 Thank you very much for inviting me to give the talk. Right, so, the gallery at the bottom really shows an evolution in our structural biology efforts. And the title is really referring to what we had to do to be able to get these initial crystal structures. And at the end of the talk, I will take you all the way through to our latest structures where actually there's not much whipping involved. So we'll see how this evolves. So I thought I'd start off with a little bit of an introduction to GPCRs, because okay, you understand that they're really important and for drug discovery, 34% for FDA-approved drugs are target GPCRs. But I think the important thing to recognize is that how dynamic they are. And this has been really instrumental in trying to understand how you can get structures. If you understand your protein, the dynamics that undergoes, then you're in a far better position to be able to engineer your protein and to get structures. Oh yeah, let's get my interruption. So in the inactive state, which is here, there's no ligand bound. Obviously the GPCR just sits in the membrane quite happily. Binding and vaginas though induces a conformational change which allows G protein coupling. And the alpha subunit then exchanges the bound GDP for GDP, you get dissociation. This is the active form goes off and does all its intercellular signaling. Beta Gamma can also do signaling. And then this signaling is turned off through phosphorylation through receptor kinases. Normally phosphorylation of C-terminus or psychopathic loop three. This then allows coupling of arresting and shutting off of G protein signaling, but arresting can signal in its own right. And I'm not gonna go too much into the pharmacology here, but what's really interesting from the drug discovery perspective is that we know that some of the therapeutic effects in many GPCRs may be driven, for example, through the G protein coupling, but adverse effects may be driven through arresting coupling. So there's a lot of interest in comparing different structures with arresting bound versus G protein and whether we can develop really tailor made drugs which only signal through a specific pathway. So there's still a lot of interest in GPCR structures. And obviously we want to get as many as possible to be able to understand this pharmacology. So I'm gonna be talking about ligands, ligand efficacy. So in terms of biological response, a full agonist is maybe 100% biological response. So think adrenaline or adrenaline, but then a lot of drugs may give a partial response. So even though they've got full occupancy, they may only turn the receptor on partially. And then antagonists can be divided into a number of subgroups depending on their activity, weak partial agonist, neutral antagonist. So inverse agonist. And it's only the inverse agonist that really inhibits basal activity and inhibits the receptor fully. Now, the thing to bear in mind is that the receptors undergo a number of conformational changes. And so the inactive state R can transition spontaneously into R star. So this isn't even in the absence of ligand, the small proportion can conform this active state. And then these states are stabilized by the binding of an agonist. But you can have agonists bound to an inactive state, it's low affinity. You have a high affinity state here. And of course, when a G protein binds, you get even higher affinity. So the key part here is that this equilibrium is going on the whole time, even in detergent. So you have to be able to shut this down if you want to get structures. And this has been part and parcel of learning how to deal with GPCRs. So let's just, sorry, I can't stop it, hold on. So this is basically a movie of a GPCR going from the inactive state to the active state. And what we have is just three crystal structures of the A2A receptor going from the inactive, active, intermediate to active. And what you see is this cleft is open that allows the coupling of the C-terminal alpha helix of the G protein actually in here. And mainly through hydrophobic interactions. And so we understand a lot about how the GPCRs are activated now from all the structures that we've done. But that gives you an idea, it's a fairly small change. It's not a dramatic change that we actually see compared with more transporters. Okay, so how have we gone about getting structures? This is basically a rather steep curve that has been happening over the last few years in terms of numbers of unique GPCR structures. So we're up to about 70 of GPCRs out of 350 or so, ligand binding GPCRs in the human genome. And what you see is that virtually all of them have been done by HVAC crystallography. And it's only in the last few years we've been getting unique structures bound by cryoEM. The first structure was done of rhodopsin. And this was a structure that was a purified protein from native sources, so bovine eyes. And the engineering required for all of these subsequent structures, virtually all of the structures have ligands bound as well. So what have been considered over the years? So if we think that most of them have been done by crystallography, I'll mention probably about crystallography initially, it still has a lot of advantages and go on to cryoEM at the end. So what we had to do was think about removing flexible regions, obviously post-translational modifications, as Aaron mentioned, these fusions to the soluble proteins to improve