 Okay, let's get started. So thank you all for coming out on this cold Friday afternoon. We started this week off with an excellent protein design talk by Tanya Kortemme. And today we get to wrap up the week with more exciting work, also coming from UCSF with Bill. So today's lecture is the David E. Green lecture in enzyme chemistry. This is sponsored by David Green Memorial Fund. David Green was a faculty member here in the Department of Biochemistry at UW Madison and became the founding director for the Institute for Enzyme Research in the late 40s. During his 30-plus year career here at UW Madison, his research group made substantial contributions to our understanding of the electron transport chain and oxidative phosphorylation. And this work really helped establish our department's legacy in enzymology and metabolism. And today's Green lecturer is Bill DeGrado, visiting us from UCSF. Professor DeGrado received his PhD from the University of Chicago, where he worked with Emil Kaiser on peptide design and synthesis. He started his career in industry where he spent 15 years at the DuPont Merck Pharmaceutical Company. He was on the biochemistry faculty at University of Pennsylvania, and a few years back he moved to UCSF. Bill has very broad interests in peptide and protein structure, function, and design. He's made major contributions to peptide therapeutics and our understanding of membrane transport proteins, such as the influenza M2 proton channel. I'm most familiar with his pioneering work in de novo protein design. Over the last several years, his group has been designing not just structures, but actual functional biomolecules. And this work really tests our understanding of protein function and lays the foundation for designing proteins with tailored properties. Bill is a member of the National Academy of Sciences and recently received the Stein and Moore Award from Protein Society. Please join me in welcoming Professor DeGrado. Okay, well, thank you so much for that very kind introduction. There's lots of seats here, guys. And for a really stimulating day, I've just enjoyed meeting so many people today, and especially the new young faculty hires, it's really exciting to be here. So what I'd like to do, and also I'm very honored to give this lecture, the green lecture, and a little intimidated because when I started my career, I thought we should be able to design proteins entirely from first principles that were as active as enzymes if we really understood the basis of protein structure. And at that time, we were starting to think we understood it because we could solve crystal structures and do mutagenesis and ruin the proteins and say, okay, now we understand that glutamate's really important. But if we really understand it, we really should be able to start from first principles and design things that are as complex as natural proteins, and first getting to just simply fold was a great challenge. And now I think we're right on the door, knocking on the door, being able to really make them bind complex molecules and the cofactors we need for catalysis. And I think it's really going to be in the next decade that we achieve the ability to design catalysts that are very efficient. And it's really gratifying seeing so many people in the field working now. So I thought I'd give you a little sort of historical perspective of what it was like to try to start designing proteins and then focus the talk on metalloproteins with the hope that that will give us some focus. And we'll talk both water and membrane-soluble metalloproteins and how we design them and what they do. So when we started thinking about designing proteins, people were very quick to tell me that this is impossible. How could you possibly do this? And the kind of arguments that they would make would be, for example, you'd have to go think about how are you going to go through 20 to 100 different sequences. That's more than the atoms in the universe. If you mutate at a femtosecond time scale, so molecular vibrational time scale, in the time of the universe you're not going to come anywhere near to this number. So how are you going to find the right one? There is a conformational problem with astronomical questions again about how can you ever find a native structure. So basically, those seem to me to be sort of vitalist arguments and I'd say, okay, then you believe in the great force that created everything and intelligent design, et cetera. And they'd say, well, no, not really. So then how did it work? I think basically what happened early in the evolution of proteins was that you got peptides that could self-assemble, often probably originally to amyloid-like structures, and then these can, we'd get gene duplication, allow things to assemble into single chains and we can create a symmetry. And so if that's the case, we really should be able to design proteins in a buildup strategy starting with peptides, allowing them to assemble building sites and so forth. And so that's kind of the approach we took early on and the question is what's going to stabilize our peptides to make something like a helical bundle or whatever. And what we hypothesized at that early stage was it was just hydrophobic periodicity. You could take a peptide with the same amino acid composition, change the sequence, and if they are arranged to stabilize a helix, then when they bound to an interface or self-associated, they would make a helix and if they had a different order than they would make beta sheets. And in fact, we found that to