Measuring Dynamics: Fluorescent Speckle Microscopy

00:00:11;28 Hi.
00:00:12;28 I'm Clare Waterman from the NIH, and I'm going to talk to you today about
00:00:15;14 fluorescent speckle microscopy,
00:00:17;16 which is a method that I invented by accident when I was a postdoc with Ted Salmon back in the 1990s.
00:00:25;03 So, what is... light microscopy is a fabulous tool for studying the dynamics of protein
00:00:32;28 macromolecular ensembles in living cells with spatial and temporal precision.
00:00:41;12 The advent, today, of multicolor specific fluorescent labeling through either
00:00:46;22 chemical dyes or fluorescent protein analogues has allowed the labeling of multiple molecules
00:00:54;19 within multiple cellular machines, and the analysis of their dynamics and their correlation
00:01:00;20 with each other and with cellular behavior, so that we can understand how proteins self-assemble
00:01:06;09 into dynamic machines that animate a living cell.
00:01:09;08 But the problem with light microscopy is that it's limited by the fundamental limits of
00:01:16;12 geometrical optics when using visible light, which is the diffraction barrier.
00:01:23;13 Which, for light in the visible range, is about 200-250 nanometers.
00:01:28;24 Which is a problem because proteins are about on the order or 5-20 nanometers, so if you
00:01:36;26 really want to understand how macromolecular assemblies dynamically come together,
00:01:43;14 you can't resolve in the individual proteins or their dynamics below the resolution of a light microscope.
00:01:50;11 So, what is a speckle?
00:01:52;05 And how does a speckle help you get this information?
00:01:54;10 I like to say that a speckle is a local probe of biochemistry and physics in a living cell.
00:02:00;08 A speckle... each speckle... and you see this image, here.
00:02:04;04 It's a... it's the speckled actin cytoskeleton in a migrating lung epithelial cell.
00:02:08;21 The focal adhesions are labeled with a focal adhesion marker, in green.
00:02:13;14 The actin is in red.
00:02:14;26 And what you see is a bunch of... millions of flickering and moving speckles.
00:02:20;08 And each one of those fluorescent speckles is packed with information about what's going on
00:02:25;17 at the sub-resolution level in the macromolecular assembly of the actin cytoskeleton.
00:02:31;16 Each speckle is encoding, for example... the intensity of speckle... the rate of change of intensity
00:02:37;19 encodes the rate of binding and dissociation of fluorescently labeled molecules
00:02:42;28 to and from the cytoskeleton.
00:02:45;04 So, this is a biochemical rate constant at the spatial resolution of the light microscope.
00:02:53;20 So, 250 nanometers, everywhere in the cell, you get a rate constant.
00:02:58;19 The motion tells you about the physics of the material, the trajectory and velocity
00:03:03;24 of the material.
00:03:05;11 You can even use the Brownian motion of the speckles as a probe for the viscoelasticity
00:03:11;22 of the material.
00:03:12;22 It's a very high-resolution method.
00:03:17;13 The spatial resolution is on the order of 10 nanometers when you use
00:03:22;03 point spread function fitting software.
00:03:25;10 And the temporal resolution, with EMCCD cameras today, is about the order of 10 milliseconds.
00:03:30;19 So, in a movie like this, you've got millions...
00:03:36;00 literally millions of speckles that are sparkling and moving.
00:03:40;03 The information density is huge.
00:03:42;17 50,000 speckles per time point, over time, giving you all this information.
00:03:47;05 So, it's really a remarkable technique.
00:03:51;06 And 50,000 data points per time point over time... you give that to a graduate student
00:03:58;06 to track them, they're gonna commit hara-kiri.
00:04:00;14 So, if... what you need to do with all of this data is extract it in an efficient way.
00:04:08;03 And to do that, you need computer vision analysis tools.
00:04:13;03 So, just to give you an example of what a speckle image looks like, what you see here
00:04:19;01 is the actin cytoskeleton fully labeled with rhodamine-phalloidin over here, all the way
00:04:26;01 to the right.
00:04:27;01 So, you see all of the actin filaments labeled with rhodamine-phalloidin.
00:04:30;22 That same cell has been injected... or... expressing a very low level of fluorescently tagged
00:04:35;19 actin in another color.
00:04:37;25 And what you see is that you miss a lot of this higher-order structural information
00:04:42;00 -- the bundles and what not -- and all you see is a bunch of little fluorescent speckles.
00:04:45;23 Now, if we zoom in to this small area at the leading edge, what you see those speckles are
00:04:51;00 are actually diffraction-limited point spread functions of varying intensity.
