 All right welcome everybody. It is my great pleasure today to introduce Barry Honig to various somebody that I met very well I I met him through his papers almost immediately When I entered this field and met him shortly thereafter He's made enormous contributions, especially to the field of protein electrostatics and implicit solvent models for proteins and macromolecules Today, he'll be talking to you about some newer work so Barry is a Investigator of the how are you Howard Hughes Medical Institute and a professor at Columbia University in Molecular biophysics Systems Biology and Medicine. He's the director of the computational biology and bioinformatics Institute there And he has numerous awards. I almost feel like I should save half of this for tomorrow because there's such a long list here, but he has the Hollander award in Hollander award in biophysics from the National Academy of Sciences and is also a member of the National Academy of Sciences and The American Academy of Arts and Sciences He has the Founders Award from the Biophysical Society the Delano Award for computational bio Computational biosciences and the Anfanson Award from the Protein Society and honestly I can't think of any other possible award that well maybe just one but you know any other possible award that somebody could win in computational structural biology and And very also I asked him for a couple of hobbies And he let me know that he is an avid scuba diver and also wanted me to let you know that he played a year of football at the University of Florida before he went to Went to Brooklyn to get his his degree in chemistry and then on to the Weizmann Institute for PhD So if you'll please help me in welcoming Barry Honig as our Steenbach lecture Thank you, Julie I Heard for many years about the Steenbach lectures one of these things you don't actually think you'll ever be one I'm really excited and delighted by the honor and thank you for attending. I this is the title of this talk and It's it represents an area. I began working on about 15 years ago and I say right here with Larry Shapiro because when I became I've done computational work all of my life, but When I became a user investigator, I was able to expand my interest to do experimental work as well and Nobody would ever fund me to do experimental work And but Larry's a accomplished crystallographer and together we set up the joint lab And I think it's perhaps a way that some modern science can be can be done We have a single experimental lab and actually share computational students as well. We do crystallography Everything that goes along with that cell assays biophysical measurements and the people working in the lab Aren't even sure who pays their salary. It's it's really a joint effort. So All of my work on adhesion is done with Larry. We publish everything together and I Just want to start with that. I also wanted in part to advertise Tomorrow's lecture in part because I sent the wrong title There was a typo that I never caught but there's a fundamental difference in what I'll be talking about today and tomorrow Today is I'd say more traditional biophysics. There's some cell biology in it It's really understanding how molecules work in a biological context Tomorrow is I'm going to be talking about the use of three-dimensional structure to Basically to predict what proteins interact as it says on a genome-wide scale It's sort of low-resolution structure biology today will be high resolution structural biology and the differences are are quite dramatic So the sorts of problems we're interested in Are summarized here and I'll be getting to some of them But the basic question is can we use the information about protein structure and protein protein interactions to understand how cells Organize how they interact and this is a picture or diagram of motor neuron pools there are different motor neurons that connect to muscle and They separate into into various pools as you see here dependent on the proteins that appear in them that are expressed in them And these are the proteins that I'll be saying a lot more about that These are proteins called nectons I'll be discussing that or play a role in organizing the inner ear and in many other roles and in both cases I'll be trying to try to convince you that there's a clear connection between molecular and cellular properties And then the second half of my talk will be devoted to some very new work Which involves the how protein protein interactions play a role in a specific problem in the design of the new in the wiring of the Nervous system particular how neurons know when to form synapses and when to avoid each other So that will be the second part of the talk, but they'll all be based on common families of proteins So the proteins that we're interested in and these are generally how adhesion proteins are organized This is supposed to represent a cell membrane. They're multi-domain proteins I'll be talking first about what are called classical cat herons. They have five immunoglobulin like domains and a Trans-membrane region and a site of plasmid region. I'll be talking about proto cat herons nectons They all look like this though multi-domain proteins With some connection to the cytoplasm This shows in the case of classical cat herons what that connection sort of looks like there's a Unstructured region the outside the extracellular domain is structured the cytoplasmic domain is not but it forms complexes with beta-catenins alpha-catenin and ultimately links to the actin site of skeleton Almost there in fact everything. I'll be talking about today involves Recognition which takes place outside the cell. I won't be talking about the various Cosmic processes that are initiated by recognition. So the basic question is how does protein recognition? translate into cellular recognition so what got me into this field was 15 years ago or so when Larry actually who's in our department