be the case on some model peptides. So arranging these peptides now to make every other residue, they would, or with a helical periodicity, they assembled into something that we thought were helical bundles and when they were every other residue, they made something that we thought turned out to be a cross beta structure. Now going fast-forwarding to the two sort of current times, there's been a lot of thought more recently that a lot of the, again, the first precursors to modern proteins were amyloids and then that they got stitched together in solenoid-like structures. And you can see residual remnants of those type of very regular structures say in this carbonic anhydrase, it's a zinc-dependent enzyme from archaebacteria shown here. And the arguments for an early evolution is that short peptides that form amyloids can have many different conformational forms and these conformations are self-replicating and self-purifying. So you can have a mix of things and only one confirmation and one sequence will come out. So to test this with Ivan Karendevich, initially when he was in the lab and then most of this work is since he's been independent and he's been carrying this forward, we made a peptide, just a heptapeptide with histidines on one side, hydrophobic residues on the other side and after, I think we screened four peptides to reach this sequence here, so certainly not large computational design project. And yet this peptide when it binds zinc has this hydrolytic activity. We're looking at nitrofenyl esterase activity. This is a function of substrate concentration. We see no activity whatsoever for the zinc ion or for the peptide alone, but we mix them together. We have robust, saturable activity just from a little peptide. And in fact that activity, at least on a weight basis, beats anything that had ever been done in de novo protein design of metalloesterases. So it shows that it's a fairly substantial rate in contribution. So we solved the structure by solid xenomar with May Hong recently and this it assembles in pretty much the way we had anticipated. We have zinc ions that are tetrahedrally coordinated with the last ligand being water, more about that in a minute. The hydrophobic sides of the peptides associate and then they make this overall assembly shown here. Now, what was interesting about this is that we're seeing a new form of stabilization of amyloid and asparagine and glutamine zippers had already been known for many years from David Eisenberg's and others work. And it was also known that metal ions can stabilize amyloid and this is the first example now. We can see how the strands are all stabilized by this amyloid structure shown here. And we're interested in this for a number of reasons. First, it's been known for some time that amyloid forming peptides can be triggered to form amyloid by metal ions. You can imagine if you have parallel peptides, it takes the residues that are able to chelate or ligate and puts them all in register to one another in closed space. So metal ions can now bridge between the chains. And it's also known that metal ions also contribute to some of the toxicity of amyloids. In terms of applications for catalysis, you can see here we've activated water molecules in this array. I think that's the basis for the hydrolytic activity. Corendovich has also shown that these can catalyze copper catalyzed reactions. And I really think there's quite a bit one could do with these amyloids and metals in a combinatorial manner for catalysis. Finally, I think this is a new type of material. It's sort of like metal organic frameworks except it's peptide organic frameworks. So again, I wanted to make sure I called out Ivan, who's really been pushing this work forward in recent years. And May Hong's group for solving the structure. So let's think again about how we came about designing helical bundles. We did this in a more or less iterative fashion, making peptides assembled and then building loops and then single chain for helix bundles. The types of principles, we're always testing principles in our designing things. It was first that hydrophobic effect was going to drive this. We had to have really tight packing of the side chains, electrostatic interactions, and then propensities to adopt different confirmations. So we made these and they were almost native, but they had sort of molten interiors and we never could solve the structure. And one of, Christ, excuse me, that's weird. So something that really made life easy was the invention of repacking algorithms already back in 1978. Michael Levitt had found that amino acids have sort of quantized rotomers or confirmations. And ponders and Richards had shown that you can now search through sequence and rotomers space to find combinations that filled sequence structural cores efficiently, days you lay in handle, first put it to use in protein design. And one of the first really de Novo's applications were from Dayant and Mayo who made a zinc-less finger. And with that, we were able to proceed much more rapidly in de Novo design. And the first things that were done were more or less coiled coils. This is one anti-parallel structure that we solved with David Eisenberg. Kim Harbury and Albert solved many structures and repacked cores and showed how that the packing really dictated the stoichiometry and topology of the structures. Moving to globular proteins, we were able to make three helix bundles that folded in a very nice manner and these were the first globular proteins that were genetically encoded