Harnessing actin dynamics for endocytic trafficking
This is a modal window.
Das Video konnte nicht geladen werden, da entweder ein Server- oder Netzwerkfehler auftrat oder das Format nicht unterstützt wird.
Formale Metadaten
| Titel |
| |
| Serientitel | ||
| Anzahl der Teile | 34 | |
| Autor | ||
| Lizenz | CC-Namensnennung 3.0 Unported: Sie dürfen das Werk bzw. den Inhalt zu jedem legalen Zweck nutzen, verändern und in unveränderter oder veränderter Form vervielfältigen, verbreiten und öffentlich zugänglich machen, sofern Sie den Namen des Autors/Rechteinhabers in der von ihm festgelegten Weise nennen. | |
| Identifikatoren | 10.5446/50885 (DOI) | |
| Herausgeber | ||
| Erscheinungsjahr | ||
| Sprache |
Inhaltliche Metadaten
| Fachgebiet | ||
| Genre | ||
| Abstract |
| |
| Schlagwörter |
9
22
28
00:00
Computeranimation
07:04
Computeranimation
10:19
Computeranimation
16:48
Computeranimation
21:47
Computeranimation
23:55
Computeranimation
25:04
Computeranimation
26:27
Computeranimation
32:11
Computeranimation
36:11
Computeranimation
43:03
Computeranimation
46:25
Vorlesung/Konferenz
Transkript: Englisch(automatisch erzeugt)
00:15
Thank you very much. I'd like to thank the organizers, Mikhail, Anik,
00:21
Nadia, and Nava for the opportunity to be here at this very, very interesting, enjoyable conference. Today I will tell you about our studies on the role of actin dynamics in endocytic trafficking.
00:41
And we haven't heard enough about the cytoskeleton here yet, so I thought I'd start off with a picture of the actin cytoskeleton, which is one of the main subjects of study in my lab. This is a scheme made by Dyke Mullins and Tom Pollard,
01:00
which shows the force-generating machine that's responsible for the motility of cells and for other types of mechanical force generation in cells that use the actin cytoskeleton. So here the plasma membrane is depicted up here. The outside of the cell is up here, and this is the inside of the cell. And what happens is there's sort of a short signaling
01:23
cascade that activates a protein complex called the Arp2-3 complex, depicted here in green, which then binds to the side of a pre-existing actin filament and then makes a branch on that actin filament. And actin monomers polymerize.
01:41
And through a Brownian ratchet motion, this actin actually polymerizes at this so-called barbed end of the actin filament, or plus end, and can generate a pushing force on the plasma membrane. And this whole process involves a cycle
02:01
of assembly and disassembly. And it's very interesting because actin assembles as an ATP-bound protein, but then a short time after assembly, a couple of things happen. A protein called capping protein caps the filament. So filaments only grow for a couple of seconds before they're capped.
02:20
And that's very important because the filaments must remain short if they're going to be able to resist the tension of the plasma membrane as they push on it. And then also, a short time after the actin assembles in these filaments, the ATP gets hydrolyzed to ADP. And that actually makes the filament
02:43
susceptible to severing by a protein called cofilin. And cofilin severs the filaments. And then other proteins regenerate a pool of monomers, which get recharged with ATP. And then the cycle continues. And so you get this constant assembly and pushing
03:02
against the plasma membrane by this actin network. And this has been worked out in great detail where we know rate constants, concentrations, binding constants for all of the key factors involved in this process.
03:21
This is through many labs, largely Tom Pollard's lab, as well as Mary France Carlier's, just down the road in Gif. And so for our work, this has been a really important framework for our studies because what attracted us to endocytosis,
03:41
because we're really a cytoskeleton lab, and many of the projects in our lab actually have nothing to do with membranes, just have to do with actin assembly, is that it turned out in budding yeast, where we began most of our studies, during endocytosis, the assembly of actin, here shown in red, is absolutely essential
04:00
to invaginate the membrane and to pull off endocytic vesicles and for the vesicles to undergo scission. And so we study this system as a way to study, in a biological context, how the forces generated by assembling actin are harnessed
04:20
to do work for biological processes. And it's proved to be a very nice model. And I should say my lab is sort of split half in yeast and half in mammalian cells. So how did we get involved in all this? Well, we were working in budding yeast some time ago, and we started identifying proteins that regulate actin.
04:45
The first one that I found as a postdoctoral fellow, I called ABP1 for actin binding protein one, because it was the first actin binding protein found in yeast. Through genetics and biochemistry, we and others found many other proteins that interacted with ABP1 and with each other,
05:04
using a synthetic lethal screen of the type Charlie Boone mentioned. We found a gene called SLA1 for synthetic lethal with ABP1. What was curious, and as we started to work out this network, is that a number of proteins
05:21
that sort of were entangled in this interaction network of physical and functional interactions, were proteins that had been implicated in endocytosis. And at that time, in the 1990s, those two processes weren't,
05:40
there really wasn't any good reason to think that they were directly involved with each other. And that was something that we found curious. We kept running into genes, for example, called end genes that someone named Harold Reesman in Geneva was studying, because he was studying endocytosis in yeast cells,
06:01
but we didn't really know exactly what to make of it. So then, this was around the time that after GFP had been found and different spectral variants of GFP had been found, it was possible to start doing live cell two-color imaging. And I had a new postdoc in my lab named Marco Kecsonen,
06:21
who was very interested in this problem of how these proteins were working with each other. And he decided to set to work by tagging pairwise combinations of proteins in this network with green fluorescent protein and red fluorescent protein and looking at them, looking at live cells,
06:41
expressing both proteins at the same time, so doing two-color imaging. And so, one of the, I think the first pair he looked at, he tagged ABP1 and SLA1. This SLA1 protein turns out to be interesting because it's an endocytic adapter. It binds directly to endocytic cargo,
07:01
to the pheromone receptor in budding yeast. And so, this is what Marco saw. Now, first of all, so other people had started to look at these proteins and they did this by looking at static images of cells. And there was a paper published, for example, that looked at actin, here the red is surrogate for actin,
07:21
and an endocytic protein here in green, and concluded that for the most part, they were present in different structures in a yeast cell. You know, occasionally you would see some yellow, which meant that the two proteins were together, but it's hard to know when it's just a low level
07:40
of coincidence what to make of that. But the real significance of this interaction and the explanation for how these proteins are functioning together in a network comes when you do two-color live cell imaging. And Marco did a couple things differently from what other people did.
