 This is going to be slightly different from the previous keynotes where I'm literally just going to follow a single protein target To everything we did to it to get actually our structural information, which is at the end of the day what we really wanted so Effectively this project It started even before I came to Paul Nissen's lab as a postdoc It actually preceded me and it continues now in my lab that has just set up here at August And we learned a lot about pretty much ourselves the protein and actually how we do structure biology by tackling this and for me one of the things was what to do when you cannot get good crystals and So as you can assume from the title we actually turn to cry we em but as an overview of the talk because a lot of How to say strategies that we have put employed in the downstream processing of the sample actually related to protein function I want to give an up an overview of lipid flip bases and then how we actually went around tackling the project From the problems and the questions we wanted to answer so As many of you know the eukaryotic membrane is quite complex and one of the complexities is the Non-random distribution of lipids across the membrane So if you would just focus on the plasma membrane for example You would have an enrichment of single mining in PC in the outer leaflet of the membrane Whereas the anionic and the anionic lipids PS and PI are sequestered in the inner leaflet along with the phosphatidin etanolamine And this is not by chance elements are actually established to lipid synthesis But it still requires a variety protein machinery to actually maintain and establish this lipid asymmetry There is three different protein families that move lipid across the membrane You have the scrambleses which drive bi-directional movement across the membrane. They're typically calcium activated You have the lipid flophases which are from the ABC transporter family that drive the exoscelular or the export of lipid to the outer leaflet And you have the so-called lipid flip bases which drive the inward How to say the inward directed translocation and both of lipid flip bases and flophases are powered by ATP and the lipid flip Lipid flip bases belong to quite a a broad family of membrane transporters called p-type ATPases They are further complicated by the fact that most of them require an accessory subunit from the CDC 50 family To actually be expressed and properly localized to its correct membrane as I mentioned the transporting unit is a p-type ATPase also called the pump and mainly from their Role as iron pumps. So sodium potassium pump and the sarcoplasmic and sarco-interplasmic reticulum Calcium ATPases are very very prominent members of this family Lipid asymmetry actually is very important. So it first of all serves to Define different compartments or different chemistries for the membrane surface. So if you have a how to say In large concentration of negatively charged lipids on the inner leaflet This serves as a docking site for polypacic proteins to actually recruit onto the membrane surface To its role as a lipid transporter It can actually form low induced local membrane curvature, which gives rise to nice and vesicle formation Movement of lipids across the membrane are very prominent in that lipid signaling for example surface exposure of ps triggers a variety of Biological processes chief amongst them blood coagulation and apoptosis This regulation of this asymmetry actually has detrimental effects on the membrane So we decided to look at a well characterized organism in this case sarcoplasmic sarcomycin Cervicia and We're looking at the lipid flipases in sarcomycin Cervicia, and they actually expresses five different P480 pieces of lipid flipases. So DNF1 DNF2 or plasma membrane localized they have a beta subunit called M3 and they have a Substrate specificity ranging from PC lyso PC and also glucosoceramide DNF3 is localized to the Galgite it itself has substrate specificity for phosphatonecholine and phosphatonectonolamine NEO1 is the How to say the black sheep of the family in the context that it does not require a beta subunit for localization And you will hear more about this protein tomorrow when you'll have a presentation by Lena Marie Christiansen Also from Polynesian lab. It is localized to the Galgite and the substrate specificity is unknown However, there are direct evidence for indirect evidence for PE translocation DRS2, it's beta subunit as a meant is from CDC 50p it's localized to the trans Galgite in early endosomes and has a substrate specificity of a Phosphatone serine and phosphatone ethnolamine and it was the anionic lipotransporter DRS2p that we actually decided to focus on in the study So This is work that we've actually been in a long-term and very fruitful collaboration with the lab of Guillaume Renoir at CES at Clay and they actually generated the protein constructs Or at least the parents protein constructs using this study. We've made some modifications ourselves along the way For specific to answer specific questions and actually so this is a plasmid that we can use Based on the PYBP60 backbone that we use for homologous expression of DRS2 in saccharomycese servicia And actually we not only explode express DRS2, but also CDC 50 on the on the same plasmid DRS2 is actually exists as a fusion with a c-terminal