Viruses From Structure to Biology

oralhist.gif (766 bytes)

dwiley.gif (1590 bytes)

Recorded and edited by
Sondra Schlesinger

April 1 and 5, 1999

wiley.jpg (29769 bytes)

second tape

April 5, 1999

SS O.K. Don - let's go back to London.

DW As I mentioned earlier, I think the most important thing we were doing at the time was trying to figure out what the oligomeric state of the hemagglutinin was, if it was a trimer. I also mentioned that Judy White and Mary Jane Gething sequenced the fusion peptide of Sendai, which totally convinced us at that time that HA was a membrane fusion protein because of the similarity of the N-terminus of the second chain of F and HA2.

I think most of the time we worked very hard and John's lab was just a place where we worked like crazy all day. It was a lot of fun. It was more fun than I've had since because I got to really work at the bench every day and we talked about viruses and membrane proteins all the time. John, because he's a virologist, knows all kinds of stories about viruses so you might say I learned the lore of virology then, whereas I had learned the facts and so on from the earlier course by Alice Huang.

I recall I went to a meeting at one point in Baden-Baden. It was virology and other things and I met people like (Graeme) Laver and (Rob) Webster and in fact I think that was when I first heard about histocompatibility antigens. One of these guys gave a talk. I couldn't understand a word and John explained that there was this problem with histocompatibility antigens and what were they doing with antigenicity.

SS That was when there was still no idea of how MHC restriction could work.

DW Yes, the dual receptor. I think it was about a year after the original paper came out by (Peter) Doherty and (Rolf) Zinkernagel and there were some people worrying about the basic premise. John and I were thinking: how do you get antigenic variation? Antigenic variation was a response in which antibodies no longer were able to respond to a variant virus and the virus differed only slightly from the previous virus. And the sequences hadn't been done at that time, they didn't come along until about '79 or '80 when DNA sequencing and RNA sequencing became possible. So there was a lot of guess work as to what was going on but there must be changes in the structure. They were thought to be in the hemagglutinin because of the antibody experiments and therefore the focus was on the humoral immune system. The antibodies were changing and antibodies were what was being escaped from.

But during this time and maybe slightly after that there were experiments going on in Brigitte Askonas' laboratory and I think they were in collaboration with John (Skehel). You could do these experiments because flu has a multisegmented genome, and you could do experiments where you isolated different genes in different viruses so that you could do genetics. You could attempt to understand the genetic basis of a trait, for example if a T cell were tracking some antigen you could figure out which antigen that was by swapping the genes around until you had a virus that had only that gene from the virus that you had raised the immune response against. What those experiments were indicating clearly was that surface proteins, the glycoproteins, were not the only thing involved in the immune response. Whereas certainly my simple-minded view and, I think that of other non-immunologists, was that it was antibodies and antigens and that was it. But if there were proteins that were inside the virus or, only present inside the cell, it was quite mysterious how they could get involved in an immune response. And so, even in those days, Skehel and I used to talk about this and wonder what was going on and speculate that - you know Strominger was working on histocompatibility antigens - whether there would ever be enough histocompatibility antigens to do structure on as well.

I think that probably is about it for England, so maybe we should come on to the U.S. and to actually solving the structure. Solving the structure was difficult. I was inexperienced. Collecting data was difficult because the crystals diffracted poorly. It took 12 hours to get a one degree oscillation photograph.

I think I mentioned last time that we had crystals almost from the beginning. The first batch of material that John had sent me I was able to get some crystals. It turned out that the first crystals were highly twined but by good luck a similar condition - moving from ammonium sulfate to sodium citrate - gave crystals that weren't twined and had a different morphology. In fact I always worried after that, because we never really understood why we got these horrible twins which you couldn't do any work on. They were the pentagonal dodecahedra and with only slightly different conditions we got these beautiful octahedra and I thought the difference was ammonium sulfate in the one case and sodium citrate in the other. But sometimes I was finding the bad crystals in the sodium citrate, so in the back of our minds, at times, was the worry that we would lose the new crystal form and go back to the old crystal form forever. And that would be it. We would never get the structure solved. Both of them could be arising in the same drop, some kinetic phenomena or something that we still don't understand. If all of a sudden only the wrong ones appeared we would have been out of luck because we did need new batches of crystals constantly because the data collection required 12 hours for each oscillation photograph. You could shine the X-ray beam on one place on the crystals only for about 24 hours before that place died, as we would say, became damaged by the beam so that it didn't diffract well. The crystals were large enough - about a millimeter by a millimeter and the X-ray camera beam was collimated to be very small - only about a hundred to 200 microns. We could actually translate the crystal and expose it in a series of places. Since it was shaped a bit like an octahedron I could go across the longest axis in the middle of it and get as many as 5 photographs and then I would translate up and get 3 and translate down below that line and get 3 more. Ian Wilson and I were doing this.

SS I have two questions: Was the choice of the '68 flu, the Hong Kong strain done by John and was that done because that was the only one in which the HA was cleaved (from the virus by bromelain)?

DW Those are good questions. You probably have to ask John. I'm pretty sure that the Brand and Skehel paper was on the X-31 strain in 1968. Shortly after that we did try a 1972 and 1975 (strain). In fact, that reminds me that part of what I was doing in England - we were making different strains hoping to get an even better crystal. We worked our way through a number of human strains. I remember '72, '75, I seem to remember '43 as well. Those all crystallized but they gave crystals that were not suitable for diffraction, either very thin or disordered or something. So it turns out that '68 was a lucky choice you might say because it was the only one that worked. It's the only one that's worked up to now and here we are in 1999 and I know that some other groups have tried different strains and haven't succeeded. It does happen that in the laboratory at the moment we have a talented post doc Dr. Ya Ha; and Ya has succeeded in crystallizing a number of other HAs from avian and other flu's. So there will be more structures in the not too distant future.

SS Now why don't you introduce Ian (Wilson).

DW Ian Wilson was a post doctoral fellow who came from David Phillips lab in Oxford and had a lot of experience in what I would call the late stages of a crystal structure determination - difference maps, refinements, model building. Model building in those days was very difficult compared to what it is now. For example when we first started building models of hemagglutinin, there weren't computer programs that could do it and people would build models by hand. We never did. We just made dots in the electron density maps where the alpha carbons were and later we had a good program here we could use for building.

SS So by 1979-1980 you were already using computers?

DW Yes, in some sense crystallography has been limited by computers at every stage. The faster the computers were, the more you could do. But to get back to actually collecting the data, what's done now in 20 minutes at the synchrotron or a few days in the lab with an area detector, with this molecule took approximately 8 months to collect the native and a single derivative data set.

SS It was all done here (at Harvard)?

