 So as I was saying yesterday, we talked about signals, mostly as ways to just dump something into the culture media and get these patterns. And today, I'd like to talk about the signaling pathways as dynamic objects that actually interpret those signals and have influences on how these self-hates form. But before I do that, I didn't quite have time to say everything I wanted to say about these patterns yesterday. So I'm going to spend the first few minutes doing that and then switch gears to talking about signaling dynamics. So just to remind you of what we were talking about yesterday, we have these micropattern culture systems. They're made by making complementary patterns of cell and protein phobic things on things that the cells like to stick to, like major gel or laminin on the cell surface. When you seed the cells onto them, the cells adhere only in these patches. And then they're confined to these patches so that the confinement both gives regularity to these systems and increases cell density so that you get these patterns emerge and these patterns contain all the germ layers along the radial axis of the column. So you have ectoderm in the center and the red is mesoderm and this is extra embryonic tissue. If you stain for more things, so this green is mesendoderm and this red is endoderm. So you have all three germ layers and the extra embryonic tissue forming on the radial axis of these colonies. And I talked a little bit yesterday about how this is the response both to the primary signal, so the primary signal, primarily signals to the exterior cells in the colony and they respond and differentiate to these extra embryonic fades and that creates a secondary nodal signal which positions these extra fades in between and then these cells in the center which don't really see any of those signals defaults sort of to this neural or ectodermal fate that happens at the center. And now I'd like to talk a little bit about what happens when you change the shapes of these patterns. So this is a nice thing that you can do with stem cells that you can't really do in other developmental systems is that I can really play with the geometry and make any shapes I want and I can target different shapes to ask different questions, right? Like there's been a lot of debate in the developmental literature about diffusion and how far can things diffuse and do diffusable signals from certain cell populations influence things from other cell populations and so we can pretty cleanly just make these colonies with gaps in them and then ask do we see any differences in the outside if the center is here or in the center if the outside is here and so on, I'll tell you a little bit about that and then if we're interested in sort of house cells sense shape and corners we can engineer corners into these things or we can engineer polygons and things like that and then ask if there are differences and what's in the corner and not. So I'll tell you some results from these things but we think they're sort of still quite descriptive we see differences in all these cases we don't really understand what we see if you have ideas I definitely welcome them. So first, thinking a little bit about corners so are there effects of corners? The answer is yes and so we've made a series of these regular polygons to ask what happens so these are sort of representative images of what you get on these different polygons and these look like cartoons but they're sort of actually data so if you take averages over many shapes like this and then threshold the cell fades so that where you get this red extra embryonic is shown in red here and where you primarily have this green is shown in green here and so on that you see that whereas in a circle everything is sort of perfectly symmetric around the circle and so the fades are all equidistant from the boundary here you're pretty close to that but you start to deviate and then in these other states you start to deviate more so you see these effects of these corners in inward shifts and widening of the territories particularly widening of the territory of this mesodermal territory. It can be a little bit more quantitative about that so if I think about taking a region which is sort of shaded in yellow here and asking what taking sort of an angle from the center and then asking as a function of that angle from the center what's the radial what's the sort of average as a function of angle of these different markers right what you see is particularly this brachiary marker so the corners are denoted with dashed lines here and this brachiary marker will peak on the corners but that becomes less sharp or almost imperceptible here and then when you get a circle you'd get a totally flat line so we see these effects of corners we think the cells are sort of sensing the corners that could be due to some kind of diffusible signal it could be due to some kind of mechanical sensing of corners we don't really know the answer to that yet. Okay so on to thinking a little bit about diffusion so what if we make shapes like this and compare them to shapes like this and so the question here is I told you yesterday that fates are formed from the edge inward so is there any difference from the edge inward in this inner circle and this outer circle and the answer is really there isn't right so if you look at these fate markers here and you look at these fate markers here which are these are the inner circle you don't see any difference so having this ring around it doesn't influence the inner circle interestingly if you do the converse experiment so you think about what's happening in this outer circle and then you either don't have an inner circle you have a small inner circle you have a large inner circle you actually do see an effect so we're seeing diffusive signals which we think are they essentially reinforce the fate in this outer circle so you get pure and better expression of the extra embryonic markers in this outer circle when you have these central circles and it's sort of a consistent effect where here you have none here you have a small circle and then it's enhanced even more if you have a larger circle and it's sort of interesting right because the patterns we know form from the inside outward so the original signal has to go inside that's what makes you know you get a sequence of fates inwards but then there seems to be some reciprocal signaling from the outside back to the from the inside back to the outside which is important for reinforcing those cell fates on the outside of these patterns okay sorry so this is the intensive sorry it's not well labeled this is the intensity of the CDX2 marker in the outside trade so this is the marker of the trophactodermal fate which is what you would typically get in these outer colonies what you see here is that it's this both results from the cells expressing more of it and it being more pure so you get less of other intervening cell fates in this outer region when you have this circle in the center so somehow there's communication from the center that reinforces the fate so they're not switching fates these are the fates that they would normally adopt but they do a better job of adopting it when there are cells in the middle something like that yeah yeah so I mean the way these experiments are done right you pattern the cells here and you drop we drop this BMP morphogen everywhere sorry and the right there's two primary sources of response to the BMP morphogen one is in the circle itself and the other is in these the outer edge of this inner ring right because the inner ring makes the same pattern it's a circle it would usually make but then when we see that happening right we see that also reinforces what's happening in this outer circle it's outside the cells as well actually so a lot later I'll show you some but the cells will take it up and the ligand will be depleted over time but that really only happens at low concentrations of the BMP so if you have a low concentration and you watch the signaling response you'll see a decay and you can tell that's due to ligand depletion because if you take that same ligand and move it to another well of naive cells right you'll see that decay just continue from where it left off so it's really like the ligand is being removed extracellularly but at the concentrations we use to get these fades that's it's not really a factor right so there's plenty of ligand around they don't really eat up the ligand outside the cells right so we think most of these effects are actually secondary signals actually I was pretty speculative but our best guess for what's happening is that actually you reinforce these fades by inhibiting the secondary signal so what happens in the central colonies we think here you make nodal but you also make lefty inhibitors and it's thought that the lefty inhibitors diffuse further than the nodal so we think these inhibitors of secondary fades get to these outer colonies and then they prevent them from adopting these secondary fades and a little bit of evidence for that which I'm not showing you is if I take these circles and I make them asymmetric so I move them closer to these things I actually see the opposite effect so I see that I get some inhibition of this fade and a little bit of up-regulation of the secondary fade so we think that the sort of central things are sources for either the secondary morphogens or inhibitors of the secondary morphogens and at certain distances you're more likely to get inhibitors of those secondary morphogens and that reinforces the primary fade I don't know if that makes sense okay so then we wanted to ask a question so it seems like the boundaries in these colonies are very important for you always get the same pattern from outside to inside and the sort of naive expectation of if you didn't have a boundary was that you would get everything adopting the inside fade and isn't that true and so postdoc in my lab it's a Hume's Kirk had a good idea which is what if we can get beads we can grow cells on the surface of the beads and so that if you have the same surface area as these normal colonies but it actually has no boundary it's just making a complete surface around the sphere and then we differentiate them in the same way and ask what you get and so here are some of these pictures of these things what you find is that these things will almost always autonomously polarize such that you get different fates on the different sides of the sphere right so roughly these are labeled in DAPI but if we were to label them they would label them they would label for these extra embryonic fates and then you get these mesodermal fates on the other side and so