 I mean, it's, it's a pleasure to talk here and actually see a number of friends listening. So, you know, it's actually interesting, I saw Chris there, and I saw Chris on the, on the, on the symposium and, you know, Chris, I think, I think when I give a talk on this, it's been a while since I gave a methods talk, but it's a good chance that Chris is also giving a talk in the same symposium. That's fun, even if it's by Zoom here. So, what I'm going to talk about is sort of the approaches that our lab has been using over a number of years to go about, you know, understanding structural details of, of SLC transporters, and for those who didn't know, the human membrane proteome has around 400 or so, so it's after GPCRs, they're ripping in the second largest fraction of the membrane proteome. They haven't been so much targeted as drug tigers, so they're, that's changing, it's been just from farmer for that, but of course, they're responsible for transporting metabolites, ions and drugs across the membranes, they're very important to cell homeostasis. And some of them, of course, are known drug tigers already. So, our interest in understanding structure of these guys, we can better understand their role in human physiology, how they work, should maybe as an idea of how they go about the function and where they do it. So, our interests in the lab have really focused on two main types of transporters, sugar transporters, those monosaccharides such as glucose, those are glute transporters, and those that involve demaintaining pH homeostasis in certain levels, certain protein exchanges. And of course, those two systems are fundamental for cell homeostasis. So, you know, when you want to think about developing a mechanism using structures to understand that, of course, we have to appreciate what we're trying to do, and to understand how something works, we obviously going to need multiple confirmations. And what we're trying to do is to understand a picture of this so-called alternating axis cycle, and this shows you that they can come in three different flavors of how you might move a substrate across the membrane. It's not really necessary for this talk, but it's just for us to remember that, you know, a structure by itself is just one snapshot, and obviously multiple confirmations to really understand how something's working. So, you know, we're going to need to combine that structure information with assays, kinetics, binding, we need something of a relevant confirmation, you know, we think working detergent, obviously stuff working membrane, things naturally dynamic, even with, you know, career, we might get states, we understand how those states are connected. To all that, we really first need to obviously work with the protein. So, you know, so honestly, being able to produce the protein, in fact, I would say with Kareem, it's more and more challenging targets. So things that we may not even consider before, because difficult to get, you know, yield to sufficient enough, and maybe stability issue. Now, of course, you know, we made things so stable because we work in multi-tergents and get structure information. And so things are definitely changing and requirements are definitely changing as well. But bear in mind, you know, to build up a model, we need multiple confirmation states and understand how they're connected. So the approach that we took when I was a PhD student, and I think I presented this for the first time in a meeting in Switzerland in 2000, I think I definitely remember presenting that post to the Chris. So it goes a while back. But the idea of the time was that Jeffrey Waldo had tagged things to GFP and shown that you could use GFP as a folding indicator for global proteins. And we thought, could this work for membrane proteins? Turns out GFP is not as good for the global proteins because global proteins fold multi-fold, I mean, two proteins will fold cosententaneously if you like, whereas membrane proteins with co-translational folding, of course, if you have the GFP at the C-terminus, the membrane protein first has to integrate to membrane before the GFP can express me folded. So then, you know, we could show that you could use the readout for the presence to give you an idea of how much protein was being expressed, at least how much ended up in the membrane versus inclusion bias in the coli. And of course, the nice thing about GFP, you know, that is that you can also detect it on SDS gels, it's very stable. And so we don't need to use Western blotting, which is also an advantage, of course, having good antibodies against membrane proteins, of course, you know, it's a little bit simpler. But I guess the sort of the bit that really sort of made sort of GFP, tagging more widely used, of course, was this fluorescent size student chromatography, the Eric's lab set up. Essentially, you could take your solid-blades fusion now before it's purified, run a size student column and see what it looks like. So you might want to test as detergents, but of course, now you can look at different homologues or different constructs. And