform. Well, the title has already been announced: Super-resolution Expansion Microscopy. So I will take the chance and just very briefly talk a little bit about localization microscopy with a particular focus on D-STORM, of course. A few slides and then I will come to discuss about technological advancements to further push the resolution of optical light microscopy. And one of these examples, at least for me, is expansion microscopy in combination with super-resolution microscopy techniques. So, fluorescence microscopy, just a single slide, I mean, a very powerful technology Thanks to people who introduced fluorescent dyes already 140 years ago, like fluorescein, and today rhodamine and cyanine dyes, we can label almost any protein of interest— sorry, almost any protein of interest and do fantastic multicolor fluorescence imaging on confocal microscopes. But there is, of course, the resolution limit you're all aware of. And so structures well below 200 nanometers we cannot resolve, and this has dramatic influence on what we see. For example, if we go into neurobiology, what is shown here is a presynapse and the postsynapse and the synaptic cleft. All these structures that are key to functioning of such a synapse, they have dimensions less than 200 nanometers. So 200 nanometers is shown on the right-hand side by this blue-gray circle here at the limit of about 200 nanometers. So the vesicles which which carry the transmitters which are released upon fusion with the presynaptic membrane, have a diameter of 40, 50 nanometers, and then the synaptic cleft, there is a distance to the receptors on the postsynaptic side of 10, 20, 30 nanometers something. So we cannot resolve these really important structures using standard microscopy, but there are techniques which have been invented by different people, different groups, and the Nobel Prize has been awarded to different groups, which allow us now to bypass this diffraction barrier and achieve resolutions down to a few tens or below 20 nanometers even, latest technologies. And what's behind single molecule localization microscopy, I guess you're all aware of, it's just the temporal control of fluorescence emission of single fluorophores, which allows you then to precisely determine the center of mass, which is called to localize them, and the precision with which you can do that and finally reconstruct the super-resolved image is given mainly by photon statistics. So how many of photons you detected per one event of these fluorophores. So what was key for the development of this technology? The idea had been around already 30 years or so, but it has been limited by the non-availability of photo-switchable or photo-activatable fluorophores. But this has then been introduced early 2000, 2002 by George Patterson, who unfortunately passed away last year, at that time postdoc in Jennifer Lipincott-Schwarz's lab. When they introduced the first photoinactivatable GFP. And then different groups have been playing around with photoswitchable organic dyes, like Xiao Wei's group in Harvard and my group that time in Bielefeld. And this has then been termed STORM and dSTORM, whether with or without an activator, that you can switch organic dyes with a mainly cyanine dyes, where have been the first dyes where this principle has been shown successfully, can be switched between an on and off state, if you add a thiol buffer. So 100 millimolar of a thiol buffer. And this was also not really new because this thiol buffer as a stabilizing agent has been introduced already by Yannick Guida in his seminal Nature paper about single-molecule tracking of cyanine dyes in 1984. So, well, having these technologies like STOMP/DSTOMP, you can label your target structure of interest with commercially available probes, like shown in the background here. This is together with LeucLav, is a photoactivatable organic dye, and, and upper left corner, these are nuclear pore complexes as we published them 2011, when we introduced these nuclear pore complexes as a reference structure for super-resolution microscopy, and where we have been able to count this 8-fold symmetry for 99% of all these nuclear pore complexes, we could really resolve the 8-fold symmetry. And we're using this photoactivatable dyes here for the actin skeleton, labelling actin with phalloidine together with Luke Lewis' group, and we achieved localization precision easily of only a few nanometers already at that time. But I mean, albeit this goes at the cost of imaging time, so these images you have to record, we have recording times of a few hours. Well, but nevertheless, so it's a very powerful technology and many people said, oh, these 8-fold symmetries, this was just cherry picking, but this was not cherry picking. We using Alexa647 or Cy5 as a fluorophore in D-STORM imaging achieves a very high reliability for really more than 99% of these nuclear pore complexes. We could count 8 or even resolve the central channel of the nuclear pore complex with diameter of about only 14 nanometers. Well, so what's the trick behind it? Why does— oh, I'm sorry, I have to restart just my presentation a second because my videos are not running, and this often is the case and they often run then again if you restart it. So sorry, I share my screen in a second again and hopefully then we can, we have these things running. They are not running, this is a pity. Okay, so the beauty of D-STORM imaging On one hand side is of course that you can buy antibodies, you can purchase antibodies from different companies, and of course they then all have different degrees of labeling. And this was then why we underwent a study of the influence of the degree of labeling on the performance of these antibodies in D-STORM images. So Dominik and Gerti at that time investigated different DOLs and the influence on D-STORM imaging. And as you see on these images, so between a DOL of 1 and 8, the resolution is almost the same. What is increasing is the background signal because highly labeled antibodies have a specific or increasing tendency to bind non-specifically. So the DOL doesn't play any role in D-STORM imaging. Why is this so? It's because of, of the fact that no matter how many fluorophores are labeled to an antibody in switching buffer, they all behave like a single photon source. So, there is always only one fluorophore in the on state, which is showing this blinking as you expect from a single carbocyanine dye as in this example. Well, if you do this imaging, single molecule trajectory imaging on a confocal microscope, as shown on the left-hand side in PBS, you see also doing anti-bunching measurements that that for a DOL of 8, of course, you have more than one fluorophore contributing to the light field detected by two APD detectors. And this is also reflected in the fluorescence trajectories. You see stepwise photobleaching. But in switching buffer, there is always only a single fluorophore residing in the onset. So, well, the last image is about, about dSTORM. Because I want to come back to the synaptoneural complex later, I will introduce it already here. So, it's that you can use now different antibodies from scratch from companies, from colleagues, and label all the proteins, for example, of a multiprotein complex like here the synaptonema complex, which is a well-preserved structure among many different species. It tethers the chromosomes of father and mother together during meiosis. So, it forms a ladder-like structure, stays in this ladder-like structure for a few hours during meiosis, and then it disassembles afterwards again. So, we have lateral filaments, central filament blue in the middle, and some transversal filaments holding the whole complex together. And while testing now all the different antibodies available directed against these different proteins on the right-hand side here of the synaptoneural complex, you see that we finally can resolve the three-dimensional structure or the molecular architecture of this synaptoneural complex with almost isotropic nanometer resolution. And the beauty of nature is really that this is a helix-in-helix complex, the synaptoneural multiprotein complex. All right, well, I mean, so far, or in most labs, super-resolution or localization microscopy is mainly done are in TIRF mode. So the specimen or the, the object you want to image has to be in close contact to the coverslip in order to excite efficiently the fluorophores in the evanescent field. But there is also a controversial discussion going on about whether this contact of a cell with the coverslip, for example, changes the organization of especially membrane proteins, for example. So we thought about, can we combine D-STORM imaging with lattice light sheet microscopy. So we adopted the lattice light sheet introduced by Eric Betzig and colleagues at that time in Science 2014, and Felix and Jan in my group, they then equipped this, this lattice light sheet microscope with high-power lasers to achieve then finally the possibility to— I will try it again because my videos are not running, please excuse me— to stop it here. And maybe load the PowerPoint again. Now, just a second.

