Transcript for:
Towards a Virtual Bacterial Cell Envelope

so uh many thanks to sarah and sarah uh for the kind invitation to share the session um uh it was on the condition that i'm relatively relatively polite i think so um i will do my best um it's a pleasure and of course norma to introduce professor simon khalid uh for her research seminar today um so just a little a quick bit of background so after obtaining her phd with mark roger um i got i first met simon and got to know her in mark samson's lab when she did her first post talk uh i think 2003 2004 or something like that yeah and uh i think it's fair to say we've been great friends and collaborators ever since who are mutual interests in bacterial membranes and also our fellow suffering as uh liverpool supporters um so [Laughter] um and uh yeah after her post talk simon's career progressed really rapidly she quickly established a hugely successful research group at the university of southampton and i think it's fair to say that she is the world leader in multi-cell multi-scale simulations of bacterial envelopes membranes and and cell component cell wall components um i'm also very fortunate to be able to be a long-term collaborator of cybers and when she's not watering her pumpkin patches in animal crossing um we are co-supervising uh several phd students together and and that's i think resulted in lots of exciting discoveries as well so i'm very grateful to have been involved in in all of that um so without further ado i will hand over to you simon and i think your talk today is entitled towards a virtual bacterial cell envelope adding the biological complexity to biomolecular md simulations so over to you cool great well thank you very much um thank you again to both sarah's and also to pete for the very generous um largely generous introduction um as he says we've known each other for a long time um okay so i'm just going to share my screen right can everyone see that no i can't yes okay uh yeah yeah okay cool let me just start with that okay so um i'm going to start off with just three slides firstly um discussing with you and describing to you heck biosim um just because it's a sister organization of ccb biosim so i'm the chair of hex biosim which is a uk-wide um consortium that promotes promotes the use of the very high-end computing resources by the biomolecular simulation community we're an inclusive open group and we we started off really by providing time on the national supercomputer called archer which i've no doubt many of you will have heard of um but in recent years i guess really over the last 18 months 18 months or so that's also begun now to incorporate the so-called tier 2 machines and i'll describe on the next slide what i mean um by tier 2. so basically there's a bunch of resource that we have that we can provide for biomolecular simulation um research so for example if your local university cannot provide you with the resource you need you don't despair likely we can do something to help but not only do we provide hard sort of time on these super computing facilities we also provide support help you with best practice and also expertise with if you've got a favorite code that you want to use how you might want to port that over for um high performance computing uh resources so we've got lots of technical support we can provide too i've given at the bottom there um uh you the url for our website which does contain a whole whole bunch of information let's give you a snapshot from that website here where we we have online forms where you can apply for time on archer you can apply for time on jade at the moment jade is a a gpu cluster which is largely used for machine learning but there's a element of it which is reserved for by molecular simulation so if you apply if you'd like if you have a code that's optimized for gpus then actually working on archer which is cpu based may not be the best thing you may prefer to go for a gpu cluster so it may be worth you talking to us about applying for time on jade i've given you there a little snapshot of the management committee group the management group is actually built up of bi-molecular established by molecular simulation groups from across the uk and the way we've put together the group is to try and represent the pop users of the popular molecular dynamics codes people who apply those codes in a range of different ways so for example there'll be applications that are very capability based and there's others that are very capacity based so for example there's those of us within the consortium who like to simulate very large systems for long simulations but then there's others who do lots of simulations of smaller systems but they also need loads of resource but just because of the sheer number of simulations they run and it turns out that different architectures are suited for different types of calculations and so the management group represents um as much of you know the breadth of the field as we can possibly get i've highlighted in particular dr james gabby rayette who's our co-sex support so he's based over at stfc so james is really i guess the um the heart of heck biasing he um he's able to do benchmarking calculations he provides support with porting code over helps he's developed i've just given a little snapshot it's a little bit small to see which i apologize for this thing called the heck time calculator so this is an applet that james has developed whereby the user can input the size of the calculation the type of