Jeanne Stachowiak on Disordered Proteins and Membrane Traffic

Jeanne Stachowiak (00:02:26):
Okay. So I’m going to start. Let’s see. Oh, no. I have to figure out how to get it… How do I get it to advance slides when I’m sharing screen? Oh. Aha. Okay. Figured it out. So I’ll start with a little bit of background on this field and then I’ll tell you about some of our earlier work, thinking about how disordered proteins as individuals function in this field or function in this pathway, and then I’ll move into how do disordered proteins when they interact with other, when they form networks, function, and that’s where we’ll come to protein droplets, which that’ll be great to hear what you guys think about that.

Jeanne Stachowiak (00:03:07):
So first membrane traffic is one of, and how membranes become curved by protein assembly, is a very fundamental question in cell biology and biophysics and many labs have made contributions in this field for several decades now. And you can see here an example of one of the most studied structures, which is a clathrin-coated vesicle. Clathrin-coated vesicles are studied so broadly for several reasons. They form at the plasma membrane, so they’re really accessible to a lot of experiments. Also, the clathrin-coat, this density that you see here around these vesicles, is very accessible and obvious in many different types of imaging from EM to fluorescence microscopy, and then many receptors are specifically internalized by the clathrin pathway. So that sort of makes the clathrin pathway relevant to many different signaling processes and disease processes.

Jeanne Stachowiak (00:04:04):
But this question of how do membranes take on these highly curved shapes, maybe 50 to a 100 nanometers in diameter, has really been the source of a lot of debate and work for a long time. And traditionally there are two main classes of answers to that question. The first is there’s a broad set of proteins that are known to have specific structures that can scaffold the membrane and really impose their curvature on the membrane through a curved binding interface. And here you see this structure, which is the structure of the BAR domain. It’s a dimer. But then we can also think more broadly about structures like the clathrin coat itself or the cylindrical coat site, dynamin, that cut off the vesicular necks, so many different scaffolds.

Jeanne Stachowiak (00:04:51):
And then the other major way of answering this question has been to think about things that increase the spontaneous curvature of the membrane. So one of the primary ways of doing that is to insert something into one leaflet of the membrane, that’s not being inserted into the other leaflet. And here are the example is an amphipathic helix is being inserted into the membrane and quite a few endocytic proteins attach to the membrane with these amphipathic helices.

Jeanne Stachowiak (00:05:16):
So clearly both of these mechanisms rely on specific structural motifs, whether it’s a scaffold with a certain shape or this inserting helix. But a really important thing to understand about many of these endocytic proteins is that they also have substantial regions of disorder. So here we’re looking at the same little helix of this protein called Epsin inserting into the membrane. And on the previous slide, I was just showing you the structured portion of Epsin called the ENTH domain. But you can see there’s also this large intrinsically disordered domain and biochemically, we know a lot about this domain. You can see some of the binding interactions mapped along it, but physically relatively little is known about it.

Jeanne Stachowiak (00:05:59):
And then so in the first bit of the talk, I’m going to talk about these disordered regions and what they can accomplish sort of individually in terms of driving the membrane to curve. But then the other thing that these disorders regions are responsible for is forming these multivalent networks among proteins. And this is an area that’s just beginning to be explored. Many of the experiments in the field have really just looked at one protein at a time, but relatively little is known about these networks. So in the second bit, we’ll talk about what we’re learning about the networks.

Jeanne Stachowiak (00:06:33):
Okay. So first let’s talk about these disordered domains. So I’ve already pointed out that proteins like Epsin contain these disordered domains, but then many of the other endocytic proteins, for example, here’s one of those that contains a BAR domain scaffold, but you can see that each monomer of that dimeric BAR domain also has a substantial disordered domain. And this is a protein called Amphiphysin. Both of these proteins, Epsin and Amphiphysin are what we refer to as adaptors. So what adaptors do is they bind the membrane. Some of them interact with transmembrane cargo, but they also interact with the clathrin coat. So the clathrin coat doesn’t bind the membrane directly. It’s recruited by these adaptors. And many of these adaptors have large, intrinsically disordered regions.

Jeanne Stachowiak (00:07:23):
Okay. So this first little story is about how those disordered regions modify the functions of adaptors. And we’ll start by talking about this BAR domain-containing protein Amphiphysin. So first I’ll point out if you’re not familiar with BAR domains that they’re a large super family. They are these sort of banana-shaped dimers and they have a range of different curvatures. So the most highly convex BAR domain are these what are called in N-BARs, and they’re involved in shaping endocytic vesicles and shaping the neck of the vesicle in particular. And then there’s the slightly less curved F-BAR domains. And F-BARs are involved in the early initiation of coded pits, and they’re also involved in other pathways throughout the cell, like driving cytokinesis. And then there’s actually these I-BARs, which I stands for inverse. So they have the opposite curvature. And as you might expect, they’re involved in forming structures of opposites topology, like filopodia or dendritic spines, so things that poke out of the cell rather than things that poke into the cell.

Jeanne Stachowiak (00:08:26):
So the BAR domains are really a structural biologist’s dream. Not only do they have a nice looking structure as a dimer, but if you let them assemble on a surface, so here you’re looking at BAR domains assembled on lipid tubules, they make this beautiful helical oligomer. And so cryoTEM can really highlight that oligomer and a lot of these structures have been mapped. So from these cylindrical structures, the main interpretation of the function of BAR domains is that they stabilize cylindrical architectures. So they’ve been thought to stabilize these necks. So imagine this is the endocytic vesicle down here and here’s the plasma membrane, so they’re sort of responsible, it was thought, to stabilize the neck of the endocytic bud and then recruit the proteins like dynamin that are going to cut the neck off and free the vesicle.

Jeanne Stachowiak (00:09:17):
Okay. So that had been the paradigm for understanding these proteins. But almost all the work leading up to that idea and all those structures that I was showing you were all solved without the disordered portions. And that’s for a variety of reasons. It’s tough to purify the full length proteins. They tend to inhibit crystallization, so they really get in the way of structural biology. But we set out to ask, “What are these disorder domains doing? Do they, in some way, modify the function of this Amphiphysin protein?” So this was the work of a graduate student, Wilton Snead, who actually recently moved on to Amy Gladfelter’s group, if you know her.

Jeanne Stachowiak (00:09:56):
So Wilton, he started looking at this. And at first, he just did some very basic stained EM. So this is what vesicles look like in a negative stained EM. They look like these crushed cans on a grid. And then if you add the BAR domain only of Amphiphysin before you put them on the grid, you’ll see these long straight-sided tubules. You can almost make out a little bit of the helical oligomer on them and lots of people have seen that in EM. But then if you use the full length protein, instead of seeing these tubules, you almost never see a tubule, instead you see all these little vesicles that have a diameter sort of similar to the tubules, but no tubules. So that was really surprising to us. Somehow this protein is breaking up the membrane.

Jeanne Stachowiak (00:10:40):
So we’re trying to get more insight into that. And one of the first things we did was just to run the experiment in another way, in a higher throughput format. So rather than using EM where you can count relatively few objects and the interpretation is a little subjective, we tethered the same vesicles on a cover slip and used some calibrated brightness techniques to estimate their diameter distribution. So here you can see the vesicles by themselves, but then if you add the BAR domain, you don’t get much of a shift in diameter. They still look like dots because the resolution is low. We have the optical diffraction limit. So even some of those short tubes are still diffraction-limited in length, but we don’t get this shift in the overall brightness of the objects. But then when we add the full length protein, we see this dramatic shift and these small objects.