08:00
One is he looked at two colors in real time. Another is he used a medial focal plane. So yeast cells are spherical. And if you'd use a medial focal plane, you're really focused on the surface of the cell, just around the edges. And you can see that all of these dot structures, we call patches, are present on the surface of the cell.
08:22
Now, if you watch in real time, you see something really interesting. And that is that every single patch has a very similar, undergoes a very similar dynamic process. When it first appears on the surface, it's green. And then invariably, that green patch turns yellow.
08:43
In other words, first an endocytic adapter appears on the surface of the cell, and then actin filaments start to assemble at that site a short time later. It happens in a very predictable order. First the endocytic adapter, and then the actin. And then if you really study,
09:01
it's amazing, a simple movie like this, a question came up the other day in the discussion, what can you learn from live cell imaging? From a simple movie like this, it's amazing what you can learn. And one of the things, if you really study this movie, that you start to notice is that just when these patches start to turn yellow, they move off of the surface into the cytoplasm, okay?
09:22
As though perhaps the forces from actin polymerization are driving some sort of structure from the surface of the cell into the cytoplasm. You can depict that nicely making something called a chymograph, where you draw a line through one of these patches and sample that line in every frame of a movie.
09:42
And then you can see that this endocytic protein here is present over time on the surface of the cell. Then there's this burst of actin polymerization at exactly that moment. The endocytic protein starts to curve off of the surface of the cell into the cell's interior, showing there's a very tight correlation
10:01
between the assembly of actin and the movement of this structure into the interior of the cell. So we've done this experiment over and over again with a lot of pairwise permutations. It turns out there's about 50 or 60 proteins that are in this network. And what we ended up with is this,
10:22
which summarizes quite a few years of work from my lab and other labs in the field. And what it shows is a cartoon of what we think is happening on the surface of the cell. We think that some endocytic proteins start to accumulate, cargo starts to get captured, and then there's a burst of actin polymerization.
10:43
The forces from the actin polymerization are harnessed to invaginate the membrane and pull off a vesicle. And then along that timeline and color-coded with the cartoon above are about 50 proteins that we have ordered in this pathway, all by doing different pairwise permutations
11:01
of labeled proteins. So for example, that SLA1 protein that I told you is an endocytic adapter is here, and ABP1 is here that I showed you in the pair. And so SLA1 arrives, and then predictably, ABP1 also arrives with actually a very predictable
11:20
and minimally variable amount of time between the appearance of these proteins. Okay, so then what we've done is built this temporal map for the recruitment of many different proteins to these sites of endocytosis. And when you do this for 50 proteins, it gives you kind of a holistic view
11:42
of this very complex pathway and series of events. And so one thing that you can do is, it's really hard to think about the functions of 50 or 60 proteins, at least for me, is you start to see that you can cluster groups of proteins together within the pathway.
12:01
For example, early proteins were things like coat proteins, and you can do that by looking at genetic interactions, physical interactions, the dynamics of the proteins, the lifetimes of the proteins, phenotypes when you knock out proteins. And what we realized is that you could cluster these 50 proteins into maybe four or five
12:22
groups of proteins, and then we developed the concept that these were modules of proteins, and that each module was carrying out a function. And so the first module of proteins, shown in green and light blue, are sort of a coat that would create the coat of the vesicle and also capture cargo.
12:41
This blue module are proteins that would link the coat to machinery that nucleates actin assembly, then machinery that nucleates actin assembly, shown in purple here, and that generates forces on the actin, would get recruited, so a WASP protein that activates this R23 complex to nucleate actin, a myosin that generates forces on actin.