calculation they'll run the code they would like to run that on and it provides a good estimate of the time of resort type of resource that will be needed which is then needed to input into the application form so james in a way i mean i guess sort of my name's on the grant but really james really runs um biosim and so he's usually a first well certainly the first port of call for me when we get queries and finally the last slide is i just wanted to um provide a brief overview of what's coming ahead within the uk landscape for high performance computing and i've called it exciting times ahead because it really is in things changing so for example the national facility archer is right now actually in the process of being upgraded to archer 2 which will give us a huge boost in terms of capabilities but also alongside that so that's the national facility is tier 1. tier 2 is the level that sits beneath that and these are smaller machines and there's a whole range of them around the uk they're smaller than the national facility but they will be larger than your local university machine and these the second round of these machines are now beginning to come online there are a whole range of architecture so there's some based on cpus there's gpus there's arm architecture there's some that are hybrids of the different um different including different components and the way things are moving is that in future it will simply be a case of matching up the type of calculation you want to do with the architecture that's best designed for that and so how that moves forward in terms of what do you the user have to do to apply for these resources is still being worked through and i urge you to keep an eye on the uk ri website and also the heck biosim websites for updates we'll be um we'll be providing you know sort of i'm on various epsrc panels as long side being on being on biosim so we are in a good position to keep our community updated so just for completeness the um and context tier one is the national facility tier two are these additional machines that sit somewhere above your local university level where tier three will be your local university machine and of course you know not all universities have access to the same facility and that's why it's really important we provide access to tier one and tier two but above tier one there's also tier zero which is sort of at the international level for example you can apply for time in praise in europe or inside in the usa where they use various and department of energy and defense machines so i guess the bottom line is that as we move forward there's a whole bunch of resources available to you if your university machine is not providing you with the um the resource you need to run the calculations you want please don't despair get in touch with us keep an eye on our website and we'll be happy to ask answer any questions and to help you sort of find the right um the right machines the right architectures for your applications okay i think i'm going to stop talking about heck biasing there but i guess i'm i mean could we take questions on this part now maybe yes no yes people want to ask questions yeah there are no questions as yet but okay well maybe that maybe gives people a chance to think about it and they can ask them all at the end then if yeah like shall i just move on to the science part okay yeah i think it makes sense okay cool so folks if you have questions sorry sorry there is one question now if you want to take it yeah sure um so is there any european-wide project so there's price right so price provides access to resources across a bunch of machines um within europe so um to be honest i personally have never applied for praise time but i know many many people who have and many of those machines are give give bigger chunks of resource than archer will so there's marinostrom in in spain for example i think they have a more formal application procedure than we have for archer but it's there's a lot of help available and my understanding is that our so the uk involvement with price will not end as we brexit okay so take a look at their website there that that project is there within europe and of course there's other things like um just trying to remember the name of um there's a scheme whereby you get access to a european machine and um ukri also provide resource for you to go that's not you it's combination of ukri and eu where you can then go and actually spend time in the country where that machine is based and they'll pay your living costs as well i'm just trying to think of the name of that scheme sorry it's just completely gone out of my head hpc europa and that's right there's there's there are projects like that available yeah and i think it's definitely worth the uh contacting the people uh go to the website and contact them because even singapore that there are various schemes and even singapore has been able to access um computers via place in some situations so it's certainly worth talking to them i think yeah absolutely i mean these machines are there and they're there to facilitate good good science so if you if you have a an excellent proposal and you feel you don't have the resource available locally then please do contact people yeah that's right this day and age that shouldn't be uh that shouldn't be something that limits limits research uh we do have one other question which is uh is tech biosim open to all universities or just universities within the uk it's i'm afraid it's open to just universities