Stefan W. Hell (2015) - Optical Microscopy: the Resolution Revolution

Thank you much for inviting me here. I think everyone of us is familiar with the saying "a picture's worth a thousands words" or "seeing is believing," and this not only applies to our daily lives, it also applies to the sciences. It's probably not a coincidence that the beginning of the natural sciences as we know them today very much coincides in time with the invention of the light microscope. Because of the light microscope, mankind was able to see for the first time that every living being consists of cells as basic units of structure and function and many organisms of course were discovered with the light microscope. However, we also learned in school that the resolution of a light microscope is fundamentally limited to about half the wavelengths of light. And if you want to see smaller details we have to resort to electron microscopy. There's no doubt about the fact that electron microscopy has achieved a much higher spatial resolution, in a number of cases even down to the size of a molecule. So the question comes up, why do we still care for the light microscope and its spatial resolution? The answer is given in the next slide where I made a little experiment. I counted the number of studies here that used a light microscope and those that used an electron microscope in this basic medicine journal. Now you see which of the two won. It is a fact that light microscopy still is the most widely applied microscopy technique in the life sciences and that for two good reason. The first reason is that light microscopy is the only way of which you can look into the inner part of a cell and material for cell with minimal invasion or even into living tissue with minimal invasion. It doesn't work with an electron microscope and there is another reason. Of course there are many thousand types of biomolecules proteins in the cell and you want to see them specifically. You want to highlight a certain type of protein, you want to know what it does, what the protein interacts with and so on and so you have to do something to highlight it. It is possible relatively easily in a light microscope by doing fluorescence contrast modalities. Meaning that to the molecule of interest we attach a florescent molecule and this florescence molecule has a very basic feature. It has energy ground state and it has an energy excited state. And of course if you put the right beam of light onto that molecule the molecule will absorb a photon, in this case a green photon. Then would be raised to highlighted energy state and there it wiggles, its atoms wiggle a bit. And then eventually the molecule comes down emitted a fluorescent photon. Because some of the energy is lost in the wiggling of the atoms, the florescent photons are red-shifted in wavelength. This makes florescence microscopy enormously sensitive because you can easily separate the florescent light out of the illumination light. In fact, florescence microscopy can be so sensitive that you can detect even single molecules, as has been discovered by my co-laureate W. Moerner and also by others. Then of course this means that you can really see individual molecules that really are in a body of a cell for example. However if there is a second molecule, a third molecule, a fourth molecule coming close to that first molecule, close in distance of 200 nanometers, then you cannot tell them apart. You see resolution is about telling things apart. It must not be confused with sensitivity. And therefore it's clear that a lot of your information is lost actually at the nanometre scale simply because we cannot tell things apart. And if we manage to overcome this barrier, this would have an impact in the life sciences and possibly beyond. In order to understand how we can overcome this barrier, of course we have to understand where it comes from. There are several ways of explaining it but the most profound way of explaining a diffraction barrier is the simplest one fortunately. Well we can say that the most basic element of a microscope is the objective lens, and role of the objective lens and nothing but to concentrate the light in space, to focus the light. However, because light propagates as a wave it's not possible for the lens to concentrate the light in a single point, a focal point. Rather the light would be smeared out on this diffraction blob here of about 200 nanometer extent in the focal plane and 500 nanometer along the optic axis. And so there's a major consequence out of this. If there are several features falling within this range, then they will be flooded at the same time with light, there's no way out. And hence they will also produce light, back scatter light or fluoresce at the same time. So if you place the detector here, the detector will not be able to tell the single features apart because they produce light basically at the same time. Even if you collect this light, say florescent light with a lens and image it onto the detection plane, even then we couldn't tell them apart. Why? Because each of these little features, say molecules, would produce a blob of diffracted light, focused light here as well and you see the neighboring one would have another one and the neighboring one again and again and again. So, these blobs of light, they will overlap in space and no detector would be able to tell them apart. Be it a photo-multiplier of the eye or even a pixelated detector such as a camera. Now the person who realized that diffraction poses a severe problem was this man, Ernst Abbe, who lived at the end of 19th century and he coined this diffraction barrier equation that is named after him. It's saying in order to be separable in a light focusing microscope, two features of the same kind have to be further away than the wavelengths divided by the numeric aperture