13:05
These proteins start to accumulate. Interestingly, there are a lot of multivalent proteins, the SH3 and proline-rich proteins here. We think that there might be a phase transition
13:21
involved in, and we published a paper last year on this, in linking the actin assembly to this endocytic coat, because there seems to be a threshold effect where having these multivalent interactions concentrates the activators of the R23 complex, because then what happens is we reach
13:41
a threshold effect of a couple of key proteins, and then there's this transient burst of actin assembly, and the reason that it was so hard for people to detect an association of actin with the endocytic machinery earlier is because this interaction is so transient. Okay, that transiently there's a burst of actin assembly. Knowing when things get recruited in the pathway
14:02
can generate ideas about what things might be doing, so the actin is recruited very late when the vesicle internalizes, suggesting it's generating a force. These bar proteins come really late in the pathway, suggesting they might be involved in scission, which was, again, verified by genetics. Now, key to a lot of this work
14:22
and our ability to make such a precise pathway was the fact that in yeast we could precisely integrate GFP and RFP into the genome, because homologous recombination is very robust in yeast, and so we could look at the dynamics of each of these proteins expressed at their native levels,
14:40
because we didn't have to do what was commonly done in mammalian cells, which is to make a cDNA of a gene you're interested in, clone GFP or RFP behind it, and then reintroduce it into cells, in other words, and then overexpress that protein on top of the endogenous proteins. What we decided to do,
15:00
so we started, our eyes started looking towards the mammalian cells, because a number of proteins in that network that we found in budding yeast had homologs, almost all of them did, in mammalian cells, and a great number of them were unstudied, and so we decided, why don't we study them? And then when we decided to look at the dynamics in mammalian cells, we decided to use genome editing
15:21
to make precise integrations. So for example, we tagged the clathrin coat with RFP and the dynamin protein that mediates the scission of the vesicle with GFP, but we did this using first zinc fingers, but nowadays CRISPR-Cas9 to make a precise integration of these tags.
15:41
And so as Tommy showed you yesterday, you can now, you can look at these events in real time in mammalian cells, and many other labs have done this, including Tommy's before us, but I think one innovation we made was to do this at endogenous levels by genome editing.
16:02
So here you could see the red clathrin coat appearing in this turf, this is a sort of early turf movie of ours, and then each of these spots, very predictably, would turn sort of yellow and green when the dynamin would come to mediate the fission event. And so it's a fair amount of trouble
16:21
to do the genome editing, and you can ask, is it worth that extra work? And so one way we looked at this question was to compare cells in which we had overexpressed clathrin and dynamin as RFP and GFP and another to cells where they were endogenously tagged. And so to do this, we made 3D chymographs.
16:43
So I showed you a 2D chymograph before, but in this case, we took a four-minute movie like this and showed the entire movie in one picture by putting time in the Z dimension. So these are basically a stack of frames from a movie, and you can see each endocytic site when we overexpressed clathrin and dynamin,
17:01
and you can see that the two colors kind of blur together. And so it's really hard to distinguish when one protein is arriving much before the other, and that sort of thing, when you overexpress the proteins. But when you genome edit the cells, you find that there's a nice period
17:21
when the clathrin is assembling, and you have primarily clathrin, and then the end of the process is punctuated when dynamin is recruited and vesicle scission occurs. So we think that having these genome edited cells, and we now have probably, I don't know, 130, 140 lines in the lab
17:40
with various proteins engineered, allows us more sensitivity to both look at the normal cells but also to look for effects of perturbations in the cells. But we also, coming from a yeast background, we wanted to try to more closely replicate some of the features of yeast cells.
18:03
And, oh, sorry, this just shows you sort of a profile of looking at the average. You can look at the kinetics of clathrin recruitment and dynamin recruitment. Clathrin largely comes first, then there's a spike of dynamin, and the two disappear together when a vesicle forms. Okay, so but what were the other sources of variation,
18:21
maybe from one lab to another, one experiment to another? Different labs were looking at cell lines in mammalian cells from different species, different cell types, fibroblasts versus whatever, liver cells. Almost all the cells that are studied in tissue culture
18:43
have chromosome abnormalities because they're cancer cells, and those cancer cells are in a cancerous state. They don't represent normal physiology. So we wanted to establish a robust system where we could study cell proteins at their endogenous level in as close to the normal physiological state as possible.
19:01
And what we chose to do was to start studying cancer cells, and we used this cancer line from Bruce Conklin's lab at UCSF called WTC. It's an induced pluripotent stem cell. We've also used ES cells. This is a karyotype of a HeLa cell, and you can see the chromosome numbers are quite aberrant.
19:23
There are five copies of some chromosomes and three of others. There are massive translocations. And this I should note is a snapshot of a cell because these cancer cell lines are not stable. When you have this kind of karyotype, it's constantly changing.
19:41
This is a snapshot of a dynamic change. For one thing, chromosome instability is a hallmark of cancer cells, whereas you can get stem cells that actually have a normal karyotype and normal physiology and by many indications are normal. And so we decided a few years ago
20:02
to start doing all of our genome editing in stem cells. There are other advantages, like you can take your stem cells and you can induce them into many different cell types. So now you can compare a cellular process like endocytosis in cells that are genetically identical.
20:23
The only difference is the epigenetics. You're differentiating them to different cell types. And at first, we've concentrated on comparing the stem cells to fibroblasts and neural progenitors. So we've made a bank of genome edited stem cells that we then differentiate into these different cell types.
20:41
And it's really interesting because if you look, my post-doc Daphne Dambrenae, if you look in the stem cells by EM, in a collaboration with Justin Taraska, we see sort of large clathrin-coated vesicles. When we differentiate to fibroblasts, we see these large structures that have been referred to as plaques
21:01
that often seem to have vesicles emerging from their sides. And then neural progenitor cells have super fast endocytosis and extremely regular smaller vesicles. And so the EM level ultra structure recapitulates very well what we see by the dynamics of looking in real time.