in the uk but i think um for example insight in the u.s is open to any country as far as i know so i used to be on the panel for that actually for a few years ago now but certainly that was not just for i mean it's it's the machines are provided by uk department of energy and defense but nevertheless the resource is available for non-us people okay yep i think that's all for the questions for the moment okay great right let's move on to the science okay so in the next i guess 35 minutes 40 minutes or so um i'll be discussing some work we've been doing in my group for for a number of years now i guess pretty much from when i really set the group up in about i think when that was about 2007 so it's sort of work that projects that have developed over time and what i'd like to discuss with you in particular is going through this all the time is when we do buy molecular simulation sort of asking ourselves why we're doing the simulations what we hope to get out of them and how do we interpret the results can we well i certainly believe we can over interpret results especially when the systems we're looking at are biological because there's so many variables and there's so many things we don't yet understand so some of this is is inevitably there'll be a few sort of success stories but littered in within that will be area for projects where we went off the wrong track and trying to sort of give lessons where i hope others will learn from and not make the same mistakes we made a quick biology lesson um i hope peter's listening in particular because he's a bit hazy on this stuff um so the e coli cell envelope it turns out certainly for gram so because e coli is ground negative bacterium the cell envelope is quite complex and that for me this is of endless fascination given bacteria are single-celled organisms and we of course have millions of cells these bacteria still manage to evade all of our pathogenic bacteria evade all our attempts to try and combat them with antibiotics for example right they're still managing to somehow come up with resistance mechanisms and much of this comes from the complexity of the cell envelope so you can think of it as a sort of sandwich where you have two membranes so there's one that's on the outside that's the sort of the first physical barrier that's encountered by anything that tries to enter the cell and then you've got the inner membrane which is on the side of the the one that's just that's connected to the cell interior and in between and if you think of those membranes as the bread of the sandwich in between the filling is an aqueous region which is called the periplasm now the outer membrane is enormously complex in that it's asymmetric in terms of its lipid composition so we all know that biological membranes have two layers or two leaflets with lipids well in the outer membrane of gram-negative bacteria we have a a very complex molecule called lipopolysaccharide it can have between six and four hydrocarbon tails it's connected by head groups which have kind of varying amounts of phosphorylation and then above that has levels and lots and lots of levels of sugars and so in a way they encompass this molecule encompasses pretty much all of chemistry it has hydrocarbon regions it has charge groups and the phosphates it has chirality in the sugars of course also in the sugars it has the capability of hydrogen bonding and so there's a lot of chemistry going on and ignoring that means that all this important chemistry is ignored the inner membrane is by comparison more simple it has these um phospholipids on both sides and of course there's proteins jam-packed in both membranes the region in the middle the periplasm as i said this is aqueous but it contains what's called the peptidoglycan layer this is also called the cell wall and this provides a sort of skeleton if you like it gives rigidity to the cell so this arrangement of the two membranes and the peptidoglycan it surrounds the entire cell okay so you have to breach this defense if you want to gain access to the inside of the cell which many antibiotics do antibiotics work by either breaking the cell envelope destroying that or by getting through it into the cell and interfering with some process inside the cell and before go on to talk about the simulations we've done i just wanted to put this slide up here given this is a training week to ask well why do we want to do simulations clearly one of the first or one of the most obvious reasons is that we want to rationalize and understand at the molecular level some very specific experimental observations often experiments can't reach the molecular or the atomistic levels they get some results and then call upon simulation experts to help understand and rationalize why they're seeing those results it could be that we want to study some system under conditions that are difficult or expensive to achieve using experimental methods so difficult to achieve pressures or temperatures for example and it's cheaper and easier to use computational methods but i think increasingly important is that actually we want to develop and put forward new hypotheses we want to do curiosity driven science just to ask the question let's see what we find i think that's perfectly okay um from coming from a very traditional chemistry background we're always told that an experiment is done to test a hypothesis but actually i think experiments can be done to put forward new hypotheses i mean many of us are here doing science because we were the kind