of the objective lens. This equation can be found basically in any textbook of physics or optics, but also in textbooks of cell biology because of the enormous relevance of light microscopy to this field. It can also be found on a memorial which was erected in Ernst Abbe's honor in Jena, in Germany where he lived and worked, and there it is written in stone. This is what scientists believe throughout many centuries that it is the practical limit of resolution. And it was the limit. When I was a student, at the end of the eighties in Heidelberg, student of physics, this was the maximum resolution that you could achieve for example if you wanted to look into the cytoskeleton structure of a cell. But now we can do much better. As the development of the super resolution microscope, in this case STED microscopy, has shown that you can overcome to diffraction barrier. There is so to speak physics or physio-chemistry in this world, that gets you sharper images. This of course is a substantial addition in the field. Now, I believe as any substantial addition, development or discovery, it has a personal story. A story behind the story and this is definitely the case in my case. Something about my background: I was born in Romania in the western part of Romania. And it was communist Romania. At the age of thirteen I realized that in this country I wouldn't have a great future of course and I convinced my parents to immigrate to West Germany, which they finally managed. Two years later actually at the age of fifteen, we emigrated to west Germany where I got my high school leaving certificate and eventually studied physics in Heidelberg, which I always wanted to do because I was fascinated with physics. Now like most physics students of course I was fascinated by the basic physics, by particle physics and all this fascinating developments that took place at the time. However, at the time I had to decide for my PhD thesis, this is what I looked like back then, believe it or not. By the time I had to make the decision, I must admit I lost courage. No surprise because my parents who had to struggle settling in the west as you can imagine. My father was threatened with unemployment, my mother was diagnosed with a life threatening disease. I said to myself: it's better not to get unemployed as well because most people were saying, oh with the particle physics you'll get unemployed most likely, and so on. I opted for something that is very applied and I signed up with a thesis advisor that had set up a start-up company, Heidelberg Instruments in Heidelberg. The start-up company investigated confocal microscopy and the use of confocal microscopes for the inspection of microlithography, so computer chips. I felt okay again. I would get a job with him or IBM and so on and I should do that. However when I did that after a year I was totally unhappy because this was not the kind of the physics that I wanted to do actually. This was too technical. There was no physical phenomenon in it. Just polarization and diffraction. Microscopy was the physics of the 19th century. So I had two options. Misery that would have been the first option and the second option to think about something fundamental. Is there still a problem left in light microscopy that is interesting and where a change could be possible? Then I realized, breaking the diffraction barrier, that would be cool. That's an interesting problem. That's a problem to work on. I thought to myself, could that be possible and I realized at some point, yes. My thinking was the following. Where does diffraction barrier was coined in 1873 and at that time, we had 1988, 1990. So much physics came up in the 20th century. Quantum mechanics was invented, molecules, spectroscopy, there must be at least one phenomenon that gets beyond the barrier and gets sharper images. There should be something although in the beginning I didn't know what it was. But at some point the idea materialized and I put it down in writing. I knew that it wouldn't work just by changing the way the light is focused. No way. But maybe if one exploits the properties of the fluorophores, the states of the fluorophores, spectroscopic changes and transitions in the fluorophores, maybe that's an option. Or maybe there's quantum phenomenon, a quantum optical phenomenon, that gets beyond the diffraction barrier. That was the strategy, the philosophy I had in mind. You see I put here far-field light microscopy. Why far-field? Because in those days there was already sub-diffraction optical microscope in existence. This was the near-field optical microscope that overcame the diffraction barrier by confining the light-specimen interaction with the tiny tip down to 30 nanometers. But that's not what I was attracted to because that's confined to specimen surfaces. And to me honestly, it did look like a microscope, like a light microscope. The assembly was getting a ton of a microscope. I wanted to overcome the diffraction barrier in a microscope that looks like a microscope and operates like a light microscope. But this was very difficult because it was a very unusual idea and when I tried to get institutional support in this country, in Germany, I failed. I was fortunate that there was a person in Turku, in Finland, who gave me some space and some freedom and a postdoctoral fellowship from the Finish Academy to work on this problem. Eventually I went to Turku, to Finland, a place I've never heard of before, very reluctantly and there of course I took my philosophy with me. About six weeks after I arrived there with this idea in mind, of course I started to screen text books, text book. Could there be a phenomenon that could be used? I opened this textbook on a Saturday morning hoping to find something in the quantum statistics of light and so on. And I opened this page and I came across the phenomenon