Steve Harrison Oral History

Recorded and edited by
Sondra Schlesinger

July 16, 1999

SS O.K. Steve, now there is a structure of tomato bushy stunt virus. What happens next?

SCH Chronologically what happens next is that we chose to look at a close relative that I had been planning to do for some time, turnip crinkle virus, for two reasons. One was a crystallographic reason that at the time I think I took as a simple assumption, in retrospect it was not so obvious, namely that one could use one structure, that of TBSV to get a phase start for determining another structure. Later on, that was done much more thoroughly and systematically both by Michael Rossmann for determining the structure of mengovirus from rhinovirus and then by David Stuart who created a hybrid molecule he called "rengo" to solve the structure of foot and mouth disease virus. We weren't able to do the computations nearly as thoroughly at the time, just because of the limitations of computing power, but nonetheless Jim Hogle did a spectacular job of demonstrating that you could do that.

Jim had done his graduate work with Sundaralingam on some lysozyme related problem at Wisconsin, and he came to my lab with the explicit mission of learning enough about how to solve virus structures so that he could do polio. He had become interested in polio, I can't remember why. Ah! - Roland Rueckert had been on his advisory committee and had gotten him interested in polio and actually, ironically enough, Sundaralingam, his graduate adviser, was a polio victim and had a severe limp because he had had polio. He grew up in Ceylon (Sri Lanka). Jim basically came with the idea that by doing turnip crinkle virus (TCV) he would learn a little bit of the game and then figure out how to do polio virus. He spent the last year or so of his time in the lab starting polio, figuring out how to get good diffraction data. A crucial issue was how to cool the crystals.

SS You had actually said earlier that you had begun to think about polio virus and had obtained some from David Baltimore.

SCH Yes, I had actually set up crystallizations in David's lab. I was supposed to do it there. The crystals never grew in those batches but Jim pretty quickly repeated the early crystallization with Mahoney type 1. But there was another reason to study TCV and that was that Ruben Leberman, working with Aaron Klug when I was a student there, had shown that you could disassemble TCV in a reversible way and put something back together. It wasn't quite clear what was put together, but it seemed like a virus-like particle. You could also under some circumstances get what were called "small particles", little T=1 particles, and so it looked like one could study regulation of assembly.

SCH Yes, the TMV dissociation-reassociaton was before I was a graduate student. That was early '60's, late '50's or maybe mid-'50's. I don't remember, but Ruben was tackling TCV as a spherical virus for studying assembly, being obviously a different sort of problem, in terms of what you thought about conformational shifts, than a helical virus. So TCV was supposed to be how we could study assembly. Now, in practice, that actually worked out quite well, although I don't think the structure of TCV helped us very much in the assembly studies. The structure of bushy stunt virus helped us totally. Knowing that the structure of TCV was the same as bushy stunt helped us, but the details of the structure of TCV, what we really learned from the crystallography, didn't help us very much, and that's because site-directed mutagenesis and related methods were still 5 years off. The notion that we ought to make TCV protein in bacteria and study assembly using recombinant protein, so that we could do things that would be second nature now - for example, how we thought about polyoma assembly (or rather how Bob Garcea thought about polyoma assembly) was in 1979-'82 out of the question. It wasn't in one's vocabulary. What we could do with bushy stunt and turnip crinkle virus was limited to what we could do by tickling it with proteases or doing other kinds of more traditional protein chemical things. Nonetheless, by doing some very elegant biochemistry and electron microscopy, Peter Sorger, who was then an undergraduate student, managed to show a number of important features about TCV assembly that, I think, remain central issues in assembly of shells. Basically, he showed that the assembly mechanism had the following characteristics:

First, there was a defined assembly unit, which was a dimer - that is the whole assembly proceeded, not through larger intermediates that then came together, but, by getting started correctly and adding a single small assembly unit to grow a complete shell; the assembly unit was the dimer we had previously characterized.

