Super-Resolution: Overview and Stimulated Emission Depletion (STED) Microscopy

00:00:11.22 I'm Stefan Hell. In the current lecture, I'm going to explain the principles
00:00:15.17 of super-resolution fluorescence microscopy with emphasis
00:00:18.29 on STED microscopy. Now before we've been told that resolution
00:00:23.08 of a light focusing microscope is fundamentally limited by diffraction
00:00:26.18 to about 200 nm. So if we're having features residing at a closer
00:00:31.13 distance than 200 nm, half the wavelength of light, it's not really
00:00:35.22 possible to tell them apart. Now this is the reason why electron
00:00:40.08 microscopy was invented, and there's no doubt about the fact
00:00:42.18 that as a result of its higher spatial resolution, electron microscopy
00:00:46.11 has allowed us to make many discoveries. However, it's also
00:00:50.20 clear that if you want to look inside the cell in three dimensions, especially
00:00:54.16 inside a living cell or living tissue in three dimensions, there's no
00:00:59.05 alternative to using focus visible light. Now for that reason, of course,
00:01:03.13 a light microscope that would overcome the diffraction barrier
00:01:07.09 and image with a spatial resolution of an electron microscope, would be
00:01:11.17 very, very important. Especially for the life sciences, but not only for
00:01:15.24 the life sciences. Now, why have people thought that the resolution of a light
00:01:21.20 focusing microscope has come to an end? The reason for that
00:01:25.03 can be put into very simple words. The main part of a light microscope
00:01:29.19 is the objective lens. And the role of this objective lens is to focus
00:01:34.13 the light down to a point. But because light propagates as a wav,
00:01:38.25 it's not possible for the lens to concentrate all of the light into a single
00:01:42.19 point. Rather, the light will be smeared out, forming a blob that is at least about
00:01:48.06 200nm across the focal plane. And as a result, all the features falling
00:01:54.24 within that spot will be flooded at the same time with light. In fluorescence
00:01:59.18 microscopy, it will be excitation light. And hence, of course, all the features
00:02:04.02 will give off signal and be collected by the lens, and impossible
00:02:07.11 to tell these features apart. Now the person who has realized this diffraction
00:02:12.12 resolution barrier, Ernst Abbe, formulated this problem in an equation.
00:02:16.23 Two features of the same kind, in order to be separate by a light microscope,
00:02:23.13 have to be further away than the distance, d, given by the wavelength
00:02:27.02 of light, divided by twice the so-called numerical aperture of the objective
00:02:31.22 lens. And this value amounts to at least about 200 nm, up to 300, 400, 500 nm.
00:02:38.12 And this is the reason why people have thought that a light microscope
00:02:42.22 will not be able to do any better than that. Now in order to explain to you
00:02:48.06 how we can overcome this diffraction barrier, let's have a look again
00:02:52.15 at the problem of focusing. Now, as I said, the lens will not
00:02:58.04 be able to make a smaller spot than that. If it could make a smaller spot,
00:03:02.27 of course we could concentrate the light in here and in here, and you could
00:03:06.03 at least discern the features that are here at the vertical distance. But this is
00:03:11.21 not possible. All the molecules that reside within this 200nm
00:03:15.27 zone, will be flooded at the same time with excitation light.
00:03:19.10 Hence, give up fluorescence at the same time. This is the energy diagram
00:03:24.10 of fluorophore excitation fluorescence, and hence the signal that
00:03:28.09 is generated in here, more or less at the same time, will be confounded
00:03:32.18 by any detector. And any separation is not possible. Now it's rather
00:03:37.19 clear, I would say, that the diffraction problem basically persists only within
00:03:43.04 this typical range. If it manages to sort out the features separate
00:03:48.03 for example, this strand from that strand, assuming these are microtubules.
00:03:51.22 Then of course, separating the rest of the object would be trivial. So
00:03:56.21 it's fully okay to just to concentrate on this critical 200nm zone.
00:04:01.21 Once we've sorted out the problem in here and managed to separate
00:04:04.19 the features in here, we are truly done. So, let's concentrate just on
00:04:10.08 this zone and ask how can we tell the features in here apart.
00:04:15.25 Now if the problem really is that all of the features are flooded
00:04:20.29 at the same time with excitation light, and hence give off light
00:04:23.01 at the same time, a solution to the problem is to make sure that
00:04:27.02 not all the features that are flooded with excitation light are in the end
00:04:33.00 capable of emitting. And this is exactly what we do in a STED
00:04:37.27 microscope. In the STED microscope, we not only use a beam for exciting molecules,
00:04:42.23 one that is focused into this 200 nm range, but we also use a beam
00:04:46.22 of light that is typically shaped as a donut. And the role of this
00:04:51.11 beam is to keep the molecules dark, turn them off, if you will.
00:04:55.27 So, how can we do that? How can a beam of light keep molecules
00:05:00.27 dark? Keep molecules off? Well, in fluorescence microscopy,
00:05:04.18 or in a fluorescent molecule, this is possible by using photons
00:05:09.20 that don't have an energy that is high enough to excite molecules
00:05:13.16 but have much lower energy. And if the photon energy fits the fluorescence
00:05:17.25 data, the energy gap between the fluorescent state and the ground state,
00:05:20.24 those photons of course are capable of sending molecules back down
00:05:26.02 to the ground state instantly, by taking away the majority of the