bad tag that can be cleaved The CDC 50 actually has a his type if we require autogonal purification or just a way of detecting the protein using blotting the strains that we were initially using actually were quite unique and They were actually generated Sorry the technology we were using was actually quite unique and it was generated by Guillaume back in 2002 if I recall and It required so the plasmid has been modified to also include galt the expression of galt-4 behind a Galactos-inducible promoter, which we also use to express DRS2 and the mindset behind this is overexpression of the galt-4 encryption factor would serve to boost expression of our gene of interest and This is actually technology that was used very early in the expression of Circa, so the calcium ATP is in saccharomycese that Rosslin mentioned this morning So this is this is the technology used to actually express the first memory recombinant membrane protein in use in saccharomycese servicia We also looked at a pep-4 deleted stream and I will mention a little bit more about that in the next slide about why we needed this and The expression system itself is actually relatively straight forward. We did Precultures in minimal media, whereas our main grow-ups were actually using Rich media so the protein We have 36 hours of 28 degrees and we actually use a two-step induction For to induce overexpression of our protein of interest in this case DRS2 So we have after 36 hours a first induction with 2% galactose Temperature is reduced to 18 degrees for expression and then we give it after 12 hours another 2% galactose before harvesting the cells after about five five to six hours and Isolating the membranes and we actually use glass beads in a How to say a rock miller to actually lyse the cells The construct I mentioned had a bad tag fused to the sea terminus and this was actually cleaved by trombone Trombone came with its own a unique Difficulties and the fact that it also actually cleaved the first 104 amino acids from DRS2, which as The initial plan for this project was crystallography We actually thought this was not a bad thing and also when we expressed a variant that did not cleave And terminus we saw no difference in the activity So the removal of these 104 amino acids had no effect on the actual activity of the protein at least that we could detect so when Miriam who's a postdoc in the lab and the Christina during her masters actually tried to express this construct in Saccharomyces we saw that the galfor strain or the galfor Strain also actually had a A new effect if you will well it had a We get overexpression of the protein You have the p2 and p3 membranes the p2 membranes are the membrane pellet. We actually isolate 20,000 G or the p3 membranes and those we isolate That in the ultrasound to feed if you will and as you can see both membranes contain the protein of interest But we also have a significant amount of unglycosylated or partially glycosylated CDC 50 whereas in the Delta pepper strain we actually The yield was a little bit more reduced, but we were able to get the fully glycosylated CDC 50 So we decided to go in forward. It was better to have the more homogeneous sample We also did detect some truncation of the protein However, when we purify downstream we will we lose the lower band because it it seems to be a c-terminal truncation of the construct Or not solubilized In terms of purification, it's a very very simple procedure We use DDM to extract the rest to pee from the membrane So we end up with DDM solubilized protein, which we then batch bind to streptapine and resin So the bad tag is the biotinulated acceptor domain. It's biotinulated in vivo in yeast and Saccharomycea is actually quite good in the fact that I think there's only three other proteins that are natively biotinulated in Saccharomycea. So in terms of background for binding onto the resin, there is not a huge amount of protein that we have to compete with However, as the interaction between streptapine and biotin is Incredibly strong. There is no way to elute the sample. So we chose to effectively cleave the protein from the resin But once bound to the resin we're actually able to We lipidate the sample if we need to and actually do detergent exchange, as we didn't have to worry about the protein leaching from the colony Trombone cleavage released our protein as a Matured version of our protein that lacks 104 residues at the end terminus, which we can then use downstream for biochemistry and structural studies In terms of yield I look back to the notes I think in our hands depending on the construct we obtained something between 0.5 to 1 milligrams of pure protein And I mean after size extinction per liter of culture So In terms of biochemistry It has already been established that this protein is Activated in a way by interaction with phosphonisotide for phosphate. It can interact with others But the strongest are the most significant activation is with PI4P This is a Typical purification or at least purity of the DRS2 sample after the bad day So on one hand, it is a very costly method for producing Protein because you cannot regenerate the resin but as DRS2 was so well expressing We didn't have to use a lot of a large amount of a resin to trap the protein so it actually became a trade-off between getting good quality pure protein