21:22
And we've begun to dissect what's happening as cells differentiate to differentiate this pathway and adapt it for these different cell types. And we think because it's an isogenic model, we now have a lot of control to do very well controlled experiments and to really get to the source
21:43
of these different phenotypes. The other thing that you can do with stem cells is to make organoids. So again, if we wanna get closer to physiology and we heard about this yesterday from Tommy
22:04
and it was beautiful zebrafish studies. Another, I think complimentary way is to make organoids. In this case, we're looking at a intestinal organoid. This is Daphne Dambournet, she's French from Paris and a collaborator in Dirk Hockenmeyer's lab
22:21
at Berkeley, Ryan Forster, who's helped us make organoids. And these epithelial cells have an apical surface which happens to be in the lumen, and then this basolateral surface. And so in real cells in tissues,
22:41
there's stuff like cell-cell interactions and polarity where different activities are happening at different surfaces. And so we think ideally what we wanna do is to be able to watch these things in organoids. And Tommy already showed you that we also collaborated with the Betzig lab to start imaging these things. And I just wanted to mention,
23:00
if you look at a volume like this, Tommy alluded to this too, one of the problems that is becoming really acute with these advanced microscopes, like the lattice light sheet with adaptive optics is the amount of data that you can generate. And so in this first frame here, we just did some simple segmentation,
23:21
just looking at the nuclear volumes and starting to look at the membranes. And commercial software, a file like this gets to be about two gigabytes, which is the point at which you start overwhelming commercial software. One of our movies from our study with Eric Betzig, so Daphne spent just eight days at Janelia Farm,
23:40
she generated 30 terabytes of data. And a typical movie like this one is 72 gigabytes. And so it's hard to manipulate these kinds of images and it's hard to do particle tracking. And so we have a team now of people who are trying to catch up with Tommy,
24:00
who can develop software for analyzing these things. So these three folks are all a student postdoc and undergraduate computer science major have all been improving our particle tracking software for 2D. And then Joe Schoenberg, who's a new data science fellow in the lab
24:21
has now got things working well in 3D. So we can start to look at these things. And this is a movie that compares different modes with the lattice light sheet. And you can see with the adaptive optics, you can hopefully start to see, you can see the individual endocytic events
24:40
and start to quantify things in real time. And this is part of a paper that Tommy and I are both authors on that will be out in science sometime soon with Eric Betzig's lab. So this is work all done in Betzig's lab and Tommy actually showed this movie yesterday. So that's now with the adaptive optics and this is with particle tracking from Tommy's lab.
25:03
Okay, so with this project, we're now making, we started with intestinal organoids but there are a lot of people on my campus who are interested in making other things. And so Joe, just this data science fellow is now making brain organoids with another lab.
25:22
And I think the organoids are complimentary to things like zebrafish because you can make many different tissue types, you can create huge banks of stem cells and there are now resources. One of the things that people in say the yeast or Drosophila field enjoy are shared resources of knockout collections,
25:42
tag gene collections and so on and so forth. The Allen Institute of Cell Science which I'm also involved with is making a big library now using the same parent cell line as we are with virtually every cell, organelle, cytoskeletal structure, signaling protein, so on and so forth tagged with GFP or RFP.
26:01
And so all of these things can now be imaged using the lattice light sheet, for example, to look at whatever your process is and then you can engineer in your favorite disease mutation like we heard from Dr. Shen before me, I don't know how you say that disease but the awful skin disease, you could start to take these cells
26:21
and differentiate them into keratocytes and look at these things. So I think this has a lot of promise. Back to actin. Okay, so this is an old experiment that Marco and Evie Sun did in my lab some years ago. Again, this is a budding yeast cell and these are these cortical patches.
26:42
And when you look just in a wild-type untreated budding yeast cell, in a chimerograph, you see these hook-like structures where the endocytic protein's present on the surface and then at the end of its lifetime, it curves into the cell when the membrane invaginates. In yeast, if you add an actin inhibitor,
27:02
latrunculin A to the cells, you completely block this internalization. Actin is absolutely essential for generating that force. And so we wanna use this as a system to study force generation but we're now starting to look at this in mammalian cells and so with our genome-edited cells,
27:22
one question was whether actin is really integral to the endocytic machinery in mammalian cells like it is in yeast cells and so when we genome-edited the cells for, say, clathrin and actin, we found that essentially every endocytic event in a mammalian cell involves a burst of actin assembly.
27:41
Again, it was eluded people for many years because it's transient and it was really Christian Merrifield who found this originally that actin, there's a burst of actin assembly late in the endocytic pathway but I think our work with these genome-edited cells added to that by showing that it's really something
28:00
that happens at essentially every endocytic event. And so, and... David, this is a look. No, I'm sorry. So what this experiment is, it's looking at actin at endocytic sites. I'm sorry. So we've labeled actin with RFP and we've labeled dynamin with GFP
28:22
and showed that essentially every site there's a burst of actin. That's on me, sorry. Yeah, thank you for slowing me down. Okay. So then using our genome-edited cells now where we think the events are more regular and it's easier to detect perturbations. So we've done a lot of sort of drug screens and RNA eyes and here, again,
28:42
whether actin was involved in endocytosis in mammalian cells and how important it was had been a question for many years and a lot of inconsistent data in the field. As we titrated latrunculin, this actin inhibitor, and looked, these are kind of graphs looking at the lifetimes of dynamin,
29:01
we found in wild-type cells, it's very regular. In our cell lines, generally around 18 seconds lifetime but as you titrated more and more actin inhibitor, the lifetime of the dynamin got longer and longer showing that that final step of vesicle formation is getting delayed and impaired in the absence of actin.
