 Go start? Okay. So What my plan was for today and tomorrow was to talk about two different aspects of development that I think are particularly interesting with respect to morphogenesis and how a fertilized egg assumes the complex body form that ultimately results from any embryo. Today I wanted to talk about how the positional information that's supplied by the mother in the egg is translated into specific developmental programs and the specificity of that process. And then tomorrow I'm going to follow and that means that today's lecture will be a little bit a follow-up on the one that art gave this morning, but I'll give you a more personal, experimental and ubiquitin-centric view of the problem. And then tomorrow's lecture will effectively follow up on the lectures that Boris has been giving on mechanics and again try to supply a embryo-level view of the process. So what I'm going to talk about today and we've already alluded to this and you've seen this already is this transformation of in morphology in the embryo that is underlayed by a pattern of gene expression that builds on this maternal gradient called bicoid. So there's a bicoid is a transcription factor as we'll see its RNA is localized the anterior into the egg. It establishes a concentration gradient of the protein along the anterior posterior axis of the embryo and that concentration gradient activates gene expression patterns in defined regions of the embryo. The first genes to come on are a set of gap genes, two examples are hunchback and cripple, and then we'll discuss this in a little bit, but somehow these expression patterns then are relayed to establish a more complex pattern of gene expression called peri-rural gene peri-rural expression patterns that are complex and detailed enough that individual rows of cells are now distinct from the cell that's immediately in front of them or behind them. So this little unstained row of cells here are cells that don't express the two either of the two genes, peri-rural genes, paired and run. And there are seven of these peri-rural genes that are expressed in overlapping domains and so that gives you an extraordinary specificity of self that fates that are defined at this period, and it's that specificity of that detail of expression that allows specific morphological events to occur at this stage. Now I'm particularly going to focus on this early step which is the transmission of information from the bicoid protein gradient to the downstream, to zygotically activate the transcription of these genes here. This process occurs in the early embryo over the first two to three hours, so it's a relatively rapid process. Remember the fly embryo starts its development with fertilization. The nuclei replicate without intervening cytokinesis, so you have an initial stage, which is a syncytial stage, until there are 13 rounds of division. After 13 rounds of division, there are about 6,000 nuclei on the surface, about 100 nuclei along the anterior posterior axis of the egg, and at that point these nuclei cease dividing. They're subdivided into individual cells by membrane synthesis, and it's during this period here that this individual, that these transcription patterns become, reach their final magnitude and appear to establish defined states. Now the way people think about that happening, and this is the kind of a textbook view, is that if you have a greater distribution of this protein, the individual genes that are activated by bicoid expression are activated, have enhancers that will have affinity for bicoid at different concentrations. So there'll be some enhancers which are relatively low affinity, some enhancers that are relatively high affinity, each enhancer, you know what I mean when I say enhancer, right? I suddenly had this horrible feeling like, okay, so that enhancers are built in a way to have different concentration sensitivities, and those, those enhancer sensitivities defines thresholds, bicoid concentration thresholds that define boundaries of gene expression. That's the simple basic model that we all work with. This is a, the bicoid is the major determinant for anterior, and for actually I believe substantially most of the patterning in the fly embryo. It is though a special case. There's certain features here that are, that you don't find duplicated in other organisms. The idea of well, there are localized messenger RNAs in many unfertilized eggs. What's peculiar, largely peculiar, is that this embryo remains a sensation, so the movement of the molecules from the source is moving through the cell rather than around cells. That's going to change how we think about the process a little bit, but I couldn't begin this without having a little session to tell you why I think bicoid is so cool. Okay, so that's why was the discovery by Christiana Nussain-Volhardt in her lab fifteen, twenty years ago, so exciting. As the picture emerged, I think one of the things that it suggested to assist development of biology was that information in biology was potentially simple, small number of molecules, and that information was also quantitative. You could measure it, and that the response of cells therefore, so when the information is quantitative and then the response of cells and therefore cell fate depends on the ability of cells to read concentration levels. And to make, and points us in the direction that the important feature is the ability of cells to read, as we will see, subtle very small differences in concentration along a concentration gradient. Another feature is that the way this gradient actually arises, you can see and measure the initial source, and the information, though, is not in the source. The information is not really present in the egg, the way the mother makes it. The information-rich content evolves during early embryonic development to produce a final information-rich distribution. And there's a feeling, and I think this is still true, although it's what we're really struggling with here, is that you can understand that how that final information-rich distribution arises from relatively simple physical parameters. We ought to be able to understand this. If we can't, in fact, I would argue that if we can't understand bichoi, we're not going to be able to understand anything. The limits, this is the ideal test system for us to test, are thinking about diffusion degradation, many of the ideas that that aren't raised this morning. The third thing that's interesting about bichoi is that actually, although bichoi is essential and important in fly development, it is a fairly newly evolved gene. Most of the dipterin, other flies, many of the other highly evolved fly species, have bichoid. But not all, but bichoid evolved during the evolution of the, what are called the advanced or eucalyptid diptera. So there's a little branch of all of the dipterin flies that have bichoid protein, and it is not used in patterning any place else. But what's interesting then is that this is a patterning mechanism that evolved about 200, I don't know, 200 million years ago or something. And as the, this family of insects has evolved, the bichoid protein and its use and the properties, the governance distribution have also evolved and changed over time. So we can, well, it's really nice since it's such a small, discreet little area of the animal kingdom, you can follow how an evolutionary solution is then modified or adjusted to allow development, as we'll see in this talk, in bigger and smaller eggs. Because this single thing that has evolved, then how does one, how does one, how do those processes organize, change over time? So what I thought I would do, and actually this is, is to just go over a couple of the experiments that we've done over the past ten years now, that were really designed to