of stimulated emission. This phenomenon I was familiar with from actually from my first year of physics studies and all of a sudden I was electrified. I felt not this could be a phenomenon, at least one phenomenon, that gets me beyond the diffraction barrier in fluorescence microscopy. This was the basic idea of course that started things going. Now, why was I so electrified? Let's get back to the situation. We have here the lens, we have here this blob of light and as a result of diffraction all the features here will be flooded at the same time with light. We cannot change that. All the molecules will produce light here at the same time because they're flooded at the same time with light. But maybe there's a solution there. What if we manage to make sure that not all the molecules that are flooded here with light, are in the end capable of producing light at a detector? So we keep some molecules in a state in which they are not able to produce light here. We should be able to separate those that can produce light from those that can't. If we manage to keep these molecules in a state in which they are dark, not sending light back, then we should be able to separate the bright ones from the dark ones. You see the idea has not changed the way the light is focused, but changed the state of the molecules transiently such that you produce two classes of molecules, bright ones and dark ones. Now you would say, this is the idea, use a dark state. Are there such things as dark states in a dye. Look at the basic energy diagram. We have here the ground state, we have the excited state. Of course the bright state is the excited state because the bright state produces the florescence. But the ground state of course doesn't produce florescence. The ground state is a dark state. Here we go. The basic states of the alpha of a dye. Give me a dark state that is the ground state. I'll guess what this phenomenon of stimulated emission does? Nothing but making dark state molecules. Stimulated emission pushes molecules from the bright state back down to the dark state and then of course we have these two classes of molecules, bright and dark, and they should allow us to get beyond on the diffraction barrier. And this is why I was electrified. This is the basic idea behind the STED microscope. We have a lens, we have a sample of course, we have a detector, and of course we have a green light for turning molecules on from the dark state to the bright state or the fluorescent state. Of course the green light will produce a blob of diffracted light in here and all the molecules will fluoresce as a result normally. You also have this red light and the red light is nothing but used stimulated emissions so this transitions back down to the ground state, so the job of the red light is to make dark state molecules. Make molecules that are dark. The condition for doing that is that you have the right wavelengths, it should be a red shifted wavelengths and that the beam has to be bright enough. Why? Because we have to ensure there is always a red photon at the molecule that is able to take the molecule down to the ground state. If the red beam is weak of course then, this phenomenon would not take place. This is displayed in here. So if you guarantee a certain intensity at the molecule, then it can turn the molecule off because that beam of light would instantly make the molecule dark, so prevent the molecule from occupying the on state, the excited state. We don't want to turn all molecules off. So what we do is actually we phase modulate, we change the beam such that it produces a ring, not just a spot of light and the role of the ring is to turn off the molecules in here. Let's assume we want to see only the single here from that the inner part of the object. This thread. So what do we do? We have to turn those molecules off and the way of doing that is just to make that red beam bright enough so that those molecules in here have to face an intensity that is higher than the threshold. And this way of course we turn those molecules off and only those molecules in the centre are left. You see we are there. Because now we can separate features that are closer than the diffraction barrier. We can separate the strands because the molecules that the belong to them, they are forced to emit time-sequentially because we played this on-off game and only these are capable of producing light by the detector. I'm removing now the donut just to make it clear, that this is a state game that we're playing, not a way of changing the way the light is, but the state game. You see this is a diffraction zone. These molecules in here, these molecules are forced to stay all the time here in the ground state, because as soon as they get up they are pushed back down instantly. And only those molecules in the centre are allowed to assume the on state. This is where we play two different states within the diffraction zone, we can tell the features apart. I'm putting the donut back on. This was the idea. You may say, oh this sounds logical, but believe it or not when I published this in 94 in Turku, hardly anyone believed it. In fact I applied to many places. To the States, to Germany, to Europe of course. Maybe 25 applications I sent out with this idea in mind, but to no avail. I didn't have the right pedigree, not from Stanford, Harvard, Bell Labs, anything. Turku was a place no one has heard of. It wasn't easy, so it was difficult to survive, both financially and of course academically. And not having a strong social background, believe me, it wasn't easy. But I worked on a fascinating problem and I was happier than in my thesis. Why? Because I wanted to do this. This was a subject that was fascinating. I was the only one who really believed in going far beyond the diffraction barrier and it was cool doing that. Of course at