29:22
Okay, so now we want to think about how actin might be working to help make endocytic vesicles and so for quite some time I've been collaborating, I'd started a collaboration on mathematical modeling with George Oster in my department. Lately, George has retired
29:41
and my collaboration's been handed off to Padmini Rangamani who's now at the University of California, San Diego and these two folks in my lab, a graduate student, Julian Hassinger and postdoc, Matt Akamatsu, have been doing mathematical modeling. Julian has been doing continuum modeling
30:01
and agent-based modeling and there's been sort of a nice synergy between the two of them. For Julian's work, he views the membrane as an elastic sheet and then starts to vary parameters to see how they affect the vesicle formation.
30:21
So for example, he can vary the spontaneous curvature of these proteins that form the coat or the surface area of the coat and he can show that he can form vesicles but then over sort of a physiological range of membrane tensions, he finds that he can stall
30:41
the endocytic process and that depending on how much tension there is, it will stall either at this sort of U shape before the U to omega transition or it can stall at a very early stage and Julian went on to show in his paper that if you stall these events at higher membrane tension
31:01
but still within a physiological range and we've measured the membrane tensions with our colleague Dan Fletcher that you can add forces that might be provided by actin to push the pathway towards completion. Now it was very satisfying to us that these structures accumulated in the stalled cells
31:23
because it fit well with work from mainly I think from Tommy's lab and from Sandy Schmidt's lab and Tommy had this very nice study where he both varied membrane tension and looked in cells where there were natural differences in polarized epithelial cells. The apical surface has high membrane tension
31:41
compared to the basolateral surface and found that you could with actin inhibitors inhibit endocytosis at the high tension apical surface but that the basolateral surface was much less sensitive to actin inhibition and basically the modeling work that Julian had done
32:01
had fit very well with some of the experimental work from Tommy's lab and Sandy's lab suggesting that actin was really required when you have high membrane tension. Okay, so then in Julian's model he varied where the actin forces might be acting
32:21
and in one scheme he had the actin sort of working like we think it works from our yeast studies by pulling the vesicle in and he also had the actin generate a pinching force and actually both of these could help drive the process towards completion.
32:41
So Matt then started to use his agent-based modeling where he's looking at every single actin filament and individual molecules in this process but in doing so we had a whole wealth of information from studies like those from Tom Pollard's about all the relevant physical properties
33:01
and physical constants and levels for actin cytoskeleton but there were a few things we didn't have at endocytic sites and so one of the things Matt did was built a really robust system for translating a fluorescent signal into a number of proteins and I think this is something that I want to become part of our regular workflow in the lab
33:21
is every time we get a fluorescent signal is to be able to immediately read out a number of proteins and so what Matt did is he adopted a system developed by David Baker where he builds these synthetic nanocages, well he engineers them, proteins that make nanocages of different numbers 12, 24, 60, 120 and then puts GFP on them
33:43
and then expresses these in cells and so these individual puncta each represent a certain number, a known number of GFP molecules and it makes a really nice standard curve over a range of numbers that's relevant for most of the numbers involved in endocytosis
34:02
and so we can use that for example to look at here dynamin and the Arp23 complex which nucleates actin, the dynamin's in purple and the Arp23's in green and we can count the number of Arp23's and so now for Matt's modeling we know that there are about 150 Arp23 complexes and so on and so forth.
34:22
So more numbers to add to this model. So Matt has built this model using Francois Nélac's Cytosim program and he models the vesicle as this object hanging from a spring which is the plasma membrane and then he puts in Arp23 complexes,
34:41
he caps filaments, he does a sweep of parameter space so that we can explore sort of within all biologically reasonable range these various parameters that are all known
35:00
or that we've determined for his mathematical model and the question is can he generate a system that self-organizes itself around an endocytic site and that generates sufficient force to overcome the highest membrane tension that we think occurs in a physiological context where endocytic vesicles are forming and so here he's plugging in numbers
35:22
that he's either determined or that come from the literature and he's done these simulations and now he can generate a network in fact that self-organizes itself around this vesicle and generates sufficient force to pull an endocytic vesicle against this spring.
35:44
Okay, so then from Matt's work we can generate this sort of self-organizing actin network sort of the one we have now is more in this mode and then some of the conclusions I already mentioned he can generate about 15 piconewtons of force
36:05
which we think is sufficient to overcome even very strong membrane tension. Can you ask a question? Yeah. Why is the nucleator in the bulk not at the plasma membrane? Okay, okay, I'm sorry, I'm going fast, I'm not explaining everything.
36:20
There's two classes of proteins. The nucleators are blue and they are generally at the base on the membrane. In the bud, in these purple proteins are coat-associated actin filament binding proteins. Sorry, and that's an important, it's not a nucleator.
36:40
So for example, there's a protein called HIP1R that binds to clathrin, it binds to PIP2 and it binds to actin filaments and it's a part of the coat. My former student Ose Enkbus Goldstein showed that. So the purple protein, thank you for asking that. The purple protein is a filament binding protein that captures filaments nucleated at the base.
37:01
Okay, so, all right, so okay, so in this model then, I said there are two modes in which Julian found actin could help to overcome the high membrane tension.