and The cost of the resin and the protein one in this context giving what we were using it for Here you can also see that in Detergent alone with without PI4P the protein is not active even when we give it its substrate This is work from Guillaume's lab summarizing this paper and once your PI4P you get a market How to say interaction sorry activity and we measure activity to the liberation of phosphate By the ATPase domains in the in the in the transporter so we don't directly monitor lipid movement here in my cells more the ATP turnover of the protein which should be coupled to lipid transport So phosphotide is clearly the substrate of choice and it needs PI4P for activation and they'll come back to this concept later So we decided to at the start we wanted to crystallize a protein called lab itself as a quite a famous and prominent Protein by a structural biology lab, but and they've also developed their own Methods for crystallizing embryo proteins. I came from a background of lipidic cubic phase with mark and calf re So we tried that method also and we were trying to make a lipid based method work We ended up also trying other in circle methodologies. However, nothing really worked well for us So highlight is a high lipid high detergent method in this case we add a high amount of So we use the high amount of detergent in the protein sample the solubilize a lipid film that has been evaporated on the base of a glass vial or a glass tube and this these lipid detergent Protein mixes can then undergo crystallization and like the cubic phase. We typically would get these layered stacking of crystals This is the method at least what the method generates when we set up crystallization and indeed we had Very nice beers and jumping up and down when we got our first crystals And we brought them to the synchrotron so you can see these are quite small crystals I think they're about 20 micron in diameter and We mounted them on a I think it was this is a diamond Light source and the best of fraction we could see you extended to about 14 angstroms No matter what we tried to the protein and the crystallization conditions the lipidation ratios We could not actually move this beyond 14 angstroms and I must say at this point the protein was in element G So moving forward we were trying to think What wasn't working so one of the Elements of the US to it has quite a long sea terminus I think it extends by about 110 residues 105 residues In that domain there is meant to be a lipid binding site at the time There was a vision to be a lipid binding site namely the pf4p was meant to bind the sea terminus And there was an auto regulation site for the protein So together with Guillaume, and he's a very talented lab We decided to look at first limit of proliolus is to get a sense of how much can we cut off the protein by Exploiting it to a variety of different Proteases and this is just one example We use karma trips and trips and but this is one example We're just problem where you can see the parent protein So this is actually a different construct which had an N terminal tag So you purify the full protein and it was cleaved with tev So we didn't have any nascent cleavage of the N terminus when we exposed to the trombone you get a very rapid cleavage of the N terminus at So the first one in forming was it's leave And also we get a slight presence of a slight lower band when you increase the Incubation time we really didn't change the ratio Maybe you can say we got this lower band a little bit more But actually once we added pf4p, we got a market reduction of the upper band to this lower molecular weight species which comprised based on mass spec residue 105 to 1290 and so this was This residue is actually even after the proposed auto inhibitory site and What we could also see was based on the phosphorylation that this very low cleaved Fragment was actually the most active form of the protein as in is the one that was the most readily dephosphorylated and These were using radioactivity And of course the truncated sample over here was the most active and What kind of struck us as remarkable at the time was it still required pf4p for activation But the overall activity was a lot higher meaning that It was right We were actually previously only detecting the activity of a very very small sub-population of cleaved protein in our actual measurements or matured protein in our ATP is measurements So because we could not Use limited proliolosis to cleave the C terminus Further we then try to actually just engineer a construct So here is our delta N104 construct where we had trombone cleave about the N terminus and C terminus And we decided to modify it to include another trombone cleave site and We we look we put this side very quite close to the To the C terminal end of TM10. I think it was around the yeah It was a residue 1247 so we remove all the pf4p binding motif or the proposed pf4p binding motif and the auto regulation out the auto regulatory site and Indeed this double truncated construct was more active than delta N104. So this species up here and Also, it's still required pf4p and that baffled us because the reported site was meant to be now removed so this opened up questions of where is pf4p now binding in the protein and So we needed to actually get more structural information and this gave us quite a