00:04:56;05 And if we superimpose this little 2x2-micron region onto an electron micrograph taken by Tatyana Svitkina
00:05:02;27 and really, you know, sort of Photoshopped, what you see is that
00:05:07;22 within one of these little diffraction-limited spots there's actually lots and lots of actin monomers.
00:05:14;16 And what you see is adjacent regions that have very few fluorescent monomers incorporated
00:05:20;22 into them, and thus are dimmer than the brighter peaks of intensity.
00:05:25;16 So, this is literally just drawn on with Photoshop, these little dots representing a probable number
00:05:33;16 of how many fluorescent subunits went into making each one of those speckles.
00:05:38;22 And what you see is that it's a very, very small fraction of the actin subunits
00:05:46;00 that are labeled in order to get this contrast between adjacent diffraction-limited image regions,
00:05:50;26 a dim one here and a bright one there.
00:05:54;23 You know, we have techniques for measuring the dynamics of protein macromolecular assemblies
00:06:00;28 in cells with a light microscope, for example, FRAP, where you bleach some subunits...
00:06:05;22 you bleach a region of a macromolecular assembly, and it reincorporates fluorescence and
00:06:11;02 you see the rate at which that happens, and that tells you about the turnover of molecules in that.
00:06:15;25 Here, we're getting that information in every single speckle, so it's like a million little
00:06:19;26 FRAP experiments.
00:06:21;17 And FRAP is limited to a small region of the cell, whereas fluorescence speckle microscopy
00:06:28;02 gives you this information across wide-field view.
00:06:30;28 It also is capable of monitoring non-steady state dynamics, of variable rates of
00:06:36;18 assembly and disassembly that a FRAP experiment couldn't give you very precise information about.
00:06:44;04 And you can detect variation in molecular dynamics at much higher spatial and temporal...
00:06:49;02 or, at much higher spatial resolution.
00:06:51;08 So, how do you make a speckle?
00:06:53;10 It's simple.
00:06:54;13 It's totally simple.
00:06:55;21 The way that you make a speckle is you create contrast by assembly of a very, very low fraction
00:07:00;28 of fluorescent subunits with a high fraction of unlabeled subunits that are supplied by your cell.
00:07:06;18 The cell is cranking out nonfluorescent proteins, and you put a small amount of
00:07:11;06 fluorescent subunits that co-assemble randomly into macromolecular ensembles.
00:07:17;16 And during this assembly, this gives rise to a random variation in sort of the...
00:07:23;04 the spatial distribution of these fluorophores that gives a speckled appearance to the image.
00:07:29;21 And it is a very low number.
00:07:31;06 So, in the case of microtubules, which we characterized extensively early on,
00:07:36;17 optimal speckle contrast is on the order of about 99.9% unlabeled tubulin and 0.1% labeled tubulin.
00:07:45;15 At this concentration, each speckle contains about 1-7 fluorophores, and adjacent regions
00:07:51;22 in the image contain basically none.
00:07:54;24 The advantages of FSM are that movement and changes in these speckle patterns stand out
00:07:59;26 to your eye and can be measured.
00:08:03;15 Because you're putting in much lower fluorescent... a lower level of fluorescent proteins,
00:08:08;09 the background fluorescence is very, very low, so you can see deeper into thick specimens
00:08:14;17 with fluorescent speckle microscopy.
00:08:16;17 You get wide-field views.
00:08:18;11 And the hardware is simple; you can pretty much do it with any properly set up high-resolution
00:08:23;17 microscope system.
00:08:25;05 And you can adapt it to any different kind of high-resolution fluorescence microscopy:
00:08:30;03 confocal, spinning disk, TIRF, for example.
00:08:34;02 So, this is just to show you an example of comparing the information that you get by FRAP
00:08:40;26 versus fluorescent speckle microscopy of actin dynamics in a migrating cell.
00:08:45;28 So, this movie that you see playing on the [left] is a movie from... a classic movie
00:08:51;09 from Yu-Li Wang that first demonstrated that actin in the leading edge treadmills.
00:08:54;22 And what you see is a small bleached spot form at the leading edge.
00:08:59;17 And that bleached spot, as it recovers, moves rearward toward the cell center,
00:09:04;09 indicating new assembly of actin at the leading edge coupled with motion of the bleached area of
00:09:10;15 the actin cytoskeleton.
00:09:11;26 In contrast, what you see over here in the speckle movie is millions of little speckles.
00:09:16;28 You can see they're not just generally going rearward at a certain speed.
00:09:20;26 They start out here.
00:09:22;05 They move a little bit.
00:09:23;05 They disappear.
00:09:24;09 There's a much slower rate of retrograde flow here.