Jeanne Stachowiak (00:11:29):
So somehow the full length protein which contains the disordered domains is driving this transition from tubules to vesicles. And this is really perplexing, but it makes a little bit more sense if you think about it in the context of this oligomer. So here’s the oligomer that was solved in a recent paper, beautiful structure, and think about where are the disordered domains. So there’s two of them on each dimer. So I’ve kind of put them into this cartoon down sort of one stripe along the helical oligomer. And then we measure their hydrodynamic radii using florescence correlation spectroscopy. So you can see about how much space one of them would take up. And then if you draw them all down this stripe, you see the problem. They really don’t fit. There’s this substantial overlap. In fact, you can calculate that at the surface of the oligomer, you need two to three times more volume than you have to fit them all in. So with that would mean is that if they’re all going to fit, they’ll have to get straightened out.

Jeanne Stachowiak (00:12:28):
Sorry. Did someone say something? No? Okay. Feel free to interrupt if you have a question.

Jeanne Stachowiak (00:12:35):
And straightening them out, as you all are probably aware, if you think about polymers, encourage a pretty substantial entropic cost. In fact, by our estimates, that energetic cost will be greater than the cost of breaking the membrane into pieces, which is actually relatively inexpensive energetically. So our hypothesis became that this steric pressure among these tightly concentrated disordered domains is what was driving the vesiculation. But that idea has several obvious implications. Basically by hypothesizing that we’re saying that the…

Jeanne Stachowiak (00:13:09):
Sorry. Give me one second. I’m going to turn off my email, which is giving us those annoying dinging sounds. Okay. Okay. Good. Sorry about that.

Jeanne Stachowiak (00:13:24):
So that hypothesis has a couple of obvious implications. One of them is that we’re just acting there as if the disordered domain is just a certain size, just a steric bulk, with no particular specificity. And if that’s true, we ought to be able to chop off the existing disordered domain and replace it with another one of similar hydrodynamic radius and see the same thing. So we did that. We made this chimera where we replaced the disordered domain in Amphiphysin with another disordered domain from another endocytic protein, Epsin. And we compared the wild type protein to the chimera. And here you see our shift in vesicle size distribution as a function of concentration. And if you plot these data together, you’ll see they match very well.

Jeanne Stachowiak (00:14:07):
But then even a further implication is that it shouldn’t matter how we attach the protein to the membrane. We should be able to just take the disordered domain by itself and if it reaches equivalent density on the membrane surface, it should drive the membrane to breakup as well. So we ran that experiment and that turns out to be true. But to get up to the same coverage of the membrane surface, we had to go to a dramatically higher protein concentrations. Before we were talking about maybe 250 nanomolar, and to drive…

Jeanne Stachowiak (00:14:40):
Sorry. Got to go back one.

Jeanne Stachowiak (00:14:42):
To drive fission of the membrane with this monomeric version that’s attaching to the membrane with the histidine tag, we needed more than 10 times that much protein. So that really illustrates that the BAR domain, which can oligomerize on the surface and is a dimer, is really potent at concentrating these disordered domains and that more than directly scaffolding the membrane. May be one of the key functions of the BAR domain is to concentrate proteins locally.

Jeanne Stachowiak (00:15:13):
Okay. So we wanted to look at some of these same ideas in cells. So here you’re seeing retinal pigmented epithelial cells, which are popular in this field because they have really big lamellipodia, so they’re really spread out and thin and you can do great TIRF imaging of the plasma membrane. So here we overexpressed the N-BAR domain, which is an old experiment that a lot of people had done in the early days of looking at BAR domains. And so we saw the same thing that everybody else saw, which is if you overexpress this protein sufficiently, you see all these little tubules, this big kind of meshwork of tubules protruding into the cytosol. And it’s thought that the reason for that is that you’ve so overcome the endocytic machinery, you have so much more of the N-BAR domain than of any other endocytic protein, that you’ve run out of all the other adaptors, and instead of making a bunch of endocytic vesicles, you just see these tubules.

Jeanne Stachowiak (00:16:05):
But then if you express the full length protein, you don’t see nearly as many tubules or persistent tubules. Instead, you see all these little fragments or maybe short tubules that extend and then get cut off or retract. So that’s sort of in line with our thinking, but a key caveat here is that the disordered domain here of Amphiphysin can recruit the whole protein scission machinery. So it can recruit dynamin, which is still in these cells. So it’s possible that we’re just recruiting the proteins that are going to cut the tubules off. So to check that out, we made another chimera where we replaced the disordered domain with the disordered domain that isn’t involved in endocytosis.

Jeanne Stachowiak (00:16:49):
Some of you all might be familiar with the neurofilament domains, which are disordered domains that have been studied for a couple of decades now. And interestingly, these domains play a steric role in cells. So they’re sort of these brushes of disordered proteins that extend off of these neurofilaments. You can kind of think of them like a pipe cleaner or like one of those bottle brush flowers. And their job is to space out the neurofilaments and therefore control the caliber or the sort of the waistline of an axon.

Jeanne Stachowiak (00:17:22):
So we took these domains, which are not in this type of cell at all and fused them to the I-BAR domain… Or sorry. To the N-BAR domain of Amphiphysin. And then so when we express this chimera here, we saw a similar phenotype to what we get with the full length protein, which is these short tubules or very few tubules, mostly just fragments. Okay. So from these data, we can observe that the presence of the disordered domains bound to Amphiphysin’s N-BAR is inconsistent with the assembly of these long range scaffolds. So in this way we… Sorry. We kind of changed our thinking about amphiphysin. Rather than-

Jeanne Stachowiak (00:18:02):
Sorry. We changed our thinking about amphiphysin. Rather than thinking of amphiphysin as a protein that stabilizes cylindrical geometries, it now seems more likely that it’s part of the fission machinery. It’s helping dynamin to cut off these endocytic vesicles and that that function relies on steric pressure among the BAR domains, or sorry, among the disordered domains.

Jeanne Stachowiak (00:18:24):
So in doing this, we were curious, which mechanism is more potent at bending the membrane? Is it the structure-based scaffolding of the membrane or this entropic crowding among the disordered domains? And to address that, Wilton had a clever idea. He said, “Well, remember the BAR domain family includes these I-BAR domains that have the opposite curvature. So we could essentially fuse the I-BAR domain to the disordered domains, and we can make a molecule that’s really at war with itself.”

Jeanne Stachowiak (00:18:53):
So here’s this crazy chimera here, and you can see that part of it would like to bend the membrane downward, like this. And then part of it, the crowding among these two disordered domains, would like to bend the membrane outward, like this. So we decided to put this thing on the membrane and see which mechanism wins.

Jeanne Stachowiak (00:19:11):
So first the control experiments here. So these are giant unilamellar vesicles, they’re big spheres of lipid that are just one layer thick. And this is what they look like before you add any protein to them. And if you expose them to the I-BAR domain, it will bind the surface and generate all these inward tubules. And then, in contrast, if you expose them to the N-BAR domain, the one we were just talking about, it will bind to the surface and give you all these outward tubules.