DW Yes, it was all done here and it was done by Ian and me. I think it fell to me mostly to collect data and we had a couple of undergraduates who helped us process films after we had set it up on a film scanner. The typical thing was you would take a picture and then after 24 hours you would have to translate the crystal. So somebody had to be here every 24 hours all during the weekends and we were doing it at 4 degrees C because the crystals seemed to last a lot better and we had these horrible cooling apparatuses that were blowing cold air on the crystals and those would always freeze up and we would have to come in and get the ice off them. Everybody helped us. Steve Harrison used to come in some mornings and squirt alcohol on the coils to de-ice them and so on. It was a long struggle. I hesitate to say boring because it wasn't boring, but it sounds boring. Every few days you had to mount new crystals in capillaries, put them on the beam, get them lined up and in those days we had to get them lined up quite accurately. We got so we could line up crystals within 5 minutes of arc and with 5 minute exposures on a polaroid camera. And you got really good at some things because you did them so many times. I must say for myself I like things like that. It's comforting to do the same thing over and over again when you know that at the end there is going to be something worthwhile. And so I didn't mind at all being involved in collecting a lot of data and then processing the data. Every film - there were 2 in a pack so there were maybe 50 to 100 packs of film so hundred to 200 pieces of film - every piece of film had to be looked at, measured, punched and put on an automatic film scanner and processed so that was also a bit of an assembly line process but the assembly line worker was me. Occasionally an undergraduate would help us. You had to keep the film scanner going mostly day and night because it took time to do each step so there was this mentality that we were factory workers trying to get something accomplished. It was more like building something because we knew at the end there would be something worthwhile so you didn't feel you had to do the same thing over and over again. You knew there was an end to it. But the end might be 6 months away or even more like a year away.

SS Just you and Ian were doing this?

DW Yes and a couple of undergraduates that helped. Once we were able to process all the data and get the so called h, k, l 's and Fs then the first problem was to solve the position of the first heavy atom. We had what we thought was a derivative which was a mercury derivative that we thought would go on an arginine - mercury phenylglyoxal. We thought that 2 mercuries would go on 1 arginine and we wanted as much scattering power as we could get because we had a whole trimer in the asymmetric unit of about 200,000 Daltons so we had purposely used this mercury phenylglyoxal which had been pioneered in Lipscomb's lab by a student named Hugo Monico.

It turns out that some of the things we thought about that molecule (mercury phenylglyoxal) were incorrect so we wanted to be sure there were 2 mercuries because when you were going to be collecting data for 6 months, you wanted to be darn sure there is a mercury in there. We used radioactive mercury to show there were mercuries and to the best of our ability we found that the stoichiometry was correct, that is, there seemed to be 6 mercuries per trimer which would be 2 at an arginine where it would have been attacked by this phenyl glyoxal. In retrospect 2 phenyl glyoxals did bind but in 2 different places quite a distance from each other and not on an arginine.

SS It must have taken a long time to figure that out.

DW Well, we were always hoping that the mercury heavy atoms would be close to each other and they turned out not to be close to each other but once we had the data for a single mercury derivative and the native form, we were also collecting data on a second derivative. We never finished that - it was taking many months - and we solved the Hg Patterson in the meantime. In solving the Patterson, I can't recall for certain but I think it took something on the order of 9 months. And again I would put some of that down to inexperience, although the Patterson map itself was extremely noisy. When you look back at it today, it's very difficult to understand exactly how we got it, frankly. I was writing computer programs to do it, because I liked writing computer programs. I was doing something called the symmetry minimum function or symmetry sum function - an idea that Steve Harrison had used on bushy stunt virus - not to find heavy atoms but actually to find the whole electron density and which had been pioneered in various labs. But I was following work that had been done in Lipscomb's lab by Simpson and Dobbrot and others. It was a lonely time in the sense that you came in in the morning and worked at the computer program and went home at night and worried about all the bugs and came in the morning and worked on it some more. And I woke up in the middle of the night and worried about the bugs, woke up at 7 AM and worried about the bugs; it's that kind of experience. At the same time we were keeping the data collection going. So there was a mechanical aspect to what we were doing as well as this computer aspect.

SS Computers were not PCs!

DW Oh no, these were computers that were at a computing center so you had to carry your cards over to the computing center to do the work. By this time, I guess we had a PDP-11 which was running the film scanners but the real computing was done remotely.

SS You were still using punch cards?

DW Yes by this time a lot was being run on magnetic tapes but I think we were still entering data and carrying it around on punch cards at those times, Not everything. For example, the film scanner would produce magnetic tape and these large magnetic tapes that you've seen - the old fashion ones that are 12 inches in diameter - and we had hundreds of those and we would be carrying them over to the computing center and they would download our film scans. We then had a small computer which would do film scanning and would produce h, k, l's and F's eventually. As I said solving the location of the heavy atoms was a real struggle and Ian and I were struggling at it for a long time. The computer programs were giving us hints as to what the solution was going to look like. We knew to look for triangles because the molecule was a trimer. Now it wouldn't have to have triangles but we thought it would. So through the mass of potential peaks that the Patterson said were possible, we would sift through those looking for things that in Pattersons - you have self vectors - vectors between an atom and itself and cross vectors between an atom and a non-crystallographically related atom. And the total number of such vectors can get to be a very large number in the space group we were in, so there was a lot of searching and these computer programs were made to search out all possible solutions and tell you that here are all possible 3-atom solutions to the problem, and here are all possible 4-atom solutions; and we're looking for all the 6-atom solutions. We never really produced the six atom solution directly by a computer program - eventually we had what looked like some good sites, but not everything, and difference displays were not showing it - in retrospect they did.

There was a contradiction happening which was part of the problem. Using Rossmann's rotation function we had located - we thought - where the non-crystallographic 3-fold would be, where the trimeric axis of the molecule would be. And we expected the heavy atoms to be around that position, but we weren't finding heavy atoms around that position and in retrospect it turns out that we were getting a misleading view point from the rotation function. The rotation function was picking up a packing vector - a pseudo 3-fold axis which was relating two of our molecules which were related to each other by a 90 degrees rotation from each other which was like a and b in a cube are related to each other by the diagonal the 1,1,1; you can go from a to b by a rotation around the 1,1,1, of 120 degrees. Our molecules, in retrospect, were located approximately on the a-b plane approximately either along the a or the ab diagonal, I can't recall. It doesn't matter, only that one molecule was located 90 degrees away from another and the non-crystallographic 3-fold that we were seeing took one of those over to another. We had other peaks in the rotation function and, in retrospect, they turned out to be the right ones but were much smaller than the ones we had been concentrating on.

But the actual solution of the heavy atoms - Ian made a key contribution of piecing it together right at the very end when we had pieces and actually he did something that was very dramatic. He left on my desk one morning, a styrofoam model; a piece of Styrofoam with some sticks sticking up out of it and on the end of those sticks, balls and the balls represented the mercury positions and they were in a triangle. That was the day we knew it - when the balls formed a triangle and it turned out that there were 6 heavy atom sites - 2 triangles and they were concentric around the same 3-fold axis and so that was totally convincing. At that point we had thought there were 2 molecules in the asymmetric unit because, that is to say, the 2 trimers we were searching for instead of only one but having found one set of heavy atoms, we thought well there must have been one trimer. That would have made the amount of solvent in the crystal very high - 78 - 80%, I can't remember. And so we spent the next week or so frantically trying to find any other heavy atoms. I wrote some computer program called the higher order image seeking function, which was one of the fun things I did at that time, which took the known heavy atoms positions and went searching for other heavy atom positions - a common thing done now and it found none. I mean I was totally convinced that it would find them if they were there. For example I could start with any subset of known positions and find all the rest and I was finding nothing. So we finally got around to calculating the first SIR (single isomorphous replacement) phase map with just these mercuries from one derivative and it showed a mess - electron density everywhere, But we had a great advantage; we knew where the 3-fold axis was and so we looked down that 3-fold axis and you really couldn't even see where the molecule was. You looked down the 3-fold axis and there was density everywhere and you couldn't even imagine its outline. In retrospect, if you look back you can't see it, but you could do a trick that Steve had done and, I think, Aaron Klug had done on tobacco mosaic virus protein. You could simply average about the 3-fold axis - a kind of illegal operation in crystallographic terms because you're averaging things that shouldn't be averaged - like the a axis with the c-axis and all sorts of things but the idea would be that anything that was the molecule would average up and the noise would average down. So you would get an increase in signal to noise and when we did that immediately the molecule jumped out of the noise. It was so dramatic - one look and you knew there is the molecule and there's no other molecule because you looked all over the electron density map. Now of course you would have averaged the other molecules away in some way- that's true. Still, once you looked at this molecule and saw how clean it was, and that it packed to form a lattice - and the fact you could find no other heavy atom derivatives it was clear that the molecule was found.