crystal contacts and also using binding partners. And then the two key parts here, reducing conformational heterogeneity and increasing thermostability. You see actually you're using the same techniques to improve both of these. However, you have to be very careful because you can add a ligand, which may increase thermostability, but may not reduce conformational heterogeneity enough to be able to get crystals. So these things have to be considered independently and you've got to find the right combination to allow both of these two to occur. So I just want to say a few things about expression before we go on. So I mean, you know, this is all about techniques. And I just want to remind you, this is about GBCRs. So hopefully I won't upset too many people by what I say again, but of course, as I mentioned yesterday, you know, you've got a huge variety of different expression systems that you can use. And basically you can get anything to express in any expression system, but it's the amount of time and effort required to be able to do that, and also the quality of material that you actually get at the end of it. So if you think of the complexity of the expression system in terms of expressing the mammalian GBCR, so I'm thinking in terms of how well does the system cope with folding, eukaryotic membrane proteins, post-translation modifications, et cetera, then clearly the best systems are on the mammalian cells. If you then look in the literature and see what has been happening, and look at the ratio of the functional to non-functional receptor. So to me, this is really important. If I'm going to start a new project, I really want to make sure that the protein I've expressed has the optimal possibility of being folded correctly with not too much misfolded material. I'll just mention that in a subsequent slide. And in our experience, it's really the mammalian cell systems which have been the best. But even transient transfection in mammalian systems can cause a lot of problems. And then the, I forgot to write this. Sorry, let me get that out of the way. I can't see one of the slides. Then you've got the potential functional expression levels of the receptor. And what is surprising is that actually a lot of the mammalian systems can express them very, very well. And same with baculope. So it's a trade-off very often between these two, these two things that we think about. That's interesting. It seems that my slides have stopped moving forwards. Maybe you can re-share once again. Oh, there we go. That's it. So the other thing to remember is that expression is receptor dependent, but also user dependent. And I speak from my own personal experience of when I was sort of around about 2000 to 2003 developing and setting up mammalian expression systems for use in the serotonin transporter. It took me a while to actually get mammalian cells growing effectively for the different systems and to get good expression. And the key thing for all of these particularly for baculoma mammalian, if your cells aren't growing healthily, your expression will be bad. So I learned that the hard way. So I just wanna say a couple of things about the transient expression systems. An inducible expression. So this is an inducible expression in the T-Rex H2O3 cells. So these are tetracycline-induce balls. We used this in 2002, 2003 for expression of the serotonin transporter. We've always found this to give very, very nice expression. And in baculoma, what you find is, you often find is a lot of stuff is intracellular and not on the cell surface. Actually cell surface expression is always a good surrogate for deciding whether your protein is folded correctly or not. What we came up with with this, so this is actually expression of a GPCR and a Tinsim receptor. And it was basically a quick way to be able to ask the question, have you got misfolded material present? And this is basically a differential Western blob. So basically you just take some cells, same number of cells, solubilize them in different detergents. So we've got two harsh detergents here, which are gonna solubilize absolutely everything, but of course it's going to kill your protein stone dead. So there's no activity there. And then we've got these detergents, very mild detergents, which will maintain your protein probably in an active state. And we should get full extraction of functional material. So what you see, if you look at SF9 cells and you just see what you extract from the membrane, you get a huge amount of material extracted with SDS and FC12. And apparently on the initial, there's virtually nothing here. And if you expose it for longer, then what you see, yes, there is some actual protein extracted. But when you do the binding assays using a specific ligand for the receptor, then what you see, yeah, SDS as you'd expect, nothing there, FC12 a little bit, but then you're getting extraction of large amounts in DDM and Digitome. And just to say, so this is your 100% level that you'd get from membranes. And you're wondering why do I get double? Well, you get double because this ligand won't cross a membrane. So when you make and freeze membranes, you get inside out and right side out vesicles. When you solubilize, of course, the side of the scopes. So you only measure half actually here in the membrane. Digitonin is not quite so good at extracting material as DDM, but you still got a lot of functional material. Now, if you