 and okay. So hello everybody. We start our QLESS guest seminar webinar series. I present to you Adrian Jacopo from the Rockefeller University. I believe also Howard Hughes Medical Institute is involved in this project that Adrian develops. Today Adrian will tell us about symmetry breaking during morphogenesis of a mechanosensory organ. This talk may slightly overlap with the previous speaker Anna Erzberger, but I think that the topic is very rich and Adrian would tell us about very different things from what we heard in Anna's talk. All right Adrian, I invite you to proceed. Thank you Roman. Thank you for the invitation and thank you everyone for being here virtually. So yes, there will be some overlap with what Anna told you but yeah I'll try to, so we try to agree with Anna so that there's no so much overlap and hopefully you can learn a few new things. So as you probably know like at least some of you from talking with Roman, we in the lab study mechanosensory petylia so the essentially the tissue that enables our senses of hearing and balance and what I'm interested in and I've always been interested in is in pattern formation and one of the reasons that I joined this lab is because these mechanosensory petylia are one of the most striking examples of patterns in nature and of self-organization. And what you're seeing here, this picture is the picture of the apical surface so it's the top of what is called a bullfrog saculose. So this is an organ from a frog that the frogs use to sense vibrations on the ground and that way they can detect approaching predators essentially. So they sort of lay on the ground with their what is would be the equivalent of their ears touching the ground and then these organs pick up vibrations and they transmit these vibrations and they transduce them into electrical impulses that are related to the brain. And they do these things to this structure which is called the herbandle and I'm going to talk a little bit more about how that is done but what I want to focus on in this picture is how remarkable these structures are. So in this organ first of all you have structures that are at multiple scales so at the cellular scale you have this structure called the herbandle. So these are sort of very well reproducible from cell to cell and from animal to animal and they're formed by these sort of finger-like protrusions called stereocilia and the variations in length and number of stereocilia in from cell to cell are controlled within like five to ten percent. So this is a very well controlled process in which cells can form these structures and these patterns and I'm going to talk a little bit more about the patterns in a second with a lot of reproducibility in this environment that is very dominated by noise so the cells the production of proteins inside the cell is a stochastic process it's very noisy there's thermal noise there's noise associated to the copy number of these proteins but nevertheless they're able to create these remarkable structures that are that are sort of hundreds of nanometers in size and they're in some sense they're better than anything we can do with current technology and the other thing that I want to uh the other two things that I want you to focus on is that these bundles are all oriented in the same direction so they have this sort of high gradient that is always pointing in the same direction and this is essential for their function because they are sensitive to to movements displacement of the surrounding fluid just in one direction in a long one axis and so they they have these so the organ has these sort of long-range polar order where all the cells point in the same direction and the other thing that I want you to focus on is how the cells are sort of space regularly forming these hexagonal patterns so you see the cell at the center and has six cells on the size and sort of these form this quasi-hexagonal crystal that is that spans the whole organ okay so these patterns occur at multiple scales and these scales are coordinated okay with one another the so in particular for example the polarity uh that is sort of long range has to coordinate inside each of the cells it has to tell each of the cells which way they need to uh to point so let me tell you a little bit more now about um Edrin I'm sorry have you changed your slides or it's always the same I changed the slide now is it stuck now um it did not work okay can you here restart the second slide so you're seeing the second slide now here in bounds and making a reduction yes so now yeah I see the second slide yeah now okay um so um as I was saying before uh these tissues and the organs that they um that they form uh are used for um hearing balance and hearing and balance okay so this the this organ um that I showed you before is uh shares a lot of similarities with the tissues in our inner year and in particular what they share is that they form by the same type of cell so these hair cells are also present in our inner years and they allow us to so they allow you to hear what I'm saying now and they also allow your semicircular canals so these three small organs here to sense rotations of your head so these are like uh essentially like gyroscopes and they sense the like accelerometer sorry and they sense the rotations of your head like your font does and then there's a vestibular organ that sends linear accelerations of the head so if you are for example in an elevator with your eyes closed you know that the elevator is going up or down um uh just because uh the this the the vestibule gets activated and senses these linear accelerations are like if you're in a car or in a bus uh you sense the accelerations thanks to the utricle and they do this again as I was saying because um these organs contain these head cells so this is a schematic of the um of uh one of the cells that I showed in the previous slide so this is a hervandal which corresponds to this green structure here this sits on top of the whole cell that you don't see here because it's covered sort of by the epithelium doesn't let you see what's um below and then um if we zoom in into the hervandal we see is that it's formed by these acting protrusions called stereocilia and these acting protrusions are connected to each other by these chains of proteins called tip links that are sort of like uh springs and these chains of proteins connect to um uh ion channels and these ion channels open when the tip links are put into tension so when there's shear motion by so there's deflection uh created in the surrounding fluid and this here is created by let's say movement of the head or by sounds uh then uh these tip links these chains of proteins are put into tension the um channels open and then ions enter into the cell the cell depolarizes and then relays through the um um so the neuronal connections relate this information to the brain so this is the way that these cells transduce um uh mechanical movements into electrical signals and just uh just to complete um this slide so this is