Gave a talk in my group meeting trying to get my lab interested in the following sort of problem. This is a Image of the developing chick embryo after six days. This is the ectoderm and This shows that the cells here express a protein called ecad here an e for epithelial And you can see that these cells are separate from this inner set of cells Which form the neural tube and these proteins express these cells express N ket here and N for neuronal and It has been they were discovered by Takaichi quite a few years ago And the basic thought then was that this is very simple ecad here and doesn't stick to N ket here And so these are the thought was that they were homophilic proteins So that when cells start expressing it get here and those cells would break off from the cells creating ecad here and giving rise to this sort of early developmental process so We had at the time models, but now we have crystal structures of E and N ket here And you can imagine the membrane over here and here so these coming out from the membrane they sort of form a banana shape and Only bind at the external domain the ec1 domain and the there's a lot to say about this But I won't today the interaction is involves what's called a strand swapping You see there's a tryptophan from one protein from one monomer that's Inserted into a pocket in the other and vice versa. So they they sort of exchange the beta strand That's how they bind and the question that actually posed a graduate student at the time was Why how are they different? How is the in N ket here and different? They're about 70 percent identical and it was The fellow working on it was an empty PhD student and they all want to get their PhDs within three years So I told him that if he could tell me why they were different he would get his PhD and it could go on to do other stuff so He he basically we had structures or models of this is just the membrane distal interacting domain of e ket here and N ket here and he built a model of E N and all he had to do was tell me What was wrong with this model? why he doesn't bind to N and How these two guys are different and after a year or two He couldn't figure it out and the models looked as good as the heterophilic models looked as good as the homophilic ones And at that point I became an experimentalist So we actually used an analytical ultra centrifuge and SPR surface plasm on resonance to Measure the binding affinities of these proteins and we found a few things number one that N ket here and sticks to N Quite more strongly than e to e. This is a hundred micro molar. This is a 20 micro molar binder and E N actually bound better than e so these molecules actually did bind heterophilically and This came as some somewhat of a surprise But it sort of set us on our path after these experiments were done He did Peter Chen his name was did get his PhD. So it all worked out for him but this Gave us a different picture of these proteins in that they were heterophilic and then the question was well, why do the cells that express e and n ket here and why do they separate from each other and The answer to that we believe I'll just set up the answer I'll just point out that there are proteins I'll be talking less about that are called type 2 ket herons that are very similar, but they have a larger interface They have insert to trip defense with strand swapping and they don't bind to type 1 ket herons And the reason I tell you that it's because we did the following set of experiments So these are cells are labeled with red and green dyes And if they both express both sets of cells express n and n ket heron as you might expect they mix together because n binds to n When they express e ket heron again They mix together and this is a control with ket heron 6 of type 2 ket heron and again You see the same of the same result now if you have some cells expressing n and some cells expressing ket here and 6 They form separate separate aggregates because n doesn't stick to 6 if you have cells expressing e and Cut here and 6 again. They form separate average aggregates But when you have cells expressing e and n together they actually form separate aggregates, but they stick to each other and in a way that's Very crudely reminiscent of what one sees here that n aggregates in the middle any around it And why is that true? Because n sticks to n has a higher affinity for n than e to e So the concepts that might work at a molecular level work at a cellular level the most strongly aggregating Cells will maximize their interactions and therefore form a central aggregate The weaker ones will aggregate around the stronger ones interact with each other and with the central core so in very very crude, I know I'm going through this very sort of Superficially, but very quickly the concepts that simple thermodynamic concepts are working here at a cellular level So that's what I want to say for type about type 1 ket herons Now another family of proteins that are cell-cell recognition proteins called Nectons and These are these actually have three domains. They're immunoglobulin domains and we solve the structures of a bunch of them and they all look alike They form an interface again at the membrane distal region and when we Measured the binding affinities of Nectons that we found that their behavior is very different than classical ket herons. These are Homophilic binding affinities and they range you can see from 0.4 to what 200 or so micro molar actually a point I want to make Membrane binding proteins tend to have much weaker affinities than Soluble proteins than than cytoplasmic proteins. These numbers are actually quite quite typical Anyway, you see these are the range of affinities. You see that Necton 1 Binds to itself say more strongly than strongly than Necton 3 binds to itself But now when you do an experiment with say a Biocore apparatus where you flow Necton 3 or other Nectons On a chipped containing Necton 