January 4, 2001, La Jolla, CA
Ian Wilson and Don Wiley worked together to solve the structure of the influenza virus haemagglutinin. Ian was a Research Fellow - the only postdoctoral fellow - in Don's laboratory during this time. He is now a Professor in the Department of Molecular Biology and the Skaggs Institute of Chemical Biology at the Scripps Research Institute in La Jolla, California. He comes originally from Scotland, and although he has been in the United States for more than 20 years, I could still detect the Scottish brogue during our conversation. Ian told me that he first became interested in science when he was in high school and did biochemistry as an undergraduate at Edinburgh University. But when I asked him about that he said:
IW I did biochemistry as an undergraduate at Edinburgh University, but they didn't actually teach biochemistry in school in those days. You had to do things like floral families, dogfish dissections, and botany. I knew I was interested in chemistry, but I really wanted to do biochemistry rather than chemistry.
SS What years were you an undergraduate?
IW Between 1967 and 1971. Then once I was in Edinburgh and was thinking what I wanted to do in graduate school, I went to talk to one of my professors and told him I was interested in structural aspects. I had actually been reading NMR papers rather than x-ray papers at the time, because NMR seemed to be the newest and greatest thing. I was advised to talk to David Phillips at Oxford and I went down and visited and became very interested in how different structures were related to one another, especially when I saw the alpha lactalbumin and the lysozyme sequences compared. At that stage David Phillips' lab had solved the lysozyme structure.
SS Had both structures already been done by then?
IW Only lysozyme. It was quite clear that this seemingly unrelated protein, alpha-lactalbumin, was related structurally because it had a very similar sequence, but a very different function. I applied to that lab and said that at that stage I was interested in the evolution of proteins and how they were related to one another, and, consequently, a few months later I joined the lab in Oxford. I started working on triose phosphate isomerase. In those days, projects took a long time. So I took this project over from two other graduate students, Ann C. Bloomer and David Banner. Greg Petsko was also working on the project. He was mainly working on complexes with substrates and inhibitors.
SS Was Dorothy Hodgkin there?
IW Dorothy was there. She was in the other wing of the building. David Phillips had a large lab in one wing and Dorothy had a much smaller group in the other wing. She was a Royal Society Professor. There was really a pretty large group of people, probably over 50 people in the lab that were doing crystallography and related protein chemistry. That's really when I started getting more seriously interested in protein crystallography. I spent the next six years there. I did my thesis work on the structure of triose phosphate isomerase (TIM) and then stayed on a little longer.
SS Did you solve it?
IW Yes. It was one of the early enzyme structures that was solved. It was an enzyme in the glycolytic pathway. Michael Rossmann was working on LDH (lactate dehydrogenase). The structure of LDH was just about to come out. The evolution part came back into it because it was quite clear from looking at these structures that they all had similar types of folds; alpha, beta repeats, and TIM was a remarkable structure at the time. It had such a uniform repeat - it was alpha, beta, alpha, beta, eight times. It seemed to have almost too much secondary structure compared to things like lysozyme. Hemoglobin was almost all alpha helical, but this was a mixture of alpha and beta structure in a very organized molecular way. It has become an absolute classic now. Well over a hundred different enzymes have been identified that have this fold. There are about seventeen different families. Luckily, that was the first of the TIM beta-barrel structures to be determined.
SS So you were at Oxford working on enzymes. Did you meet Don (Wiley) at that time?
IW No. The first time I met Don was at a meeting in Alpach, Austria organized by Max Perutz. He had organized these protein crystallography meetings there in the mountains and Don was speaking at that meeting. I was still a graduate student. He was an assistant professor. There were very few protein crystallography meetings, so when someone like Max Perutz held one people flocked from all over the world to go to it. There were not that many crystallographers compared to what there are now, and so almost everybody tried to go to that meeting. As a graduate student, I managed to get to go to that meeting. This was quite an influential meeting in those days.
SS Do you remember what year it was?
IW This must be about 1972-73, around that time. Don was talking about ATCase (aspartate transcarbamylase). I didn't really meet him. He was at the meeting.
SS How did you decide to go to work with Don?
IW I decided that after I had done TIM, I would probably stay on at Oxford and work on some other TIM, either a thermophile TIM or a plant TIM. I had set that up and then just by chance a postdoc from Martin Karplus' lab, Christoph Kratky, was visiting. He was visiting David Phillips' lab and gave a seminar. I started chatting with him afterwards and he mentioned Don's project on hemagglutinin. I thought that sounded really interesting. He told me that Don was looking for a postdoc and asked if I was interested. I said: "This sounds like a really good project". When he went back, he obviously had talked to Don. Don sent me a copy of his grant which he had just submitted, and they had just got some of the first crystals then. I looked through the grant proposal and thought this is a project I would really like to work on. It sounded an awful lot more exciting than doing another TIM structure. And so, quite sheepishly, I took the grant into my boss, David Phillips, and said: "Could you have a look at this and tell me what you think of it because I am very interested in it. I know I am committed to being here, but let me know what you think". So he said: "Well come back in a couple of hours". When I went back in and sat down he said: "Ian, this is the sort of thing we should be doing here" I took from that, it meant approval and this was something worth doing. So for about 2 or 3, I think, torturous months after that I had to make a decision as to whether I would leave Oxford and go to Harvard or not. Fortunately, I decided to take that plunge.
SS Can you remember what made you think it was so interesting? You didn't know any virology or did you?