compare your inducible system, what do you see? This is the band of interest, these are contaminant bands. So virtually the same amount of material is extracted with both systems. And again, no functional material here in SDS, FC12, nice functional material in DDM and Digitone. So basically this gives you an idea of how much misfolded material you actually have in your expression system. And this is really important. You wanna be purifying good stuff for structural studies. So we found this actually very, very useful in our work. And this is something is also worth remembering. At the end of the day, membrane protein folding is complex and poorly understood. So I'm always a great advocate of expressing it in a system where folding is going to be most likely to occur in the most optimal way. And I'll be happy to answer questions more about the, of why I think the mammalian systems are better than say bacula virus or membrane proteins. And certainly from our perspective, we found it's more simpler just using bacula and mammalian systems as much faster. Okay, so where did we start on this story? Of course, we didn't start on the river punting, although that's always quite nice. We started with the beta receptor. This is where we started in Gephardt-Schertz's lab. And when he started in Cambridge, what was about 1990. And this was expressed in the bacula viruses. But in terms of X-ray crystallography, now with hindsight, you just look at it and you think, well, this is not going to be a good one. You've got potentially flexible regions, N-terminus and the C-terminus, perhaps other loops as well. And you can see from a disorder prediction that these regions are actually quite distorted. We've got multiple phosphorylation sites and end glycosylation. So lots of potential problems. This is Tony and he actually spent seven years trying to get this into shape. And what he did was for the beta one receptor constructs, he made up to 36 constructs, deleted the N-terminus, C-terminus, ICL-3, he could purify a fair amount from insect cells, from each of insect cells. I think you'll agree that he can purify it quite nicely in a truncated form, nicely non-dispersed and non-gel, it's nice and clean, fantastic. However, what is clear from this slide, no crystals. So why? And what is apparent now is that the receptor is basically too unstable in short-chain detergents. So I'll come on to why that's a problem in a minute, but if you just put this, take it from DDM and do the same prep in some, like, decommaltes on it, then it'll just crash out overnight in the fridge at four degrees, so it's that unstable. So this was a real problem. So subsequently, we've done some work in looking at the, of why these GPCRs run stable in short-chain detergents. And so these are ND simulations, which have been done either with dodecal multicide or with octal glucoside. And as you watch the dodecal multicide moving about, you know, it's jiggling about a little bit, but it's fairly static. If you compare that now to the OG, you know, the OG is on speed. I mean, it is just going around all over the place. And if you look at some of these individual molecules of OG, they are really rotating in the plane of the hydrophobic, where the membrane would be, they're jiggling about. And now at the end of the simulations, they're starting to intercalate between the helices. So this is like the first steps of denaturation. So it's because you've got these enhanced movements of the receptor in short-chain detergents is probably why they're so denaturing. Okay. So how did we get round this? So we got a team of postal. So this is Tony. We've got Yoko, Maria and Francesca. And they worked on developing this conformational thermostabilization. So the idea was to make the receptors more stable in short-chain detergents. And this was a very straightforward, simple process. So we did alanine scanning throughout the receptor. So basically you change every amino acid into alanine. If it's already an alanine, you change it into a lucid. You then express each of those receptors so you can either express them in manian cells or actually originally we decided to use E. coli. We thought it might be easier, but it actually isn't. It's just as fast doing it in the manian cells. And then after expression, you then do the assay. So this is the heart of the whole process, the thermostability assay. And so what you're doing for each of your mutants is then saying, asking how thermostable is it in relation to your wild-type receptor? And so when we say functional receptor, this can be measured in a number of different ways. So initially we did radioligin binding. So we had a radiolabeled antagonist or agonist and you just measure how much is bound. And you heat the sample to different temperatures and quench and then just measure how much is bound, draw a curve and that to your parent. So very, very simple, albeit somewhat tedious. So a heptare is now they have a whole suite of different methods for doing this. So like aggregation assays, fluorescent assays because in many cases we don't have a radiolabeled ligand or the properties of the ligand are good for these assays. But what you end up with is basically round a banal, I don't know, 6 to 9% of the single amino acid point mutations being thermostabilizer. And then what you have to do is to be able to put them together. We've worked out a number of ways of efficiently being able