00:09:26;22 And there's actually a forward flow up toward the cell center, here.
00:09:30;16 You get a... it's like doing a million little FRAP experiments to get the motion dynamics
00:09:34;18 of the actin cytoskeleton, in one movie, basically.
00:09:38;21 So, I'm going to give you a little history, because it's a good history lesson.
00:09:42;02 We discovered this technique by accident, okay?
00:09:44;26 I was a postdoc in Ted Salmon's lab.
00:09:46;16 I was studying the dynamics of fluorescent microtubules in migrating epithelial cells.
00:09:51;24 And we noticed in... we had a 100x microscope objective lens and new cameras,
00:10:00;06 new cooled-CCD cameras with a small pixel size that allowed us to get full diffraction-limited resolution
00:10:06;13 in the camera image.
00:10:07;15 And what we saw was, instead of microtubules with a continuous label, there were
00:10:12;05 these dim spots along the microtubule.
00:10:14;11 And we noticed in this time series that the dim spots actually... marked by this black arrow...
00:10:19;11 they actually stayed static relative to the microtubule, where the front end of
00:10:24;21 the microtubule moved forward, the back end of the microtubule moved forward,
00:10:28;19 indicating that this microtubule wasn't being pushed around in the cell by molecular motors,
00:10:32;26 but it was actually assembling at one end and disassembling in the other end, and undergoing
00:10:37;18 this treadmilling through the cytoplasm.
00:10:40;06 So, we... you know, we distinguished between two mechanisms of motion of this microtubule
00:10:47;22 just by looking at the... this... what looked like an artifact to a lot of people.
00:10:51;25 It was like... ooh, those are ugly, speckly microtubules... let's inject more tubulin...
00:10:56;07 but we thought about it.
00:10:57;07 So, we thought really hard about where these fluorescent speckles come from and whether
00:11:01;21 they are, in fact, an artifact or not.
00:11:03;24 Ted Salmon developed a model to try to explain how the assembly of fluorescent and non-fluorescent dimers
00:11:12;02 would result in a microtubule that looked speckled in a light microscope image.
00:11:16;20 And what he assumed is that microtubules are a polymer with 600...
00:11:21;04 1625 dimers per micron, and that they assemble only at their ends, from a dimer pool
00:11:30;23 that randomly incorporates dimers into the microtubule at a rate that's according to the concentration
00:11:36;22 of the dimer pool.
00:11:38;02 So, when a very small fraction of that dimer pool is fluorescent, what that results in
00:11:43;02 is just a random incorporation, every once in a while, a stochastic incorporation,
00:11:48;10 into the microtubule.
00:11:49;24 And when you think about how that would look in a microscope image, what you have to think about
00:11:55;21 is what the microscope is going to see, which is the diffraction limit.
00:11:59;04 And in this case, we were using red fluorescent tubulin, so that was about 275 nanometers.
00:12:03;00 So, within one diffraction-limited image region of that microtubule, there's...
00:12:08;24 270 nanometers... there's about 400 tubulin dimers within each diffraction-limited image region.
00:12:15;17 So, what... if you just assumed that every 440 dimers there was one fluorescent dimer incorporating,
00:12:27;05 that would give you a mean fluorescence of the brightness of one dimer,
00:12:31;04 okay?
00:12:32;04 Now, for 1% labeled tubulin, that would be 4.4... or 4 dimers per 270 nanometers.
00:12:40;05 So, you might expect that you get a microtubule that's just as bright, continuously along
00:12:45;10 its length, as 4 dimers per...
00:12:48;02 4 fluorescent dimers per 270 nanometers.
00:12:51;02 But when the fraction of fluorescent dimers is very low, the chances are that...
00:12:56;06 that you're actually going to have a fluorescent dimer are very low, and more likely
00:13:01;20 it's gonna be an unfluorescent dimer.
00:13:02;22 So, the standard deviation around that mean is very high relative to the mean.
00:13:07;28 And as the mean goes down and down and down, the standard deviation of the number of fluorescent dimers
00:13:12;18 per diffraction-limited image region goes up relative to the mean, giving rise
00:13:17;08 to, basically, a speckled contrast, where, in this diffraction-limited image region,
00:13:23;04 you might have 4 or 6 fluorescent dimers, and in this adjacent one, this adjacent region,
00:13:31;04 you might have 0.
00:13:32;25 And this high standard deviation of the number of fluorescent dimers per diffraction-limited
00:13:40;13 image region, relative to the predicted mean fluorescence based on the fraction of fluorescent
00:13:45;25 and nonfluorescent tubulin, is what gives rise to the fluorescent speckled appearance