a view it's sort of a side view of this hervandal and this is a view of a similar cell where we remove all the stereocilia so the these um uh dots here represent show where the the stereocilia were anchored on the surface of the cell and what you see is that they also form this hexagonal arrangement that is very precise so the distances the variation in distance between um between two of these locations is uh within uh one percent and these are like these are like 10 nanometers apart so this is very very well controlled up to a nanometer in distance and this is something I also work on and we collaborated with Roman also in a paper trying to understand how these patterns are formed and how is the noise uh controlled well so what are the limits in uh to the um noise uh the noise in the expression of proteins that would allow for um such a precision to be achieved but so this is not what I want to talk about today we can discuss it later if if anyone is interested um so uh just to make the the point further there's these patterns are multiple levels so you have the pattern at the subcellular scale that forms this hexagonal arrangement of the stereocilia and also this staircase that is also very well controlled and then sort of at the cellular level you have what I was talking about before so you have this long range um order of cells that point in one direction or the opposite and this makes the cells sensitive for example in this case this is a human vestibule so this is um um it's a mouse vestibule so this sorry is um these cells are sensitive to movements of the head towards sort of the to forward movements of the head and these cells are sensitive to forward movements of the head okay and and so this is another example which is the organ I'm going to talk about today um this is in a in a fish and you can also see how sort of this green this black dot so sorry this the other thing I need to explain so this black dot uh shows the position of this so the tall end of the um of the hervandal and this um shows what is the the um direction of sensitivity okay so these black dots makes these cells sensitive from the left of your screen towards the right and these in this case these cells are sensitive in this direction these cells are sensitive in the opposite direction um and so on and this is an organ that I'm going to talk about more it's an organ in the fish that I'm going to talk about more in detail um and the fish use this organ in a very similar way to detect um flow of water okay and I'm going to tell you more in detail how it's done but then let me move also to the tissue level so this is the other thing I was mentioning before that like if you look at the whole tissue and this is a uterical in a mouse so this is again an organ that the mouse used to detect linear accelerations of the head the cells in red which are the hair cells are sort of in a this pepper spray pattern and they're always uh surrounded by these other these green cells around them which are um the supporting cells the so-called supporting cells and this is also uh reproduced in our in the cochlea this is a mouse cochlea again but it's very similar to a human one where you can see the cells sitting in this hexagonal pattern and they're surrounded by these sort of black unlabeled cells that are the supporting cells and this is always um the case in this mechanosensory pitella so as always you get these um arrangements in which um you get you never get two hair cells in contact with each other they're always separated by a um support cell this is something also I worked on I'm working on trying to understand how you get this uh how you get this arrangement but now let me focus on uh the story I want to tell you today which is this organization um of this polarity reversal in the in the zebrafish so again so this is a schematic of a two-day three-day old zebrafish and these green dots form what is called the lateral line which is formed by these organs called neuromast that the fish used to detect water current so these organs are right underneath the skin so this is sort of these uh uh pinkish uh cells are the skin layer and they have this very long kind of cilia so these which are represented by these black dots that stick out the um the skin of the fish and act as antennas okay so these big vibrations the vibrations get transmitted to the hair bundles which are these green structures here and then the hair bundle hair bundles get uh deflected and this is how the fish knows which direction the surrounding fluid is moving okay and they can use these organs to detect many things so one is proprioception so they can detect uh the direction and the speed at which they're swimming uh and they can also detect uh predators and prey in the surrounding water okay they can detect vortices in the surrounding water they can detect flow if there's a predator trying to suck them in and eat them and and so on and they do this because uh this organ has two varieties of hair cells so there's one uh half that is pointing towards that is sensitive to movement towards the tail of the fish these cells here and there's the other half cells that are sensitive to movement towards the head of the fish these cells um here okay and so this is an this is a schematic and this is sort of a real picture of some of these cells uh where you see three of them pointing towards the tail three of them pointing towards the head and so the the thing that we want to understand is how is this arrangement achieved okay so this is a sort of a reduced version of what I told you before with this long range order so in the in the newer mass there's not that many cells so the order is sort of local and you get this sort of line that divides the cells pointing in one direction and the other and we want to understand how this comes to be okay and this organ is very robust so you can kill all the cells and it will generate the same arrangement of cells over and over again and you can kill a few cells and the new cells will always be um sort of uh having this ratio of 50 50 and the cells um the new cells uh facing each other so the first thing we want to do uh if we want to understand how this organ self-organizes is just look at it and see what happens okay see where these cells come from okay so what you're seeing here is um a confocal picture so we're looking at the cells in the microscope um and we have um the we have a genetic genetically modified fish where the hair cells and only the hair cells so the there's surrounding supporting cells here that you are not seeing expressed a protein that is green fluorescent so express these green fluorescent protein um and we're looking at a focal plane so that is sort of cutting in the middle of the of the of the neuromask so what you're seeing is sort of the nuclei of the cells and these very bright cells and what you're going to see now when I play the movie is that there is a stem like cell here that makes a decision that it wants to become a hair cell so it starts up regulating these