Cell Motility and the Cytoskeleton

Transcript of Part 3: Evolution of a Dynamic Cytoskeleton

00:08.0	Hello.  I'm Julie Theriot from Stanford University,
00:11.1	and I am back for the third part of my iBio Seminars presentation.
00:15.0	Now, in the first two parts,
00:16.3	I focused on well-understood systems
00:18.3	with strong experimental evidence
00:20.2	supporting all of the conclusions that I was telling you.
00:22.3	For the third part,
00:24.1	I'm going to venture into the field of wild speculation,
00:27.1	and in fact I'm going to wildly speculate
00:28.3	about the biggest questions of all
00:31.1	in the context of biology,
00:33.1	which is how things evolve.
00:35.1	And in particular what I'd like to discuss
00:38.1	is some big picture ideas
00:40.2	about how the dynamic cytoskeleton
00:42.2	has evolved throughout the history of life on Earth
00:45.0	in order to give us such an incredibly diverse array
00:48.1	of cell types, of structures,
00:50.1	and of behaviors of cells.
00:52.2	And just to illustrate that,
00:54.1	here on the left I have an image
00:56.2	of one of the most fabulous of protozoa,
00:58.2	called Stentor,
00:59.3	which is an absolutely enormous cell
01:02.0	with, as you see, a great deal of very detailed subcellular specialization
01:04.1	in terms of its structure.
01:06.1	And then on the right,
01:08.0	a luridly colored electron micrograph
01:09.1	of some bacteria,
01:10.3	which are quite diverse in terms of their
01:13.3	molecular components,
01:15.2	but fairly similar and fairly simple
01:17.2	in terms of their shape.
01:20.0	Now, one of the things that's become very clear
01:22.0	in the modern era of sequencing
01:25.2	is confirming something that of course Darwin had
01:29.1	originally speculated about some time ago,
01:31.2	which is the proposition that
01:33.2	all of the organisms that are currently living on our planet
01:35.1	are descended from a common cellular ancestor
01:38.0	or a common pool of cellular ancestors.
01:41.1	And certainly, one of the most undeniable pieces of evidence
01:44.0	in support of that hypothesis
01:45.1	is the observation that there's
01:47.1	a tremendous number of molecular components
01:49.3	that are found highly conserved
01:51.2	throughout all branches of life.
01:54.1	This particular illustration
01:56.3	shows the sequence similarity among ribosomal RNA
02:00.1	for bacteria,
02:02.2	for archaea,
02:03.2	and for eucaryotes (eukaryotes).
02:05.1	And the fact that it is possible
02:07.1	to arrange all of the known living cells on Earth
02:09.2	in this sort of tree,
02:11.3	where everything looks related to everything else
02:13.2	is a very good indication that in fact
02:17.0	all of the ribosomal RNAs
02:18.2	that exist in cells on the modern Earth
02:20.2	all came from some common ancestor
02:22.3	back at some distant point in the past.
02:25.2	The common ancestor on this diagram
02:27.1	is indicated with the green circle,
02:28.3	and it's fondly called "LUCA",
02:30.1	which stands for the last universal common ancestor.
02:32.2	Now, this was not the first cell,
02:34.1	this was just the last cell
02:36.1	that gave rise to everything that's living today,
02:38.1	and it was already very complex.
02:40.0	It already had ribosomes, it already had DNA,
02:42.2	it already had most of the components of central metabolism,
02:45.1	including, for example,
02:46.2	all of the enzymes involved in glycolysis,
02:48.2	and it also had cytoskeletal components.
02:51.1	And we know this because
02:53.1	those common components
02:54.2	are found in every single branch
02:57.2	of currently living organisms.
02:59.3	Now, given that everybody started off
03:02.0	as part of the same family,
03:03.1	it is interesting, I think,
03:05.1	to speculate about why it is that
03:08.1	some parts of this tree of life
03:10.0	have ended up behaving so differently
03:12.1	from other parts of the tree of life,
03:14.0	and in particular I'd like to draw the distinction
03:16.3	between what I will loosely call the prokaryotes,
03:19.2	which are on the top part of the tree,
03:21.1	and that includes the bacteria and the archaea,
03:24.1	which are very distinct branches,
03:26.0	versus the eukaryotes on the bottom part of the tree.
03:28.3	Now, although bacteria and archaea
03:31.0	are no more closely related to each other
03:33.1	than they are to the eukaryotes,
03:34.3	they nonetheless are quite similar morphologically,
03:38.1	and in fact we didn't even know that the archaea
03:41.1	existed as a separate kingdom until sequencing data started coming along,
03:44.0	because if you just look down a microscope,
03:45.2	it's very hard to tell the difference
03:47.0	between a bacterium and an archaeon.
03:49.2	Whereas the morphology is very simple
03:51.3	in those prokaryotes, the metabolism is extremely diverse,
03:55.0	so almost any interesting chemistry
03:57.2	that's done in living organisms
03:59.1	was first invented by the prokaryotes.
04:02.1	So, what we got going for us down on the eukaryotic side of the tree?
04:05.2	Well, although our chemistry is not
04:07.3	necessarily as interesting or as complex
04:09.1	as what's going on with the little guys,
04:11.1	we have morphology.
04:12.2	We get very big,
04:13.2	we get very complicated,
04:15.0	and we get very multicellular.
04:16.2	So, within this context of everybody
04:18.3	being related to everybody else,
04:20.1	and in fact all of these different organisms
04:23.0	having something that looks like cytoskeletal proteins,
04:25.1	I think it becomes a very interesting question to ask,
04:28.0	well, what is it really then that makes eukaryotes different from prokaryotes?
04:31.0	Specifically, what is it that gives eukaryotes
04:33.0	the ability to have such extremely diverse
04:35.1	and complicated cellular morphologies,
04:37.0	based on cytoskeletal structures?
04:41.0	Just to, again,
04:42.3	give you a little bit of a taste
04:44.2	of the kind of diversity you find
04:46.1	even within a very closely related group of eukaryotes,
04:48.2	these are some illustrations
04:50.0	from one of the great microscopists of the 19th century,
04:52.3	just showing a group of fairly closely related
04:55.1	ciliates and flagellates.
04:56.2	Each of these are unicellular organisms...
04:58.1	they're not all drawn to scale with each other...
05:00.0	this is Stentor,
05:01.2	that I showed you a photograph of on the very first slide,
05:03.1	and a number of other familiar organisms,
05:06.0	including, for example, Paramecium, here,
05:08.3	and Giardia.
