 I had a very, very old talk on a memory disk and I also had a paper that somebody had recently sent me on some of the work in order to read whilst I was here. So I kind of cobbled together a talk which you see it's got the wrong title based on a number of different things but I'm going to use what I am going to call hybrid technology which is I'm going to use some slides and I'm going to use the chalkboard. And what I'm going to try and do is spend about the first third of the talk talking about work that we've done in the past and work that is published or largely published. And then I'm going to switch to do the last two thirds of the talk and they're going to be on all these pulling experiments that we've been doing. Some of which have just been published but many of which have not yet been published. And the first two slides just kind of give an overview of the fact that my group is interested in a number of different aspects of biological self assembly and in particular in protein folding. And this is my history of the protein folding world in which covers about more than 20 years. So for those of you who don't know very much about protein folding what happened was is there a laser pointer? We started off kind of in the middle here looking at the folding of relatively small proteins. This is ubiquitin. It's 76 residues long and it folds on the timescales of seconds. And then the whole of the, well not the whole of the protein field. An awful lot of the protein folding field went in this direction and they went to smaller and smaller systems and faster and faster folding systems in order to try and get systems which were computationally much simpler to simulate and which folded on very, very fast timescales. So whilst this was going on I was going backwards. I was going in this direction and actually asking the question from all these small proteins how do large proteins that actually fold on the minutes and even hours timescale fold? So proteins with complex topologies, how do they fold? And this is what got me into the area of knotted proteins. So you will hear or at least I've heard a lot people say that the protein folding problem is solved. It's done and dusted and we can move on. And I would just like to point out the protein folding problem has been solved for things that are small and fold fast and don't need any of the cellular chaperone machinery to help them fold. That problem we have solved, how these complex structures fold and of course most proteins are much larger than 76 amino acids is actually a huge problem still and a big challenge. So I'm kind of contrasting these small proteins with these large and sometimes knotted proteins. So this is what got us into the area. So what I'm going to do is give a quick overview of the experimental work we've done on two classes of knotted proteins. That's the three one, three four knotted methyl transferases in the structures shown here. And then also the five two Gordian knotted C-terminal hydro lasers. So I'm going to get some water and this is now where I think I have to go to the chalkboard. So for those of you who don't know, there are four different classes of knotted protein. Three one, four one, five two and six one. And I'm only going to talk about three one and five two. And that is not and so I don't think that four one and six one are not interesting. We just haven't had the time or the resources to study them to study them yet. So and Peter yesterday, he talked about the conference that was here in 2008. And that got me thinking about how much we knew then and how much we know now. And actually what I generally go through life thinking we haven't got very far with this problem. And actually I thought about it. I went back and I saw which of which the papers were published before 2008 and which one. And actually now I realize that actually in eight years we've done quite a lot. So that kind of made me feel better apart from the fact that it did make me wonder where eight years had gone. But anyway, so I'm going to do a quick. This is also a quick history lesson in quarter of eight years. I hope people can see the board in 1992. People speculated on whether you would ever find a protein which had a knot in it. And basically they said it's impossible. They won't exist. Now I'll just qualify that for a second and say that they actually did find one example of a protein with a knot in it in the protein database at that time. And I'm not very good at drawing systems and I'm talking about proteins. Of course I'm talking about open systems, which can be closed. But essentially the example that they came up with, which was carbonic anhydrase, had a very, very shallow knot. So you only needed to thread a couple of amino acid residues through a loop in order to create the knot. So there was one example of one shallow knot. So basically people said impossible. In 2000 and possibly before, there was a rigorous mathematical approach to determining whether proteins in the database were knotted. And the one I know was done by Willie Taylor and the answer is actually there are quite a few. And I think that maybe on Thursday, Joanna Salkowska will also talk a little tiny bit about the number and the number that there now are in that database. So I can't remember when I got interested sometime here and in 2008, when the last time this meeting was held here, we had begun to do some experimental work on these knotted 3-1 methyl transferases. And we knew a few very basic things. So we knew when we get hold of our protein, we get hold of our protein in a native state. And in that native state it's an open chain but it's knotted and if you pull then