1 you see that 3 binds to 1 more strongly than 1 binds to 1 So these proteins are heterophilic binding proteins primarily. They're also Homophilic and what they doing all the measurements you get some kind of a matrix and binds in 1 2 and 3 and 4 Etc. And how this works is is ah So these are experiments done in our collaborators lab Sergei Troinovsky and what's nice is I'll explain this in a minute You see these molecular properties manifest at a cellular level So these red cells express and Necton 1 and the green cells express Necton 1 and you can see At the interface between these between red and red Red and green and green and green you see a build-up of proteins those proteins are diffusing to the cell-cell Interface because they stick to each other and you see what are called will call junctions from now on at all interfaces Now if you express Necton 1 with Necton 2 Necton 1 doesn't bind to Necton 2. So you only see junctions Within a single color within a single type of protein. You don't see anything in between them But now when you express Necton 1 with Necton 3 all the sudden you see the strongest junctions between the red and the green Because Necton 1 Necton 3 is the strongest binding interaction. So these proteins. It's not surprising They're diffusing they're finding their partners where they can gain the most interaction sort of free energy That's where they're going to show up. So again the properties of the individual proteins are being Reflected by their behavior on cell surfaces, maybe not surprising, but it's it's nice to see it By the way, I should say I'm happy if anybody wants to interrupt at any stage. Please please do The way these proteins are designed is very clever and very simple If you want to design a heterophilic protein put a charge there and on the protein it binds to most strongly have an opposite charge So if these two proteins dimerize you'll have this sort of red negative region interacting with another Glutamate there'll be some repulsion, but in the case of Necton 3 there's a lysine There'll be an attraction. There's a very nice way of designing Homo heterophilic proteins and there's the way that the Nectons work What's very nice at a cellular level this these are cells Which Contain either Necton 1 and Necton 3 and the and this is these cells are in the inner ear. I Admit I don't remember of what organism, but I assume it's mouse But you see they form a checkerboard pattern and when you knock out Necton 3 the checkerboard pattern basically goes away And this is exactly what you expect from heterophilic proteins if you want to max Maximize the favorable interactions, then you're going to have their heterophilic at the molecular level They're going to be heterophilic at the cellular level and that the way to optimize this is through a checkerboard so again the properties of the individual proteins are being reflected by No by the by the behavior of the cells that contain them and This is sort of I guess An example of someplace. We're going. It's just something I I like showing because I never in my life when I started As a computational chemist dreamed I'd be showing a picture of a stained kidney But this is what this is These the kidneys being stained for different types of cat here and this is called a nephron, which is a filtering unit in a kidney and Some regions of it contain ecat here and which is a type one Cat here in six, which is a type two. I already told you that Type ones and type twos don't bind to each other So what happens in between? There's a region that contains both type ones and type two cat here And so we combine in both directions so Again what at least very qualitatively, and I know I'm hand-waving a lot We can understand some of the simpler aspects of this development of a nephron based on the adhesion properties of the individual proteins We are hoping to now using CRISPR technology to design artificial kidneys By changing we can change affinities of these proteins almost that will and that's something that we're hoping to do and this Work is being done with rosemary symponia a nephrologist and a PhD at Columbia So that's a Sort of almost an introduction. There's the properties of proteins on cell surfaces are reflected in the properties of the Behavior of the cells now. What happens at the interfaces between cells? This is it's not enough to say recognition. There's complicated things going on and this is a Picture I took from a textbook central cell biology and what you see these are cells stained with ecadherin and See the ecadherin is distributed all over the surface of the cell and as soon as these cells come in contact You see basically the proteins diffusing to the interfacial region and you could see the increase in fluorescence intensity So that's that's what's going on well It turns out that in the case of classical cadherins these interfacial regions called a form eventually fixed junctions called adherence junctions and We're interested in the properties of those junctions and what that means for signaling So we when we solve the structures of E and N cadherin And also another cadherin C cadherin We found that they formed they all formed a very distinct crystalline lat looking lattice Crystal lattice in the case of the structures Where this is the trans interaction trans in the language I'll be using in this talk now means interactions of proteins in different cells sister interactions of proteins in the same cell So this is the interaction. I already showed you but in addition. There's a second interaction. You can see going This way. This is trans. This is cis You can see the lattice It's a it's a two-dimensional lattice seen in a crystal and we Suggested that maybe this lattice is what forms between cells, but before I get to that What's very interesting is when we I already told you we can measure the binding affinities of catherins And they're depending on which one 10 to 100 