and solved. And we were able to make domain swap dimers and hexamers and ultimately molecules that assembled around carbon nanotubes and templated gold particles, bucky balls, etc. in the nano world. And David Baker came along and just really had dominant contributions in this area of de Novo protein design through his application both of Rosetta and his training of many, many people, terrific people. And so at this point, I think we can safely say it's possible to design small single domain proteins of a variety of different topologies. And the biggest question for us as designers is what is the question? What are the things that we want to accomplish? What are the functions that we want to build in? And so I sat down with Mike Therrien at some point and he said, Bill, it's nice that you can make these proteins. But let's look at a real protein that does something important. And so this is in photosynthesis the protein that shines light and reduces oxygen and it's really complex as you saw. I just cut it back to just the business part and got rid of the antenna. And it still is really complex, right? Let's look in at the cofactors it holds together to do this amazing job. And what we find are porphyrin like structures here. We have quinons. We have an oxygen evolving complex. We have things that can form radicals. It's remarkably complex. And what it does is, of course, you shine light, you separate a hole and electron. And this charge separation ultimately is what generates the power. Now, we would like it really though to be very simple. And we think of things under control of evolution. We want it to be like the statue of Darwin, right? Very simple, very elegant, very functional. Instead, what we see nature has handed us is this God awful music man, right? And so but it also suggests a modular approach where maybe we could start by trying to design a drum, right? So here's a drum. Maybe we could deal with that. Maybe we could make something that binds quinones. Maybe we could stabilize a radical. Maybe we could make a multi nuclear complex here. And so I'll show you in the next few minutes our progress towards that. Putting it all together, of course, is going to be the great challenge. And I'll tell you a little bit about how we design membrane proteins as we move along. So let's start with our little multi nuclear center here. How do we design proteins that bind multiple metal ions together for oxygen binding and activation? We've made a little progress over the years. And this goes back to some really early work that we did in which we started. We became fascinated as Brian has in these dye iron proteins. Is this your structure, probably? Anyway, but basically what we see is they're highly complex proteins but deeply embedded in them are these very simple helical bundles. And we can really describe the structure to about an angstrom RMSD using very simple equations. So this is just a coiled coil approximation. And that then suggests that it has symmetry. We have a glutamate from each of the helices, hydrophobic residues to make the structure fold. And then a little more detail is needed. So here, this is the structure we first designed with glutamates to bind the metals, both bridging and chelating. Then we added histidines just one side, leaving another side for oxygen substrates to approach. And what's going to hold it there? But we needed to have second shell ligands, so hydrogen bond networks. And we can build them here. We could also build second shells here. And all of this done with simple side chain repacking algorithms. And then finally, we have to extend the helices enough that we can now build hydrophobic residues here and here that are well packed and allow the structure to fold. And then the outside is just basically polar things. We solved the crystallographic structure. And indeed what we found was that we bound the cofactor precisely as expected. We have this very large network of hydrogen bonded structures. We also solved the structure without metal ions. And what we found was that hydrogen bond network was completely retained in the APO form. So it's all pre-organized waiting to bind the metal in a very specific geometry. And you can also now design these helical hydrogen bond networks in Rosetta. And there's modules to allow that to happen. Now what does the protein do? So the first reaction that we were able to show is similar to an alternative oxidase in which we take an amino phenol and it gets oxidized in a two electron process, two. So from dipheris to dipheric. And we found it followed Michaelis-Menton kinetics and really Kcat on Km was getting within a couple orders of magnitude of what we see for very simple enzymes such as alternative oxidase. Now what you can see here, looking into the active site, is that there's two irons, one here and one here. And so it's very good for symmetrical two electron processes. But what happens if we block access and change the energetics by making this here? Now we've added a histidine, you can't even see one of those irons. And what happens now is that we can focus our chemistry on one of the two irons. It makes a, when we add oxygen, it makes a peroxo group here. And then that can react directly on an aniline. So we have a completely different type of chemistry going on. Hydroxylate and then further another two electron reduction. Regenerating the iron two. So in all, we have a four electron process. And we both designed and analyzed this in collaboration with Ed Soloman, a theoretician and spectroscopist, and