05:10.3	And all of these cells,
05:12.2	to a first approximation,
05:14.0	have the same components.
05:15.1	If you look at the protein components of these cells,
05:17.2	there certainly are things that are a little bit different between them,
05:20.0	but the major components,
05:21.1	certainly the major structural components,
05:22.3	are pretty well conserved,
05:24.0	and yet somehow all these different organisms
05:25.3	are able to make drastically different cell shapes
05:28.0	with very different motile behaviors,
05:29.2	very different kinds of division behaviors,
05:31.1	simply by rearranging this common set
05:33.2	of molecular components.
05:37.0	Now, if we contrast that to the bacteria,
05:39.2	you certainly will hear
05:41.2	proponents of the prokaryotic side of the tree
05:45.0	claiming that there is a lot of morphological diversity among bacteria...
05:48.2	and to a small extent that's true.
05:50.1	So, if we look at this illustration
05:54.1	from a really enchanting review by Kevin Young that came out a few years ago,
05:57.3	this shows essentially the diversity
06:00.2	of the different kinds of shapes
06:02.1	that have been described for isolated bacterial species.
06:04.3	And here in the zoomed-in version,
06:07.1	we find a couple of the fairly familiar shapes,
06:09.2	so we see things like E. coli here,
06:11.3	which is a rod shape,
06:13.2	there are some crescent-shaped bacteria here like Caulobacter,
06:17.1	there are helical-shaped bacteria,
06:19.2	but they're all sort of fairly simple variations on a theme.
06:24.2	And then all of those guys are also quite small.
06:27.2	And then we zoom out to higher magnification...
06:29.1	there's really very few bacterial species
06:30.3	that get much bigger than just a few microns
06:33.1	in characteristic size.
06:34.2	One interesting exception
06:36.2	is illustrated here by this yellow dome
06:39.0	that stretches across this drawing,
06:40.3	and this is actually to scale
06:42.2	with the other bacteria here,
06:44.1	an illustration of an organism called Thiomargarita,
06:47.0	which is one of the largest bacteria know,
06:49.0	and these guys actually can be almost a millimeter across.
06:52.2	And this shows a photograph
06:54.3	of Thiomargarita next to a Drosophila
06:58.1	#NAME?
06:59.2	and you can see its absolutely enormous size.
07:02.0	But the trick here is that
07:04.1	even though the cell itself is very large,
07:06.0	its cytoplasm is very small.
07:07.2	Thiomargarita is basically
07:09.1	just a hollow shell of a very thin layer,
07:11.3	a couple of microns thick of cytoplasm,
07:13.3	that's stretched an enormous vacuole.
07:16.1	So, even when bacteria get to be pretty large
07:18.2	and, relative for bacteria,
07:20.0	morphologically complex,
07:21.2	they still have diffusion-limited dimensions.
07:24.1	And it is obviously observable that,
07:28.0	overall, if you take a random eukaryote
07:29.2	and compare it to a random prokaryote,
07:31.0	the eukaryotic cell is bigger
07:33.0	and is much more structurally complex.
07:35.2	So, this just pulls out a couple of those things
07:39.0	as examples.
07:40.1	Here's Paramecium compared to a bacterium to scale,
07:43.0	and then that bacterium blown up,
07:44.2	and here I've chose Vibrio cholerae.
07:46.2	And you can see, again, there is subcellular localization,
07:49.1	subcellular specialization
07:51.0	in the context of Vibrio cholerae.
07:52.1	It's got its flagellum growing out of one pole, for example,
07:55.1	so it somehow is able to know
07:57.1	which end is which
07:58.3	and know the difference between the poles and the sides,
08:00.1	but even at that level the complexity
08:02.2	is just vastly less than what you see in something like Paramecium.
08:06.1	So, if you think about,
08:08.1	what are all the observable differences
08:10.0	between eukaryotes and prokaryotes,
08:11.2	there's a pretty long list that is...
08:14.0	you know, there are some exceptions,
08:15.1	but this is reasonable typical,
08:17.0	reasonably representative of these different kinds of life.
08:19.3	Eukaryotes have, of course,
08:21.1	a membrane-enclosed nucleus,
08:22.3	that's the definition.
08:24.1	And so that means they have separated,
08:25.3	spatially and temporally,
08:27.1	the process of transcription
08:28.3	from the process of translation,
08:30.1	which gives rise to many options
08:32.1	for very complicated regulation
08:33.3	that basically can't be present
08:35.1	in bacteria or archaea.
08:38.0	Along with the membrane...
08:41.1	err, sorry, the nuclear membrane,
08:43.2	the eukaryotes of course also have lots of other internal membranes
08:45.2	making things like the Golgi apparatus
08:47.1	and the endoplasmic reticulum.
08:48.2	They typically have much, much larger genomes
08:51.1	than our bacterial friends,
08:53.0	with often with multiple chromosomes that can be very large.
08:56.2	As I mentioned, they have much larger cell size.
08:58.1	Instead of being characteristically
09:00.1	just a few microns long,
09:01.2	they're usually a few tens to even hundreds of microns in size,
09:05.1	and they have a very high degree of subcellular compartmentalization.
09:08.2	In addition, eukaryotes have endosymbionts,
09:11.3	so they have mitochondria and chloroplasts
09:13.1	that used to be bacteria
09:15.2	that have now been captured and domesticated
09:17.2	by the eukaryotic cells
09:19.1	in order to perform energy functions.
09:21.0	And then, finally, you know,
09:23.0	the reason we're even able to talk about this
09:25.0	is because eukaryotes make multicellular organisms.
09:27.1	Now again,
09:29.1	just as in the context of the shape,
09:30.3	it's not the case that bacteria
09:32.2	never make multicellular organisms,
09:34.0	it's just that they're not very complicated.
09:35.2	So, for example,
09:37.3	the best understood multicellular bacterial organism
09:39.3	is this thing called a stromatolite.
09:42.1	We see a bunch of stromatolies here,
09:43.3	growing in an ocean next to Australia,
09:45.2	and these are basically big piles of colonial cyanobacteria.
09:50.2	They grow and deposit
09:52.0	some extracellular matrix,
09:53.2	and then the next generation will grow on top of it,
09:55.1	making these very characteristic layer cake patterns
09:58.2	that are seen not only in the kinds of stromatolites
10:00.2	we have growing today
10:02.0	but also are found in stromatolites
10:03.2	that are several billion years old in the fossil record.