we think that the patterning system that makes these fates is sort of capable of totally self organizing these patterns without an asymmetry from the boundary but that if you do have a boundary that it will bias that asymmetry such that you position particular fates in particular places yeah so you take like a polystyrene bead I think it is and then you coat that bead with laminin and then you soak that bead in a cell suspension so that the cells stick all around the bead and so I'm actually I'm not showing you the initial distribution but initially the cells are uniform around so you'd see a flat this sort of even cross section around and then as the cells differentiate they actually do what they do in these 2D cultures which is where they express brachiary they kind of pile up where they have these extra embryonic fates they're flatter and more spread out and so they've sort of recapitulated at least some of this cell fate pattern on the surface of these beads when they differentiate so that makes sense the yeah so I think these beads might have to be a little bit bigger to get that but I think we will yes we think yes and so an interest so well okay I'll show you something this is kind of a bigger bead so we've actually sort of fortuitously come across cases where you've got two beads glommed together and then what we always find pretty much is that you recapitulate the whole pattern but one cell fate forms on one bead one cell fate forms on another bead and the mesoderm forms in the middle right so here this is actually the epidermal fate on this bead here this is the mesoderm in the middle and this is the extra embryonic fate on this side and so this kind of right so here there's no sort of intrinsic bias there's no real boundary but there's there's kind of geometry constraining this thing and then these things somehow pick one sphere to be one fate one sphere to be another fate another thing in the middle yes sorry what oh it's polystyrene so it's like a plastic bead basically so it's basically the same stuff we would grow cells on in the culture surface but in a bead of I think it's 200 micron diameter something like that and so the cells are if you code it right the cells can just grow on the surface of it right so I mean you can think of this is a cross section through the beads so these are not it's not a solid ball of cells right it's a sheet of cells growing on a curved surface of a bead right so and so the idea was to as closely as we can mimic these 2D culture systems but not actually have a boundary so we could see if the cells could pattern themselves without having the boundary okay the last thing I want to talk about about this is we also got interested in is this a general strategy so we've done this for early embryonic patterning we've made different germ layers but can we take the same strategy and pattern later fates in development right so is this only applicable to the sort of early stages of development starting from stem cells or would it be possible to start with a different progenitor population which has a different spectrum of fates and differentiate them and get similar patterns and so to start out we were interested in whether we can get patterns of pure ectoderm within the germ layer and so the idea so I told you yesterday that if I take stem cells the default fate of those stem cells is a neural fate right so if I just take stem cells I abrogate all signaling those cells are all going to become neurons and then within the ectodermal germ layer actually it's the same BMP signal that we're using to differentiate the different germ layers which will push cells to different fates within that germ layer so if I take cells which are committed to ectoderm and I treat them with BMP based on results in other model systems I should get epidermis and I should get neural crest in between the epidermis and the actual neural tissue and so the question was can we push cells towards neural differentiation hopefully they get restricted from becoming these other germ layers and then hit them with the BMP and then see these sulfate patterns emerge but now we should get different sulfate patterns they should be patterns within the ectoderm instead of patterns between all the germ layers so our first attempts at this were almost successful but not quite successful so if you do this protocol the cells actually retain the ability to be diverted to mesoderm for quite a long time which surprised us right so if you do three days of neural differentiation and then hit them with BMP you get still a little bit of mesoderm and this SOX2 marker at this stage just sort of pan ectodermal so we get almost pure ectoderm I'm not showing you the subdivisions within the ectoderm but we get a little bit of mesoderm and it's interesting that actually the extra embryonic fate that you usually get at the borders of these colonies is totally restricted so you'll first lose the ability to form that extra embryonic fate but retain the ability to form mesoderm which is somehow a secondary response in these colonies but most of your colony will be ectoderm and we said okay this mesoderm must come from some residual nodal signaling so we'll do nodal and BMP inhibitors for three days then we'll treat with BMP but we'll keep inhibiting nodal so you can't get mesoderm and that worked to give us pure ectoderm and that actually also works to give us patterns of cell fate within the epidermis so now these look quite similar to the patterns I showed you earlier but the cell fate identities are totally different so this middle expresses PAC6 which is a marker for neurons this ring of red which is where I would have seen the mesoderm in my other colonies expresses SOX9 which is a marker for neural crest and then this outside expresses AP2 alpha which in the absence of this SOX9 neural crest marker it's typically a marker for epidermis it's a marker for non-neural ectoderm and you can make radial average of these things and see what you expect which is that you get this sort of series of fates along the radial axis in almost all of these colonies the BMP whatever structure is in the we haven't actually checked in this system yet in the pluripotent stem cell system there's no gradient in the supplied BMP but there's a gradient in the response to the BMP we've received results from a couple of factors that I talked about yesterday one is we know it's dependent on secreted inhibitors to the BMP pathway which we think accumulate in the center of these colonies and the second is there's this process where the cells acquire apical basal polarity which is particularly pronounced in the centers of these colonies and the receptors get sequestered basally and so that the cells don't even though the BMP is everywhere the cells are not BMP responsive in the center because the receptors are sequestered on the bottom of the cells and they're not accessible to the BMP so the cell in this here I don't know for sure right it's possible that similar mechanisms work it's possible not what I'm nearly positive of is that the center cells don't respond very well to the BMP because if they did they wouldn't have committed to the neural they wouldn't adopt the neural fate I guess technically there is a second possibility here which is that we're pushing them towards neural differentiation for three days before we do this and so it's possible that there's a sort of gradient of commitment right so maybe the cells in the middle see the BMP but they've already committed to the neural fate and they can't adopt a different shape that's possible and we have to do the assays for looking at the BMP signaling to see whether they actually respond or not if that makes sense yeah I mean it is BMP is the initial signal which kicks off all these processes I actually think both of these both of these systems are very similar as I'll show you in a second in the sense that I have the primary BMP signal I have a differential response to that BMP signal and then I induce a secondary signal and so the sort of intermediate fate in both of these colonies is dependent on a secondary signal not on BMP this plot you're saying yeah so this is if I well it should have made better axes but yes this is the center of the colony this is the edge of the colony and then for each data point here I've basically taken a bin at that radius taken the intensity of every cell and averaged all the cells in that bin so it's an average of all the cells that distance away from the center of the colony okay so we can look for secondary signals that influence these patterns and of course since people know quite a bit about extradermal differentiation we had some idea about what these secondary signals are so if we inhibit the wind pathway which is thought to be important for this neural crest specification you don't get any neural crest anymore and this looks sort of like what happens when you inhibit nodal in our other patterns and so you get some of these outer fates you get some of these inner fates you get absolutely none of this red so we think that there's this in the same way that you have this initial response to BMP and then upregulation of nodal in the pluripotent colonies in these colonies you have the it's initial response to BMP then upregulation of wind in particular areas that creates these central territories and then interestingly if you inhibit notch you get all the fates but you actually don't make proper borders between the fates particularly at these outside colonies so these neural crest and epidermal fates they get all mixed up on the outside of the colony and well this is sort of loosely consistent with a roll for notch and border specifying so that we don't totally know how it works yet okay before I move on I just wanted to sort of mention for completeness that other people are sort of modeling development in vitro in a lot of different systems and there's sort of a lot of remarkable sort of organoid systems that recapitulate the morphology of developmental systems in 3D so this is sort of one of my favorite examples where people have grown essentially entire eye cups starting from pluripotent stem cells so this is an in vitro grown cup of the eye and I'm not gonna get into the details of what these different things are but you have the proper markers for different fates you have the right basement membranes made between cell fates and they have this sort of remarkable organization and this is totally self organized another system where this has become pretty advanced is growing sort of crypt like structures that you would see in the gut and these are structured where you have stem cells at the bottom and then the cells proliferate up