00:05:29.18 energy in this red-shifted beam. So the beam here is red-shifted,
00:05:37.21 because this is lower photon energy. And the role of the beam is just to silence
00:05:42.21 the molecules. How can we make sure that the molecules remain
00:05:47.15 silent? Well, first of all, as I explained, you have to use the right
00:05:51.08 wavelengths. But that's not all, you also have to make sure that
00:05:54.10 there is enough red photons in the red-shifted beam. And why?
00:06:00.22 Because if there is enough red photons, we can be sure that
00:06:03.22 once a molecule gets excited, there's always a red photon out there
00:06:06.26 that will instantly kick the molecule down to the ground state.
00:06:10.00 And this is actually shown here in this graph. Here, we have plotted the
00:06:14.21 probability to be on or to be fluorescent, as a function of the intensity
00:06:20.09 of the red beam. And you see that typically, after a certain threshold,
00:06:24.25 this threshold intensity Is, the molecule is not capable of emitting
00:06:29.14 any more. Because there's always a red photon, so to speak, in the area
00:06:34.03 that will kick the molecule down to the ground state. And hence,
00:06:37.13 shut off the ability of the molecule that are up in the fluorescent state.
00:06:41.16 So those molecules in here cannot assume the fluorescent state,
00:06:45.16 they have to reside in the ground state. They have to stay dark.
00:06:49.29 So, you've seen that we've shaped the STED beam into a donut
00:06:55.07 pattern. Why that? Because you don't want to shut off all the molecules,
00:06:59.00 you want to see some molecules. You want to discern features like
00:07:01.20 you want to discern this microtubule from that microtubule.
00:07:03.25 So you want to keep an area where the molecules are still capable of emitting.
00:07:08.21 Now this area is the area in which the intensity of the donut is
00:07:13.16 smaller than the threshold intensity, Is. So here the molecule is still
00:07:18.05 allowed to emit. Despite the fact that all of them are colored with excitation
00:07:24.09 light, only these are allowed the emit. The rest are silenced.
00:07:27.26 This is the basic principle of STED microscopy. Only a
00:07:31.07 subset of molecules is capable of emitting within this 200nm
00:07:34.27 zone. Namely a subset that is located at the specific location
00:07:39.25 in space that is predetermined by the shape, by the pattern, of
00:07:45.22 the STED beam. Okay, now the question is, once we have signal from here
00:07:52.29 generated, we only have one intensity signal, but we want to see also
00:07:55.27 the rest of the features. What do we do? Well, the simple answer
00:08:00.16 is we move the beam across the 200nm zone. And now these two molecules are allowed
00:08:05.13 to emit. Of course we can't separate them because they produce
00:08:07.29 signal at the same time, but they can clearly be separated from the molecules
00:08:11.16 on this strand. Why? Because these ones emit when these ones are off.
00:08:16.08 And so we can separate them because they are sequential in time.
00:08:19.18 We always know where the signal comes from, because the
00:08:23.19 coordinates of emission are preset by the red beam. By the position
00:08:27.23 of this minimum of intensity of the red beam, of the donut.
00:08:31.19 And so we scan that beam across that 200nm zone. And so
00:08:37.00 it can generate a signal from very tiny regions, and so can
00:08:41.03 separate these features, although they are very close together.
00:08:46.02 Now the strengths of such a concept of STED microscopy that
00:08:49.12 determines where the beam of light, where the molecules are allowed to emit,
00:08:52.10 where they're not allowed to emit, is that you can tune the spatial
00:08:55.25 resolution just by tuning the spatial extent in which emission is possible.
00:09:00.22 For example here, we have made a large area in which emission is possible.
00:09:05.21 How can this be done? Well, the intensity of this donut is now slightly weaker.
00:09:11.03 And so, this area which is too weak, and beyond or below the
00:09:16.10 threshold has now become large, and the resolution is not so good.
00:09:19.24 So, now the resolution is higher and you can make in principle,
00:09:23.15 you can make it very high. Such that only one molecule fits into
00:09:27.29 this very tiny little area. So in principle, down to the size of a molecule.
00:09:31.12 If you get a few photons here, we already know where it is, and we know there is
00:09:35.19 a molecule, we can have very high spatial resolution, in principle.
00:09:38.23 And we can of course separate this molecule from the adjacent molecule
00:09:42.17 just by moving that little donut beam, such that adjacent molecules
00:09:48.26 are capable of emitting only sequentially in time. And this gives us
00:09:54.04 the separation. So in any case, if you do it like that. Defining with a beam
00:10:00.16 of light where the molecule's allowed to emit or where they're not allowed
00:10:04.19 to emit, in that case, the resolution cannot be given by Abbe's
00:10:09.06 equation anymore, like this expression. We have to use a modified
00:10:13.10 expression, which is actually shown in here. And in this modified
00:10:17.10 expression, this threshold intensity, Is, will have a critical role
00:10:21.25 you see it in here. Moreover, the brightness of the beam, the total
00:10:26.02 brightness of the beam will also have a critical role. Because the
00:10:29.10 larger it is, the intensity I, the higher the resolution will get because of the
00:10:33.26 beam will become very small. And now you see in this expression
00:10:37.14 we have here, this ratio of I over Is. It becomes very large, d becomes
00:10:43.15 very small and hence, the resolution very very large. And so we can
