Single-Particle Cryo-EM of Biological Molecules – the Sky Is the Limit
This is a modal window.
The media could not be loaded, either because the server or network failed or because the format is not supported.
Formal Metadata
Title |
| |
Title of Series | ||
Number of Parts | 340 | |
Author | ||
License | CC Attribution - NonCommercial - NoDerivatives 4.0 International: You are free to use, copy, distribute and transmit the work or content in unchanged form for any legal and non-commercial purpose as long as the work is attributed to the author in the manner specified by the author or licensor. | |
Identifiers | 10.5446/45070 (DOI) | |
Publisher | ||
Release Date | ||
Language |
Content Metadata
Subject Area | ||
Genre | ||
Abstract |
|
Lindau Nobel Laureate Meetings301 / 340
3
5
15
16
18
22
23
27
32
33
34
35
47
54
58
67
69
70
72
73
78
79
83
84
85
86
87
90
92
95
100
102
103
104
105
106
114
115
116
118
119
120
122
126
128
129
130
131
133
137
142
143
145
147
148
149
153
155
156
159
162
163
165
168
169
174
176
178
181
182
189
194
198
201
202
203
206
209
213
214
217
218
219
220
225
227
228
237
240
241
244
245
250
254
257
260
261
266
273
278
284
285
287
291
293
297
302
308
310
317
318
319
321
325
327
328
333
00:00
NobeliumChemistryAreaMolecular beamChemical structureMoleculeCrystallographyX-ray crystallographyLecture/Conference
00:40
CrystallographyMoleculeCrystalChemical structureConformational isomerismElectronProteomanalyseX-ray crystallographyMoleculeZigarettenschachtelThermoformingBinding energyQuartzSample (material)WalkingChemical structureTiermodellHemoglobinConformational isomerismCrystallographyRogue waveComputer animation
02:10
NobeliumChemical structureElectronFormaldehydeThin filmSample (material)MoleculeArzneimitteldosisElectronic cigaretteMoleculeActive siteElectronSample (material)SaltArzneimitteldosisLecture/ConferenceComputer animation
03:15
NobeliumPlant breedingElectronLecture/Conference
03:54
QuartzRiver sourceGesundheitsstörungPlant breedingSpawn (biology)Lecture/Conference
04:57
Cell membraneProteinBakteriorhodopsinElectronArzneimitteldosisSeafloor spreadingMoleculeCrystalGlucoseHope, ArkansasRibosomeFatty acid methyl esterAcidMoleculeCell membranePuddling (metallurgy)CollectingElectronQuartzBurnCell (biology)GlucoseSchwermetallElectronic cigaretteFatty acid synthaseWater purificationRibosomeRhodopsinHelixChemical structureGesundheitsstörungExtractSynthetic rubberTool steelOmega-6-FettsäurenSecretionSolutionGasolineFunctional groupComputer animation
09:34
CrystalMoleculeNobeliumNanoparticleConformational isomerismAlumSequence alignmentTidal raceElectronMultiple chemical sensitivityMutilationMicroarrayPeriodic acidMoleculeHoneyWalkingBallistic traumaElectronic cigaretteChemical structureNanoparticleFunctional groupLecture/ConferenceComputer animation
12:02
NobeliumLecture/Conference
12:30
Modul <Membranverfahren>Process (computing)NobeliumKohlenstoff-14Sequence alignmentRibosomeProtein subunitHemocyaninMoleculeConformational isomerismHexamereBase (chemistry)Shuttle-VektorMultiple chemical sensitivityChemical propertySolutionMoleculeShuttle-VektorRibosomeFunctional groupNanoparticleHomogeneous (chemistry)LevomethadonAtomclusterProtein subunitMultiple chemical sensitivityChemical structureSchmerzschwelleHemocyaninSunscreenAdenomatous polyposis coliModul <Membranverfahren>Wine tasting descriptorsDeterrence (legal)OperonHorse meatHydrophobic effectTiermodellGrowth mediumWursthülleEconomic sectorThermoformingQuartzElectronic cigaretteTransformation <Genetik>Lecture/ConferenceComputer animation
22:17
NanoparticleNobeliumRibosomeProtein subunitFlintProcess (computing)NobeliumEpidermal growth factorFreezingLecture/Conference
23:00
CryogenicsMoleculeRefrigeratorIon channelRibosomeCalciumOctopus (ride)Golgi apparatusRhodopsinAtomElektronentransferWaterKorngrenzeMoleculeElectronic cigaretteCryogenicsGesundheitsstörungThermoformingColourantChemical structureQuartzProcess (computing)Wine tasting descriptorsGene clusterHuman body temperatureFlüssiger StickstoffWursthülleFreezingStickstoffatomSong of SongsSetzen <Verfahrenstechnik>Octopus (ride)TuberculosisEthaneRibosomeCalciumRapidSample (material)IceMeeting/Interview