to put them together. And then you end up, in our case, found about four to six mutations being put together and gradually increases thermostability. So all of this, we've summarized all of our experience in this publication here if people are interested in the details. So this is the first example that we did, the beta one receptor. This is when we got into the stage of having single point mutations and we're putting them together in a whole number of different combinations. So these are just six mutations put together in different ways. And then what you see, when you look at the TM, so this is the wild type, all of these are all much more thermostable but there's one which is absolutely amazing compared to that, so it's a 21 degree improvement in thermostability. And what does this mean? Well, the original wild type receptor, which is shown here, we'd only measure thermostability curves in dodecoma altoside and in decoma altoside. Whereas the thermostable receptor in 23, not only could you measure it in DDM and DM, which these curves, so you can see that they've shifted significantly to the right, but now you can actually measure thermostability in octal glucoside and non-al glucoside. And of course, these are short-chain detergents and ideal for crystallography. But of course you have to think about what you've done to the receptor. And what we found was very interesting. So what we've done is biased the confirmation of the receptor towards the inactive state. And so these are just pharmacological, pharmacological gases binding by agonist or antagonist. So antagonist binding is unchanged. Agonist binding is shifted significantly to the right. So it's a key thing. Your structure tells you something about the antagonist bound state, but it won't about the agonist bound state. So yeah, be wary of that. So in the end, it crystallizes in octal-fired glucoside. And so in summary, these are all the things that we had to do to be able to get these crystals. So we have the deletions in orange boxes. We've got the thermostabilizing mutations in red. We've got another mutation, PSS116L to improve expression. We've deleted the palmatulation site at 358. And yeah, so lots of engineering actually required to be able to get that. But what's been nice about this is that once that engineering has been done and we understand it, what we've been able to do is to be able to get a whole multitude of different structures with ligands of different efficacy. So this allows us to understand why it is that a full agonist is a full agonist and why a partial agonist is a partial agonist. We can understand why it is that when you go from the inactive state to a state coupled to a G protein, why you see this increase in affinity of the agonist and also why that increase in affinity is weaker when you actually have a resting bound as opposed to a G protein. So the structures have been incredibly informative. So, as Adrian has mentioned, you can also do these gene fusions, T4 lysozyme. It's another way of doing it. Actually, this methodology was developed by Gil Privey and for trying to get crystals of lack permeate, which never worked, but it worked unbelievably well when put in GPCRs. But the position of these fusion point is absolutely critical. You can change these by one amino acid and it can be the difference between getting a crystal and not getting a crystal. So typically people may be screening a hundred or so different constructs to get the best crystals here. And all the crystals have to be grown lipidicubicite. You don't get good crystals in detergents. You also need high affinity ligands to be able to stabilize the center. And now what people are doing is basically combining the methodology here, the T4 lysozyme and the thermostabilizing mutants and having good ligands. And all of that has come together in a very robust system for structured termination of GPCRs by X-ray crystallography. And these are just some examples. So this is the same receptor. It's been crystallized in a variety of different ways, either with a fat fragment bound, either with one of these gene fusions. This is T4 lysozyme or by thermostabilized. But here you need very high affinity ligands. The advantage of thermostabilization is that you can now actually crystallize the presence of very small ligands with very low affinity. Okay, so I'm running out of time. There we go. So we can do exactly the same thing for looking at GPCR G-protein complexes. So in 2011, I'm sure you realized the, we have this amazing structure came from Brian Kabilka's lab in collaboration with Roger Tsunahara. And yeah, it won a Nobel Prize. No question, it was a real milestone in the field at the time. But nobody has been able to repeat this for any other receptor. And obviously people wanted to get active state structures. And what we noticed was that when you looked at this structure, virtually all of the interactions between the G-protein and the receptor are driven by GES. So I'm sorry, by the alpha subunit, not by the beta and gamma subunits. And virtually all of that is driven actually by the GTPase domain, right? And not the alpha helical domain. So we thought, well, we can simplify this. And so we started engineering. So we don't need T4 lysis, and that was required for contacts. We don't need beta gamma or MB35. We don't need the alpha helical domain. And that's what you end up with. This is the minimal unit of the G-protein