that's been published. Now, to move further, we would really like to be able to build four metal ions that would be redox active. And so then they can deliver each one of these one electron as we move to thinking about things where we take oxygen, I mean water and oxidize it to O2. And so as a step in this direction, what we decide to do is to bind tetranuclear zinc at this site, which would be easy for us to characterize by NMR and X-ray spectroscopy. And so what we did was, again, built second shell hydrogen bonds. We have four histidines, four carboxylates as the ligands. We organize everything such that the helices, when the side chains are the appropriate rotomers and bind appropriately in a geometric sense, are well-packed. And then it's simply the same process that we've gone through over the years of side chain repacking and building the structure up. So let's just look at the crystal structure because it came out so very close to the structure that we had designed. And it really shows, I think, a new level of mastery of being able to design highly complex metal environments. So we have layers of hydrophobic residues that provide the driving force to really push all these highly polar residues close together in the right geometry to bind to the metal ions. Let's look at the site close up. And what we can see is we have the four carboxylates, and then we have two bridging water molecules in addition. And that's really important because ultimately, we'd like to be able to take protons on and off. And we finally have plenty of second and third shell hydrogen bonds. So here's, you can see the cuboidal like arrangement. And here's the bridging water receiving additional hydrogen bonds. And now you can see that we built second shell and third shell interactions in large hydrogen bonded networks that really are critical to holding this whole thing together. Okay, so that's about where we are. It would be great to be able to get manganese centers in here and really start pushing this towards functional chemistry. Let's look at some of the other things that we can do. For example, can we bind quinones and stabilize radicals? And here, the challenge that we set ourselves on was to take a cataclysm. And a cataclysm will go to the quinone form via a very, high-energy unstable intermediates, which is called a semiquinone. And it's a radical form. And the question we said is, can we bind this so tightly into our protein that we would stabilize and that would become the dominant species? And so what we did was, first of all, we switched from redox active metals to dyesink. And then we knew that the binding of the hydrophobic residues in a cavity, which we have here, would drive the affinity. So we're going to use the dehydration free energy and packing free energy to drive the stability of the semiquinone, which then could bind very tightly because it's a good key later. And so indeed, we're able to do this and we're able to stabilize the semiquinone by a minimum of 10 to the sixth fold. So it's many Kcals per mole. And that became the dominant species and it's been published. I won't go through all the data, but we made this protein that can stabilize a radical for many, many weeks in solution. So, now let's think about how are we going to deal with binding proteins that bind to porphyrin cofactors with redox active metals? And so we started again to design helical bundles that bind to porphyrins. And this really is a major difficulty because when we move away from simple metal ions, we're really thinking about the problem of binding ligands, organic ligands that can associate with a protein. And we really need to be able to achieve subangstrom accuracy. And this is a major unsolved problem in de novo protein design. Now, to be sure, for the last decades, there have been many, many proteins designed. This was originally work we did in collaboration with Les Dutton. And he's continued along these lines as of many others. And there's literally hundreds of papers on proteins in which cofactors have been non-covalently assembled or tethered into helical bundles called maquettes. But there are no crystal structures and no NMR structures for these in the reconstituted state. And that's because they have highly molten interiors. And so it's sort of like a hydrophobic dye binding to a very nonspecific environment. And yet it's been possible to achieve a modulation of midpoint potentials and some very simple O2-dependent chemistries. But if we really want to push beyond this to doing a useful catalysis, we're going to have to be able to design things that really adopt one single structure. And that then brings us to a major unsolved problem in de novo protein design, which is can we make things from first principles that really design, that bind to ligands? And the only work in this area where we've seen some success comes from Baker's lab. And that's usually after trying many, many different scaffolds. They can find some of the binds a hydrophobic ligand. And then using yeast display, they can tune it up to make it high affinity. But just being able to go in and make something show that it binds, solving the structure, showing it's exactly what you have had not been accomplished yet. So what's the problem? Well, first, when we think about designing the protein, we usually think just about this region here around the binding site. And yet, we know that proteins really finish their folding only when the ligand binds. That is that you build in repulsion between a lot of the residues