26:11
RibosomeHost (biology)Conformational isomerismNobeliumMedroxyprogesteroneQuantum chemistryElectronPeriodic acid-Schiff stainIce sheetHelixMoleculeFatty acid synthaseHybridisierung <Chemie>RNAMotion (physics)Cell (biology)Ion channelCalciumAlpha-1-RezeptorAmpicillinCryogenicsNanoparticleCrystalSample (material)Chemical structureGesundheitsstörungProteinCell membraneCrystallographyDiamantähnlicher KohlenstoffBase (chemistry)RibosomeSample (material)QuartzDeterrence (legal)Ion channelSingulettzustandProcess (computing)OceanChemical structureTranslation <Genetik>StockfishElectronic cigaretteNanoparticleDiamantähnlicher KohlenstoffWursthüllePotenz <Homöopathie>MoleculeTool steelFoodHydrophobic effectWaterCrystallographyBranch (computer science)ÖlElectronFunctional groupGesundheitsstörungPharmaceutical drugProtein subunitCell membraneProteinX-ray crystallographyConformational isomerismCalciumTransfer RNAMetabolic pathwayWine tasting descriptors
32:19
NobeliumLecture/ConferenceComputer animation
Transcript: English(auto-generated)
00:15
So, when it comes to looking at the structure of biological molecules,
00:23
the reference is, of course, to X-ray crystallography. This has been a technique, a wonderful technique, that has worked for decades. So, why do we need a new technique? So, in X-ray crystallography,
00:40
many copies of the molecule are arranged in regular order. And when the crystal is exposed to the X-ray beam, we get a diffraction pattern, and then, with phasing methods, we can get structure determination. And you might remember Max Perutz and John Henry
01:03
wrote a spectacular contribution to the field in 1982 with a model of hemoglobin. And then, that's essentially when proteome X-ray crystallography took off. The X-ray has to be high intensity,
01:22
the crystal has to be almost perfect. And today, more than 100,000 structures solved by X-ray crystallography are in the public data banks. Now, to the downsides. The crystal packing means that the molecules are not visualized
01:44
in all conformations and binding states that are functionally irrelevant. And some molecules, or many molecules, in fact, don't form highly ordered crystals. So, they're not amenable to X-ray crystallography.
02:02
And the sample quantity can be an issue as well. So, this is where electron microscopy steps in. It can be used to solve structures, as I will show you. The projection images are formed at roughly 30,000 magnification.
02:23
And to reconstruction an object, one has to combine multiple angles. The sample must be thin. Electrons are readily absorbed by matter. And one of the downsides is also that
02:44
electrons strongly damage the molecules and there's a need for a very low dose. What it means is that any single image doesn't give us a lot of information. It's salt and pepper.
03:01
So, the minimum should be around 10 to 20 electrons per square angstrom. And with these, one doesn't get a very good statistical definition. So, the weekend before I received this wonderful call,
03:21
I was in Central Park and took a picture of my dog. And strikingly, it shows the relationship between a 3D object and its projection. But in transmission electron microscopy, we don't get just a shadow cast,
03:42
but we get true line integrals. So, instead of the shadow, you have to imagine that you see actually some kind of an outline of the total projection of the dog's head. Now, the pioneers in three-dimensional reconstruction
04:02
were Aaron Klug and David Derosier at the LMB in Cambridge. And they reconstructed the tail of a phage, of a bacteria phage that had high crystal order
04:25
and had almost perfect helical symmetry. Under these conditions, one can gather the information on the total angular range just from a single picture.
04:42
And they made very clever use of the helical symmetry in a way that essentially the math followed the symmetry of the object. In a similar way, Tony Crowther came out with reconstruction of spherical viruses with icosahedral symmetry.