of kid who would pick up the pebble to see what was underneath right we didn't know what we would find there wasn't a specific question that was being asked and so there's this idea and to put forward new hypotheses where you don't actually have a clue what you might find and you do the simulations to get to that point and that leads us to the that leads us to another point which is it's equally important for us as simulation people to drive projects for us to come up with the hypothesis that we then give to our experimental colleagues and say you do the experiment to prove me right or wrong or to tell me why this is happening rather than it always being this other way around where the experimental guys go and do the experiments and then come to us and ask us to rationalize and i think as computing power improves we're increasingly in a position where we can we can lead these projects okay so back to the outer membrane um as i mentioned right at the start this molecule lps which sits on the outside of the outer membrane is horribly complex it has these hydrocarbon tails this sugar you can see just one here in the golden red which has got a phosphate group and then levels and levels of sugar sugars so when i started my group there was only one model available for lps developed by turk stratzma and theresa suarez this was using the amber force field um the sugars were very accurate but i think it's fair to say there were some inaccuracies in the tail group at that time my group and independently of us one pill m and jeffrey clouder started developing models of um of lps em and clouder use the charm force field whereas in my group we use the gromos force field and what happened actually is a i think very nice is we're at a stage now where the amber guys have also fixed the issues with the lipid tails such that charm there are now charm amber and gromos force fields available for lps and they by and large agree very well i'm just putting up this one image here of some work we did with jeremy lakey's group in newcastle and i hope this will convince you that our model is pretty accurate so jeremy did some experimental studies where they used chelating agents to remove divalent cations from the head group regions of these lps molecules so it turns out that lps is held together by calciums and magnesiums in the head group region okay so they ripped out those dive cations and replaced them with sodiums keeping the ionic strength the same what happened was the membrane just lost its integrity no longer formed a nice double layer and indeed the same thing happened with our simulations you can see here that after in just 200 nanoseconds when we've removed the divine cations and replaced them with sodium can you see here that the water shown in cyan is beginning to seep in and we no longer have a nice double layer and i think this is a fairly convincing pictorial argument for the um the model being quite accurate i won't bore you with all the sort of more quantitative um models we have okay so that that was all well and good we you know one of the first things we did i said is we developed this model but model's no good unless you use it to learn some new biology and so to do that we um turn to our favorite protein omphae this is a very well studied um outer membrane protein um i've done simulations on this when i was a postdoc our chair pete bond did simulations on this for his phd so we knew this protein very well it turns out that despite it being studied a lot there's um there is some controversy in the literature of how what what its um polygameric state is in vivo is it a monomer is it a dimer and indeed there are still arguments out there from higher olympus i collaborated at oxford cal robinson's group did some mass bank studies and where they found um evidence for dimers of the full length protein so here i'm just showing you the bit that sits in the membrane okay there's a transmembrane sort of barrel that sits in the membrane but actually it turns out that the full length protein also has a large linker region and then a soluble domain that sits in the periplasm and crucially also non-covalently binds the cell wall so we thought well you know this has been looked at by mass spec studies it would be useful for for us to have a look at this in in context with the membrane present and the cell wall to get some insights into how it binds the cell wall so we luckily for us there was a crystal structure of the c terminal domain orbit from a different organism of this protein bound to a small portion of the cell wall i should just say here that the cell wall is composed of it's a polymer composed of repeating sugar um and peptide units so we built in um we built in the missing bits just to give us a complete monomer of cell wall and we ran some simulations in particular the crystal structure showed two salt bridges the two electrostatic interactions between the protein and the cell wall and we found as we run our simulations we we picked those interactions were very very stable which gave us some confidence that you know the simulations were accurate and we were binding the protein in the correct regions interestingly we found that as we run the simulations of the monomeric protein after a while the linker region of the protein contracted and the c terminal domain which we know is bound to the cell wall and remember the cell wall is a big sheet that covers all of the bacterium this small portion of the cell wall that we had was being lifted up such that it was