micromolar or so We couldn't measure the this interaction which we thought might be a crystal contact Because we could we when we just tried to measure this in solution. We couldn't measure it. So it's very weak less than one millimolar but when we Connect these Cadherins to liposomes first you see that the liposomes form a there's a there's a clearly ordered structure between the liposomes, but when we make mutants when of this Once we have the cis interaction which has some hydrophobic groups We simply replace one or two hydrophobic groups with aspartic acids. We disrupt this interaction and you lose structure So we're disrupting an interaction. That's too weak to measure and Yet we're seeing a strong effect on structure, but more interestingly this is a image of Adjunction formed in live cells Transfected with wild type e-ket here and when we mutate again. We the system interaction We no longer see aggregation or there's very little aggregation in the cell cell Contact region so again disrupting an interaction. That's too weak to measure has very significant effects on what happens in cells So how is that possible? I'm going to go through this quickly. There's a whole theoretical story, but I couldn't resist Telling you about it quickly what you have of this is I'm going to show you a simulation a minute but these are supposed to be cadherins moving on cell surfaces and If they formed a cis interaction on their own as I'll just tell you that would signal to the cell that something has happened But so they can't be allowed to form a cis interaction until they've seen another cell So there has to be a mechanism whereby this interaction becomes strengthened by interactions with other With proteins from another cell and the way we think this happens and caught myself here You have two proteins that are Connected to the surface. I'm not telling you how I do these simulations, but Have linkers that are fairly flexible between them and they move around quite a bit if they're connected to Another protein from another cell surface then they have much less conformational freedom because they're linked at both sides to a membrane in the middle to one another and See I'm not good at this And the point is these guys don't move so much once they form a connection with a protein on another cell and We believe that it's this trans interaction that strengthens the cis interaction and this is This sort of leads to the next step. I'm about to show you is The question the question we're asking is can these proteins form a two-dimensional assembly on their own of this of the sort I showed you on the cell surface and This leads to sort of all kinds of theoretical issues, but the basic point is that when we calculate or think or measure binding affinities in solution these proteins are moving in three dimensions and When proteins interact on cell surfaces, they move in two dimensions Sort of two dimension quasi two dimensions so that the affinities We measure in three dimensions if we have somehow relate them to affinities in two dimensions And we have a whole theoretical framework that we use to Calculate though to make the transformation But I just want to show you a another simulation before I get to the second part of the talk So here's here a cat here ends Green and red means different cell surfaces here When they in contact we draw them blue so there's three colors here green red and blue You don't see the blue yet and the only difference between this lattice on the left and the lattice on the right is we've made the cis interaction a little stronger Consistent with a theory. I haven't told you about and this light yellow region is a is a contact region and I just want to show you the two sets. So we're moving these proteins around whenever they form a contact This is a Monte Carlo simulation. You see something come up in blue So here on the left the cis interaction is weak the proteins diffuse to the interface, but There aren't they move out because there's nothing keeping them there here. There's some sort of a phase transition Where the stronger cis interaction the trans interactions the same in both cases? But the cis interaction is strong enough to lead to some ordered structure Which is at least consistent with a self-assembly process we think is going on on Cell surfaces so in general and this is a Point I want to take for the second part of this talk which begins now is that these proteins have the capability of forming ordered structures on cell surfaces We know that once these clusters form they they involved in signaling to the side of plasm and ultimately connect to the Act inside of skeleton We don't know if the ordered structure is important is if there's a connection between order outside and order inside Or whether order is just a way to assemble lots of proteins in one place We don't know that yet, but we are what we do know that ordered structures form that can be disrupted Experimentally by disrupting the cis interaction. So that's sort of an overview of What we've worked for many years And now I'm going to discuss a totally different problem which will relate closely nevertheless to what I've been talking about and I mean, this is a presumptuous title. How is the nervous system why I'm not going to tell you how but This tells you how many synapses and how many neurons there are in different organisms and that these Neurons wire in precise ways is certainly absolutely remarkable and it involves different features of neurons and The one I'll be concerned with is a phenomenon of self-avoidance So here's an example of a cell body in Drosophila and these are Neurites extending from the cell body and what you see is that they don't Touch each other they avoid each other And here's an example of neurons in vertebrate in mouse and you see the same self-avoidance phenomena So the question is that I'll be addressing is what is the basis of this self-avoidance? For