of course, the bit to remember, and this is a bit that something to forget, is just because something's in a membrane doesn't mean it's well-folded, right? So obviously, looking for things that look like they will be folded, but they're still not functional. So the GFP tag gives us readout stuff that gets into a membrane, won't really tell us about the quality of the protein that's there. But even that is very useful. And we've been working with pipelines and E. coli and Saccharomyces. And the idea basically is, you can think of the GFP, is we still can't predict what proteins will be able to be expressed, right? So we need something that's going to enable us to go through this process as fast as possible, allow us just to get information that's useful that we can go back and redesign a system, redesign a construct. So with that in mind, in fact, this was reason with Saccharomyces, all of the time, Pikki was quite important, was that Saccharomyces does one thing really well, and that's a mulch recombination, which means we can adjust synthesize our gene, and we can add it to a linearized vector, and Saccharomyces will happily recombine our linear DNA fragment with a vector, and the cloning becomes trivial. And so that's a real advantage, because within two days, in fact, we typically order gene strings for live technologies, linear DNA fragments, and they'll come and we could order 20 clones, they come to the tube, you can actually add water to those overnight, and then directly transform those, and you get colonies in a couple of days. And so the idea was to have this fast system where we could quickly look at expression, and then we thought, then we'd go to Pikki, but it turns out that most cases, we give enough material to work with, so we actually end up sticking with Saccharomyces. And so we have our GFP tag, it has an 8-Hist tag, I was interested to see if the 8-Hist tag was quite popular. I think, I made an 8-Hist tag in E. coli because I failed to get the 9-Hist, and I gave up on cloning, but it's interesting to see how the 8-Hist then sort of became more popular, there was no logic actually behind the 8-Hist tag, but in any case, this is our vector use, and we have a cleavable site, we used here, but of course you use precision or whatever. And so what we typically do is, if you have many constructs, and you just say we can quickly transform, but these are quite a tough cell wall, so we need to add some sort of glass beads, break the cell debris, spin them down, and actually you can just spin down the supernatant on even a normal desktop centrifuge, it's enough to isolate membranes that we can solubilize in a gel and do some STS page and in-gel fluorescence. And so when we did that, this was now, when I was a post-op with Sewater, and that should have been Simon, and you stayed to join me on this, and working with eukaryotic membrane proteins, and actually it worked fairly well, and we managed to express actually quite a number of drug transporters, I can't remember exactly the ones that are here, but it works surprisingly well enough for us to be able to, when we're just taking a selection of SLC transporters, number of homologs, enough that we'd find few that to work with in that case. And some of the control experiments we did was to check that if you're taking 10 mils of a culture, when you scale that to fermenter, would you get similar yields? And you did, if you're taking wholesale counts, and then you isolate membranes here, do you get roughly the same amount of membranes? And you see there's a pretty good correlation. So if you could take a wholesale estimate and then roughly estimate what that meant in terms of membranes. One sort of technical point, which may be interesting to some people, I don't know, is we always wondered what didn't spin down, right? So you do the auto centrifugation step, we use a Ti45 rotor, but it was interested in stuff that is still fluorescent and supernatant, is that free GFP? Actually, if you run a gel quite often, it's actually still fusion. So I think it's actually vesicle that are just too small to spin down. So if you spun longer or use a Ti70 rotor, you might be able to isolate slightly more material. In any case, there's a little bit of loss there, but this is our approach. And this is an example of just what this means. And so this is LACY purification. This is what you might get. Some solos membranes, it's well-expressed protein. LACY, why? You flow through what's bind in nickel column and your elution. And then in GELF resins, you can, you know, you can might be able to pick up my finger and there's a camera there in the natural light. I'm just taking a picture of this in GELF resins. Obviously, you can see everyone's really concentrated GFP is there. And that's very simple, of course, of course with the blue light, you can go to lower detection limits. And this is our recovered protein, which is fine for crystallization, et cetera. And so we, you know, use that pipeline and we're interested in this case. We started working on glucose transporters. Well, I got too much