put this model onto a more physical or dynamic basis. Basically, the idea of these experiments was to try to establish physical measurements for bichoid concentration, half-life diffusion that would allow us to actually test our cartoon models. And what I'm going to do is just going to go through some of those measurements, just to give you an idea of the kinds of things that we know or think we know about the molecule. Also, the kinds of measurements and our best estimations for any of these values that we have right now. And what I, I would hope to do is depending on how we manage time, is at the very, at the end of this, I'm going to, I'm going to aim to end early. And then by end early, I would like to have a discussion that is totally unprepared, but totally triggered by Art Landers lecture this morning of Can we make a list on the blackboard comparing a morphogen gradient, say the DPP gradient in that operates in an epithelium in the wing and bicoid gradient, and compare different features, different parameters, different anything that we can think of to try to begin sorting out what elements are going to be important in one system, what are going to be important in the other. So I'm going to go through, therefore, a kind of a limited subset of things that we can measure in bicoid. But I tried to pick the recent ones, and that's the first half of the lecture. And then what I'd like to do in the second half is really focus in on information transfer to ask how much information is in the bicoid gradient, how much position, can we define that in some quantitative kind of way that tells us allows us to quantitate the information content in the gradient and then ask how that information content is transferred to the first tier of genes in terms of transcriptional responses, that is activation of gap genes. And then what I'd like to do for the last 15 minutes of that experimental part is to talk about some recent experiments that we've done to really get in and look at the enhancers that respond to the bicoid gradient and try to get a sense of how it is that these individual enhancers are able to measure concentrations accurately. What are the features? If you're in a biological system, we're going to build genes that would respond to different concentrations. How do you do that? We don't have any idea. Really, in terms of how what features of an enhancer would be essential for it to measure concentration, and that I think is one of the really powerful places where we can use bicoid. So I'm going to talk about and end up with a model for what are the features and then we can ask whether those features are sufficient to actually account for the accuracy and concentration measurements that we see. So that's the plan Are there any questions? Good. So what I had to say that I've focused mostly on is time scales. Remember this gradient arises in about the whole process is over in about two to two and a half hours in embryonic development. So that's our big time. That's the time, the big time scale we're working. But I'd like to kind of break it into little units and say what do we know about any of the individual components? How soon can we see protein? Things like that. And so I'm just going to show you some pictures and give you some numbers and give you some idea of what's the normal dynamics of this process. Okay, if you look at embryos right, look at eggs, before they're fertilized, they're actually hard to get out intact and attractive. But the bicoid RNA is localized as little clusters tightly bound to the plasma membrane, the anterior plasma membrane of the egg. And females can make these eggs and store them and keep them for two to three weeks with this RNA bound at these sites and not being translated. It's just a bound RNA. When the female lays the egg the egg is activated, actually goes through the final steps of meiosis, is fertilized by a sperm, and those events lead to a dispersal of the RNA and a release from the translation block. And it's a fertilization that the RNA begins to be translated into protein. We know that, well, I won't go into the numbers, I have to tell you there are about a hundred thousand RNA molecules from various estimates in the egg, and they probably make 50 times as many protein molecules. I didn't check those numbers, I'm kind of making them up, but that's the order of magnitude that we're going to deal with here. So now these are images of fires. Once I was stained with DAPI so we can follow the cleavage divisions. And what I wanted to show you was that by the time there are about 32 nuclei in the egg, you can see it's still a syncytium and the nuclei are still kind of scattered around here. We can begin to detect bicoid protein in nuclei at about to about 40% egg length. And we can follow that. So within about 40 minutes of the release of this RNA and the beginning of translation, we begin to see RNA all the way out to this point. This is obviously a detection problem. Yes. That's the protein only comes from the messenger RNA that is anchored at the anterior end of the embryo of the... No, no. Absolutely not. It is a single source. We can show that genetically. You could imagine how it might be able to show that genetically. The synthesis of this RNA depends totally on the genotype of the mother. Removal of this RNA, any embryo does not transcribe. Overall, I should also say that transcription is very, very low until the very final cleavage divisions. So the embryo is not capable of making any protein other than from maternal RNAs that are supplied. Are there other questions? Yeah. Before fertilization, do you think this... I would say this is less diffused than this. And so there is a transition. Other questions? It's just a blow-up so that you can see in these two nuclei that we're beginning to accumulate bicoid. Now we can follow. Those were actually antibody staining and that is an easier. We can follow the synthesis. Once nuclei reach the surface and you can see here the establishment of this gradient. You can see that it's graded. You can see that the protein being a transcription factor localizes to the nucleus. By this point, we've entered cycle 14. The cells have read and can distinguish these concentrations. The concentration differences between individual nuclei along the anterior posterior axis are in the order of 10 to 15 percent from one nucleus to the next to the next. This is an exponential decline, but individual nuclei are. The difference then that cells are going to determine a maximally precise boundary. That's the concentration difference that they must be able to detect. Thomas Greger was in the lab, developed this, the first of the usable bicoid GFP's, and showed that by nuclear, certainly by nuclear cycle 11, if you measure the intensity of staining and therefore the concentration of bicoid protein in the nucleus of individual nuclei, that concentration was stable by cycle 11, 12, 13, 14. The protein goes into the nucleus and as you saw in the movie when the cells were divide, the protein would leave the nucleus, go out into the cytoplasm, we can actually see this little increase in cytoplasmic levels during mitosis, and then a return rapid immediate, actually one of the fastest returns I've seen of nuclear proteins into the nucleus after the completion of mitosis. So the patterns are stable. Another way of looking at that, Thomas did was to measure the bicoid concentration in the individual nucleus anywhere along the length of the egg, and then allow the cells to divide and then measure the concentration or the intensity in the daughters. And that's what's plotted here and what you can see is that they lie on a diagonal, this is one, which means that essentially the daughters have the same intensity, bicoid concentrations as the mothers. Okay, so