00:10:47.20 tune the spatial resolution. Notice that the rest of the sample,
00:10:51.29 the rest of the features in the 200 nm zone is simply obtained by scanning the beam
00:10:57.19 further. We always know where the signal comes from because the position
00:11:01.18 of the on-state, of the emissive state, is determined by the beam that
00:11:06.26 we put on the sample. So the photons that go into the sample.
00:11:10.10 And so we can separate features and disentangle the features within this
00:11:16.21 critical 200nm zone. Okay, so once we have done our job. We've seen that
00:11:22.14 200nm zone is done for that region. So what happens with the rest?
00:11:25.15 One option is of course to move further with the beam and go like that.
00:11:30.11 Like that. Like that. And do the rest of the on/off game, so to speak.
00:11:36.03 Also on the rest of the sample. Of course, one could imagine having many
00:11:40.12 donuts or stripes, like structured illumination and so on. But
00:11:44.10 the simplest way of doing it, of course, is just move that excitation
00:11:49.01 and co-aligned STED beam across the specimen. And if you wonder
00:11:54.17 how this is done technically, I'm just showing you a little view graph.
00:11:58.06 Here, you see a typical arrangement of a STED microscope. This is the objective lens.
00:12:02.10 And of course we have here the excitation beam, producing green light,
00:12:06.05 that's focused into this area, obviously. But it's overlaid with the red beam,
00:12:11.28 that turns the molecules off. And in order to produce a donut, the red beam
00:12:19.26 passes a little piece of glass that has a varying thickness. It's variation
00:12:24.22 of thickness produces this donut shape. So it's technically red
00:12:27.29 and it's simple to produce donuts. And the net result is for certain intensities,
00:12:32.20 only molecules residing in this small region are in the end capable
00:12:37.13 of emitting. Because in this area here, where the molecules are shown
00:12:42.00 in dark color, they are not allowed to assume the fluorescent state
00:12:46.16 and hence, they are switched off. You could flood them with excitation light
00:12:50.17 but it would stay off. And this concept of course has led to higher spatial
00:12:55.25 resolution. On the left hand side, you see a confocal recording
00:13:00.07 of individualized molecules on a surface. On the right hand side,
00:13:04.24 you see the same recording, or the same sample recorded, but now with
00:13:09.17 STED microscopy. Here, we can clearly separate the molecules.
00:13:13.11 Why? At the time this one was emitting, this one was turned off.
00:13:17.07 And the time this one was emitting, this one was turned off.
00:13:20.04 Here, on the contrary, the molecules were allowed to emit
00:13:24.16 simultaneously and this is why we could not separate them.
00:13:26.29 Now this on/off switching by exciting, to turn fluorescence on
00:13:31.26 or de-excite and turn the fluorescence off, is done just by using
00:13:35.08 basic molecule transitions of a dye. Excitation and de-excitation
00:13:39.21 or de-stimulation are very basic fundamental phenomenon that are
00:13:44.21 in basically any fluorophore. Now, I would like to show another imaging
00:13:50.06 application. Now confocal microscopy has been the gold standard
00:13:54.00 that for at least 20 years, I would say again that the 21st century, it provided
00:14:00.12 the best possible spatial resolution in fluorescence microscopy.
00:14:03.27 But again, you see here the confocal microscope is not able to
00:14:08.05 discern the units of the nuclear pore complex indicated in here.
00:14:13.09 So, because the spatial resolution is about 250 nm, now we do a STED
00:14:18.03 recording. And this really tells a different story, because the resolution
00:14:22.06 is fundamentally increased as you see in here, the STED microscope
00:14:26.02 is capable of discerning the 8 individual subunits here of the nuclear
00:14:31.07 pore complex. Why? Because the spatial resolution is improved
00:14:35.07 by an order of magnitude. It's roughly about 25nm in this case.
00:14:39.09 So we see 8 individual subunits, but confocal microscopy
00:14:43.17 couldn't tell the difference. No way to see this with confocal microscopy.
00:14:47.25 So an order of magnitude resolution really means something.
00:14:51.17 Now, keep in mind the STED recording is done like anything
00:14:57.15 a confocal recording. You place the sample, you push a button,
00:15:01.18 the beams move across the specimen, scanning over the sample,
00:15:05.03 and out comes the image. There's no further requirement
00:15:08.28 on image processing or anything. Just the preparation of the beams
00:15:13.02 of light with regard to transitions of the fluorophore in the end
00:15:17.25 rendered image. And these are the strengths of the image, everything
00:15:21.10 is done basically by molecular transitions. Now, I would like to show you a few
00:15:27.18 applications. So, here STED microscopy has been used in order
00:15:31.25 to unravel this spatial arrangement of proteins in the presynaptic
00:15:38.12 actizone in drosophila neuromuscular junctions. So, protein called
00:15:43.13 Bruchpilot and another protein, RIM-binding protein are arranged
00:15:46.19 in space. They cannot be resolved because the resolution is not good enough
00:15:50.04 in the confocal case. But here in STED microscopy, the resolution is good.
00:15:57.04 It's of the order of 30-40nm, and so one could instantly get an
00:16:02.23 idea of how these proteins are arranged in space. So you get distances
00:16:07.29 of 38, 57, 98 nm between different terminal ends of the proteins.
00:16:15.21 And so, this helps coming up with the model of the presynaptic
00:16:22.18 actizone there. Now in that application, in neurophysiology,