Or, Markus, if you have them stored on your computer, you can just play them, share the screen. Because I know some of these videos are really awesome.

Usually it works. It's just, I mean, I don't know why it would take too long to find them on my computer now. Ah, now it looks better. I see some videos running.

Yeah, it seems to be working now.

Come on. Oh no, I don't know what's wrong. But anyway, uh I mean, finally we combined then lattice light sheet microscopy with D-STORM imaging, and by doing that, using a lattice light sheet with high-power lasers, we achieve lower localization precision of only 15 to 40 nanometers in XYZ, but we can visualize then, as in a few examples demonstrated in this paper, receptors on the plasma membrane of cells. In 3D on the basal and on the apical membrane, or here's activated primary tumor cells, CD3 clusters unperturbed by any surface effects. So this was the status quo where we have been a few years ago, and we were still using these technologies together with people from the university hospital here to diagnose or to identify cancer cells, going to improve immunotherapies, for example. So, but what is puzzling us in the last years is how we can potentially further improve the spatial resolution of localization microscopy. Or can we finally really achieve true molecular resolution? And there have been techniques developed and advances have been made to push the localization precision of STORM, dSTORM microscopy, for example, using or developing MinFlux. Technologies or microscopy that achieves a localization precision of a few nanometers only, ideally 1-2 nanometers or something. But I mean, still this high localization precision has not been successfully translated to biological samples, to cells for example. And problems associated with this are mainly that, I mean, as long as we label with antibodies, we are limited in labeling efficiency. Linkage error is dramatic.