and so each one of these things is an individual crypt and they grew this giant sort of gut like structure and the red march proliferating style so you can see that the stem cells are typically only located in these crypts and then this other region is non-proliferative and I think a sort of challenge for the field going forward is that people can really do amazing things with these organoid cultures but they're much, much less reproducible than the kinds of things we're doing where you throw your stuff in culture sometimes you get something really beautiful like this sometimes you get things that look like three eye cups fused together and so I think it's sort of a bioengineering challenge and both for us for pluripotent systems to think about them going into 3D and for these organoid models to try to take these systems and move them into places where we can really do these kinds of beautiful patterns but make them reproducible and quantifiable in 3D so we can understand what's happening. Okay, any questions on that before I move on? Okay, so this is what I actually wanted to talk about today so today I'm gonna talk more about the signaling dynamics and the dynamics of these pathways that give rise to these structures so I'd like to start out by giving you a theoretical example of why you should care about signaling dynamics and then I'm gonna give you a sort of experimental example which is the NF Kappa B pathway which is probably one of the only pathways where the signaling dynamics have been sort of well studied for quite some time and then I'd like to tell you about some of our own work relating signaling dynamics to cell fate in a highly simplified system and then some of our work looking in more complex cultures at the signaling dynamics of morphogen pathways in five time. Okay, so to motivate this theoretical example I wanted to show this movie again of gastrulation and to point out a couple of features of this so you see this dark spot here where the gastrulation is going to begin. This is the dorsal lip of the blastopore and this is also a signaling center so secreted signals from this dorsal lip are very important for generating the fates in the cells as they move by this dorsal lip and the other thing I want you to notice as you watch this movie is that you'll form this ring all around where cells are invaginating but cells invaginate first and much more aggressively at the dorsal lip than at the ventral side of this blastopore. You see the cells invaginating here, they're invaginating here too but not nearly as strongly and then eventually gastrulation is over and you go through this neuralation process. So you have this system which I think is somewhat common to other gastrulating systems where you have a source of morphogen, you have cells moving by the source of morphogen but depending on their spatial position the cells move by the source of morphogen more or less quickly, right? And so I think it's important to think about the dynamics of these cases because cells are seeing the source, they're seeing the source depending on how they move and how they interpret that signal as they go by will influence the cell fate pattern. So I wanted to sort of think a little bit about sort of theoretical example based on this kind of idea and the idea is that if I have a sheet of cells and it's gonna move through a morphogen gradient, right, so the morphogen gradient is high where it's purple and low where it's yellow so these cells are all gonna move here and let's imagine that there's some kind of velocity profile to these cells moving through these morphogen gradients, right? So the ones with a big arrow are moving quickly and the ones with the small arrow are moving slowly so this is not obviously a perfect model for anything real developmental but it's sort of a rough approximation to what's happening in cast relation in a number of systems where you have a source, you're moving past that source of morphogen and you're moving past the source of morphogens at different speeds depending on where you are and then what you wanna think about is okay, you have cells and they're moving through this field of signal and then you wanna ask does it, if I move through this field of signal at different speeds, how is that gonna influence how I see the signal? And it turns out that the answer depends rather dramatically on how I interpret the signal as I see it and so I can imagine a couple of different scenarios for how I interpret the signal so a number of pathways and I'll get to a pathway that does this later have been shown to essentially be adaptive which means that they respond to changes in the signal but if the signal were to remain constant they don't respond, they don't interpret constant signals so basically they respond to increases in ligand. The sort of simplest idea is that you would respond to the level of ligand that you see so as you're going along you see changing levels of ligand and that's what you care about, you care about the current level of ligand, not some derivative and then another simple possibility is that you might care about the total amount of ligand that you see over time so as I'm moving through this gradient I see ligand, I continue to see ligand as I move through I just integrate that over time and however much my integral is that's how much I care and so if you just go ahead and... Okay, so it depends... Yeah, but I actually think that the time to move past is so our own work has shown that for example the nodal signaling pathway is adaptive in response to these ligand so it actually does this kind of thing in red here and that the time scale for that adaptation in mammalian systems is on the order of four hours and that the time that it takes cells... Well, I'm not sure exactly how long it takes an individual cell to move through the primitive streak I wish Cat was still here but I think it's also on the order of probably slightly longer than that, right? The whole gastro-relation process takes like a day and a half. So the time scale of the process can't ignore this? I think we can ignore this and so I'll show you also a little later we've looked a little bit in Xenopus embryos during my postdoc and the time scales of the adaptation are much faster and that corresponds with the speed of the gastro-right in Xenopus gastro-relation takes a few hours but the adaptation also takes like less than an hour and so I think we actually do have to care about this in this example. I think there are almost certainly... Yeah, I think there's basically lots of cases where the signaling dynamics matter for what you see. Okay, so back to this, I have these different scenarios and then I remember I have cells that are moving quickly and cells that are moving slowly, right? So it's sort of intuitive but well, so first of all, the sort of total profile of ligand that I see, if I just see the ligand, right? It's the same, it's just compressed in the fast moving cells, right? That's these blue curves. If I care about increases in the ligand, right? I'm gonna see a faster increase in the ligand if I'm moving faster because I'm moving through this morphogen field and so the red curve is higher here than here but if I care about integrated ligand, right? Then as I move through this field, if I move slower, I'm just gonna see more integrated ligand because I spend more time moving through this morphogen field and so this green curve is lower here than here and so actually if I think about this, right? So well, then I have to take these curves and come up with some rule for the cell fate and I've done that in the simplest possible way which is just to take the maximum of these curves over time and so if I think about that, right? If I only care about the ligand, well, every cell is passing through the same morphogen rate at some, same morphogen field at some rate, right? So if I'm just taking the maximum of that, I actually don't care how fast I move through it and I don't make any spatial pattern whatsoever. On the other hand, if I'm integrating the ligand, right, the slowest cells are gonna be seeing the most ligand so I'll get some profile of integrated ligand that looks like this and maybe I'll make some, you know, French flag cell fate pattern that looks like this. In the sort of reverse scenario where I don't care about the integration of the ligand but I care about how fast the ligand is changing, right? The cells that see the most ligand or the cells that are moving fastest because it changes the fastest, these cells see the least ligand and so if I think about my cell fate pattern it will be exactly reversed, right? So, I mean, I think this system is sort of a toy model, it's probably not very realistic but it has a few surprising features which are first of all that you, all these gradients are made from the gradients of the cell motion and so they're all sort of orthogonal to the morphogen gradient and secondly, you get opposite trends in which way the pattern forms depending on whether you're integrating your signal or depending on whether you're using this adaptive response and essentially differentiating your signal. And so hopefully this sort of convinces you that we really should care about what the dynamics of ligand are and how we think about these things. Okay, so now I want to tell you a little bit about how we measure signaling dynamics through an example which has been well studied in a number of labs which is the signaling pathway NFKB, this is not really thought to be a particularly important pathway at least in mammalian systems for developmental processes but it is important for a lot of things like signaling during immunity and inflammation and things like that. And I don't want to get too into the wiring diagram of the pathway but essentially what happens is that you have some input, it activates this IKK and then the important signal transducer is this NFKB which is usually sequestered by this IKB molecule. The activated IKK destroys this IKB and frees this NFKB to move to the nucleus. So if you don't want to worry too much about the details, basically the input leads to the destruction of this inhibitor NFKB moves to the nucleus that importantly these inhibitors are also transcriptional targets of the pathway. So you allow this to move in but then you make these inhibitors which will relocalize it back to the cytoplasm and eventually the NFKB goes back to the cytoplasm. And I think it's important to note that we can, wiring diagrams like this, we know enough information to basically make wiring diagrams at this level or much more complex wiring diagrams for basically any signaling pathway you care to think about. DDF-beta, BMP, notch, FGF, whatever