00:16:28.00 so this is a stretch of a dendrite in a hippocampal organotypic slice.
00:16:33.20 A living hippocampal organotypic slice. So some of them dendrites
00:16:38.25 here, some of the neurons here, express the yellow fluorescent protein
00:16:42.06 this is why we can see them. What you can see here is dendritic
00:16:46.18 spines, with a spatial resolution that is about 3-4 times better than
00:16:50.26 in a standard confocal or multi-photon microscope. And you can see
00:16:56.17 how they change over time. For example, here, if you concentrate on this
00:17:02.05 little spine, we can see how this little spine actually evolves
00:17:07.17 in time, forming this little cup in the end. And so this is quite
00:17:12.05 interesting, because one can learn about the morphological changes that go on
00:17:15.16 in neurons. And of course morphology is connected with function,
00:17:20.27 this is very well-known. A strength of the STED concept is that it can
00:17:26.01 be neatly integrated with the confocal microscope. So you can use that
00:17:30.19 lowered resolution to improve the confocal microscopy very easily.
00:17:34.00 And that means one can image deeper down in the specimen.
00:17:39.03 And this is demonstrated here, like 10 um deep, 63nm, 25um deep,
00:17:45.06 65 um deep, 115 um deep, and still, the high spatial resolution
00:17:50.23 is fully maintained. And probably in the future we'll be able to go even
00:17:58.11 deeper down with some special optics. But the point that I'm making here
00:18:03.08 is that you can focus of course into deeper regions of neuronal
00:18:11.12 tissue, in this case. And still have the high spatial resolution
00:18:14.16 in a living neuronal tissue. Now, quite exciting is the fact that you can even
00:18:23.07 focus into the brain tissue of a living mouse. So here, a mouse
00:18:29.10 was anesthetized and the skull was opened. And a coverslip was placed atop of it
00:18:34.13 and we focused the light onto the molecular layer of the somatosensory
00:18:38.16 cortex in a transgenic mouse that expressed in some of its neurons,
00:18:43.12 the yellow fluorescent protein. And now you see again the stretch of
00:18:47.06 dendrite. And this is a sequential recording, so in other words, a movie.
00:18:51.15 And what can be seen in here is actually how these little dendrites
00:18:55.12 change over time, how they change their morphology and during
00:19:00.27 anesthesia. And actually, the mouse was living, so during the process of
00:19:06.23 life, if you will. It's interesting to focus here on this small little
00:19:11.27 area, it's possible to see again how these subtle changes, these
00:19:16.19 morphological changes going on in the brain of the living mouse.
00:19:21.13 So this tells us that subdiffraction resolution imaging by STED really
00:19:25.18 has the potential to open up a totally new avenue for
00:19:31.02 investigating brain. For example, seeing proteins in here on the
00:19:36.16 synapse and so on. Now, after showing a few examples of
00:19:43.20 applications, I'm coming back to the basic principles.
00:19:46.12 Well, in STED microscopy, we separate by making sure that adjacent
00:19:52.16 features are not capable of emitting at the same time. Because this one is
00:19:56.29 off, and this one is on, and vice versa. And we do that by applying
00:19:59.24 a beam of light that determines where the features are on and where
00:20:03.20 the features are off. Now this principle is of course a very fundamental
00:20:07.01 principle, because there are several ways of playing the on/off games
00:20:11.22 so to speak. Because there are several molecular mechanisms
00:20:14.25 that are imaginable for separating features like that, by on/off.
00:20:20.21 Now, STED was the first method probably because if you have
00:20:28.08 a physics background, then you know the most fundamental way of turning
00:20:31.14 off a fluorophore is to send the molecule from the fluorescent state down to the
00:20:35.07 ground state. There's nothing more fundamental than that, you just
00:20:37.10 send it back down to the ground state and you silence the fluorescence.
00:20:40.17 Because of that, the concept of STED microscopy is very universal.
00:20:45.07 You can apply basically to any fluorophore. But this universality comes at a
00:20:50.27 price that we have to pay. In order to switch off a fluorophore, or turn
00:20:55.12 off a fluorophore by this phenomenon of stimulating emissions
00:20:58.12 by a beam of light, beam of exciting molecules, you have to make sure
00:21:02.05 that the beam is bright enough. Because we have to have many
00:21:07.03 photons, redshifted photons in the area, so to speak, ready to
00:21:11.11 shift molecules down to the ground state indicating that it gets excited
00:21:15.12 We have only 1 nanosecond of time, so we have to make sure there's enough
00:21:18.17 photons in the air, so to speak, to be ready to shift the molecule
00:21:22.21 down to the ground state. And this is the reason why we have this megawatt
00:21:25.05 per square centimeter. So, because this can be a drawback in a number of
00:21:30.27 applications, especially if one goes to a short wavelength and the very
00:21:34.15 very high intensities, the question comes up are there alternative on/off
00:21:39.14 mechanisms that allow us to play a STED like on/off switching
00:21:44.29 at much lower light levels. And surely there are. And this is the idea
00:21:49.24 behind this generalization of RESOLFT microscopy. Now in a
00:21:55.11 typical say, RESOLFT microscope, which is in principle a STED-like microscope,
00:22:00.02 we use not stimulated emission to off-switch molecules, not the
00:22:05.23 excitation by light, but we use two states. Metastable states
00:22:11.12 of a fluorophore, like a cis state and a trans state. Now usually