Jeanne Stachowiak (00:19:36):
So then, what’s going to happen with our chimera? It turns out to be something of a tie. We see what we call these frustrated fluctuations where you’ve got this membrane that’s bending outward and as soon as that bends outward, it’ll snap back inward. And that’s telling us that these two mechanisms, one structure-based, one coming from disordered proteins and more entropic in character, are on equal footing and can oppose each other.

Jeanne Stachowiak (00:20:04):
But then, when you think about a molecule like Amphiphysin which couples them together toward the same goal, you can understand why that protein is so potent at remodeling the membrane.

Jeanne Stachowiak (00:20:15):
Okay, so that’s what I wanted to tell you about BAR domains, but in this next little bit, we’ll move on to talking about, if disordered proteins are able to bend membranes, then maybe they could also sense membrane curvature.

Jeanne Stachowiak (00:20:28):
And sensing membrane curvature is really important because it’s the mechanism that’s responsible for bringing together a critical number of remodeling proteins at the same place, is that they can sense the curvature and selectively assemble together at curved sites.

Jeanne Stachowiak (00:20:43):
So this is the work of Wade Zeno, a postdoc in the lab who’s about to leave us. And Wade did some clever experiments where basically he ran experiments where he tethers vesicles of many different diameters to a cover slip using this chemistry that you can see here. And this technique has been around for a while and has been used to measure curvature sensitivity by structured proteins.

Jeanne Stachowiak (00:21:06):
And then, he just looks at the partitioning of proteins among the small vesicles, medium vesicles, large vesicles, and that turns out to be a really effective means of measuring curvature sensitivity. So when Wade first started working on this, he started by using some structured proteins with well-known curvature sensing abilities.

Jeanne Stachowiak (00:21:27):
So this, again, is that structured domain of Epsin and it’s been known for at least a decade now that this amphipathic helix is capable of sensing curvature. And if you take this helix and replace it with some tag that doesn’t insert into the membrane, so here we’re just using histidine tag, you’ll lose that curvature sensitivity, and you can see that in these pictures.

Jeanne Stachowiak (00:21:50):
So if you look closely this is the lipid channel, and then here’s the protein channel, and then let’s take this protein where we don’t have curvature sensitivity. When you merge these two together, all the vesicles are a similar shade of yellow.

Jeanne Stachowiak (00:22:05):
In contrast for this one that has the helix and senses curvature you can see that some of the smaller vesicles are a bit more red and the larger vesicles are a bit more green, meaning that protein to lipid ratio is higher on the smaller vesicles, which is a hint that curvature sensing is going on.

Jeanne Stachowiak (00:22:25):
So then, if we apply one of our disordered domains, so this is the disordered portion of another adapter protein in the clathrin pathway called AP180 and you can see again here when you merge the two images, you get this heterogeneity in the coloration that tells you that the protein to lipid ratio is not the same on all the vesicles.

Jeanne Stachowiak (00:22:46):
So all this can be quantified by analyzing the images and applying some statistics. And so, here you can see that the control protein here that doesn’t have the inserting helix isn’t very sensitive to curvature, pretty much the same partitioning across vesicles of all these different diameters. Whereas in contrast, this structured protein, as well as the disordered protein, are both roughly eight or 10 times more likely to bind to a very small vesicle with, say, a 20 nanometer diameter than they are to bind to a flatter, bigger vesicle with a 200 nanometer diameter.

Jeanne Stachowiak (00:23:23):
Okay, but what then is the mechanism of curvature sensing? So to think about that, let’s look at minute at this AP180 domain. There’s two interesting characteristics physically of this domain. One of them is that it’s pretty long, 571 residues, makes it one of the longer disordered proteins in the clathrin pathway, and then it has a mild net negative charge, 0.06 negative charges per residue on average. So not as strong charge but some charge.

Jeanne Stachowiak (00:23:53):
So we can think of at least two possible mechanisms. So, one mechanism might just come from the long length of this protein. You can imagine that a polymer with that many segments is quite flexible, has a lot of conformational entropy, and when you take it out of solution and ask it to bind to the membrane surface, you’re depriving it of many of the orientations that it could have. So there’s a substantial entropic cost to this event.

Jeanne Stachowiak (00:24:18):
But then, the more you’ve been the membrane away from the protein, the more conformations the protein recovers, and so that entropic cost should be reduced. So it’s less costly to bind this highly curved small diameter vesicle than to bind this flatter larger vessel.

Jeanne Stachowiak (00:24:36):
Then a second possible mechanism could potentially arise from electrostatic repulsion. Many membranes, including the plasma membrane, the inner plasma membrane of cells are substantially negatively charged. So if this domain is also negatively charged, you can imagine that there will be a repulsion between the domain and the membrane. And the more the membrane bends away from the protein, that electrostatic repulsion will be reduced and there should be a preference, therefore, for the more curved membrane.

Jeanne Stachowiak (00:25:13):
So an interesting way to potentially differentiate between these two mechanisms and figure out which one is responsible is to vary the salt concentration. So if you make the salt concentration very high, you’ll screen out the electrostatics. So you would expect this mechanism here to become weaker at high salt concentration.

Jeanne Stachowiak (00:25:33):
In contrast, making the salt concentration very high, makes the polymer floppier because you’re screening out the electrostatics. So it increases the entropy of the domain. So doing that ought to make this entropic mechanism stronger.

Jeanne Stachowiak (00:25:48):
Okay, so down here there’s a little table to tell you, what do we expect if this mechanism is dominant versus if this mechanism is dominant? So we ran this experiment where we varied the salt concentration between 10 nanomolar and 450, or sorry, 10 millimolar and 450 millimolar and measured the curvature sensitivity.

Jeanne Stachowiak (00:26:07):
And indeed, you can see that for this protein, the higher the salt concentration, the higher the curvature sensitivity, which leads us to believe that this protein, its sensing mechanism is primarily entropic. And this was in line with simulation results from Dave Thirumalai’s lab. Some of you guys may know Dave, he’s also here at UT Austin and has been a great collaborator for us.

Jeanne Stachowiak (00:26:31):
But is this always true? Are disordered proteins always sensing curvature through an entropic mechanism? We thought not necessarily. There should be conditions under which the electrostatic mechanism could dominate. So to try to create those conditions, we took just the first one-third of this domain, so that first one-third happens to have most of the negatively charged residues. So the charge per residue is about twice as high in this one-third, and then it’s shorter, so it has less entropy, fewer configurations compared to the full length domain.

Jeanne Stachowiak (00:27:04):
So here we did see the opposite. So here, the highest curvature sensitivity is achieved at the lowest salt concentration. So this seems to be a regime in which the electrostatic mechanism dominates. And then, as you might expect, if you look at some intermediate portion, so here’s two-thirds in terminal fragment of AP180s C terminal domain, which has an intermediate electrostatic charge per residue as well as an intermediate length, of course.

Jeanne Stachowiak (00:27:31):
And here you can see that the curvature sensitivity is almost invariant with salt concentration, probably because of a balance between the electrostatic mechanism and the entropic mechanism. So to summarize that, it seems that long disordered proteins with relatively low net charge will likely interact with surfaces and sense their curvature through an entropic mechanism, whereas in contrast, if a disordered protein is relatively short and charged, it’s likely to interact with membranes through an electrostatic mechanism.

Jeanne Stachowiak (00:28:05):
And I think this is an unusually large disordered domain with a rather low net charge. So probably the second electrostatic mechanism will be applicable to more proteins outside of this one field. Okay, but you can also see this curvature-sensing mechanism in cells. So here we’re looking at filopodia, these little tubular projections off the edge of the plasma membrane. And we expressed this disordered domain on a transmembrane protein chimera and we compared it to these chimeras that don’t have a disordered domain.