10:06.2	And in all that time,
10:08.1	look how far eukaryotes have come,
10:09.3	but bacteria are basically still making the same structures.
10:12.2	Eukaryotes have been able to make things like Redwood trees,
10:15.2	they're been able to make things like coral reefs,
10:17.2	with this extraordinarily morphological diversification.
10:21.1	Okay, so this very fundamental question of,
10:23.3	what is the difference between eukaryotes and prokaryotes
10:26.0	that gives us such extraordinary morphological complexity,
10:30.1	is a question that I would say, today,
10:31.2	we don't know the answer to,
10:33.1	but a few decades ago
10:34.3	we thought we did know the answer to it,
10:36.1	and before 1990 or so
10:38.2	the nearly universally accepted explanation
10:41.2	for why there's such big morphological differences
10:43.1	came down to the eukaryotic cytoskeleton.
10:45.2	If we go through all of those characteristics I described
10:49.1	about what's special about eukaryotes
10:50.2	as compared to bacteria or archaea,
10:52.0	all of those things can be attributed
10:54.0	to functions of the cytoskeleton.
10:55.2	So, for example,
10:57.1	the existence of the membrane-enclosed nucleus
10:59.1	and all of the complex internal membrane systems
11:01.1	can be ascribed as being due to
11:03.3	intracellular membrane transport
11:05.3	with motor proteins, for example,
11:07.2	pulling vesicles along microtubules.
11:10.1	Nuclear lamins,
11:11.3	which are essentially a form of intermediate filaments,
11:13.2	stabilize the nuclear membrane.
11:16.2	The expanded genome in eukaryotes
11:18.3	of course divided by the extraordinary machine
11:21.1	of the mitotic spindle,
11:22.2	which with all of its self-organized microtubules,
11:25.2	and its massive amount of flux
11:28.2	and force generated by motor proteins
11:31.0	such as kinesins and dyneins,
11:32.2	is able to divide a lot of chromosomes,
11:34.2	and it's no big deal for a eukaryote
11:36.2	to generate an extra chromosome,
11:38.0	split a chromosome in half,
11:39.2	vastly expand the size of its genome,
11:41.2	because the mitotic spindle
11:43.1	is able to handle a lot of capacity
11:46.0	in terms of segregating chromosomes.
11:48.2	The much larger cell size of eukaryotes
11:51.1	can be attributed to the fact
11:53.0	that they're no longer diffusion-limited.
11:54.2	Because of the direct action of motor proteins,
11:56.2	you can actually have transport and mixing
11:58.2	within the cytoplasmic compartment.
12:01.2	Similarly, the subcellular compartmentalization,
12:03.3	the existence of endosymbionts,
12:05.2	and even the ability to have cells come together
12:08.2	and make multicellular organisms
12:09.3	with very complex extracellular matrices,
12:12.2	largely involves coupling between microtubules, actin, intermediate filaments,
12:17.3	and other components within the cell
12:21.2	or from one cell to another.
12:22.3	Endosymbionts, for example,
12:25.0	are thought to be the product of a phagocytic event,
12:27.2	where an ancient eukaryotic ancestor
12:29.2	ate a cyanobacterium
12:31.2	and turned it into a chloroplast.
12:35.1	Now, the reason I say we don't know the answer today,
12:37.0	even though we did know the answer in 1990,
12:39.1	was because it became clear in the 1990s
12:41.3	that bacteria actually do have a cytoskeleton
12:44.1	that in many ways is remarkably similar to that
12:46.3	of eukaryotes.
12:48.1	And so the first real nail in this coffin
12:50.1	was the discovery of
12:52.1	a tubulin homologue in bacteria
12:53.3	that goes by the name of FtsZ.
12:56.1	FtsZ had been originally identified genetically
12:59.2	as being a protein that was required for cell division,
13:01.2	and some beautiful work in the 1990s,
13:03.2	combining biochemistry and cell biology of this protein,
13:07.0	has established that it really acts as
13:09.2	a legitimate cytoskeletal protein in bacteria
13:11.2	that's involved in cell division.
13:13.1	It forms a ring right around the middle of a bacterium
13:15.2	that's about to divide,
13:17.0	and as the bacterium divides,
13:18.1	that ring actually becomes smaller.
13:20.2	Furthermore, the purified protein
13:22.1	is able to self-assemble into filaments
13:24.0	in a manner that depends on the presence of hydrolyzable GTP,
13:27.2	which again is a very cytoskeletal behavior.
13:30.1	And any remaining doubt
13:32.1	had to be squashed
13:34.0	when the crystal structures came out
13:35.2	for both tubulin and FtsZ
13:38.1	at about the same time,
13:39.2	back to back in the same issue of Nature,
13:41.3	and it was clear that the structure of one
13:43.3	could be almost perfectly superimposed
13:45.3	on the structure of the other.
13:47.2	So, it's clear that these are not only sort of analogous proteins,
13:50.1	but in fact true homologues,
13:51.3	in the sense that they are descended from a common ancestor,
13:53.3	and by far the simplest explanation for that
13:56.1	is that the LUCA, the last universal common ancestor,
13:59.2	also had a protein like tubulin or FtsZ.
14:04.1	Okay, so bacteria have something like tubulin...
14:06.2	then it turned out a couple of years later,
14:08.2	bacteria also have something like actin,
14:11.1	and just like with FtsZ,
14:13.1	these proteins
14:15.0	-- they go by several names, there's MreB and Mbl
14:17.1	and a few other family members --
14:18.3	were originally described as being genes that are necessary
14:21.0	for some aspect of bacterial shape,
14:23.1	a fundamental cytoskeletal-type function.
14:26.0	And so, for example in Bacillus subtilis,
14:28.0	disruption of any of these actin homologues
14:30.0	causes really abnormal, overall,
14:32.2	types of cell shapes.
14:34.3	And these proteins, when purified,
14:36.3	can assemble into filaments,
14:38.1	again like FtsZ or like any of the eukaryotic cytoskeletal proteins.
14:42.2	Now, one thing that's I think been particularly amusing
14:45.2	in the context of the bacterial actin homologues,
14:48.1	has been the proliferation
14:50.1	of such a large number of them
14:52.0	that appear to have slightly different specialized functions.