Actually, going back to the heavy atom part, I had actually gone on vacation for a week down to the Cape and it was during that time that Ian finally found the last idea of where all the heavy atoms were. Despite all this work we had done for such a long time, it was actually the morning that I came back from vacation, Ian had put this thing on my desk. And interestingly enough within the last couple of years, Peter Rosenthal and Xiaodong Zhang had an equivalently difficult problem. They had the influenza C hemagglutinin called hemagglutinin and esterase-fusion protein and they couldn't find the heavy atoms. It was just a nightmare. Even having lived through it all before, we still couldn't make use of all the tricks and everything we had done then and it was still difficult and the reason was the signal to noise was extremely low and you were mostly looking at noise. There was a signal and you knew that when you finally got the answer you would know you had the answer because it had to have this certain criteria - it had to form triangles. In that case they did get it and I came in one morning in my office and they had this little Styrofoam model with these corks where you can see 2 triangles of mercuries (One triangle at the receptor binding sites and a second triangle of the esterase active sites). I don't know what became of the model. It is probably around here some place. It was obviously a key moment. Ian and I had really worked together, while in the Flu-C case which was almost 20 years later, I was only a supervisor; at least a very interested supervisor.

Anyway, back to the electron density. We then used the method of Bricogne which had made practical the idea that you could determine phases by using non crystallographic symmetry and by the time we used it, I think that it had been used only twice before - once by Aaron Klug on the 17-fold, 2-disc aggregate of tobacco mosaic virus protein and once by Steve Harrison and his colleagues on bushy stunt virus. And with Steve right next door I had followed closely how that was done and really knew something about all the computer programs and so when it became necessary to write these long job control language scripts to get everything to run, it was very important that Steve was right around to give us hints as to what was likely to work and what wasn't likely to work. And we started that process which is just a phase refinement process which averages the electron density around the 3-fold axis and flattens the electron density outside the molecule and iterates this process. And in a short time, although it would take a few days to get the computer programs to process all the data. Still in a short time, the map was improving dramatically. I can't remember how long it took before we had a map that was sufficient that we could see the whole structure. But it was pretty quick, with many false starts.

SS When you say see the whole structure, do you mean the amino acids?

DW Yes, in fact I still have the map here. What you would do, is you would calculate the electron density and then we plotted out on these pieces of acetate and immediately you could see that there was a triangular-shaped object and if you piled the maps up on top of each other, you can actually follow the chain. And these dots that we put on here, they're numbered. They correspond to each carbon alpha position of each amino acid. And it was actually Ian who did most of that and I must say that Ian was a wizard at looking at the electron density maps and seeing the connectivity. I mean in all fairness to the map - it was a very good map and you could just follow it. In fact I think the map was entirely connected that is to say there was just one long strand but there are lots of side chains and it’s a confusing business to get the whole thing connected up. It took some days to get everything connected into place but it was apparent as soon as we saw these maps that it was going to be possible to solve the structure. There was never any doubt in our minds. I guess we just didn't know enough crystallography to know that people sometimes fail at this early stage and the electron density isn't good enough to see the structure. But it was good enough in these maps, as a result of the iterative non-crystallographic real-space averaging.

SS What was the resolution that you had at this time?

DW It was only 3 angstroms - but again the important thing was that because there was non crystallographic symmetry, the map was of such high quality that even at 3 angstroms resolution you could see each individual amino acid. You would be hard-pressed, we couldn't see the carbonyl oxygens for example on the main chain which is a key that helps you keep the peptide planes in order, but we could see the side chains so it was possible to just go through and - well the first thing you do is go through and put dots wherever you see a side chain coming off the long main chain. The main chain, as it were, is a long snake going up and down through this thing and there's a side chain every few angstroms and every place you could see a side chain we would put a dot.

SS Was the sequence known by then?

DW Yes, the sequence was determined just about the year before. Well there were sequences, whether the X31 sequence was actually known at the time we were putting the dots on - it was known by the time we published. It was being done in Fiers lab in Belgium and by Ward and Dopheide in Australia. I think that the Ward and Dopheide sequence may have been the only one that was finished at the time. The first thing we did was to go through and put all these dots and then we began to look at all these dots and try to piece together the sequence and I would say it went reasonably quickly but it's been a long time and maybe it didn't.

It seemed - I would say we were euphoric relative to how we had felt all the years before that - we really were. We had been working on it since 1974 or something and here it was 1980 - 6 years - and now we were finally seeing it, so it was very exciting. One of the things that was immediately exciting was that it was strange looking - it was long and thin. Proteins up until that time were mostly spherical globs of one sort or another and we knew we were looking at something interesting and different and that helped to keep you motivated, but I don't think we needed any help to be motivated. We were extremely enthusiastic to get the whole thing built and see it. And once we had it all dotted in and figured out where all the amino acids were - we were able to read off the co-ordinates by using some fancy computer gadget that was hooked up at the time. So at that point we had a set of coordinates but only of the carbon-alpha positions and not of all the side chains because we hadn't built them in a picture system, yet. You could see them, in the sense that you could see the phenylalanines, but we hadn't actually built the structure and I think when we first published it that was the state it was in. It was not uncommon in those days to be at that stage.

SS Is that the Nature paper?

DW Yes it is.

SS You had also done the antibody paper at that time.

DW Well, that's right. The antibody studies: mostly that was taking information which was available and making sense out of it on the structure, that's right. Certainly when we started to think about writing the paper - I must say in those days we took it rather leisurely. I remember sitting at home one summer writing the paper for 2 or 3 weeks. That was a long time after it had been solved and a long time before it actually got published.

SS But you knew nobody else was doing the structure.

DW Exactly, but also in those days there wasn't the intense pressure to immediately publish things or at least not in that field. With reverse transcriptase, there might have been intense pressure to publish. In crystallography in general there wasn't that pressure to publish immediately and so instead you tried to figure out what the structure meant and studied it. And so we spent a huge amount of time just studying the structure.

SS You wrote in the paper that there were two obvious difficulties - that the length of the HA seemed to restrict membranes from interacting and that the location of the HA2 amino-terminus (the putative fusion peptide) was 100 angstroms from the distal tip. You actually wrote " a major conformational change would be required after receptor binding".

DW That's right. The molecule is very tall and thin and you could immediately deduce where the binding site for the cellular receptor was likely to be because there was a pocket on top of the structure and all the residues in the pocket were conserved in all the sequences so that oriented the molecule top to bottom. Another thing that obviously oriented and that was that the C-terminus was at the bottom and the C-terminus was known to extend on into the membrane. So it was pretty clear that this 135 Angstrom tall molecule stood up on the membrane and had a globular domain at the top facing the cell's membrane. The fusion peptide was only 35 angstroms from the bottom - that is it was down by the cell's membrane and it was tucked in between the subunits in a nicely buried location where it looked like a good place to put such a hydrophobic sequence. On the other hand, it seemed difficult to imagine how that sequence could get out of there and get involved in membrane fusion.

SS So was your first response that maybe this wasn't the fusion peptide?