you want. We know the receptors, we know the cascade, we know the target genes but you can look at this and still not know anything about the dynamics. I can look at this and then I'll ask you, okay, now I'm gonna stimulate this pathway, what should the dynamics look like? You really couldn't tell me. It really depends a lot on the parameters of how you parameterize this pathway and also might vary from situation to situation. And so it's sort of instructive to think about how people have measured this and there's been a real split in the field of people measuring this dynamics and how they like to do it. And so there's sort of the classical biochemists who will do things like this. So sorry, this is not well labeled. These are different times basically. So this is moving forward in time. So they'll take cells, they'll fractionate the nuclei, they'll ask how much NFKB protein is in the nucleus and they typically see a profile like this which is that it gets activated by the ligand. It decays a little bit. This is not a real quantitative assay so it's hard to tell how much it decays then it comes back on and then maybe it's going down kind of slowly. Then people using approaches that I think people here are more sympathetic to have done things like this which are to tag the protein with RFP or GFP and watch it move in and out of the nucleus. And if you're watching, we'll play that one more time. If you're watching this movie kind of carefully you'll see these kind of dramatic oscillations where you see these nuclei fill in and then empty and then there that fills in and empties and fills in and empties. So you see these kind of dramatic oscillations at the single cell level of this NFKB pathway. So just to sort of briefly contrast these two techniques, right, the nice thing about this technique is that I don't need to modify cells here. Obviously I have to make my fluorescent cells. Here I have to worry about overexpression artifacts and this has been the source of a lot of controversy in this field of whether these oscillations are real or not or they're just overexpression artifacts. But of course here I get dynamic single cell information which is what I think a lot of the people in this room are after and here I just get bulk information and only these dynamics from collecting multiple samples and no single cell information. Of course there are techniques that are sort of somewhere in the middle, right? So if I were to do a immunofluorescence on these cells I could get single cell information but not necessarily dynamic information unless I collect multiple samples again. So people have really been sort of fighting back and forth about which of these are better in the NFKB field. And the question is, do they really give the same answer, right? So if you look at this sort of profile, right, this is sort of what you see in these biochemical assays. And if I look at this, this is what you see in the single cell assays and then one might argue that actually the mean of this is not too different from the mean of this, right? What's happening is here, right? You have an initial synchronized oscillation and then you have asynchronous oscillations and the average of those asynchronous oscillations tends to result in something decaying as they become more and more asynchronous and here you're just looking at the single cells as synchronizing. But I think it's still quite controversial whether anyone really believes these single cells actually oscillate or whether actually what you really have is a population of cells that do this. And so, whichever view you take of that, one of the most interesting things that people have studied about these pathways is to look at the target genes of these pathways and to ask, okay, I have this NFKB translocating to the nucleus, this NFKB is a transcription factor and what are the consequences of the NFKB dynamics for turning on target genes? And so here you see these NFKB dynamics that I was talking about and here you see two NFKB targets, it's not really important what they are, but you see that one turns on rather quickly and one takes a long time to go on and so if I give it just a brief pulse of the NFKB, this one that takes a long time to go on basically doesn't whereas I still get this one that turns on quickly. So you have targets that turn on selectively with sustained activation and targets that turn on selectively when I give it a pulse and so the pathway, the target genes in the same cell can sort of tell the difference between different dynamics of NFKB that I've stimulated with. They have ways of removing the negative feedback and then actually you still get these sustained targets turning on even with a 15 minute pulse. But the interpretation of this is quite different so the people who say this oscillates actually have a story about how you require sustained oscillations to turn on particular target genes whereas people who do biochemistry are saying okay well you still have some sustained baseline level signaling. And then just a last slide on this, so my own lab got interested in this because in addition to development we do some work on cancer and this is an important pathway in cancer and so these are some ovarian cancer cells and something that really hasn't been done because it's sort of relatively recent technology is all the studies with single cell dynamics have all been done with these overexpression studies and we're now capable of just taking CRISPR, fusing CRISPR to the endogenous thing and then yeah you still have to worry about whether the fusion impacts function but you don't have to worry about the effects of overexpression. So these are cells where we've used CRISPR to fuse YFP to the endogenous locus of these things and the ovarian cancer cells which is a model that we're interested for other reasons then we can make these movies. Eventually they'll be stimulated. You'll see it's pretty obvious there. That goes to the nucleus and then you'll see it kind of decays over time and you don't see any of these dramatic single cell oscillations. And so if you quantify this at different doses you see more or less what you see in these biochemical assays which is you see this initial response, this kind of wiggle and then this slow decay. This wiggle is sort of less pronounced than I would have expected from the biochemistry but what I think this shows is that you can get sort of more accurate results more in line with what you would expect from the sort of experiments done on unmodified cells if you're making better reporters and using CRISPR to do it. This actually hasn't been done in the same cell lines that were done in those papers so we really need to go back and do that and see if it's a difference of cell lines or a difference otherwise in dynamics. This is heterozygous. So there's one copy that has YFP fuse to it and one copy that's... We've actually, we made a few lines, one of them turned out to be homozygous, the dynamic cell and we, I'm not showing you but we've done quite a bit of controls just doing immunofluorescence at different time points and then asking if we get the same, we can basically ask at static time points do we get the same result from immunofluorescence and the results look much better in the heterozygote than the homozygote, I don't know why. So we're using the heterozygote. Yeah, maybe it affects feedbacks and so here if the heterozygote, if the fusion transcribes a little less efficiently, you're rescued by the other allele or something like that, yeah, so that's possible. Yeah, so we need to do that. So we are working, the easy thing to do is to do QPCR for a bunch of these target genes which is what people have done but what would be really nice to do is to engineer reporters and single cells to look at these target genes which to my knowledge nobody has done yet but I think that's worth doing and actually not terribly difficult now with RISPR but we don't have the data. Okay, so that's nice, right? That shows you something about signaling dynamics but we're really interested in in the long term as relating signaling dynamics to cell fate. I first wanted to show you a movie to just give you some idea of how challenging this is. So my lab has been working a little bit with a neighboring lab at Rice with Dan Wagner's lab and we're interested in generating signaling reporters to some of the same pathways we're interested in the zebrafish because it's a really beautiful system for imaging and it allows us to study some of the same signaling pathways. So this is a TGF-beta signaling pathway reporter in Fish which you're watching during the process of epibole and gastrulation and these dots are nuclei where the signaling pathway is active and so you can watch this sort of process of gastrulation and the activity of the signaling pathway and well what you can see from this is that you can see things, you can measure things but this thing is sort of a tremendous mess, right? We've had actually a great deal of difficulty following individual cells, extracting temporal information from individual cells and then trying to correlate that with the final position of the cells where the fate ends up and so we're still working on this but we really wanted to get to situations where we can actually really dissect what the signaling is and what the fate is and correlate those things and even in colonies of stem cells we're working on that too that has proved to be somewhat difficult so we tried to come up with as simple a system as possible where we could address these kind of issues and so we're thinking a lot of the complication coming to the fact that if I imagine a sheet of cells and I think about some cells sitting in the middle of this sheet of cells there's all kinds of influences, right? If I'm thinking about a stem cell culture as we've talked about I'm gonna dump some media and growth factors in here that's gonna influence what the cells do and so the cells are maybe gonna see that growth factor but they're also gonna see what all their neighbors see and disentangling these two effects is quite difficult and if I really wanna understand the input output relationships of signaling pathways and how they relate to cell fates I really have to disentangle how is the cell responding to what I'm supplying and what I'm telling it to do and how is the cell responding to paracrine signals and so the idea that we had is well we can use the same micro patterning technology we use to grow cells in these colonies of well-defined size and shape to grow very small colonies of cells so then I could have a colony with only one cell now it has no paracrine interactions with its neighbors and