05:06
Again, the math sort of followed the symmetry of the object. More simple, symmetry is the symmetry inherent
05:21
in a crystal, in a two-dimensional crystal. It has to be two-dimensional electron microscopy because, as I said, electrons don't penetrate very deeply. And Richard Henderson and Nigel Unwin came out with a very spectacular work in 1975
05:44
in which they reconstructed a helix bundle of the bacterial rhodopsin for the first time. And they are highly ordered in the purple membrane. And they came out with a way how to collect different tilts
06:04
from different crystals and then combine them all into a three-dimensional reconstruction. And they came out with two additional innovations. One is to reduce those to an absolute minimum,
06:21
even less than one electron per angstrom square. The other innovation is the introduction of glucose embedding where, up to then, negative staining was used. And negative staining is a way of sustaining the molecule
06:46
in a puddle of heavy metal. This is a very poor way of preserving the structure of the molecule, but it was the only way that existed at the time.
07:08
The underlying theorem in reconstruction is the projection theorem or central slice theorem,
07:21
which says that the projection in transmission microscopy is equivalent to a central slice in 3D Fourier space. So in order to fill the 3D Fourier space with information,
07:43
one can essentially collect projections in different directions. Now, my mentor, Walter Hoppe, was the first who posed the question, why didn't we need crystals at all?
08:02
Can't we just image single molecules? And his idea was to put the molecules on a grid, and he used fatty acid synthetase in one instance and ribosome in the other instance. And then he tilted the grid by many angles.
08:24
And this is so-called electron tomography. It's very well understood how to combine all this information. However, when you're doing this this way, you get an accumulated electron of more than a thousand electrons per angstrom square.
08:44
Under these conditions, the molecule sort of burns into cinder, and whatever you see in the end doesn't have any resemblance to the object that you want to study. So essentially, this provoked the ideas in me,
09:08
who was a graduate student at the time. So my ideas really developed in juxtaposition to my mentors, and I give him great credit for this.
09:25
So this is the other approach. The other approach is the 3D reconstruction of single molecules. The idea is that in a solution, after purification from a cell extract, we get obviously millions, billions of copies of the same molecule,
09:47
and they are in random orientation. They're all separate, and why not take advantage of this? So at the time, as I said, we did only have negative staining,
10:01
but this was the only way of doing this. So we get random projections, and the question is how to make use of it. So the concept is simply single-particle techniques. The structural information is collected from images of single molecules.
10:23
Single means unattached, not one molecule only. And the molecules, it's very important, they are free to assume naturally occurring confirmations. Molecules are randomly oriented, so there's easy access to many, many projections,
10:47
and so a single snapshot may already give us a large number of projections in different angles. And as we collect more snapshots, we can gradually fill the entire angular range
11:03
and we just stop when it's convenient. So these ideas were then published. I published them in ultra-microscopy, which had started, just started publishing, and was desperate to get articles,
11:21
so I was sort of lucky to get in this first volume. And so I make a specific reference to Unwin and Henderson, who were able to get seven angstrom resolution. And the article essentially says,
11:41
you know, why not do the same thing with single molecules and had certain ideas on how to align them from my doctoral thesis, the use of cross-correlation function and so forth. So then a year later, there was a commentary in Science,
12:03
and they somehow got wind of this and said, if such methods were to be perfected, then in the words of one scientist, the sky would be the limit. And I only go, he went back to this quotation, prompted by the recent success of the technique.
12:21
So I was really reminded that this is a very good way of characterizing the situation that we have now. No, but at that stage, there were many problems to be solved. There was only the start of the entire work because the concept was very nice, but how to actually get a structure out of this?
12:46
One needed to have a method to align images. Well, I had some clues from the way, from certain results in my doctoral thesis. One needs a method to estimate the resolution of the average or reconstruction.
13:03
That's not straightforward because we are dealing with non-crystalline objects, so they don't have some kind of an inherent pointer to how far the resolution extends in Fourier space. We needed to have a method to sort and classify images
13:22
because the assumption that they all were the same was really only approximate. And then the most important thing is to find the projection angles and then reconstruct in 3D. Well, reconstruct in 3D is not trivial either because all existing methods,
13:42
all existing mathematical techniques made assumptions of regular arrangements of angles. And here we had a completely random arrangement of angles. So this also had to be solved. And faced with a daunting program,
14:02
like a research program like this, it's impossible to just make a big computer program and then sleep overnight and then wake up with new ideas and make the program bigger and bigger and so forth. It just doesn't work.