00:13:51;20 of an individual microtubule.
00:13:53;10 There were several predictions of this model.
00:13:55;01 A random incorporation of labeled and unlabeled tubulin should result in a random pattern
00:13:59;23 of fluorescence along the microtubule.
00:14:01;17 So, we measured the fluorescence along the microtubule and ran a power spectrum of it,
00:14:07;00 and we saw no... no peak.
00:14:09;04 There was no periodicity of this fluorescence.
00:14:11;04 It was a random pattern.
00:14:13;07 Speckle patterns should be dependent on growth.
00:14:15;15 When a microtubule regrows, you get a different random incorporation of fluorescent subunits,
00:14:21;08 and you should get a new... a new speckle pattern.
00:14:23;12 So, here we have a single microtubule that shortens.
00:14:27;22 You can see it... you know, it's long here, it shortens back down a little bit here,
00:14:32;04 and then it regrows.
00:14:33;04 And what you can see is that, after that regrowth, the pattern of fluorescence intensity is different
00:14:38;23 here than it is here, indicating these speckle patterns are dependent on growth.
00:14:43;13 We additionally showed that microtubules... it's not some cellular factor that contributes this.
00:14:50;01 You can do it with pure tubulin in vitro, so it's not MAPs, it's not organelles,
00:14:54;07 it's not fluorescence quenching by some molecule in the cell... it's intrinsic to
00:15:00;09 fluorescently labeled and unlabeled tubulin when they co-assemble.
00:15:04;06 And this speckle pattern, the degree of speckly-ness, depends on the fraction of fluorescent tubulin.
00:15:11;03 So, at 50/50 fluorescent/nonfluorescent, they're basically... the mean fluorescence is high
00:15:16;28 and the standard deviation is low relative to that mean, so you have very little contrast.
00:15:21;15 And as you go down, to about 1%, the mean is very low but the standard deviation of the mean
00:15:29;19 is much higher, and this gives rise to a speckly-er and speckly-er microtubule
00:15:34;06 as you reduce the fraction of fluorescent subunits.
00:15:36;20 Okay, so what do you need to do to make a speckle in practice.
00:15:41;22 So, really, speckles don't...
00:15:44;14 I mean, they come from the microscope in the sense that you have to make an image from them,
00:15:48;02 but the principle of making a speckle comes from the specimen and how you...
00:15:54;26 what kind of fluorescent proteins and the amount of them that you put into the cell.
00:15:58;14 So, what you really need to have good speckles is you need well-labeled, bright, very functional
00:16:04;12 fluorescent protein.
00:16:05;26 So, you want it to be bright so that the... the brightness of regions with maybe 3 fluorophores
00:16:13;21 is high enough, relative to the camera noise, adjacent... to adjacent regions that are very dim.
00:16:20;13 So, you want bright, well-labeled, functional protein.
00:16:23;09 You don't want a nonfunctional protein, fluorescent protein, that contributes to a soluble, diffusible background.
00:16:29;22 We've messed around a little bit with putting multiple fluorophores per subunit on.
00:16:33;17 You want a very low level of the fluorescent relative to the nonfluorescent, about 1%
00:16:38;24 or lower labeled relative to 99% or greater endogenous unlabeled.
00:16:44;18 Obviously, you need difference in fluorescence intensity between adjacent diffraction-limited
00:16:49;18 images... image regions.
00:16:51;10 And you need these differences to be stable on the timeframe of image acquisition.
00:16:55;13 So, if it takes 50 milliseconds to get an image, that... those differences
00:17:00;23 between two adjacent 250 nanometer regions has to be stable, so the fluorophores have to be relatively
00:17:07;04 immobilized on that timeframe.
00:17:09;04 So, this is just an example of using a crippled CMV promoter to get low-level expression.
00:17:16;16 What you see here is the endogenous vinculin.
00:17:19;14 These are vinculin in focal adhesions in a fibroblast cell, the endogenous vinculin
00:17:24;05 by immunofluorescence, and you see a relatively continuous labeling along these focal adhesions.
00:17:29;10 And this is a GFP-vinculin expressed by this cripple CMV promoter, incorporating into
00:17:36;02 the same focal adhesions, but now you see that because there are so few labels that
00:17:40;07 the focal adhesions are speckled.
00:17:42;10 Okay, so what do you need hardware-wise?
00:17:45;00 Again, it's not that complicated.
00:17:47;11 You basically need to prevent photobleaching, because you want these very small number of
00:17:54;16 fluorophores to remain bright through a camera exposure, so you need to shutter illumination
00:17:59;14 between exposure times, and have oxygen scavengers, if your cells will tolerate them,
00:18:04;04 to reduce photobleaching and photodamage.