DW No - definitely no. Our first response was that it was going to come out and then the issue was is it going to go into the virus's membrane which would have been 35 angstroms down toward the bottom or go into the cell's membrane which would have been a 100 angstroms up to the top. And despite the fact that we couldn't imagine how it would go up to the top - I think it was always thought- at least 50-50 that it was going to happen and probably more. It always made so much sense that the fusion peptide would somehow get into the other membrane and that would be the key event. I don't know where that idea came from but it was shared by a lot of people, it seems to me. I doubt if we said anything explicit about that in the paper. We didn't think that explicitly at the time. The only problems that were solved immediately were: here's where the binding site for sialic acid is, here's where the fusion peptide is, but that doesn't tell you how fusion takes place or how you bind sialic acid and then the third thing was where were the antibody binding sites that were involved in antigenic variation.

It was really the sequence work that made that possible because by this time sequences were coming out at a fast rate on all the different strains and you could see that the variation was primarily in HA1 which comprised primarily the globular domain at the top as opposed to HA2 which was involved in fusion. And even in that first paper we were able to use the variation in sequence to show where the binding site was and the sequence for the 1972 virus and the 1975 and a few others were known at that time - it was 1981. So you could see that there had been about 14 amino acid substitutions between the Hong Kong 68 epidemic and the 1972 MRC-11 strain and the 1975 so-called Victoria strain. So the issue was where were these 14 amino acids and you could go and put little dots on the surface of the protein and by this time- I don't recall- I guess it's a little later when John and others were making single amino acid variants - I think that comes later. But at that time, we said something like: here are the five binding sites for antibodies and it looks like at each binding site you have to have at least one substitution, so that the antibodies that you raised against the last infection wouldn't be effective against the next infection. One problem with that idea, is that it is probably almost the entire surface that is actually antigenic - the sites tend to group in 5 places because antibodies can get a good grip on those places, is how I would view it. I think when more substitutions were known they covered almost the whole surface of the hemagglutinin with the exception of the conserved binding pocket and the carbohydrate sites. So the idea that the binding pocket would not be subject to direct antibody assault, so to speak, was around. The basic idea would be that as long as it's smaller than the binding site of the antibody then substitutions around the edge would knock the antibody off. More recently, Marcel Knosow and John showed that was in fact the case by showing the structure of the antibody bound to the influenza hemagglutinin that does bind right over the binding site, but of course it's touching all the rim of the binding site and that what's changing year to year. Peter Coleman had enunciated this idea at some point and I don't know how much we were influenced by that, but I do know that Peter had had this idea about antigenic variation in general.

SS When you first presented this structure what were the responses?

DW People were wildly enthusiastic. You have to ask John - I can't remember. John and I were talking over the telephone all the time and the papers that we wrote - there was a very integrated writing of the papers as there still is now. One of us tends to write the first draft but by the time the paper is ready to go out it is difficult to find a sentence that belongs to one or the other. I would say that probably the wooden ones all belonged to John.

When you are writing a crystallographic paper a lot of what you are trying to do is condense an awful lot of information into the paper because the structure is so rich with information. After all the trimer has more than 12,000 non-hydrogen atoms. So you're looking at this thing and you have all sorts of stories in mind - look at this neat part - look at that neat part. And in the end in a paper in Nature, you get 5 or 6 figures so you have to think pretty carefully about what those figures are going to be and what they are going to show and even more - who can understand them. If you make figures for crystallographers; the crystallographers look at everything in stereo and see everything in 3-dimensions. They pretty quickly get the idea that you have in mind but we wanted to reach virologists as our primary audience and so we wanted the virologists to understand what we were talking about so we tended to have drawings as well as stereo figures next to each other so that someone could look at the details in the stereo figure or they could look at the drawing and get the idea. And back in those days, the idea of having 2 papers back to back in Nature was not outrageous at all. It would happen all the time and if you had something important to say - obviously it wouldn't fit into one Nature paper and you had the license as it were to write it in full. Nowadays, it would be squeezed down to one Nature letter. So the 2 papers were written in full, one describing the structure and the other describing everything we could think of that it explained at the moment and pointing to things that it might explain in the future.

SS Do you remember any interesting comments on that? I didn't see anything in New and Views - there wasn't anything that month.

DW I don't know whether there were News and Views. As is typical in a field, all the people who are in the field immediately write review articles so as to establish their claims or reestablish their claims in the field since here we came as a new force in the field, you might say. So it wasn't too long before all the people who had been making major contributions in the field were writing review articles that had pictures of our structure in it, so it was very gratifying. I don't remember, I mean, I got a lot of invitations, and so did John, to give talks. In fact, prior to that time, I had traveled very little and talked very little but within a couple of years of the structure - coming up to it - people knew we were about to have it and so on.

SS Was that the first membrane protein?

DW I think it would be fair to say - the first membrane glycoprotein, maybe, I don't know, we didn't emphasize that. Certainly the purple membrane work had been done before that in the mid-70's and that was what people would call a real membrane protein since it still had the part that stuck in the membrane. I think of it as the first surface antigen or something like that. I think it was the first molecule that had a lot of carbohydrate on it, maybe that had any. To tell you the truth, I don't know. It was somewhat disappointing that the carbohydrate didn't seem to have any obvious role. It was just decorating the outside of the molecule and mostly disordered. I recall when the issue of whether or not I should get tenure came up that one of the people who had supported me (and I found this out later when this person told me this) had made a statement like, "well, it seems worthwhile to solve a structure like this. How are we ever going to know something about carbohydrate if we don't just go and look". Well, we went and looked and I don't think we really learned a lot. What we learned was mostly about implications for virology. It really is for other people to decide how much impact it had. It had a large impact in my life in the sense that it made me very visible and I went around and gave a lot of talks and showed the structure to a lot of people, both virologists and crystallographers, and also cell biologists were interested in it. After all it was what was on the surface of membranes. I think it seemed interesting to a lot of people.

SS In fact, it probably was one of the first molecules that brought in a lot of people from different areas. Previous to that, it (atomic structure) was mostly enzymes, the muscle proteins.

DW That's right. The point you're making is that it was one of the early proteins that had a kind of cell biological message as opposed to a biochemistry message. Enzymes after all were catalysts and they were catalyzing essentially an organic chemical reaction and so up until that time people had been concentrating mostly on molecules like that. Now one major exception to that would be hemoglobin where Perutz had known from the start that what was interesting about hemoglobin was the way it could respond to different physiological conditions with different affinities for oxygen. So that it was a little machine. I always thought that Perutz was such a hero because not only did he set out to do the nearly impossible - solve the structure of a protein, but he had the courage to pick a difficult one right at the beginning. Four times larger than the little ones that other people were working on at the time, because it had a more interesting biological lesson to be learned in the long run and that made it possible for him to continue working on that molecule for his entire career. Because it had so many interesting aspects to its physiology - the allosteric aspects of how it worked. In fact, I still like to teach it as an example of how complicated a protein can be and how subtle the structural effects can be that give you different physiological effects. In a way, the hemagglutinin does have that characteristic because it has different activities and I must say we knew that when we started.