uh green fluorescent proteins so when it makes a decision so all the genetic regulatory network that creates a hair cell gets activated and this activates this gene that produces the green fluorescent protein so it's becoming green there and now you see that the cell went on and divided so now there's two cells there's one there and one there and you see sort of the division between the two cells I hope you can see the video um clearly and then this says go on and do these rearrangements that is this is what Anna I think Anna told you about the in her in her seminar yes yes so you're gonna see uh the cells sort of move and rearrange and then they go their separate ways and now what they're gonna do is they're gonna form this structure at the top the hair bundle so now we're gonna move the focal plane up um so that we can see this happening and you're gonna see the her bundles of all the other cells that are already mature and then you're gonna see the her bundles of these newly formed cells sort of starting to form they becoming you see now they're becoming sort of more prominent uh because they're uh collecting more of these green fluorescent protein and initially they're unpolarized but then they sort of they break the symmetry and they form these polarized structure and what you see is that these two bundles are facing each other okay so this whole process of division starting from stem cell creates two cells which are oppositely polarized okay so the question now becomes how do they coordinate these so how does one cell know that it has to point towards the tail of the fish and the other cell know okay my sister is pointing towards the tail so now i'm going to point towards the head okay so and this is um what i'm going to talk about for the rest of the talk uh but first let me tell you a little bit more about how these cells know essentially head from tail um what is known is that they express a set of proteins one of which is the set this protein called vangle two and these proteins form um part of what is called the the planar cell polarity pathway so there's a bunch of cells that get polarized on the surface of the of the cell in either one side or the other and and these um in this case for example vangle two is always polarized on the side that faces so the posterior end so the the tail of the fish okay so this is how the cells know have this global order that says okay the tail is one side the head is the other side okay but all the cells have the same pattern of expression of these proteins okay so this is not what makes a decision if they're going to point away from this pattern of expression or in the same direction okay and it's uh been shown recently that there is um another protein called emx2 which is a transcription factor and with this protein is expressed it's known that the cells face so the the bundle the her the the her bundle or the the kind of ceiling of the cells faces the same direction as vangle two and when it's not expressed it faces the opposite direction okay and we know so the cells come in equal numbers of what i i'm going to call them from now on codon the ones that point towards the tail and roster the ones at point of the the head so there's equal numbers of codon and rostered cells the codon says always express emx2 the rostered cells don't express emx2 okay and we know that this is sort of a master switch because um if we over express so if we flood all the cells with emx2 so we create a transgenic fish where all the cells express emx2 then all of them point in the codon direction all of them point towards the tail and if we mutate the gene so that the protein is no longer present all of them point towards the head of the fish okay so this seems to be a necessary condition so the presence of emx2 seems to be a necessary condition for the cells to be codon and when it's not present the cells are rostered but this doesn't tell us much so this only changed the question from a morphology question of the cell so which way the cells are pointing to a genetic question essentially when do they express emx2 or not but it doesn't tell us how the cells coordinate with each other and to understand how this happens we need to look at what is called the notch signaling pathway okay so this path this pathway is a set of proteins that the cells use to talk to each other and they do this in a sort of a first neighbor way okay so these proteins the cells the these proteins are expressed on the surface of the cells and the cells need to be in contact with each other in physical contact with each other so that they can these proteins can interact and can relate signals to the nucleus and the way that they do it is by expressing what are called the notch receptors so these are the proteins that give the the pathway its name and delta the corresponding ligands called delta so when delta of one cell meets notch of a neighboring cell there's a there's a conformational change in the protein that makes that the the internal part of notch so there's a part of notch that is inside the cell this part of notch gets cut so gets cleaved by a by an enzyme and this part called the notch intracellular domain or NICD for short can then migrate to the nucleus where it can activate or replace the production of many genes and in particular it activates its own production so the when you get NICD you activate the production of notch and it and it inhibits the production of delta okay so having a lot of notch makes the cell produce less delta okay and this has consequences but for the way in which the cells communicate with each other that I'm going to tell you about in a in in a second okay if they they're not very obvious now they'll become obvious when I show you equations but first let me tell you sort of a little bit about how this story started was so we realized that this pathway was involved in the setting of the polarity of the cells because we could use a chemical inhibitor called the APT that what it does is that it prevents the it blocks the enzyme that cuts notch and sends the signal to the nucleus so when you do this the enzyme is blocked so and notch and delta meet but there's no release of the notch intracellular domain to the nucleus so the communication is essentially cut the cells are are essentially interacting with each other but there's no signal sent to the to the nucleus and because there's no signals sent to the nucleus there's no activation or inhibition of anything so nothing can happen in the in the cell well it what sorry what can happen is that the notch intracellular domain gets depleted so you don't produce anymore because you're not cutting it and then you produce then the cell becomes low on notch and produces a lot of delta okay so that's essentially the the idea of what would happen and therefore these cells that presumably will have low levels of notch intracellular domain because nothing is being cleaved now nothing is being cut and sent to the nucleus if you treat the the the fish with this chemical inhibitor what