some point somebody has to rescue you and so I was fortunate that the Max Planck Institute for Biophysical Chemistry became aware of me and gave me a five years probation period actually to prove that I'm on the right track. So it was. After some development it was clear that you can overcome the diffraction barrier. This is the standard resolution, confocal. This is a STED resolution. And you see now the gain in spatial resolution is very obvious. Now we can see this molecular sub-units here of this nuclear pore complex in the eight-fold symmetry but in the standard case, confocal, which used to be the best of course, there's no way of seeing them. Of course if you have this high spatial resolution you get new information. I would like to give you just one example or two. Say virology, this is going to be important because viruses are between 30 and 150 nanometers in diameters, are smaller than the diffraction barrier. If you want to image say protein distributions on the virus with a normal microscope you just a blob of something smeared out, but if you do the sub-diffraction imaging with STED in this case you see that they form patterns dependent in this case on the maturation state of the virus. What has been found out here is that the mature particles, those that can infect the next cell, they have their Env proteins in a single place. So this is something to show that you can get inside of course with the high spatial resolution. Surprise, surprise. Well, I think the strengths of a light-focusing microscope is that you can of course image living cells. And living cell imaging at the extreme is of course to go straight to a living system. This is about 20 microns of the upper molecular layer of the brain of a transgenic mouse where some of the neurons expressed a yellow fluorescing protein. Now you see we can record the images. See these dendritic spines with much higher clarity than before, subtle movements. And the seeing of these movements requires of course the high resolution because otherwise it would smear out. Of course there's more work to be done. There's no doubt about it. In fact, however, I have all reasons to believe that in the end we will be able to see the distribution of molecules here at a synapse and perhaps give a visual cue to the mouse and see what's going on in the brain at the molecular level. With that I'm coming back to the resolution question. Being a physicist of course I want to know what is the spatial resolution that we can get now with this method? What is the end of it? What's the conceptual limit? What's the practical limit? Now the conceptual limit can be answered very easily. Of course if we increase the brightness of this beam, then more and more molecules will be turned off on average so the region in which the molecules are allowed to assume the on state would become smaller and smaller. And in fact the size here would depend on the ratio between the brightness of this donut or ring at the crest and the threshold intensity which is a characteristic of the molecule of course, and this means now that we have to modify Abbe‘s equation so that d is not diffraction limited but this ratio of course is here in the denominator. You see if this becomes large then it goes down to zero, meaning that the conceptual limit now is the size of the molecule. And this is not a surprise because we separate using molecules, and of course then the conceptual limit has to be the molecule, the size of the molecule. We can have conceptual molecular scale resolution. Now the practical resolution depends of course on how well you can play this game, this on-off game. For getting molecule for STED we are at 20 to about 40 nanometers depending on the molecule, depending on the molecular environment. Just to make sure that this concept can be worked out and developed further, if you use a fluorophore so to speak, that allow themselves to be turned on and off as many times as you want, for example, these color centers in diamond in this case, you can show very extreme resolution potential. For example here 4.2 down to 2.4 nanometers, so given the fact that this is wavelength of 775, that's a small fraction of it. Now this may look just like a proof of principal experiment. But there is also applications for this because this color centers, the charged nitrogen vacancies are very popular with physicists because they can be used for example as qubits for quantum information at room temperature or for magnetic sensing at the nano-scale, so there's more than life science. It's not just the life sciences. With that I'm coming back to the basic concept again. I said the name of the game that you play is on-off. We have the offset here, we have onset here, this is a more basic state, and this is the most basic transition imaginable. Going up by light, going down by light. But of course, once you realize this is about on-off, not just about making simulated emission in here, then you know, maybe there sorts of other states in the molecule that allow you to do the same game. For example, you can push the molecule to a meta state with dark state, like a triplet state. Or if you look into chemistry books then you will find such thing as cis-trans photoisomerisation. You can relocate the atoms so to speak and play on-off in this way. And now you must say, why do you care for this transitions if you have a very fundamental one here that applies to any fluorophore so to speak? Well, there is a very strong reason. Don't forget: we separate by transiently putting the molecules into a different state. On versus off. Then this on then this off. Then of course the lifetime of the state matters and it's obvious that the longer the lifetime, the easier it is to create a difference in states. But here the state disappears after nanoseconds. Here lasts longer, microseconds, milliseconds. Here