37:21
And so we wanted to know what actin actually looks like at endocytic sites. And actin is really hard to see in the EM compared to say microtubules. But Tatiana Svitkina's lab has done, I think the nicest study looking at actin around endocytic sites. And what she does is she does platinum replica shadowing
37:40
of unroofed cells. And she sees something that looks much more like that first model where the actin is helping to pinch off the vesicle because the actin is concentrated around the base of the endocytic vesicle. Now this work is beautiful and I love it and I think it's showing you how some actin is organized.
38:00
But there are a couple problems with this kind of analysis. One, it's EM and you can't look at a very large number of events. You can't vary parameters like membrane tension very easily just because there's so much work needed to generate these images. And then also because these cells are unroofed.
38:21
So in order to get these kind of images, they did something very violent to these cells. They ripped the top of the cell off and then they looked at what was left behind. So if part of the machinery that's associated with these endocytic sites is more tightly associated with what's being ripped off than what's left behind, it will be gone.
38:42
So for us, for our modeling and for understanding this process, it's really important to understand how actin is organized. And so we decided to strike up a collaborator with my colleague, Ke Xu in the chemistry department at Berkeley. And my postdoc, Charlotte, collaborated
39:01
with Ke's postdoc, Sam Kenny, and generated by super resolution imaging, storm imaging, thousands and thousands of images of actin around clathrin sites. And Joe has come in and helped us
39:21
to do some quantitative analysis of these imaging. But what's really interesting now is that we find basically two classes of structure. So clathrin is shown in red and actin is shown in teal or whatever that color is. And in many sites and in the majority
39:40
of just normal growing cells, what we see is that the clathrin is higher than the actin. So the actin is around the base of the clathrin coated vesicle. This is exactly like what was seen in that EM study from Tatiana Svitkina's lab. However, we also see this other kind of figure where the clathrin is completely engulfed in actin.
40:04
And I should tell you, for their data set, they use a surrogate timer to figure out where they are in the endocytic pathway, which is they used labeled dynamin. So dynamin appears very late in the pathway. So they only analyzed endocytic sites
40:21
that had dynamin associated with them. And what they found is that normally the endocytic vesicle, the clathrin, is higher than the actin. And that's shown here because the blue is the height of the actin and the red. And I can tell you, one of the things that's really cool about this super resolution imaging
40:41
is it's done in XY, but if you put this spherical lens in the path, you can actually generate Z data. So this is actually a Z projection from data that was collected in XY, which I think is really cool. Anyway, that's an aside.
41:01
So anyway, this is the height of the actin and the red is the height of the clathrin. Now, if you do a transient osmotic treatment to raise the membrane tension, what you see is that you shift this distribution. So now the majority of the endocytic sites, the clathrin is engulfed in actin.
41:20
And so we think that the cell responds to high membrane tension by assembling more actin and actin with a different geometry. And it's interesting to think about what might be the tension sensor that's sensing the higher tension and whether it's a different nucleator that is nucleating the actin around the top of the clathrin pit
41:41
as opposed to the bottom of the pit. And so this data is really, really rich and we think we can do things now like look at class averages and actually get some more structural information from this and we can look at other perturbations to the system. So in the last couple of minutes,
42:02
I just want to tell you about one short little vignette about a project we've had. So a long time ago with George Oster, we shared a postdoc named Jin Liu who collaborated with my longtime postdoc specialist, Edi Sun, and did a theory paper
42:24
which one of the notions in that paper which is I think very commonly discussed now but I think was a little less common then was that there's a crosstalk between the geometry of the membrane and the biochemical reactions that are occurring throughout this endocytic process.
42:42
And so you can think of endocytosis as a sort of cascade of events where the curvature, the geometry is constantly changing for the membrane and that these geometries are being read back by the biochemical reactions. So for example, there are proteins that bind specifically to curved membranes,
43:01
bar proteins for example. And so if one of those proteins binds, that will make the adjacent membrane have the ideal curvature so more proteins can bind and then also enzymes that act on the bilayer.
43:21
If the bilayer is flat, they may have a hard time accessing bonds but as it becomes curved, they could act more quickly and so on and so forth. So with this notion that there's crosstalk between the curvature and the biochemical reaction rates. So we liked that notion but it just kind of sat there for a while and thinking here's some words from George.
43:42
George says that modeling can tell you how things might work. Modeling can also tell you how things cannot work but modeling cannot tell you how things do work. For that, you need experiments. And so what we wanted was a way to experimentally test whether curvature
44:01
was affecting this process and we've seen some elegant studies here from pulling out tongues from GUVs. But we came up with another way to look at curvature in live cells which was I met this woman named Bien-Xiao Cui who's a material scientist at Stanford University
44:22
and they make nano arrays and we saw micro arrays in the previous talk. These are arrays that she makes by etching on a nano scale on quartz glass. And what happens is you can sit cells down on these nano arrays, you can also put supported bilayers on them and the bottom of the cell actually tightly conforms
44:42
to the curvature of the pillars and you can dial in all sorts of curvatures and then we have all these genome edited cells so we could put the cells on these substrates. This one happens to have bars, sometimes they're pillars. The bars are flat in the middle and they have very high curvature at the ends.
45:03
If you put our genome edited cells on them, here's two of the bars. It turns out that the highly curved ends of the bars become hotspots and this is now a chymograph and you see clathrin, dynamin, clathrin, dynamin, clathrin, dynamin. So they're just streaming off vesicles on these highly curved sites.