IW No, but it just seemed like this was a significant crystallographic challenge, for one. It did appear that you were going to learn a lot. Obviously, we didn't know how much until we solved the structure, but it seemed like this could answer a really significant question. It was sort of at the next level of what one was able to do with crystallography, and I just felt that if you did this, it was going to be enormously important. I had to feel that way in order to convince me to leave Oxford because I was really comfortably placed in Oxford. I had a junior research fellowship in one of the Oxford colleges and had a very nice place to live, good food to eat, and I liked my lab environment. It just seemed like an opportunity that wouldn't come very often and I didn't feel that it would be worth waiting and wondering later about what one should have done, so I decided to do it.
SS Don had really not done very much in structure, probably no more than you had done, actually.
IW In a sense. He had worked on a very difficult project, on ATCase, but it hadn't come to fruition as far as the structure was concerned. I had worked on a slightly smaller protein. It was a dimer, his was a hexamer of both regulatory and catalytic subunits - I guess, much more complicated. We had worked at different ends of the game in a way and so it actually worked out that we were quite complementary in our skills.
SS You must have realized you were going to a very different environment in terms of the size of the lab and the expertise.
IW I didn't really in the sense that it was a shock once I got there. I realized it was a smaller lab, but I had run into Steve (Harrison) before at a meeting in Erice and so I knew what he was trying to do with the virus and that impressed me very much. I thought, and I think we all did in those days, that it was absolutely crazy for somebody to try to do something that large. I knew that Lipscomb was there and so I assumed that there was quite a big community of crystallographers and so I didn't get too concerned about that at the time. However when I went there, I realized it was Don and myself. We obviously had support from Steve. Art Olson was there, he was a postdoc. Then upstairs from Lipscomb's group were people, such as Doug Rees, whom we could talk to. It was quite a frightening experience in a way coming from this rather large lab with enormous expertise in crystallography to a place where there was only Don and me and that was it. That was quite a different situation.
SS What stage of the project did you come in on?
IW This is now 1977, I came in the fall of '77. Don had already got crystals of the native hemagglutinin and was just starting to collect some data. So it was a very opportune time to arrive. I arrived basically at the start of the native data collection. John (Skehel) was sending protein in glass vials and we had to purify it, concentrate it, and grow the crystals. I started growing crystals and Don always liked to set up crystals on a Friday and so he always insisted I set crystals up on a Friday - superstitious, he was very superstitious about crystallization. He felt like if we didn't look at the trays until Monday then that was probably a good idea. That was assuming that you weren't going to be there on the weekend which of course was not true. We were there every day. I began then to set crystals up and grow crystals for native data collection, but Don would always do the mounting of the crystals. He didn't in those days trust me to do that. I was less experienced with the sort of data collection that they were doing. They were doing everything on film and I hadn't done film data collection before.
SS What had you done?
IW We had done it on diffractometers and so this was something that was basically completely new to me. Also, there was a lot of maintenance to go on with the data collection. They had mirrors on the X-ray generators so there was quite a lot of heavy maintenance and Steve was really the expert on that aspect of it. It was pretty high maintenance to actually collect data. Gradually, I got more and more involved in data collection and when it came to doing derivatives, I was allowed to mount crystals myself and collect the data!
SS Did you work together? And was it just the two of you?
IW Yes, there was just the two of us and an undergraduate, Frank Escobar, who actually helped process the data. My primary responsibility in those days was actually to collect the data and hand it over to Don or to the undergraduate to process the films and then I would take the data back and then go over and merge the data and actually work on the heavy atom refinement.
SS Was the computer in the lab by then?
IW Yes, there was a computer in the chemistry building. All the data were collected on these small tapes, very tiny Dec tapes, and then it would be transferred over to the main frame computer which was relatively small in those days. Most of the actual processing of the data, and almost all of the structural determination that we did, was done over the telephone line to Columbia University. We would take the tapes over and then send the data down over the wire. One other thing is that we were using cards for running the jobs. The problem was we could only do that in the evenings. Don was still relatively junior in those days and he didn't have the grants that Martin Karplus did, and basically if you ran jobs during the day, it cost a lot more than running them in the evening. So we had to run them every night. Don and I had to gear up every night to run our jobs, basically we were there from early evening through to 6 AM.
SS Were you living by yourself?
IW Yes, I was living in one of the Harvard houses so I could actually eat there as well if I could actually coincide with their meal hours. But it turned out that it was an all night slog. I had just come off that sort of project in Oxford and thinking I never wanted to do that again because we were only allowed to use the computers at night there as well, but for another reason. They were developing computer graphics during the day. I guess it was Tony North and Dave Barry.
SS Dave Barry, I remember him. He spent several years at Washington University.
IW Dave Barry was there doing all the computation during the day and we were allowed to get on about 7 in the evening until about 8 the next day. These were the sort of similar hours we were working at Harvard. At Harvard, it was really frustrating because a lot of the time the line would go down and it would be in the middle of the night. We would try to call and sometimes we would be able to get people to answer and sometimes we wouldn't. We never knew whether we should wait there and the line was likely to come back up or not; it became frustrating some evenings trying to work. However, that was the schedule that we were on basically - it was all night.
SS Was it difficult to get the heavy atom derivative? Did that raise problems?
IW Yes, we had tried this phenylmercury glyoxal derivative which was supposed to be two mercuries attached to a phenylglyoxal, that looked like the most promising derivative. We tried other things, but nothing seemed quite as promising. That was really the one that we focused our attention on and that turned out to be a successful derivative because that was what we used. But we were working under quite a lot of assumptions which turned out to be wrong.
SS Tell me about some of those.