you see is that these cells now have a bias towards the rostrum tie you get more of these cells that have a rostrum direction than cells of the coda direction okay so the idea here is okay you block the communication you block the production of the notch intracellular domain you get one polarity okay so now our rationale was okay what if we make a lot more of this notch intracellular domain we should see the opposite okay the idea is low notch intracellular domain gives you one direction high-notch intracellular domain gives you the opposite direction so what we did is we created a line of fish a transgenic line of fish so you don't need to understand the details of this but what this says essentially is that the hair cells and only the hair cells produce a lot of this notch intracellular domain that is tagged with the protein that then we can go and find okay this protein is called mech so essentially what you do is it's just turn switch that turns on the production of of nicd and the cell doesn't care about what notch is doing or what delta is doing or where anything else is doing it's just stuck in this state in which produces a lot of nicd okay and then we can go and look for this nicd by looking for this protein mech which is plot here in in cyan so you see all these cells that express this extra nicd and when you count the polarity of the cells that produce nicd you see that unexpectedly so against what our intuition was telling us these cells are also rostered polarized and they're even more rostered polarized than in the case of the the apt treatment okay so the the idea here is okay we block the communication between the cells we block the production of nicd and the cells are rostered polarized and now we flood the cells with nicd and they're also rostered polarized and even more so this was very confusing we didn't understand what was happening and this is sort of when I decided okay we need some modeling because it's clear that in this system which is a non-linear feedback between the two cells there might be things that are not so intuitive that are happening and we're just not seeing because we're thinking in this sort of simplified linear terms our other system so I went and looked for in the literature and the notch signaling pathways a pathway that is relatively so considering the state of the art for most of most of the signaling pathways is where it's been modeled in quite a lot of detail so I looked for some some models of your signaling and I adapted them for this situation in which you have these two cells interacting with each other okay so what these equations tells tell you is you have equations for the concentration of notch of delta and the notch interstellar domain for cell one and cell two and then you have this term here that shaded term shaded in blue represents the production of notch by the binding of notch interstellar domain okay so not the not just a domain activates the production of notch and inhibits the production of delta okay then we have this term that is called cis inhibition of notch and delta and is represented by this picture here so it can happen that notch of one cell binds to delta of the same cell but it's known that when this happens both proteins get degraded but there's no production of notch interstellar domain okay so this is just a lost term so the you just lose a molecule of notch and a molecule of delta and you don't gain anything here for the notch interstellar domain and then you have this term which is the transactivation it's called transactivation so notch of one cell binds to delta of the opposite cell and this now creates a molecule of the notch interstellar domain and then we assume that these molecules can all degrade at the sort of a constant rate okay and so now we can try to use these to understand what happens in our system so we can use these first to get a simplified picture of this positive feedback between the cells so essentially what happens is that this creates a winner takes all situation with a positive feedback in which if if you start from random initial conditions where and notch of two both cells is close to zero but you have fluctuations the cell that has more notch will produce less delta and therefore will force the other cell to produce less notch because this cell won't be able to cleave because it it doesn't have anyone it doesn't have enough delta to cleave against so because it can cleave it doesn't it doesn't release notch interstellar domain so it can produce notch and it produces more delta and because it produces more delta than this cell can cleave more of its notch and sort of repeat the circle okay so this is a winner takes all situation in which the cell that because of the random fluctuations has a little bit more notch initially it will suppress the other cell and it will win the competition so this is what you see here and this time traces of a simulation of the the equations that I showed you before so the sets start with very minimal amounts of notch and then in this case the blue cell that cell number one wins and goes to steady state and suppresses the production of notch in cell number two while at the same time activating the production of delta in cell number two and this sort of this gets leads to the release of a lot of not in terms of domain in cell one and non in cell two okay so so essentially yeah what you have is at the cells can in this way um you get a very stable steady state in which fluctuations will take you to a situation where cell one or cell two are high notch um and so and vice versa uh and these sort of polarizes the cells spontaneously in in these two in these two states and you can probably and already see how this can be used for the cells to make polarity decisions in a robust way okay so now if you have the level of nicd for cell one or cell two at the steady state then the cells have an activation threshold if you're above the activation threshold then you get one polarity and if you're below the activation threshold you get the opposite polarity and at this point uh we don't know which polarity corresponds to uh so the model is degenerate in this case we don't know if high notch corresponds to a robust polarity or a causal polarity but we can sort of backtrack and go back to the experiment and we know and so we can look at what happens when we overexpress the notch intracellular domain so in this case this um is equivalent in our equations to add a source term to the production of um the notch intracellular domain um so we have a constant term to the equations and we can sort of look at the bifurcation diagram for what happens if we increase the rate of production so what you see is obviously that the amount of notch intracellular domain increases for both cells and what happens is that you uh for even small rates you very soon move cell number two above the threshold okay and then you get even to a point where the viscability is completely destroyed and all the system is just dominated by this term and you just get both