you don't completely extend the chain. You end up with a Titan knot in the middle. And we knew by using chemical denaturants, which is our standard method, that we could unfold this. We could unfold it in terms of, we could unfold all the secondary and tertiary structure to get to some denatured state. And then we could rapidly dilute out the denaturants and we could watch it folding back. And so we did this with our 3-1 knotted methyl transferases and we found out a number of things. So I just will use some general nomenclature. So D represents denatured or unfolded state. N represents native, folded state. And I represents intermediate, partially structured states. So we knew that we kind of had a complex pathway where there were multiple intermediates and there appeared to be heterogeneity in the denatured state. And probably none of you at the, probably none of, well maybe you can just about see that. Anyway, we established that there were, there were parallel pathways and there were a number of metastable intermediate states. And the whole thing was much more complex than for those small little proteins like shoulder chaging. And we went on and I went, we didn't at that point. I'm coming on to that. So I've deliberately not, so we assumed, I assumed that the whole thing would unfold and then the knot would fall off the end of the chain. So we assumed back then that actually what we were looking at in this reaction, which is quite complex, is the unknotted denatured state going to the native knotted state. So that takes me beautifully on to the next point, which is actually, and I won't go into how we established it, but a number of different ways. I know people found this quite surprising to begin with, but we established that for these methyl transferases we unfold the protein and we do not get rid of the knot. The knot is kinetically stable. And I should have just said that for these methyl transferases, they're quite, instead of the shallow knot, they're quite deep knots. So some 40 amino acid residues have to thread through the loop to form the knot. So, yeah, so this was surprising. So it's a kinetically stable knot. I think one of the points I perhaps make to this audience is, you know, we all draw our polymer chains, whatever our polymer is, like this. But for proteins, they're not at all like that. And maybe for DNA, actually, it's a better approximation. But for proteins, I think you have to remember that there's lots of things. There's lots of side chains. And those side chains can still make interactions in the denatured state. And actually it's known from other studies on other proteins that denatured states aren't always completely random coil and unfolded. And so actually it's not quite as surprising compared to considering just a very smooth chain. But we didn't actually know that in 2000. So in 2008, we knew this, but actually we were looking at it going from a denatured knotted state to the native knotted state. So then we had to, so we couldn't actually get rid of the knot. And another group showed that it takes months and months and months and months, you can eventually get rid of the knot. So we came up with some other experiments in which we actually looked at the folding of nascent chains of these methyl transferases. So essentially what we did was instead of starting with the protein and denaturing it, we started with a DNA template. And then we used an in vitro transcription translation system. So it's got all the cellular components needed to transcribe DNA and translate it in order to make our knotted proteins. So first of all you get your mRNA and then you get the ribosome and you get the protein synthesised on the ribosome. At some point it's released from the ribosome and one of two things can happen. Either it knots whilst it's on the ribosome or it actually then is released from the ribosome in an unknotted state from which it can then fold. So essentially by comparing our rates of folding from our chemically denatured state that we knew was knotted versus the rates of folding that we got from these experiments, we found that the rates of folding in these experiments were some 20 times slower. So we're pretty sure that actually it doesn't knot whilst it's on the ribosome, it comes off the ribosome. So there may be some folding on the ribosome but there's no threading to form a knot. And then what we're looking at in the in vitro translation experiments is this reaction here, which is the reaction that we actually want to study. OK, now this is where my memory goes. This is where having slides is really helpful because you just put up the next slide and you go, that's what I was going to say next. OK, so these experiments turned out to be really useful. So we could show, I'm going to use knotting loosely or threading, that knotting or threading itself or some event associated with it, I'll come back to that I think, is rate determining. So essentially knots slow the folding. Oh, we've shown that and a lot of computational studies have shown that too. We were also able to then put in various bits of the cellular chaperone machinery and show that for these treefall methyl transfers, that one of the key components of the bacterial chaperone machinery, which is called GroEL, accelerates the folding actually considerably, at least 20 fold. And that was very interesting. This is very interesting. Then we were also able to go back and do some experiments, these experiments, but with fusions where essentially what we'd done was we'd taken our knotted methyl transferases and perhaps I should look at what the next slide is. No, no, no, go back. And what we'd done was we'd