Jeanne Stachowiak (00:28:43):
So here the red channel is always reserved for the one without the disordered domain. So when neither protein has a disordered domain, you can see the filopodia just look yellow, and the relative partitioning between the plasma membrane and the filopodia is the same in both channels. So neither one of these proteins is more curvature sensitive than the other, which is what you’d expect since they’re both just fluorescent proteins tagged to the membrane.

Jeanne Stachowiak (00:29:08):
But in contrast, when you express this disordered domain, you see some positive partitioning to the filopodia, which makes the filopodia look green. And then, you can also think about, well, what would happen if you put the disordered domain on the other side of the membrane? Now, we can’t do that with these proteins, with these endocytic proteins, because they would bind all their endocytic binding partners.

Jeanne Stachowiak (00:29:31):
So we use that same neurofilament chimera and we put it both outside as well as inside. And when it’s outside, it does the same thing as we saw with AP180, it makes the filopodia look green. So it’s sensing their curvature and preferentially localizing to them. But if we put it on the inside, instead it wants to move away from the filopodia, and you see them here looking a little red.

Jeanne Stachowiak (00:29:53):
So it seems that even in the complex environment of the native plasma membrane, these disordered proteins can express a curvature preference. So you might be curious, what will happen if you combine a structured protein and a disordered protein since that’s what most of our natural endocytic adapter proteins do. So again, here’s Epsin. It’s got its disordered domain and it’s structured domain, and the same thing with the Amphiphysin.

Jeanne Stachowiak (00:30:21):
So we measured some of these interactions and interestingly what we found is some pretty strong synergistic effects. So here, for example, with Epsin, you can see how much curvature sensitivity the structured domain has, here in this red curve, and then the green curve is showing you the disordered domain alone. But then, when you put them together, you’re on this purple curve, which gives you more curvature sensitivity than you would get from just adding these together.

Jeanne Stachowiak (00:30:49):
And then, it’s even stronger over here with Amphiphysin, that neither one of its component parts is all that sensitive but the full length protein is highly sensitive. And we don’t fully understand that, but our hypothesis is that this dimer brings these two disordered domains so close together that now not only do they have to avoid the surface, they have to avoid each other. And so, there’s a really strong preference for binding a really curved vesicle that will provide more volume above the membrane surface for these disordered domains to get away from the surface and get away from each other.

Jeanne Stachowiak (00:31:25):
Okay. So that brings me to the second part, where I’m going to talk about endocytic protein networks. Maybe I’ll stop for just a minute and see if anybody has any thoughts or questions before we move into the second bit.

Speaker 4 (00:31:42):
Is there any role of ATP or GTP in this part?

Jeanne Stachowiak (00:31:47):
Ah, that’s a good question. So endocytosis is a pathway in which all the hydrolysis of ATP and GTP occurs at the end. So the machinery responsible for uncoating the vesicle consumes energy in that way. So you have these chaperone proteins that go in and take apart clathrin and they also convert…

Jeanne Stachowiak (00:32:09):
So most of these proteins that I’ve been talking about bind the lipid PI(4,5)P2. So energy comes into the pathway directly by two means. One of them is uncoating and the other is converting the lipids into phosphoinositides that the proteins no longer have affinity for. So all those costs are paid at the end to take the vesicle apart. But the process of assembling the vesicle is all spontaneous, they’re all energetically downhill. There’s no ATP hydrolysis involved in the vesiculation. Does that answer your question?

Speaker 4 (00:32:49):
Yeah. Thank you.

Jeanne Stachowiak (00:32:50):
Sure. Okay.

Speaker 5 (00:32:54):
Um, the-

Jeanne Stachowiak (00:32:55):
Sure, go ahead.

Speaker 5 (00:32:56):
Yeah, I’m just curious if you ever tried to swap, instead of their disordered domain by another disordered domain, the disordered domain with and ordered domain and see if you have the same phenotype where you have smaller vesicles or if it’s something that is a property of the disordered domain?

Jeanne Stachowiak (00:33:16):
Yeah, that’s a good question. So for the first bit about bending, before we ever got interested in disordered proteins, we had studied how these steric or crowding effects can drive vesiculation using always structured domains. And then, at some point we realized that the largest volume occupancy, or the largest domains, in terms of how much space they take up and how much steric pressure they could generate, in this pathway are the disordered domains.

Jeanne Stachowiak (00:33:48):
And that was actually what got our lab interested in the disordered proteins is their large hydrodynamic radii, but you can certainly see the same effects with structured proteins if you can find structured proteins that are large enough. Many of these disordered proteins, for example, Amphiphysin’s C-terminal domain has a hydrodynamic radius of almost 10 nanometers.

Jeanne Stachowiak (00:34:10):
So it’s rare to find a structured protein that’s that large, but I think that the effects would be the same if you could. They might even be a bit stronger because the disordered protein is flexible and cannot change its shape, of course, even if that’s entropically disfavored, whereas a structured domain is, of course, more rigid. Then, in terms of curvature sensing, there, I think the disordered domain really does bring something unique. So of course you can just imagine that if you have a vesicle that’s small versus one that’s large and you attach a structured protein of a certain diameter, that the smaller the vesicle is, the more proteins you can pack around it, right?

Jeanne Stachowiak (00:34:56):
It’s like runners on a track. If you happen to get the inside lane, you have not as far to go, right? So there’s more and more space the further out you go. So you should have some curvature sensitivity by very large structured proteins and we can measure that in our experiments, but it should be very predictable and quite linear with decreasing radius.

Jeanne Stachowiak (00:35:20):
Whereas the disordered domains all seem to sense the curvature of the membrane in a way that’s more significant than that. So I think there, these electrostatic and entropic effects are really enhancing curvature sensing.

Speaker 5 (00:35:35):
Thank you.

Jeanne Stachowiak (00:35:35):
Great. Good questions. Okay, well if you think of more, just let me know, but we’ll move into the second bit here, which gets us into condensates, which is the area where you guys probably have the most direct interest. Okay.

Jeanne Stachowiak (00:35:51):
So I should tell you, if you don’t regularly think about endocytosis and membrane traffic, that it’s an incredibly complex pathway. There’s at least 30, possibly as many as 50 proteins involved.

Jeanne Stachowiak (00:36:03):
There’s at least 30, possibly as many as 50 proteins involved in forming a clathrin-coated vesicle, and almost all of those, n-1, are adapter proteins, and clathrin coat being sort of the same regardless of the mixture of adapter proteins. It’s long been a question, why do we have so many adapter proteins? What are all these adapter proteins doing? And it’s really tough to think about reconstituting this whole complex pathway with that many proteins involved in it. But at the same time, by just studying one protein at a time, we’re obviously missing very important, possibly dominant interactions. So we really wanted to do more in terms of trying to study protein-protein interactions, but we were kind of in a quandary about how can we do that meaningfully given that there’s so many proteins in the pathway?