we have something that is optical bistable and if it's bistable there's more time. Meaning that we need less light per unit or less photons fuel force by using time to do this kind of game, meaning that this threshold intensity goes down, meaning that the intensity required of course goes down so you can break the diffraction barrier at lower light levels if you do for example the cis-trans isomerisation. This was of course fascinating because you can do this also in switchable fluorescent proteins. I would like to show you one example where this concept has worked out. I couldn't have it called STED anymore because there is stimulated emission in it. If you did little light, as is the case here, then you could produce many rings or holes or donuts or whatever, and this has been done in here. You can parallelise the whole thing and then recording a large area becomes very quick and even at relatively low light levels, because this is a metastable transition with long-lived states. Here actually this living cell, these filaments, were recorded with a 100,000 donuts in less than two seconds. I'm just showing development goes on, not only that it's the molecular transition that is decisive. That in turn is the performance of the system. With that I'm coming back to the last question I'm putting up now. What does it take to get the best resolution? Let's assume you would have asked this question in the 20th century. What have been the answer? Well, good lenses, why? Because the separation of features is done by the focusing of light and then if you produce small spots you can separate better. Producing a spot of light here, as sharper spots would be producing a spot of light here, that determines the resolution. But that is of course diffraction limited. If you have several features falling within that spot of light, we cannot tell them apart. What was the solution to this problem? The solution was not to separate just by the phenomenon of focusing but to separate by a state transition. Prepare the molecules in two different states within this diffraction zone and then you can tell them apart. I'll give you a few examples of state transitions that can be used. And so in this concept STED and is derivatives, we use a pattern of light so to speak that determines where the molecules are on here and where they're off to predetermine actual position and play this on-off game. You control actually the states and the position of the states using a pattern of light. How does it compare with the other seminal concept that was done first and suggested by Eric Betzig, based on the discovery that you can detect individual molecules by W. Moerner? Well, in this case, the on-off game is played individually. Individual per each molecule and this way of course makes the features different. This is a fundamental difference. It's really a fundamental difference because now you can play the on-off game on a single molecule basis which has a number of advantages. If you do it stochastically, if you do it stochastically in space, then you don't know where the molecule is located. What's the solution? Well you have to resort to states that produce many photons in a bunch, here at a camera, and a pixelated detector then gives you the coordinates and eventually you will know the position as well. Of course, you need always many photons to either find out the coordinate or define the coordinate because one photon goes astray here, here, here or there within the diffraction zone. Only with many photons you can deal with coordinates. But that's just the coordinate which is critical and needed but for the separation which is needed to tell features apart you need the state transition. As a matter of fact, in order to separate features now what you need is a state transition which in the simplest case, on and off. You flood everything with light but only one molecule is allowed is to emit or only a group of molecules at a certain location is allowed to emit and this is how you overcome the diffraction barrier. It's the state transitions that is a critical element that gets us super resolution. It's not a coincidence that it was fluorescence and non-fluorescence that allowed us to overcome the diffraction barrier because that's the easiest to play with, to turn on and off. In the 20th century it was the lenses, but now what determines the resolution, what determines performance, it's the molecules. So it has to kind of become a chemistry topic. You want to push this you, have to push the molecules here. The story goes on, because you can imagine not only on-off in fluorescence but you can also imagine absorption, non-absorption, scattering, non-scattering, spin ups, spin down, binding, non-binding. There are different ways of playing out this idea of transiently placing the molecules into different states within the diffraction zone. With that I'm coming to almost last slide. It's very easy now to expand Abbe‘s equation. We just have to put in this square root factor and this square root factor tells us that we can go down to the molecular scale. What does it stand for? What does it stand for? It stands for a state transition obviously. So what I'm saying is light microscopy resolution is not just about waves. It's about waves and states, and if you put in the states into your equation literally then you change the game. And this is what changed the game in the resolution in a focusing light microscope. With that I'm acknowledging the people who have worked and contributed to this development in my lab. We had a lot of fun even at a time when we were about the only ones who worked in this field. But one little point I would like to make at the very end. Well, it wasn't this guy who started the development, don't forget it. It was this guy and actually it's him who deserves the prize. Thank you very much.