Drosophila much of our understanding at the genetic level and the neuro biological level comes from work from No, I don't have his name here Larry's a Pereski's lab who Worked with talk and study the properties of a family of proteins called these scams down syndrome. So he's in molecule Flies don't get down syndrome, but these proteins are related to vertebrate proteins and here's what happens. This is a Drosophila neuron when you knock out this gene cluster you see now that the Neurites bind to each other So this establishes the role of these scams, which I'll be talking a little bit more about this is again work from Larry's a Pereski's lab in vertebrates This is Sort of a starburst anocrine cell Again when you knock out a gene cluster You see now the dendrites are sticking to each other and the proteins that do this are Called press clustered proto ket herons, and they're related to the ket herons. I've been talking about And what I'm going to do is briefly tell you At a molecular level how the d-scams work Again, it's a Pereski's work and then get to our own on vertebrate neurons, but there's a very very sort of counterintuitive thing going on So here's a diagram of two neurons Red and blue and they can form synapses with each other But they don't see redness binding to blue But they don't synapse within the same cell and it turns out each neuron has a unique barcode That defines it and tells it that it's different or and some other neuron, but where it where it gets counterintuitive is that the proteins both in Drosophila and Invertebrates are homophilic adhesion proteins And it's adhesion that leads to repulsion and Why and in fact it must be that way because what you want is a specific mechanism whereby two Cells will repel each other and I don't I've never found anyone who could tell me how you could make two proteins repel each other specifically We have there are lots of ways proteins can bind specifically, but how do you make them repel specifically you have positive charge negative charge There's not a lot you can do so specificity begins with adhesion and then adhesion Leads to repulsion Through activation of cytoplasmic phenomena So when you have two neurots from the same neuron they bind and then repel When they're from different neurons, they don't bind therefore they don't repel That's the counterintuitive part of the story But you need to remember that because it can be confused other I get confused all the time so in the case of these scams there are The proteins have three domains which are alternatively spiced IG 2 IG 3 and IG 7 we know that the structures of These proteins from David Eisenberg's lab and basically they IG 2 binds to 2 3 to 3 7 to 7 there's 12 48 and 33 spiced forms if you multiply out all the possibilities you get 19,000 distinct proteins and What Zaperski's lab showed is that these proteins are all strictly homophilic they only bind to each other and That leads to very interesting protein design questions How do you design 19,000 closely related proteins only to bind to each other and? Obviously nature has done it and I can Maybe isn't all that complicated. This is how they this is how they work and this leads to sort of an issue which is talked about but It's essential that we think about there aren't clear answers for these scams you have 19,000 distinct proteins and 10 to 50 isoforms per cell so what that means is through stochastic splicing Each neuron say neuron a will select 10 to 50 of the possible 19,000 choices it has Let's say in this case. They're only six To simplify the figure so this shows that there are six different isoforms chosen There may be a thousand copies of each but six different isoforms chosen Here's another near a neuron with six isoforms chosen of which two are the same The yellow and the blue so let's say they were all the same then you'd have Homophilic interaction the cells would be the same But by having this opportunity to choose six in this case from 19,000 you get differences, but then you get to a question. Well, you don't want these cells to stick to each other What if five out of six were the same? You know, what if 19 out of 20 were the same you'd expect them to bind You know the one that doesn't bind would get out of the way so to speak and one doesn't really know What the number is and the people working on drosophila haven't worked this out but it raises sort of a Term the tolerance for common isoforms how many isoforms can be different and the cells still don't bind to each other They they assume 10 20 percent But putting that aside It's clear that given a choice of 19th given a choice of 19,000 isoforms It's not that hard Intuitively to imagine Selecting a few so that the probability that any two neurons would have the same set is very small So drosophila solve the problem with 19,000 proteins Okay, without now. This is what's known today now in the case of vertebrates There's actually Gene clusters of protocadherans. There's there's an alternative Splicing going on here, but I'll turn stochastic promoter choice And there are different clusters But the point is that in the case of vertebrates and sake of mouse for example, they're only 58 proteins not 19,000 So mice accomplish Being having more complicated brains than flies mice accomplish this problem, you know solve this problem with 58 proteins So the mechanism working in drosophila can't be working in vertebrates But again, this is to remind you that if you knock out a gene cluster You lose self-avoidance. So We want to figure out I'm going to try to tell you how I think mice accomplish the same task so working with Tom maniata's lab with postdoc in his lab a to chan we first study the binding properties of these proteins and These are cells expressing Different protocadherans red, you know, and they're red and green Fluorescing cells and you can see this when if we look at the diagonal you see that All the cells aggregate