detail about those. Obviously, they're very important to genome-wide, mainly in blood glucose homostasis. And so they're one of the systems we're testing in yeast and we could, you know, get nicely well-expressed protein. We see, you know, here, you might get a bit of free GFP that's quite normal and it's typically protein that's been degraded, but the GFP is still stable. But something like this we wouldn't be worried with because the main protein peak is monosperse and then we have a HISTag and we'd go to the purification, cleave off the GFP and go to the perparsal reverse column. And we end up with the purified commoner protein. In this case, we could compare its transport of the glucose and its vitamin inhibitors. And they were similar to the same protein. In this case, that could be produced from red blood cells. And this was worked with Michiru Kazuhara, which actually did the original reconstitution of these transporters in the 1970s. And it was as good as material that he had worked with. I mean, it's nice confidence that at least for the glute transporters, this system was working well. And so we ended up that we could actually express a number of glute transporters in saccharomyces and things like glute four doesn't express as well. And glute two and some expressed better words, but we screened a number, not extensive, I would say, maybe a little bit more than this, but maybe two or three times more than this. But just at that time, CDNA, maybe we could get a hold of and see how it went. And so what we ended up working with actually was, I can't even quite remember how we decided to work in Glute Five. I think, you know, it wasn't one of the best expressed, but maybe we tried a few in this one, ended up giving me some crystals, just to note that the wild type protein has a single inglycosylation site, but removal of that glycan site was quite unusual. Mutant was based on previous literature. I think that's why I made it mutant, but it was well-folded and it didn't seem to change in terms of functional activity. But this is what post people realized was crystallization, obviously. We have crystals that are poorly diffracting crystals at the time. And we could add lipids and we could prove the stability, but we got stuck at this five unstrongs, which is obviously very typical for this sort of project. And so this is when I was finishing up, the postdoc was sort of trying to, if you like, tabulate where we had got to and what other people were doing was screening a number of SLC transporters, mainly. And basically, if we were just interested in things that were month spurs and that we could express neither yeast or coli, and we could purify in DDM, it wasn't too bad in terms of getting crystals, right? Obviously, we know what we're doing a little bit and then we're making mutate, but it wasn't like this was the hurdle, of course. The hurdle is getting something that's crystallized to actually solving its structure. So what did we need to do to improve the crystals that we were getting? And so the question came, of course, and this is really work to be headed by Chris that looking at GPCRs was actually how stable do these transporters need to be? If we want to get them in some other detergent, what do we need, basically, was the question we had. And so at that time, and this is a number of years ago, but maybe it's still, I think it's still important and relevant to think about, especially if you want to be in crystallization, what things you need, what requirements you need. And certainly look at the data at the time, the short similar, that things in my, smaller micelle detergents tended to go to high resolution, of course, because the smaller micelle, the more protein packing they can get, the lower the solvent content, the better the crystals can attract. Obviously, the point is to work out how stable do you need something to get them in those small micelle detergents. And Chris will be, tomorrow, I think about whipping GPCRs into shape, essentially where we want confirmationally stabilized proteins that obviously more than stable detergent. So we can get them in those small micelle detergents, so they're gonna diffract well, and you can tell you about that tomorrow. But at the time, we were just interested in something that we could sort of do and quite easily do. And we had, so we ended up using this CPM assay, which is one of these thermal shift assays, like a supper orange or something, except for membrane proteins, more hydrophobic melomide dye, that when it forms a conjugation so a hydrogroup becomes fluorescence. So we can just look at increased fluorescence over time as a readout of stability. And so, just to bear in mind all the tags we're looking at with Montes first, we could crystallize them. They had multiple cysteines in TM segments. And the idea was just to see how stable these things were. And in a sort of a systematic way of what we could sort of handle at the time as well as trying to get our own structures. And so this was actually quite useful as we decided to get a couple of master students to work