you were seeing, pardon? My cell cycle 1, 2, 3, 4 has been able to resolve. You can see it arising, I showed it to you at cell site, but we don't see that we have not been able to easily resolve the protein that early. It's just, obviously what happens is that initially it cannot be stable. You have a large amount of RNA, but a small amount of protein that you've just made. The protein is produced everywhere? No, the protein is only produced where the RNA is. The protein only happens? Exactly, the diffusion only. And the RNA is probably, we've also measured the distribution of the RNA in about 90% of the RNA, even though it becomes rather diffuse, is localized in this area. So protein moves from the site of synthesis through the rest of the embryo. Other questions? Yes? Yeah, yes. If I didn't have that, but if I were to only plot the red one, I would see some sort of peaks then the site would be directed to the red one. You would see more and more? Yeah. It's probably true because what's happening here is that we're increasing the number of nuclei by a factor of two with each nuclear division. But we are maintaining the same concentration per nucleus. The nuclei are actually getting slightly smaller, but the total amount of bicoid signal is going up still. And this is one of the complicated things about the process that the total signal measured in nuclei on the surface of the egg continues to rise during the cleavage divisions, although the concentration in each nucleus, and the concentration each nucleus falls along the anterior-posture axis, but in a given region of the egg, that concentration is constant. So what you'd like to know though, the ability to establish a stable gradient from a stable source of RNA, the amount of RNA is constant, so you're going to be making protein at a constant rate. And that protein is the shape of the gradient, the distribution of the protein is essentially going to depend on the diffusion of the protein from the source of synthesis and its lifetime. So a critical thing to measure then is whether it is in part the lifetime of the protein that will determine how long it will take for this gradient to achieve stability. So we wanted to measure the lifetime of the protein and the way that we did that was with photoactivatable fusion proteins and the one that we love is a protein called drampa and it has a peculiar property in that it's actually a photo-switchable protein that switches from a dark state to a light state depending on the wavelength illumination. And that's opposed to a protein that simply switches to a light state for a minute. So what's attractive about drampa, and I'm giving you a sales pitch on drampa, is that this is a slip of a position of a metal and you can do this switch back and forth a hundred times and the protein just goes dark, light, dark, dark, light, and if you use a photoactivatable form, the standard photoactivator will require much more energy to do the flips and they're not as easily reversible. So what we do is we turn all the protein to dark and the last thing is that it still flips in fixed material. So this is actually a fixed embryo that Oliver Grim who made the stock wrote Bicoid on the embryo and photoactivated the Bicoid in this embryo, in this fixed embryo. So on the ease of use it's ideal. The only disadvantage is you can't see it until you're done with your experiment. You have to believe that it's there. It's not like a switching from green to red where you saw it and you made it red from red and it was green. You have to start with an embryo that's dark and do stuff. That is a psychological, at least a psychological disadvantage. Okay, so two interesting features about degradation is we can measure. So the way you do these experiments is you would use the photo switchable to change the Bicoid in the embryo to a dark state or measure it, change it all to the dark state, and then change it back to a light state and ask how much is still there or subtract one population from the other. Figure out the lifetime, figure out how much has been degraded during the minute or two minute or five minute or ten minute interval that you've allowed degradation to occur. Two interesting features is that degradation follows first order kinetics. That's just what you would expect of a protein that has an inherent lifetime. There's nothing out there degrading Bicoid other than just the overall milieu and it is uniform along the entire axis of the embryo. So we're not going to be able to use degradation as measured by these qualities to shape the gradient. Lifetime is the same and this makes us believe there's no obvious feedback on concentration and the establishment of the gradient. Again, source of RNA makes protein diffuses and degrades. Questions? The rates are the same at all cycles. This is its early cycle 14 but we measured in some of the... It gets slightly more difficult in the earlier cycles because they're shorter. So if the cycle is only eight to ten minutes you have less time. But overall there doesn't seem to be any change. The lifetime is this actually. Our measurements during the early cycles 12, 13 and 14 where we can do this suggest that the life, our best calculation is something like 50 minutes for the protein. And then when the embryo enters cycle 14 and the gradient is used we see a rapid change in the lifetime. So this is the only developmentally regulated point in bicoid lifetime that we see and that the lifetime falls to about 15 minutes. And actually what you observe in the embryo is very rapidly the levels of bicoid fall. We think this is just by this point and we believe probably at this point but the concentrations are red and you don't want the information around anymore. A lifetime of about 50 minutes. What was the relationship between lifetime and being able to establish stable... It takes a couple of lifetimes. Yeah so we're in the ballpark of about the right lifetime to get us a stable gradient by cycle 12, 13, 14. It's not perfect. Lifetimes a little bit too long by about 10 minutes. And we don't know whether... I think overall we're just going to run with this number and say this is probably close enough and we're in the right order of magnitude. Okay. Pardon? No one knows. At least I don't know. We've tried to do some things. If you look at the sequence you can find regions that are called pest sequences that are thought to govern lifetime in other proteins. But no one knows how to interpret the sequences into a lifetime. We don't know who degrades the protein. We don't... We've looked at some modifications. Sometimes proteins are ubiquinated or otherwise simulated or things that change their lifetime. We haven't found... We're dumb. We don't really know. Question. We don't know that. Do we know that? We might know it and I could have forgotten it. If the egg is laid and activated but not fertilized by sperm it's called an unfertilized egg. It will initiate pathway translation but there's no nuclei there. And we could and I believe Jeff Draco in the lab did measure lifetime of the protein under those circumstances. And I honestly can't remember whether it's very different. It wasn't like, oh, change our world view of how the process is occurring. Question of that implies that if you can't unfertilize that there can be a gradient of the protein then that means the mRNAs are released even when the sperm hasn't made contact with it. Yes. So the mRNAs are released into this translation. And this is something that's known for general translation actually that is an activation process. And this is also true for many embryos. That is not entry of sperm but some process that's called activation that involves the calcium influx most times in most embryos that somehow activates and occurs simultaneously