together Which means that these are homophilic proteins beta 4 to beta 4 you see there's mixed aggregates Any other combination they don't bind to each other You see the set they form separate aggregates So we're really using cell aggregation as a probe of binding and what you see again as is the true for drosophila. These are Completely homophilic proteins They don't bind heterophilically But the really interesting and surprising observation we made in studying these cells is that Well, you know, let me just take you through it You have to start on there on the right the way this is organized These are cells that express three different protocadherans. Don't worry about their names and in the figure on the right All three in the red and green cells are the same and they form nice aggregates But if two are the same and one is different the cells don't don't bind to each other So one incorrect protein is getting in the way of two correct proteins Is that clear I hope I hope it isn't so there's what we call an interference mechanism You might think intuitively if you have three different classes of proteins on the cell surface What I said a few minutes ago if one of them doesn't bind It'll just diffuse away and the other two guys can bind to each other But they don't one incorrect protocadheran is enough to disrupt cell-cell contact and we've done experiments with up to five Expressed in the same cell one incorrect one interferes with binding and We believe that it's this mechanism that lies at the heart of how protocadherans work So we solve the structure I I Say we solve the structure. It's I couldn't I don't know how to solve a structure say But Larry does and we have this common lab So I always feel a little comfortable saying we we solve but I'll say it we solve the structure of protocadherans. I see this figure is not so clean we solve Structure with few of them. This is what it looks like as opposed to classical cadherans These guys form anti-parallel contacts these very similar proteins to the cadherans. I've been showing you yet They interact in a completely new way instead of forming a banana shape, they're very straight and You can see they form this anti-parallel sort of head-to-tail structure Where ec1 the membrane distal region connects to ec4? ec1423 this is how this is the interface they form But that doesn't tell us anything about interference I guess we've mapped we we understand how they bind we understand which residues lead to specificity I Won't be discussing more of that this slide is crucial So I want to take you through it We made constructs with different numbers of domains So and then assay they're binding both in a biophysical assay basically Analytical ultra centrifuge and in cell assays and it's this combination of information that taught us something very different So if we have a construct of only three domains It's a monomer in solution. It doesn't form a dimer because you need all four and it doesn't affect cell aggregation If you have a construct of four domains they form a dimer, which is what we see in the crystal structure if we Transfect cells with four domain constructs the cells aggregate nicely So that shows you that the four domains that we see in the crystal structure are enough to aggregate cells And we do that with large larger constructs as well The critical observation is this when we cut out. Well, we move EC one so we have a construct with two to six it forms a dimer, but it doesn't aggregate cells and From this and other evidence we know that this forms a sys dimer So sys dimer on its own. We've killed the trans interaction. We've only left the sys interaction and See it doesn't aggregate cells it forms a dimer when we have full constructs in solution We already know what aggregate cells and we know it forms a tetramer So this series of experiments tells us that the recognition unit of a protocad heron is a sys dimer which then interacts homophilically with proteins on the other cell surface and There's two ways this sys dimer could interact could Could form a tetramer and solution it certainly this is this is what happens in solution But there's also a possibility of some sort of a zipper Which I'll show you in a second where it forms an extended assembly Which I'll get to in a second, but the basic notion then is that this is what happens This cartoon is what happens on cell surfaces. We know that the sys interaction I didn't tell you this is is promiscuous any protocad here and can interact with any protocad here and in sys So if you have different isoforms shown here in different colors, you're going to get a combination You're going to get some statistical distribution of dimers But the way to maximize interactions is see some of these dimers are heterophilic like this one here It's red and blue the sys is the same, but they're different trans so this this If you start with red to reds it can bind to a red and a blue which can find another blue and in this way You can get maximum number of interactions between cells and This is at least consistent with the notions. I've been talking about a system trans interactions and lattice assembly So what's nice is that this gives us a mechanism for self-avoidance. So let's pretend That we start with I'm just going to show you how lattice might grow red and blue Red blue green If all the isoforms on both cells are the same any protein can always find a partner of another color Because it's going to be there some in some dimer and therefore these aggregates can grow at will But now let's say we have a single mismatch When you start trying to grow chain you quickly get to a situation where this green has no partner on the other cell and the Assembly stops sort of a chain termination mechanism So if we if you actually calculate which we did here The size of it the average size of an assembly. This is a Monte Carlo simulation Let's say there are a thousand copies per