Actually, I might take an aside quickly and say something about motivation to be a scientist. I mentioned to you earlier that I found it very difficult to work on the protein-DNA complexes that I started to work on when I first became an Assistant Professor. Frankly, I was not that interested in them despite being in an environment where everybody else was interested in that sort of thing. Part of it could be that it was obvious that everyone else knew a lot more about it than I did and were always going to know more about it and it wasn't clear to me how I would ever add anything. They were learning how protein binds to DNA and were learning all the details. I thought I would always just be illustrating things that they were doing. Whereas with flu - the antigenic variation - if one thought carefully enough about it you would probably come up with the answer but it wasn't my bent to think so carefully, I guess, and no one else was thinking that carefully except maybe Fazekas de St Groth and so pictures were needed. We needed something to look at to figure out what was going on. We needed it for antigenic variation. There was also no known example of a protein binding to a receptor and in the end that turns out to be hemagglutinin binding to sialic acid. Not so exciting. Not like human growth hormone receptor binding to human growth hormone, but at the time I couldn't make distinctions like that. Then membrane fusion struck me as completely outrageous that a protein could do it, and that Sendai virus could cause cells to all fuse together. The idea that you would have a protein structure that had the capacity to do something like that clearly meant you were into a rich vein - so to speak- and could keep mining it for some time.

SS I was surprised actually when you said last time how much you had thought about membrane fusion because I remember in the late 70's that you were thinking about that even before you said you knew that flu could do it.

DW That's right. It was really Judy White, having Judy White as a graduate student in the lab working on Sendai that is in some sense proof that I'm not just dreaming this. We had really decided that fusion and viral entry was what interested us most deeply about the molecule. What I really mean when I say most deeply interesting is that it's not just pushing electrons around. I would leave that to people who knew more about organic chemistry. It appealed to me to work on a protein that seemed to do something that was macroscopic: it took two membranes-brought them together and caused them to fuse. I didn't have any picture at that time as to how it was happening, but it just appealed to me that it was doing something that was like cytology. I still feel that way. It helps motivate you. It helped motivate me to think that I can't see the form of the answer. In the case of an enzyme - the form of the answer, to me, was I'll draw some arrows and show how the electrons move. Now for somebody deeply interested in that thing, they actually care whether the electrons move this way or that way. I couldn't find it in myself to care so much about it but I could see the form of the answer and I could see that I would not be excited by being able to explain to other people- ah, the electrons went this way, they didn't go that way. Whereas, I couldn't even see the form of the answer in hemagglutinin. That was particularly appealing to me. It seemed that I might otherwise leave science if I couldn't find something that was mysterious. And proteins binding to DNA at that time - even that didn't seem mysterious in the context of the people at Harvard - Wally Gilbert and Mark Ptashne. They were learning so much about how they bound but there didn't seem to be much mystery left. It seemed that you could plot out what was likely to be found. Of course new things were found and they are much more interesting than you thought. But hemagglutinin had an element to it that was impossible to just say - oh well it's all going to be simple.

SS How do you feel now? Because now there's an awful lot that is known before you actually know the structure. I agree with everything that you are saying and I wonder about that for somebody who is starting now.

DW We should not despair - having learned this much about the structure and still not understanding how membranes fuse.

SS So you didn't have to worry about your lab doing all those changes (changes by genetic engineering) because I think everybody in the field must have jumped in to try to use that structure.

DW Yes

SS Did you get involved with other people at that time to do some collaborative work?

DW Mostly it's just work between John and me. I should be careful because there have been some other things, for example to figure out how sialic acid bound into the active site. The binding of the different sialic acids was shown by Jim Paulson and others. There is an interesting aspect of that biologically, although it's still a bit of a mystery to me. Although Dr Ha's work going on in our lab (with John) now may be revealing how this might work.

SS I was thinking about that in particular.

DW The viruses that bind to human cells and infect humans seem to bind to sialic acids that are linked to the next sugar by a 2,6 linkage and in animals and birds they bind slightly more tightly to a 2,3 linkage. Getting carbohydrates to be able to study this was difficult. Bill Weis first did it in the lab and showed how sialic acid bound and that was a good start but in the end what was important was to see the difference between 2,3 and 2,6 where the oligosaccharide was long enough so that the difference was meaningful. And that was only done a couple of years ago by Mike Eisen in the lab. You can really see that with the 2,6 and 2,3; in one case it causes the molecule to bend back almost on itself, and in the other case the molecule goes out more or less in a straight line. What the overall significance of that is, remains to be seen. We would like to do the structure of the protein that preceded the 1968 Hong Kong flu, the duck-Ukraine 63. Its sequence was done long after it was originally discovered and found to be only 17 amino acids different from the 1968 Hong Kong flu. So in other words, its almost as close to the 1968 as the 1972 and 1975 were and yet it comes from a duck in the Ukraine. So it is clearly the hemagglutinin that jumped into the Asian strains to give us the Hong Kong flu or it's clearly an ancestor of that hemagglutinin. And it would be nice to see whether there is any mystery to be unlocked by seeing its site that binds sialic acid since it's one of these sites that prefers 2,3 over 2,6. (Just now being completed in the lab by Dr. Ha.)

SS It usually is the HA that changes but I wonder if in some cases now if one went back and looked at the other proteins whether they could have been changed to cause the epidemics.

DW Right, but in the 1968 case, the biological, as John would like to say, the biological experiment was done. The Asian flus that were around from 1957 to 1967; all the genes in that Asian flu are the same in the Hong Kong flu except one - the hemagglutinin. The neuraminidase is the same as the Asian. So if you could be protected by antibodies against the neuraminidase you should have been protected in 1968. Now that isn't to say that you won't be protected by some future neuraminidase antibodies but the ones in 1968 obviously couldn't do the job.

SS So the hemagglutinin has to change, so that the question is whether other things change so it(the virus) becomes not only not resistant(to antibodies) but it's more pathogenic.

DW Right, questions about pathogenicity are really difficult. In the case of the Asian flu and the Hong Kong flu, I don't know whether the 1968 Hong Kong flu was really more pathogenic or just infected more people as the result of lack of immunity.

I wanted to mention quickly that it was a prodigious amount of work to actually build the whole hemagglutinin structure using a picture system program that was available at that time. It was called Builder and had been made at that time by Bob Diamond at the MRC. And Ian did all of that and it took months and months of patient building to build up the whole structure. So there are really some thankless tasks in this business. All that has become a lot easier but back in those days to build it on one of these picture systems was a long, slow, painful process. And so after the structure was completed and published it was still a long process to be able to get a good molecular model built.

I think the next major event in my mind was when Bill Weis showed how sialic acid bound into the binding site. Bill Weis was a graduate student. I'm remiss at not keeping track of who was my first student and who was my second student. I know Judy White was the first whether Bill was second or whether Pam Bjorkman was second I'm not sure but Bill was one of my earliest graduate students and he was casting around doing various things and finally settled on getting sialic acid into the binding site which was difficult in those days because you couldn't get sialosides. You could get sialic acid but sialic acid is in the wrong anomer to fit in the binding site. You have to have sialic acid attached to another sugar and those were hard to get. But he was able to get a trisaccharide. I think he was finally able to buy it but it was a bit of a struggle and anyhow the first things we had weren't working well and it was the usual uphill struggle. But he did show how sialic acid was bound and that was nice because it led us to start thinking about drug design. Could you get something into the binding site that would bind much more tightly making use of what you could see about how it was bound? In particular, the idea was to stay in the binding site because if you go outside the binding site, obviously antigenic variation is happening "out there".