that you actually need. We call this mini-GS. And you think, well, let's make this. So to simplify the crystallography. Easy to say, difficult to do. So first of all, we devised an assay. And this was basically an assay which allowed us to measure how good a mini-G-protein is. We know that an agonist shift happens when you couple to a G-protein. So this is the inactive state. So anything that we engineer has to do exactly the same. And this is what Byron did, Byron Carpenter. And I'm not going to go through all the gory details because I don't have time. Suffice to say, at the beginning, this is the best he could purify it. Because if you purified it anymore, the band disappeared because it's so unstable. You could only do the assays at four degrees. You could see an agonist shift, but at 20 degrees you couldn't. So this was very, very unstable. But you could do engineering. And you could do engineering based on the structure of a small GDP age. You could compare it and engineer the protein. So about 300 or so mutants later, you ended up with this mini-G-protein, 28 kilodolton protein. It's got eight mutations, three deletions. It recapitulates all of the pharmacology that you would want for the receptor. So you see an increase in affinity when you go from the inactive state to when you actually bind the mini-G-protein. And even better, you actually increase the thermostability of the protein compared with when you do the G-protein itself. It's absolutely ideal for crystallography. And we ended up being able to solve a structure here. So, you know, G-protein engineering worked just as well with the soluble proteins as with membrane proteins, and they can really help you be able to get structures of membrane proteins in different states. So in the last few minutes, I just wanted to say a little bit about the, about cryo-erea. Because, you know, what we've worked on initially of the simplest proteins, this is the way biology works. You work on a simple system, but there are some real beasts out here. So this is a sort of proteins we're sort of interested in now. The ligand itself is 28 kilodolts. The ligand is, it's got 11 disulfide bonds, 3-ended-guide-cosylation sites. This is just the ligand. When you look at this, you've got a large extracellular domain. This has got 6-ended-guide-cosylation sites, 5-disulfide bonds. There's a cleavage site in here. If you're a mediate cleavage, which requires for activity to remove the C-peptide, this is a nightmare. So how would you actually do this by extra crystallography? I don't know. You need a lot of work. I'm sure you could do it, but it'd be a lot of work. But the key thing with cryo-EM is that, as I'm sure you're aware, you can use any detergent. So that means instead of worrying about trying to crystallize things in O-G-N-G, or using lapidic cubic phase and optimizing the packing in there, you can just go straight for very mild detergents, very, very large. So you've heard all about FSEC before yesterday. So I'm not going to dwell on this. Suffice to say, it is a real, really important tool in our lab. It allows you to assess the receptor quality, the receptor quantity. It allows you to assay a whole variety of different detergents and stability, and you can get, in an afternoon, a huge amount of information, which really helps you be able to take your structural biology project forward. So for example, in this case, this is a typical of what I mentioned about the problems with transient transfection. You transient transfect, you get nice green cells, what would you do? But you see you've got a lot of misfolded material here, and the peak, this peak doesn't look too good either. You use lentiviral transduction, you can make stable cell line here, and now you look at the peak there. That's what I want to start with, is to look at the purification and not this, much more dispersed, much less aggregate there as well. And then you can go through and test a number of things. So for example, we know from high-resolution structures that we've done, sodium is now a steric antagonist. You can look at different effects of different concentrations of sodium, and you can pick the best concentration for purification. You can look at the best used FSEC also for looking at many G proteins. Okay, we've done structures, they all look amazing, and the resolution is getting better and better in terms of just the sheer number of particles you can now actually get. I think this is worth thinking about, and really the importance and use of cryoEM, because when you think about it, you've spent so much time engineering the G protein required for X-ray of removing flexible regions, post-translation modifications, optimizing crystal contacts, thermostability, all of these things we don't have to worry about now with cryoEM. Protein dynamics, yes. Here we want to lock it down. CryoEM maybe, maybe not. In some cases we've been able to separate out different conformational states, but you don't want too many of them and you don't want to continue them. So this is something you may need to think about, but of course the real advantage, you can use virtually any detergent. Size may be a little bit of an issue. I think at the moment, we're around about 70 to 80 kilodolton in size, which of course for 35 kilodolton, GBCR is a bit challenging. And yeah, there's no worries about shape or the complex dissociation perhaps isn't come up in the discussion