cell shown here in black The assembly if if there are no this is the number of mismatches if there are no mismatches The assembly comes up to be this basically related to the size to the number of proteins if there's even a single mismatch then the size of the assembly goes down to What is this 50 or so and then the decrease is very sharply after that so The argument we're making the suggestion we're making is that this effective poisoning of this assembly is The recognition mechanisms how cells tell Self from non-self it it translates into the size of the assembly they form which we would then Suggest effects what happens in the cytoplasm That the model is consistent with the with all the data at least it explains this Remarkable ability of 58 proteins to do the job of 19,000 proteins As shown here these scams have 19,000 they basically function by Having this great molecular diversity We think at least we're suggesting that protocol adherence function with an interference mechanism Which allows them to generate the even greater diversity by the way than protocol adherence Do we're obviously in the process of testing these ideas and there are number of ways to do it But the critical connection then between binding and repulsion remember ultimately binding leads to repulsion is Is is we're claiming in the size of the assembly? We'll see if that's true or not This is just to sort of want to summarize that From point of view of protein design it's sort of remarkable that these proteins there are other Conforms so many different interfaces and it's just one of the wonders of nature that this that this can happen I'm not going to be discussing them So let me just finish by mentioning sort of key people I say we have a a large lab and lots of Wonderful collaborators. We have people doing solution biophysics Computer simulations the protocad here and work the people in red worked on the protocad here and wrote them Rubenstein Carrie Goodman did the crystallography large protein production group the Simulations were done together without being no one been shown in the Hebrew University in Jerusalem and the protocad here And work is done in collaboration with the money out of slab and I briefly showed you some Kidney stains from Rosemary Symponia's lab. So thank you very much for your attention Mechanism So it may well be we don't yet know how many isoforms are expressed we know it's on the order of 10 to in Amicron cells is I think it's 15, but we don't really know the actual number. There may be more isoforms in Invert in invertebrates. We're not sure yet Well, so you have this is where there's there's a random selection of different proteins that takes place at the RNA level or at the DNA level so That's the mechanism for choosing these different proteins is is these different isoforms is unclear, but The point is in the same cell there's no problem because they're all the same They're dendrites coming out from the same cell body, but the dendrites come from the same cell. They have the same Protein, so they'll always Stick to each other and then repel each other so the trick for any other cell is Not to stick and what we're suggesting is that even a single mistake is Enough to make sure you don't stick and that happens randomly. It's just the probability that the same cells will be Produced the same proteins will produce in different cells is very small so Yes The monomer we would argue that the monomer isn't there there may be small clusters formed So but but once you're starting us, maybe I don't understand your question Implicit implicit in this is that diffusion on a surface is fast enough that things will happen out at the proper time There's there's so many many assumptions in this Model, but implicit is yes is that is that if something is there it will find its Place in a short enough time to let the chain to grow Yeah, I mean there's a there's a there's a lot of work to be done I say it's consistent with everything we know about these proteins, but maybe there's something we haven't thought of that's always possible Well, you did notice I had a simulation of the molecules moving yes, so the the cat herons The classical cat herons all the cat herons have three calcium's between each domain that make them semi-rigid Nevertheless, they're fairly flexible. So if I do if I in that simulation I showed you If I fixed the membrane here and just asked what's the region that the External domain moves it's about 2030 angstroms so but you have five flexible regions So they're quite flexible the protocol here is that bind this way are far less flexible because they have a much larger interface So they're far less flexible. So they are You could so one question is does what effect does flexibility have on their properties? We actually think that Flexibility interdomain flexibility affects binding affinities. We've shown that There are alternate Conformations of binding that we think there's an entropic driving force for some of these proteins, but I don't I Think I think beyond affecting binding affinities. I'm not sure what else but again, we're in the earliest stages Unless unless as I say unless this size Unless the size of the cluster is what's responsible for signaling in Which case no it wouldn't make any difference in the case of the classical cadherans what one of the experiments we're Doing now is I said that you need cis and trans to get this assembly So what we're doing now is we're making the trans interaction much stronger in Which case perhaps we can assemble as many proteins in the interface as we do with weaker Trans and stronger cis and the question we're asking is is the order important So if we if we make strong trans interactions, there'll be no order, but there'll be enough the same number of Proteins, so that's the kind of game. We're playing so we we don't really know the answer to that yet But yeah, I mean that's the kind of experiment thinking