Let's—

if we use a primary and a secondary antibody, you displace the fluorophore 10, 20 nanometers from your protein of interest and introduce such a large linkage error. Then the labeling density is limited only due to the sheer size of these antibodies. Of course, there are smaller labeling methods like nanobodies versus proteins, but still they have a size of a few nanometers. And then all these methods usually work good or really good only with a single dye. So because these dyes have to fulfill demanding requirements like this photo switching, and the microscopes we're using are usually expensive and require expertise. So, and this is maybe the reason I'm coming back to Kitty's paper here, about this, this, this, this debate which has been sparked about the resolutions achieved by these different single molecule localization microscopy techniques. So, from the first introduction of the nuclear pore complex 11 years ago by us here, then Jonas Ries achieved a much higher localization precision and also nicer images than 8 years later here using SnapTag technology to label nuclear pore complexes.. And but then MINFLUX, even though the localization precision is dramatically higher, they came up with these two color images which show that, I mean, for some of these nuclear pore anchoring nuc proteins, the resolution seems to be super high, but the image quality is not better than in these previous works. So we thought about, first of all, can we achieve a higher resolution in 3D and in multicolor mode? And the idea to make this accessible was to use expansion microscopy in combination with super-resolution microscopy. So, for example, if you just imagine 10-fold expansion in combination with structured illumination, should allow you to access 10, 20, 30 nanometers. Spatial resolution in 3D with different colors. And structured illumination microscopy is not very demanding concerning fluorophores which can be used. And the other thing is, how can we reduce the linkage error? And one possibility to decrease this linkage error and increase also the labeling density might be to label proteins after they have been expanded. And I will talk about that in the next 10 minutes or so. So what is expansion microscopy? It has been introduced by Ed Boyden in 2015 in the seminal Science paper here. So it's also not a completely new technology he introduced that time because clearing has been used in pathology for tissue imaging already for decades. Already more than 100 years ago on the right lower side, there's the, the glassy woman, which she's still in the museum in Dresden in Germany. So by changing these protocols of these clearing methods, David achieved that samples can be expanded after gelation. So what you first have to do, you have to link your protein of interest into a charged polyacrylamide gel, and then you digest or destroy in a way the interprotein or intermolecular interactions in this sample, in this cell. And then, because this polyacrylamide is negatively charged, then you put it into water overnight, and then it's swelling. Yeah, and your fluorophores are still attached ideally in the polyacrylamide meshwork, network. And then this negative image has of course then, depending on the expansion factor, a much higher spatial resolution. So you can achieve super-resolution on a standard confocal microscope easily. And immediately after this first paper, we tried to share forces together with Ed Boyden then and another group in Geneva from Paul Guichard, Virginie Hamel, and that time postdoc Davide Gambarotto. We thought about these people that are experts in cryo-EM, so they are working with purified centrioles usually, and we thought about can expansion microscopy be used to isotropically expand also such delicate or complex multiprotein complexes like the centriole while preserving their ultrastructure. So we tested different protocols that time already. So the pre-expansion protocol shown here in the upper part, so that you fix your sample, you label, you link your labels into the polyacrylamide gel, you digest or denature or destroy in a way the protein interactions, and then expand your sample and do imaging. Or that you do the labeling after expansion.. But this requires that your epitopes, your protein epitopes, have to survive the digestion process. And then having a centriole, a centriole in the top view looks like a 9-fold symmetry with 9 microtubule triplets. And from the side, we have long tubulin filaments. So, and by using standard confocal microscopy of labeling two of these tubulin filaments, poly-E and alpha tubulin in two different colors, you see that confocal microscopy we cannot resolve anything in an unexpanded sample. Using the standard first published protocol, eXm, the expanded centrioles looked like shown here in panel B. And using the MAP protocol, the post-expansion labeling protocol, it looked like that. And this all shows that the ultrastructure was not, is not preserved or could not be preserved during this expansion process. And therefore, we tested and further optimized these protocols and came up with this ultra-expansion microscopy protocol. And the trick was that you have to almost completely avoid fixation and use a lower formaldehyde and acrylamide concentration. And by doing this, you can preserve the ultrastructure of centrioles. This is what exactly what you expect from poly-E and alpha-tubulin and the long filaments. And you see here also in front of you exactly how a centriole should look like. Coming back to the synaptonemal complex, this time Fabian worked on that together with Ricardo Benavente's group and Marie-Christine Spindler's from the BioCenter. And what I show here are three images. Left is a standard structured illumination microscope image of one of the lateral filaments labeled with antibodies. And these are expanded views here. This is a D-STORM image of the same. You see this helical structure of the lateral elements, how they twist. And on the right-hand side, we have a threefold expanded, only threefold expanded sample and post-expansion labeling of one of these lateral element proteins by structured illumination microscopy then. And this allows you of course then to label 3 different proteins here, let's say a lateral element, a central element protein, and transversal, in this case at the N-terminus, antibodies for the N-terminus. And in fact, then for the central element and the N-terminal end of the transversal filament, this protein signals overlay in a 3-column MAP-SIM, 3-fold expanded structured illumination microscopy image, and you can see that you can image and visualize large field of views here of synaptonemal complexes in 3D. But the most convincing part came then that we could now for the first time visualize details of the synaptonemal complexes which have never been shown before by optical microscopy using antibodies. So by first expanding the sample, not digesting the sample, but heat denaturing the sample to preserve the epitopes, we can do post-gelation immunolabelling. And then we could see things which had never been seen before. On the right-hand side are two electron micrographs of synaptoneural complexes. And you see that these lateral element proteins from time to time split into two or more sublateral filaments, and that they show this fraying at the ends. And now we can show this fraying at the ends for the first time by optical microscopy, and also the splitting into sublateral filaments in the structures of synaptonemal complexes. So why can we visualise them? We think