Jeanne Stachowiak (00:36:46):
What we decided to do was look at some of the very earliest proteins. Very early in the pathway, there’s relatively few players involved, so we thought that’ll be the easiest way to start thinking about protein-protein interactions. Fortunately, since this pathway has been studied for so many years, it’s really been nicely mapped where we know which are the first proteins to arrive, then what’s sort of the second wave and the third wave and so on and so forth. In this pathway, the first wave, which has been characterized in all these references you can see on this slide, is the arrival of what we call this pioneer complex or the initiator proteins, and this complex consists of a couple of proteins. One is Eps15, and the other is Fcho1, or Fcho1 has a second isoform Fcho2, and Eps15 has another isoform called Eps15R. But these two families, Eps15 and Fcho.

Jeanne Stachowiak (00:37:42):
Both of these proteins are dimers. Fcho, we’ve already talked about the F-BAR, so that’s that more shallow BAR domain, and it has a morphology that looks a lot like Amphiphysin. It’s got this disordered middle domain, and then it’s C-terminal domain, rather than being an SH3 domain like it was in Amphiphysin is this thing called a muHD domain. That muHD domain interacts with the disordered portion of Eps15 in this complex that you can see over here. Then Eps15 we haven’t introduced yet. It has these EH domains, and they are pit binders, as is the F-BAR, so these domains all stick to the membrane by interacting with PI(4,5)P2. But then Eps15 has this coiled-coil domain, and then it has this disordered domain that interacts with Fcho, and the disordered domain also can bind back to the EH domains, so this protein has multiple ways of generating a multivalent network.

Jeanne Stachowiak (00:38:41):
Okay, so we started looking at these proteins and trying to figure out what they would do if they were together on the membrane surface, and this is the work of two people in the lab, Kasey Day, who’s a postdoc, and Grace Kago, a graduate student. So Kasey and Grace, they started doing this work, and at first they made these giant vesicles, and they purified full-length Fcho and Eps15, which at least from mammalian cells had not been done before, so that was a feat in itself. We have a lot of help on that from our collaborator, Eileen Lafer. These proteins by themselves are not very interesting. We add them to a giant unilamellar vesicles. You can see a cross-section through the vesicle as well as a 3D projection, and there’s not really any interesting arrangement going on here. Really just a uniform coating of the membrane. But then if you add them at the same time to the membrane, you get this interesting partitioning between a denser phase here and a more dilute phase, And you can see that in the projection. You see kind of a little cap, like a baseball hat on the vesicle there. Sometimes rather than seeing just one cap, you’ll see a bunch of little spots that wiggle around, but no matter what, you see this partitioning between a brighter phase and a less bright phase. If you reduce any of the multivalent interactions between these, so if you mutate this disordered domain to make it not bind in the muHD domain, or if you remove the ability of Eps15 to dimerize or make various mutations that reduce the affinity here, then you’ll not see these domains form nearly as often. It seems that they depend upon these multivalent interactions.

Jeanne Stachowiak (00:40:27):
Then giant vesicles are nice, because you can see these large domains, but they make it harder to really monitor what’s going on in the membrane surface. We also do these experiments on flat membranes, supported bilayers. Again, on supported bilayers, no assemblies without adding both proteins, but when you do add both proteins, you see these assemblies. When the assemblies bump into each other, they’ll fuse together, indicating some sort of a dynamic behavior. You can also FRAP them, and they’ll recover. This made us think that there’s a lot of the properties we think of in terms of condensates, so we’ve got intrinsically disordered proteins, dependency on a multivalent network. There’s some merging going on and some molecular exchange, so we started to wonder whether these two proteins might be forming some sort of condensate.

Jeanne Stachowiak (00:41:16):
I’ll say on our part, this was a reluctant conclusion. We had, as you’ve seen, studied disorder domains independently for years, and we really think about them as sterically repelling each other, so the idea of thinking of them making this condensate was a little foreign to us. But you can see that when we take these proteins in the absence of the membrane and mix them together with a bit of PEG in solution, you do see these droplets raining down under the cover slips. Here’s a closer up of the droplets fusing together, and you can FRAP the droplets and see them recover, so they seem to have the properties of a lot of the other droplets that you see in the literature.

Jeanne Stachowiak (00:41:56):
Then once you have this droplet system, you can think about mapping phase diagrams, so you can heat the droplets up, and they’ll dissolve and go away. Then when you cool down the same solution, you’ll see the droplets come back. We, like many others, have recognized that you can use these images of droplets at different temperatures to make a relative phase diagram, so you can measure the approximate protein concentration inside compared to the approximate protein concentration outside. That will let you map this arbitrary units phase diagram here. Remember I told you that Eps15 has an ability to make a multivalent network all on its own and, indeed, if you cool Eps15 sufficiently, so here we’re below physiological temperature, you can get Eps15 to form droplets. But then if you add a bit of Fcho, the critical temperature will go up substantially. The more Fcho you add, the more the critical temperature will go up, so it seems that these two proteins are stabilizing each other in a mutual way, which is in line with what we saw in the membrane surface that we couldn’t see protein aggregation or protein condensates on the membrane surface without having both proteins.

Jeanne Stachowiak (00:43:09):
It was really important to us to test what are these structures doing in the cell, if anything, and we had to think pretty hard about how to do that. But Kasey had a clever idea, so she said, “Remember,” back from that slide where we were doing all that mutations on Eps15, “that this coiled coil domain that makes this protein a dimer is really essential. If we don’t have that coiled coil domain, no matter how cold we go, we never see this Eps15 form any condensates.”

Jeanne Stachowiak (00:43:40):
So, she thought, “Let’s borrow this clever idea from this paper by Cliff Brangwynne.” I’m sure many of you have seen this, where you can use this CRY2 domain to make a light-activated clustering of IDRs and see droplets show up in the cell. So Kasey adopted that, and she replaced this coiled-coil domain with CRY2. So this is still in vitro, but she took this purified construct, and when she shines light on it, she can get these condensates to appear in the presence of the blue light. Then, she decided to take this tool into her mammalian cells. Here I have to give you just a moment or two of background on how we look at the dynamics of endocytic events, since a lot of you probably aren’t in this field of studying endocytosis.

Jeanne Stachowiak (00:44:27):
Here is the plasma membrane under TIRF illumination of a mammalian cell, and this is a [som] cell, another cell with this really broadly spread, another epithelial cell, and it’s also one that’s pretty easy to gene edit, which is the reason we chose it for this experiment. Here, this cell line originally came from Tommy Kirchhausen, so we thank him for it, and he had used gene editing techniques to put a halo tag on AP2. AP2 is one of the major adapters of the clathrin-coated pit, so it’s sort of a clathrin-coated pit marker. So you can see all these coated pits here, and then we’ve labeled Eps15 with mCherry. If you fix your eyes on any one of these little dots, you’ll see it show up and get brighter and then eventually go away. Here we’ve zoomed in on just one, and you can see it go through that cycle of appearing, getting brighter and then going away. These are these endocytic events, and at the plasma membrane, the average endocytic event lasts about a minute, but they can vary from 20 seconds to two or three minutes.

Jeanne Stachowiak (00:45:33):
Endocytic pits can basically be categorized into one of three outcomes; some pits are there at the plasma membrane so briefly, usually under 20 seconds, that it’s been shown that they don’t actually capture any cargo and they never form a vesicle. They’re just stochastic assemblies of adapters and clathrin that then disassemble. Then at the other extreme, clathrin-coated structures that stay at the membrane for more than about three minutes, if you follow them, you’ll find out they’ll stay there for tens of minutes. These are what we called stalled structures, so they’re not really productive. They’re just staying at the plasma membrane. There have been some recent work suggesting that these structures have roles in signaling, maybe doing things other than endocytosis, but then the majority of the structures are in the middle and they’ll form a productive vesicle and depart anywhere from 30 seconds to a couple of minutes.