made fusions where either on the C-terminus or the N-terminus or both termini, we'd put another protein, which is very, very, very stable, a little rock on the end of the chain. And with these in vitro experiments on those systems, we were able to show that the threading occurs through the C-terminus, which is what had been predicted, because that's the closest end to the knot, but we're actually able to show, interestingly, and again, you're just going to have to believe me at this point, we were also able to show that if you put a very stable domain on the C-terminus, it hinders its folding. If you then put an N, and then of course what might happen is it might just thread through the N-terminus, but if you put an N-terminal fusion on, so you put a block on the other end, it has very little effect, so that actually we think it still undergoes C-terminal threading even when the C-terminus is blocked. And we have other evidence for that. And then also looking at the effect of GroEL, GroES on these systems, we also began to get some ideas for how GroEL, GroES has this dramatic effect. And again, without any data, I'm a data junkie, so the idea of giving a talk without any data is shocking. You can think of a number of different possible mechanisms by which GroEL, GroES may help these proteins fold, not unfold. One of which is GroEL, GroES is essentially it's just a big barrel with a lid on it. The lid can go on and off and inside the barrel there's a large cavity in which folding can take place. So one idea was just it's simply one of confinement. If you can find the chain, you increase the likelihood of knotting and therefore folding. And I'll just say that actually I think the one thing that our experiments did show was that they didn't categorically show what mechanism was happening, but I think they ruled out the fact that it's simply confinement. And one possibility, which is what I think is the most likely and we haven't shown this yet, is that what happens during the fold and some of the computational work on this is wonderful, but which shows that actually some of the time what happens, let's put some intermediates in there, is you end up in some misfolded state which is quite close to the native knotted state but which isn't knotted and you have to backtrack from there in order to get to the native knotted state. And I think that my best guess at the moment would be that what the chaperone machinery does is it essentially aids in that unfolding step. So it's unfolding, misfolded intermediates. Right, I'm going to switch now and talk about UCH which is the other family of proteins that we've studied. And I'm not going to talk about any of the work that we've done beforehand. I'm going to go straight in and talk about the pulling experiments. I've used a lot more board than I thought I would, so I hope this is even vaguely understandable. So essentially, so we did some optical tweezer experiments with Matias Reef's group in Munich and the paper has just recently been published and essentially what we did with the optical tweezer experiment, I'm going to try and try and you can have your folded protein that knotted. If I draw something that's not knotted, just excuse me. So what you do is basically you can tether your protein between two beads in an optical trap and then you can apply force, you can move the beads apart such that the protein structure will unfold and you can pull it apart sufficiently that you completely, this now looks, you extend the whole chain and you can basically measure this distance here. So you can measure the forces needed and you can measure the distance that you've pulled it. And if you have a knot in your chain then what you should find and what we do find and others have found as well is that what happens is you don't get the full extension. The contour length that you expect is different and it's different by the amount of chain that's in the knot. And this had already been done on 4.1 phytochrome so we did it on our 5.2 UCHL1 and the nice thing is is we can put the pulling attachment points at any place we want in the structure. So what we were able to do which was the idea between using the technique is we could take our native 5.2 protein then depending on where we attached the tethers we could pull it and unfold it so that we got to a 5.2 knot but we could also change the position so that we could unfold it and get to a 3.1 knot and then we could also unfold it such that it was completely unnoted. So we did this and what you do in the experiments is so you pull, you unfold, you remove the force, you let the system relax back and refold and then you pull again to see the extent to which it's refolded. So essentially what you can do is you can look at unfolding but then actually you can also look at refolding and we could look at the rate of refolding from these three different states. And that gave us a wonderful meta because otherwise it's very hard to do this experimentally. I'm sure computationally this is actually probably reasonably trivial. So what we were able to do is show the differences in the folding rates and essentially going from D5.2 to D3.1 this is about three times slower and this is about three times slower than this so this is about ten times slower. So again the actual knot and the threading are really making a difference. So I will also just say a few other things that we found out because they're quite interesting. One thing we found out was we knew that there were several intermediate states that we could observe through the techniques that my group uses but what we managed to ascertain in this and perhaps not very surprisingly is that