that are going to bind here in the bound state, which becomes, let's say, relaxed when the ligand binds. So you might have a lot of dynamics here, ligand binds, and then we finish the folding process for the protein. That's how I like to think of it. And then the other thing that we tend to lack, and we're working on, and I won't speak about today, but we're making a lot of progress, is the cooperativity of the binding. So once you bind one ligand, let's say, make one interaction here, it polarizes the system so that the second one can be more favorable. We know this about hydrogen bond formation for many years. And so we have to start thinking about the cooperativity of the whole process. But again, that'll be the next time I give a talk. Now, so how do we rethink this whole problem with this sort of insight into designing a protein that would bind a porphyrin, a metalloporphyrin, such as this? And the way we've always approached it was in some ways sort of inspired by antibodies so that you think you've got these variable loops. You're going to vary the sequence so that you can bind something. And pretty much the same thinking has permeated de novo protein design. We've got a core, we've worked hard to make this core. Our nature worked hard to make the core. And then we're going to rebuild a binding site inside what we call a scaffold. And yet we know that in antibodies, we have the process of somatic mutations, whereby residues that are way down here can contribute to the binding of something way up here. We'd like to think of it in terms of allosteric networks these days. And so it's well known that things really very far away contribute to binding affinity. And so you have to think of this whole process of designing a protein to bind ligand as a unified process in which the folding is very tightly coupled to the ligand binding. And if we do that, we think we're going to succeed. So we took as our goal binding this very electron deficient zinc tetrafenyl porphyrin with a highly ruffled structure. And we chose this because of its optical electrical properties of interest to Mike Therrien at Duke and said, can we design something that binds us? OK. So how do we do that? Well, first, we have to have one metal ligand interaction. That's going to help us position this helix. And then I mentioned second shell interactions as being very important. So we have a threonine that we're going to build here as a second shell ligand. And this really is what we've done in previous work, where we're trying to build four helix bundles that bind porphyrins. The other two helices, initially we can just place them as symmetrical replicates of these, but lacking the ligand and so forth. And so that's the original positioning of them. And then that creates an overall structure in a series of folds so we can create a large ensemble. And then we treat the process of repacking of this to create, go from here to one with side chains that have been repacked in the center is a unified problem. So while we're trying to maximize the interactions with the porphyrin, we're also trying to repack this core. And we do this with flexible backbone design using Rosetta. Tanya Kurtemi was here and she probably just said more about the back rub and approaches for flexible protein design, but it allows the protein to breathe and change its confirmation during iterative cycles of design and repacking. And so at the end, we have a sequence. And oh, there's one more thing that I think was crucial is that we avoided putting a large number of hydrophobic amino acids to pack against the porphyrin. Instead, we had backbone interactions between the C alpha H and the porphyrin ring and glycines in this region as well. And the reason I think that's important is if we add a lot of hydrophobic residues, the whole thing could collapse in the APO state and become much too nonspecific in terms of its geometry. So we designed this and we made one sequence. So we didn't make large numbers, we just made one. And this is the solution NMR spectrum that we got. Here's the proton dimension. Here's the N15 dimension. When you see this large number, this large dispersion in the chemical shifts, it immediately tells us that we have a very nicely natively folded protein. Each amide is in a unique environment. And we also saw a very good spectrum, though, the less dispersion, to be sure, for the APO. So we were able to solve both the APO structure and the holo structure. And these are the structures shown here. So the holo structure has the protein bound in it. That should say, I don't know what's happened here. Anyway, when we bind the heme, we basically have very well-structured protein. And in the APO state, we see that this region down here at the core is really well-structured. But it's more dynamic up here. We can do, we found that there were two confirmations. One is closed in the APO state, and one is open in the APO state. And we think that possibly that motion is really important to allow the cofactor to come in. Once it comes in, snap. It's shut. It never lets it go. And the protein has been soluble for over a year now in solution, despite the fact that that cofactor has low nanomolar of any solubility. So it keeps it in solution, never lets it go. So how did we do in terms of the accuracy with which we can design something? It's about 0.8 angstroms in the binding site relative to the design. And so it's extremely, it's on the order of less than a single bond. So I think we're really getting there. We can really specify exactly