The second thing he showed was there was a defined initiation event. He characterized that as 3 dimers associating with the RNA, and we still think that's probably the right number, but was certainly limited by the accuracy with which one could do the measurements at the time. I wouldn't be surprised if our model for the initiation structure turns out not to be completely accurate. But the notion that you get started with a defined and extremely stable complex of protein and RNA is certainly true. Indeed, what he showed was that when you take TCV apart there remain very tightly associated with the RNA three dimers of the protein subunit, which can serve as an initiation complex for reassembly. The bound dimers remain stable to dissociation if you go around a second disassembly/assembly cycle. So there seems to be one particularly stable association of protein and RNA. That was consistent with the notion that I think I talked to you about at the end of the last session, namely, that the packing of RNA made us understand that all that matters is one particular initiation site and that all the rest of the RNA would pack in, in a way that simply depended on the particular pattern of stems and loops that were present and probably a variety of random events during assembly.

SCH O.K. (laughs) The third important characteristic of TCV assembly was a more special one. Peter Sorger carried out an extremely clever experiment, which he did while I was on holiday with my parents in Woods Hole, so he did it all on his own. I remember hearing about it on the phone in the summer, that was between his junior and senior years. The experiment was to look in the electron microscope at various time points. Basically what Peter did was quickly to dilute the protein and RNA of disassembled virus into assembly conditions and then to make grids as fast as he could by negative staining. He then looked at the time points in the electron microscope; it was a kind of stop-flow EM experiment. Everything was over in about 5 minutes. You didn't need stop-flow, and there was reasonable reassembly in 5 minutes under the conditions he had chosen. What was very interesting was that at intermediate time points you never saw incorrect structures. You saw partial shells always with the correct curvature, that is as the dimers added, they knew whether to switch into the flat CC conformation we had seen in the structure. Probably I said this in the last interview, but if not I'll remind you that the structure shows there are two conformations of the dimer, flat and curved, if you wish, CC and AB. They differ by ordered arms on CC and disordered on AB, and basically you could conclude straight away from those simple electron micrographs that as the dimer added, it knew what to do. Either it ordered its arms up and became flat or kept its arms disordered and imposed curvature, since as the shell grew it was growing correctly. And when you got double nucleation on an RNA, you saw two partial shells each of the right curvature. I called these objects telephones, because they looked like telephone receivers with correctly curved things at either end and a bit of mess in between.

These observations also reminded us why specific nucleation is so important, even if in the infected cell the major RNA that would be packaged is the right RNA. There probably isn't a huge pressure to package viral RNA specifically, there would be little penalty to accidentally occasionally package some cellular message. It is very important to have a unique and particularly stable nucleation event, so you nucleate only once on any piece of RNA. Otherwise you get these dead-end structures that would have two inconsistent nucleations and would start to grow toward each other, incompatible growing shells. I think these principles are likely to be true of many icosahedral assemblies, for that matter non icosahedral shell assemblies, for example the assembly of a clathrin coat. Because we couldn't follow it up in quite the way we would now, and because one doesn't tend to think about working on plant viruses anymore, those points have never been reestablished in a modern context, where you could go and do the mutational experiments that in a certain sense would be required for modern rigor.

SCH Ah-good. That's important. They're working on flock-house virus. We did show that if you cleave the arms off, you got small particles, that the arms were essential for establishing a larger curvature.

SS Speaking of the arms, as I remember when you cleave the arms off no RNA was incorporated.

SCH That's right. Again, the other thing we never actually worked out properly was what recognized the RNA. Remember, you and I talked about whether we should do experiments on RNA viruses and the capsid protein. Since then several things happened. Jack Morris' lab finally did properly identify the sequences that were involved in nucleation.

SCH No, he hasn't determined the protein side of that reaction. I'm confident that they will turn out to be elements of the R domain. Now we understand more about RNA recognition - that RNA makes structures which can then specifically interact with initially unstructured proteins – a notion that was not in our heads at the time. We assumed the protein would have to provide the specific structure for a recognition event. Oddly enough, we figured that the R domain had to be well structured or it couldn't be the recognizer, whereas my guess now is that just as recognition of TAR by TAT seems to involve an arginine-rich peptide that is templated onto a prestructured piece of RNA for the specific recognition event, so I suspect the same is true in viral packaging. Interestingly enough, the first efforts in Jack Morris' lab to identify the nucleation or initiation sequence were overly complicated, but they got what, I think, is the right answer from later experiments in his own lab.