the reason why we can visualise them is because we use post-expansion labelling.. So, and we, in principle, this demonstrates that after expansion, there are more epitopes accessible. And exactly for the splitting of these lateral elements, this into sublateral elements, if the splitting is only 10-15 nanometers, there will no antibody bind in between in an unexpanded state. But upon expansion, antibodies can bind and therefore you are able to resolve also those sublateral filaments or the fraying of these lateral filaments at the end. Well, just to complete this expansion of cells, we also came up with expanding membranes. Membranes are far more difficult to expand because they do not have amino groups for in order to link them into a gel, into a polyacrylamide gel. But there Jürgen Seibel from the organic department of the University of Würzburg came up with a new functionalized ceramide, which is a membrane, a lipid probe, which automatically inserts into membrane if you add these lipids to the membrane. And this is now a very special sphingolipid, the ceramide shown here. So, it has a clickable azide group, so you can attach a fluorophore there by click chemistry, and an aliphatic amino group for linkage of these lipid into a polyacrylamide gel or for fixation.. And then we undertook a few studies with also bacteria where we joined forces with Thomas Rudel and Tobias Kunz. And from my side, Ralf and Jan have been working on it. Okay, so to cut a long story short, this— by feeding these cells with this ceramide, this finger lipid, and afterwards fixing and expanding these cells, here fourfold expanded cells, you see that these ceramides insert into all membranes, not only into the outer membrane, also the nuclear envelope membrane, and all other membranes in the cell. But still, it allows you to label proteins by immunolabelling. And this is what we did here. We labelled the mitochondrial matrix protein peroxiredoxin 3 by immunolabelling and using tenfold expansion now, we investigated now bacterial infections of these cells. So we infected these cells with Chlamydia infection bacteria, and they are then visualized inside of the cell here in green, because this ceramide also inserts into the bacterial membrane. And in magenta, we have these mitochondrial proteins. So we see here mitochondria. And from time to time, we see these contact sites between these intracellular bacteria and the mitochondrial protein. And we see even an uptake of these proteins, of these mitochondrial proteins into the double membrane of this Gram-negative Chlamydia bacteria here. Tenfold expansion structured illumination microscopy. What finally convinced us about the power of this technology is that we could also for the first time by light microscopy then visualize the double membrane of this Gram-negative bacteria of Chlamydia.. As shown here, these are chlamydia bacteria inside of a cell. On the left-hand side, this is the cell. And these are the chlamydias inside of the cell, an infected cell. And if you go into more details, you see that these are double membrane bacteria, and you can then determine the distance between the outer and the inner membrane of these gram-negative bacteria. And we came up with a distance of about 27 nanometers from electron microscopy microscopy, we know that the distance between the outer and inner membrane is about 25 nanometers. So, just a few examples that you can also use expansion microscopy in neurons, and this is what Stefan, Jana, and Christian are doing in my lab. So, we can combine it with this PanXM from Jörg Bewerstorf published that you label all proteins by NHS. And we also do that post-gelation, so after expansion we label with NHS all the proteins. You see mitochondria, mitochondria, you see the folding in mitochondria in neurons, you see other membranes together with microtubules here together with immunolabelling or together here with bazoon and Homer to identify synapses in the sample. And only from the NHS labelling you can also already identify sometimes synapses because there the protein density is much higher. All right. Lipids can also be fed to neurons and expanded together by with immunolabelling. And you can finally visualize also vesicles on the presynaptic side and HOMO on the postsynaptic side, the synaptic cleft together in such a fourfold-only expanded sample by three-colour structured illumination microscopy. Okay, well, very powerful together with structured illumination microscopy, but how is it about can we achieve real molecular resolution using, using this technology? So intuitively you might think, well, let's simply combine it with D-STORM. However, there are some problems associated with that. One of the problems is that fluorophores are destroyed during this gelation process. So, but we can use post-gelation labeling, so after expansion. The other thing which seriously destroys your efforts is that these expanded gels of a negatively charged polyacrylamide gel, these gels shrink dramatically if you add salts. So a photo-switching buffer containing PBS and 100 millimolar of thiol, the expanded gel will shrink back to the original size. So, but therefore an idea came up about how about if you re-embed these expanded gels in a neutral polyacrylamide gel after expansion. And this is what then Fabian and Sebastian in my group tested on microtubules. So, here in the lower part, you see microtubules labeled by Alexa 532 and D-Stom imaging immunolabeling. And if you do cross-sectional profiles here, you see that there is a dip in the structure, that means the resolution is quite good. And the peak-to-peak distance here is about about 36.2 nanometers. Okay, if you do expansion re-embedding of a pre-expansion, so we do pre-labeling, yeah, we also expand the antibodies during the process, then you end up with a peak-to-peak distance here in the cross from the cross-sectional profiles of about 137 nanometers, a threefold expanded sample. So you expand the microtubules, but you also expand the primary and the secondary antibodies. And then you can model these contributions of your labels to the size after expansion if you do this pre-labeling, all these different pre-labeling options. So here, for example, in green is just immunolabeling, primary, secondary pre-labeling. If you modify your secondary antibody with different DNA constructs of different length, you should end up in even larger diameters or peak-to-peak distances. And this is what exactly is the case. So for immunolabelling here in this experiment, we end up with 134 nanometers. For two differently modified secondary antibodies, we end up in 200 and 226 nanometers. So this experiment we simply did to demonstrate that small changes in the probe size are directly translated into dimensions after expansion because you also expand your fluorescent probe. So of course then again the trick is to go to, to use post-labeling, and then you end up with microtubules with a diameter of about 80 nanometers. And this means that your primary-secondary antibody complex now, which is about 17.5 nanometers, melts down to a linkage error of about 5 nanometers for a 3.2-fold expanded sample. And using epi-illumination, because we cannot use TIRF here, we are in a hydrogel somewhere in the middle at a height of 10-20 micrometers and a water immersion lens, we can achieve here imaging of microtubules as hollow structures, which demonstrates the potential of this technology to achieve high, almost molecular resolution. And then we again translated this for centriole imaging of isolated