Jeanne Stachowiak (00:46:26):
In wild type cells, this is a typical distribution, so a small fraction of coated pits are in this abortive category, a few are in the stalled and most of them are productive. This is the cells expressing in the wild type version of Eps15. But then when you knock out Eps15, the main change that you have is that there’s an increase in these abortive structures. That kind of makes sense because Eps15 is thought of as one of the main catalytic proteins for getting endocytosis started, so if one of those catalysts is missing, you’re going to have more structures that don’t get off to a successful start and ultimately abort.

Jeanne Stachowiak (00:47:09):
Then, Kasey expressed this CRY2 chimera here, but she was careful not to expose the protein to any blue light. Here, you’ll see that that abortive phenotype is maintained. In fact, it’s maybe even increased a little bit, possibly because this protein is blocking sites that could be occupied by compensatory proteins. But then when you shine some blue light on these cells, you can recover a fairly close to the wild type distribution of dynamics, so you can see the abortives go back down close to wild type levels. At this point, we thought, “Oh, this is exciting. We’ll keep shining this blue light, and we’ll drive the abortives down to zero and make endocytosis more efficient.” We thought this would be really fun, but interestingly, when we turned up the light, we didn’t quite get what we expected.

Jeanne Stachowiak (00:47:58):
We did get a small reduction in the number of abortive pits, but most notably we got this huge increase, about a four-fold increase in the number of stalled structures. You can see the stalled structures in this video here. You see these large endocytic structures that, some of them, if you fix your eyes to them, you’ll see they persist through the whole movie. This is, I think, a 10-minute movie, so they’re clearly lasting a lot longer than a normal endocytic structure. So what’s going on with that? Why is more not better? We were really perplexed by this, so we went back to our droplets and we put different levels of blue light on them, and I’ll let this movie cycle around again.

Jeanne Stachowiak (00:48:39):
What we saw is that the level of blue light that had been giving us a rescue of dynamics gave very fluid droplets, whereas this level of blue light that caused the stalled structures was giving us still nice rounded droplets, but they didn’t fuse. They formed these gel-like structures, where you can see two that have stuck together never re-round over a long period of time of multiple minutes. Okay, so we also saw that if we FRAP these same droplets, the recovery time of the droplets formed under low light was very similar to what we get with wild-type Eps15, whereas in contrast, when we applied that strong light, we got a much slower FRAP recovery time.

Jeanne Stachowiak (00:49:26):
We began to think of this problem kind of like a classic catalyst working in a reaction. So Eps15 and Fcho are these early catalytic proteins, and like all catalysts, they really have two jobs they have to accomplish. They need to bring together reactants, and we’d been kind of focusing on that function thinking more blue light will be better, but they also have to release the product, and in this case, the product is the coated vesicle. If the coated vesicle can’t be released, then the catalyst is not efficient, so we began to wonder whether this fluid light catalytic complex really gives us a balance between these two functions that were still efficient in bringing together reactants, but we’re allowing the rearrangements that are necessary for release of the vesicle.

Jeanne Stachowiak (00:50:12):
To think a little bit deeper about this, we looked at the next adapter that’s recruited after Eps15 and Fcho assemble, and that is AP2, which I depicted here in this little yellow cartoon. Here, we FRAP’ed AP2 in the presence of these different light levels that had driven assembly of Eps15-CRY2, and what you can see is that the wild type recovery of AP2, so AP2 in the presence of wild type Eps15 is this black curve, and if we put in CRY2-Eps15 and don’t apply any light, we get to rapid recovery of AP2, as if AP2 isn’t being drawn to the pit long enough to drive a productive membrane vesiculation event. But in contrast, the low light level brought us back to the wild type level of exchange of AP2. But if we keep going and make Eps15 and Fcho2 solidly assembled, then AP2 spends too much time at the coated vesicle, indicating that it probably can’t rearrange enough to drive the membrane remodeling and the formation of the vesicle.

Jeanne Stachowiak (00:51:23):
Okay. All right, so that was a lot. I hope I didn’t overwhelm you with too much information, but just to conclude here, I’ll say we’ve talked about how disordered proteins can can drive and sense membrane curvature and how coupling them to an ordered protein like a BAR domain can really lead to some synergy, where the structured portion and the disordered portion are working together to bend and remodel the membrane. Then, that these protein droplets, we’re really still learning about what their function is in endocytosis, but they may be providing a fluid, and therefore really efficient and flexible, catalytic platform for initiating endocytosis. I’ll stop there and thank my students, who really have made all of this happen. The first bit was from Wilton Snead, who’s not in this picture because he recently moved on, but then we talked about the work of Wade Zeno and primarily Kasey Day. All right, thanks guys. Please let me hear your questions. Oh, well I don’t know. I’ll see if we should stop sharing or not. Maybe we can for a minute and see if you want to go back to any of the slides.

Diana (00:52:34):
Thank you so much. That was wonderful.

Jeanne Stachowiak (00:52:37):
Thank you.

Mark (00:52:41):
Yeah. Beautiful stuff. Really wonderful. I keep thinking about the binding energy that drives the formation of these dimers. I wonder, because each monomer is already embedded in a membrane, so entropically they’re near each other to begin with, so I wonder if they’re fairly weak.

Jeanne Stachowiak (00:53:05):
They seem to be strong.

Mark (00:53:06):
They are strong.

Jeanne Stachowiak (00:53:08):
We can take either protein and dilute it. If you’re familiar with FCS, you run those experiments under very dilute conditions, 10 nanomolar, and we’ll still measure a big hydrodynamic radius compared to if we truncate that domain and measure the monomer. Everybody has always said that, well, you can kind of see it in the structure of BAR domain. You’ve got these long alpha helices that are really too long. A BAR domain as a monomer, those helices would bend but or would kink, but because they overlap each other, they’re stable. You have such an enormous overlap. I think that’s what makes the KD so low for the dimer.

Mark (00:53:49):
Yeah. It’s really nice. It’s so fundamental in the way you studied these systems and so beautiful. These are such complex systems, and you’ve dissected them so beautifully.

Jeanne Stachowiak (00:54:00):
Oh, thank you. I appreciate that.

Mark (00:54:02):
… beautifully.

Jeanne Stachowiak (00:54:02):
Oh, thank you. I appreciate that.

Speaker 8 (00:54:04):
I have a question about cargo.

Jeanne Stachowiak (00:54:08):
Sure.

Speaker 8 (00:54:08):
So when a cell needs to form some kind of vesicle, I’m thinking the purpose must be transporting some kind of cargo, right?

Jeanne Stachowiak (00:54:21):
Yes.

Speaker 8 (00:54:22):
Mainly protein. Maybe also use that to transport some lipid.

Jeanne Stachowiak (00:54:26):
Sure.

Speaker 8 (00:54:27):
But it’s mainly protein. So how does the cargo play a role in forming the clathrin-coated pits? How does it know, “Okay, I have a vesicle that is full of angiotensin recptor or full of [inaudible 00:00:47]? So how does cargo …

Jeanne Stachowiak (00:54:49):
That’s a great question. Yeah, I’ll tell you what I know about it. So these cargo adapters, most of them, they bind three things: lipids, a cargo, which is a transmembrane protein, in the terminology of this field, and clathrin. We didn’t talk about cargo, as you pointed out. So most cargo have a tail that extends into the cytoplasm, and that can be anything from a very short motif to a large, folded domain.