cells having high levels of uh notch intracellular domain but what it means is that very rapidly you go to a system in which the cells are flooded and both will have the same polarity and we know from the experiments that the cells that overexpress uh the notch intracellular domain are mostly rostered polarized okay uh so this would mean that a high amount of notch intracellular domain means a rostered polarity okay so now we can go back and try to think what happens in the situation of um the notch inhibitor so why now when we inhibit um the the cleavage of notch um why do we get these also these rostered bias okay so what we have here is bifurcation diagram in which I'm changing this um the cleavage rate constant kt okay kt here which is a so the constant for the term that uh describes the interaction of notch of cell one with delta of cell two and the production of the notch intracellular domain so if we add this inhibitor what it's effectively doing is lowering the value of kt okay so lower values of kt means we have more of this inhibitor and if we look at the bifurcation diagram what happens is that for no inhibitor so the wild type values of kt we have this polarized uh bisable steady state where cell one is high cell two is low or the opposite in which cell two is high cell one is low but then when we increase the um uh the amount of inhibitor or decrease the value of kt there's this middle branch this symmetric middle branch that appears in which both cells have the same amount of notch intracellular domain but initially this branch is above the um the decision threshold so this means that these two cells although they have an intermediate value of nicd this intermediate value is enough to put them in the rostral configuration so this will give you an excess of rostral cells and then the model says that if you keep moving so if you keep increasing the amount of inhibitor you should reach a point where the this middle branch is below and all your cells should be colored polarized so at this point when we have these results uh these theoretical results uh we couldn't increase the amount of the this dpt inhibitor in the cells because it uh was little for the fish okay so we were sort of stuck in this area and if we try to use more the apt the cells with the fish will die and we would see no cells so there's no bias but the thing we realize is that we can go to this end of the uh of the diagram where there's no nicd okay and there's no notch at all so this equation is zero there's no notch there's no nicd so everything is zero in this system but this says that if you have zero nicd you should have all the cells being um colored polarized and the way we could do this is by looking at notch mutants okay so these fish have the gene that produces a notch protein is mutated so the protein is no non-functional it's not there so this seed right at this end of the bifurcation diagram and what you can see is that all the cells point in the color direction and we can count these cells and see that like a hundred percent of them so this doesn't add up to a hundred percent because there's some young cells that we can't score but all the cells that we can see are all colored polarized you don't see a single cell that is rostral polarized and so this gave us sort of the idea that we were on the right track but it would be nice if we could see sort of this transition going from an even number of cells to a rostral excess of cells to a colored excess of cells so we um look uh in the literature and we found that there is another inhibitor uh of the uh of this cleavage of notch and delta that is called ly44175 or ly for short um that uh that is it has a much higher affinity and less side effects than the APT so we could try to use it to see if we can sort of uh um span a longer range of the bifurcation diagram so we started by our control so these controls have only this molecule called dmso that is just a solvent and when you start so there's nothing there so you see the wild type you have 50 percent of cells uh colored 50 percent rostral and then if you add a little bit of this ly chemical you see a rostral bias if you keep increasing the concentration you see now that the middle branch here is sort of at the level of um the the decision threshold so like random fluctuations will take you above or below with equal probability so you see uh an equal number of colored rostral cells and if you increase it a bit more then you flip the ratio and then you get cells that are colored polarized um that have a colored bias okay so um and we can go much further because yeah again it starts becoming toxic and just for comparison this is sort of the range where we think uh the APT is in this middle range where uh you get again an excess of rostral cells like it's like it happens here um so but in this way we can sort of transverse the whole bifurcation diagram and see where and see this flip that we predicted with the theory so um there was a one more question that we wanted to answer which is that this is very nice i mean it shows that notch is involved in setting these uh these bias but there's a question of whether this is really um spontaneous in the sense that the cells initially are really equal and it's just random fluctuations that give you one fate or the other or the other possibility is that there's something called asymmetric cell division that might happen in which one of the cells preferentially inherits something that makes it already different and then this notch partway is slowly amplifying those initial differences okay um so to test these um the hypothesis is that if you have asymmetric cell division so the cells are already predestined to become either codot or rostrat and you kill them at random so randomly you kill one of them and you let the other one survive what's going to happen is that you're going to have because you don't know which cell you're killing and this so you can kill with equal probability either the rostrat or the codot cell and what you would get is with equal probability so 50 50 uh percent of the surviving cell being codot or rostrat whereas if the division is symmetric and this is just a purely dynamical process that is amplifying stochastic fluctuations and that is uh yeah and that so and it's a dynamical process that depends on the interaction between the cells then if you kill one of the cells the surviving cell doesn't have a partner to signal with and because it doesn't have a partner then it won't clip any notch um and because it doesn't clip any notch it will have a low level of notch and therefore will with higher probability acquire a codot state than a rostrat state okay and the other thing that this experiment would allow us to do is that if we kill the cells at um uh different times after the division we can see how long do they need to be in interaction with each other to um to adopt a fate that you can change okay so with uh three very nice summer students uh that I had last year uh we did these experiments and um so