centrioles. And here you see now these post-expansion re-embedded D-STORM 3D images of centrioles. And from time to time, you even see here the whole structure of these microtubule triplets. These are projections. So therefore, it's difficult to see from inside view. They see these filaments which have a distance of about 15 nanometers. And you can easily resolve these 15 nanometers because the sample is 3 times, threefold expanded. So, and this is an unexpanded 3D DSTM image just for direct comparison. So, well, I mean, there are other fluorophores available on the market. They do not require the addition of any buffers. For example, this spontaneously blinking HMSRR silicon rhodamine derivative which shows spontaneous blinking at pH 7. So we also tested this, but the quality of images we could generate never outperformed the images we achieved using this re-embedding. Due to the re-embedding, the 4-fold expanded gel shrinked to 3.4-fold about. Without re-embedding, we have close to 4-fold expansion, but we also have a pH slightly below 7 in pure water, and there the switching performance of the silicon rhodamine is not optimal, and therefore also the image quality is not optimal. If you then add PBS, the gel shrinks to 2.2-fold, the switching performance, the blinking performance increases, but the expansion factor is lower. So this brings me to the conclusion of this part here, that I think molecular resolution imaging is achievable combining localization microscopy with expansion microscopy if you use clever and tricky post-expansion labeling strategies, because this melts your linkage error down by the expansion factor about, and it increases the labeling efficiency and the epitope accessibility. So, but Let me come back to this debate which is going on about, can we achieve molecular resolution imaging of unexpanded sample as well? And this is again from Kitty's preprint on bioRxiv where he compared here these— I already showed an image 20 minutes ago— compared here nuclear pore complex signals of unexpanded samples of, as we published it, by labeling a nuclear pore complex anchoring protein by immunolabeling. Then Jonas Ries here with his cell line where he, with the SnapTag NUC96, and these beautiful two-color images. And then the MinFlux images. And you see here that MinFlux indeed achieves a higher spatial resolution, but the images always look incomplete. Even though they are using the same labeling technology. So something must be wrong then at the end with the localization probability. And this is what we then started 3, 4 years ago. We started to investigate this in more detail. So whether there is— what's the problem if you want to go below 10 nanometer resolution? I mean, it's intuitively easy to understand that below 10 nanometers fluorophores start to communicate with each other. There are several energy transfer pathways. I mean, multichromophoric systems have been studied already decades ago by single-molecule spectroscopy and singlet-singlet annihilation processes, singlet-triplet, and so on. All these energy transfer pathways have been in detail investigated and unraveled. So we thought about what's going on then for these, in these MINFLUX and other patterned illumination localization microscopy technologies in this sub-10-nanometer regime. And I mean, while with these nuclear pore complexes we can still discuss about the labelling efficiency, but I mean, we can use DNA origami and there quantitative labelling is no issue.. So it remains then open why out of 10 fluorophores only 5 or 6 are detected using these technologies. So this points that the lower image quality achieved might be mainly due to that the localization probability of these fluorophores decreases substantially in the sub-10 nanometer regime. And these are the people here left Below, Domi and Gertie, the driving forces of this project, sure, and helped a lot with data analysis, and Danish and Mara played important parts in this, in this story. So, in order to understand what's going on, we have to revisit the D-Storm photoswitching mechanism just very briefly. I mean, what everybody is agreeing worldwide is that the triplet state belongs to the on state of these cyanine dyes if they switch on and on, and then depending on the triplet yield, they undergo inter-system crossing, end up in a long-lived triplet yield, which is potentially quenched by oxygen, and the singlet manifold is repopulated. And then there is the thiol, which reacts in a reductive way. So a reduction process is going on and produces a long-lived intermediate state with a lifetime of a few seconds for these cyanine dyes. And then and just by chance the single ground state is again repopulated, most probably by reaction with residual oxygen something. Importantly, this mechanism, it has been discussed over decades now, but just recently there was this checks paper by Yaser Gidi and also Martin Schneermann and Gonzalo Cosa finally came up with a really refined transient absorption and theoretical study about the nature of these off-states in Si(V) under different conditions. And they came up with these absorption spectra. So the cis, so a cyanine can isomerize between the trans and cis. The cis state has a strong overlap with the emission of trans Si(V), the triplet as well.. And the reduced Cy5 in the presence of switching buffer shows the main absorption band at 310 nanometers. This is conformed with the study of, of Roger Jiang and Xiaowei Zhuang that the product is a Michael addition chemically speaking. So thiol attacked the double bond. But there is also still some absorption here in the longer wavelength range. Low, but there is some absorption. Yeah, and this is why we speculated, if there is some absorption, then this means that— I mean, this is fact, everybody who is doing D-STORM experiments knows that UV irradiation is in most experiments not required to activate, to reactivate these off-states and to repopulate the on-state. 640 nanometers is usually sufficient to reactivate these off-states into the on-state, to repopulate the on-state. So there must be some absorption of this off-state, also at 640 nanometers. So we hypothesized that this off-state might serve as an energy transfer acceptor for the on-state. The consequence of this would then be, if you have more than one fluorophore in, or interfluorophore distances of a few nanometers only, and let's say 4 fluorophores, 1 is in the on-state, 3 are in the off-state, that these 3 off-states can serve as energy transfer acceptor for the one fluorophore residing in the on-state. For a single fluorophore, you have typically— or for fluorophore separation distances or inter-fluorophore distances of larger than 10 nanometers, you have typically on-state lifetimes in these DOM experiments of about 10 milliseconds and off-state lifetimes of several seconds. However, at distances shorter than 10 nanometers, a certain chance might come up, that there is energy transfer between the on and the off states, and this then repopulates the on state. So we hypothesized that this should result in faster blinking of fluorophores then, yeah, at least during the first few seconds. So we designed DNA origami then to demonstrate that this is really the case, or to investigate this behaviour and did some just blinking analysis of 4 different DNA origami using D-STORM and compared it with DNA paint experiments. DNA paint, so we have DNA origami, 4 fluorophores separated by 18, 9, 6, and 3 nanometres. D-STORM fails to resolve the 9 nanometre DNA origami, 18 we can resolve