Jeanne Stachowiak (00:55:20):
But specific biochemical motifs, they’re usually four-letter codes. The most classic one is YXXPhi. So you’ve got two amino acids that don’t matter, and then the Phi refers to something hydrophobic, and the Y on the other end. So that motif will recognize AP2, which is the best studied adapter in the clathrin pathway.

Jeanne Stachowiak (00:55:43):
There have been studies that have shown that if you overexpress a cargo, that will increase the recruitment of AP2 to the membrane and give you more endocytic events per area of the plasma membrane. So, in that sense, endocytosis is cargo-dependent, but the increase you’ll get is not particularly large, maybe 20, 30%. You’re not going to double or increase the number of endocytic events by tenfold. Endocytosis is largely constituitive. It’ll go on and on, even if the number of cargo per vesicle is relatively low or if the number of cargo is relatively high.

Jeanne Stachowiak (00:56:20):
I think that is less understood. Exactly how is it that endocytic vesicles are as cargo-independent as they are is an ongoing area. We had a project for a while where we were trying really hard to see what it would take to stop the coated vesicle. So we made these cargo with these ridiculous disorder domains displayed on the outer surface of the cell, and we would find vesicles where they were very crowded together. We were trying to generate some steric pressure to oppose endocytosis, and it turns out that’s very hard to do. Mostly what you get under those circumstances is just fewer cargo per vessel. It seems that this sort of interwoven network of adapters on the coat side, there’s such a high energetic barrier to disassembling that that almost nothing a cargo can do, whether it clusters or is very large will really even slow down a coated vesicle, which is kind of curious. We’re still kind of trying to understand how that pathway is robust as it is.

Speaker 8 (00:57:30):
Thank you. That’s fascinating.

Carolyn Sayer (00:57:39):
Thank you. Oh, go ahead, Avi.

Avinash Patel (00:57:43):
Oh, sorry, Caroline. You go ahead. You were first.

Carolyn (00:57:45):
No, you go first. You’ve got your video on.

Avi (00:57:48):
Okay. The phase diagram is quite striking, right? So if you look at the shape, it looks not kind of a like a bimodal, as you would expect. Right? One side looks more close to a spinodal straight, and one is more gradual, more like a bimodal. Any comments on what might be determining such a …

Jeanne Stachowiak (00:58:11):
You guys are the phase diagram experts. Maybe I’ll pull it back up again here, and you can tell me what it means to you, because we have noticed that and not really known how to interpret it. But I’ll bring it back so you can get a proper stare at it. Let’s see. I have to figure out how to efficiently advance my slides. It’s not too far back here.

Jeanne Stachowiak (00:58:41):
Oh, wait. There it was. Sorry. Got carried away. Okay. Yeah, there you have it. So you’re commenting on this side being very straight, whereas this one is much more tapered.

Avi (00:58:56):
Yes. Exactly. So I don’t know how to interpret that. Right? It seems like you somehow … and this is in the presence of … Are these just binary components, both the Fcho1 and Eps15?

Jeanne Stachowiak (00:59:09):
That’s right, or no, they are both dimers.

Avi (00:59:12):
Yeah.

Jeanne Stachowiak (00:59:12):
Yes, yes. But you’re right. This is binary. So there’s only two proteins here, and this isn’t the CRY2 system.

Avi (00:59:19):
Yeah.

Jeanne Stachowiak (00:59:20):
I guess this over here is telling us that the droplet itself has just sort of one concentration. Its concentration doesn’t change very much as we increase temperature …

Avi (00:59:34):
Exactly.

Jeanne Stachowiak (00:59:35):
… whereas the concentration of liquid’s varying quite a lot.

Avi (00:59:38):
Yeah. That’s pretty interesting. That is something I haven’t seen before.

Jeanne Stachowiak (00:59:45):
If you know what to make of it, tell me. I know an alternative. When I see something interesting, I first ask, “Can it be an artifact?” Here, we really just wanted to know the shift in the critical temperature. But the way we’re measuring this is just by comparing relative brightnesses. So I guess I don’t immediately see how this can lead us astray, but it’s worth thinking about. Would we expect our camera to be … though I think all these images are pretty … None of them are anywhere close to saturating the camera, and we have a nice, linearized EMCCD. So it’s not immediately apparent to me how we’re measuring it wrong, though I want to think more about that. Should we really trust how straight this is?

Avi (01:00:35):
Yeah. Maybe one other way would be maybe complementing it with alternative non-fluorescence-based measurement. Right? Something like that.

Jeanne Stachowiak (01:00:46):
Yeah. We have trouble getting enough of this protein to do a more bulk measurement, but we could do FCS, though all experiments are tricky. That one will be a little tricky to get it just dilute enough in this phase, but that can be accomplished with enough fiddling around.

Jeanne Stachowiak (01:01:06):
Another possibility maybe is maybe we need to … Well, no. I was about to say maybe we need to plot this differently, that there is a slope to these, but it’s just so overwhelmed by this little bit. But that in itself is what’s odd here.

Avi (01:01:18):
Yeah.

Jeanne Stachowiak (01:01:20):
Yeah, yeah. If you have any theory about what that could mean, let me know.

Avi (01:01:25):
Yeah, sure.

Jeanne Stachowiak (01:01:26):
I’d be glad to learn from you.

Avi (01:01:27):
Yeah. Okay.

Jeanne Stachowiak (01:01:29):
I’m going to go out of sharing.

Carolyn (01:01:36):
Thank you so much for this talk. It was really interesting. I wanted to ask a general question. In the first story, you were talking about the lipid that all the proteins are interacting with.

Jeanne Stachowiak (01:01:46):
Yeah.

Carolyn (01:01:46):
Can you tell us a little bit about how you identified that lipid or how you confirmed … Was it P(1,5,4) or something like that?

Jeanne Stachowiak (01:01:55):
Yeah, PI(4,5)P2?

Carolyn (01:01:55):
Oh, okay.

Jeanne Stachowiak (01:01:56):
Yeah. So it’s not me who identified it. So this is quite an old field. So people have been studying these adapters for at least 20 years, maybe 30, and the early experiments that were done, people would take different phosphoinositides in that pathway, PI(4)P, PI(4,5)P2, PI(3,4,5)P3, all the ones that make up this pathway of the metabolism of PIPs. They would basically do Western blots against PIPs. Okay? You can still buy these things today, the so-called PIP strips. Avanti will sell them to you.

Jeanne Stachowiak (01:02:36):
Also, it’s known that the predominant PIP at the plasma membrane is PI(4,5)P2. So it also kind of makes sense that these proteins would bind PI(4,5)P2. Then, as the vesicle progresses, so we think of every organelle, whether it’s the plasma membrane or the endosome or the lysosome as having its own PIP signature. So, gradually, the PI(4,5)P2 is turned into PI(4)P to make this thing look like an endosome, and then endosomes don’t bind these proteins that are plasma membrane proteins, which is all the adapters. So the adapters will hop off, and then these chaperones take apart clathrin. That’s how you uncoat the vesicle.

Jeanne Stachowiak (01:03:21):
So yeah, PI(4,5)P2 has, yeah, been studied by a lot of folks and is sort of the plasma membrane PIP.