00:22:15.05 such states have a relatively long lifetime, like milliseconds or seconds.
00:22:20.01 And because they have a long lifetime, this on/off state can be maintained for
00:22:26.26 a longer period of time. So there is not much reason to hurry up and
00:22:30.27 put in many photons, so one can reduce this very important intensity
00:22:35.09 required to play the on/off game by orders of magnitude. And
00:22:40.00 the same can be done by using reversibly switchable fluorescent
00:22:43.25 proteins, because usually they also have cis/trans isomerization as a
00:22:48.09 basic mechanism for going from an activatable state to a deactivated
00:22:53.04 state, and vice versa. So how do you play a STED-like game
00:22:57.19 here with cis/trans? Well you use, for example, blue light, for example,
00:23:01.17 that pushes the molecule from cis to trans. And as you can see here,
00:23:05.14 what happens, it pushes these to trans, all the molecules are turned off
00:23:09.03 are in the trans, but only these are in the cis. And so we flood everything
00:23:13.28 with excitation light, no way out. All the molecules will be covered.
00:23:17.07 But only these are allowed to emit because they are in the cis.
00:23:19.29 The rest is in trans, so they can't emit. So you can separate this
00:23:23.00 feature from that. Of course by doing the same thing with the rest of
00:23:26.15 the molecules. Now, the good thing is, as I said, the Is, the threshold
00:23:31.08 is reduced by orders of magnitudes. And this equation of course means
00:23:35.27 that d can become very small, even if the intensity I is not so high.
00:23:41.03 It's reduced by orders of magnitude. So can we play this game?
00:23:45.03 The answer is yes. We can play this game. How do we do that?
00:23:50.00 Well, we turn off molecules like deactivate for the switchable
00:23:53.16 molecules, like switchable fluorescent proteins. And then readout those
00:23:57.20 that have been left over. But then we have to activate them again,
00:24:00.12 so we turn them on again. And then again, we turn them off without
00:24:05.17 those, activate everything again. We have to activate everything because
00:24:09.03 the beam with which we activate is also limited by diffraction, so
00:24:12.06 all the molecules will be activated. Then we turn it off again.
00:24:15.13 And then these are leftover. And you can discern the features
00:24:19.05 like in the STED microscope. But now, as you can certainly imagine,
00:24:23.25 if you play this on/off game, like switching on, switching off,
00:24:30.07 we have to force the molecule to go between the on/off state
00:24:36.29 many, many times. There's no way out, because in addition to the
00:24:40.24 activity of the molecules in the 200 nm zone, many of the molecules
00:24:45.12 will go on/off without having been readout. So there will be
00:24:48.14 many on/off cycles required, in order to play the on/off game.
00:24:52.10 So this concept RESOLFT has become powerful recently because
00:24:56.28 new photoswitchable fluorescent proteins became available
00:25:01.27 that allow us to accommodate many on/off cycles. Like the ones that are described
00:25:07.02 here in this literature. And so, you can see that we can have here
00:25:12.27 a higher spatial resolution, a STED-like resolution, although there is no
00:25:16.08 STED in it. It's all done, in this case, by a cis/trans isomerization.
00:25:21.02 This has allowed us to do a STED-like recording at orders of magnitudes
00:25:24.23 higher than light allows us. Now this does not mean that it's going to replace
00:25:29.00 STED entirely, of course. STED maintains a number of advantages
00:25:33.04 such as, it allows you to stop the fluorescence instantly.
00:25:36.16 Here you always have a kinetics between the on and the off state.
00:25:40.06 But still, when it comes to imaging very gently, RESOLFT, this method clearly has an
00:25:47.25 advantage. Why? Because it can have a STED-like recording at very, very
00:25:52.24 low light levels. And here we have recorded neuronal tissue
00:25:56.12 at very, very low light levels over extended period of time. So
00:26:00.18 it's a living hippocampal organotypic slice, and after two hours of recording,
00:26:06.14 there was no visible photodegradation. And this is really remarkable
00:26:10.21 because it means that one can break the diffraction barrier, stay in
00:26:15.25 the physiological conditions, even work at regular light levels because
00:26:19.09 we take advantage of the long lifetime of the two on/off states
00:26:24.20 involved. Now, how do these concepts, STED and RESOLFT
00:26:31.02 compare with PALM/STORM? Especially since RESOLFT and PALM/STORM
00:26:36.06 both use photoswitchable fluorescent proteins. Now, in the STED
00:26:41.05 and RESOLFT cases, it's very important to understand that we determine with a beam
00:26:45.12 of light where the molecules are on and where they're off. So we use the
00:26:49.16 blue beam of light to turn, for example, the proteins off in here.
00:26:55.02 And you see they're allowed only to be on in here. Or with the red beam
00:26:57.28 of light, in the case of STED. And so, we use a pattern of light
00:27:03.10 to define the coordinates with very high precision in space
00:27:07.13 where the molecules are capable of emitting light. So this is a
00:27:10.27 hallmark. We don't have to find out where the emission takes place,
00:27:13.14 we know already where it is because we have used this pattern of
00:27:16.23 light to determine where the emission comes from. What has to
00:27:20.27 come from to be precise. With PALM/STORM, this is different.
00:27:23.26 No structured illumination, no donut or stripe or anything is used
00:27:28.15 in here. So molecules are turned from on state to off state
00:27:33.07 randomly. It's uniformed illumination, such that only one molecule