SCH That didn't work, but they ultimately did mutagenesis or deletion analysis or something like that, but looking back I discovered that somehow some early experiments that Peter Stockley had done with RNA protection, I think, had identified the right fragments. But we hadn't the confidence at the time - when Jack did much more thorough experiments and seemed to get a different answer - we were not confident enough in our own answer to question those results.

SCH Peter Stockley was a postdoc. He had come from Jean Thomas' lab. Interestingly, he's wound up doing, with the RNA phage, the mature version of the experiment that we had hoped he would do on TCV. He has collaborated with the Swedish group in studying RNA phage structure and assembly. He was able to get for them a crystal with 90 copies of the stem-loop that is recognized as an initiation sequence, incorporated into a virus-like particle. There is one copy bound against each of the 90 dimers in the particle.

He managed to get the RNA into the empty shell and bind in a way we believe is likely to be specific; mutagenesis suggests it is. So he has succeeded ,in collaboration with the Swedish group, in showing what the basis of RNA recognition in that case is. It does depend critically on RNA structure and also on an ordered part of the protein, because there is no R domain. The protein is all ordered. My guess is that in TCV, if we were now to try to crystallize or study by NMR the specific initiation or nucleation sequence, assembly origin, or whatever you want to call it, on TCV RNA, together with a fragment of protein, we probably would find that the RNA-interacting fragment of protein would be a piece of the so-called R domain, at the N-terminal basic region, and that bit wouldn't have any structure on its own, but would somehow be configured onto the RNA the way the relevant piece of TAT is on TAR. That's never actually been properly established, however. Alan Frankel's studies of the TAT/TAR interaction were, in my view, the experiments that validated this sort of picture. It certainly changed my view of how initiation of viral assembly would have to work.

SS Was Alan's contribution the idea that the protein was not structured until it interacted with RNA?

SCH It was the observation that led to that idea. What he focused on was the fact that arginine was enough. I felt he might have overstated that case, but the key notion was that the preformed structure in the interactions between RNA and protein, - or at least largely preformed - and the most complex folded 3-dimensional information, came from the RNA, which was then recognizing as a "substrate" in some sense the peptide, a piece of relatively flexible protein. We had always been thinking of a case like TMV, which is the other way around. In TMV, you have flexible RNA that fits into a groove on the protein, so TAT/TAR (and by implication TCV) is sort of anti-TMV. To the extent that the extension to TCV and other spherical viruses is true, then there was a really novel extrapolation from the NMR observations that Jamie Williamson and Frankel made. That is they did the NMR structure and, bingo, in my view there was an interesting conclusion about a very different system.

SS If it had been done many years before it would have really been difficult for people to accept, but it came just during the time that people were beginning to realize that RNA could act as a protein.

SCH That's right. So it didn't seem particularly revolutionary, and it wasn't, in the sense that it was another piece of RNA structure. But what I found most remarkable was that it was RNA acting, if you wish, like an SH3 domain recognizing a bit of previously loose peptide, just like an SH3 domain takes a bit of loose proline rich peptide and configures it into a polyproline 2 helix.

SS This is slightly tangential but I think relevant. It must make you now very uncomfortable to look at the protein without its RNA or nucleic acid substrate. We talked about this before but if you look at a protein structure that is supposed to be binding to an RNA, you might get a very different picture.

SCH On the other hand, there aren't dramatic changes in the RNA phages, MS2 or whichever one was done, they were all pretty similar. There aren't major changes in the protein. There, the RNA docks into a nice groove. I guess the same happens with non-sequence specific RNA binding domains, the ones found in splicesome proteins and so on. Again, that's a folded domain of a protein that configures a bit of single stranded RNA across it, and so there is a protein configuring a piece of RNA rather than the other way around.

SS Proteins can still do a lot of nice things. So, now at this stage, were you thinking of continuing with plant viruses, or I should ask what directions were you thinking of going?