on control proteins. So we ended up going back and crystallizing proteins with structures that we really know in, you know, at the time, we needed this control. But I think it was very informative for us to understand to compare to our own samples what we needed here. And so we did that. And then we measured unfolding rates in this case, that's what better for us because at the time we didn't have a great RT-PCR machine. So we ended up just hitting at one temperature looking at unfolding given time period. And then from that, we could basically, from the exponential decay, we could look at the relative unfolding rate of our protein in different detergents, essentially. And so when we did that, we found that as you might expect is that the bacterial proteins, obviously a little bit more of those, we could crystallize the more stable and large micelle detergents, and as the detergent micelle gets smaller, the average stability comes obviously worse, it unfolds more quickly. The Euclid proteins, we didn't see the same trend. I can tell you why, essentially, they're on the limit being stable enough already. So, you know, those that we could produce and get some crystals for, you know, they're just stable enough indeed more ready. So we're actually not seeing that trend because we're already on that, so limit stability. This is just something like, you know, I did it and, you know, it's obviously a small data set, but I was just curious. So I looked at the calculated micelle sizes by light scattering and probably against the average half-life we measure, and we see a fairly good correlation. I mean, I'm sure if you add more data sets, it's not gonna be this good, but, you know, it seems that, you know, to generally the size of the micelle rather than the hair group suggests to be a more dominant factor in terms of driving instability in the protein. And so, of course, what we want to do is look at these half-lifes that we're monitoring and compare them to the modusperity of our sample. And so one of the things that we found, for example, and this is what we did, so we took with that the stability, the green is a harsh detergent, LDO, DDM is a mild detergent DDM. And basically what we could very much clearly see is if this, if a protein you're working on, it's gonna look good in DDM, it's gonna look good in LDO, which is a very harsh detergent, and you superimpose the two exit traces. If they look the same, you're good to go. It's a very stable target. And this will quite, it's very nice, if you go around and purify the protein, then calculate its unfolding rates in half-life. So if you wanted to screen, hey, you could just screen in LDO, but remember to do the comparison to something like DDM. It's always difficult when you just look at one detergent. And then we could basically, it was very clear that the structures that had got high resolutions of mice detergents, these were these gray guys, our control proteins, they were definitely coming up the most stable in our dataset. And then a few others were obviously were interested in some of the structures though, which is obviously good for us. And over time, we've actually could focus on these few of them, we actually managed to determine their structure. So it turned out to be very useful, although it didn't have a work on these glucose transporters, which were obviously, and a lot of the eukaryotic proteins, which are the black guys, which have still course still unstable, haven't solved our problem. It's just said, if you want to screen enough and you were looking for natural homologues, you could do that in bacteria and you could probably get away some of you, that is useful. But for the eukaryotic proteins, there's an interesting, we haven't really found those targets still. And so one of the things that we did was to say, well, okay, let's just go back and keep screening. And we, rather than the homologued rat, the Glut5, we then found, you know, a different version of one from Bison, actually, I think it was, that doesn't express so well, but the Essek trace in the house, it was looking pretty good, was as good as the rat one. So it was like, oh, okay, at least there's something most of them are worse. And in fact, that actually went really well. And this is, you know, still you need the hands. This was Gregory Vadom, who had actually just come, took this and with actually within a half a year, we actually got a structure for Glut5 and using now focusing on the homologued, that was slightly more stable. The other structure we obtained was actually collaboration. So's lab in Japan at that time, it was what I crystallized, but we never got to work well, but of course they raised antibodies against it. I was actually Fee fragment that managed to co-crystallize. And we were lucky that both were different confirmation. And then, you know, it's now this to how interpret how we, you know, that the confirmation changes are undergone for glucose transport. Of course they are from different species, right? So, you know, there's always the caveat there, they're not quite the same protein