what we want to happen. So what did we learn? I think explicit design of both the folded core at the same time that you're designing the binding site is a major step forward if we're trying to design ligand binding proteins. We have to treat this as a unified problem. If we don't, I like to think of it as dominoes. Each one is influencing the other. We take one out, we separate the problem, and now we've completely dissociated what's happening here from here. You've lost that influence. Secondly, what we found was we had something that was flexible enough to allow the porphyrin to come in. But once it was there, it holds onto it forever. We can boil the protein at 100 degrees. If we look at the spectra, they're entirely identical, showing that it maintains the porphyrin in the same environment and remains folded. OK. And I think the field of cofactor binding proteins really is now positioned to progress in a much more predictable and high throughput manner. OK. Now, the last thing that we just got, I got data from my postdoc, or a graduate student, is shown here. And so I don't have any explanation for it. But she's really excited, right? And so when we first cloned and expressed we called gator in the protein, we found this little teeny bump at the Soray band, suggesting that it had just a teeny, teeny amount of heme that was incorporated from the E. coli. And if we went through it, it was about 0.05% incorporation. But we then said, well, what if we tried to design this purposefully so that it would bind to protoporphin 9 iron in the cell? Would we be able to make a protein that really expressed, that really used that cofactor, could drive the biosynthetic machinery and load the heme? And so again, it was a matter of repacking the structure. This time, we have to stabilize a lot of polar groups, the propionates from the hemes. And indeed, when we do this, the first protein that we made, when we break open the cells, we find that it binds to heme very, very tightly. And indeed, since we had to overexpress the protein, the E. coli were pink. And so this is the first example where somebody's been able to design a protein that really loads a natural cofactor without using like a covalent modification of C-type cytochrome. So it opens the door in the future to doing in vitro evolution, to looking at a large number of different types of catalytic reactions. So I think it's going to be a lot of fun. Now let's go back to our theme, though, of putting things together and making them work in membranes. And just give you a little overview of where we are in terms of being able to make functional membrane proteins. When am I supposed to, how are we doing this? It goes to 4.30, is it 10 minutes? OK, good. So one project we've worked on is asking ourselves, can we design a protein that will take transition metals and cause them to go from one side of the bilayer to the other in response to a proton gradient or to go down a gradient, create a proton gradient? So can we make a transporter that uses ion gradients to drive another substance? So that's the question. This concept's been around for a very long time. And we see now many protein structures that show it. You have what's called alternating access. One side opens, closes off, and then something can go in. And that can be coupled to a gradient. And this was first predicted before. I'm sure most of us were born in nature. This was from Mitchell, the idea of the alternating access mechanism 1957. So why does nature want that? Well, you don't want to make just a channel. Otherwise, things will just diffuse through. You dissipate the ionic gradient that you worked so hard to create that drives most of the processes of life. And whereas if you have a rocking mechanism like this, then the channel is never open entirely to make a channel. Instead, it goes through discrete steps. And we can start to thermodynamically link binding of multiple factors to the rocking motion. So that's what we wanted to do. But it required a large number of challenges. First of all, we had to de novo design a membrane protein to show its structure was right by high resolution methods. That had not been done before. Secondly, if we want to bind zinc, as you've seen, we're really going to have to hold on to these metal ions. So we had a large number of charged residues that we had to stabilize within a membrane. And then the next part was relatively easy, we thought, that you have to link proton binding to zinc release. And again, if you can imagine if you protonate a ligand, it's going to not be a good ligand for zinc. So we didn't think that was too hard. And then we had to design an energy landscape, think about what could go wrong. And what we were worried about was double occupancy of this structure in a symmetrical conformation, with exact symmetry, falling into a deep energy well. And so we have to destabilize that, as shown here. So going from there to the design of an actual protein sequence, we first, to make it easy, we said, let's just make a single peptide that combines zinc, disink, either here or here, if it's an asymmetric conformation and if it's a symmetric conformation that can bind in both. And again, we have to destabilize this conformation relative to the other rocked conformations that are partially asymmetric. To achieve that, Gevor Gregorian did the computational part of this, using a negative design algorithm in which we think about maximizing gaps, rather than simply thinking about maximizing stability. And any time we're thinking about