SCH Well, that was an interesting question. I was sort of agonizing about what I wanted to do. Jim Hogle had come to the lab specifically with the notion of learning virus crystallography so he could go on to do polio virus. Picornaviruses were an obvious next step, so that one thing that then began to bother me in terms of overall lab planning strategy (over the trajectory of 5 to 10 years) was the following: had it been unwise to "give away" picornaviruses, which were the obvious transition from the RNA plant viruses to studying animal viruses, or structural animal virology? So Jim had gotten launched, and by 1985 he had the structure of polio virus, which he did at Scripps. A key question was whether I should continue to work on assembly, but until mutagenesis and recombinant proteins came along, that was obviously going to be frustrating. If I had known it was going to come along, I suppose I would have jump-started such a project, but one didn't know that recombinant proteins were going to come along, quite quickly, or at least I didn't. So I decided to think about working on polyoma virus.

I had several reasons for turning to polyoma. I had actually crystallized polyoma virus by accident when I was doing small angle scattering studies as a graduate student. I think it turned out that Murakami had previously made (perhaps even published) an observation that he had crystallized it, so it wasn't new. But I had noticed little spots on a small angle diffraction pattern and gone and looked at the sample and sure enough there were tiny crystals there. So it was always in my mind that one could go on to somewhat larger viruses. I had hesitated to do so around 1983-84 because Don Caspar was actually working on it, doing crystallography of polyoma. Ivan Rayment (I don't remember the exact date, but it's '83, '84, '85) had come to Don's lab and they had started to work on polyoma virus. Indeed Ivan had made the observation - or not observation but conclusion - surprising to everybody at the time, that polyoma was made entirely of pentamers rather than a mixture of pentamers and hexamers as had been guessed - or not guessed but interpreted - from lower resolution electron microscopy by Klug and coworkers and quasi-equivalence theory.

SCH Yes, so just jumping back for a bit, what drove Klug to start thinking seriously about how to treat electron microscopy quantitatively and led ultimately to the entire foundation of the field of molecular image reconstruction from electron microscopy was an effort to try to understand whether various viruses obeyed the notion of quasi-equivalence that he and Don Caspar had put forth in 1962. There was, in particular, a key paper on human wart virus and also a paper on polyoma or SV40 - I don't remember, I think it was polyoma - that John Finch had done in the mid to late '60's in which they interpreted their images as consistent with quasi-equivalence theory, because it was impossible to tell whether the six coordinated morphological units were hexamers or pentamers; they interpreted them as hexamers. Indeed, there was also a tubular array of polyoma protein, which was interpreted as a hexamer tube; in some ways that was the best evidence for hexamers, since it was clear that the resolution on human wart viruses or on polyoma from negative staining wasn't sufficient to conclude that the hexa-coordinated bumps were pentamers. It seemed to fly in the face of thinking about specificity, that the hexa-cordinated objects would be pentamers. So what Ivan did was collect data to about 25 to 30 angstrom resolution, as I recall, make a model based heuristically (if you wish) on lumps at the positions indicated by electron micrographs, use that model to obtain phases, and then use non-crystallographic symmetry methods to refine. And he kept getting pentamers at the hexa-coordinated positions, even when he used spherical lumps. But everyone, myself included, was pretty skeptical of the results for quite awhile.

SCH Well Don got all excited because Don actually has an incredibly flexible mind, and he was convinced. Don, obviously had the data under his nose, and the rest of us didn't have the data under our nose, so we didn't know how confident we should be in the data. Don believed that the data were good and therefore, pretty quickly came to the conclusion that there were indeed pentamers. This violated the simple extrapolation from quasi-equivalence, and Don got quite excited about the answer. Klug resisted it strenuously. I have some correspondence, actually, in which Klug's group tried to go back to the analysis of the hexamer tubes, which all of us agreed would be a clincher. That is, if the protein could make hexamers in tubes, then we suspected it was likely that it would also do so in the virion. A reanalysis of the hexamer tubes seemed to confirm that they were actually hexamers and not pentamers. But the clincher was when Tim Baker, working with Don Caspar and Ivan, showed that the "hexamer tubes" were in fact made of pentamers. The so-called hexamer tubes were just another lattice of pentamers.