IW Well, again we divided up the tasks. I did a lot of the processing; Don was really good at writing little jiffy programs or bringing in programs and getting them implemented. I tended to do most of the analysis work using those programs. He had found from a Patterson function that there was a threefold axis and we expected that the hemagglutinin would be a trimer. So we were looking for heavy atom sites that were three -fold related. Don had found what he thought was a very clear indication of what the molecular threefold would be, that was not a crystallographic threefold, so that would specify the threefold axis of the trimer. When I was producing possible heavy atom solutions to the Patterson function, I would take my solution and rotate it about the threefold and try to look for the other expected positions and I could never find them or they would be just off. It wouldn't be perfectly threefold, it would be like two would be all right but one would not be perfectly 120 degrees from the others. Now this could have been possible because there were slightly different environments in the crystal and so there could have been some asymmetry in this three-foldness. Also, we were looking for two sites that were close together. In the searches we were doing, I can't remember exactly the distance, let's say it was 12 angstroms apart that we expected the maximum distance, then you should be finding vectors that were 12 angstroms in the Patterson. And so I was trying to use that information to also limit my solution. However, I could never find any two atoms that were that distance apart. Anyhow, I managed to find a couple of solutions right away that seemed to be good overall solutions, but I couldn't find the threefold related ones. Eventually, I got to four solutions and I remember Don looking at my notebook after it seemed like months of work and saying: "Are these the only ones you've managed to find?" as if I hadn't been looking for any others! I said "These are the only ones that match. There are multiple other solutions but they are not consistent with each other and there are no cross vectors that would actually account for these being real solutions if I believed the first two". So I got up to four and then there was still no obvious sign of this threefold relationship. So in some frustration, I took the Patterson map and again got an undergraduate to help me, in this case it was Andrew Cherenson. We just basically manually read off all the peaks of the Patterson map that were not accounted for and that was hundreds and hundreds of peaks. Then we surmised that some of these peaks have to be cross vectors with existing solutions. So we read off the peaks and, by some mathematics, we should be able to find a consistent set of cross peaks that would identify other solutions. In that way we managed to pull out two additional heavy atom solutions. It was at that stage that I realized that we actually had two trimers now and these trimers were not related by the threefold that we originally thought, but were related by a non-crystallographic threefold. There were two trimer solutions - they also had to be in the same molecule because they were relatively close together but were at a different radius and, at last, I knew that we had actually solved the Patterson.
Now, from crystal studies, from density measurements, and from thinking of the crystal packing and how much solvent is there, we had made the assumption that there were two molecules in the asymmetric unit, two trimers, and this only specified one. At that stage, Don was concerned that there was another solution. So he had set about writing a program in order to look for these other solutions, to do it by computational approach rather than by this painstaking way that I had gone through by using a sort of hand approach. In a way, these other solutions should have come out the way I had done it previously, but he was just doing it in an automated way to make sure that nothing was missed. Basically, he quite nicely verified the six solutions that I had found and no others. At this stage, I think that Don was at last convinced that in fact there was only one trimer and that we had all the heavy atom sites and so we could now proceed. I remember that I was absolutely ecstatic that day that I had figured out that the six heavy atom solutions specified one trimer and importantly the true molecular threefold axis. Don was away at the time. I took it over and showed Steve and said: "Look this is it. I'm sure we've got it now". My notebook was very thin. Although there were months and months of work, I had shelves full of computer output, I was really pleased because only the correct solutions had been added in one at a time and eventually there they were and these were the six solutions. If you look at the original Pattersons, you can't spot some of the solutions in the Harker sections. It took some other methodologies or some other ways of analyzing the Patterson other than Harker sections to be able actually to interpret it.
SS People don't do that anymore.
IW Not by hand. They tend to use programs. I think it is a lot less fun and you learn a lot less. You worry about teaching people in your lab that way. They don't have to understand as much of what they are doing.
So at that stage then Don took it up and then realized we could calculate a map but there were no anomalous data. The data were really good enough actually to measure the anomalous data and so we actually thought: well, they've used non-crystallographic symmetry averaging for viruses, so perhaps we could use this threefold symmetry. It wasn't really much used except for the virus cases, but perhaps we could use this, and Don got the programs from Gerard Bricogne to actually start working on. I remember Gerard being extremely skeptical at the time that this threefold would be sufficient to determine the structure. He was happy to help, but we were probably wasting our time, at least that's my recollection. But undaunted, Don went ahead and produced the maps. Then we had a look at them and down the threefold axis, and they pretty much looked uninterpretable. But there was some indication of where the protein boundary would be. We couldn't really interpret the map unless we could see where the protein was. So bad as these maps were, and the phase error was over 75 degrees which is close to random, 90 degrees being random, I took a stab at tracing the envelope around each of these maps that we had drawn onto plastic sheets. I drew by hand what my idea of what the protein boundary would be. Then, the problem was of transferring that information into the computer and I know we got some help from some of the people upstairs who wrote some jiffy programs to be able to basically read this envelope that specified inside there should be density; inside there was one HA trimer and outside there was solvent. Then Don was able to average the map within that boundary and flatten the density outside - these were the early days of solvent modification. When we continued to do the averaging and the flattening, then we could see immediately features appearing which sort of looked protein-like and it looked like a molecule now.
SS What do you mean by protein like?