what you're seeing in this movie is a newly formed cell here that is going to go and divide and then you'll see one of the two newly formed cells uh disappear so you see now it goes pop and it disappears and now the same thing is going to happen up here so you're going to see uh another cell sort of dividing and then you see one of the cells disappear and then now we can again move up and see the um the apical surfaces of the cells and you're going to see how this cell and these other cells here form their hair bundles and these bundles will polarize and will point in uh given direction and what you see is after they're finished is at both points so these bundles and these bundles here they're both codot polarized okay and these we can do this systematically repeat the the experiment many many times and this again is a great work from uh um my summer students Ashley um Isabel and uh and Isabella um and you can see sort of that depending on how long do you wait uh to kill the cell initially you get all the cells pointing in the codot direction and then you start getting more and more that point in the roster direction meaning so we can sort of count how many so if you kill them within the first half an hour a hundred percent of them are going to be codot polarized and then if you wait a little bit more you go to a situation in which the decision has already been made and now you have equal probability of killing a cell that is at the roster of the codot so this uh probability goes down to 50 percent okay which is sort of the uh as I explained before is a situation which the decision has already been made and you don't know which cell you're killing so you kill them with equal probability so in this way what we show is that it takes sort of about half an hour to an hour for the cells to make this decision and and if you kill the cells before that the system always defaults to this sort of low notch state because you don't have a partner to signal with so they go to this low notch state in which they're they become codot polarized so um let me go back now sort of to the beginning of the talk I mentioned that there's this transcription factor called emx2 that is responsible for um uh the the polarity of the cells so what we see is that the the notch interstellar domain inhibits the production of emx2 so in the sense that you have high notch you have low emx2 um and therefore this is what produces these uh roster bias and in the sense that have no notch no notch interstellar domain you have a lot of emx2 which produces the codot bias okay so this can uh and we know also that uh if we look at the expression of delta uh we know that delta is expressed earlier in the cells than emx2 so it seems like notch comes first and then it makes this um uh positive feedback between the two cells it polarizes into a high notch state or low notch state and then this directs emx2 to be either high or low in um the cells so the idea here is that notch just inhibits emx2 we can sort of write this in the form of equations we can write for we can do an adiabatic approximation we don't need to write the dynamics of of emx2 we can just say okay emx2 is just um under the control of what nicd is doing and a high nicd cell will be low in emx2 uh sort of low low in emx2 and this will make the cell being codot polarized so this is all um seems all nice and fine but there's a prediction of this model that says that now if i if i upregulate the production of nicd so if i overexpress nicd but at the same time i overexpress emx2 the cells should only care what happened about what's happening with emx2 okay because i'm just setting this i'm setting a switch that floods the cell with emx2 so what the nicd is doing i don't care so all the cells should have a lot of emx2 and therefore they all should be codot polarized okay so because we want to make our lives lives hard we went and did this experiment and we overexpressed nicd and emx2 involved cells and the expectation here was okay emx2 is going to dominate all the cells are going to be codot polarized but what you see is that most of the cells are actually rostered polarized which um is sort of what is similar to what happens when you overexpress nicd okay so this means that the the the situation is not as simple as not regulating emx2 and that's it but there's some other interaction that we we are not aware of that is happening okay so after playing around with a lot of different models the more parsimonious explanation we could find is that nicd does not regulate emx2 because this is something we see in the experiment but then both nicd and emx2 compete for the regulation of a yet unknown gene or set of genes so something that is the actual polarity effector this is what regulates the actual polarity of the cells okay so now if this gene is has a high expression then the cells will be one polarity and if it has a low expression will be the opposite polarity okay and sort of these are the equations that we came up with that regulate the the expression of this putative gene so in this case we chose the model is to generate on on the signs of these the regulation by nicd or emx2 they just need to be opposite and what we decided here is to use a situation in which emx2 app regulates the production of the polarity effector and nicd down regulates the production of the polarity effector okay and they compete for this regulation so now we can simulate these equations and see what happens so in the wild type what we will have is we have the so this is the state diagram for the steady state concentrations of emx2 and nicd and i run a hundred simulations changing the values of the parameters in the equations by 10 in each so randomly by up to 10 in each so in the in the case of the wild type we have these cells that are low in nicd and high in emx2 and because of that they're above so this is the the line of the polarity effector that if you're above you are called if you're below your roster okay so this line we decided as being the half value of this p0 constant okay that you can calculate this half value as being so this is the shape of this line in the emx2 nicd space okay so these cells have a concentration that is higher than p0 over 2 so they become colored polarized these cells have high nicd and low emx2 and they become roster polarized and these in this way so this is a wild type there's so much to say here you get 50 percent of the cells colored 50 percent of the cells roster okay but now we can see what happens if we overexpress only emx2 what happens is that you push these two states you push them up okay you're just adding a constant to both state states and you push them up so now most of the cells regardless of what happens with nicd most of the cells are high in emx2 and therefore they're above so emx2 wins this competition and activates the polarity effector more than what nicd can inhibit and then you're above the the decision line and you become colored polarized okay if you do the same now and activate nicd over activate nicd