45:22
And so this is very exciting to us. You can see it actually happening in EM. Here's a cross section of a pillar. There's an endocytic vesicle budding from the edge of the pillar. And the only problem with this system is that it was very hard to make these nano arrays. Fortunately at Berkeley,
45:40
we found that we have another way of making these. We now etch molds and then we use a polymer and we can stamp out these nano arrays where now this one happens to have ridges. And you can see looking at the dynamin for example, that endocytic vesicles are coming from the crowns of these ridges. And so it's nice now because we can make lots of these
46:02
and so now we can do things like RNAi screens and see if we've bypassed certain steps in the process. Our hypothesis is that there's some rate limiting step where we have to generate initial curvature and then you attract more curvature sensing proteins and so on and so forth.
46:21
If you're interested in this, I just have some iBiology talks that were just posted online. So these are the people who did the work. I tried to mention everyone as we were going along and I don't see anyone that I didn't mention. This is the lab.
46:42
There's me and I'm here. It's a joint lab with my wife, Giorgia Barnes, who's here in the audience at the meeting and I thank you very much for your attention. Okay, thank you very much for the beautiful talk.
47:01
We have time for questions. What do you think determines the size of the vesicles? Is there much variability in the size of endocytic vesicles you can make and what determines? So naturally what you mean? Naturally. Yeah, well it's interesting.
47:22
You know, when we differentiate these, so the stem cells actually have sort of a variable and there's kind of a range, I'm gonna say 80 to maybe up to 200 nanometers. Is that too big? Yeah, too big? A little small. Okay, but you know, in these neuro progenitor cells, they're very, very regular and very small
47:43
on the small end of that range. So you can vary them and in fact, Tommy's done these beautiful studies with viral infections showing how adaptable clathrin is. The coat can actually encase something quite large. But that's given by what is being taken.
48:00
So naturally you mean? Naturally. And if you play with your membrane tension, do you change the size? Yeah, that's a good question. Tommy says no. So I don't think, I haven't, yeah, I looked at the absolute numbers. So what, make sure that they don't collapse.
48:23
Clathrin self assembles, it's a rigid molecule. It's rigid enough. As it builds, it puts hexagons and pentagons. So the overall curvature is defined by the ratio of hexagons against the pentagon. What exactly does that sound clear? As David says, the neurons inside the similar guys,
48:43
the cooler vesicles that you have in the circuitry pathway, in all cells, which are small, but you have a huge pressure inside the cell. No, no, no, the vesicles that are forming from internal vessels inside the cell, the angiosomes,
49:02
they're all small guys. No, but at the plasma membrane. At the plasma membrane, it doesn't, we have never seen changes on the overall shape just by changing tension, right? What does it, it's a little bit of a cargo, so they adapt a little to the size of the cargo with an upper limb.
49:21
We go beyond 100 nanometers, they stall. We've also done mass spec to look at the differences in the clathrin-associated proteins in the three cell types that we've compared, the stem cells and neural progenitors and pyrolysis. We've found some interesting differences, and at least one of them, if we modulate it,
49:42
we can change the geometry of the vesicles. So, AP2, actually. Have you looked at how lipid composition changes in your different differential cells? No, it's very interesting. No, we have not done that.
50:01
And again, there's some nice probes available now, so we haven't done that. The young guy. The young guy. This might be a basic question, but do you know why are those rigid hot spots for endocytosis? Yeah, that's what we're trying to figure out now.
50:22
Our idea is that the curvature is a signal that's attracting, there's some limiting step where some initial curvature may help to recruit proteins, and so there are proteins
50:41
that bind specifically to curve memories. Like clathrin, for example, likes to make a cage. And so that by giving the cell curvature, you're attracting those proteins. And so we have a strategy. We haven't been able to do many experiments because it takes, the quartz substrates we have
51:02
take a day, they're made at Stanford. It takes a day to make each one at a cost of about $1,000. We can make about 15 of these in two hours now, so we're just in a place, it's taken us two years, but we're just in a place where we can stamp these things out and ask that kind of question. Maybe I can continue on this question.
51:21
So now, normally what you're expecting, the first protein which would be sensitive to that kind of curvature, which is at that high, would be the F bar, right? So if you deplete F bar, you change the localization. We're setting up to do those experiments. Literally, we've got these substrates. You know, we had to go through all these polymers
51:40
to find things that weren't autofluorescent. Work, we had to, my student learned the CAD programs and how to go, at Lawrence Berkeley National Lab had to learn how to etch everything. So now, he's, Bob, he's ready to go. So those are good questions, yeah. I'm gonna continue on this.
52:00
There's, in epithelial cells, you have all kinds of specializations in foldings and brush borders and all those. And endocytosis is actually preferentially coming from the pit, from the curvature or not? No, I don't think so.
52:20
I mean, no, not in the, I think that it's possible the cell might exploit some natural fluctuations and trap a state, but one of the things we're going to explore now is we're trying, we wanna set up some kind of systematic study. I think you heard today on the micron scale about septons recognizing curvature
52:42
and the clathrin pathway being influenced by curvature. We think there's probably a lot of other things that are happening in this other, affected by curvature, so we wanna use this system to look systematically at a lot of processes and see what else is responding. But I don't know, so in a natural setting,
53:01
yeah, I don't have a comment on that. Otherwise, there's something funny about this. At least in the fish, we haven't seen that. Oh, that there's a preference, or where they come from the, yeah. I think she can say that. Are they at all? Yeah, well, this must be an old question for you.