IW Seeing things like alpha helices and things like rod-shaped density for alpha helices and things like that, also density that was connected to one another. It wasn't immediately apparent what the structure was, but it was apparent that this was really a very long molecule. It wasn't this globular molecule that one was used to seeing with enzymes. It was a very long thin, and when we actually looked at the packing, we could see how in fact it did form this lattice. I remember Don saying it's just like Lincoln logs. I didn't know what Lincoln logs were because this was an American term. But basically you could see that it was a long molecule and if you put one down and put another on it at 90 degrees and then kept packing going upwards around a fourfold axis, a 4-fold screw axis, you could see how you could generate a three-dimensional lattice. You could see also how remarkably tenuous those lattice contacts were and how amazing it was that we even got a crystal. It also explained why something that was so heavily glycosylated could crystallize, and this was the other challenge in those days - there was an enormous amount of glycosylation, more than 25% by weight. There were seven glycosylation sites, something like that, and so you could see why with all these solvent channels, that the variable glycosylation wasn't interfering with the packing because there were only slender contacts at either end. There were protein-protein contacts that were initiating the lattice formation. So once we had that, it looked like we were going to get the structure at that stage, and the threefold averaging would be enough. That again was a huge surprise - only threefold averaging. We also knew that we were likely to be getting an awful lot from the solvent flattening because with 82% solvent, that's a lot of potential phase information. When we put that information in computationally saying that is solvent there and protein there, then that gave us additional information. It was like having another heavy atom derivative that was sufficient to determine the structure. So once we had done a few cycles of refinement, we slightly refined the threefold axis as well, and, basically, again we were in a way fortunate in that the heavy atom positions were so exactly threefold symmetric, this 3-fold axis didn't waver very much. So we were always averaging correctly around this non-crystallographic threefold.
SS I can't tell how long a period this was that you are talking about.
IW We started collecting the data in '77 and I can't remember exactly when we finished data collection, probably mid to late '78 or something like that
SS Then you said you knew you were going to be able to solve it.
IW That wasn't probably until '79 to '80. I worked on that Patterson for probably nine months at least.
SS How did you feel during that time? Depressed? Optimistic?
IW In some ways, it was depressing and, in some ways, it was encouraging because we knew that there was information content there. I was really convinced that the limited solutions I had were correct, but it was frustrating because I couldn't finish the thing off. All the pieces didn't fit together. It wasn't until the end that I could see why they didn't fit together especially because the data weren't that terrific. The crystals weren't that good, the diffraction was sort of "poorish". It was around 3 angstrom data, but it was a fairly large unit cell and with a lot of solvent content. It was really difficult to actually make very good measurements. In retrospect, it was understandable and in fact it was amazing that we actually got a solution out of it. But some days it was sort of quite depressing just to keep plodding on. It seemed like one was banging one's head against the wall for week after week after week with apparently making very little progress.
SS No virologist or even a protein chemist would be interested at this stage in what you were doing.
IW No, no, but John (Skehel) was always interested. He was on the phone all the time. Don and he spent hours and hours on the phone talking to each other. We always knew that because I think Don felt that he had to shout loud enough for John to hear him across the Atlantic without using the telephone. When they were on the phone we always knew, because we could hear him no matter where we were in the lab, we knew when he was talking to John on the phone. John visited fairly frequently and was always very enthusiastic and encouraging and he was always very good at sending material. We were never short of material to be able to generate new crystals.
SS So tell me what is was like when you were really able to see the protein and how you were thinking about it.
IW Once we decided that we had refined things as well as we could, Don calculated the map, and I stacked it up. You may or may not have seen that map, I don't know if Don still has it. In those days, there were almost no computer graphics capabilities so everything had to be done by hand. This was when my experience working with TIM paid off because that was a very tricky map - the TIM map, and hemagglutinin was no different. Don remembers it being a relatively easy map to trace. That's not true!
I was tracing it and some parts of it were easier than others. The bit that formed the triple coiled-coil down the middle, that was the bit that was instantly obvious. But the beta sheet structure at the globular head was very, very difficult to trace and especially connecting up all the ends of the beta sheets. Gradually, I filled in all of the beta sheet strands but I didn't know how to connect them and the density for the connecting loops was very tenuous. And so I would spend hours just staring at the maps.
SS Did you have a sequence then?
IW Yes, we did have a sequence. What one was trying to do then was to connect the sequence onto the protein and so if you could get a long enough stretch with enough aromatic residues then you could actually just go through the sequence and figure out where you were on the sequence. We did know where the triple coiled-coil was expected because of the work from Ward and Dopheide. So we knew roughly what region to look at; we knew it was in the HA2. That was the easiest bit to get clued in initially in the sequence. It was a little bit more difficult to trace the sequence in the globular head, trying to figure out where one was. That's why it was essential to make all the connectivity so you could actually see where the start was and where the end was and then trace it all the way through, or at least try to use the sequence to figure out any ambiguities in the size of loops you were looking for to connect to the next secondary structure element, if you could identify it from those little stretches of sequence. You build it up piece by piece so you know that this corresponds to that piece of sequence and that corresponds to that piece of sequence and you know roughly how many residues you have in between and then you try to fit it all together. It's not as if you trace it all and then you say here is one (the first residue) and you go through to the end, that's not going to happen. So this process took several weeks, the tracing of the map, because it was a big protein. Initially, we were trying to figure out where one monomer was of the trimer and, ultimately, we did actually build in some of the other ones. But we were basically trying to figure out where one monomer was because we didn't need to build all three because they were symmetry-related by non crystallographic symmetry. Then when that occurred we figured out we had almost the entire structure that was put in. We knew then that we had it right back to the bromelain cleavage site.
The question then was how to build a model. Now up until then most building had gone on by using Richards' boxes using wire models. I had built TIM that way; it was 250 residues times two so this was a 500 residue protein which was similar in size to the hemagglutinin monomer. Steve had actually been building one of the subunits of TBSV in a wire model and when I initially came I helped Art Olson a little bit to get started on that model and then Art built that. But it became clear that some graphical methods were becoming available for building things on the computer and Bob Ladner was there and, in those days if we didn't use Harvard programs, we tended to use ones from MRC Cambridge. Bob Ladner was working on "Builder" and he was working on modifications of a program of Bob Diamond's in Cambridge (England) and so we were trying to get it up and running on the computers. Then, we actually got our own computer at Harvard that could actually do that. So we could actually put the graphic device on there. We got our first Evans and Sutherland but again, thanks to the Karplus group, I was destined to build all night. They didn't want me building during the day because I slowed their jobs down and, in fact, they weren't too happy with my running at night either because I was slowing their jobs down then as well. So I got into a few verbal scuffles with some of the people there for how much time I was using building. We were actually developing the Builder program at the same time, trying to make it useful to do such a large structure as well. I worked fairly extensively with Bob Ladner trying to make the program more user friendly and put things in that would be useful.