14:20
Instead, one needs some kind of a workbench. One needs a workbench of programs which are essentially modules. They already provide solutions to something that is sort of easily done. For instance, autocorrelation, cross-correlation,
14:41
Fourier transform, rotate, and so forth. You can very easily come up with 400 operations in this way. And they form modules in what you can then put together in a way of bricks.
15:03
And so this is what I set out to do. And the program was called Spyder. And there was the first specific program for electromicroscopy.
15:21
And so it has the advantage that people who wanted to do similar things could just take advantage of existing routines that didn't have to sort of go very deep into the code or that didn't even have to have mathematical or computational background.
15:45
So images can be aligned with a higher precision as I found out already in my thesis project. So this was sort of a checkmark that was already done. So the proof of concept was a paper in Science in 1981
16:09
that showed 40S subunits of HeLa ribosomes that faced a particular direction, the L-faced projections.
16:26
The subunit was either falling on one side or the other. It's very characteristic for a negative staining which produces preferential orientation. But this paper really showed in a very dramatic way
16:41
what kind of improvements could be gotten by using these kinds of correlation averaging methods for single particles. And so this was really the start of all grant proposals
17:01
that have sort of kept the whole group alive until now. But there is a problem of heterogeneity. What if the assumptions of homogeneity are not fulfilled? You can already see this here in these examples,
17:23
L and R views or flip and flop of the HeLa ribosomes or flip and flop views of the hemocyanin. So we realized very early that aligned images can be considered vectors in an n-dimensional space
17:41
where n is the number of pixels. So for 100 by 100 images, we are now in a 10,000-dimensional space. And if we look for clusters of images, these are simply clusters of these vectors in this high-dimensional space. And they are routine methods that already existed at the time,
18:02
multivariate statistical analysis that it works by reducing the space into an appropriate subspace. So here the subspace is just two-dimensional and immediately shows you the clustering that we were looking for.
18:20
So this was essentially a really big breakthrough and that made people sort of listen because it indicated the whole thing might be feasible. So the other problem that I told you about was finding the angles.
18:41
And we came out with the random conical tilt reconstruction, which again only existed as a concept for a long time, since 1978 in fact. That's when a footnote was published. Here's an overhead that I found recently
19:01
in which the method is sort of sketched out. Then later I got a very fancy description from the people at American Scientists, the professional illustrators.
19:23
And so the idea is that here we have molecules that are essentially facing the grid in exactly the same direction. So what you see here is essentially the same view
19:44
only in different asymptotes. And that doesn't give you any more three-dimensional information. However, the situation changes entirely when we instead now tilt this entire grid with the molecules on top.
20:03
Then we see unique projection directions here. And in the geometry of the molecule, it looks as if the electron microscope had walked around in a cone. And that's why we call this a conical tilt reconstruction method.
20:25
And we call it random because these angles here in fact are not equidistant, they are random as the asymptotes here. So here we have everything known because we have the very large tilt angle known
20:41
which defines the tilt angle of this cone and we know the asymptotes of the angles. So this is the first way that we found in order to collect the entire information. And we call this kind of thing a bootstrap reconstruction because this is not sort of the final word,
21:02
but it's the first way of getting something done even for a completely unknown structure. So with this, we got our first reconstruction in 1986, published it as a short note in Journal of Microscopy,
21:20
impact factor of 0.6 or something like this, and then later on in Ambo Journal. And here is a series of reconstructions, or it's the same reconstruction of the 50S ribosomal subunit of E. coli
21:42
displayed at different thresholds. And the hero here is Michael Rademacher. He was also a student of Walter Hoppe and he was specialized in conical reconstruction geometry. So it was a very fortuitous find.
22:02
So he joined my group in 1982 and then within four years, he managed to develop all the appropriate programs in order to get to this point. So that essentially brought me to the idea,
22:26
when I was asked by the Nobel Museum to bring something that was related to the discovery process, I thought the most appropriate thing was to give them a contour stack
22:40
that was mounted in a foot and frame with correct distances. That was one of the ways how things were displayed at the time. The whole thing weighed 11 pounds and I didn't carry it with me. I sent it ahead of time. And here we are sort of presenting our various objects
23:05
and presenting this. And next to me is Jacques Dubochet with his plunge-freezing apparatus and Richard Henderson with a now atomic model of the bacterial rhodopsin.