There was a pentamer tube that Kiselov and Klug had characterized; they thought there were tubes made of pentamers and another set of tubes made of hexamers, which made sense if they were both morphological units. But Tim Baker, by showing that the so-called hexamer tubes were actually made of pentamers - by getting really good electron micrographs and doing a nice reconstruction - pretty well "iced the cake". Any little lingering doubts you might have, that somehow pentamer bias could have crept into Ivan's crystallographic phase refinement in a way we didn't understand, went away, because Tim's reconstruction said that there was no evidence that the protein ever made hexamers. So I found that the crystallography was very good, but you just always worry when you find a totally surprising result. You want to see something else that confirms it. If you then find that even a lattice that ought to be a packing of hexamers is made of pentamers - well then -

SCH So there are no hexamers. And then a brand new puzzle arose, namely, how do you fit a five-fold peg into a 6-fold hole, without needing to invoke something that I was very uncomfortable about, namely, non-specific greasy surfaces. Don Caspar appeared to be thinking that way. There was the word "non-specific" that appears, I think, in the abstract of the Rayment et al. paper.

SCH This is the Nature paper, and that word really bothered me. It seemed to me that we were going to have to have some way in which the structure of the protein didn't violate specificity, in which it still had nice lock and key interactions in some local way, but managed nonetheless to assemble correctly. Suddenly, the structure became a lot more interesting for that reason. I hesitated a bit to work on it, because I thought Don would probably move forward on it but when it became clear that Don wasn't going to make rapid progress on it, Tom Benjamin and I decided we would see what kind of crystals we could get and whether we could get crystals that would actually diffract to very high resolution. The crystals that they looked at in Don's lab, although they were the same crystals, didn't appear to diffract very well.

SCH The difference, I think, was ultimately just that we took them to a synchrotron and that we kept pushing. In other words, I think the difference was looking for weak high resolution diffraction, finding it, being optimistic, growing bigger crystals and perhaps subtle differences in how we handled them.

SCH Actually because Jim Hogle had left the lab and there was no one to carry on the virus crystallography, I started to work on polyoma myself. There was a postdoc from Tom Benjamin's lab, who helped grow the virus and actually grew the crystals, but all the crystallography, crystal mounting and all of that stuff I did until I got help from two different people. A guy named Javad Moulai, an engineer who came to the lab for a while and whom I took on, in part because I was desperate to have some help on the project., Then Youwei Yan, who arrived from China in about 1986 or so, also undertook to help. He was originally working on something else. So the goal was to see whether we could get high enough resolution diffraction so that we could really turn polyoma into a high resolution structure problem. I was reasonably confident that Don's group was going to focus more on low resolution issues and that we probably could work things out so there wouldn't be a conflict.

SCH We didn't make any formal contract. It was one of those situations where in retrospect I think things worked out wonderfully. I don't think Don would have carried it forward in that way. It involved major league synchrotron stuff at a time when very few people were doing it, and Don's own way of thinking tended to focus more on low resolution questions. I think it was actually better to push forward in a quiet way, to see whether we could set ourselves up to do something that would be different enough from what they were doing. Meanwhile Don was doing very pretty things with Bob Garcea about assembly - discovering polymorphic forms in assembly by electron microscopy and so on. There was a nice paper in Cell about reassembly of polyoma. What Bob and Don had shown was that somehow you could get under the right conditions pretty good reassembled particles just from recombinant VP1, unmodified. It said that the rules for getting an appropriate assembly - one in which the 5-fold pegs "knew" to be in 5-fold or in 6-fold holes in the right way - were probably related to those I mentioned for TCV - namely as the assembly units add, they surround themselves correctly as things go.

SCH When we found that with TCV I was excited by it, but since it seemed like a recondite concept to more biologically oriented people around me, I never was able to sort of share the excitement very much. We published a paper that I am still very proud of in JMB about assembly of TCV (J. Mol. Biol. 191: 639-658, 1986), but I have never found a way of getting it across in lectures.

SCH The biologists said evolution figured out how to make sure it assembles accurately. It's built into the protein. How it gets that way - well that may be your business. I'm willing to believe that evolution could figure out how to do it, but there still is a real physical chemical puzzle, namely how do you describe a pathway of assembly in which there is "local templating" as you might call it - I have never found the right word. Sometimes I've called it "autotemplating", Don Caspar used a word like that. It would probably help if I found a really good word to describe a process in which you get started with a specific initiating structure, followed by successive steps in which an assembly unit- in this case a pentamer - is added in such a way that conformational switching is determined by the position at which it adds.