That was really the first step in the computational part of the project that we actually did on site at Harvard. The rest was done over the phone lines at Columbia and so I basically sat there and built the structure for several months and then obviously we got to the end. It was gratifying too because we actually saw some of the carbohydrate. It was difficult to build in carbohydrate in those days because there was very little experience. There was only one protein structure with an extensive amount of carbohydrate from Hans Deisenhofer (the Fc fragment from an IgG). He came and visited the lab and we looked at the electron density maps trying to figure how one could build the carbohydrate in. So I actually had a stab at building that carbohydrate in. The first molecule actually had carbohydrate on it. We also tried using NMR information from Jeremy Carver at Toronto and he also came down and brought some NMR structures to try to fit.
SS Could you actually see the carbohydrate?
IW Yes, that started to give us clues about what the carbohydrate might be doing, what its roles were. It was quite clear it was covering a lot of the surface probably protecting it from proteolysis;. it was also masking some of the antigenic sites; it was providing monomer-monomer interactions. I think we started to get some glimpses as to what the role of carbohydrate was, in particular when we looked at the evolution of the Hong Kong virus and saw that the carbohydrate started to vary in position so it exposed some sites and covered others. You could see that that was used for masking and unmasking parts of the protein surface and that could be important for antigenic variation. So then we had everything in the computer and it's very hard to actually see the structure. It's such a big structure when it is actually in the computer; so Don thought it would be really nice to get one of these Labquip models.
SS What are those?
IW These are plastic models from a company in Britain. We decided we would have a go at building a trimer so we co-opted John Mack from New York who was a model builder. He came up to Harvard and we constructed a dimer as one unit. This was a three-dimensional plastic model, 1 angstrom per cm. We could actually look at this thing to see what it really looked like and we could obviously trace it out. There were some limited computer-graphics programs that you could use to trace out the molecule, but it was really hard to get a feel for what it was all like without seeing everything at once. So that (three-dimensional) model turned out to be extraordinarily useful for trying to understand the structure. The third monomer was also built and we thought we would be able to assemble and disassemble the trimer, but they were so intertwined that it was impossible.
SS I am trying to recall now what would have struck you at that time. You already knew about the peptide that was similar in sequence to the Sendai fusion peptide. But you tell me.
IW OK, so the immediate striking thing was the odd shape of the molecule; however, it did have a globular head and a fibrous tail, so it was like a fusion of a fibrous protein plus a globular protein.
SS So what questions were you thinking about? That is, you had electron micrograph pictures of the molecule so you must have been had some ideas about its shape.
IW We knew it was going to be a long-shaped molecule but we didn't know how the secondary structure was going to relate to that. When I say globular head I mean it sort of packed together with a core like a globular protein, whereas the fibrous thing looked something more like collagen or something like that with this big repeating coiled-coil alpha helical structure. That looked more like a fibrous protein than a globular protein although it did have aspects of it that were globular such as the beta sheet down at the base near the membrane from the association of both HA2 and HA1. In that sense, yes we knew the shape of the molecule and it was consistent with the shape, but it was really seeing how this shape was manifested from the secondary structure elements.
We looked to see where the fusion peptide would be and that was surprisingly buried, inserted in between the helices. That was not what we initially expected but, on reflection, that was sensible because it is very hydrophobic sequence and we immediately knew was that there had to be a conformational change that went on in going from HA0 to HA1 because there were 17 or 18 angstroms from where the clipped bond took place. So clearly there had to be some rearrangement. We also surmised that because the globular head was further from the membrane that it had to contain the receptor binding site and one of the first things we did was to look at the various HA sequences. By that stage, we were accumulating sequences every year and we were keeping, in the lab, a record of all of those by alignment. Then we plotted the antigenic variation - first to look for the sites that were mutating fastest, but more importantly for receptor binding, which sites weren't mutating. So when we actually plotted all of this, it became obvious that there was a site at the top of the molecule that was likely to be the receptor binding site. This wasn't some huge big pocket that we normally would expect to see for a binding site, like we're all used to seeing in enzyme structures. This was actually a very shallow pocket, but it was surrounded by all these hypervariable loops and so then it started to become pretty obvious that this was how the virus escaped from neutralization. There was conservation of residues that were recessed from the protein's surface, but were surrounded by all these loops and the loops kept changing all the time. The loops had no structural requirement to remain even the same shape, although they clearly probably do, but you could actually put all the sequence variation in and still keep the same framework and the receptor binding site. Don got very excited about this and started plotting at all these antigenic variants. So the next stage was actually to get a schematic and again there were no programs like there are today and we didn't know which view would we use. I don't know how we heard that Hidde Ploegh could actually draw cartoons of protein structures. I had a similar experience with TIM. We actually co-opted an artist, also a scientist, to draw the structure.
SS I couldn't tell when you acknowledged Hidde Ploegh in the paper, if he had drawn it by hand.
IW Yes, but from stereopairs of the alpha carbon backbone trace. I would produce the stereo pairs and we tried all kinds of different views. Jane Richardson was starting to develop some of her renditions of other proteins following on from hemoglobin diagrams that Irwin Geiss had been doing as hand drawings of proteins for awhile and this was taking it to an additional level of trying to represent alpha helices and beta sheets by ribbons and arrows.