nice tool Now where we actually have a construct where we don't have an order Auto inhibitory domain so we can now between our constructs start trying to map the transport cycle of the protein with a focus on the Auto inhibition and regulation so as I mentioned crystallization studies didn't Work out as well as we hoped So what could we do so we knew we had the protein solubilizing DDM and we knew we could actually exchange that to lmng with With no impact on at least function that we could detect and it was still quite well behaved based on gel filtration we get a monomeric protein and Using negative steam EM we were getting predominantly monomer As a result we were also detecting some Dimer formation, however the predominant species was monomer. So this was one avenue we could exploit Around that this was around the time of the resolution revolutions. So this gives you a time frame for when we were actually tackling this and This is right off the back of sure shares publishing the beta galactus. Sorry beta gamma secretase structure and in their paper they actually used amphibals to Act is a non-detergent like my soul across around the protein and we actually established a Exchange for DDM and amphibals will be just incubated the protein with an excess of amphibals and Used bio beats to actually remove the excess detergent and we're able to get very nice stable Particles and amphibals, which we actually were able to turn into 3D reconstructions using negative steam that approximated the size of our protein and Indeed we also looked at Membrane based nanodisks. So in this case I looked at solar pro nanodisks to try and generate a Version of the protein that is in the membrane environment Because if you remember these proteins are not already regulated by lipids, but they're also transport lipids So we thought using a lipid based approach will give us the best chance of actually tracking such a complex so based on Yeah, I'm sorry negative steam EM analysis we couldn't quite determine which sample to go forward with and Based on timing we decided to look at the amphibals sample first And I will not say this is the best grid, but it's actually one of the better images we had for this sample and To a lot of very hard work by Malena In collaboration with Arnold Moller who was originally at August University We moved to the Max Planck in Frankfurt at that period of time We're able to get the best resolution you could get was around five angstrom and that was a global EM resolution There was regions of the protein that were a little bit better But most of the regions was a lot worse than five angstroms meaning the novel building would have been a challenge. So we could have Approximated the ATP is domain or the ATP is subunit based on public structures of circa Heavy metal pumps, sodium potassium pump. However, the beta subunit here would be de novo. So we needed more More information to actually help drive model building so at this time we were actually shipping the protein to Frankfurt and they were making grids and We decided to try and do it ourselves because we had our own vitro bottom or microscope slowly came online with a K2t Camera and actually during the course of our optimizations. We actually noticed that the LMNG sample was by far the best sample we had so we actually Put this on a grade. We only needed really low concentrations. So we were Taking it from the column at about point two point three makes per mil As the peak fraction we concentrated perhaps even two-fold to get the concentration You needed to see this particle density here, which is quite remarkable in this context And I think it's a testament to the low CMC of LMNG Even from the 2d classifications We're able to see elements of secondary structure for a protein that got us Gd with excitement. I won't go into that and we're able to see all the features that we would expect of our protein And this is work that was summarized quite beautifully in a major paper But also equally more impressive in malena's iphd thesis where we're able to actually solve the first structures of the lipid flip is And if you indulge me for a couple of slides I want to show how we what we learned from the protein and also how we modified our experiments going forward And as you can see you have the cdc50 and pink the transmembrane domain in tan Phosphorylation domain, which actually gets translated into phosphorylated during the transport the actual domain which drives the phosphorylation And the nucleotide binding domain, which of course binds ATP for phosphorylation and in green here We actually resolved the auto regulatory part of the c-terminus Not only did we get this auto inhibited form Because em we didn't have to make crystals once you have a good condition We actually can exploit different conformational states We got an intermediate resolution for a intermediate complex if you will where we tried to just add pf4p the cc could be destabilized auto inhibitory domain And actually just adding pf4p on its own did not Give us the desired effect. So we actually went back to this C-terminally truncated construct. I mentioned earlier that we made for crystal studies that amounted to no better crystals than 14 angstrom. So what we could actually see was We were able to trap this at a very good resolution that allowed the innova building