so all the cells become very high in nicd nicd represses completely emx2 so you have very very low levels of emx2 and therefore nicd represses the polarity effector and it all the cells are below the line and then you have a roster bias okay and now what happens if we combine these two results and activate nicd and emx2 at the same time so what what happens now is that you push the cells sort of diagonally on the diagram and now there's this competition where depending on whether the cells express a little bit more nicd they will become roster polarized if they express a little bit more emx2 they will become colored polarized polarized but there's this double inhibition of nicd to the polarity effector okay so nicd can inhibit the polarity effector directly or it can inhibit the polarity effector production by inhibiting emx2 and preventing it from activating the polarity effector so more often than not nicd wins and the cells have a tendency to be rostered polarized instead of colored polarized okay so just to finish the talk I want to connect a little bit with the talk of Ana and tell you essentially where this story sort of lands to where what Ana was telling you in her talk and is that this as I mentioned before this is an unbiased mechanism that amplifies stochastic fluctuations and creates a robustly creates a cell that is colored polarized and a cell that is rostered polarized but this can happen in two possible configurations okay so you can go to a situation in which the colored polarized cell is to the left or the colored polarized cell is to the right and in this case if you look at the cells at the bundles in this case the bundles will be facing each other in this case the bundles will be facing away from each other okay but if you look at a neuromast you never see the situation in which the her bundles look away from each other okay and this is because as Ana explained during her talk these cells undergo these rearrangements so if you see the cells dividing here when they in the wrong configuration where they would be looking away from each other they there's this coupling between the internal not state of the cells and the direction in which they're going to move that creates these rearrangements so when the cells are trapped in the wrong configuration these um uh polarity directed movements will make them flip and they will put them in the sort of right configuration that is the the one that you uh see always uh which is the one where the bundles are facing each other and we see that these rearrangements happen 50% of the time which agrees with the idea that this is this is a completely unbiased mechanism and 50% of the time you get it wrong and then you need to rearrange the cells to get them in the in the right configuration and and these sort of just to summarize and sort of connect with what Ana said in her talk is that you have this biochemical step that gives you this sort of double well um potential in which the cells can be either in the sort of positive or negative dipole but then there's this mechanical step that makes the negative dipole sort of flip and form a positive dipole and sort of these consistently and robustly creates uh this organization uh of the neuromast and with that I would like to thank uh Jim my mentor for all his support, Agnique and Ana for being great collaborators and all this work that we did in these two papers that described sort of first the biochemical step and then in detail the rearrangements and Kim who's a grad student who was in the lab that started sort of this project with me and she taught me all I know about dealing with syrophage and doing all this this imaging and uh my funding sources and thank you for your attention. Thank you very much Edrin, very interesting talk how mathematical biology can help explain what is going on with the very naive predictions right? I invite the audience to ask questions please you can directly connect to their audio and to unmute your microphone and ask questions directly you can also ask in the chat if you have any issues with your microphones and okay while someone comes up with an idea I will have a question okay um you so just in the end of your talk you discussed this idea that there is a putative uh intermediate step right with the putative gene that there is a intermediate step which determines if PMX or NICD would dominate and lead to the determination of the final polarity. So my question is so this putative gene comes up to fix this prediction that follows from a less elaborate model but have you thought of any alternative so if it's not a putative gene right what would be so if putative gene is hypothesis zero then what would be the hypothesis one that it is not? I mean there's there's a possibility that there's some complicated interaction just between NICD and EMX2 that we don't understand. I'm a bit sort of hesitant to think that's the case so I try a lot of different models where you don't assume any other variables and it's very hard to represent you need to do like very contrived things are just the parameters in very narrow ranges to get all these results to be able to reproduce all the results that we know so what happens when you when you have a wild type when you overexpress one of the genes the other gene when you do chemical chemical inhibition and things like that so it always breaks down at some point if you don't assume that there's something else but what you're saying there is in this diagram of the parameter space there are ranges of parameters that would reproduce your original prediction. Yes yes so if you assume like some interactions between EMX2 and NICD you can find some terms in the equations and like some ranges of parameters where you can see it but they're not very robust so so the problem is that they will work for like finally two parameters but when you try to do something like what I'm doing here which is like you just do a robustness analysis and just just vary the parameters by let's say 10 percent or something like that you allow them to vary and just do random simulations with the parameters you you can reproduce all the the proportions of cells that you see yeah I see yeah all right and well and so suppose so that alternative model would introduce some new interactions between NICD or EMX2 but what other ideas other any other ideas except these two just in case you thought of anything? No I think so I mean that's I would say the most obvious is either there's some interaction between the genes themselves that form the the sort of the the regulation the regulatory motive or there's some extra genes that we that we don't know yeah that would be sort of the more parsimonious explanations I can think of yeah okay thank you let me then ask a question from the audience so Luciana Bruno asks so she tells that it was a very nice talk thank you so the question does the lateral inhibition occurs between pairs of cells or it can be involved more than or it can involve more than two neighboring cells um so we think it only