53:23
I was wondering whether you think, try and thinking about whether force is the only way acting with endocytosis, or there might be other mechanisms, for instance, I'm thinking of a paper by G2 mayor, and look at Johannes, and you still propose that actin plays a role in phase separation.
53:43
Oh, yeah. And that will be part of the procedure, or maybe other ways. We had a, so we had a paper in eLife this year where we actually proposed something like that, that this multivalent interactions make a phase separation to nucleate the actin.
54:00
But there is, your question also about whether actin's doing something else, in budding yeast, it's really clear that actin does something else, which is it sends back a negative signal to take everything apart, to turn off actin nucleation and to uncoat the vesicle. If you remember, I showed a image from yeast where we've used this latrunculin A,
54:21
and when you look at a coat protein like clathrin, when you block actin, it assembles on a surface and it just stays there, which normally it forms and turns over. And we've actually found a couple of the proteins, which may be yeast specific, I don't know. One is a protein kinase that phosphorylates an actor with ARP2-3 complex and turns it off.
54:42
Another is a synaptojannin, which binds to, it binds indirectly to actin to recruit it to facilitate the uncoating step. So there the actin not only generates a force, but it also negatively feeds back on the processes that set up the site.
55:01
So I was wondering about the type one myosins that are involved in yeast endocytosis. Do you know what they do there, and are they also involved in mammalian cell endocytosis? Yeah, so there are type one myosins in mammalian endocytosis. In yeast, they are almost essential
55:22
for the invagination step. And so I have a very good student, Ross Peterson, working on that process. And they're interesting proteins because they both nucleate, the one in yeast nucleates actin assembly and it has a motor domain. And we have a collaboration with Michael Ostap at Penn,
55:42
who's a single molecule biophysicist. And we're characterizing the type one myosin in budding yeast. They're two kinds of myosin motor. You can roughly classify them. Some are tension sensitive clasps that when you pull on them, they just bind very tightly to actin.
56:02
And the others are force generating motors. And so with the single molecule experiments, you can differentiate which type it is. We were, or at least I was really betting that we had a tension sensitive clasp that the myosin was actually holding everything together. But in fact, from the kinetic profiles, it looks much more like a force generator.
56:23
And so, but budding yeast, according to the theoretician that Fred Chang collaborates with in fission yeast, says that the amount of pressure you need to make an endocytic vesicle in yeast which grow under tremendous turgor pressure
56:41
would be equivalent to pushing your finger into the tire of your car. So to put it in sort of real world sense. So I think it may be more acute in mammalian cells, but we've cloned it and tagged it. And it's the same person doing the super resolution work is studying the myosin in my lab.
57:01
We just got an inhibitor through the mail from someone in Germany of the type one myosin. And we're trying to figure out what it does in mammalian cells. I don't know if there's time for one more. We want to take coffee. There's this guy in the back who's been patient, but. So, David, this is probably philosophical.
57:23
I mean, COP1 and COP2 vesicles are roughly the same size. They don't use actin. They don't use bar domain proteins. I mean, is this actin driven process entirely due to perhaps the surface properties because there is so much of the cytoskeleton
57:43
to begin with? Is it to do with the tension? Although Tommy says tension has no role. What? No, no, no. He said the opposite. He said the opposite. Tommy actually did a nice study, I think, to clarify the role. Is that it? Because the membranes are so tense
58:00
that you need extra force to imaginate. I think that might be true, but you know, actin is also used for some intracellular trafficking events. It's an AP1 with the... The AP1 without actin doesn't work. And I mean, basically, the lipid barrier is not tense. I'm talking about COP2 and COP1.
58:20
But the lipid barrier is not tense and the plasma membrane is anyway. Is that it? Is that it? It is not. It is not. You have a lot of excess area, so the plasma membrane is not tense, they say. Well, this guy says it is. But sometimes, there is some events that probably, you agree that the plasma membrane normal condition, the lipid barrier is not tense, right? So, in fact, if you follow up,
58:41
I mean, one point I wanted, right, you said that all the inclusivity clathrin codes, in your opinion, are recruiting actin, right? The events, well, the 90% according to my... So, what we did there, we did the structure illumination, in turn.
59:03
So, there, we were looking a bit, right? The actin only appears in about 60% of the endosyclin. The pits, the other ones don't have it. They are perfectly functional, they're coming in. How did you tag actin? Excuse me? I'm sorry, how were you looking at actin? What would you use for actin?
59:22
I don't remember, but it was actin. Okay. I can't remember, it was... Go get it. I should Google, I should look at my own paper, right? Yes. Why is that important? I'm just kidding. I'm just kidding, I'm just kidding. We see 90, you saw 60, I don't know, but, you know, maybe... No, but there was a difference. The guys that had the actin were shorter-lived,
59:43
and were, yeah, they were faster. Okay. There was no difference in the size of the pit. I see. But those were faster, the kinetics was faster. And the other ones were a bit slower. That's interesting, yeah, yeah, yeah, okay. Well, you know, in our latrunculin titration, we saw it slowed.
01:00:00
Yeah, but that's the problem that when you do the trunking, you're disturbing the whole mastasis. We tried to look acutely, but now we've also, we actually get exactly the same thing with ARP2-3 inhibitors. Can we continue for the discussion first? Yeah, yeah. We should have a coffee and then discuss. Thank you, Dave, before the coffee. Yeah. Yeah.
01:00:20
Thank you, Dave.