14:54.3	So, for example,
14:56.1	one of my favorite examples
14:58.0	is in magnetotactic bacteria
14:59.2	that are able to orient themselves with respect to the Earth's magnetic field.
15:03.0	They do so my lining up these little crystals of magnetite
15:05.3	along a structural element inside the cell
15:10.1	that is in fact made up of an actin-like filament.
15:14.0	Now, this is not the same actin-like filament
15:16.2	that is contributing to cell shape,
15:18.0	it's actually a copy of the gene
15:20.1	that was historically duplicated,
15:22.1	and then diverged for this specialized function.
15:25.0	So, bacteria not only have actin,
15:26.2	they have a bunch of different actins,
15:28.0	but they can use them both to determine their overall shape
15:30.2	and to determine the distribution
15:33.0	of intracellular organelles.
15:35.2	So now I think we're set up to ask ourselves
15:37.2	a slightly different question.
15:38.2	It's clear that bacteria do have a cytoskeleton,
15:41.1	so the existence of the cytoskeleton
15:42.3	cannot be the thing
15:45.0	that distinguishes eukaryotes from prokaryotes.
15:46.2	So, we might ask then, well,
15:48.0	if bacteria do have a cytoskeleton,
15:49.2	then why don't they do something more interesting with it?
15:51.1	And by that I mean in this very eukaryotic-centric world view,
15:54.2	why don't they have morphological complexity
15:57.0	and why don't they make
15:59.0	big, fancy multicellular organisms?
16:01.0	So, here we get to the point of wild speculation,
16:03.2	and my specific hypothesis is that
16:07.2	the thing that all of the cytoskeletons have in common
16:10.0	is the ability to perform large-scale cell organization
16:13.1	based on self-assembly of helical filaments
16:15.2	that are highly dynamic.
16:17.0	That's true among all the different branches of life.
16:21.3	I am proposing that was is special about eukaryotes
16:24.2	is not the cytoskeletal filaments per se,
16:26.3	but rather two specific classes
16:29.0	of cytoskeletal associated proteins:
16:31.1	the nucleators and the molecular motors.
16:35.1	And I'll go into some detail about why I think these are the things
16:38.2	most likely to really be different,
16:40.0	but so far there is no strong positive evidence
16:43.0	that prokaryotes have
16:45.2	any classical molecular motor proteins
16:47.0	or any regulated nucleators for cytoskeletal filaments,
16:50.0	which raises a second question.
16:52.1	You know, if the prokaryotes don't have them,
16:53.3	why don't they have them?
16:55.0	That's something even I'm not going to be able to speculate about,
16:57.3	but as far as why I think these two things
17:00.0	are really the key to this morphological difference,
17:02.1	I'll give you some of the information
17:04.2	that has led me to that proposition,
17:06.2	and suggest a few very specific ways
17:10.0	that these ideas could be tested.
17:13.2	Okay, so the overall basis for this hypothesis,
17:17.2	that you need nucleators and molecular motor proteins
17:19.2	in order to get morphological diversity,
17:22.2	comes from a series of observations
17:24.2	about the kinds of structures
17:26.2	that can be made by cytoskeletal filaments
17:28.0	in bacteria versus in eukaryotes.
17:32.1	In particular,
17:34.1	any self-assembling helical filament
17:37.0	can make either a structure with a single filament
17:40.1	or it can make a bundle or a raft of filaments,
17:44.1	where the filaments are oriented in random directions
17:46.1	with respect to each other.
17:47.2	All that requires is helical self-assembly,
17:49.2	which as we'll see is a very universal feature
17:52.0	of proteins that are able to interact with themselves.
17:56.1	On the other hand, the kinds of structures that really require localized nucleation
18:00.1	or localized motor activity
18:01.2	are illustrated here on the bottom.
18:03.0	These include things like microtubule asters,
18:04.3	which are familiar from the organization
18:07.0	of overall organelles inside of a cell.
18:10.2	These include things like parallel bundles,
18:12.1	which we find inside of eukaryotes flagella,
18:13.3	or we find in muscle...
18:16.0	you know, the sarcomeres that are able to give large-scale muscle contraction...
18:19.3	and also things that have bipolar morphology,
18:23.0	like this mitotic spindle shown here.
18:25.1	We know enough about how these different structures are formed
18:27.1	that I feel fairly confident in stating
18:30.3	that the easiest way to make these kinds of structures
18:32.3	is either by localizing nucleation of filament growth,
18:36.1	or having molecular motor proteins
18:38.2	that are able to sort of filaments of different orientations,
18:41.1	or some combination of both things.
18:43.2	And what I will assert is that
18:46.2	no structures of the type shown under here in class B
18:49.2	have yet been found in bacteria,
18:51.3	specifically have yet been found in the cytoplasm of bacteria,
18:54.2	with perhaps one or two interesting exceptions
18:57.0	that I think actually sort of prove this rule.
18:59.3	So, if nucleators and molecular motor proteins
19:02.1	essentially enable you to take your
19:05.0	helical, self-polymerizing, self-assembling
19:07.2	cytoskeletal filaments
19:08.2	and make large structures out of them,
19:10.2	you might ask, well,
19:12.1	what evidence is there that eukaryotes can do this
19:14.2	but prokaryotes cannot?
19:16.2	So, I'd like to now again take a step back
19:19.0	and say, if we are willing to accept the premise
19:21.3	that something about the activity
19:23.2	of the eukaryotic cytoskeleton
19:24.3	is the key difference,
19:26.2	then the question is, what is so special
19:28.1	about the eukaryotic cytoskeleton?
19:30.0	And if you look at the most abundant
19:32.0	and most studied eukaryotic cytoskeletal proteins
19:34.3	-- we have tubulin making microtubules,
19:36.2	actin monomers making actin filaments,
19:39.2	and then intermediate filaments
19:40.3	made up of their own subunits --
19:42.1	and what you immediately see,
19:44.0	looking at all these things,
19:45.1	is that they're all basically helical self-assembled structures,
19:49.3	but I would argue that helicity
19:52.2	of a self-assembled structure
19:54.1	is not something that's particularly difficult to evolve.
19:56.2	And in fact there is a very strong argument
19:58.2	made first by Crane
20:01.1	and then by Pauling, back in the 1950s,
20:03.1	that pretty much any protein
20:05.1	that has any tendency to associate with itself
20:08.1	is more likely to form a helix than it is to form anything else.