23:25
And around that time, the new technique was popularized, a technique that was developed by Jacques Dubochet in 1981 and first applied in 1984 in a spectacular way by Adrian
23:44
to get the structures of viruses. The molecules can be embedded in vitreous ice and the contribution, first of all by Robert Glaser, was that he thought about this way of rapidly freezing
24:07
but he used liquid nitrogen. And the problem with immersing a grid in liquid nitrogen is that air bubbles form at the interface between the sample grid
24:21
and the cryogen. And these bubbles prevent rapid heat transfer and that allows the water to form crystals. Crystals expand and damage the molecule in the process.
24:46
So the big invention by Jacques Dubochet was to replace liquid nitrogen by liquid ethane on liquid nitrogen temperature. In this case, the formation of air bubbles is prevented
25:03
and heat transfer is almost instantaneous and under these conditions, the water becomes vitreous. So it becomes glass-like, almost the same structure as in the liquid water.
25:21
And so this now gave the method that developed a very big boost because now we could really look at molecules in their native states and here are some images obtained in various collaborations with people.
25:41
One is the E. coli ribosome in 1995. For the first time, there was a very detailed structure that had not been seen before. It showed such things as inter-subunit bridges and so forth. Octopus hemocyanin, calcium release channel,
26:02
these were all sort of pioneering contributions at the time with this new technique in various collaborations. And here's another view of the reconstruction of the E. coli ribosome in 1995
26:21
that was published in Nature. There are a number of images that I show you that showed the power of the technique. However, it showed also the limitations that were relatively low resolution. Here, in collaboration with Winter Global,
26:44
we were able to show the translocon attached to the eukaryotic ribosome. Here, in collaboration with Jennifer Doudna, we show how hepatitis C virus invades and hijacks the ribosomal subunit.
27:08
OK, and then there were further breakthroughs because we were able to show that the ribosome changed its conformation in a very radical way in the process of translation.
27:27
So this movement was found to be necessary for the ribosome to perform its work. This inter-subunit rotation, ratchet-like rotation,
27:45
is necessary in order to advance mRNA and tRNA by one codon. So it's really impossible to tell you over all the milestones that happened.
28:01
It was a very interesting, very adventurous path that got us to 2013. And I'm showing you this particular reconstruction because this is the best one that we ever obtained on film.
28:24
So it was the fruit of two years of work, of going through 260,000 images and with things like defocused correction
28:41
and all the difficulties that come with using film. We got stuck at 5.5 angstrom resolution. There was essentially a wall. And that wall was overcome seven years ago
29:01
by the introduction of the first commercial single-electron detection cameras. And they show you here, the MTF shows you that this is a very, very big advantage. And that brought this breakthrough that you now have experienced.
29:26
And in our case, it meant that we got the structure of 2.5 angstrom resolution of the ribosome. We see a water molecule, and I couldn't believe it when I first saw it. After all this drudgery work for 40 years,
29:44
I see a water molecule image by Cryo-EM. It's really unbelievable. And then another very, very beautiful result is that we see multiple structures all in the same sample.
30:06
With maximum likelihood methods, we can extract all the coexisting reconstruction all at once. And these are just examples for what I call story-in-a-sample
30:21
from our recent work. With this, I just wanted to show you where we stand right now. I was always concentrated on ribosomes
30:44
because they were essentially a very good demonstration object for the development of the technique until I discovered that I could actually contribute to the biology. But more recently, many people have been knocking on our doors for collaborations on ion channels, calcium release channels.
31:06
So, very spectacular results have been obtained in the meantime. Then, of course, the method has spread out everywhere. So, in conclusion, single-particle Cryo-EM
31:22
really opens a new era in structural biology. There's no need for crystals. Very small sample quantity is needed. Resolutions in the 3-4 ernstron range are now routinely achievable. Best resolution are in the range of two ernstrons.
31:43
Multiple structures retrieved from the same sample, which means that we get clues on function. The molecules are enclosed to native conditions. They really have functional significance. And the solving of structures of membrane proteins
32:01
is much easier than with X-ray crystallography. And there is going to be a huge expansion of the structural database relevant for molecular medicine. And we just see the beginning of this. At the end, I'd like to thank my wife, Carol Sargano,
32:25
for her loving, long-lasting support and to many postdocs, students, colleagues for their hard work and for contributing brilliant ideas and to many collaborators for their enthusiasm but also for their patience. Thank you very much.