SCH The polyoma pentamers - or the dimers in the case of TCV - have, if you wish, alternative ways in which they could add. Some of which lead to aberrant assembly - for example, the tubes in the case of polyoma, or the small particles in the case of TCV - and others which lead along a productive pathway that correctly gives you hexa-coordinated pentamers where you need them in the shells in the case of polyoma, or CC vs. AB dimers where you need them in the case of TCV. Actually one reason why it's been hard to describe these processes or to get people excited about them is when you reproduce it in vitro there are a lot of mistakes, and yet when you do it in vivo, just making VP1 in insect cells for example, you get pretty perfect particles. So I think that like protein folding, there's probably editing by chaperone-like processes or chaperone-like particles.

SS In fact I remember there was a lot of discussion also by you about protein folding.

SCH Right, so it's high time to come back to it and figure out what chaperones might intervene in virus assembly. Why do you get beautiful 72-pentamer shells of polyoma out of insect cells when you express VP1, but when you try do it in vitro, the pentamers are clearly trying to make the right thing, but invariably there are mistakes in almost every particle. An in vitro assembly is sort of a semi-mess in the EM, where almost every particle is about right but is damaged in some way or incomplete in some way. Whereas what you get out of cells is obviously better. And that still is an important question. That's actually the most important question now, in terms of the biology of assembly, and as far as I know it's not being worked on. It's not easy to work on. It's an important problem to work on, and we've talked about it, but I don't see any easy way to do it.

SCH There are so many chaperones, and there are probably ones we don't know about yet. That may not be the right approach. Anyway, by about 1987 or '88, Bob Liddington came to the lab and agreed to take on the polyoma project.

SCH Right, but we quickly decided we also should look at SV40, since those crystals were better.

SCH At this stage we had polyoma crystals from which we could easily get 6 angstrom data and could often see spots beyond and were collecting data from them but not as regularly as we hoped. We were recording data on film, and this was being done on the old A1 beam line at CHESS which was somewhat weaker than the beam line in England on which we finally got the data. The old A1 was a good 25- to 50-fold weaker than the beam lines we now use for these purposes, so I think we could have gotten data to reasonably high resolution on a contemporary synchrotron source.

SCH That's probably a good idea. The most dramatic demonstration of the utility of synchrotron radiation for these kinds of projects was the rhinovirus structure, which was done at CHESS and basically helped put CHESS and the importance of synchrotron radiation on the map. So it was perfectly clear that any of these structures, in particular the large ones, were going to require synchrotron radiation. And we were in fact doing some serious stuff with polyoma, even before Bob arrived, at whatever synchrotron source we could get a bit of time on.

SCH I did some synchrotron work simply as a hired hand the year I was in Heidelberg in '70-'71.

SS In fact we talked about that. By then people already knew that it could be used, but CHESS wasn't available.

SCH Oh no. Keith Moffat started thinking about developing CHESS as a possible place to do biological experiments - let's say circa 1980 - I don't remember exactly - and I persuaded him to put in a proposal to CHESS. CHESS was an independent organization, and Keith, who was at Cornell, was beginning to think it would be interesting to try to develop the possibility of doing biological diffraction at CHESS - protein crystallography. So I persuaded him that he ought to put in a proposal to CHESS on behalf of a consortium of people, each of whom would come and try to do something. I think Michael Rossmann was probably part of that.

I brought a bushy stunt crystal or a bunch of them and an oscillation camera in the back of my car. We set it up using some modeling clay on the table at CHESS and got some pretty striking photographs, actually. I didn't at the time have a project that was likely to need that. Michael had the first really important project. That was the rhino project, but it was perfectly clear by the time we were working on polyoma that was the way to go. And it was useful that Bob Liddington had actually worked for six or nine months running a beam line at Darsbury (England) between two different stages of his educational career. I don't remember exactly when but he had actually been at Darsbury for nine months as a beam line scientist so he had an enormous amount of experience.

SCH He didn't know anything about viruses, that didn't matter. So we decided we had better look at SV40 in parallel. We made a prep of SV40 at the MIT cell culture center and observed rather better diffraction and so switched to SV40. In retrospect, I think the SV40 crystals were just a little bit more stable, because subsequently we were able to get almost as good data for polyoma, but not quite. We switched to SV40 and moved forward then fairly quickly and by about 1990 -

SCH OK and then we didn't publish it until early '91. But we probably had most of it done. Anyway that got the whole virus crystallography technology, if you wish, going again in a major way in my laboratory. Actually it set the lab in some other new directions because they were synchrotron based.