interacting with the lipid head groups of the membrane that's just puzzling because we wouldn't expect this to happen we know from tomography tomography data what the separation is between the membrane and the cell wall what we were seeing would lead to distortions of the cell interestingly when we sat around simulations of the pro protein in its dimeric form we did not see the interaction of the cell wall with the membrane and the reason for that is that the dimeric interface is is hold the dimer is holding these c terminal regions together so they can't lift up individually and interact with the membrane so that led us to well hypothesize that well could it be that when so when according to our results anyway when you have the protein in its monomeric form you see the cell wall being lifted up and interacting with the membrane so if it was if it was indeed a big sheet you'd get something that looks like the um schematic diagram on the left there a huge distortion in the cell wall whereas when we have the dimer we don't have the distortion in the same way so what does this mean does it mean that binding of pectidoglycan by the monomer causes in real life causes severe distortions of the cell wall does it mean that unpaid when it's in its monomeric form doesn't bind the pet doesn't bind peptidoglycan at all i mean looking at tomography data we hadn't ever sort of seen such big undulations or such big distortions of the cell wall so it was difficult to rationalize or to really that the protein would be causing these severe distortions but of course in the back of our minds was always that is what we're seeing an artifact of the fact that we've used really short strands rather than a large sheet of the cell wall and i guess this takes me back to what i said right from the start where it can be easy when doing simulations of biological systems to over interpret what one sees so we've taken a tiny portion of a huge polymeric mesh and we'd simulated it and we were seeing results so unexpected it's tempting to say that perhaps we've you know we're we're seeing distortions of the cell wall or when it's a monomer it doesn't bind the cell wall at all but we have to bear in mind that what we simulated was not particularly realistic at this point we then moved on to make a bigger model of the cell wall so now we did indeed make a big mesh and we bound it to itself across periodic boundaries so it is actually an continuous three-dimension sorry two-dimensional structure um if that makes sense so it's a flat structure it's continuous in 2d we set the system up where we take a monomer of the protein and we started the simulations by placing it at about 90 angstroms from the cell wall we took that distance from an estimation from cryotomography data we ran the simulation and frustratingly nothing happened right we didn't see any interaction of the protein with the so we didn't see any interaction of this bit of the protein with the cell wall at all even though now we've got a more we've got a more complete model of the cell wall but then when we started thinking about it we realized that we had made a big emission we'd missed out this protein here called bronze lipoprotein now this is the most abundant protein in e coli so there's shed loads of this everywhere and we hadn't put it into our simulations so when we went back and did the simulations again now adding in bronze lipoprotein what we saw was that this protein sorry i should say this protein is covalently bound it's the only protein known that's covalently bound to the cell wall in e coli okay so it's linked to the cell wall covalently those bonds don't break on the other side it's anchored via lipidated regions into the membrane as we ran our simulation we found that braun's lipoprotein was able to tilt a little bit it was able to kink a little bit and in doing so it led to a reduction in the gap between the membrane and the cell wall so you know between the protein on bay and the cell wall and spontaneously the protein sensed the cell wall by electrostatic interactions the linker region expanded and this c terminal domain of the protein landed on top of the cell wall and rather pleasingly it picked out the same electrostatic interactions that we'd seen in the crystal structure being really important okay so that so that that was a really you know we felt a nice result with by putting in the bronze lipoprotein this crucial ingredient we've shown that ompay as a monomer does but indeed binds to the cell wall and we did not get that big distortion we were seeing when we only had tiny portions of the cell wall intriguingly however when we looked at the um simulations of the um the dimer the diamond didn't seem to need braun's lipoprotein it could by itself just um if you like drop the anchor of the um c terminal domain onto the cell wall i can only rationalize that as maybe now you've got twice as many electrostatic interactions as you have in the monomer and there's some kind of threshold level and so that's enough for for the protein to sense electrostatically the cell wall and and change its conformation to interact and you know i guess one's tempted again to maybe make a hypothesis that could could this be a in vivo role for the dimeric protein that may be in when the absence of bronze lipoprotein having these two monomers together provides a tight um interaction with the cell wall but it could be that you know we're missing something else in our simulation so again i'm hesitant to um to say that in any with any