with sperm entry. But sperm aren't necessary for the right, absolutely right, for activating this process. Normally they occur simultaneously. And so one of the things that is obviously interesting from the standpoint of where degradation occurred, one of the features is nuclear that when we look at the protein except during mitosis we generally see the protein in the nucleus. So the sense is that you're establishing a gradient of nuclear protein. And so the question that Oliver Grimm wanted to ask is does this diffusion pattern, this accumulation, the final shape, does it depend in some way on Bitcoin being, say, trapped in the nucleus? Does that slow down its diffusion? There were two, in a way, I'll show you this result. So what Oliver did to address this was he made a mutant form. We don't know very much about Bicoids, the control of Bicoids nuclear entry. It doesn't have an obvious NLS, nuclear localization sequence. But what was done was somewhere in the region of the protein actually fairly close to the homeo domain, which is the DNA binding domain of Bicoid protein. You could delete that, those bases, and you would still make a stable Bicoid protein, but the protein no longer partitions specifically into the nucleus. So this is a wild type normal protein that's been tagged with CFP, and you can see the standard Bicoid gradient. And this is a YFP tagged version of the mutant form, and you can see that it's cytoplasmic and nuclear. And we know that this is something to do with the entry rates, the impact of this mutation on entry rates into the nucleus. Again, we don't know enough about Bicoids of the actual process of nuclear localization to say what's really going on here. But again, you have a messenger RNA, it's translated into a protein. And now what we can do, because these are differently tagged, we can put these two constructs into the same fly and follow the levels of the CFP versus YFP. And ask, is there any difference? And along the entire length of the entry into the nucleus makes no difference on the shape of the gradient. Now this is kind of like a surprising result, because it must be spending all of its time in the nucleus, but that is based on looking at the staining patterns in localization, that's what you might think. Actually, in retrospect, you can do a FRAP experiment, where you can photobleach a nucleus, and then ask how long does it take for that nucleus to recover. And that kind of photobleaching experiment is a little bit more reliable. Some of the issues that art raises, but when you do that you find that even though Bicoid is localized clearly to the nucleus, the entire population of Bicoid is moving in and out of the nucleus with a half-life or a half-time of about 60 seconds. So Bicoid is going in and out of the nucleus into the cytoplasm becoming dispersal. We don't see it so well, but constantly being partitioned probably in a concentration-dependent way that's similar to how you would phase partitioned solubilities between two domains. The cytoplasm of the domain is much bigger, but you're making... Okay, so I realized... You make the gradient, you probably make the gradient in the cytoplasm. It partitions into the nucleus and supplies information. At a given point, when it's above a threshold, we can look and say, Hunchback is over-saturated, as you'd see here, Bicoid would bind to and activate transcription of this Hunchback gene. Now, what I'm showing you here is actually a slightly different view. We're not focusing on the surface, we're focusing through the optical midline of the embryo. And what I just wanted to show you is that, and I didn't point out to you, is that during these syncytial stages most of the nuclei migrate to the surface, but there's a certain fraction of the nuclei that are just left in the cytoplasm. You can't see them there, you could have potentially seen them in the earlier figures. And what I wanted to show you is that you can measure Bicoid concentration, and it's partitioning here as well. And we don't see...it's tricky to measure whether the internal gradient is exactly the same as the surface gradient, but they are very, very similar. And in terms of transcriptional responses, they also activate Hunchback, again in a border that corresponds to the border at the surface. So we think basically the protein's going through the cytoplasm, it's there, and somehow partitioning into the nucleus. Questions? Yes. What becomes...these nuclei here, there's Bicoid RNA here that we don't see, that's being translated into Bicoid protein that is accumulating in these nuclei. There is Bicoid RNA only here. So, yeah, are you asking that if you have a diffuse mass of Bicoid RNA, could some of the Bicoid protein doesn't have to go too far to get to those nuclei? That's true. That's probably true. It doesn't have to go too far to get to these either. It has a harder time getting to these or these. And so we think that it's diffusing probably with the same kinetics. One of the reasons for that, and I didn't show you, is that if you take a gap gene-like cripple that would normally be expressed in this area here in response to Bicoid, we see Bicoid coming up, we see cripple coming up here. That would have been a better demonstration that the protein is, again, moving away from its source, establishing a concentration gradient. Are there other questions? In cytoplasm, periphery and in the yolk. So you say they are almost the same? I would say they're almost the same. Yeah. I have not measured that. I measured it. And what did you find? I find there is nothing in the yolk. That's exciting. I mean, for a variety of reasons. One, we're turning on hunchback or cripple, known Bicoid responsive genes in a pattern that... So in nuclei, you have the hydrogen concentration. And you go to cytoplasm, the concentration is way too low. And as I do measurement in yolk, it's just very good to that. And I'm not ready to eat in the yolk. Yeah. So very possibly what you need to do is look in nuclei in the yolk. Because if you look at this figure, I would say exactly that's what I see. Let's see. There's more protein here and then less as you're going through here. So you do an overall, you say, ah, no protein. And when you hit the nucleus, you would see Bicoid protein. I bet. I guess it's worth it to maybe to look at that again. This is at least the way it looks like for us. Now, okay. So overall, we think that the measurements that we've gotten. I think, did I not include? No? Okay. So things that we've gotten are basically compatible with a really simple model for how you produce this gradient. You have a constant amount of RNA, translated at a constant rate. It has a constant degradation, lifetime, and moves with a certain dynamic and makes a gradient. And that works in flies. And I come to an interesting thing that we've never really fully been able to understand. And that's that, as you all know from your own personal experiences, flies have a very wonderful and diverse species of animals. And they come in different sizes. So they're really attractive, beautiful, you know, Drosophila wonderful flies that are wonderful in the lab. And your heart immediately goes out to with great affection. I said that to make you guys laugh. Thank you. And then there are these ugly annoying ones that show up at picnics or in your house and buzz around and are huge and ugly. And they're partially ugly because they're so big. And well, you know, it's interesting not only are the flies big but the embryos are big. So the species like blow flies and house flies make these. This is a calyphera egg. This is a house fly egg. These are Drosophila melanogaster eggs. There are even some little, there's another species of Drosophila called Drosophila buski, which makes eggs that are about 300 