Selection by drug would happen as well and you would get resistant mutants. And that led to a whole series of experiments being done in the lab by students and post docs, notable among them, Nick Sauter who developed the first assay using NMR so that we could measure a binding constant. It was only 2 millimolar - a pathetically weak dissociation and in fact, in this 2,3 - 2,6 business that I mentioned earlier with Paulson, we found that the difference between 2,3 and 2,6 was on the order of a millimole: nothing. Essentially very small numbers so it would be very difficult to study and make any sense out of, because you are never going to look at a crystal structure - what I mean is you can't now explain the difference between something that binds by only a factor of 10 different from something else, certainly not by a factor of 2 or 3. Still this got us started synthesizing small molecules. Various people in the lab at various times synthesized lots of small molecules and also in Jeremy Knowles' lab and George Whiteside's lab.

SS So you had an organic chemistry lab.

DW We had a little bit of an organic chemistry lab here. There was a postdoc, John Hanson, who also worked with Nick Sauter. We synthesized lots of molecules, there are still lists of them hanging around here but we never really got things that bound substantially better unless we went outside the binding site and then people in Jeremy Knowles lab were getting things that were even nanomolar. But they had huge non-polar objects sticking outside the binding site and to my way of thinking they are not serious.

That side of the subject has become less interesting to me because I guess I don't see that my own abilities have much to do with whether or not it would succeed. And so the binding site is there and people try things. In the end it was very frustrating to have ideas that you wanted to try and not to be able to get them tried very easily - to take literally years to be able to synthesize a few compounds and then have them bind marginally. It's clear that's a subject which is still very difficult. The success that Peter Coleman has had with neuraminidase indicates that there is no real necessity to have something that would block the hemagglutinin binding site. On the other hand, it remains to be proved whether blocking the binding sites for viruses as they bind to cells would be a viable way of preventing viral infection. I had always hoped that we would do that with hemagglutinin and we haven't succeeded.

So there became 2 sides to the work, you might say. One was focusing on the binding site and trying to think about sialic acid which as I say has not gone very far toward the direction of producing a tight binding of a small molecule inhibitor. And the other was to follow up the membrane fusion activity and John's lab in particular was very active in characterizing the molecule at low pH, or let's say the low pH induced molecule. Because you only have to lower the pH - John showed in his first paper in 1982 that the structure was irreversible. When you lower the pH and raise the pH to 7, the molecule didn't go back. There were various possible reasons for apparent irreversibility. It could have been it was all stuck to things and couldn't come unstuck and so on. But slowly, during the 80's a number of papers were published, mostly by graduate students and postdocs in John's lab and I was involved with them - I would say quite a bit on some of them and less on others. The idea was to characterize what the new molecule looked like and in actually the very first experiment that John did, he discovered a very important thing, and that was you could cleave the fusion peptide with thermolysin once it had become exposed. That opened up the possibility of doing a crystal structure by lowering the pH, raising it again, treating with thermolysin to now resolubilize the molecule.

So the result that John had was that thermolysin simultaneously removed the fusion peptide and resolubilized the low pH form. So, that immediately led to the idea that the hydrophobic tail had been exposed and that it probably went into the cell's membrane. And it also made possible, in theory, doing a crystal structure. Although it took a very long time to produce enough of the protein - many, many years and it was an off-again on-again kind of thing - to try to get the whole thing to work clearly enough to get enough material to do the crystal structure. Finally John began to supply us protein and Per Bullogh, a post doctoral fellow, started to work on it.

He actually got crystals almost immediately which often happens but they were unsuitable. They didn't diffract well and so on. Eventually he got good crystals and he was joined in the work by Fred Hughson a post doctoral fellow. Hughson had come here to do other things but had always been interested in the membrane fusion problem.

SS As I remember you ended up doing something that was quite small, just a piece of it.

DW What we ended up doing is called TBHA2, what does that mean? That means you lower the pH of the hemagglutinin, raise it to 7, treat it with thermolysin or trypsin and two things happen. HA1 dissociates. It actually doesn't completely dissociate, there are 27 amino acids, the first 27 of HA1 are disulfide bonded through amino acid 14 so that stays on. It's part of a five-stranded beta sheet which exists on this structure as well as on the low pH form.

SS Did anybody raise the problem of looking at such a small molecule?

DW Yes, definitely, I'm sure everybody raises it but we don't hear it very much. I think that what we were doing here is dissection. There are two ways of looking at it. One way is to say we are doing it because it can be done. And the other way is that we were very cleverly dissecting away the parts that are causing trouble. In this case, when you lower the pH, the top globular domains appear to dissociate in some way and are apparently flexibly linked to the rest of the molecule. Well, when something is flexibly linked it doesn't crystallize so here we are faced with something that doesn't have a structure because it is flexible and if you look in the electron microscope after you've lowered the pH, it’s all a mess. The flexible part can be in all different places. As soon as you cleave them with trypsin which cleaves only a single bond and yet causes the HA1 from amino acid 27 on to dissociate. What dissociates? Marcel Knosow in his lab in Paris eventually did the structure and it looks just like what's on the original pH 7 structure but it's monomeric.

SS So what dissociates can be crystallized.

DW Yes, the structure's been done; its monomeric. It has the binding site for sialic acid, it binds antibodies - all these things John showed in his original paper but the structure sort of confirms it for the non-believers. So the globular part of the top of the structure, we actually know what it looks like at low pH: it looks the same. The only parts we don't know are about 30 or 40 amino acids, a little piece of HA1 that goes up toward the globular head and the little piece that comes back down. We could see everything from that part of it. Of course to say all this to people every time and convince them and eventually you start using shorthand because you know what the facts are. If they care enough they will have to find out. But we came to the conclusion that leaving HA1 off was a perfectly legitimate thing to do because it had dissociated and become monomeric. Now the next issue would be where did you get the courage to chop off the fusion peptide? Well, where did Skehel originally get the courage to chop off the transmembrane domain? Born of necessity, if we didn't chop off the fusion peptide, the molecule aggregates into rosettes and wouldn't crystallize. So we chopped it off. Actually, the molecule is very chopped up and I have often thought that if anyone was going to get into trouble on the issue of - does protein crystallography really tell you what the real structure is in biology - then we might be just those people because we're always pushing the envelope so to speak - trying to go further. Now, we're not doing this blindly. We're doing it because it's the only way we have been able to find to move forward. If we didn't take off the top of the structure we wouldn't know what it looks like today. We’re trying to crystallize it with the top part still on but we did some tricks which we've published. And Jue Chen, a graduate student, did a trick where we left the fusion peptide on and then put some very hydrophilic peptides N-terminal to that to help keep it in solution and that molecule is actually soluble. Despite her best efforts it never crystallized. It will some day. Ultimately, we need to know how these molecules look. I don't know whether we need to know it, but crystallographers would imagine that knowing how the molecules fit into the membrane will help understand membrane fusion. All the other people working on membrane fusion, or let's just say scientists working on membrane fusion wouldn't necessarily feel that knowing that answer is important. They would use more indirect experiments. I would call the crystallographic experiment of doing the low pH form a more indirect experiment. It's an attempt to learn something about the process without actually having all the facts at hand.

SS Given the answer, you feel you knew what you were doing.

DW Yes, it is quite a dramatic looking structure. Just as we were about to finish the structure, Carr and Kim had a dramatic publication (Cell 73: 823-832, 1993) where they showed that a small helix and a part of a loop could form a triple-stranded coiled-coil and therefore they predicted part of what we eventually found - that the fusion peptide could end up on the far end of a long rod that was a triple stranded coiled coil growing upward out of the original coiled-coil.

Carr and Kim found that the loop region had the capacity to be a coiled-coil using a computer program. It has heptad repeats and in fact, interestingly, Ward and Dopheide had actually pointed that out in their paper in 1980 when they published the sequence of the structure and they had said it would be a coiled-coil of the following dimensions and so on. And when we published the structure in 1981 we said they were wrong because of course we saw the thing bent over. So they were right obviously.