from time to time. DNA paint achieves a higher spatial resolution and can also from time to time even resolve the 9-nanometer DNA origami, but not the smaller ones. But then if you do some analysis of the on and off rates, you see that for the— shit, the videos are not running. Well, I can explain it. This is DNA experiments. On the left-hand side, it's a DNA origami carrying just a single docking strand, and you see blinking throughout 30 nanometers imaging time. For the 18 nanometers, you see also blinking throughout the 30 nanometers. For the 9, for the 6, and for the 3, you always see constant blinking of the sample because of binding, unbinding of imaging strands. And of course, you see fourfold more binding events for the fourfold labeled DNA origami. But for the DStORM samples, on the other, on the other hand, you see dramatic differences in the blinking behavior. If the videos would be running anyway, you see that for the 18 nanometers and 9 nanometers and for the reference, a single labeled DNA origami, we see this blinking of Cy5 fluorophores throughout the 10-minute acquisition time.. But for the 6 and the 3 nanometer origami, we see at the beginning of the video, we see super fast blinking during the first few seconds, and then after 5 minutes almost there's no more blinking. The whole sample is really dead. There is no more blinking events. Yeah, and this appears partially, it even doesn't appear as blinking, it appears like flickering of the sample. But only through the first few milliseconds. If you analyze this, these storm videos or movies then, and you track, for example, here color-coded, 18 nanometer dark blue, 9 nanometer light blue, red 6 nanometer, orange 3 nanometer, the localization statistics with time in this 10 minutes when you record, when you detected all these localizations. You see that for the reference and for the 18 and 9 nanometer, we localize fluorophores throughout the 10 minutes with almost the same probability. But for the 6 nanometer separated and 3 nanometer interfluorophore distance, we see that the majority of localizations occurred during the first minutes, seconds. On the right-hand side, we then plotted for different individual origamis after which time 80% of the localizations have been detected. During the 10 minutes. And you see for the 18.9 at reference it's this bluish-grey time, but for the 6 and for the 3 nanometers, 80% of all localisations are detected within the first minute or so. I mean, now we did these D-STORM experiments, we do it typically with a few milliseconds, 5 milliseconds integration time per image, So we thought about, let's go to confocal single molecule microscope and investigate this detail, this in more detail here. And on the left-hand side, this is the reference structure, it shows this typical blinking of Cy5, and then the 18 nanometers more blinking because we have 4 fluorophores, 9 nanometers, and then for 6 and 3 nanometers, you end up with what we see here in this expanded views of the first seconds. There is a lot of blinking going on in the first few seconds. Super fast blinking. And we then hypothesized that by investigating this photo switching, what we call fingerprints, that we can retract or reveal information about the underlying interfluorophore distances. The problem now for super-resolution microscopy by D-STORM is that typically as you start your D-STORM experiment, and you irradiate the sample for a while with red light until you see homogeneous blinking, and then you start data acquisition. During the first few seconds of— during this first few seconds, however, most of the fluorophores will be photobleached for short interfluorophore distances, right? So you will miss these fluorophores and the localization probability will decrease dramatically. Can we really demonstrate that this is energy transfer. So this should be reflected then in the lifetime of the donor, yeah. And these are the lifetimes of this DNA origami for 18 and 9 nanometer distance, for 6 nanometers, and for 3 nanometers. Color-coded on the right-hand side for FLIM images of this DNA origami, and in the middle for a 3 nanometer origami where we recorded the trajectory, the intensity with time, and measured the fluorescence lifetime during the first few seconds, then here, and at the end. This means here at the end there is only one fluorophore that survived, and here we have 4 fluorophores or 3 fluorophores, I don't know, but at least then energy transfer is reflected directly in a quenched donor lifetime here of about 600-700 picoseconds. Well, what are the consequences? I already said that. The consequences for microscopy, for localization microscopy, is definitely that the localization probability will decrease. And for a highly densely labeled sample, it is difficult to bypass these problems because all the Cy5 fluorophores at the beginning of the experiment, they will stay in an on state and you have to push them or transfer them into an off state upon irradiation with light. Yeah, and this, you will lose therefore during the first seconds all the information encoded in the photo switching fingerprints during the first few seconds, which is this fast blinking of the sample. Last but not least, the last two slides here, we thought about can we translate or use this time-resolved photo switching fingerprint analysis now advantageously is basically to extract information from distance information, spatial resolution information from samples from cells. And for that we use genetic code expansion. So we introduced unnatural amino acids into GABA receptors in the plasma membrane and to the GABA subunit. So in order then to have the possibility to label only a monomer out of this pentameric arrangement, or into the alpha subunits, two unnatural amino acids, then we have a dimer in the pentamer in the plasma membrane, or the kinase receptor, we have introduced into this tetrameric membrane protein, an unnatural amino acid in each of the four monomeric units. Here, the dimer, the alpha-gamma, the dimer has a distance of about 5 to 5.5 0.5 nanometers between the two sites, and in the kinase about 6.8 nanometers. If we then do click labeling of these sites in the plasma membrane of cells and do DStORM images, we see no differences in the DStORM images. They all look identical. If we analyze from DStORM movies the off-times, we see that for one labeled site in the protein,. And for 2 label sites and 4, we see that the off times decrease. Of course, we have 1, 2, or 4 fluorophores that show blinking. Also, we have more on events for the tetrameric receptor than for the dimeric and the monomeric, so which tells us we achieved efficient labeling of these sites in this multiprotein complexes. And if we then have again a look on the localization statistics with time, we see that something is going on that the tetrameric glucA2 receptor at least. The same we saw for the 3 and 6 nanometer DNA origamis. Yeah, so, and also the lifetime data then incorporate that information about the number of fluorophores and the distance can be unraveled by analyzing these photo-switching fingerprints. So the blinking pattern during the first seconds and the lifetime of these samples, this allows you then to reveal information not only about the numbers and interfluorophore distance of also spatially unresolvable fluorophores. So with that, I'm at the end. I'm sorry, 5 minutes too long or 6. These are the people who did the work. I hope I mentioned most of them on the respective slides. And here are the people who give us the money.