then I can compare that to a colony with two cells where I do have some paracrine interaction and compare that to three cells where I have presumably more paracrine interactions and so on and do I see trends in how cells signal and differentiate depending on how many neighbors they have so we've made these arrays so we don't actually make only one cell colony or two cell colonies what we do is we make these arrays of tiny colonies that can't touch because they're separated and then you have a one cell colony here and here are two cell colonies or colonies with more cells and then we can computationally separate the number of cells in the colony and do analysis and ask does it make a difference is this two cell colony different from this six cell colony and so on so this is just a distribution of colony sizes in a typical experiment so we have reasonably well represented from say one to about eight cells and past that the data gets sketchy so we're gonna take these tiny colonies and treat them with the same BMP4 signal that gave us these patterns in these larger colonies and see what we get and here we get something kind of interesting and what we see is that perhaps not surprisingly we get conversion to a single cell fate and that single cell fate is what you would get at the edge of these bigger colonies so all these cells somehow know they're near the edge of a colony they all respond to the signal and they adopt a single cell fate they do this in a dose dependent manner so you see here I have all basically here it's not terribly important what the markers are but green marks stem cells and red marks these extra embryonic fates so when I have no ligand I'm all green if I add a little bit of ligand I start to see some red cells if I add a little bit more ligand I see more red cells if I had enough ligand I basically converted all the cells so I get this kind of switch like behavior just between these two cell fates that's correct yeah so we've looked for correlations between like neighbors and we don't see them so we think that they're far enough apart and the cell numbers are small enough that they just don't secrete anything enough to make it to the neighboring colonies yeah they can't bridge this gap and touch it all because they're divided by these areas of material that they won't stick to inside the colony yes so you're only looking at the nuclei here so everything I'm showing you is in the nuclei but they're touching each other sorry I didn't I'm gonna tell you that in a minute okay so this is just to convince you that basically we only have two fates in this colony so here's the percentage of cells that show the pluripotent fate in blue the percentage that show this extra embryonic in red and the percentage of cells that are one fate or the other in purple and so it accounts for in all cases it accounts for 90 plus percent of the colony so you really have this switch like thing where you're turning off one fate and turning on the other fate and it's interesting this is in contrast right to the experiments where the same ligand the same treatment will make these patterns of multiple cell fates in larger colonies and it's also interesting how sensitive the cells become to this ligand in small colonies so to get these patterns I actually need to add 50 micrograms per ml which is way out here right to get a pure population of cells I can do it with almost a hundred fold less yeah about a hundred fold less ligand here so the cells in these tiny colonies are very very sensitive and they differentiate quite homogeneously to one fate so the first thing we notice when we did this is that if you look at these sort of heterogeneous conditions right you get some fraction of red and some fraction of green cells but they're not randomly distributed at all right in this picture I have four or five red cells but they're all within the same colony right and here I have some red colonies and some green colonies so cells are making a decision at the level of the colony as a whole and not at the level of a single cell so if you scatter plot single cell data right so here's the intensity of the pluripotent marker here's the intensity of the stem cell marker right what you see and it's color coded by the number of cells in the colony right so you can see this most clearly here right the larger colonies all cluster together at higher expression of the stem cell marker and lower expression of the differentiation marker in larger colonies in sorry higher differentiation conditions you see the reverse right the larger colonies are all differentiating well the colonies that fail to express this marker are all single cell colonies right so you get this sort of reinforcement of the fate that the whatever fate is more common is better reinforced in larger colonies than smaller colonies so if you look a little bit more quantitatively at distributions of single cells and now this is in pluripotent conditions right so this is the pluripotency marker so what I see is that if I have say seven cell colonies essentially every cell expresses this pluripotency marker but if I have one cell colonies I have a basically a two peak distribution where I have one peak that corresponds with expression which is where these seven cell colonies are and then I have one peak which corresponds to basically cells that have differentiated right so this is showing us that some fraction of cells in these one cell colonies have differentiated and this never happens in these seven cell colonies under pluripotent conditions and if I look at the differentiation marker I see the same thing in reverse right so these seven cell colonies all keep this differentiation marker off but they have this long shoulder and the one cell colonies of things that have started to turn this differentiation marker on if I look at differentiated conditions I see the same thing exactly in reverse right so differentiated cells will in larger colonies will turn off these markers but in smaller colonies will keep will some fraction of them will keep the marker on and for the differentiation marker they'll in larger colonies they'll all turn it on and in smaller colonies some of them will fail too so you have this two state system the larger colonies are actually very good at interpreting which state they're supposed to be in so in pluripotent conditions they're uniformly pluripotent in differentiation conditions they uniformly differentiate the one cell colonies sometimes get confused basically and they'll sometimes differentiate in pluripotent conditions or fail to differentiate in differentiation conditions yeah so we consider it as having seven cells and the reason is because when we've looked with live cell imaging at the rest you have a little colony it has seven cells in and cell rearrangement within the colony over the scale of the two days that this differentiation takes are quite frequent right so I don't it's not you know if I pick any two cells in the colony it's not correct to say that they're always neighbors or always not neighbors the colony is always shuffling around so I think that there's enough cell mixing in this colony that we consider it to be one colony one sort of colony where every cell is kind of the neighbor of every other cell so okay I'm not sure I understood what I mean each data point that goes into this histogram is a single cell but it just they're just color coded depending on whether they came from colonies that they're the only cell or whether they came from colonies where they're seven no so there's no cherry picking for the colony of the colony of seven cells will contribute seven data points to this histogram right so every cell is accounted for and histogramed it's just that the seven cell colony will contribute seven data points the expression of every single cell in that colony and the one cell colonies contribute only one data point because there's only one cell responding to the gradient but they're also dropping each other and forcing the position another one it really has to do with the biochemistry about the region to find it it's possible that the region just binds and unbinds and essentially you get a slightly higher concentration when you have more cells if not all the binding get like that successful you get an ability to find and unbind and don't generate any signal but kind of trapping and I don't know if this has been reasonable from what we know about the K-1, K-2, your concentration I wonder if you can say anything about I mean is this like just a geometrical effect or is this really a sign that you're talking to each other and how can you explain this to me? I'm not sure we have really good data I guess what I would say is that I think if it were in effect of just ligand trapping if I were to increase the concentration of the ligand itself I should see the effect go away because then in the one cell colonies I should just have enough ligand floating around and that doesn't happen so whether I'm working right over here where I've just gotten to sort of a hundred percent or whether I'm working at an order of magnitude plus greater than that I see the same effect so I agree technically I think it seems possible that if ligand is sticking to cells cells sort of trap ligand and then present it to their neighbors but I guess the only data we have that really argues against that is that it's not the effect we see doesn't depend on the concentration in this range it is yeah okay I don't think that is thought it is internalized into the cells that take it up I don't think it's been described to be passed to neighboring cells although I mean I think that would be a biochemical mechanism but it's non-trip right and if you have more cells that are able to make a collective decision because they're sort of passing a common pool of ligand between them I think that's interesting but yeah I don't have any evidence that is or isn't the case here you put that more than able to I think that's probably the case if you want to construct nuclear to at least focus on this yeah okay okay so historically this was reminiscent to us of some old work from John Girdin who won the Nobel Prize as I talked about yesterday not for this but for reprogramming of frog nuclei to pluripotency but what he described in a sort of very qualitative way was that if you take these animal cap cells which again are sort of like the pluripotent cells of the frog you dissociate them and you put them between the vegetal tissue which is the tissue on the bottom of the frog which would induce them to become mesodermal fates and then you look at how well they're introduced he noticed that if you put them in sparsely they wouldn't induce