00:27:37.06 fits into this 200nm zone. Of course, if this happens randomly
00:27:41.19 at any location within the 200 nm zone, we have no clue
00:27:45.05 where it's located. But, we can find it out and that's the good news
00:27:49.27 here. If the molecule is capable of emitting m photons in a
00:27:54.15 bunch, and many photons in a bunch, then we can locate
00:27:58.11 with very high precision where the molecule has gone to the
00:28:02.21 on state. And so we can locate with very high precision where
00:28:05.29 the molecule is located. And then we make a tic mark in here, in the
00:28:11.13 PALM/STORM case and turn it off, this is very important, and go to the next
00:28:14.04 and turn this one off, locate it, and go to the next. Randomly. And this is how
00:28:21.00 we can separate things again. All the molecules here, flooded with
00:28:24.25 excitation light. No way out, but only this one is allowed to emit.
00:28:28.20 And this is why we can separate, of course, this feature from the other
00:28:32.16 feature. Because the molecules are allowed to emit only sequentially
00:28:35.29 from adjacent features. Again, here, everything is flooded with excitation light.
00:28:41.01 But only this one is allowed to emit, and that is why we can separate
00:28:44.01 this feature from that feature. And so we go on and reconstruct,
00:28:48.09 this is why it's also called STORM, stochastic optical reconstruction
00:28:52.07 microscopy. The whole 200nm of molecules at random positions
00:28:59.00 in space. Now, a good question is, how do the two concepts compare?
00:29:03.16 What are the requirements for each them? For example, in terms of
00:29:07.05 fluorophores? Some fluorophores will be good for this one but not
00:29:10.00 be good for that one, and vice versa. So what are the conditions
00:29:13.13 actually for this one or that one, in terms of the fluorophores?
00:29:18.05 Now, it's very clear that a method such as PALM/STORM requires many
00:29:23.06 fluorescence emission cycles, so many photons to be emitted because we
00:29:26.14 need those photons to locate where we are. Here the positioning,
00:29:29.10 the precise positioning is done with the photons that come out.
00:29:32.15 Here, this is not the case, because here we use a beam of light
00:29:38.03 which has many photons inherently to determine with high precision
00:29:41.16 where the molecules are on and off. So here, we determine the position
00:29:45.19 and the coordinates of emission with the photons that go in
00:29:48.13 that are coming from the laser, and here we do it with the
00:29:51.17 photons that come out from the sample. And of course, there are
00:29:54.19 many photons here that go in, and of course, there must be many
00:29:57.22 photons that come out. Because otherwise, we don't get precision.
00:29:59.25 But then, where this sounds like a disadvantage, this one has a
00:30:06.00 big advantage, PALM/STORM in this case. It's enough to have the molecules
00:30:10.03 go only once, from on and back to off, to record an image. That's enough.
00:30:15.09 Because once you have recorded the molecule, you go to the next
00:30:18.10 and you don't have to return to that molecule. So one switch cycle is
00:30:22.04 enough. This is not the case in here. Here of course, a few
00:30:25.24 emission cycles, fluorescent emission cycles, are okay. If
00:30:29.14 2-3 photons come out of this say many molecules, we know that
00:30:33.01 there is a strand of microtubule or whatever, we can go on.
00:30:35.27 But you need many on/off cycles in order to make an image.
00:30:40.00 So if the molecule basically dies simply because it's unable to accommodate many
00:30:45.13 on/off cycles, we have a problem in here. So, you see there's a
00:30:49.25 complementarity in the way that it works and also its requirements
00:30:54.09 for getting an image using fluorophores. And some fluorophores will be better
00:31:00.17 at this, and some fluorophores will be better at that. Now, if you ask me
00:31:04.22 since I'm talking about this family of concepts, what is some kind of unique
00:31:08.18 strengths of this concept? Well, for one example, you don't need many
00:31:12.08 photons to know that there is something. Because we know
00:31:15.20 already the position of emission, because we predetermined
00:31:18.14 it with the photons from the laser. And because of that, and because this
00:31:23.18 takes time of course to get all those photons, and works on a single
00:31:27.03 molecule. You can imagine that this one is quite good at making fast
00:31:30.10 images. And this is actually what is illustrated here. You see on
00:31:35.16 the left hand side, you see basically the STED-like recording.
00:31:39.12 On the right hand side, you see the PALM/STORM-like recording.
00:31:43.14 Now if the sample moves, obviously, in the stochastic method,
00:31:48.11 in PALM/STORM, you can imagine that the whole recording time
00:31:53.26 starts on the first to the last molecule. And if something moves
00:31:56.28 of course, the pattern will be smeared out. Because a temporal difference
00:32:01.05 between two adjacent features can be very large. It can be
00:32:04.16 as long as the typical duration of the whole recording.
00:32:07.16 But in a deterministic case, where we have preset the coordinate
00:32:11.00 like in the STED case, the situation is different. Adjacent features
00:32:18.16 are recorded a very small time intervals, and because they're recorded
00:32:21.25 at very small time intervals, the movement doesn't matter that much
00:32:24.19 for the separation. This is why we can still record in the STED
00:32:27.29 microscope and see features separately. Whereas in the other
00:32:33.21 case, in the PALM/STORM case, they will be more or less a blur.
00:32:38.15 So this is an important point. Now, this is actually the reason why