again, but yes, this is a real problem. So this is just one example of a one problem we have, preferred orientation. This is a nightmare and you can spend a lot of time. This is equivalent to your crystal problem in X-ray crystallography. Your crystal is not diffracting properly. You've got a lovely sample. You only get one view and that's it. You can't get a structure from that. And there's a whole variety of things you can try. And in one example that I'm going to show you, you've got to collect tilted data sets, which is a bit of a nightmare, but it was absolutely essential to be able to get this particular structure. And this really shows the power of cryo-EM. Could you ever get a structure like this with a nanodisc, liquid nanodisc, which is essential for arrest in binding? You're not going to be able to do that by X-ray crystallography. I don't have time to go through all of the purification stuff, but I just want to finish off with, you know, the revolution which is occurring in cryo-EM and make no mistake, it is a revolution. It's going to get faster and faster. There are loads of more improvements which are coming online and small membrane proteins, people are now working on these 50 kilo-dolton membrane proteins or smaller, just putting on a fab or a nanobody, you can increase the size and do the structures. This really shows the way forward if you like the GPCRs. So this was done by a PhD student. He started without having ever worked on a membrane protein and within 14 months he had gone from absolutely nothing. We'd never worked on this protein before to the density for this structure. And what is really nice is that this is a wild-type receptor. So there's absolutely no deletions, no mutations, and all the post-transitional modifications are present. We've got 125 of amino acids hanging off here, which is unstructured. And actually, if we tried to engineer it, we probably would have got everything wrong because this was thought to be a really unstructured and terminus and actually is absolutely essential in a domain-swapped method of forming the dimer interface between this GPCR dimer. So there you go. It's my favorite cartoon. You have to think, you have to work hard and be meticulous and never ever give up. And well, as the frog found out, there's always a way forward in the end. So thank you very much for listening. I'm sorry I ran over a little bit. And I'd like to thank everybody who was involved in the work and of course all the funding over the years. Thank you very much and I'll be happy to take a few questions. Thank you, Chris, for this interesting talk and showing us so many strategies to purify and mix table GPCRs. So we can take a few questions. And the first question is like, what do you think about like, what is the reason for better yield of functional protein with stable expression rather than with the transients expression in mammalian cells? Yes, a good question. So with transient transfection, you cannot regulate how much plasmid goes into each individual cell. And there's probably a range of over 10 to the 5 difference between 0 to 1 molecules and 100,000 plus molecules of plasma. And what that means is that you can have a vast overproduction of messenger RNA. And that just totally overwhelms the secretory pathway in the folding systems for the membrane protein. And so all that happens is that if that happens, you end up with misfolded protein. Whereas if you do a lentiviral transduction, you can determine you have one, two, three on average. Lentivirus is going into the cell. So that means that you're not overwhelming the cell with mRNA. You get proper folk. So we have another question. At what stage in purification can you test the receptors for thermal stability? Oh, yes. That's a good question. So actually we test them for thermal stability all the way through. So you can actually just take cells, solubilize, because they've got GFP on, an F-set, you don't need to purify them. So you can do a thermal stability asset. Oh, sorry. Yeah. Sorry, I was thinking about something else, not F-set. So you can use F-set, heat F-set, which David mentioned yesterday, but you can do it immediately on solubilization. You can actually do it in membranes if you want. And you can do it after purification. And very often what you see is that your most stable protein, maybe in the membrane, when you solubilize it becomes less stable. And when you purify it becomes even less stable, presumably because you're removing lipids in the life. So yeah, you can do it whenever you like. Okay. Then one last question. What is the physiological relevance of thermo-stability? I think it's quite similar. The physiological relevance of thermostability. Well, obviously the mutations we put in have no physiological relevance whatsoever. They're just basically a tool to be able to get the structure. But the corollary is, is that some receptors are more thermostable than others. So for example, we worked on the Turkey beta-1 receptor initially because that was far more thermostable than the human beta-1 receptor. Why that is, I have no idea. And I don't think there's any particular correlation between an organism and thermostability. So it's just what evolution has come up with. You know, people have looked in thermophilic bacteria for, for transporters that are much more thermostable. Some are, some aren't. So yeah, I'm afraid there isn't anything. Thank you.