rather than 500. So you get this whole range of egg sizes. And yet all of these are fairly closely related diptera and they all have bicoid and they all use bicoid to pattern the embryos. So you could ask, how does this work? Well, they're close enough that actually you can use many of the standard Drosophila reagents and say, do they have gap genes? Do they have pair world genes? Do they show, are these equation patterns similar in the different insects? And what you can see is that, yeah, eggs can be big, eggs can be small, and you establish the same patterns of gene expression. The other thing that's really interesting, especially for some of the things that Stefan talked about, is that the early development of these different species are very, very similar. So they all take about two and a half hours to reach the cellularization level. They all go through 13 rounds of division and stop. The big ones make blasted irms that are big, but they have the same number of cells. And they're all very similar in all these respects. One thing that's different is that they're eggs of different sizes. And so what we would like to know is, how do you do this? Obviously you can make a gradient and you can make a gradient and if you have a longer egg, you keep a gradient with the same properties and it would extend out farther but had lower concentrations. So you would adjust the cis-regulatory, the enhancers of all of these response genes to allow you to adjust for the size of the egg. That would be a cool way of doing it. Or you can somehow change the distribution or the length scale of the big gradient. And the answer is that when you look at the gradient in Drosophila booski, which is 300, or Drosophila melanogasterin green, which is 500, you look at the shapes of the gradients or Lucilia, which is one of these big flies. The gradients are bigger in the bigger eggs. And you can show that if you collapse, if you plot the data not in terms of microns, but in terms of percent egg length, you collapse all of these gradients back onto themselves. So what that means is that the distributions are what scales here. Is the distribution of big weight. You can go back now and you can say, well, what is the properties that the big weight gradient depends on? The distribution is like half-life and diffusion. We know pure diffusion. Thomas Gregor injected different fluorescent probes into the eggs. You could measure diffusion. It's the same in all of these. So it almost has to be something special about the proteins that have different half-lifes in these species. And that might be maybe the best model, but Alastair McGregor did was to then clone out the big weight genes from these other insect species. And you can plot them out and look at them here. It is pretty much, you know, there's a certain degree of conservation, certain things that are different between them. Nothing truly very helpful. They all have this conserved pest degradation region. They all have the homeodomain. They all look like bicoid. There's no doubt this is bicoid. These are the bicoid proteins from these. And what he then asked was if one took these bicoid genes from other species and you tag them with GFP and you make transgenic Drosophila melanogastera by putting these big bicoids that make big gradients into the little tiny egg. You could imagine one thing. We thought we could get big gradients and little tiny eggs and they would just fill up the whole egg with bicoid and that would be really interesting. And he spent the next year just looking at interesting phenotypes and it would all have been very exciting. But actually what happens is that you put the big bicoid from the big insects into the small eggs and the size of the bicoid gradient scales to the size of the egg. What that means is that it's comparable to Lambda's... Yes? The sequence of the bicoid promoter. The bigger eggs have a stronger promoter. See if you made more protein. One of the ways that you could do this is actually making more protein doesn't work. It gives you a higher rate of synthesis, presumably, but the shapes are not going to scale. The Lambda... I have to work this thing through. Can we say yes, would you believe me for a second and then we'll go back to this discussion? So changing rates of synthesis doesn't help. The problem has to be an odd conserved property of the protein. And... Oh yes, of course. That would be cool to do and we haven't done it. That's a great thing. Or you could even... So part of the problem is working with these ugly flies because we are not able to grow them in the lab. They have long... And so we order them. So one of the things... We couldn't do the reverse transgenesis and put the... But we could always... You're right. We could always get the flies, get an extract, and just measure lifetimes. I think that would be a great thing to do. And then you could chop the protein in half so that you could begin a structure function analysis to begin to address what it is that controls the lifetime and what the degradation process is. We don't know. So what it does suggest, though, that Bitcoin as a protein is built... This protein is conserved, the property. So what's conserved is Bitcoin protein, as the egg changes in length, a property of the egg must change in this model that would change the lifetime of the protein. And it would be interesting to know if those are the same... How the properties relate to egg length. Okay. I'm going to stop there. That's probably the coolest experiments that we have in the past 10 years at Bitcoin. I think they've helped us to think about the process, but they're still kind of open-ended in terms of our understanding mechanistically or biochemistry. Again, yes. Excellent. That would do it for you. In that sense, that's kind of what I was getting at, that the same property that increased the size of the egg is the property that decreases the lifetime of the protein because you dilute out that particular degradation factor. But it would mean that Bitcoin would have a degradation factor that was specific to it. Because the cell cycle degradation things are everything else is running at the same time clock. And so all the other degradation processes are still running at their normal clock in the bigger eggs, but Bicoids in this model would be going at a slower. That would be cool. And that could be a property that is conserved in the Bicoid protein, evidently, and would be a property that could be recognized from the Bicoid sequence potentially if one recognized what was found. Yeah. The same. Yeah. They go as fast. They go as fast, but an argument would be that if the protease was more diluted, you would change the lifetime. Yeah. And that's a great model. Yeah. Why did you pick up mutants in that? You would think. Yeah. Right. So what you would have, what we don't pick up is mutants that aren't specific for oogenesis or for that. So if it were really specific for Bicoid, which is what I was arguing, remember, because we wanted it to be specific because we want all the other processes to go at a normal clock rate, but Bicoid to slow down, then the protein should be specific if it's really specific. Then you knock it out. It shouldn't be homozygous lethal. One should have obtained it as a maternal effect mutation. The only way of saving that argument is to say that there are some other things that are going on in the making of the egg, a big egg, that happened during oogenesis. And that same degradation property is going to be, that same degradation gene is going to be used for those and would be slowed down or sped up in the big, making the big egg involves diluting out a particular factor and that when you made a mutation in it, you messed up oogenesis and never got an egg. And that might be harder to... The other more likely explanation is that when Trudy and Yanni did the screens for maternal effect mutations, they stopped too soon. They may have hit 90% saturation, but there's still that interesting... 