Anyhow, we heard that Hidde Ploegh was an artist in his spare time and liked to draw things. So I started working with him and first we had to decide on the view and eventually we got a view that Don realized we could put dots on to represent the antigenic variation. He would take it and photocopy it - color photocopying was only just starting in those days - we got some color photocopies and he set a dot on every place that there was an amino acid change from one year to another. So we went from '68 to '72, from '72 to '75, from '75 to '77 etc. and then we could actually start to see what was emerging was that there were discreet areas where there were more mutations than in other areas and that was when the structure really started to talk. We could start to understand antigenic variation. We could understand how you have this conserved receptor binding site. We didn't really know what to make of fusion, but we knew from some studies that came later that the fusion peptide had to be extruded. I remember making a model of the fusion peptide arm extruded because we knew that then you got aggregation at low pH. These N-terminal fusion arms had to come out and could be enzymatically chopped off at low pH. John Skehel had showed that if you chopped the arm off, the HA became soluble again. We had some sort of idea, not really of the detailed changes that were found later, but that there had to be some conformational changes that were going on.
SS It's interesting to read the papers and see how people go back and forth between saying there had to be big changes but the changes weren't big. Now everyone says they knew they were going to be big but that wasn't what they were writing.
IW I don't think we had any idea how big the changes were. One other thing we knew, from the EM, was that there seemed to be some elongation of the structure and that suggested that there might be some change in the HA1/HA2 interfaces somehow that made it longer. The only way we could see that could make it longer was for the helices to rearrange in some way and that was what we were seeing, but we didn't know that in those days. We did know that we had the Ward and Dopheide prediction of where the triple helix would be and, when we determined the structure, we realized it didn't coincide with what they had predicted as being coiled-coil. In fact, the N-terminal residues they had predicted as coiled-coil were not coiled-coil in this form of the structure. I guess we just put that aside and thought it was a missed prediction or something like that. But the rest of it was predicted correctly but actually it went on a bit longer. So the prediction was about the right length, but it was moved relative to what we saw. You could understand why it wasn't predicted so well at the other end because the helices started to splay apart so they no longer had all the hydrophobic leucines and valines in the center here but actually started to have charged residues and it even looked like there might be some water molecules or ions in the middle between the helices. The helices were not tightly packed at one end, but were splaying out and that was below the site in which the N-terminus of the HA2, the fusion peptide, was inserted.
SS Don tells the story about somebody challenging you at a meeting.
IW I was actually giving a seminar at Princeton. I was taken out for lunch and over a discussion, one of the postdocs there started saying that he thought that the structure we determined was wrong and that I had basically got the trace backwards. He said that there was no way the structure could be right; it didn't agree with his prediction of coiled-coils. I was a little taken aback at the time because I sort of felt that I had been setup for this meeting. Anyway, I tried to point out to him that his coiled-coiled prediction wasn't that different from Ward and Dopheide's. If you go back and read Ward and Dopheide, they in fact had predicted that the coiled-coil would extend backwards past that and his prediction was no different and there was nothing surprising about it. He was in fact correct that there was a region which was predicted to be coiled-coil that we didn't see. But at that stage we were starting to get an inkling that there were some conformational changes going on and I think I pointed out previously that there could have been some changes that actually extended the helix based on the EM work. At that stage we were becoming aware of the work that John was doing showing that there were possible conformational changes going on. His studies with fusion mutants also suggested that something could be rearranging. It wasn't clear in any way what was happening. We didn't expect the massive extent of the changes when that structure (the structure after exposure to low pH) got solved in Don's lab. It seemed to me, at least that it wasn't inconsistent with that disputed stretch at some stage being helical, although I couldn't possibly have known that. I was more irritated about the way it was done rather than the actual challenge because I knew damn well - I had sweated with that structure - that the structure was OK. I was upset that somebody would actually challenge that I had actually misinterpreted it. I had spent a long time on that. I was very careful with the analysis and map interpretation. That's one aspect I really enjoyed doing was the building and I knew there was nothing wrong with that.
SS I would like to get some feeling for how knowing this structure influenced your thinking.
IW In those days, protein sequences were almost as hard to come by as protein structures and the most important thing was that we had all this sequence information. That was really the fortunate thing. So it was immediately apparent that we could actually relate that sequence information to the structure and map. I pointed out how Don stuck these dots on the schematic pictures of the HA. It became instantly obvious that we could actually relate this sort of information to virologists. This is something they would understand.
In those days, you really didn't know what an antigenic site looked like. We didn't know whether the whole surface was antigenic or only part of it, whether there was dominance or how the virus escaped from neutralization. These were problems that were still raging in the '80's, never mind in the '70's, and this was really the first glimpse of what antigenic sites looked like and how the virus evolved. It was actually thrilling to be able to realize that we actually had a structure that talked and that this wasn't going to take months or years to understand. It became instantly obvious what some of the lessons were that we were going to learn. I remember this very distinctly because we used to get asked by the CDC every year to look at the structure and tell them if we thought there was sufficient new changes that would warrant a change in the vaccine.
SS This must have been after the structure was determined?
IW After the structure. We had proposed that there were at least 4 antigenic sites, A, B,C and D, maybe five and that two of them clearly seemed to be quite dominant, the ones at the top, this loop at A and the one at the top of the helix at B. For a new epidemic to occur, you needed mutations in each of these areas. Then when we started to see that carbohydrate started changing, we could see that some areas were being masked and some were opening up; then some areas would have to evolve whereas others didn't have to evolve because they were being covered by the carbohydrate. Since then I have been fascinated by the role of carbohydrate and I'm still working on the roles of carbohydrate with the proteins of the immune system. It just gave us that opportunity to see what was going on at a molecular level. That was really a stunningly important observation. For several years after that we kept in touch with Alan Kendal and, in particular, with Nancy Cox just looking at sequence variations to try to figure out whether these were substantial enough to think that a new epidemic might or might not emerge. Later, I did an Annual Review article with Nancy. I think this was some very practical use we couldn't have anticipated from the structure.