Thank you very much for your attention. I'm happy to to answer questions. Thanks, Marcus. This was an excellent talk. And what we will do is we will quickly do the quiz, and then we take questions and answers. There are already quite a few questions in the chat window. OK. So if everyone can log in to the quiz, we'll just quickly do the quiz and, and I'm going to share my screen now. Yes, so we have quite a few people joining in for the quiz, about 26. Let's wait a couple of more seconds. So I think that's it. So we will start with the questions. And yeah, the link to the quiz is in the chat window. I have a sore throat today, so Everything might not be understandable, so let's start with the first question. And faster you answer, the more points you get. So what is the— what limits the spatial resolution achievable in super resolution microscopy? Labeling density, irradiation intensity, the wave nature of light. Time's up. I think this one was a bit tricky, but everyone who has answered A and C both, you will get the point. So this was a nice warm-up question.

Now we go to the next one.

I mean, we could discuss about intensity, but well, I think, uh, Jonas Ries has a good paper on it. So yeah, but let's, uh, skip that. And so we have egg and nano frog. Frosty. I think first 5 until Ash. I think all of first 10 are pretty close, like, so anyone can be a winner at this point. So by the way, people who don't know, the winner of the quiz gets a Foldscope. So it's from Manu Prakash lab in Stanford. It's pretty cool. So which are the best dyes for D-storm? Rhodamine-6G, Alexa Fluoro-4, 405, Alexa Fluoro-4. 647. I guess this is the easy one. Yep, let's see who is leading the race. I think Nano Frog has taken a lead. Nano Frog and Frosty. So GB was the fastest, so he gets extra points. So now we move on to the next question, and please continue to post your questions. We'll try to take all the questions that are posted in the chat window. So expansion microscopy requires functional groups to be linked into a hydrogel: amino groups, thiol groups, alcohol groups. Yes, if you I'm not a chemist, this might be a bit tricky, but if you listen to the presentation, this should be easy. Wow, so about 20 correct answers. And let's see who is leading the race. It's getting a bit close. So as you see, all our top 10 are very close, within 200 points. So Still, anyone can win, and Nanofrog is maintaining the lead. So this is second last question, and I thought this one— why does the post-expansion labeling in expansion microscopy achieve a lower linkage error? Because the label probe is not expanded, because the dyes are not destroyed during gelation, because antibodies appear smaller in expanded gels. Again, a tricky question and you might have to hear it. Also, let's see, 15 people. I guess it's the top 10 people who are getting all the correct answers. It seems NanoFrog is still leading the race. NanoFrog, GB, Frosty, and Jenny has come into the race now. So the last question, please answer fast. It's touch and go. Why does an expanded hydrogel shrink in photo switching buffer? Photo switching buffer destroys acrylamide groups. Hydrogel is negatively charged in addition of ions. Oh, sorry for this, uh, this got cut off because our answers only have limit of 200 characters or something. So sorry if this option, uh, second was not correctly— will, will, will reduce or something— was the full thing. Let's see who won the race. After this, we take on the questions. Jenny and NanoFrog, let's see who's the winner. Okay, so Jenny came out of nowhere in last two questions, so congrats Jenny! And, uh, If you email your email to either me or any of the organizers or Jess, we will send you a full scope. So thanks everyone for participating in the quiz. It was fun, very nice questions. And, and what we'll now do is we take on the, the questions and we'll start from the top, the first question. And if you want to ask the question yourself, please unmute and also show your video if you can and ask the question yourself. So I think the first on the list is Ignacio.

Ignacio, are you still there? Yeah, I'm here. Awesome. Go ahead. Yeah, I'm just reading your question. Yeah, well, I mean, our experience with, for example, with— to answer the question, the question is, how critical have you noticed the fixation? How critical have you noticed our fixation buffer to conserve cytoskeleton, for example? Do you know if it varies a lot depending on the problem of interest between ER and cytoskeleton, for example? Our experience overall with cytoskeleton, in particular with actin filaments, are not very good, to be honest. But this might be due to the labelling method. Phalloidin is not very efficiently binding. It's not useful for post-gelation labelling at all. And pre-labelling, we think that it often falls off of the actin skeleton. So there are no really convincing images about expanded actin filaments. I'm really sorry for that. I think we don't know why this is so exactly. But I mean, there has been a paper by Christian Egeling and colleagues in ChemBioChem, they investigated the expansion isotropy or the isotropic expansion of different organelles in cells, and came up and showed that different organelles expand differently. I think there was a second part also, Markus, how much time? And how much time can you image these with the imaging buffer used in? I mean, for these re-embedded samples, we use the standard photo switching buffer with 100 millimolar thiol, and there is no large difference between the imaging times. So this photo-switching buffer stabilizes the whole behavior dramatically, we have to say. But it's of course an additional step, this re-embedding in non-charged polyacrylamide gel.

And well, it also shrinks a little bit, 10, 20, 30% your gel, depending on the conditions.