10% interesting genes out there that haven't been identified. Okay, so what I'd like to do is step back a little bit and try to address the question of information content in the Bitcoin gradient. So we're showing that it's stable and what stable means is between many different embryos, if you see this particular concentration, you know you are at this percent egg length. And that's at the 10% level, which is basically enough to distinguish one cell from its immediate neighbors. Now, historically, if you look at how the fly field looked at the original process of... thought about the original process of segmentation after the initial gap genes and peril genes and segment clarity genes and maternal effect genes, they were kind of arranged in a hierarchy, a conceptual hierarchy. And the view was that these genes function to gradually establish precise patterns. And in that view, you know, a simple version of that view is that what maternal effect genes are really doing is setting up blocks of gap gene expression and then the gap genes kind of interact with each other and make more complicated patterns and then you get eventually peril patterns that are built on the... and so that you gradually establish a precise determination of cell fate that is, therefore, allows you to say the information content is low and that's how people were used to think before Thomas Greger actually did the careful measurements of the bicograde gradient. You know, it was enough to give you some kind of a polarity, it wasn't a lot of information then, that information complexity increases in this kind of model. And there'd be less or little correlation between the initial input and the final pattern or the final response. If you're going through many steps and you're generating pattern and each step is subject to noise and so you're going to be... there's going to be less correlation between input and output. One of the few... one of the outlooks particularly in the Princeton bicograde consortium which is, I guess, Bill Bialik and Thomas Greger and myself and the participants in the lab a view that arose about the time that we became aware of how precise the gradient was is the sense that the initial gradient is a very rich source of spatial information, reproducible, and it is translated into a rich pattern. So this is essentially an information transfer. Information is in a greater distribution... and we want to take the same information but convert it or translate it into a system of discrete gene expressions without losing any information. And that's the way to think about it. Information complexity would be stable or potentially even decrease with steps and there would be still a strong correlation between this... based in part on this observation of Thomas's that the gradient was so incredibly reproducible. So... but it really depends on... and this gets to the real theme of the second part of this talk is how well do... even if the information content is really high because the information can be put from the same moment of RNA into the egg and has the same lifetime and fusion properties aren't very different and you really get the same gradient out. That's physics. And so there's not necessarily going to be a lot of variability in... but can cells really use all the information that's there. And so the approach that... Julia Dewey, who's a graduate student is working kind of with all of us but more specifically with Thomas and Bill did was to actually go back and try to measure the precision in the information content in the gap genes at 15 minutes into cycle 14. You realize there are four different gap genes. What we did was we generated antibodies and different animals to all four so that we could simultaneously in the same embryo characterize all four expression patterns and then characterize those... the expression patterns and initially we used a statisticians once you get to how you're going to express that you can... Julia Dewey says, which is quite beautiful you should read it if you have a chance translates this spatial information into bits of information that allows you then to ask how much... you can look at how many bits of information are in the big grade gradient what's the information content of the gap genes the total information content mutual information... Yeah, there's no time... there's one time. We picked the time where... But this thing is very dynamic. Yeah, so there could be more information but our point is really that if we identify a single time where we have enough precision to specify cells different from each other then it could be that if we'd looked 10 minutes earlier or 10 minutes later we would have the same level of precision or less or more but if there is a time where the gap gene expression patterns contain enough information to specify single cells then the embryo would be really smart to use that time point for the next step. Yes, and Julian didn't do it the next version of the manuscript version of this is kind of doing this but just the conclusions as you see from Julian's paper was that if you ask and if you measure... these are non-normalized well they're basically the conservation... normally if you do an expression profile you would anchor or you would normalize with the top highest expression or lower these are... our strategy is to always include wild type embryos not normalized to use absolute values and I told you if you tell us the absolute intensities of these four genes along the axis of the embryo all four can you say precisely where you are or how precisely can you say where you are and to do this you have to change your view of the gap genes from genes that are expressed in blocks to normal distributions with graded edges because the information content is really going to be in the graded regions of gap gene expression and if these are reproducible and the positions are reproducible if you can get information from all four of these genes you will know where you are to a precision of about 1% which is again sufficient for us to tell cells one cell from the next so if the embryo is as good as we are in measuring concentrations and reading concentrations the embryo can use that gap gene information to set up different phase now this allows us to do take a slightly different approach and this is the approach that Mariela Petkova has using to try to use this data to build what are essentially lookup tables so a pattern of gene expression and then we'll plot it out here so if a cell is actually sitting at position 5 of 50% egg length it sees a particular combination of concentrations of these four genes and it makes a prediction of where it thinks it is so this is the a picture of what we think the embryo is doing and then it activates at it's estimated depending on what it estimates it will activate a pair of genes or not so you can then use this is kind of an internal representation of what we believe the embryo is doing in sitting in a position seeing a given concentration of gap gene estimating what you could call position which is really do I activate this pair of genes or not we can use these lookup tables in a certain sense the width of this little bar here this is real data so it doesn't show up here and I'm cutting lots of stuff out but gives you a sense of the accuracy of a cell's ability to measure so what I could have showed you it would have made me helpful is that there would be one gap gene you would a cell sitting there and if you could only see one would in some places this would be very very very fuzzy because it would have no idea where it was and in some