20:12.0	And the argument is pretty simple:
20:13.3	if you have a globular protein
20:16.2	where every aspect of that protein's surface
20:18.0	has slightly different physical attributes
20:20.2	-- slightly different charge distribution,
20:22.1	slightly different shape --
20:24.2	and that protein has some tendency
20:28.0	to interact with a second copy of itself
20:30.0	in some particular orientation
20:31.2	where, you know, part A of one protein
20:33.2	binds to part B of another protein...
20:36.0	if the protein is going to make a dimer,
20:37.2	you know, those two things that interact with each other
20:40.2	have to be distributed symmetrically on the protein.
20:42.2	But if the part A and part B
20:44.2	are in any orientation
20:46.2	other than just directly opposite each other,
20:49.0	then essentially what's going to happen
20:50.2	is those two subunits will come together
20:52.2	and there will be an interaction between the two
20:55.1	that will give a dimer that's got,
20:57.1	you know, some slightly off-center asymmetrical structure
21:00.1	where, if part A and part B are engaged here,
21:03.1	this subunit still has a part B available
21:05.1	and this subunit still have a part A available.
21:08.1	Now, if you think that through, as these guys did,
21:10.1	what you see is that you can then
21:12.1	add another subunit on this side,
21:13.3	you can add another subunit on that side,
21:15.2	and you're able to make helices
21:17.1	out of this very simple binary interaction rule.
21:20.2	And in this really creative
21:23.0	and beautiful illustration
21:24.1	done in this original paper
21:26.3	using little matchboxes,
21:28.0	you can see you can actually build all sorts of different kinds of helices
21:30.2	just by slightly changing
21:32.2	the orientation of the interaction between those two subunits.
21:36.1	So, making a helix is not a big deal,
21:39.0	and in fact a really great piece of evidence
21:41.3	to show that that's true
21:44.0	is the effect of the sickle cell mutation on hemoglobin.
21:47.0	Now, hemoglobin is highly selected to be very, very soluble
21:50.2	#NAME?
21:52.2	can get up to on the order of 300 mg/mL --
21:55.2	but just one mutation in hemoglobin
21:58.1	can give it the tendency to precipitate.
22:00.3	And when it precipitates
22:03.0	it forms these incredible looking
22:05.1	helical filaments that look, really,
22:07.2	almost exactly like microtubules.
22:09.2	If I told you this was a microtubule,
22:11.0	you would probably even believe me,
22:12.1	but it's actually a polymer of sickle cell hemoglobin.
22:15.2	So, when you take a very soluble protein
22:18.1	and you make something go just a little bit wrong
22:20.1	with one part of its surface
22:21.3	so that it has a very minor tendency to self-associate,
22:24.1	the consequence is you get a helix.
22:27.2	Okay, so forming a helix
22:29.1	is not a special property...
22:31.1	what is it about cytoskeletal filaments
22:32.2	that gives them their ability to give us morphology?
22:35.2	Well, one other important thing
22:37.3	that is true of all of the eukaryotic cytoskeletal filaments
22:40.3	is that they have this very interesting dichotomy
22:44.3	that they have to be stable in order to have physical strength,
22:48.0	but they also have to be unstable
22:50.1	in order to allow the cell to change shape
22:52.1	or to move or to rearrange its elements
22:54.2	as it's dividing.
22:56.0	And at least for actin
22:58.1	and for tubulin,
23:00.0	part of the reason that they're able to
23:02.0	exhibit this dichotomy
23:04.0	is because those subunits are also able to bind and hydrolyze nucleotide --
23:07.2	ATP in the case of actin,
23:09.1	GTP in the case of tubulin.
23:11.2	And as you see in this illustration here,
23:13.3	the fact that you can have
23:15.3	different conformational states of a protein
23:17.1	based on the identity of the nucleotide that's bound to it,
23:21.1	can set up a situation where you can have one conformational state,
23:24.3	in this case illustrated as being the thing binding ATP,
23:27.3	which is a conformation appropriate
23:30.2	for forming one of these helical filaments,
23:32.2	and you can have an alternative conformational state,
23:35.1	which is shown here in this case
23:36.3	as being the nucleotide free state,
23:39.0	but it could also be the ADP state,
23:40.2	that will tend to disassemble.
23:42.2	And so having that switch
23:44.2	between an assembly-competent form
23:46.2	and an assembly-incompetent form
23:49.0	can enable the dynamics
23:51.0	that are characteristic of eukaryotic cytoskeletal filaments,
23:53.1	and in fact it's very interesting
23:55.1	that actin itself
23:57.1	is a close structural relative of hexokinase,
24:00.0	the glycolytic enzyme which is best characterized
24:02.0	with respect to its very large conformational change
24:04.1	on binding and hydrolysis of substrate.
24:09.1	And in fact it's true
24:11.3	that if you look at the cytoskeletal elements of eukaryotic cell,
24:15.0	they're all highly dynamic,
24:16.2	and this is something that's been appreciated
24:18.1	for many decades in the context of microtubules,
24:20.2	actin, and intermediate filaments...
24:22.1	all of them turn over very fast inside of cells.
24:24.2	And in fact for things that require
24:28.2	the cytoskeletal filaments, like for example forming the mitotic spindle,
24:31.1	if you inhibit either the assembly
24:34.2	or the disassembly of the microtubules
24:37.3	in a mitotic spindle,
24:39.1	the net effect is the same in both cases,
24:41.1	which is that the cell fails to divide.
24:43.3	So it's not just the helical structure,
24:45.2	but also the turnover is clearly key.
24:47.3	Okay, well let's turn back to the prokaryotes --
24:51.2	do they have dynamics associated with their cytoskeletal filaments?
24:54.3	Well, actually, as it turns out, they do.
24:56.2	So, as I mentioned FtsZ,
24:58.2	that's a close structural homolog of tubulin,
25:00.1	also binds and hydrolyzes GTP,
25:02.3	and in vivo turns over very, very rapidly,
25:05.2	as can be seen in this photobleaching study.
25:08.2	So, here there are two cells
25:10.1	that have assembled FtsZ rings.
25:12.2	Here, one half of one of these rings
25:14.1	has been bleached out,
25:15.2	and you can see, actually within just a few seconds,
25:18.1	the ring is able to come back.