and for the MFS transporters, I don't think we have a single transport and all different confirmations. So that's obviously that's a caveat there, but this is actually a very well conserved structural fold. So in comparison to other family members, it's maybe a little bit easier to do these structural comparisons. And another protein that we worked on, excuse me, just checking my time here, was familiar parasite. So we got these structures and Yang had published other structures of glutes. And I think SGC, and I think I'm not sure if you're gonna publish this at that time, but in any case, there were a few glutes. So I said, let's work on something that's a bit distant for a number of reasons in terms of actually this guy can transports many different sugars very effectively and quite interested in how I could do that compared to our more specialized group transporters. So we did that. One issues we had is that we couldn't synthesize the genes that had to be cut and optimized. We prefer not to cut and optimize because actually in the cases, we have tested it's actually been worse. And I have say it's probably about four years ago since I tested the last latest algorithms. And we can talk about that later, but I think you just remember that membrane protein is full co-translation, the mRNA structure is important. They have natural pore sites. So obviously current optimization depends on how good the algorithm is to optimize it. Of course, it's just not the frequency. The codons, which are the issue here, lots of different factors. One of the factors is important is initiation. So you want an mRNA structure that's a little bit, we say unstructured in the beginning. And so, and people found this to be important. So we ended up, what we did here is, because we had to cut and optimize, it was actually expressed quite poorly in use. So we ended up replacing the first five amino acids with those from the group transporter, group five, which knew expressed well. And that improved the expression about three-fold or so. And that's ended up the construct we used. It wasn't, the intern of us is not so conserved. So I think it was a fairly, you know, minor thing to do. And this was for it to work in saccharomyces. It was well-behaved. And then typically, right, and the same sort of thing, you get some optimization over time, it was as DDM plus some other short-chain detergent. I can't, if we use HGG or we, it's one of those, yeah, detergents that we could add. You know, you have to combine multiple crystals. So, you know, not the best, not the easiest data set, difficult to refine, but, you know, it surprisingly gave a very good signal for the bound sugar, even its resolution. It was a very nice single mit-map and it allowed us to build the structure. And we mapped the coordination sugar based on where it had been in homologous structures. And I think that's very logical because the binding site is so conserved. So the takeaway message there was that we could use saccharomyces. We might have, in this case, the sequence because it was poorly expressed. But, you know, we could sort of follow the same steps we did for the glue transporters and it actually didn't take us as long as we'd like to optimize this protein for structural work. But then it comes to the question is, you know, how, why, if you like, why these euclid proteins are more unstable than the bacterial transporters? And it was actually coming up in discussion there because I was thinking, well, of course, acroporins are quite stable, why? And, you know, and I think that the role of lipids is very important in terms of, you know, if the proteins evolved to need specific lip interactions. And so this was the idea we're thinking about. And one anecdotal evidence for this, I think, is the, of course, the thermostabilizing utensil receptor, which is so thermostabilized that it can be crystallized in a coli. I remember Chris, and Chris can correct me, but I mean, remember that the A2 ways, there's a few GPT-Rs expressed in all different hosts. And it's easy to see those ones less fussy in terms of the lipids, I'm not sure, but at least, you know, they'd be interesting to look at that. I mean, the nutrients and receptor-based, you know, maybe the idea is that it's so thermostable now, it's so thermostabilized that it's insensitive to what lipids it needs to be stable. So we looked at lipids, and we're interested in if this is something that was maybe an issue here. And then we use, so in this case, we use the heat-fsec method. And so basically, it's your fsec, but you're heating up your sample, right? So GFP will start to lose its presence around 76 degrees. So you can't look at the stability of everything that has the nothing temperature that is higher than 76 degrees, because obviously you lose your reporter now. And so the idea is you just heat up your sample, spin it down and see the change in the peak height as a function of temperature. And from that, you can calculate relative melting curve in the dirt that you're interested in. So this has been used for Eric's lab, look at