SS As the structure was coming in now - we have several different things that must have happened. First, how did this differ from Don Caspar's original idea? And then, it must have gotten you thinking about other problems of virology, because now this moved you into animal virology.

SCH That's right. So the structure of SV40 showed that the way you fit a 5-fold peg into a 6-fold hole without, in this case at least having to have ill-defined general greasy surfaces, was to have arms, even more elaborately extended than in bushy stunt, emanate from the pentamer and invade subunits of neighboring pentamers. The dramatic differences of the symmetry mismatch were compensated for by a difference in direction of the arm. Once an arm got to its target, however, it was doing the same thing, no matter what direction it came from. So in a certain sense SV40 has fully equivalent contacts among all the pentamers. Since there are so few contacts between the folded bodies of the pentamers, the arms make very different conformational transitions as they go out of their pentamer of origin and into their target pentamer, but once they are inside their target pentamer they are doing the same thing wherever they are. Ironically, it's almost totally fully equivalent - all the non equivalence, if you wish, is built into just how the arm gets from one pentamer to the other.

SCH And then embodied in the structure in different arm conformations in all the different locations.

SCH Oh, Don Caspar immediately recognized it was a really neat solution to the problem and wrote a lovely little piece in Current Biology, a news and views type thing, in which he drew a picture in which all of the pentamers were little octopuses - he called them pentapuses - and did a whole story about pentapuses. Don, as you probably remember, used to like to do rather bizarre drawings. There was one of hands.

SCH An array of hands. He did an SV40-like array of pentapusses, that is five footed octopuses, and it somehow has the same sort of slightly "yuk" and squeamish quality. It made the point very nicely.

SS The structure of SV40 and polyoma are structures that should have interested a number of virologists. What kind of feedback did you get?

SCH The problem has been that SV40 and polyoma have largely been used by people that study their early functions, that is as models for tumorgenesis, and so not very many people were paying attention to issues of viral entry and assembly. I think we are finally getting people - including our own collaborator, Tom Benjamin - interested in the entry problem, for example. The only person really working on assembly has been Bob Garcea, who's done some pretty things but hasn't followed it up extensively. So in some ways the impact had been more modest than it might have been.

SS At the time, the question of entry had already been addressed. That is there were people thinking hard about endocytosis. But not with polyoma.

SCH Well, Richard Consigli had done some experiments with polyoma, and in 1990 Ari Helenius published quite an interesting paper on SV40, in which he showed that, like Consigli's earlier work on polyoma, the particle appeared to bud inward from the cell surface. That is, it was wrapped in a bit of membrane as it was engulfed or endocytosed, whatever word you want, with the membrane fitting really tightly around it. I think that actually suggests that it simply buds inward; that is, an abundant receptor is involved at least in the initial step, and in the case of polyoma, we know that to be true. The abundant receptor is any appropriately sialyated glycan. So the virus can simply bud its way inward. The real question is how does that little vesicle, a sort of reverse-enveloped virus, get transported - targeted to the right intracellular membrane. Helenius' results suggested that it was targeted to the ER, because he saw a dramatic outcropping of the ER that contained virus particles a few hours after adsorption. We repeated those observations, or rather Tom Benjamin and coworkers at the Medical School repeated them with polyoma virus and could see the same thing. So it's a real observation, no question - what it means isn't clear yet. There are now known to be pathways for toxins like ricin that go from the plasma membrane to the ER. They're clathrin independent, of course. Polyoma and SV40 may actually be taking advantage of a route that exists in the cell for various things.

SS Could we go back to the time of the structure and get some more thoughts about how your thinking changed. For one thing, it must have gotten you back into thinking about virology because you were doing the structure. Now once you had the structure, how did it change your thinking or influence your thinking?