25:20.1	And this particular experiment
25:23.2	also had a really, really nice direct demonstration
25:28.2	that those dynamics are associated with nucleotide hydrolysis,
25:31.3	which was they were able to compare the dynamics of wild type FtsZ protein
25:34.1	with an FtsZ protein
25:36.1	that had a point mutation that slowed down GTP hydrolysis,
25:39.0	and looking at the turnover of the wild type protein
25:41.2	versus the turnover of the slower protein,
25:45.1	in vivo,
25:46.1	they were able to see that there's this direct correlation
25:48.1	between how fast the filaments turn over
25:50.2	and their ability to hydrolyze GTP.
25:53.2	Okay, so the prokaryotic filaments,
25:55.1	they do the same kinds of dynamics
25:57.1	as the eukaryotic ones do,
25:58.2	so it's not the helical structure,
26:00.0	it's not the dynamics in terms of turnover,
26:01.2	and it's not even fancy dynamics.
26:03.3	So, microtubules, for example,
26:05.1	famously do this dynamic instability,
26:07.1	where even in a uniform chemical environment
26:10.1	an individual microtubule end will grow for a while
26:12.0	and then shrink for a while,
26:13.1	and then grow for a while and then shrink for a while,
26:15.3	and it's not at all obvious why this should be possible.
26:18.1	This is something that people have been studying for many years
26:21.1	and we have a pretty good overall chemical understanding
26:23.1	of how this can happen,
26:25.0	but it's definitely dependent on
26:28.1	the ability of the structures to hydrolyze nucleotide
26:30.2	and change their conformational states.
26:33.2	So, this illustration up top
26:36.0	shows the classic direct demonstration
26:37.3	of this dynamic instability on microtubules,
26:39.3	it was done in 1986,
26:41.2	and then almost 20 years later
26:44.1	another group was able to show with one of the bacterial actin homologs,
26:47.0	this one called ParM,
26:49.0	that they were able to see exactly the same kinds of behaviors
26:51.0	of these filaments growing for a while
26:52.2	and then shrinking for a while,
26:54.1	even when the chemical environment is uniform.
26:57.0	Okay, so it's not the helicity,
26:58.2	it's not the ability to hydrolyze nucleotide,
27:00.3	it's not the ability to turnover rapidly in vivo,
27:03.1	and it's not even the ability to do fancy things
27:06.1	like dynamic instability.
27:08.2	What it is, then, that's different?
27:10.1	Well, looking at bacterial cells,
27:12.1	it's clear that the way they're able
27:14.2	to get some sort of overall cellular organization
27:16.1	definitely depends
27:18.1	on this tendency of proteins
27:19.3	to make helical filaments.
27:21.1	And in fact if you look at bacteria
27:22.2	and you look at different aspects of their structures,
27:24.2	what you see is the helices are all over the place,
27:27.2	and you can make lots of different kinds of helices,
27:29.2	ranging from different things like the overall helical pattern of a spirochete
27:32.3	to the twisting helices associated with bacterial flagella,
27:36.2	to even the much more, sort of uniform helix
27:42.0	associated with something like a bacterial pilus.
27:45.2	So, it's clear it's not just the classical cytoskeletal proteins in bacteria
27:49.2	that are able to do this helical self-assembly trick,
27:51.1	but in fact lots of other proteins as well.
27:55.2	And looking specifically at the cytoskeletal proteins,
27:59.1	I think one interesting example
28:01.1	is the bacterium Caulobacter crescentus,
28:03.0	which is this very cute little banana shape
28:05.1	with structural differentiation on its two ends.
28:08.1	It's got not just actin and tubulin,
28:10.3	but it also has something that looks very much like
28:13.3	an intermediate filament,
28:14.3	and each of those things assumes slightly different helical patterns
28:17.2	inside of a growing cell.
28:19.1	And mutations in either the actin homolog,
28:21.1	the tubulin homolog,
28:22.2	or the thing that looks like an intermediate filament
28:24.1	will affect the overall shape of the cell
28:26.2	in very different ways.
28:28.1	So, lots of helical organization,
28:30.3	lots of function associated with helical organization,
28:33.2	but still no great morphological diversification.
28:37.3	Now, like everything else in biology,
28:39.3	there's going to be some exceptions,
28:41.1	and there are some bacteria that have really,
28:42.3	particularly spectacular shapes,
28:44.0	and this just shows a couple of my favorites.
28:45.2	This is a star-shaped bacteria called Stella humosa.
28:48.3	It's flat, but it has either five or six points
28:51.1	depending on what stage of its cell cycle it's at.
28:54.2	This is another weird-looking flat prokaryote
28:58.1	that in this case is an archaeon.
29:00.2	This is called Haloquadratu walsbyii
29:02.3	and it was first cultured in pure form
29:05.1	only a few years ago,
29:06.3	and this is actually flat like a floor tile,
29:08.3	and when it divides
29:11.0	it divides across the middle and then again like that,
29:13.0	so you have one floor tile going into four smaller ones.
29:16.1	And then there are things like this,
29:17.2	this is Epulopiscium fishelsonii,
29:19.2	which competes with Thiomargarita
29:21.1	as being one of the biggest bacteria known,
29:22.3	and this is absolutely huge,
29:24.1	it can be half a micron long,
29:26.1	and it has many copies of its genome
29:28.0	that are actually distributed
29:29.2	in a very regular pattern
29:31.1	throughout the entire cell,
29:32.2	and it moves its genome around depending on what time of day it is.
29:35.1	So, I think for things like this,
29:36.3	there's gotta be something else going on
29:39.0	besides just our usual bacterial helical self-organization,
29:41.2	and I think it's an interesting challenge
29:43.1	to now try to go into these organisms
29:45.1	and find out what it is that's determining their ability
29:47.2	to have these very specific morphologies.
29:51.2	Okay, but getting back to the main thread,
29:53.2	I've been arguing that
29:56.1	you can have simple helical structures
29:58.1	without motor proteins and nucleators,
30:00.0	and in order to have more complicated things
30:02.1	that make eukaryotes what they are,
30:03.2	we need to have either nucleators or motors.
30:06.0	So, this raises the question,
30:07.3	well, why don't bacteria have them?
30:09.0	Where did they come from in eukaryotes?
30:10.1	And why does it seem like it's so hard for the bacteria
30:12.0	to pick them up?
30:14.1	Well, for the nucleators,
30:15.2	this is a particularly interesting problem
30:17.2	because the way that eukaryotic cells tend to nucleate
30:22.1	their cytoskeletal proteins
30:23.3	is very often by taking the subunit,
30:27.0	copying that gene,
30:28.2	and then having it diverge a little bit
30:30.3	to have it assume a specialized nucleation function.