00:18:07;00 You need very highly efficient photon collection.
00:18:09;13 We're talking about each speckle having 3-5 fluorophores in them.
00:18:12;26 You need a low noise, high dynamic range, high quantum efficiency camera.
00:18:17;13 You need a very simple light path, removal of DIC components, extraneous mirrors,
00:18:22;26 extraneous lenses, and... keep it simple.
00:18:27;13 What do you need?
00:18:28;13 You need focus stability, because you're talking about keeping diff... basically, the size
00:18:33;11 of single molecules in focus.
00:18:36;09 You need high magnification and high resolution.
00:18:38;24 So, the highest magnification and highest resolution lenses will give you the best speckles.
00:18:44;13 So, you usually use 100x 1.4 or 1.49 NA objective lens.
00:18:49;04 And very critically, you need the resolution of the microscope and the detector to be matched
00:18:56;00 to satisfy the Nyquist sampling criterion.
00:18:59;17 So, that's basically all you need hardware-wise.
00:19:03;14 Your microscope has to be set up properly and aligned.
00:19:07;02 I mean, these criteria are... they sound simple, but they're very, very specific.
00:19:13;13 If one thing is messed up, you're not gonna get a speckle.
00:19:16;14 So, it's simple criteria, but it's very specific.
00:19:20;13 So, speckle microscopy, like I say... it's primarily coming from the specimen,
00:19:25;15 so it's really amenable to any mode of fluorescence microscopy that's capable of this high magnification,
00:19:31;11 diffraction-limited, high signal-to-noise images.
00:19:34;08 This is an example of combining total internal reflection microscopy with fluorescence speckle microscopy
00:19:41;15 to see awesome speckle contrast at the coverslip surface.
00:19:47;07 So again, it's a focal adhesion protein, GFP-vinculin.
00:19:51;17 This is a low level of vinculin expressed in the cell.
00:19:54;28 This is the total vinculin, back here, by immunofluorescence.
00:19:59;11 And this is a comparison of a wide-field epifluorescence speckle image to a TIRF fluorescence speckle image.
00:20:06;09 And what you can see is much more distinct speckles in the TIRF image compared to
00:20:12;13 the wide-field image.
00:20:13;14 So, just much brighter speckle contrast because of the very thin evanescent excitation field.
00:20:20;27 This is an example of doing multicolor speckle microscopy.
00:20:24;03 We've got the actin cytoskeleton labeled in red, undergoing this beautiful retrograde flow
00:20:28;11 with all the millions of speckles flickering and moving, and the microtubules undergoing
00:20:33;11 dynamic instability, labeled with a green fluorophore.
00:20:37;07 And from this we were able to show that the motion of the microtubules and the actin
00:20:42;12 are coupled, suggesting that they are linked by crosslinking proteins.
00:20:47;00 So, where can we go from here with speckle microscopy?
00:20:50;11 Well, what I've told you about is that you get rates of assembly and disassembly.
00:20:54;15 These are relative rates; they are not absolute rates.
00:20:56;22 We don't know the exact number of molecules.
00:20:59;25 When a fluorescence speckle increases in intensity over time, we don't know the exact number
00:21:04;14 of molecules that are incorporating into that.
00:21:07;02 So, one thing we would like to do is calibrate our quantitative speckle analysis software
00:21:12;08 to give absolute numbers.
00:21:15;27 We would like more people to adopt this technique.
00:21:18;01 It's a very powerful technique.
00:21:20;00 I think it would be a great technique for studying, for example, what's going on
00:21:24;10 inside the nucleus with DNA binding proteins when transcription gets activated, and things like this.
00:21:30;17 Cell surface receptors, other cytoskeletal elements such as intermediate filaments,
00:21:34;08 signaling molecules.
00:21:35;18 We are... now that... with structured illumination, 3-D superresolution microscopy, we think that
00:21:42;24 we're gonna be able to actually start tracking speckle dynamics in 3-D in living cells
00:21:50;04 as they move, perhaps, through living animals.
00:21:53;20 So, that's where we're going.
00:21:55;08 The people who helped us get where we are and are taking us where we're going in the future...
00:21:59;19 I owe a lot to my former mentor, Ted Salmon, who co-discovered speckles with me and
00:22:05;01 told me what to do, and I actually listened to him at that time, which was out of character;
00:22:10;04 Gaudenz Danuser, who is my collaborator in computer vision, and I wouldn't be able to
00:22:14;05 get any quantitative information out of my speckle images without him;
00:22:19;02 and various members of my lab who specifically contributed data and/or images
00:22:24;11 to the talk that I've given you today.