designing and moving towards enzyme-like structures, this is going to be increasingly important. We don't want to stabilize one state. We want to get the right balance of stabilization of various states. And they had developed, with Amy Keating, algorithms to really dial in energy gap in protein design. So what came out through this work is published now. So I won't go through the details of how we designed it, but what came out was a very tightly packed Helix-Helix interface shown here and a very loosely packed Helix-Helix and wider Helix-Helix interface shown here. And so what this allows is a rocking motion about the loose interface and very tight assembly about this dimer interface. And when we solved this structure crystallographically, what we found was that the dimer was precisely as designed, but it had dissociated in my cell, so we only saw the dimer. And so then that caused us to need to go and use solids NMR with May Hong again to solve the structure of the overall tetramer in phospholipid bilayer. So we could show it made the tetramer. We could also get distance measurements between some of the Helix-Helix's and show that the intact structure, although dynamic, was as designed. Now, the next thing was to look and see whether it actually worked. And so we look at the diffusion of zinc into vesicles. And as zinc goes in, proton should go out. That's pretty so first we see zinc does go in and concomitant with that protons do diffuse outward in this experiment. And the coupling is about two protons to a zinc. We also could show that it actually drives zinc up a concentration gradient, or protons up a concentration gradient if we allow zinc to come out. So it was showing that the hallmarks of a transporter, we also were able to show that it was most impressive in terms of Kcat on Km. If we looked at cobalt efflux through the channel or influx through the channel, again, we have the same Km observed for zinc or cobalt irrespective of whether we're measuring protons moving or the metal ion moving. So this clearly shows that they're linked processes. Okay, and finally, what we're finding is Kcat on Km, which is within about a factor of 100 of some of the natural proton linked metal transporters. So the next question we asked though is, does it need to be this complex? Remember we had two sites, one here, one here, and we placed the site right in the middle of the protein instead and have a transport pathway here. And one of the things we wanted to achieve was to have a hydrophobic gasket on either side and again, it's still a single chain that can rock back and forth in different confirmations. So we designed that and we made two versions, one with an extension from the membrane and the other being shorter. This is the one that actually we've had the best luck with in terms of structural characterization. And we made two versions of it. One is water soluble, the other is membrane soluble. And by putting water solubilizing groups on the outside, our intention then was to be able to more easily characterize it crystallographically. In membranes, we were pleased to see that perhaps due to the extension that it formed a tetramer in density matched micelles. So that was good. Okay, and the crystallographic structure at high resolution of the water solubilized structure shows that indeed we have metal bound. It seems to be dynamically held in the center because we don't see individual lobes of density. And the membrane structures coming along, this is at about four angstroms resolution. But it was nice to see tetramers. Finally, we see that it's, oh my. Anyway. Good luck with your talk, Bill. It is what that said. But anyway, it does work in terms of flexing ions. So basically, I think we're at the point where we can start to design native proteins that bind cofactors. I think we have a lot of ways to go to really get them to do exactly what we want to do for challenging reactions in membranes. I think we can move things in a purposeful manner. And I think the future is really great for protein design, not just because of what we've done, but more importantly, what everybody else is doing in the area. And so, thanking finally my coworkers for the last part. This was Nate Joe and Givore Gregorian, who was a postdoc, but did this work while he was at Dartmouth as an independent faculty member and is now a tenured professor there. OK, so, and OK. I think I'm going to stop there because I'm out of time. So, thanks. I haven't really thought a whole lot about that other than the obvious things that if you want to bind copper, you would want to put a lot of histidines there. And the proteins we've designed with both carboxylates and histidines follow the Irving Williams series reasonably closely. I think if we're going to bind manganese, we really want to think about water and bridging water. And so, we'll have to get into the nuances. I've never been clear how much of the selectivity that we see in biological proteins to have just one metal ion there is due to chaperones relative to binding affinity. Maybe that's a solved problem, but I'm not sure. What do you think? Yeah. OK. This thought would be interesting to hear a couple of things. Yeah, we've in every case always started with the simplifying symmetry just because it's sort of an evolutionary argument. And it used to be that that really was a great help in protein design. Now I think it's more that we think of it as a draft and then go directly to more asymmetric structures. So the protein I just showed on the bind's