very well but if you put them in solid clumps they would induce much better and he called this a community effect in development right so cells that are sort of isolated don't read this whatever signals are emanating from these vegetal cells well but cells that are in a solid clump do and so you know we think that the stem cell systems are sort of a system where you actually see this effect also and you can study it quantitatively so here's just a sort of picture which we sort of cherry picked to visualize this event right so all nuclear are marked in green the differentiation marker is marked in red and you can really see right this individual cell doesn't differentiate but right next to it you have these colonies of more cells that are differentiating very well yeah yeah that's something we're working on right so it is possible that right so they've been growing on these dishes for two days at this point so it is possible they're related to each other so we're actually trying to do simple things where we you know independently sort of barcode or just like put an independent fluorescent protein on the cells so we can tell if they came from the same thing or not there's actually when we watch the movies there's significant cell division but there's also significant cell death so my guess would be these one cell colonies it's actually very unlikely that it would just sit there as one cell the whole time it probably divided once and then died what we think from the live cell data that the parameter that probably matters is the number of cells that are in the colony at the time you stimulate it with the ligand that's sort of one or at least close to that time that's when they're reading it out but we don't have good proof for that either yeah so this cell fell so differentiate whereas these cells differentiate very well in larger colonies but just to spend a minute talking about theory this sort of reminded us of the situation you have in magnetic materials which you can model very well with the icing model where you have some external magnetic field which tends to align these spins in one direction and then in the icing model the neighboring spins tend to want to align in the same direction and so these are sort of the same qualitative effects that you see in the sheet of cells is that you have some ligand and the ligand sort of an external thing that pushes all cells to the same fate and then the cells which are neighbors want to synchronize with the same fate and so you can write down a very simple two parameter model essentially where one parameter quantifies the strength of this field and the other parameter quantifies the strength of the interaction between cells and ask how well it fits your data right so here we're looking at instead of just looking at histograms of say one cell versus seven cells we're looking at the fraction of cells in the say here in the undifferentiated population as a function of colony size and fitting that to this icing like model with the curve in black and you can see it fits so these two parameter models actually fit quite well right here is the sort of reverse if I differentiate the cells now my pluripartent marker goes down with the same trend and then if I look at the differentiation marker I see the opposite trend as it goes down and pluripotency and up and so for fits with two parameters I think these fit quite well you can look at a bunch of other data and compare experiment and theory for like these are distributions within the three cell colony how likely are they to all have the same fate to have one cell of one fate and two cells of the other fate and so on and then without adjusting parameters basically these all look the same between the model and not so we think that you can basically sort of almost quantitatively explain everything that we see at least at equilibrium in these colonies just with two factors one is that the ligand pushes you to some particular fate and the other is that you have some some preference of cells in the colony having the same fate and if you do the equilibrium statistical mechanics on that you explain the data quite well so now we wanted to understand how this works the first thing we thought about which we were hoping it didn't work this way is that people thought it was okay well maybe single cells don't divide very well they get stuck in G1 and then when they're stuck in G1 they'll differentiate differently and so we looked at whether they're actually stuck in G1 and we saw no differences in sort of total DNA material based on Dapi staining we also made a Fucci line and looked at the fraction of cells in D1 and saw no differences so there's no cell cycle differences between the smaller and larger colonies is there a question? so we've basically just subsumed the temperature into the other two parameters right so I mean you basically have right it's basically the ratio of your interaction parameter to the temperature and we don't have an independent temperature parameter well it's an equilibrium model so that the output of the icing model is basically the likelihood of any particular configuration it doesn't have dynamics and actually yeah so go ahead it's just not a dynamic model right it's a thermodynamic model which means that it tells you the likelihood of any particular configuration but it just has nothing to say it just tells me if I observe a bunch of different copies of the system at equilibrium which you know equilibrium is basically just a distribution of those different copies it tells me the likelihood of any of those right it doesn't say anything about dynamics right so if I were to perturb the system my icing model would not apply until it re-relaxed equilibrium right it basically can't model dynamics especially within a single colony it can't model dynamics and actually what we I think what we we actually see it where it doesn't so we're back here but if you think about these situations I think what happens so over here in this region you have a kind of phase transition and your dynamics get a lot slower and actually the icing model doesn't fit as well for these couple of data points and actually what I think we're seeing here is that the transition from pluripotency to differentiated in some of these cases just happens much slower and so we have non-equilibrium cases and in those cases you can model them with this kind of model that makes sense look that you are from the right you look that most of the cells are actively proliferating yeah most of the cells are actively proliferating yeah yeah the seamen are actively proliferating in the right hand of the following time but it's dying as soon as they're dying I mean it definitely is a kind of proliferating well no I mean that the one that bother them is big and he looks at most of the cells yeah I think that's that's because I think your colony distribution reflects your seeding so you seed at some no we could do that but then we would you basically have a spectrum of different colony sizes yeah yeah yeah yeah okay so it's not colony size we're interested in the hypothesis that it's related to signaling I'm not going to go through the details of all these signaling pathways but basically there's a number of signaling pathways active in pluripotency as I talked about a little bit yesterday so active in and FGF promote pluripotency and WINT and BMP promote differentiation and so we first we're interested in starting in this in the pluripotent state where we see reinforcement of the pluripotent state can we either recapitulate or remove this effect by inhibiting any of these pathways and so the question is what would we expect to see if we did that so we use the theoretical model to predict what we would expect to see and right one thing is that you would lose this colony size dependence so if you remove the community effect right every colony size should look the same and just the cells are rich any independently and one thing is that you should have you should see the emergence of these mixed colonies right which have some cells of one fate and some cells of the other fate rather than a reinforcement of the same fate in all the colonies and so we tested basically small molecule inhibitors to all the pathways I showed you on the previous slide to see if we could find something that met these predictions we actually did so this is a MEC inhibitor and it works pretty well we were sort of disappointed to find that I actually don't think this has anything to do with the MEC pathway so we tried another MEC inhibitor it didn't work we tried an FGF receptor inhibitor which should be upstream that didn't work also we did a bunch of other stuff and so we actually have a small molecule that can completely remove the community effect and we still don't know how it works yeah so but we're pretty sure this effect is not specific to MEC so that's all for what's happening while the cells are still pluripotent while the cells are differentiating we see this community effect and here we have a better chance to have a handle on what's going on because we know what the signaling pathway is that's mediating these events we're dropping in this BMP signal the cells are signaling they're adopting some fate and so we're interested in can we follow the signaling in these cells and can we use that to understand what's happening and so we use this reporter so here we've again used CRISPR to knock GFP into this MAD4 locus MAD4 the signal transducer for this pathway and should go to the nucleus when this pathway is activated so here you see some cells with this reporter here's some nuclear markers so we can track these cells these holes again are nuclei and so you'll see when the ligand is added that these holes fill in so somewhere there the ligand is added and the holes fill in and we can track the cells and we're signaling as a function of time it was actually surprising to me how I thought this was going to be super easy I was like you know there's no more than five cells in a colony we can totally track these for as long as we want it turned out to be astonishingly difficult to track a colony like this every ten minutes for two days if you insist on losing no cells and we had to do parts of it manually anyway what we found was that if you watch one cell colony signaling the signaling is quite variable so you'll see cells that signal and then go down you'll see signals that signal in a more sustained way follow cells through division and sometimes the daughters diverge and sometimes they don't but if you look at colonies with more cells the signaling between the cells is basically always the same so it's correlated between each other but it's just always