there are many, many different intermediate states and they're intermediate states which are populated on the unfolding pathway or intermediate states that are populated on the refolding pathway and there are probably intermediate states that are misfolded and off pathway. Now we can't say that absolutely categorically. What Matthias's group has done is reconstruct whole folding pathways from the data in these experiments. We can't do that because apparently our system is too complex and it has too many intermediate states. So we'd love to be able to say that this intermediate state goes to here and then this one goes off pathway or this one then goes back on etc etc and it's a single molecule experiment so it's potentially really powerful. So we can say generally there are a huge number of different intermediate states. We can also say something about the Titan knot. The Titan 3 1 knot has about 12 to 13 amino acid residues associated with that state and that's exactly what's been determined beforehand for other systems for 3 1 knots and what's predicted. With our 5 2 knot we saw something quite different. When we did it we saw about 40 amino acid residues were associated with the Titan knot or as I now call it because you'll go on to understand the loose Titan knot. So again this seems quite surprising and this is well over what is suggested from computation and again I think what it just illustrates is the fact that actually we don't have this smooth chain, we're not tightening a smooth chain we're tightening a very very rough chain that has lots of interactions there are bulky amino acid residues on there so you can sterically hinder the passage of the chain. But what we found out was if we really apply for high forces and this was as high as they could achieve we can tighten the knot. Something I think hits this, I'm forgetting work I've just published we can tighten that knot but there's actually quite a high energetic barrier to tightening the knot for this protein and I think that's probably true maybe for 5 2 knots in proteins. Right, I am now so I think that's all I wanted to say about the optical tweezer experiments and now. So now I'm just going to do the last part of the talk and there are slides again here because this is the paper that somebody sent me yesterday when I asked them to. So one of the things that's I think been of interest in the field and people have speculated on is whether the cellular machinery that degrades proteins can essentially degrade a protein with a knot in it and there has been some suggestion that might not be possible. So I'm just going to draw one more. Oh God, this is like teaching. Sorry, I love teaching. So essentially cellular degradation machines come. They are also kind of this shape. Sorry, that went through. But essentially they come in two parts. So there's the top part and it's known that that exerts force on a protein substrate to unfold it. And then there's the second bit and this has got the enzyme activity in it. So essentially this is the part that proteolytically cleaves the protein and degrades it. So you need a tag. So there's always some recognition tag on proteins that are going to be degraded and that's recognized by the top part of the machinery and then you have your protein which starts off native and then the machinery itself as it engages and pulls it into the machinery will unfold. It will unfold the protein so that the protein can go all the way through and be recycled into short peptides and amino acids. So this is the bacterial machinery which is called clip P, clip X. It's actually very similar to the mammalian machinery, the proteasome. We have done some work on the proteasome but we haven't got any preliminary so I'm not going to present it. Well I didn't have the data to present it. And essentially you can just see what I've just said on this slide here. This is the action. So the question is can so I'm... Okay it doesn't matter what order. So we looked at the three one knotted proteins and we looked at the five two knotted proteins and our initial results so essentially all you do is you can just easily measure how much protein is in a native form and hasn't been degraded and you can plot that over time and you can run some controls which is shown here. So here this is just a standard substrate. Sorry this is a standard substrate here and you can put in UCHL1, the five two knotted protein and you can put in a C terminal tag so that it's the C terminus of the protein that engages with the machinery and pulls is pulled on and you can show that it's hardly degraded at all and you can also show that actually if you put it in with a substrate it slows down it slows down the degradation of the good substrate so it's engaging with the machinery it's acting as a competitive inhibitor and you get excited and you think that these things can't be degraded by the machinery and then you put the tag on the other end and it's all gone really quickly and you think, gosh, how did that happen? So we thought about this a bit and one of the things that's known about these degradation machineries is that the local stability, the site of pulling, matters and also the amount, if you like, the amount of chain that the degradation machinery at hold off to pull also matters so it doesn't of course pull very well on something that's very short it's kind of like a tug of war the more men you have on the longer the rope the stronger you can pull and it happens that in UCHL1 that the C terminus which is buried right in the middle of the protein not buried but it's part of a very very stable beta sheet core is in a highly stable region whereas, oh, it's