00:32:43.20 it has been easier with STED microscopy to get very high
00:32:47.17 speed subdiffraction microscopy super resolution images, and I would like
00:32:51.08 to show you an early example here. This is a snapshot of a moving
00:32:56.27 synaptic vesicle in a living hippocampal neuron. And as you can see
00:33:00.19 the confocal resolution cannot resolve it, but in this case STED
00:33:05.25 has allowed us to see the movement of this synaptic vesicle.
00:33:08.16 So you can go past the diffraction barrier at the same time very fast
00:33:11.11 because we know already the signal come from. A few photons are enough
00:33:16.25 to know there is a vesicle. And this is an inherent little conceptual
00:33:22.17 reason why STED allows you to do fast imaging and of course,
00:33:28.03 3D recording of moving samples. Now, finally, I'm addressing
00:33:34.20 the question we are now able to take images, but 10 years back
00:33:37.24 we couldn't take these images. So what is the reason, what is the key
00:33:42.08 point behind the recording of these features? Clearly it's not the fact that we
00:33:48.01 localize molecules with high precision, because the localization
00:33:51.17 just gives you the coordinates, it doesn't give you the separation. The separation
00:33:53.28 is critical. Without separation you cannot see features, you can't take
00:33:58.07 images because you have other molecules around in here.
00:34:02.14 So if there's just one molecule, you could take an image, you could
00:34:04.29 locate it. But we could not take images. So, the point is, as you can
00:34:11.05 imagine that we transiently prepare the molecules into the
00:34:15.04 different states. So, while we flood of course, all the features
00:34:21.07 residing within the 200 nm zone, at the same time with excitation
00:34:26.02 light, only one molecule in this case is allowed to emit, whereas
00:34:30.23 only this small subset of molecules, if there are any, are allowed
00:34:34.13 to emit. And this is what allows us to tell the features apart.
00:34:38.07 So, transiently bringing the molecules into an on state and having the rest
00:34:43.05 in an off state or vice versa, allows us to overcome the diffraction
00:34:47.12 barrier. Here, the coordinate of emission is predetermined.
00:34:51.17 Here, it has to be found out. Both with very high precision.
00:34:54.26 But the separation is done by the on/off, and that's the critical
00:34:58.02 element. So, off means there is some mechanism in there that
00:35:03.04 doesn't allow fluorescence to come out. Although, all the molecules
00:35:08.13 are covered in excitation light and this gives us the separation.
00:35:11.22 Now on/off is of course represented by states A and B.
00:35:16.00 And if you think about that, A is the emissive state, B is the
00:35:20.29 non-emissive state. It becomes obvious that it's not fundamentally
00:35:24.22 limited to fluorescence imaging, you can imagine all kinds of
00:35:28.22 states, different states A and B. It could be for example, fluorescent/non-fluorescent,
00:35:34.14 as we've shown here. Or it could also be absorbing/non absorbing,
00:35:38.13 cis/trans, green/blue, heating/not heating, or you can imagine
00:35:44.13 scattering/non-scattering. One can imagine many, many things.
00:35:46.28 And this is why this concept is just at the beginning. Super resolution
00:35:51.13 microscopy, as I believe, is still at the beginning. It will evolve
00:35:54.18 into a large field, because so many on/off states or A/B states
00:35:58.26 are imaginable to overcome the diffraction resolution barrier.
00:36:03.08 With that, I'm of course acknowledging and highlighting the
00:36:06.13 people who have contributed to this idea development.
00:36:10.06 And finally, I'm coming to my very, very last slide
00:36:12.29 Abbe's equation, of course, cannot explain the entire situation
00:36:17.10 because now, we're able to do better than the gold standard
00:36:21.29 of confocal microscopy. STED, raw data here, resolves much better than
00:36:27.20 a confocal microscope, which has been the limit for many
00:36:30.17 many years. And so the question comes, what do we do with this
00:36:35.00 equation? Well, you've seen it already. We can easily expand it
00:36:38.04 by plugging in the square root factor. And now it's worthwhile
00:36:42.15 spending a few thoughts on this equation. Of course the resolution
00:36:47.19 scales with the wavelength, it still scales with the wavelength,
00:36:50.20 it must scale with the wavelength because we use diffracted beams.
00:36:55.00 Every beam is still diffracted. We do nothing about diffraction.
00:36:59.01 But it's not limited by diffraction, the resolution is not limited by
00:37:04.18 the wavelengths anymore. Why? Because we have this factor
00:37:09.01 in here, and if this number becomes very large, this value of
00:37:12.15 d can become very small. So although diffraction is maintained,
00:37:16.28 the barrier set by diffraction is overcome. And that is the point.
00:37:22.07 And how is it overcome in the end? I think it's very clear by
00:37:25.10 now. Why can we separate, for example, this fiber from that fiber?
00:37:29.22 The answer is very simple, at the time, the molecules of
00:37:33.19 this fiber emitted light, these ones were dark. At the time these
00:37:37.27 ones emitted light, these ones were dark. So there were two
00:37:43.13 different states, so we resolved in modern super resolution microscopy
00:37:47.24 not by making the light beams narrower or finer, or more focused,
00:37:54.07 but by transiently placing the molecules that we look at in two different
00:37:59.13 states. Placing the molecules transiently in two different states
00:38:02.26 has allowed us to crack the diffraction resolution barrier, fundamentally.
00:38:09.09 Thank you very much.