the same they respond and they stay on in a sustained way so you can trace this sort of inhomogeneity in these one cell colonies back to sort of these heterogeneous responses to signaling where sometimes it's sustained and sometimes it's not this is the same way we differentiate them so right here you're dropping in this signal the signal is floating around but the signaling especially in one cell colony is very noisy as you get to higher colonies okay so if I have time I'll get to so both active and nodal and BMP use this signaling molecule what we found is the active and nodal signaling is adaptive and the BMP is not so if you have bulk cultures with BMP the signal is quite constant TGF beta is active and nodal branch so the division is at the level of the SMAD signaling molecule so BMP signals through SMAD 158 together with SMAD 4 and we find those have constant signaling the active and nodal TGF beta ligands signals through SMAD 2-3 and those show the adaptive signaling okay so right here is just an example right here is before here is after stimulation of single cell traces and here are some averages so we find that the averages diverge between the one and two cell colonies and this mostly results from the one cell colonies that lose signaling whereas the two cell colonies have persistent signaling if you make distributions of the signaling in single cells at different time points during the signaling and we are sort of interested right here they should be similar and here they should diverge right on that sort of what you see the pre-stimulus levels are similar then you turn it on and these distributions are similar but in the one cell colonies you have some significant fraction of cells which revert back down to low signaling and these basically don't exist in the two cell colonies the two cell colonies are almost all at this sort of ratio of one or above and so the hypothesis is that the cells that fail to differentiate we know that one cell colonies will sometimes fail to differentiate properly now we know from watching the signaling that some of these cells have these low signaling colonies and we are asking whether the low signaling colonies are the ones that actually fail to differentiate and so to test this more directly we went ahead and we looked at these low signaling versus high signaling colonies just in the one cell colony case we looked at the differentiation marker CDX2 and asked whether there's a difference and in fact there's a fairly large difference between these low signaling and these high signaling colonies so the low signaling colonies technically don't express the CDX2 so they fail to differentiate although some fraction of them do whereas the high signaling cells differentiate to CDX2 or not and if I were to just been on CDX2 levels in some unbiased way and then ask what's the signaling and the cells that failed to differentiate and the cells that did differentiate I also see a difference here although we still don't know all the details of how it works what we think is happening is we stimulate the BMP4 pathway in these cells somehow interactions between these cells reinforce sort of high signaling state that stays on in these larger colonies and that causes more uniform differentiation of these cells whereas when you have a single cell they get initially stimulated and that leads to variable outcomes in differentiation afterwards okay questions okay so in the last few minutes I have I just wanted to tell you a little bit more about our work on signaling dynamics of morphogen pathways and particularly about the signaling dynamics of this nodal pathway which is important for forming mesoderm as I've shown in the other assays things so just to remind you of the structure of this TGF beta-super family the details are not very important but you have two branches right so the BMP branch is what I was just talking to you about at the time I'll show you more data that basically this shows a sustained response but we've also used these reporters to study the response of the active and nodal branch right so these have different upstream signal transducers with here the prototype being smet 2 and here the prototype being smet 1 they both bind to smet 4 and make complexes and go to the nucleus and activate target genes and just to remind you both of these pathways are extremely important in this gastrulation process that happens in embryos right so there's a basically the trigger for this gastrulation process is a positive feedback loop that involves this BMP signaling through via wind signaling turning on nodal which turns back on BMP and so we ultimately we really want to understand the dynamics of how these pathways are interpreted and what's happening in these cells as they differentiate you probably don't need this slide but just to remind you right the way all these assays work is that you have some fluorescent protein all these things translocate to the nucleus upon stimulation and so it unstimulated looks like this and stimulated looks like this and I should also say I'm going to sort of the interest of clarity be a little bit unclear about the cell types of the following experiments so the sort of history is that during my postdoc I did a bunch of these experiments in a myoblast cell line called C2C12 in my own lab now postdoc it's a hemskirk has been repeating and expanding a lot of these in human embryonic stem cells and so I'm going to and we've actually found almost no differences between those two cell lines so I'm going to show you a mixture of the data between those two cell lines without being terribly clear about which line I'm talking about but if you're unclear or you want to know more about it just ask me I just don't want to confuse you okay so when we started thinking about this pathway everyone thought I've sought the important thing was this MAD2 which interacts directly with the receptor gets phosphorylated goes to the nucleus drags this MAD4 along with it and so we asked okay what are the dynamics of it when I add the ligand so I add the ligand the ligand is in red the signal is in white and basically this thing turns on there's a slight blip but it stays way above baseline for as long as you care to look so you get some sustained signaling through this pathway interestingly if you just go ahead and look at the dynamics of target genes so this is just done by QPCR in these cells and you ask do they do the same thing they don't do the same thing at all right so we know the ligand turns on this MAD2 this MAD2 turns on and stays on but when I look at target genes they turn on rather quickly over the course of a couple hours and then decay just as quickly so this pathway is adaptive over the time scale at least in transcription of four to six hours and so we were interested in whether this adaptation is mediated by a different component in the signaling pathway and so we looked at this MAD4 which at the time was sort of thought to be a passive cofactor for this MAD2 necessary for transcription but indeed if you do the same assay so you tag MAD4 with GFP and you make these movies there goes the ligand MAD4 responds but almost as soon as it gets to the nucleus it turns around and goes back to the cytoplasm and the dynamics of this look exactly like the dynamics of transcription right so we concluded from these series of experiments that the dynamics of the TDF beta pathway are adaptive that they're primarily reflected in the cofactor MAD4 and not in the dynamics of this sort of primary response MAD2 we've actually spent a lot of time over the last couple of years trying to figure out how this adaptive dynamic works so far without terribly much success one thing we do know is that it's a feedback inhibition that removes this thing from the nucleus so if you treat cells with cyclohexamide at the at the same time that you stimulate them cyclohexamide is a protein synthesis inhibitor so you basically don't allow new proteins to be made you don't allow feedback to happen and you make a movie and just to warn you these cells are going to die horribly after we see the result we want because of the cyclohexamide so you see this very strong response it never adapts the cells will stay on for as long as they're alive and then they well then they die basically not very happy but this told us and we've seen the same results in C2C12 cells which live for a little longer right as you compare the result with cyclohexamide in blue so you can see your normal curve and maybe okay well the pink is quite close to the normal curve right you can see that you get a much stronger response and that it essentially never decays until the cells die no there it becomes if you stimulate it to saturation it becomes totally refractory to later verses of the ligand if I have time I'll show you a little bit of that yeah and I actually think that is important because there are cases where cells see the ligand and then see the ligand again and depending on how saturated they are from the first time it can totally prevent them from seeing the ligand the second time so other features of this response the amplitude of the response is dose dependent the time scale is basically not at all so if you stimulate with different concentrations of this you increase the amplitude you don't change the time scale at all but getting this 4-hour window is dependent on having continuous signaling during that 4-hour window so if I inhibit the signaling so SB is a small molecule inhibitor of the receptor if I inhibit the signaling sometime during this 4-hour window or if I don't inhibit the signaling I do this if I inhibit the signaling after 2 hours I drop with this red curve I inhibit it after 1 hour I drop with this blue curve so this pathway is kind of constantly sensing outside the cells and that's necessary to keep this in the nucleus but even independently of that once you get past this 4-hour time point even though the ligand is still there and presumably still sensing it it will adapt I'm wondering do you have any mutant that is able to find the amniotic that is just in the process of getting the nucleus could look different could the eucalyptus have a way to see the state of the car get past 4 or past 3 or 4 yeah that's a good experiment we still see the signaling is acting on the amniotic signal out because they are able to find the amniotic yeah I agree that's a good idea we haven't done that but that would be cleaner than the cyclohexamide because then you'd prove that it's a SMAD4 target and also presumably the cells would stay healthy during the time that's a good thought we've looked a little bit about whether this happens inside actual embryos this adaptive signaling