kind of certainty so at the moment we've looked at half of the sandwich we've looked at the outer membrane and the cell wall but what's still missing is the inner membrane and any proteins in that membrane so just to complete the picture we took a protein called tolar which is sits in the inner membrane and has been shown to bind to the cell wall the protein binds to the cell wall in what's called its open state but the crystal structure shows a protein in the closed state so we excuse me but there was a model of the protein in its open states made by erin cuts in oxford so we asked if we could take a look at this and they they didn't run any simulations of the model they just produced the models so we asked if we could test that for them which they were very happy to um to give us the model so we ran some simulations and these were the first ever reported simulations in which if you like this you've got a little portion or a chunk of a virtual cell envelope we have the outer membrane the inner membrane the cell wall and proteins in both sides and the sort of actual more detailed images shown here where you can see the proteins in a bit more detail all i've done here is plotted in the z coordinate of the center of mass of the protein um on pain in the outer membrane in blue the cell wall in red and this tolar protein in the inner membrane in green and you can see there's three different simulations shown here this is why there's three curves for each color you can see that over um over these simulations they're fairly stable they don't move we're not seeing any distortion of the system they seem to be pretty happy in their orientations and indeed what was really nice was we were able to identify residues in this talar protein that are key for binding the peptidoglycan and these are the same residues that have been shown to be important by the experimental groups so again the simulations and experiments are agreeing which is nice but i think what perhaps what's nicer is we were able to show how the binding occurs so when we run the simulations we place the protein just beneath the cell wall so it wasn't really it wasn't interacting as a simulation progressed we saw these terminal regions of the protein sort of if you like almost reach up and grab hold of the cell wall and then pull that pull the rest of the pro bulk of the protein up a little bit so they form their um full complement of interactions so this had never been seen before right the experimental guys knew which amino acids in the protein were important for binding to the cell wall but they'd never seen the process of binding which these simulations showed so they sort of agreed with the experiments but added on something extra and interestingly we found that um if we look at the morphology of the cell wall itself when we only have the prot on pay bound from the outer membrane side you can see you've got a slight distortion in the cell wall i mean nothing like we were seeing when we only had these short regions but nevertheless it's not flat and equally if only tall r so the protein from the inner membrane side is bound again we see a slight undulation in the morphology when both proteins are present we get a really nice flat cell wall and so again we're tempted to say but we do so hesitantly that it it could be that at any given point we know these proteins are diffusing around laterally in their respective membranes when you get a point at which proteins from both sides are evenly distributed you get a nice flat cell wall but if there are regions of um regions where they're not distributed evenly and that can happen transiently it would lead to some slight um slight distortions which over the fullness of time would likely even out and so i think now we have we have both membranes we have the cell wall so we're pretty pleased that we've got this portion of um a little portion of the virtual cell envelope but i guess one could ask well is anything still missing and it turns out there is it turns out that the um the crowding is missing right so far we've only got one protein in the outer membrane we've got one in the inner membrane the periplasm which contains the cell wall that aqueous region 8-2 is jam-packed full of proteins indeed in the past it's often been thought of as being in a gel-like state and we haven't included those all we've included is sort of the big structures so now what this is and this is unpublished work is under review structure at the moment we've um we've built a system where we've put again on pay in the outer membrane but unborns lipoprotein is in of course but now we've put some um additional proteins into the periplasm um one a couple of them um so pal and lol b are also lipidated so they're anchored into the um outer membrane but the bulk of them sits in the periplasm and this one lol a sits completely in the periplasm we've thrown in to make things interesting we've thrown in this antibiotic polymix in b1 so this is again is an antibiotic we've looked at studied in the past before and again with p bond so this antibiotic is thought to cause cell death by its action on the inner membrane right so to get to the inner membrane it somehow has to go from the outer membrane across the periplasm to get to the inner membrane so we threw whole bunches of this into the periplasm and actually not only are there proteins in these systems but there's loads of smaller molecules right so there's these molecules known as osmolites where we can have yeah there's there's sort of ions