30:33.1	And we've seen this in two cases,
30:35.1	completely independent cases,
30:36.2	for actin nucleation and for tubulin nucleation.
30:39.1	So, actin nucleation can be performed
30:41.2	by this complex that's called the Arp2/3 complex,
30:44.1	where Arp stands for actin-related protein,
30:46.3	because there are two different proteins
30:48.3	from different genes,
30:50.1	both structural homologs of actin,
30:52.1	that come together to make this regulated nucleation complex.
30:56.1	In the case of microtubules,
30:58.0	the nucleation is often carried out
30:59.3	by what's called the gamma-tubulin ring complex.
31:02.1	Now, the lattice of the microtubule
31:03.3	is made up with alpha-tubulin and beta-tubulin.
31:06.0	Gamma-tubulin is specialized only for nucleation function.
31:10.2	Why can't bacteria do this?
31:12.1	Well, I think they clearly could if they wanted to,
31:14.3	because we do see, as I mentioned,
31:17.2	among the actin homologs in bacteria,
31:19.1	an individual cell may have
31:21.2	several kinds of actin homologs
31:23.1	that are all present within that same genome
31:25.1	that have all been evolved for slightly different functions.
31:28.1	And yet, so far, as far as we know,
31:31.1	none of them have been specialized specifically
31:33.1	for filament nucleation.
31:35.3	I don't know why they don't want to,
31:38.0	but it seems like they could,
31:39.2	and yet they don't.
31:43.1	Turning now to the molecular motor proteins,
31:46.0	this is something where it might be a little more clear
31:48.2	why bacteria don't have them.
31:50.1	If we think about where the really good molecular motor proteins come from in eukaryotes...
31:54.3	there are of course three different classes,
31:56.3	there's the myosins, the kinesin, and the dyneins,
31:58.1	that all hydrolyze nucleotide
32:00.0	in order to undergo a conformational change,
32:01.3	and of these the kinesin and the dyneins move on microtubules.
32:06.2	The myosins move on actin filaments.
32:08.1	Now, these different motor proteins
32:10.0	had all been purified,
32:11.1	had all been studied biochemically,
32:12.2	and sequenced,
32:14.1	and then eventually their structures were determined,
32:16.0	and at the point when the structures were determined,
32:18.0	there was this huge surprise,
32:19.3	and this, again, was back in the 1990s,
32:22.0	where it was found that
32:25.1	the fundamental of the catalytic core of myosin
32:27.1	was almost identical
32:29.1	to the structure of the catalytic core of kinesin.
32:31.3	Even though myosin walks on actin filaments
32:33.2	and kinesin walks on microtubules,
32:35.2	they still seem to be homologs
32:37.2	derived from a common ancestor.
32:39.2	And in both cases, the fundamental way that ATP hydrolysis
32:42.2	at the core of the motor protein
32:44.2	results in the conformational change
32:46.0	that gives you a step
32:49.0	is very similar at the heart of the motor protein,
32:50.3	even though all the details about how it couples to its filaments
32:53.2	are quite different between the two.
32:56.1	So, this gives rise to, you know,
32:58.2	another field ripe for speculation
32:59.3	about where did the motor proteins come from.
33:01.2	Well, if kinesins and myosins are related to each other,
33:04.1	it seems like there must have been some sort of motor precursor,
33:07.2	and we don't know what its substrate was,
33:09.2	if it was actin or microtubules,
33:10.2	or if it might have been something else.
33:12.2	We don't know what the complement of motors was
33:15.2	in the earliest thing that became a eukaryote,
33:18.1	and whether this might have been something
33:20.1	that actually drove it to separation
33:23.1	from the prokaryotic branches of the tree
33:25.1	in terms of morphological diversification.
33:28.1	One thing we do know, though,
33:30.3	is that those motor proteins, kinesin and myosin,
33:33.1	are both derived from a particular branch of
33:37.2	the superfamily of proteins that's called P-loop NTPases,
33:40.1	that includes a lot of ATPases and GTPases.
33:43.1	And this fairly complicated diagram
33:45.2	is a summary of sequence diversification
33:49.1	among many, many different family members
33:51.2	of this large protein superfamily
33:53.2	across the entire tree of life.
33:56.1	And what this group,
33:58.1	this is Eugene Koonin's group that's done this,
34:00.0	has attempted to do
34:02.2	is look at what the distribution of what all these protein types
34:05.1	in currently living cells
34:06.3	and then project backwards in time
34:08.2	to see when they might have evolved.
34:10.3	And so there's some classes of proteins
34:12.3	that you see here with these pink lines
34:14.2	that are present in bacteria, in archaea, and in eukaryotes,
34:17.1	and so they must have been present in LUCA,
34:19.0	in the last universal common ancestor.
34:21.1	But there's one particular protein family, right here,
34:24.1	that's shown in brown,
34:26.0	that has both the myosins and kinesins in it,
34:30.0	and then also has a whole lot of other GTPases
34:32.1	that we associate with specifically eukaryotic functions
34:34.3	-- Ras, the Rab proteins that are involved in membrane trafficking,
34:38.3	the Rho proteins that are involved
34:40.2	in large-scale organization of cell polarity --
34:44.0	all of those things come from
34:46.1	the same relatively narrow branch of the P-loop NTPases.
34:49.2	So, one possible explanation
34:53.2	for why eukaryotes and prokaryotes are different from one another
34:55.3	is because the proteins that are best poised
34:58.1	to become stepper motors
35:00.2	of the kind that are familiar from eukaryotes
35:02.2	happen to be in this branch of proteins
35:05.1	that didn't evolve until the eukaryotic ancestor
35:07.1	had already split off.
35:09.3	It's very tempting to speculate
35:11.1	that this particular class of proteins
35:12.2	that includes not only the myosins and kinesins,
35:15.3	but also all of these regulatory GTPases,
35:19.0	might be part of what makes eukaryotes what we are.
35:22.2	Now, I'd like to point out, of course,
35:24.3	there's plenty of other protein families
35:26.2	that are not shared between eukaryotes and prokaryotes
35:28.1	-- about 50 other classes of proteins
35:30.1	have been identified to date --
35:31.1	however, I think, mechanistically,
35:33.0	this is a particularly intriguing group
35:36.1	for generating specifically morphological diversification
35:37.3	at the level of cells
35:40.1	and at the level of whole organisms.