places it would be very precise but maybe there would be two places where you'd have that precise combination and you couldn't decide and so you get complicated patterns with individual gap genes that become increasingly less complicated as you add in more and more gap genes until you get to a perfect lookup table now this becomes interesting useful for us I'm going to use it a little bit later because it allows us to evaluate the information content of the eggs in terms of gap gene how good is the how we produce how good are the cell fate decisions how good is the cell's ability to estimate where it is in the embryo based on the fuzziness or sharpness of the line for us so one ensemble is variability from embryo to embryo so there's a whole different thing so this is a different experiment actually compared to this experiment there are about 300-400 embryos evaluated at 5-minute intervals during cycle 14 to and then the 15- to 20-minute interval was chosen as the most information rich in that the line here was the least fuzzy and allowed the tightest predictions for where carotid genes would form yeah so it would be yeah it's really absolutely true that you could get more information maybe if you looked at time or maybe you wouldn't get any better that's an interesting question to ask would you get better can you get below 1% if you integrate over time that's a question that could be asked now with this data set that way it is no I think it my personal view is I wrote in 1984 or 5 this paper where I argued that there were many more meaningful positional cell fates than there were even cells along the anterior posterior axis of the blastetor so I would love your result I it would be interesting to know if we can do that now we can actually ask whether integration of additional time points gives us greater precision or not yes the direction that this project took was not in that direction but the data is there to do that and that is the right thing to do that is among the right things to do that's an interesting one I'm just wondering could be feedback is there an opportunity where cool questions yes unanswered okay so I want to use this just what I want to do is in the last 15 minutes go back to this question of how the embryo does this and we've been talking about this this is bicoid but a worrisome thing of course is that I've already told you yesterday that bicoid is not the only anterior posterior patterning system in the embryo there's another one that's made by an RNA localized the posterior end of the embryo which regulates the translation of a maternally supplied hunchback RNA which is the same protein the bicoid is regulating as well there is also a receptor tyrosine kinase signaling system that regulates the localization of another so all of these systems govern or have great impacts in certain regions of the embryo bicoid in terms of its phenotype is missing mostly anterior and thoracic structures so people in the past have kind of partitioned these genes into different roles what we would like to know so I've just told you that the embryo is the embryo precise because it's using all of this information and if we were to reduce the embryo down to having only bicoid information how much what can it do and what time do we must be running we have ten minutes so we are running out of time so I will go through two quick experiments this is the experiment that I did and I'll do it in one slide I put 18 slides about my experiment into this talk and we're going to cut them all out but to make embryos which have only bicoid content you have to also not only remove the initial inputs you have to remove the downstream transcription factors that would be inhibitory for bicoids patterning effects I can't explain that to you other than to tell you that that involves making genetic mosaics in a complicated process and this is the only picture that you really need to look at this is a wild type larvae it would hatch at the end of embryonic development this is the pattern of canyreps and giant and even skipped if you have only bicoid what you have is an embryo which has patterns out to two abdominal segments six which is a lot of the pattern and is gap genes basically the whole run of gap genes and I can tell you that this patterning depends on bicoid because if we now remove bicoid from these embryos just looking here this is what you get you go from this to this when you remove bicoid so this is what bicoid is doing in these embryos and they do it pretty accurately but I won't so now we're going to go back and we're going to say well we have this bicoid gradient we have it monkeyed with it it's the same in these embryos and somehow the embryo is able to pattern out to 60 or 70% egg length in a gradient that's as flat as this area is here so how does it do this how does bicoid pattern up to this point most people would say bicoid is important for the anterior region of the embryo when you remove it you still get an abdomen and that tells it's somewhat redundant with patterning but this bicoid can pattern an abdomen we don't know whether it does it directly or indirectly and we don't know that because what the targets are and we don't understand directly presumably means that bicoid binds to the enhancers of target genes and activates them the enhancers of genes that are expressed all the way out here so now I'm going to talk about what I believe is an exciting set of experiments that really begin to address what is the nature of bicoid binding with the nature of the concentration dependent binding if bicoid is able to pattern and pattern directly how do you build enhancers that can respond to different concentrations these are experiments of Colleen Hannon and Shelby Blythe in the lab in the simple model you know what it is and the idea is that anterior targets should have enhancers of affinity and they would only bind bicoid at high concentrations that would be the standard textbook model we'd like to be able to first test whether that's true and then ask if that is true can we understand the molecular basis for it the approach that Colleen has taken is first to try to identify bicoid we've talked about these four gap genes but there are many other genes that are expressed in anterior post-year patterns in the embryo bicoid governs all of these directly or indirectly so what she wanted to know how many sites are there where bicoid binds to do that she used chip seeker chromatin immunoprecipitation using bicoid as an antibody to pull down the sites in the entire genome and bind bicoid DNA and in her hands the reproducible number were using cutoffs that she chose to use are 127 sites in early cycle 14 embryos these are hand selected visually selected embryos somewhere between 50 and 100 embryos that bind bicoid and these 127 sites bind bicoid in wild type embryos they also bind bicoid in embryos where we removed all the other maternal inputs and altered all the other gene expression pattern they don't change and so we think that binding even though expression patterns change when you remove the other genes that binding is a property that's kind of independent of the actual the downstream output of the gene so now she has a bunch of targets but what you really want to know is do these targets show different concentration dependencies and so that's hard to do in a wild type embryo where every cell has a different concentration of bicoid so what Colleen then did was to develop a set of lines in which bicoid RNA is expressed in the mother but with a 3 prime UTR that prevents its localization so it is made in the egg and uniformly distributed to the whole egg cytopacin and she drives the expression of these RNAs with different promoters that drive different levels of RNA production and therefore different levels