There is another amusing anecdote on that subject. One time Ian Wilson - much later, sometime before 1994, in the early 90's - went to a meeting and he was accosted by someone who said "you know that structure you published is wrong" meaning the 1981 hemagglutinin. "I know that there is a triple-stranded coiled-coil in the loop region and you guys call that a loop. It's not true. I study coiled-coils" and so on. Ian wasn't shaken up at all. Actually, I might have been. Ian said "Well you don't understand. There is another conformation and maybe it's part of that". And Ian called me to tell me this and interestingly enough thereby anticipated the structure in some sense. And I remember he and I discussing it and his saying it was really troublesome because this person hounded him as though we were covering up having published the wrong structure. There was going to be an exposé some day that was going to show that the structure we had published was wrong because studies of the sequence show it could not possibly be that.

SS So the sequence studies said that it should be a coiled-coil?

DW Yes, he had done the same thing as Kim and Carr. He had seen that the sequence would make a nice piece of helix. We didn't pay enough attention to the fact that there was embedded in the structure a piece of sequence that would rather be a helix and that's what had really been pointed out by Ward and Dopheide.

SS Could you have known that "if you'd paid enough attention" in the 80's because there weren't that many structures known at that time?

DW Certainly the idea that that stretch had the potential to be a coiled-coil was clear from Ward and Dopheide. On the other hand - potential. I hadn't yet appreciated the concept that a structure could have the potential within it to be another structure and in fact I think in the late 80's when people in John's lab were doing the experiments on the low pH form of the molecule, they were clearly doing experiments that were leading them to believe that the molecule might shoot up into a rod or something. The experiment was: They were looking in the electron microscope and were seeing after low pH that the rod was skinnier, after you took away the top globular domains that there was a skinny rod. (These experiments done by Ruigrok et al. are also discussed in later section.)

The problem was it wasn't longer and so if it wasn't longer, how could this thing be shooting up. And I would always say to them - and for this I deserve what I will get - I would say: "No that's completely impossible, protein's don't unfold, this is a solid protein. Look at all this beta sheet, look at this helix holding on here and so on. This will never happen". Whereas in fact what has to happen is that 2 strands of beta sheet have to unfold from a 5 stranded beta sheet, a helix has to detach, a loop has to - and so on and so forth. Dramatic things had happened in the serpins. Huber showed that a protein structure changed so dramatically that a beta strand got put down in the middle of a beta sheet in the serpin structure. So, I became less resistant to the idea that something very dramatic was happening.

If we back up a bit, in 1987, or '85 I guess or '87, there were a couple of papers by Rod Daniels, a post doc in Skehel's lab. I had a lot of involvement in them because they were the selection of mutants which raised the pH at which the hemagglutinin fused. Those mutants gave you a clue as to what the conformational change would be because they were at interfaces which were going to move in the later structure. And one thing they clearly showed was that it was happening all up and down the whole trimer. The top globular domains - there was the hint in that and also from some other antibody experiments, that the top globular domains would dissociate but you could see all sorts of other things happening. At first we thought it meant that the trimer was going to come apart - the trimer at low pH of the HA2, but chemical crosslinking studies here - actually Bill Weis did those - showed that that wasn't the case. It was staying together, trimeric. But it indicated that it was going to be very dramatic. So from that time on, we didn't think it was just going to be an oxy-deoxy hemoglobin sort of thing, we expected something dramatic. I don't think anybody expected it would be as dramatic as it was.

To get back to what the structure looked like when Per Bullough and Fred Hughson finally finished it from material that had been supplied by John. It not only had this shooting-up part, but it also has a bend back. That is the bottom of the long helix of the original molecule actually bends back, reversing the direction of the polypeptide chain by 180 degrees and sending what had been the C-terminus at the bottom of the structure up toward the N-terminus at the top of the structure. That in fact is an equally dramatic idea: what had once been a simple helix 80 angstroms long, broke approximately in the middle, unfolded for 4 amino acids allowing the production of a turn or the folding of a turn and the helix which had been one piece now ends up as 2 helical pieces. What had been one long piece going in one direction is now 2 pieces with one of the pieces antiparallel to the first one, heading back up in the direction of the N-terminus. The long and the short of it all is that the N-terminus and the C-terminus seem to be very close together in the low pH form of the molecule. And that's what you might imagine would be necessary for membrane fusion: the N-terminus and C-terminus are eventually going to end up in the same membrane and in some intermediate or transition state when membranes are fusing, they are going to be brought together. This was the first hint and we didn't fully appreciate it because two of the strands - out of the 3 in the trimer - were headed back up. We could see them to amino acid 162. We had to amino acid 175 in the crystal so we were missing 10, but one of them didn't go as far back, it stopped at 152. So it's possible it was an artifact. We talked about that and there were some in the lab that favored the idea that it was an artifact that it was going back. Certainly John and I never favored that and I don't think any of the major authors favored it. We thought we were seeing the beginning of something and if we had longer pieces we would eventually see more - which has been borne out.

In that paper (Nature 371: 37-43, 1994) we concluded that if one had thought about it enough - the molecule would have to have these features because after membrane fusion, the ends would have to be together because there is only one membrane at the end of the reaction, so there was some formal way in which the molecule looked like something that could do membrane fusion.

SS At what point did you realize that you had what you had?

DW We knew early on that the long rod was in the middle and the outside was coming back up and the problem was to figure out how far the outside was coming back up and things like that. Maybe a general point one could make here is: you have to stare at X-ray structures for a long time to get the message. We used to sit around looking at X-ray structures for a very long time. As soon as they start being generated, that's when I start getting deeply interested. Because I know that it's going to take a long time to have looked at it enough before you are going to be happy writing about it. So you have to start looking early because there will be pressures to get it written by the time everything is done. And we spend a lot of time staring at the picture system; looking at the molecule from different angles, thinking about it, looking at sequences, looking at all the sequences that we could find to see what is likely to be conserved, trying to figure out how it could have happened, which parts were originally holding onto what and what do they then end up holding onto and all kinds of things like this. Trying to figure out what is the message as it were.

If you are going to write a paper: what's the point of the paper? It shouldn't just be here's a picture of the molecule. It should be: We were clever enough to think that making this structure would teach us something and here is what it taught us. And the question is what it teaches you isn't immediately apparent. I would claim, in fact, you struggle for a long time to figure out what it is, Then once you figured out what you think it is, then the question is: Can you illustrate that in such a way that someone else can understand it? And even more, can you at least be sure that the facts get into the paper because people may not get it for the first few years, but if they are still there when you look back then they'll believe that you saw it. For example, a third thing that happens in that molecule, which I think is under appreciated up to now, is that a whole section of it gets extruded; sections of it which had been packed nicely. The C-terminal sections which had been packed nicely into the high pH form of the molecule end up extruded no longer bound to anything and like toothpaste sort of squirting out. They are the parts that make it up to the other end so the structure almost undergoes a structure - no-structure transition. So that in the molecule there are three interesting things that are observed:

there is the popping up;
there is the bending back so the two ends head toward each other:
there's this idea that the part that got extruded into long flexible segments may yet be determined to have something to do with how the membranes are actually brought together.

And the folding back part is as dramatic as the popping up part, it turns out. The popping up part causes changes of about 100 angstroms in the location of residues and so does the folding back part. In fact, in the long run, the folding back part is going to end up - when we finally know - we have another structure that isn't published yet - but there are parts of it now that go back more like about another 150-160 angstroms.