SS That has not gotten as much attention these days at it probably did at the very beginning because of all the other things that are happening with the hemagglutinin and because there are other proteins that you can look at the antigenic sites. I'm glad you brought that out.
IW I think that now it's routine because they know what to look for. It's been over 20 years that we have been able to take advantage of that sort of information. That was a very important consideration in those days and it was a real practical use for trying to decide if there was likely to be a new epidemic strain, or was it something that was tangential and wasn't likely to be important. It appeared also to let us see that it took maybe 2-3 years to evolve enough mutations to be able to evade the previous immune response and that you had to have mutations in each of these exposed areas.
SS Either you had to have four changes or you had to have changes in the carbohydrate that covered them up.
IW Or some variation of that.
SS At that same time people, including me, were using tunicamycin, an inhibitor of glycosylation, to show that carbohydrate didn't seem to have much of a role in replication and assembly.
IW We published a paper, I think in about 1980-81, in which we did bring in the roles of carbohydrate. I made a lot of figures showing the glycosylation and trying to understand it.
SS And carbohydrate could be playing a more important role in evading the immune response than it does in the actual virus structure itself.
IW Yes, although there are some carbohydrates that are between subunits that might help in the stabilization of the subunit contacts. That was perfectly obvious with one of the ones in the globular heads that seems to be very conserved suggesting a role for that too.
SS So after this first structure came out then you said you spent another year and a half in Don's lab, what were you doing?
IW We started a collaboration with Jim Paulson. He is now at Scripps, in those days, Jim was at UCLA. He had been noticing differences between the 2,3 and 2,6 linkages (in sialic acids) and we were trying to understand what the difference in specificity was and were looking at differences between horse and human preferences for the linkages and things like that. And so, I was trying to soak the (hemagglutinin) crystal with sialyl-lactose to get the 2,6 forms into it and was doing this along with a student, Andrew Cherenson. I actually had this huge crystal I had grown and was reserving that for doing the soaking experiment for when we were going to the synchrotron. We didn't go to the synchrotron very often. The only time I went to the synchrotron in Don's lab was once when we went to Cornell and this crystal didn't diffract! It was the biggest damn crystal; it didn't diffract at all. It was very frustrating because I thought we would get a complete data set. We did collect some of the data sets there. We did have some low resolution studies so Andrew Cherenson wrote an undergraduate thesis for the Biochemistry Department. He was an undergraduate doing a senior thesis project. We did have some initial indications of where binding sites might be, but the density wasn't very good. That was really what I was doing - trying to identify the binding site - although we had pretty much identified it from the sequence comparisons. We had that prediction in the Nature paper too; we thought we knew where the receptor binding site was just from the conservation of residues and also we knew from residues that changed the specificity from 2,3 to 2,6 linkages or vice versa and they turned out they were in that location.
SS At this time you must have been thinking about what the next step would be in your scientific career. Tell me what you were thinking about.
IW My intent at that stage was to go back to England and I thought I had a job lined up at ICRF (Imperial Cancer Research Fund) which was sort of discussed before I even went to Harvard as a possibility. There was a lot of hemming and hawing and I couldn't get any decision as to whether they wanted to do structural work or not. But they wouldn't say yes, and, more importantly, they wouldn't say no! This was Walter Bodmer and Mike Crumpton. Then I started looking to see if there was anything else. In those days, amazingly, the crystallographic labs were almost decreasing rather than increasing. I remember at that same stage Don was getting tenure. There was this huge discussion of whether you needed another faculty member in crystallography at Harvard. There was Lipscomb in Chemistry and Steve (Harrison) in molecular biology. It was like that all over - whether you needed more structural people. It was really not a good time to go looking for a job.
I went to give a seminar at a virology meeting in Salt Lake City in 1980. This was shortly after the structure had been determined and I was not talking until more than halfway through the meeting. Everybody had given my talk by the time I got to it! Everybody showed our slides. Actually, it was wonderful from that point of view. This wasn't something where "here is this structure guy we'll listen to him for half an hour and then forget what he said". Everybody was showing our slides; they had taken the pictures from the paper. I remember running into Richard Lerner there and he was very enthusiastic about the work. There were some challenging questions and he was very supportive. I didn't know who he was then. He had done some work on flu and that was all I knew.
I went back and thought I had better start looking for a job someplace else (other than in England) and, well, California would be a nice place. I lined myself up some job interviews. Don was talking all over the globe at that stage but hadn't yet talked in California. So I thought I had better talk out there before he got there - not that he was looking for a job. I lined up interviews at UCSD (University of California, San Diego), UCLA, UC Berkeley, UCSF and started to go on the circuit. I started off in Berkeley and San Francisco and worked my way down. I had run into Richard Lerner again when he visited Harvard. He was at Scripps and had become very interested in antipeptide vaccines, and was working with peptides of hemagglutinin. I told him that I was coming to San Diego and he said: "you have to come and visit". At that stage, I only had the Saturday left before I felt I had to come back as I couldn't be away for too long. I said that I would go and see him on a Saturday morning. He was at that stage Chair of the Committee for Molecular Genetics, I think it was. I thought I would go and pay him a courtesy call on Saturday morning after I gave my seminar at UCSD on the Friday and get to the zoo in the afternoon. To my astonishment, I was there all day and I had a job offer in my hand that I took back. I really was not expecting it. I remember going back and showing Don and we were all amazed. That set me thinking, well maybe I better consider this, but it was completely out of the blue. I tried to pursue the ICRF possibility with even more vigor and tried to get in contact with them. I was met with the response - well this is typically American, an unseeming rush into doing something quickly. I didn't think it was much of a rush for ICRF because we had been going through this process for five years or at least, I thought we had. Eventually I had to give them a deadline because Richard Lerner was pressing me and I managed amazingly to hold off for several months. We didn't have faxes or e-mail then and I got a telegram from ICRF saying they still were unable to make a decision - so they still did not said no, and to this day they have never said no.