ligands, we're interested if we could look at lipids. And so what we looked at was a certain part of an exchanger where we looked at with Cal Robinson's group, native mass spectrometry, that this guy liked to bind, this one liked to bind catalipin, whereas this other homologue didn't. And so we could detect that, it was very nice, thermostabilization by the catalipin for NHA. And then just because just adding lipids could give general stability. So these lipids are thermosoblised and detergent, and then we add them. So this is the control, it's just detergent added, detergent plus lipid added. If you add like PG's, in this case, it was really clearly with catalipin, it was clearly added rather than PG, PE gets the same, doesn't give any clear stabilisation either. In this case, if we make a monomeric mutants, the idea was that the catalipin bound between the dimers and we made a monomeric mutant, we saw that we indeed abolished the binding for catalipin. And so then what you can do then is of course you can do a titration of catalipin at increasing concentration of lipid and then look at the change in thermostabilisation. And so you get this sort of smoidal binding, if you like, so they're just in cooperative binding of the lipids to several multiple lipid binding sites. And then we can look at what these samples look like by F second sort of slight long amounts of catalipin, high amounts of catalipin. You see at this concentration of catalipin, the protein once it's heated is, it's a little bit quite unstable here, you've got a bit of a dimer form and then you go to this monomeric form here, whereas it based on what we know where the monomer is based on the monomeric mutant in this blue trace here, but just adding a little extra catalipin here where it's more stable, we could see that predominantly the protein was more stable and probably migrating as a dimer. So we could really say that and using this assay that the catalipin was thermostabilising the protein, the thermostabilised was thermostabilising the dimeric form of the protein. And indeed there's the dimerisation interface is a lot of positive charges, where it has structure with DDM and two software anions, but you could model catalipin in that case. And if you actually subsequently we mutated the arginine resus interface and we could now abolish the catalipin binding. So the point there was that we were, you know, could use this assay to look at lipid binding and lipids are important for thermostabilisation, of course. In this case, biologically it's quite interesting because it's time as group has shown if you actually look at deletion strain from the coli, NHA is monomeric and non-functional. So we think it's in this case to the regulatory mechanism because under salt stress, the coli cells increase the catalipin content and this transport is required for salt stress. And indeed evolutionary, these two proteins I talked about this was dependent, this wasn't dependent catalipin, this was, they looked the same, but they evolved a different dimerisation interface. And so this is what Carol was also looking at. And so sort of, if you like, popular light or sort of, you know, sort of the idea, if you like, that interfaces that were weaker become more dependent on lipids for organisation and that might be some way to regulate activity through organisation, whereas interfaces that were more protein mediated were less lipid sensitive. So you could think of some sort of scale here for that. So getting back to sort of the, what that meant in terms of structural work, we're also looking at, in terms of a mammalian sort of for an exchanger, just to point out to, in this case, not for crystallisation, we'll be tripe crystallised, but it didn't crystallise. In this case, we expressed in yeast, we screened at tripe and homologues, the one we worked on end up working was from horse. We liked to use a twin strip tag on the GFP. Actually, sometimes we have a twin strip and a HIST tag and the version that worked actually is a slight truncation here and maybe interesting for some of you, you can even get a protein migrating at 30 kilodaltons, but mass metrometry confirms that's still our protein, even though it's monomeric sizes around 70 kilodaltons. So they can, as you know, they can run strange on SDS gels. Of course, this one was with screen, so the idea was screened with stability as something we might need for crystallisation and that was good enough in detergent, in this case, LMNG, we switched from DDM because of the difference, of course, in the CMC from LMNG DDM and this was good enough for us to get a structure. If we kept this large seton domain, which, sorry, if we kept large seton domain which we actually couldn't model in the end, which was site disorder anyway, the resolution was worse unless particles could be added to the final 3D construction, whereas the truncated version, which is obviously a little bit more thermostable, more particles could be added and of course we could have a better map. And I'm going to skip these details, but the bit that I wanted to sort of