35:43.1	Now, I don't mean to say that bacteria don't have motors;
35:45.2	bacteria have amazing motors.
35:47.1	Bacteria have the flagellar rotor,
35:49.2	which is a very complicated structure,
35:52.0	it's much more complicated even than dynein,
35:54.1	that's able to spin at 200 Hertz
35:56.0	and is made up of more than 40 different gene products.
36:00.1	Bacteria also have extremely strong motors
36:02.2	like the motor that drives twitching,
36:05.3	by retraction of these type IV pili,
36:08.1	and this is something where a single motor
36:10.1	interacting with a filament
36:12.2	that's extended from the...
36:14.1	that's extended through the bacteria envelope
36:17.1	is able to generate up to 100 picoNewtons of force,
36:20.0	just from movement
36:23.0	of a single extended filament structure.
36:26.0	So, the motors in bacteria can be very powerful,
36:28.1	very efficient,
36:29.3	and very complicated,
36:31.2	and yet all of them seem to do things on the surface of the bacterium.
36:34.2	None of them seem to do anything in the cytoplasm.
36:37.1	So somehow this question of morphological diversification,
36:40.1	about how you get complex intracellular structures,
36:43.1	seems to be at least correlated
36:45.1	with the presence of linear stepper motors for cytoskeletal proteins
36:50.2	in eukaryotes,
36:51.2	and no real equivalent in prokaryotes.
36:54.0	And, you know, again, I can speculate wildly
36:55.3	about why there seems to be this difference,
36:59.0	but nevertheless this is a fairly compelling correlation.
37:02.3	It doesn't come down to complexity or anything like that,
37:05.1	but really only down to the presence or absence
37:07.1	of this one particular protein family.
37:11.1	Okay, so this is, overall, the hypothesis, again,
37:14.0	that we need nucleators and molecular motors
37:15.3	in order to drive things like the mitotic spindle,
37:18.2	or like microtubule asters,
37:19.2	or like parallel bundles of cytoskeletal filaments
37:23.1	that give eukaryotes morphological complexity,
37:25.1	and I've made the claim
37:28.2	that no structures of this kind, again,
37:30.0	with just very minor exceptions,
37:32.0	have yet been found in the cytoplasmic compartment in bacteria.
37:37.3	I'm focused, of course,
37:39.1	on the role of the cytoskeleton in all of this,
37:41.1	but it is worth think about maybe other kinds of explanations
37:43.2	for the difference,
37:45.0	and one thing that might really be connected
37:46.2	is the fact that it's been observed in bacteria,
37:48.2	where of course the chromosome
37:52.0	is just embedded in the cytoplasm
37:53.3	and there's no nuclear membrane separating the chromosome
37:56.2	from the cytoplasm,
37:57.3	in the cases where it's been well documented,
37:59.1	the bacterial chromosome is actually highly organized.
38:02.1	And one of the first hints of that came from these studies in Caulobacter,
38:05.1	showing that a circular bacterial chromosome
38:08.1	that was marked with different fluorophores
38:11.2	at specific locations would actually organize itself
38:14.1	in basically every cell within a bacterial colony
38:17.2	such that those markers on the chromosome
38:19.1	are all found in the same order.
38:20.3	In other words, the bacterial chromosome
38:22.1	is packed into the cell
38:24.1	in a highly structured, highly regular way
38:26.2	that can provide spatial information
38:29.1	for targeting other kinds of subcellular structures,
38:31.2	for example, to the poles
38:33.2	or to the nascent septal zone.
38:36.1	So, you know, one possible way
38:39.1	to bring these two ideas together
38:41.2	is that at some point in the development of the eukaryotic ancestor
38:43.3	something happened
38:46.0	that either enabled the nuclear envelope to come in
38:49.1	or somehow otherwise separated the chromosome
38:51.3	from the cytoplasm,
38:53.1	and if bacterial cells and archaeal cells
38:55.1	primarily rely on their chromosome
38:57.1	as being their major organizing principle
38:59.0	as far as where to put things in the cytoplasm,
39:00.3	then when that separation happened,
39:03.2	the little filaments that were left out in the cytoplasm,
39:06.0	that in bacteria had only been doing
39:08.0	fairly trivial things
39:09.2	of figuring out where to divide
39:11.1	and how to determine the overall shape
39:13.1	of where to lay down the cell wall,
39:14.2	those guys were left by themselves out in the cytoplasm.
39:16.1	They had no landmarks,
39:17.3	they had no chromosome to tell them where anything was,
39:19.1	and so they somehow had to figure out
39:21.2	how to make larger-scale structures
39:23.2	like asters, like bundles, like spindles,
39:26.2	that could help them organize themselves
39:29.1	in the cytoplasm,
39:30.2	in the absence of information from the chromosome.
39:34.0	Okay, like I said, this was wild speculation,
39:36.0	but one of the things that I think is fun about this
39:38.3	is that it's wild speculation that is absolutely disprovable.
39:41.1	So I'd like to give you a personal challenge
39:43.2	to look for and try to find
39:46.2	either a cytoskeletal stepper motor
39:48.1	that walks on a bacterial cytoskeletal filament
39:50.2	in a way analogous to what kinesin or myosin
39:53.1	does in eukaryotic cells,
39:55.0	or else to find a regulated nucleator
39:57.2	of any of these bacterial cytoskeletal filaments.
40:00.0	A lot of very smart people have been looking in various ways
40:02.1	for at least the past 20 years
40:05.0	and, as far as I'm aware,
40:06.3	nobody has yet found any proteins in one of these classes.
40:08.2	Now, that certainly doesn't mean they're not out there,
40:10.2	but it does mean that if either of these things were found,
40:13.0	that would be absolutely definitive proof that my big, crazy hypothesis
40:16.2	is wrong.
40:18.0	So, please try to prove me wrong,
40:20.1	and if you do find one of these classes of proteins,
40:22.1	a bacterial cytoskeletal stepper motor
40:24.1	or a bacterial regulated nucleator,
40:26.0	please send me an email and let me know.
40:28.0	I would love to hear about it.
40:29.2	And if you do, I personally will send you
40:32.0	a large bouquet of flowers
40:33.2	and also hearty congratulations
40:35.1	for taking another step forward
40:37.1	to trying to understand one of these biggest questions
40:39.1	about why cells on Earth look so different
40:41.1	from one another.
40:42.2	Thank you.