Carolyn (01:03:30):
Great. Thank you.

Jeanne Stachowiak (01:03:31):
Sure.

Diana (01:03:41):
Hi. I had a, sort of a technical question.

Jeanne Stachowiak (01:03:44):
Sure.

Diana (01:03:45):
So in your CRY2 experiments, the in vitro ones where you overilluminated the drops and they fell into the gel-like state …

Jeanne Stachowiak (01:03:56):
Yeah.

Diana (01:03:57):
… once you remove the light, do they dissolve or they stay as a gel?

Jeanne Stachowiak (01:04:05):
The ones that are more fluid, if you remove the light, they’ll disassemble.

Diana (01:04:10):
Right.

Jeanne Stachowiak (01:04:10):
But the ones that are more gel-like, when you remove the light, they don’t disassemble nearly as much. They might get a little smaller or dimmer, but you won’t really see them totally go away.

Diana (01:04:22):
That’s interesting. So could the stall sort of be acting as kind of a sink, just sucking out all those proteins that are in the vesicle experiments?

Jeanne Stachowiak (01:04:35):
That’s a good thought. Yeah, that’s true. So when we run the experiment, we have sort of a fixed duty cycle of the light. Right? Meaning how often we’re flashing and for how long, and that is one question we’ve had. Our movies are relatively short, so I’m not sure whether we’re reaching an equilibrium. With the more fluid one, the lower light levels, you might think, “Okay, if you go on long enough, you’ll probably reach some equilibrium level of assembly.” But if you have a protein that’s forming a gel-like structure, then sucking all the protein to the membrane might be your equilibrium.

Jeanne Stachowiak (01:05:22):
That’s probably part of this stall, is not just the slow recovery, but also just the growth of the thing. You can see some large endocytic structures by the end of that movie. So yeah, I think that probably does play a role.

Jeanne Stachowiak (01:05:39):
One experiment we’ve wanted to do is to flash the light for half the movie and then turn it off and see what happens. I bet we would find that with the low light level, we’d go back to our underlying dynamics that we had before, which would be the dynamics of the knockout, whereas with the high light level, we at least wouldn’t relax that far, if at all. That makes me more motivated to try that experiment.

Diana (01:06:11):
Thank you.

Jeanne Stachowiak (01:06:12):
Sure. Thank you.

Mark (01:06:13):
Sorry. One other thought, really listening to the whole talk, is I’m certainly not remotely an expert on this, but I’ve heard other people who’ve thought deeply about the physics of a tumor talk about the pressure that tumors are under. So I was reading some work, for example, from a professor named Liam Holt, who’s up in New York at NYU, who studied, for example, how, in pancreatic cancer, the actual tumor cells are under tremendous pressure. So I’m just thinking out loud here. I’m wondering whether the mechanics of membrane maintenance are somehow different.

Jeanne Stachowiak (01:07:08):
That’s an interesting thought. So one field in which that really intersects with this is yeast. So yeast, as you know, have this turgor pressure. They would explode if they could, if it weren’t for their big coat. Yeah? So endocytosis was probably first studied in yeast because of the ease of genetic manipulation in yeast, and there’s so many people who study endocytosis in yeast.

Jeanne Stachowiak (01:07:37):
What they have found is that to overcome that turgor pressure, actin becomes essential to endocytosis. So the bud starts to form, and then actin assembly is what pushes the bud in and allows it to actually complete. In yeast, if you inhibit actin, endocytosis is dramatically knocked down, whereas in mammalian cells, you can inhibit actin, and at least in the short term, endocytosis barely notices. Though, further, it’s been observed that if you stretch a mammalian cell, so people have actually plated cells on these flexible substrates and stretched, then you can induce an actin dependence, even in a mammalian cell.

Jeanne Stachowiak (01:08:19):
So that’s interesting to think about that. Yeah, it’d be probably pretty simple to see. Has anybody ever measured actin dependence in tumor cells of endocytosis? Of course, all of us are studying tumor cells, whether we like it or not, unless we’re trying really hard not to, but I don’t think our cells have that same pressure. Right? That’s not just being a tumor cell. It’s the physiology of the tumor.

Mark (01:08:46):
Yes, yes.

Jeanne Stachowiak (01:08:48):
With that pressure, I’ve heard of tumors under pressure, and if you try to … A part of my work is in drug delivery, which I didn’t talk about at all, but when we’re thinking about that, we worry about, oh, that the particles we’re creating are going to be flushed out of the tumor. So fluid’s flowing out of the tumor. Right? So I’d have to think about the geometry. What’s that doing to the membrane? Does it stretch or compress?

Mark (01:09:12):
Yeah, it’s some interesting tensegrity balance.

Jeanne Stachowiak (01:09:16):
Yeah.

Mark (01:09:16):
I don’t know exactly how it works, but it’s funny, because …

Jeanne Stachowiak (01:09:20):
I think you can imagine, yeah, maybe the tumor, if there’s fluid that’s trying to get in there, the whole thing is sort of stretched. Yeah. So membrane tension probably is higher in tumors. I wonder how much is known about that.

Mark (01:09:31):
Yeah, because I think it probably is a whole separate field. It’s amazing how everywhere you look in biology, things are more complicated than you think. I mean, a membrane seems like it’s such a simple thing. But, of course, [crosstalk 00:15:45].

Jeanne Stachowiak (01:09:43):
Oh, yeah, there’s this guy. Oh, where is he? I think University of Iowa. He’s got a Turkish name, Kural Comert or Comert Kural. I always mix up his first and last, but he was out of Tommy Kirchhausen’s lap, and he has been studying endocytic pit dynamics during development. So, in that context, he sees cells that are undergoing morphogenesis. So the membranes are stretched while these cells are moving, and he’s measuring endocytosis during that. He has found some interesting effects there.

Mark (01:10:18):
Yeah, that’s beautiful. I’ll look him up.

Jeanne Stachowiak (01:10:21):
Yeah, I think he’s Iowa. Is that right?

Mark (01:10:24):
Okay.

Jeanne Stachowiak (01:10:25):
No, Ohio, Ohio.

Mark (01:10:27):
Ohio. Okay.

Jeanne Stachowiak (01:10:27):
The Midwesterners will be annoyed with me to mix up Iowa and Ohio.

Mark (01:10:31):
Yeah. So other questions? Diana, anything else?

Diana (01:10:44):
I don’t think so. Yeah, I think I’m out of questions.

Jeanne Stachowiak (01:10:55):
Well, thank you guys so much.

Mark (01:10:57):
Oh, it’s been wonderful. Thank you for doing this.

Jeanne Stachowiak (01:10:59):
Sure.

Mark (01:10:59):
It’s been a load of fun.

Jeanne Stachowiak (01:11:00):
It’s a lot of fun.

Mark (01:11:01):
We may have other questions that we’ll follow up with you.

Jeanne Stachowiak (01:11:04):
That’d be great. Yeah.

Mark (01:11:04):
But for now, I think we’re good. All right. Thank you again.

Jeanne Stachowiak (01:11:07):
Great. Have a good one.

Diana (01:11:09):
Thank you so much.

Mark (01:11:10):
All right. Do take care. Okay. Thank you.

Diana (01:11:12):
Thank you.

Jeanne Stachowiak (01:11:13):
Bye.

Mark (01:11:14):
Bye-bye.