I decided I would take this job. It seemed like at this stage it would be quite a challenge, maybe a bit risky because there was no structural community out here. There was no biophysics. There was no molecular biology even and that was what Richard was trying to build up. There were only a handful of people in his Department. He had just gotten money from Johnson & Johnson to build a molecular biology building and to become Chairman of the Department. He decided right away he wanted structure and graphics. I remember talking to Art Olson about possibilities of graphics here because I thought his job might be finishing up in Berkeley.
SS When Richard Lerner was interested in you, he must have been interested in having you work on influenza.
IW Yes on peptides, trying to understand how it was that you could make antibodies against the peptides. When I decided to come here, I switched into looking at antibodies against the hemagglutinin and that was really the main project I had for the first few years: trying to figure out how to determine an antibody structure with a peptide, to look and see what conformation that peptide had, and whether it had any relationship to what it had in the hemagglutinin. In those days, it was not understandable how you could make an antibody against a peptide that would react with the hemagglutinin and a lot of people were skeptical about it.
SS Was that because they assumed it required a conformational change?
IW Or that it would require the peptide to fold. In retrospect, what we now know is that you can often make very good antibodies against peptides that would react with intact viruses if the peptides are in extended loops, if they mimic a loop. It turned out for some reason you could never really make a very effective neutralizing antibody against flu by that methodology compared to, for example, foot and mouth disease virus or poliovirus where there were some exposed loops that worked fairly well.
SS I didn't realize you had actually continued with hemagglutinin when you left Don's lab.
IW On the antipeptides. I was also trying to crystallize one of the other hemagglutinins. I think it was a B/Lee/40 strain. We had some material from John at some stage and we were talking to people at CDC about possibly getting other ones but none of that really panned out. We did get crystals of that and PR/8/34, but they didn't diffract very well.
SS As I understand it, it is only now that they (Don Wiley's lab) have gotten crystals of another one.
IW It's taken a long time. It just shows how fortunate Don and John were to get the initial crystals of the first one because it has been difficult compared to the neuraminidases where many have crystallized.
SS I asked John how the hemagglutinin was chosen. I think that was the one that cleaved. The others don't cleave as well. You were going to go back and tell me something about when you finally finished the structure and had to write it up.
IW We had a lot of material. It was a huge structure to describe and then we had all the mapping of the antigenic determinants. In those days it was hard, as it still is, to get a paper into Science and Nature. There was never any doubt that we would send it to Nature and Don thought that we could actually divide it into two papers. I guess that I was quite skeptical that they would actually take two independent articles but we decided to write it up that way because that seemed to be a logical way. I focused mainly on trying to figure out the structure paper and Don tried to figure out how we would write up the antigenic variation paper. I made the diagrams; in fact, some of them were actually hand drawn including the carbohydrate binding site.
So I drew one of the hand drawings myself but we used the template that we had from Hidde Ploegh for putting on all the dots for the antigenic variation paper. The papers went back and forth several times, like any paper, as we were not quite sure what the punch line should be. I remember Steve (Harrison) coming in and telling us we were missing the point and throwing away a great structure at one stage. Steve didn't feel that we were highlighting the structural aspects that were the most important, or, at least, not making the points in the clearest way. I remember focusing then on the antigenic variation paper where I felt there was too much virology in it. It should have been more of a structure paper. I was happy with more virology in the structure paper but more structure in the antigenic variation paper. It seems odd, but that seemed the best mix for explaining what we had learned.
We went back and forth with several drafts. We would send stuff over to John and he'd send it back and we would send it back again. Eventually we came to some consensus and so we sent these papers off to Nature. Then we didn't hear anything for awhile, and then we got a letter saying that they were both accepted without change. It was just like - "WHAT". They accepted them both right away. It must have been around Christmas when the proofs arrived and Don was away and I was at home in Scotland and they wanted the proofs returned as quickly as possible, they were going to send them to us. There was no faxing or anything like that available. They put them on a train in London and shipped them up to me in Perth. There was nobody else to talk to, Don was away and I can't remember where John was. I proofed the things and sent them back and then we just waited for them to appear. I got a call a week or two later from a friend of mine in Oxford who had been looking out for the paper. He is not a scientist. In those days they sold Nature on the street. You could go to a newsagent and buy a copy of Nature. He phoned me up to tell me it was on the front cover. I said: "you must be joking, they hadn't talked to us about being on the cover. You must be wrong". And I wouldn't believe this. He got a copy, he must have photocopied it and sent it to me. That was the first we actually knew they had reconstructed the figure we had sent them and put it on the front cover. So this made an even bigger splash. It was really nice that they has actually selected this and put it on the cover.
The following report comes from my e-mail conversation with Graeme Laver:
Don Wiley was in Australia visiting Gillian Air's lab the day the Nature with the HA structure on the cover arrived in Canberra. Graeme remembered taking the library copy to show Don (who was unaware of the cover illustration ) and hearing the typical Wiley exclamation of "Holy Cow"!
| Some historical highlights: structural virology
and virology |
| Solving the Structure of Icosahedral Plant Viruses | Picornavirus Structure | Poliovirus | Polio
The Influenza Virus Hemagglutinin | The Influenza Virus Neuraminidase | Issues of Science and Society |
contributors| Home |