Next question is from Lothar Schermelin. Lothar, are you still there? I'll load her. How well, from your experience, how well controlled is the expansion factor in XM? Is it homogeneous throughout the sample? Are there substructures that expand less well? Yes, it's not homogeneous throughout the whole cell. As I already said, depending from organelle to organelle, from multiprotein complex to multiprotein complex, We don't know yet why these structures expand differently, but it might be related to the efficiency of linking of this protein into the polyacrylamide gel, the linkage efficiency, and of course the intermolecular forces in multiprotein complexes that hold the proteins, for example, together. It really, it has to be investigated in detail and maybe using different buffers and recipes for different structures or cellulose structures. Of course, then next the question will immediately come up: so how do you know for an unknown structure then? How can you optimize the expansion protocol for it? I mean, there is no real answer to this question, but for example, I mean, as long as you know, for example, that a structure forms a circle and you see a circle after expansion, it's already quite a good sign. Next question is: Anna Malkowska, I was just wondering what are your thoughts on using STED instead of— Yes, STED has been used in different papers. Helge Evers, for example, published two, I think, two nice papers on expansion STED and also others. It's just again the limitations that microscopy has concerning dyes, but I mean, D-STORM is worse, worse, yeah. D-STORM, you have to use either these, a few of these Alexa dyes, or in the ideal case, the Alexa647 or Cy5, and you have to re-embed the sample. So, step microscopy, you do not have to re-embed, you just have to take care that your fluorophore survives that imaging. And then next one. Joana, Oliviera, you want to ask? Did you ever experience increased photobleaching? That's a very good question. Well, from time to time people in my lab report about pronounced photobleaching in these experiments. But I don't know. I really have no answer to this question. It has to be investigated in detail. We have no clue what's going on. Sometimes we really observe increased or more efficient or faster photobleaching in these expanded samples. That's still a point to be investigated. Then there is— isn't cryofixation— yes, there has just been a paper by Guichard in Nature Methods, a really spectacular paper about combining cryofixation with expansion microscopy. And I think at least for neurobiology applications, brain sections, brain slices, or whatever, this will be the way to go. Yeah, because I mean the preservation of ultrastructures like vesicles in brain slices and other delicate structures, this is really key for— I mean, this is the same fight electron microscopy had to fight over 50 years to preserve ultrastructure, and the same is true for ultra— for expansion microscopy now. How can you control in these new EM protocols that the expansion is isotropic? You cannot control it. You can play around with your crosslinker concentration and the ACX concentration, so these, these, these NHS acrydite functional groups to link the proteins into the gel, all this plays a role, but there is no common knowledge available in the moment about how to best preserve ultrastructure and achieve isotropic expansion. I'm sorry, unfortunately not.

It has to be optimized from sample to sample.

One more question popping up just now. Yeah, sorry, following this question, this question, um, I meant to ask about the procedure you do for following the expansion. Like, what do you measure to Oh, to see the expansion? Usually we always try to combine an unexpanded— so to image the same cell before and after expansion. It's really a pain in the neck, especially for the students doing the experiments, to find the same cell. But I mean, most reviewers want to see it at the end of the day. Yeah, I also try to do that, but it's very difficult. It's really difficult, it's really limiting. Yeah, but there have been other strategies have been used. So, but this is the definite proof of isotropic expansion that you really didn't introduce any artifacts during your expansion.

Because we have one last question coming from Alex. Alex, you are still around?

You want to ask the question yourself?

Sure, thank you.

Yes, I'm still here. Hi, Marcus.

Hi, Alex.

It was great, great pleasure to hear you again. Now, my question was, you know, with biomolecular condensates, there's a lot of interest in complex condensates which usually form multiple shells and resolving them. I know there have been some attempts to visualize them, but they're quite hard. So my question is twofold. Have you tried expansion microscopy to visualize shells? And what sorts of fixation— no, you haven't, I guess. No. And how would you go about fixing them to preserve those layers?

Because they're different from all the structures you presented so far. I mean, first of all, I would start with standard fixation protocol using formaldehyde, glutaraldehyde. This is always how we start. And then You have to optimize.

Yeah.

Sorry, I can't say anything more about that.

Eva has a hard time always to optimize the fixation conditions for all these samples. Yeah, yeah. No, I guess that'll be probably my guess as well, because we also know they're so sensitive and delicate as well, even Yeah, exactly. But my hope is now with prior fixation.

Yeah, that'll be the way to go, I guess.

Yeah, absolutely. Thank you. You're welcome. Sorry, Marcus, I will just take one last question which popped in, which is from Johanna. She wants tips on how to attach gel to the slide during imaging. Say it again, please. So she has troubles attaching the gel to the slide during imaging.

Do you have any tips for her? Oh yeah, I mean, we silanize the covers cover slips, for example. That helps. Yeah. I mean, there are protocols in these two Nature Comp papers about this microtubule expansion D-STORM and the synaptoneural complex. If you have a look into the protocol, you can buy, you can purchase a few silanes from different companies and just treat your cover slips.

This helps a Thanks, Marcus, for the excellent talk. This is a bit awkward moment to say goodbye to our speakers. Stephanie, for the last 2 years, has been planning a lot how we can send a small memento souvenir to our speakers, but she hasn't converged on one. But hopefully, we can send you some coffee from your nearest cafe so that you re-energize.

But thanks again for coming to the talk.

It was excellent. Thank you very much for inviting me. So maybe we should target Georgina now for souvenirs for the speakers. I'll put my thinking cap on. Bye everyone, thanks for coming. And yeah, so this is the awkward moment. All we can do is just clap. Me, Alex, and Georgina are left.

But yeah, that was quite good. All right.

Okay then.