porphyrins, of course, lacks any symmetry. But it still has a primitive symmetry at, I don't know, one and a half angstrom. RMSD, it's D2 symmetric. But as soon as we start to think about working on a substrate that's asymmetric, we can't be symmetric. And so I'll always take as much symmetry as I think is necessary or is consistent with function because it simplifies our understanding. And then rocking off of symmetry is a well-known way for alternating axes. I feel like I'm just babbling, though. Yeah, sir. So as to know what design is it fundamentally harder to do that with beta proteins as opposed to alpha proteins? So beta proteins have, the question is, can you do de novo design of beta proteins? And that was a really difficult problem for years. And I know Brian Coleman, well, one of the first proteins that was ever designed was by Jane Richardson, and it was supposed to be a beta protein and they just made amyloid for a long time. Brian Coleman, I think, worked on this for 10, 15 years. Finally, the Rosetta community, just by thinking about principles, what are the principles that are required? We're able now to design beta barrels. So, and then alpha beta has been easy from the early days. Tim barrels have been done. So I think most of the topologies have been done. I think if we go to all beta structures, you really gotta be careful or else you'll end up with amyloid. So I think you really need to have a lot of negative design thinking what's it gonna take to end up with amyloid? Do we have enough polar things on the edges of the sheets? Do we have pro-lanes appropriately placed? And then I think we'll succeed. Yes, sir. And if you did that, it might be limited in itself, but then you could do an evolution experiment. You just grow the bacteria for some period of time and you could see how far away from the ideal you really are. It might teach you something about what you're doing. Right, right. I think there's eminent scholars thinking about this. He's working on it. But we made a very simple attempt and didn't work much on it, putting loops between the helices so that we could optimize it in vivo. And it was toxic. And one of the issues we've had with this is that it's a little leaky to protons. And the coupling of metal ion to proton flux is very good. But the opposite coupling of proton to metal ion isn't very good. But basically the killer experiment here I think will be to make a zinc transporter that has low affinity for detoxification because zinc is toxic to E. coli. There's ZIF proteins in E. coli. These are transporters, two of them. So you could knock them out and then look at loss of toxicity. So that was something I should do sometime. Yeah. You showed the complex and highly evolved excender. Yes. I wonder if you've thought about maybe the most primitive and simplest catalyst which might have been the dipeptides or tripeptides that enabled condensation of amino acids. In other words, enable lengthening of amino acids. So the question is have we thought about templated synthesis or but dipeptide, yeah. The chemistries have formed long. Yeah, yeah. But basically we play with that a little bit. But my former postdoc, Yvonne Korendavich is driving that aspect of the work. But it's fun to think about. Yeah. Yes. Yes. Wait, fusing onto artificial membrane proteins. Certainly have thought about it and there's a lot of important questions you could ask. That has sort of at the maquette level where we don't have structures. Les Dutton has done that where he's put hemes at various locations into the membrane, out of the membrane and so forth. And it would be interesting to do it with some of these structures where I think they're quite possible to get three dimensional structures and then look at how it affects first the midpoint potential and then transfer of electrons between centers. We're doing that and it was a lot of fun protein design exercise to put multiple binding sites together because the porphyrins have very different topologies than the dimetal centers. So trying to go from one topology to the next was an interesting design challenge and we're just getting structures of those. But yep, so many things to do. Yes. We didn't want to have a whole bunch of hydrophobic residues because we were concerned about collapse. That it would just make a collapse structure or that we would have a lack of binding specificity geometric specificity in the native or in the reconstituted state. And if you just have glycines for example along a helix coming against the porphyrin ring that's reasonably hydrophobic, right? And gets the water out. And it also provides flexibility I think in the porphyrin free state. So it provides enough dynamics I think to allow things to come in. So these are the sort of the considerations we always go through in designing a protein thinking about every state and then trying to optimize. We don't always demonstrate that everything that we put into it was necessary. So I haven't done that experiment to make it more hundred. The APO we have the NMR structure and it's got two confirmations open and closed. But we don't have crystal structures of these. We had very, very soluble exteriors. So we had way too many charged residues I think and they resisted crystallization. So we're redesigning the outside so that we can crystallize them because it's a pain to solve it use do NMR structures all the time. It'd be nice to have both. That also allows us to see where water and other things are. Okay, well, let's take our speaker one last time. Thank you.