mention here is, again, we had an idea that lipids could be metering, optimisation here, and with Carol's lab, she could look at the lipids by natural mass spectrometry, but then you get these peaks, and you don't know what these peaks are. The peaks were large peaks, which would be consistent with something like PIP2, but we really, you know, it didn't confirm the lipids. So then we could screen the common lipids with this shift assay, and we could pick up the specific thermostabilisation with these PIP2 lipids, and then we could validate this by looking at delta TM shift, and we can see that we keep clear thermostabilisation with PIP2. This is other lipids. In this case, we can idea where the lipids might be binding. There was this loop here that had number lysine residues. Again, we could make a triple lysine mutant, which now by native mass spectrometry flies to the monomer, rather than all the monomers that were coming out didn't come with lipids, and if we use that thermal shift assay, we could see that the mutant, indeed, abolished our binding. So we could use combination by native mass spectrometry and a thermal shift to sort of start carrot crisis lipid interactions. And then that allows us to sort of think of a model here where this lipid interface is naturally, if you like, less stable. And the lipids are required to stabilize the interface by these proteins which are working by an elevator mechanism where you might need some scaffold. So just in the last few minutes, what we've done is then seeing if we can get away from having to the sixth step. So one of the advantages, of course, with the CPM assay, it's relatively high throughput. A lot of thermal shift assays, but the disadvantage is, of course, you need purified protein. The advantage of f-sec, f-sec, of course, is you don't need purified protein, but it's a bit more time consuming. And so the idea was, can we get this into a sort of format a bit quicker? And so what we found out end up doing is if we heat our sample in DDM, but then add OG to the sample before heating in DDM, we can push those aggregates to precipitate so we can spin them down. So the OG will destabilize the protein, but it also means that we can look at the thermal stability without having to do side-scrubing comatography. So this is just the same sample here. This is looking at the fluorescence and supinated. And if we resuspend the pallet, of course, we see the opposite correlation up to a point in which we use its fluorescence. And so we took the same control proteins I showed you earlier with where crystallized. We took them again and we compared their melting temperatures, but heat f-sec versus which we now call GF-PTS and we're giving the same correlation. In fact, they gave the same correlation in terms of their half-lifes. We had also measured with purified material. So in a way, the unfolding rates we could see from the two stabilized membranes were consistent with unfolding rates and purified proteins in different assays. So this is essentially it. We have a fusion in membranes, a purified fusion. We heat it up. We add OG and we spin down. And then from the thermal shift, we can try to obviously look for stabilization from the ligand. And so one thing that's interesting for us is when we did that for data set, we found that, you know, the eucalypt proteins were of course these are small numbers, but it's sort of interesting that the median stability wasn't that far off the bacterial proteins. But after purification, eucalypt proteins just become far more unstable, which is what everyone's experienced. But before purification, they weren't too bad. Whereas the bacterial proteins, they didn't seem to change too much in terms of their stability. So in other words, this was the same trend we'd get for comparing this essentially the same fold. This guy was behaving like an average protein data set, and this guy was also a kind of protein set. And so essentially the idea is that the eucalypt proteins which was anecdotally essentially evolved to recognize certain lipids. And so when we go about purification and remove those lipids with detergent that become more unstable whereas the bacterial proteins are less lipid sensitive. And of course that means that you can then use this as an ideal stabilizing lipids. Some lipids can actually be destabilizing interesting. Of course you can screen lipids, crude lipids or of course cholesterol hemisuccidate, which to be honest tends to be more generally the most stabilizing lipid anyway. But at least it gives you an idea of understanding lipid preferences of the protein. So in the last minute we'll realize I'm sort of losing time here. Of course you can use the same assay potential substrates. And this was a protein we're trying to characterize with porcelain bicellic acid and this is the human version versus the version from plant. And it gave the same profiles when we add different potential substrates. And then we could also calculate the binding affinities and compare those to IDC. So actually the affinities for both the human and the plant one gave