here, sorry whereas the end terminus, I can't find the end terminus is actually in quite an unstructured which is in, no, no, this is, sorry this is from a paper I've just been said this is really good, I'm editing it as I go along oh no, that is, that is, I thought that was the end anyway C terminus in a highly stable area of local structure end terminus, not so is it just a question of local stability and one thing to remember in contrast to the optical tweezer experiments where you're pulling at both ends is here you're just pulling at one end and the other end is free so what we did was we did some mutations in the C terminal region which really destabilise that region so the protein still folds it's still functional, it's still knotted and we put that in the experiment and here is that result so then we can pull from the C terminus but we've destabilised the local structure in the C terminus and it's degraded so this system does not have a particularly difficult time it can degrade knotted proteins so I'm just going to talk a little bit about 3.1 knotted proteins so we did the same experiments with 3.1 knotted proteins and the answer was yes it can degrade 3.1 knotted proteins and then we were interested in what was happening in this case so there are two mechanisms by which it can do this so one of which is possible that it can just pull it can actually pull a large piece of chain through that the knot is intact and is pulled through the machinery so that's one possible scenario and then the other scenario is it unfolds it there's a loose end and the knot falls off so we tried to design an experiment to distinguish between these two and I should say it's known that more than one piece of chain can go through the machinery at one time so this wasn't implausible by any means by looking at other systems so we went, well I haven't spoken about it, I missed out this bit but we went and we did the experiment again and we put our little solid block of a protein called Thys onto the end terminal end and this is the degradation tag at the C terminal end and we said does this degrade and the answer is and this is where I have data because I got sent the paper and the answer is sorry so this is a bit difficult to see basically this is the protein here this is the protein here and it's really not degrading very much and much much less than just the knotted protein on its own and interestingly over time you start to see an intermediate accumulate so what's happening in so you can see this intermediate I1 accumulate and then over time that is converted into another intermediate which is smaller so we've been able to do some accurate sizing and end terminal sequencing and we can show that this intermediate is the Thys domain plus a length of chain which seems to correspond to the size of the knot and the length of the tunnel that it has to go through whereas this intermediate there's just some protease activity around general protease activity and then this eventually will just get clipped off so this is slightly different so what happens in this case is that you see much less degradation because essentially the whole system gets blocked so this now is this fusion protein being degraded and you can see it goes down but it doesn't go down to zero and that's because the whole machinery is getting blocked and it's going down then you get accumulation and decay of I1 and accumulation of I2 so if we put a very stable domain on the end then essentially this doesn't get well essentially what happens is half the protein gets degraded but the rest of the protein remains intact and acts as a block to the degradation of competitive substrates and one might get really excited about this and say under some circumstances not if proteins can block the degradation machinery I will tell you that you don't need the knot there for this protein to block the machinery what's interesting is the fact that the knot clearly slides along until it can't get any further and then everything stops so the knot is still there the knot's not being degraded so at least that allows us to speculate that actually this doesn't happen this happens so we're fairly confident about that okay so I was really running out of energy this morning when I made this slide so I'll just do the Oscars thank everybody I know I should really point out all the people in my group also Matthias Reef and his group for collaborating on the optical tweezer experiments Danny Shu and his group in Taipei who've done a lot of NMR and I haven't really spoken about that and then Laura Idzaki and her group we've been working with on the Clip XP system I've put up this is not the last slide so I've come to these conferences I guess for a while now and I always end up thinking that in our protein world our knots are relatively simple and potentially not interesting so I also asked myself the questions do mathematicians have more knots than nature and so you might not know this but there's a little tiny bird it's migratory but it spends some of the time in the UK for some bizarre reason and it's called a knot so how many knots exist in nature certainly one knot and in fact you can get two knots and if you want to you can have a bigger knot and when you get a huge collection of knots together they swarm and they're called murmurations which is just a fantastic word you get a murmuration of knots and they perform these beautiful thousands of them sweeping through the sky you can go and see them just a couple of hours from Cambridge something I should do and so at least with nature you can get thousands and thousands of knots they're just not connected that is the last slide