SS Are these changes more dramatic than in the serpins?

DW Well, "dramatic"? Serpins undergo a very weird change where they have a beta sheet and a piece that wasn't in the sheet comes and inserts into the middle of the sheet. So, that's very dramatic because the sheets have to slide apart in the middle. I guess different people like different types of drama.

Is there anything more to say about the low pH form?

SS Well, I wanted to ask you about the form produced in E. coli.

DW Once we had seen the structure, Jue Chen, a graduate student in the lab made the same sequence that we had made by this dissection of the viral protein. She made that piece containing the same set of amino acids in E. coli. Jue Chen obviously thought that you could make this in E. coli because it was stable and it was the most stable form of the molecule. Ruigrok had shown in '87 or 89 in Skehel's lab that if you heat the hemagglutinin, it melts. The high pH form, the regular form on the surface of the virus melts at a lower temperature than the low pH-induced form. The low pH induced form is more stable. Now there are caveats with experiments like these because they are not reversible and I won't go into all those details. I hope we said them in the paper. We were convinced that the new form of the molecule was the stable form of the molecule and, therefore, if you made it in E. coli, it would form spontaneously.

SS So that was not going to be a surprise?

DW No, it definitely wasn't a surprise. Whether it was a surprise to Jue Chen, I would say no. I don't think we do things where we get surprised very often. It's hard to get somebody to do an experiment. It's hard to slog through all the work of doing the experiment. There has to be a pretty strong idea of what the outcome is going to be. And if the outcome that we were looking for had been that you'll never get anything, we wouldn't have started the experiment. Clearly, there was the strong expectation that it would fold and that it would lead to the possibility of getting better crystals, which in fact it did. John (Skehel) had showed earlier that if you heated the molecule, it also underwent the low pH conformational change. There really is some sort of barrier to the conformational change and that any way of jumping over the barrier - by heating it up to get over it or lowering the barrier by lowering the pH.

The E. coli experiment was just sort of a demonstration in the Anfinsen sense that the sequence prefers to be in the low pH induced conformation. But that was not a surprise.

The X-ray structure of the low pH induced conformation was a surprise. That predated Jue Chen's experiments, of course. We had certainly come to believe that the conformational change was dramatic as a result of the low pH experiments; and it was going to be something very strange. We knew in the electron microscope that it was a long thin rod.

SS Where was the EM work done?

DW In Skehel's lab and they are published (Ruigrok, RWH, Wrigley, NG, Calder, LJ, Cusack, S, Wharton, SA, Brown, EB and Skehel, JJ EMBO J 5: 41-49,1986). How could it be a long thin rod and have the fusion peptide exposed at one end of it? That was clear also because rosettes were formed with the fusion peptide on one end. I would say I'm somewhat at fault in the sense that I don't try to think about the answer of what the structure is going to be. I like to figure out what structures are and that could be a fault in the sense that maybe you could figure these things out by thinking and do other experiments rather than determine structures. But we were already doing the experiments (the structure) that were going to find the answer and we just had to get it finished. The fact that someone else could do another experiment that hinted at the answer- who knows - until our structure was done - you couldn't really tell if their answer was right or not. In some sense we tend not to be diverted by experiments that don't give you the answer or that are dependent on someone else's experiments. In some sense, John and I have a long history on the molecule. We were in it for the long run, we wouldn't like to have it said about one of our experiments: "well, why don't you just wait 6 months, somebody else's experiment was going to nail that down anyway". We're doing experiments that slowly but surely are grinding away toward the answer, I hope.

SS When you said you were in for the long haul - what do you think about the future. What are the questions? You did tell me one actually: what is the structure in the membrane?

DW Obviously to see the protein in the membrane and maybe even on the surface of the virus. The Influenza C work that we did: What really motivated us is that the virus doesn't have a neuraminidase so it makes a nice open lattice on the surface of the virus where you can clearly see, hexagonally arrayed, the protein, the hemagglutinin, that does fusion. So you could imagine lowering the pH and possibly following the whole thing because the structure is very distinct. I would love to do things like that using cryo-electron microscopy. Maybe it all becomes disordered and maybe you aren't going to see anything, but maybe you just have to think of the right conditions. I guess I would say the goal really is to try to approach more closely the real phenomena. We have sort of taken a biochemical reductionist approach and at some stage you keep getting closer and closer to the real phenomena. And in the case of hemagglutinin, you try to do that by having more and more parts available at the time and so on. There are a lot of experiments we would like to do but I don't want to read about them before we've done them. So I won't tell you. I think mostly because they're idle speculations. You can dream all sorts of things but one experiment changes your whole view and then you start narrowing down and going to what you can do as opposed to what you dream you might be able to do. And so, the experiments that are underway now, what they are doing mostly for us is they are narrowing down what we think we will do as opposed to all the grand schemes that we always carry around - most of them don't work.

SS Did you want to say something about HIV?

DW HIV is just another hemagglutinin from our viewpoint. Steve Harrison and I got interested in HIV as soon as it was shown that it was the cause of AIDS and I remember we actually decided - it's kind of mildly dramatic - we decided that we were going to go out and have lunch together some day and we were going to just talk about that and what we were going to do about it. Because we have so many other things we always have to talk about trying to run a lab, we never talk just about a simple scientific thing. So we actually went, we may have even gone to the faculty club, and decided we've got to get away and we've got to go and sit down and start talking about it.

We decided to do what we can, working on the HIV envelope glycoprotein is a practical extension of working on influenza. It may be harder in retrospect, certainly because it has so much more carbohydrate. But we would expect to relearn some lessons and therefore it's not as dramatic as starting out on an entirely new subject. On the other hand, doing something in parallel with what you are already doing had the advantage that sometimes the parallel system shows you something more clearly than the system you are working on. On the other hand, doing parallel work is also - it's dangerous because it's the easiest thing to do. What’s easier than picking up another membrane glycoprotein from the surface of a virus for our lab and starting to work on it. It doesn't take a lot of forethought. On the other hand, in the case of HIV, the so-called - I call it the low pH form of gp41 but the form of gp41 that we're able to find really takes off from the experiment of Jue Chen expressing the hemagglutinin in E. coli All the forms that have been done of all the different retro- and other viruses, filovirus, that are being done now are all basically that E. coli experiment. In no case have people taken it from the virus. In no case are they seeing it in anything except this radically dissected form and actually made in E. coli

SS So presumably you are getting the form that is post fusion.

DW Yes, it is certainly post the conformational change that is required to cause fusion. That could still be prefusion if this is what the thing looks like when it does fusion. And the good thing that the HIV structure showed was that the part of the chain which was coming back - the antiparallel part on the outside of the structure came back so far that the remaining amino acids which we couldn't see, which are flexible, could not possibly reach back to the other end of the molecule where they started out. This forced us to realize that in the case of the hemagglutinin the message of these structures is that the N and C terminus are heading for the same end of the rod. So in this case of what was really just a bit of "me too" - doing HIV gp41 after the Flu TBHA2 structure - it did actually show something new. I think that's the advantage of more than one system, although I wouldn't want to push that too far.

I guess another thing about HIV is the conformational change isn't triggered by protons. It is triggered by conformational changes that occur after the binding to receptor proteins. So that's interesting, inherently. There will be some changes and whether the conformational change is like the HA change or something completely different will require seeing the conformation before the receptor induced conformational change. I think it will be extremely interesting to see the relationship of the refolding events.

Pt.1 of Oral History

| Introduction | 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 |