Subsequently malena has now actually captured most of the rest of the states in the transport cycle and actually the paper came online on Friday in jmb So if you want to see how this protein binds lipids and how we reset how it actually gets phosphorylated These states are now present in In this jmb paper So in terms of what we saw we actually were able to resolve the auto regulatory domain and we're also able to Show that okay, so this has a motif This this motif here that is roughly conserved in a select human variance And actually subsequent structures of ATP-8a1 from the Nureki lab actually supported the binding pose of the C-terminus against the P, A and N domains in particular this these aromatic residues Sampled the ATP binding pocket just arresting the ATP activity of the protein What was also interesting is we actually found that the binding site for pf4p Was not present on a this ordered part of the C-terminus But actually integral to the protein. So we actually see a binding between the cdc-50 subunit and tm-10 and the core of the protein And what was interesting is it actually served to order or at least Maybe fold but definitely order a amphipatic helix Of this that is present on the C-terminus And selection for the for phosphate is actually via both this tyrosine and histidine residues So these were mutations done by a collaborator and we could actually see that it completely abolished ATP activity Even in the presence of pf4p in inferring that these were what was making this selector for pf4p So as you noticed I didn't talk about the transported substrate And the best we got from our structural studies was the truncated form of the protein gave us an open configuration So this is um something we trapped with a phosphate analog called bringing fluoride And this actually arrested the protein in a conformation where tm-2 has moved away from the helical bundle here To make an opening for lipid binding However, we did not resolve any lipids in the cavity even though we actually Provided them. So this beg the question. How can we push this? um, so bear in mind this was before the Papers from the noreky lab. So we decided to Look at as I mentioned none of this reconstitution um, in this case we use solopro, which is came from uh against frownsville I think we use at the carolinska and the technology is really reliant on the binding of supposing a so it's lipid binding properties to bind lipids that surround the membrane protein In this case, we actually pre incubate the protein with a mixed micelle of lipids and detergents So the lipids in this case brain polar lipid extract will equilibrate with our protein of interest I chose brain polar lipid extract because it was a blended lipids But was also quite enriched in phosphatidylserine, which is the preferred substrate for the For the protein that we were studying And when we incubate these lipid detergent mixes with the protein, we subsequently add supposing a Then we dilute the sample to below the cmc Concentrate and inject for size exclusion And this is a typical preparatory scale Reconstitution for another lipid flipase. So this is the bovine ATP82, which is an ortholog Of doris 2 I have similar How to say chromatocrones also for doris 2 but the subsequent story that I'm going to show I do not have that data yet for em So we are able to get a stable and complex. There's a monomeric peak Negative stain analysis was good as was cryo analysis. So you can actually see here We have the similar features for the monomeric protein in a in a nano disc So this is the data that we have which roughly around 2.4 angstrom for the structure, which is substrate band You can see all the features that I talked about before with p-type ATPases and they're color-coded as previously So you have the ATPase domains here in yellow blue and red the cdc 50 protein is in pink And the actual lipid the ps is here in green. So you can actually see it's Penetrating the cavity that's formed Um between tm2 and the bulk The bulk helical bundle of the transporter and so it's predominantly only the head group that's encapsulated And likely that makes sense because the specificity is conferred by the head group and the axial change remain continues with the membrane Uh during transport and this figure over here. You can see the extent Of the nano disc. So it's actually quite a tight nano disc around the protein Um, it there is also partial ordering of the supposing molecules in the nano disc But not enough to model them convincingly Um, so I think uh in closing I can show that T-urus 2 and cdc 50 has been overexpressed in the sarcomaic esophageal. We get modest but yields for structural studies For definite em that's more than enough crystallography required Large-scale expression it can be purified by a bad tag which is cleaved from the strapped avenue resin And we evaluated a number of different memory memory memetics for urus 2 and related protein for structural studies And in terms of the biology of the um Target, we actually divided the first structure and insights into auto regulation We have now moved into looking at how substrate is bound and how the transport mechanism actually Um, how we undergoes transport for these substrates And um, of course, uh, this project was a testament to the collaborations So first you have paul's lab. I'd like to definitely