This Talk

Speaker: Clare Waterman

Audience:

Researcher

Recorded: July 2012

Talk Overview

Fluorescent speckle microscopy is a technique that allows monitoring of dynamics in polymeric structures by doping in a very low level of fluorescently labeled monomer. The small number of fluorescent molecules make fluorescent speckles that show up as diffraction-limited bright spots in the image. Here, Clare Waterman, the inventor of this technique, describes how to make and image speckled samples.

Questions

The optimal ratio of labeled protein to unlabeled protein for speckle microscopy is about?
1. 1 : 100 000
2. 1 : 100
3. 1 : 10
4. 100 : 1
Speckled microtubules
1. Result from random incorporation of labeled protein
2. Can only be seen in vivo
3. Result from structural variation of the microtubule
4. Result from binding of other proteins to the microtubule
Which of the following CANNOT be measured with speckle microscopy:
1. The rate of binding of labeled protein
2. The trajectory and velocity of labeled protein
3. The rate of dissociation of labeled protein
4. Overall structure and distribution of labeled protein
Which of the following objectives would you choose for speckle microscopy?
1. Plan Apo 10x/0.45
2. Plan Apo 40x/0.95
3. Plan Apo 40x/1.15 water immersion
4. Plan Apo 100x/1.4

Answers

Speaker Bio

Clare Waterman

Clare Waterman

Dr. Waterman is chief of the Laboratory of Cell and Tissue Morphodynamics at the National Heart Lung and Blood Institute, National Institutes of Health and she has been co-director of the Physiology Course at the Marine Biological Laboratory for the past 4 years. Waterman’s lab studies the interactions between actin and focal adhesions, taking advantage… Continue Reading

Leave a Reply Cancel reply