of protein production what that means is she can make a large collection of embryos hundreds of embryos that mimic the concentrations at specific points along the gradient and then use these embryos with defined concentrations based on calculations from westerns or however you want to approach that to ask if you have a set of 100 enhancers or regions of DNA that bind bicoid in wild type embryos do those enhancers partition themselves into groups depending on their concentration their bicoid binding in these different concentration lines and the answer is yes they do the answer is that the lines are interesting and functional but the important answer is that with the three lines that she has she can identify four classes of DNA binding this is a very crude division you could I'm convinced you could probably parse out more classes here but for our practical purposes statistically was easiest to identify break these genes into four classes some which are we call high affinity because even under all bicoid concentrations regardless of the promoter that's used they are bound others are low affinity because they're only bound at the highest concentrations and intermediate low affinities and intermediate we have four classes so far this is just classes based on concentration what you want to know is there do these we imagine that these function as enhancers so you'd like to ask do these pieces of DNA show what are the expression patterns associated with these pieces of DNA if you hook them up to reporters and there are different Drosophila databases we've gone through four different databases here but the low affinity ones tend to be associated with expression in the anterior ends of the embryo and the high affinity ones have expression throughout these are the ones which are expressed even at very low concentrations and they tend to be expressed and these pieces of DNA tend to drive expression we can do that by really with a really large data sets were generated in Starks lab in Vienna and you can show that there are significantly different behaviors for all four classes in terms of where they activate expression so now we're excited because we have for the first time a big enough set of enhancers that we can begin to say characters that would determine high affinity versus low affinity so all of these what are the motifs that are present and they all you can ask the computer what are the most common motifs the three most common motifs are two different bicoid sites that are known as the high affinity and low affinity site a gene called Zelda which is a chromatin architecture or which is a gene that's involved in early activation of transcription in Drosophila those are the three most abundant motifs if you type the DNA though and you look at these regions you can indeed yes it's a co-active it argues that these are probably early enhancers that are driving early expression in the embryo so they have Zelda sites but they have bicoid sites that's the next most abundant site they all all these pieces of DNA if you take them out not all of them, Colleen's done 10 they bind bicoid in what are called gel shifts experiments which is a physical measurement of bicoid protein binding to the DNA sequences that's interesting but they all bind with the same KD so we can't whether they're in the high affinity or the low affinity because we can't distinguish that at all that's telling us that naked DNA doesn't mimic the concentration specificity that we see in the in vivo what distinguishes the peaks a really sensible easy model whoop that if you counted the densities of the bicoid sites you looked at that high affinity sites would have lots of bicoid and the low affinity sites would have few bicoid binding sites and so we did that you know that type of and that's not true and we also don't see differences between different kinds of bicoid sites but we see one thing which is really amazing is that in the opposite of what we expected that the intermediate and low sites that have the lowest affinity for bicoid are the ones that have the most bicoid sites and then we ask is there any other so that suggests that there's something complicated going on here they don't do it in vivo maybe there's something about chromatin and the availability or accessibility of these sites and so we can measure all the sites have a property that's called accessibility when you do chromatin assays it means the DNA is available to some reagents the reagent we use is a taxic protocol and much of DNA is wrapped around his stones and not easily accessible to the reagents and regions of naked DNA sitting out there waiting for proteins to bind are often accessible and these sites are predominantly all have high degrees of accessibility but there's a peculiarity the way chromatin is made is that you have something called a nucleosome which is a bunch of histone proteins that are balled together and you wrap DNA around and that protects that DNA if you will I used to think that DNA is wrapped around his stones in a random way but we know now there's a sequence specificity in this wrap that has to do with how easy the DNA is to bend and also the preferences for certain bases to lie in certain grooves in the nucleus so you can take DNA and you can ask on a very small scale how likely is this DNA to be wrapped in his stones wrapped around his stones and when you do that you can get a measure and what we find is that that turns out to be one of the best criteria for distinguishing between high and low affinity sites even if the low affinity sites have a very, very high propensity to be wrapped around his stones that are unavailable and inaccessible and the high affinity sites are just have lower tendencies and tend to be tend to be open and so the model that I'm going to give you high affinity sites are open and accessible, low affinity sites are closed high affinity sites these have moderate densities of a high density of bicoid so what we're arguing now is that if you want to build and the way that the animals build concentration dependent enhancers is that they balance a proclivity to wrap his stones and be inaccessible with a number of bicoid sites that will make that DNA accessible and so in the interior region low affinity sites will tend to have his stones high affinity sites maybe also because they are enriched for Zelda which I didn't point that out but that was this co-factor high sites will have the bicoid binding motif immediately available there so you don't low concentrations are able to bind here high concentrations we would have to argue then have the property of excluding nucleosomes opening up the DNA so these high affinity, the low affinity sites for them to be expressed you have to open them up and what we're going to argue is that the high concentrations of bicoid is in fact the feature that opens those those proteins up those regions up so then concentration dependence then becomes the balance between the nucleosome preferences and other potentially other factors as well that make these sites inaccessible plus the high frequency of bicoid sites and so you can test I'll just tell you that and we what this model implies is a unusual feature of that you don't normally assign to homeodomain proteins that they remodel chromatin generally we think of transcription factors as sloshing through the nucleus looking for naked sites of DNA to bind to and to activate transcription but what we're arguing for these concentration dependent cases is that the transcription factor this is a homeodomain transcription factor as typical as you can get has the property that it opens chromatin and it uses that property to drive concentration dependence so I'm going to stop there I went over I'm sorry but thank you good audience okay we're going to kill that