mention Milena who I pointed at a number of times Jacob uh, also started the project about a year before and he was focusing a lot on the crystallization So a lot of the optimization and and engineering was actually done in yakup's hands Miriam actually was the post-op who previously had in Milena Marie's working on other lipid flip faces to get over paul and myself And she will talk tomorrow um Our project is was built through a very strong collaboration with the lab of guillomene and what c a sec l'ay I would like to highlight cedric and especially t-bow some of the slides are even from t-bow today And t-bow was actually currently in augs working with paul and of course the latter part I went to the show was our collaboration with um Jens peter anderson in biomedicine on the bovine lipid flip face my cross to be was done between um How to say the max plan for biophysics in frankfurt referring to hubrant and anna moeller and uh A lot of what was actually done together with the bill post-op and anna's luck And em at home is done under the direction of thomas bosun who's our facility manager Ramon is maintaining microscope in yesper if anyone knows yesper You know that you cannot do any structural studies without yesper. Um, he maintains all our computation here at opus um, we also had initial screening at the mykosby facility at coban hagan and Results, I didn't show but we have some nice, uh, Electrodynamic studies on self-trace specificity done by young mac when he was with uh, christin lindoff larson As a shameless plug my lab is actually open for business So if anyone's interested in doing post-op, let me know and I thank you all again for your Attention and forthcoming questions. Thank you Thank you joe for that very interesting journey through the one flip face journey We have a few questions here. Um Saba wants to know can you comment? Bit more on really pedation the of the brown protein So how it affected the activity of the complex Okay, um, how it affected the activity is very hard to tease out the x because we actually just equilibrated it with, um Phosphatillic serine just a small amount of phosphatillic serine And helped preserve at least to what we saw early in In our early experiments and the stability of the protein But uh, as it went on once we improved and optimized the expression We could actually do the step without the bedation. So we just added back a small amount of ps if I recall There were two and a half I I would be surprised if there was a liquid molecule for myself based on those numbers I think it was a very very trace amount ps And as ps was always present then the substrate the basal activity of the protein will always be a little higher than without substrate, of course Hopefully that helped we also equilibrated with pi4p and there we actually use quite a lot because we need the pi4p to So we when we had the engineered construct with the multiple trombone sites We first cleaved it off the protein. Sorry, you can cleave the protein off the resin Captured that and then incubated with a large excess of pi4p to be able to destabilize the auto inhibitory domain for subsequent further cleavage so So yeah, and then you have pi4p presence and it was quite slow to diffuse in lming g So it was always going to be affect the activity of the complex. Okay Uh one more question Martin Keffre wants to know do you ever lose the accessory protein? um good question, um to our extensive EM analysis, we have never found A protein that was lacking cdc50. I don't know if that is a direct answer for the question They also don't stain equally on sds space. So it's hard to quantify based on just simple sds page analysis But based on our handling of the protein or experience with gel filtration and the variety of negative stain and cryoem studies We've never seen just the isolated um Pump if you will so I can I would say with some confidence. We never use the cdc50 subunit um sabba asks how the EM behavior of the complex salipro sample really really good So we were able to use concentrations that were even slightly lower than drist 2 in lming g So we're freezing. So this is total protein concentration of 0.4 mix per mil and we're getting similar density of particles and the lipid types we use was brain polar lipids you can buy them from a Sigma so it's I think it was the Fulc type 4 fraction it was the original lipids used in the supposing methodology from jens fraunfeld Yeah, then can you shortly Elaborate the methods to identify native lipids as this would be the primary requirement for highlight method If I understood correct Yeah, that's a good question. We actually didn't look at evaluating native lipids But uh, we do have an ongoing collaboration with carol robinson. We actually look at native lipids They at least co-purify with the protein But in terms of the actual lipids, we use the screen. We primarily look at Phosphatidylcholine so either native extracts of soy pc or the synthetic do pc as a As at least a basal lipid for highlight. We then depending on the How to say informational state we were looking for we added the substrate and or regulatory lipids to the mix But we always have kept the background of pc Okay, I think we even tried to bring polar lipids at one point and nothing came from it. So, okay Okay, thank you joseph for your contribution to the links workshop