during my postdoc we use enopus embryos so again we use this animal capped tissue for your reporters like the SMADs cut the animal capped off the embryo put it in a dish and do imaging here there's some SMAD1 signaling but the more important thing is the SMAD4 signaling and you see these flashes of SMAD4 in individual cells so we we haven't really proven this but we interpret this as sort of these adaptive signaling that happens in individual cells whereas upstream SMADs are more constant okay so in the last few minutes so I'd like to tell you about how based on these enopus results where we see these sort of more interesting dynamics we got interested in we know this pathway is adaptive can we probe what happens to the cells when we stimulate with more interesting dynamics if we have finer control over the dynamics to raise and lower them slowly or to do pulses or things like that can we sort of understand the input-output relationship in the pathway and so this was a collaboration with Ben Wasaur when I was in Ali's lab he built this cell culture chip which is based on a design from Steve Quake's lab which you know in my own lab now we use much simpler microfluidics because this is really overkill for what we need to do but it's fun anyway so here there's 96 independent culture chambers you can fill any one of these culture chambers with anything you like and so right here you're filling them all blue or half red and half blue or here you're filling an individual chamber at a time and so this allows us to grow cells and then to in a multiplex way stimulate them with whatever dynamics of log N we want this just gives you some idea of the scale of these microfluidic chambers right so here's some cells growing inside this microfluidic chamber and though you do things correctly they'll grow happily for a long time and fill up this chamber and so the idea behind this was to take instead of always doing our experiments this way where we just drop the TGF beta in to stimulate them with different types of input which will probe different aspects of the signaling response to measure the signaling response in real time to think about how these fit to some modeling and then to think about the consequences for patterning I'm not sure I'll get to hold this I'm going to skip this actually let me do this so the other thing that's enabled us to do was to have a transcriptional reporter so there's actually a live cell transcriptional reporter based on luciferase imaging and what you see is if you stimulate these cells right they it turns on and turns off in the transcription just like the signaling does the half life here is significantly longer and that reflects the half life of luciferase protein itself right so the cells produce luciferase while the signaling pathway is active and then the decay that you see here is the decay of luciferase protein thereafter and so we can measure both the signaling and the transcription in these cells so we can do things like pulse the stimulation and we were interested in sort of mimicking these pulses of stimulation that we saw in the Xenopus embryo right and so we saw you can repeatedly pulse the pathway on and off and get a response to each pulse which is almost the same length and then interestingly if you look at the transcriptional reporter what you see is because of this longer half life which I think is comparable to the half life of a lot of targets what you see is that if you have repeated pulsing that the transcription essentially stays on right the signaling is going on and off at each step if I were to just have a long step of signaling the transcription would go on and off but if I have pulses of signaling and some reasonable half life for the luciferase molecule what I see is the luciferase molecule just build up over time and so we think this might be what's happening in vivo in the Xenopus embryo where everything happens very fast if you need to keep target on sustained by this pathway the way to keep the target on is to just constantly pulse these ligands and so we see these individual pulses and we hypothesize that these are correlated with sort of sustained expression of differentiation genes the other thing that we were interested in doing is thinking about the response to a spreading morphogen in time and so if you think about these are sort of cartoons from this paper where they thought about how the active and nodal signaling spreads through the Xenopus embryo in time so it's not terribly important what the profile is but if you think about any cell sitting here it's basically seeing increased ligand as a function of time as this morphogen signal spreads through it and so we ask okay we can mimic a cell sitting here and seeing this increased ligand in time by using microfluidics to either rapidly or slowly increase the ligand in time to make a long story short what you see is that if you increase the ligand rather rapidly where you see a response which is not really distinguishable from just dropping the ligand in which is that you get this burst and you get this adaptive response if you increase the ligand slowly enough essentially this adaptation process will keep up with this process of negative feedback which mediates this adaptation is able to sort of keep up with your stimulation so you can keep ramping up and up and the cells never notice that they're being exposed to this ligand and so what you see here is that you end up going to what's essentially a saturating dose of TDF beta and the cells have not responded at all to what's in there running out of time just finished with this so we were interested in sort of probing what are the consequences of this adaptive signaling for sulfate patterning and we've so far only done this theoretically and so the idea is that we made a model where you have some morphogen signaling the morphogen spreads by diffusion and decays and then the response to the signal depends on this adaptive model so you adapt and so you really care about the derivative of the signal rather than the signal itself and the color coded some positions along this axis here to enable us to think about what's happening at each of these positions and so if you think about the ligand as a function of time at each of these positions so at the black curve you get something like this at the red curve you get something like this and so on and so if I wanted to pattern this tissue I would I could pattern it was sort of a traditional model where I think about how much ligand I have the cell reads the ligand level and then if I wait till I get to the end of this time course this is different from this this is different from this and you could imagine cells correlating sulfate with those ligand levels but on the other hand if I wanted to make a decision much more quickly if I think about the derivatives the slopes of these curves are different right from the outset when you're tracing its information on these adaptive pathways it's going to be able to make decisions about patterning much earlier these cells are also by playing we've been talking about robustness a little bit by playing with parameters of this model this is much more robust to certain parameters one that's easy to understand is the decay rate of the ligand so when you form gradients like this through diffusion and decay if your ligand say in the extreme doesn't decay at all at some point your gradient fills the entire space so in that case there's actually no information in the steady-state gradient the ligand has filled the space and there's no information in the steady-state gradient but even so how you get to that steady-state will be different depending on whether you're here or here so this sort of derivative model will make a perfectly good pattern in that case whereas your steady-state model will totally fail to the pattern okay Swith can you sense here that that is true but you have to assume that you're caring about obviously as you go to steady-state your derivative goes to zero so you have to assume that by the time you read steady-state your patterning is over we don't have evidence that that's what cells do yeah I just want to understand better what is the second cost plate we can reach but I think actually more realistically so let's see if I can go back quickly if you actually look at these responses to different doses they're adaptive but they also have some constant baseline right and so what I think is probably more realistic is that you have some targets that care about this adaptive phase and are sensing the derivative and you have some targets that actually care about these smaller differences that happen in the baseline and those would be sensitive to these more steady-state things and so probably in real life you have some things taking the derivative and making decisions early making the decisions here so I don't think they're totally mutually exclusive yes so I think these are actually cartoons I think they're for nodal but it doesn't really matter so this paper what they actually did was just stain for activated SMADs, beta-catenin and phospho arc as a function of time and just make maps of what the signaling is so it's maybe not a great assumption but if I assume that there's a cell sitting here this is what its signaling level is in phospho SMAD 2 as a function of time and what our experiments show is that what's happening in SMAD 4 and therefore the transcription of at least some targets is some function of how this phospho SMAD 2 is changing so if this phospho SMAD 2 changes quickly I see this adaptive response or not so in some sense I think we can use this data on these upstream readouts to say I'm not saying anything about where the sources of ligand are and BMP actually, a lot of the important dynamics come from the inhibitors secreted from the spaymon organizer would sort of carve out a territory but all we're saying is okay, I look at the all of our data suggests that looking at the proximal responses is a good proxy for looking at the engagement of the receptors so I'm looking at how the receptors are engaged in time and then thinking about how intercellular that would be interpreted I think I'm basically out of time so I'm going to skip some of this and just acknowledge some of the pitot, they're not on there but I should acknowledge Eric Cigia and Ali Brevenlu, I started this work in their labs all the stuff I told you about the micro colonies was the work of Anastasia Nemeshkalo the stuff on ectodermal patterning was the work of George Britton the stuff on shapes was Sapna's work stuff on beads and also on some of the active in dynamics with its work and funding agencies and I'm happy to take any more questions