involves the these two cells because the the cells um let me show you a little bit uh maybe here we go um so if you see for in the movie uh in the in the initial movie that I show how the the cells um sort of appear they sort of the cells are created in pairs and these pairs are um sort of they appear in different parts of the of the neuromast so there's never like two pairs of young cells that are uh close to each other and it's only these cells when they're in this immature state that express the uh proteins of the notch pathway okay so if you look at for example um the expression of of the delta which is one of the few that we can see so this is uh sort of kind of problematic to see these proteins uh but uh delta we we can sort of see uh in snapshots and when you look at it it only appears in this immature cells that means if I I think I have a picture here where we uh quantify this um so I can show you yes um so these are uh two young cells and you see that these are sort of expressing delta there's a little bit more delta in these cells that are um a little bit older but they're sort of becoming already um mature and they're down regulating delta and then you don't see it anymore so so we think that there's like this this pulse of signaling that happens when the cells are young and because these divisions are sort of sparse and spread out they only interact with each other and then the other thing is that they seem to move so the this organ is three-dimensional and the cells seem to move sort of down first to do this interaction and then move up again and we think that they do this also to avoid crosstalk with other cells and so we are pretty certain that this is just uh um at like only two cells interacting with each other all right thank you I hope this uh this clarifies yeah so the author of the question uh agrees that it clarifies the question um are there someone else in the audience I've seen yes of course okay um so can you go back to the face diagram at the of the model lateral inhibition uh yes yeah so if I understand there's a there's a subcritical pitch fork here right yes now and I understand it's not easy to do the experiment it was just to make clear so you moved away from the from the bifurcation point is there any chance you could do an experiment where you see the the bifurcation itself for example by hysteretic effects you go up and then you show that the system has hysteresis um it's it's quite hard yeah these experiments are um uh because you are the chemical inhibitor so this sort of disrupts the fish um so uh I mean one could I guess I yeah I could think sort of of a way maybe you can add count the cells and then if you're lucky to be able to keep the fish alive after you count the cells uh reduce the amount and then see it again it would be it would be quite hard but I guess it's it's an interesting question because uh depending on also on the parameters of the model the pitfall could be subcritical or supercritical and there's there's nothing to say um that it needs to be one way or the other and yeah maybe yeah I I think maybe we need better tools to um to do it I think we with the way that we do the experiments right now would be very hard to see because the experiments are very noisy you see that the differences uh they're not super high also and you need to and to get sort of down to these differences you need to count a lot of cells well but I would I would kick the system up to to the upper branch for example and then the differences will be high right yes yeah so you could um well but that's so that's the the thing so it's hard to yeah that's a hard part because it you I mean this uh for example here when when you go to this part um presumably there should be a range that you should be able to reach with a chemical inhibitor where you see all the cells of the same polarity but the system is very noisy in itself and when you are it's not super clear to me also that all the cells see the same concentration or react in the same way to the concentration so what you're seeing sort of uh in the experiments is uh either a combination of a lot of these bifurcation diagrams with slightly different parameters for each of each pair of cells or some dynamical noise also that we don't understand so it's I think it's it's a really great question and I thought about it a lot and I'm not super sure I think it would the experiments will require a lot of time and a lot of effort to do and I'm not entirely sure we will see the answer we want at the end and if we don't see it I'm not sure we can conclude anything because the noise we don't have a lot of control on the noise of the on the system so we don't know where is it coming from and why it's obscuring the the results yeah understood thanks thank you for the question all right I actually seen that other some other people in the audience tried to connect the microphone so if you are still around yeah there's a question from Valeria also all right uh yeah okay so the question from Valeria great talk could support cells have a role in this polarity decision yes and so we yeah we think they don't and because again the the support cells the the support cells do use the notch pathway at some point and because this is the pathway that they use to decide whether they're going to become herds or support cells so there's a possibility that there's some crosstalk there but the expression of notch in the support cells is it seems to be biased in certain so they express more notch in certain parts of the near mass than others and we never see a bias of the of the polarity depending on where the herds appear okay so that seems to suggest that the the herds are doing their own thing and they're not talking to the supporting cells and we think they do this again as I was saying before because they move sort of down in the organ they move more basically in the organ they do this communication during a short period of time and then they move up and what we think it might be happening is that there is some spatial compartmentalization of the signal so that the support cells are talking sort of on the apical part of the organ and the herds are stopped more basically and and then this is the way they avoid interaction or the other possibilities that they are using different variants of the notch proteins to do this or and the delta ligands okay that let's say the herds use notch one and the support cells use notch three and this in this way they avoid the interaction so every so all the evidence we have which is not a lot seems to suggest that the herds us do this in isolation but I couldn't sort of completely discard that the support cells don't have a role and this is something I want to study in the future but we don't have the right tools to do it right now so this is something I'm working on to try to get at you are muted Roman thank you thank you very much for the question yes and we we may also we may still answer any other questions if there are well right I think that yeah we already over there