there's these osmoregulated um glucans there's urea putricin spermadine that's basically a whole ton of these small molecules so we took realistic mixtures of these um these small molecules and added them into the mix and before i go on to talk about what we found i just wanted to give you a little bit of background onto two of the proteins we've looked at and that's the this protein lol b and lol a now just very briefly the job of these proteins is to ensure that bronze lipoprotein and similar proteins get a lot get to their correct localization get to their sorry get to their correct location they're responsible for the localization so lol a picks up braun's lipoprotein at the inner membrane it carries it through the cell wall and delivers it to low b which is sat here just anchored to the outer membrane lolby then somehow flips it to its correct location the way this is achieved is bronze lipoprotein has got lipid it's got lipid regions there's a hydrophobic cavity in the both proteins where these um lipids are carried okay so the lipids bind to the hydrophobic cavity you can just see here with simulations we did in the past you can see bronze lipoprotein in yellow and in space filling representation you can see the the lipid regions and in both cases you can see they're spontaneously bound into the hydrophobic cavity of lol a now one thing we know from our simulations is that the cavity in lol a is fairly indiscriminate other hydrophobic molecules are shown in red here this is an inhibitor molecule can also bind into that cavity okay so it's just a greasy cavity anything that's hydrophobic can bind into this cavity we know that already okay and that's important to keep in mind when i show you the results that we have for these simulations i'm just going to show you a gratuitously sort of pretty picture here where you can see a large system where we've got all these bronze lipoprotein is in purple in green you can see we the um antibiotic red is the cell wall and yellow are all the other proteins so we've the mixture we have has a realistic crowded volume okay so this is the level of crowding one would expect to see in the periplasm in vivo whoops interestingly what we see is that the antibiotic which is also a lipopeptide which means it has a lipid region we find it also binds into the cavity of lol a um you can just see that here and indeed it does so when it does so it binds to the same amino acids as we were seeing bronze lipoprotein binding to okay so in hindsight this makes sense if you've got an antibiotic with a lipid region why wouldn't it bind into an indiscriminate cavity of another protein that is designed to carry lipids right it kind of makes sense but as far as we know no one has ever considered this for this particular antibiotic and we saw it happen multiple times in our simulations and what we're saying here is could this be one way in which polymixin can sometimes move across the periplasm so we're not saying this is the mechanism what by which the antibiotic moves across the periplasm but we're saying given the complexity of the systems the sheer number of different um interactions and the number of interaction partners perhaps sometimes this does happen maybe it and we're calling this hitchhiking maybe this antibiotic hitchhikes on this um on this protein sometimes oops okay so to conclude i think what i've shown here is that we found braun's lipoprotein plays a key role in maintaining the distance between the cell wall and the outer membrane but we think you know and that was known before that i wouldn't you know i'm not claiming that we're the first to say that we saw it happen in our simulations but experimentally this has been known to but what we have found which i think we're the first to say is that it also plays a key role in facilitating the non-covalent interaction of other proteins in the cell wall right so we needed the bronze lipoprotein to be present for the monomer of ompay to bind to the cell wall we've seen that tolar form stable non-covalent binding interactions with the cell wall and we've discovered various details of that we think polymix in b1 may be transported to the pyroplasm sometimes by hitchhiking on the lipoprotein carrier lol a um and i think the take-home message really from all of this is that the biological details of complexity of the individual molecules but also their crowding should not be neglected when we consider bacterial cell envelope systems especially now given we have access to greater computer power and these bigger and more complex simulations are becoming increasingly feasible and with that and i said we're working towards a virtual bacterial cell envelope model so my plan ultimately would be to have one of these such model at the atomistic level and at the coarse grain level i haven't spoken to you about coarse grain stuff yet sorry not yet that didn't give you um give you heart attacks if she's still going to go on and on um today is what i mean um but it's a key part of what we do in the group and ultimately they allow longer time scales so the goal is to have the the sort of horses for courses really the methods that are appropriate for the systems we're looking at and finally i'd like to finish by um acknowledge acknowledging various collaborators in addition to the ones that i mentioned as i've been going along various members of my group sources of funding computing resource and to leave you with a picture of my group on a normal day at work thank you very much i will stop talking