the same binding affinity as IDC, but you know this stuff has been measured from detergent cell-wise membranes, right? And so in fact IDC was not able to measure c-picyclic acid binding but we could with this assay because it was this bind with two-law affinity picked up by IDC. One of the things we've also found is adding monolien, which is roughly lipid-cubase crystallization. I'm out of time, I'm sort of very quickly here. And of course we can't assay the concentration to be used for lipid-cubase but treating it like any lipid assuming to be female-olien in LCP crystallization condition can see it was actually quite destabilizing but by adding a ligand that we know stabilized the protein could restabilize the protein that's able to come in the structure of this protein. And then later on once we've got the structure of course then you can use the same assay to look at the effect of the point mutations but without having to purify all these proteins through the binding assay to ease the plant to the human one. So I'm just going to read this for a different transporter in HA. We have the same sort of idea. We're looking at the effect of lipids to stabilize the protein that enables it to come in the structure of LCP. And now we've recently put this into a 96-well format so this is an idea there's a small nucleotide library and then if you can pick up the control substrate versus a very similar non-substrate versus all the other nucleotides and so the idea is to now use things like Green Lantern more for the rest of the GFP to try and get the 38-well format for small molecule screening and for those interested we have a recipe paper coming out where we can show you having GF-PTS for lipid interactions, ligand interactions from purified material but also stuff in unpurified membranes. Real lastly for those that don't work in yeast we go to Hex-Sales and that's mainly for crime rate at the moment due to cost and that's transfections and Hex-Sales. I don't think we do anything special we're using cleavable GFP twin-strait TAG and for us it's just a question of cost. Anything that looks okay-ish in most person saccharomyces is always going to look slightly better in Hex-Sales. So for us that's still our initial platform is yeast, we're still screening yeast there's humble loaves in yeast and those where we need to go to will go into Hex-Sales. With that I'd like to thank you for your attention and sorry I don't have time for acknowledgments and sorry I'm going over time. Thank you. Thank you David for a very good exciting and informative talk. We have quite a few questions in the chat. Let's take a couple of them and then whatever hasn't been answered we'll come back to in the general discussion at the end if that's okay. So we have a question about seeing a higher molecular weight band for a purified one protein at 68 kilodolton 68 kilodolton specific band there is some so in the yeast there's you do get a non-frescent a fluorescent contaminant around there it might be a flavour containing protein it's actually quite useful because if you compare it to samples you can use it to make sure you're loading the same amount on the sample but yeah there is okay so you don't think it's good one because this is purified sorry it's a purified one purified protein looking at bands on gels as you know they sometimes get higher molecular weight species of course the unfolded form will run more towards correct molecular weight it's still a little bit higher for glute 1 I can't I can't remember I don't pay too much attention to STS bands I'm a bit the same I also find that they often appear to be smaller than they are they migrate faster obviously they migrate faster for NH9 that was a bit too fast it was definitely a protein so yeah okay so we'll take one more question now and then we'll leave the rest so Julie wonders how important the protein stability in small micelle detergents is for cryoem studies compared to crystallization you know so I mean I won't yeah very important I think I mean essentially I mean obviously it's alignment issue right and so the question is how much is alignment issue and how much is conformational stability so you know I can say anecdotally in our lab the samples that have gone the best for cryoem are the ones that would optimize for structural work for crystallization right so the stuff we've been from even though the 2D classes they still look like we haven't quite managed to get good 3D structure constructions whereas the stuff we've been able to express in yeast a little bit more stable there's definitely work for us so you know and I suspect you know obviously if in conformational stability by antibody of course then you have something that's going to aid in alignment and also of course you know stabilize your protein that's going to make sense for small membrane protein but I'm sure right the same sort of thing if you've got some of that it's more stable it's more homogeneous there's more of one class it's probably going to go much better for cryoem even assuming you don't have alignment to issue okay thank you