This Talk

Speaker: Stefan Hell

Audience:

Researcher

Recorded: May 2013

Talk Overview

Historically, light microscopy has been limited in its ability to resolve closely spaced objects, with the best microscopes only able to resolve objects separated by 200 nm or more. This limit is known as the diffraction limit. In the last twenty years, a number of techniques have been developed that allow resolution beyond the diffraction limit. Here, Stefan Hell, who invented many of these techniques, gives an introduction to these super-resolution microscopy techniques, and a detailed discussion of two such techniques: STED (Stimulated Emission Depletion) and RESOLFT (REversible Saturable OpticaL Fluorescence Transitions).

Questions

What is the definition of super-resolution?
How does STED produce a super-resolution image?
1. It uses switching on of molecules located randomly in the field of view
2. It uses multiple objectives and interference
3. It uses high frequency patterned illumination
4. It turns off molecules outside the center of the focus by stimulated emission
True or False: STED and RESOLFT are widefield, camera based techniques.

Answers

View Answers

Speaker Bio

Stefan Hell

Stefan Hell is the Director of the Max Planck Institute for Biophysical Chemistry and head of the Department of Nanobiophotonics. Hell was the first to conceive of and develop a fluorescent microscope with a resolution of nanometers (a fraction of the wavelength of light)- a feat previously thought to be impossible. For his work, he was… Continue Reading