SCH It further focused my attention on something I said before, namely, that protein interactions in interesting complex, non-trivial assemblies, not just dimeric enzymes or something like them, involved much more elaborate kinds of evolutionary playing around with protein structure than simply two surfaces that matched. What we wrote about SV40 was that instead of being glued together across complementary interfaces, it was roped together with strings or tied together with strings. And actually one consequence for me was that it got me excited about thinking about looking at things like SH2 domains which looked like they would be subcellular assembly modules.

SCH Yes in some ways it was my further excitement about the notion of trying to understand peptide recognition rather than surface recognition. So that just at about this time our Children's Hospital lab started, and I thought it would be exciting to look at signaling assemblies. We decided to think about the inside of the T-cell receptor and started working on LCK, and that led to src via SH2 domains and so on. So one consequence was actually to make me think that there would be surprising novel and interesting aspects of that kind of subcellular assembly which was just then clearly emerging as a key basis for much of signal transduction. And that it would be exciting to do structures about that.

SS I wouldn't have thought there would be a connection between DNA viruses and SH2 domains.

SCH Well it was a connection only in my mind and my thinking about motifs in assembly. It also convinced me that there would continue to be interesting surprises and important things to learn from looking at larger and more complex virus structures. We had already- in 1988 or so - probably earlier than that - '86 or '87 - crystallized reovirus cores, but they were impossible to work on. Clearly it was before 1987 because it's written into my first Hughes research plan - which would be in '87 - so that those crystals we must have done from cores that Max Nibert gave me in 1986.

SCH Not yet. We were working on rotavirus by '89, because Dick Bellamy had heard we had crystallized reovirus cores so he thought we should try to crystallize the corresponding structure from rotavirus.

So another consequence of the SV40 and polyoma structures was simply to remind me that there were going to be exciting and interesting structural surprises in these larger more complex structures and one shouldn't agonize too much,- but just go ahead and try to do some of them. In the case of polyoma, the reason to do polyoma once we had done SV40 was the receptor question. The receptor for polyoma as I said was an appropriately sialylated oligosaccharide.

SCH Paulson, originally, actually by studying hemagglutination by polyoma virus presuming that had to do with receptor recognition. He showed first that you had to have alpha 2,3 linked sialic acid on the red blood cells in order to get adequate hemagglutination.

SCH With polyoma- and ultimately he did do the key experiment - showing that if you treated cells with neuraminidase you could restore infectivity with the alpha-2,3 sialotransferase and not with the alpha 2-6 sialotransferase. And so we went ahead and determined the structure of both polyoma and polyoma with the appropriate oligosaccharide and showed the basis of receptor recognition.

SS Since this is a history, let's talk about some of the people working with you because there were some very interesting things going on in the world at the time. Weren't there several Chinese working on polyoma?

SCH Let's see. I had gone to China in '84 whenever the Sendai (International Virology) meeting was. Jiahuai Wang had come to my lab in '81, I think, but I don't remember the exact date, as one of the first Chinese visiting scholars to come to the US.

SCH Oh yes, a long time. Jiahuai had come and worked with me and with Mark Ptashne on lambda repressor; actually he worked with Carl Pabo. He then went back to China and started his own lab and actually started putting together quite a good group, working on a ricin-like protein called tricosanthin which the Chinese were quite interested in because it was used in abortions to control Chinese population. There was a large tricosanthin project around China, people doing the sequence and so on. He was doing the structure and had quite a decent laboratory built up. He decided to come to my lab in the fall of '88 for a bit of a sabbatical. I had visited him in '84 and then as a result of that visit Youwei, whom I met in Shanghai, came to my lab in about '86. Jiahuai and his wife Jinhuan Liu, who's a chemist and works with him doing crystallization, came in the fall of '88 ostensibly for about 9 months. When Tiananmen Square blew up in June '89, they decided to stay. By then, he was working on CD4.

They decided to try to stay, but the real problem was to get their kids out of China, which took us until March 1990 ? about 9 months. They were denied a visa three times by our consul, and the second two times with intervention from Senator Kennedy's Office, first in Massachusetts and then in Washington. Finally I was told that maybe I should call Tony Fauci's office because he was working on AIDS and that Tony Fauci could do better than Ted Kennedy. I was given the name of the guy I was to call. It was Fauci's administrator for those kinds of affairs, whatever they may be, and he told me maybe they could help out through the Executive Branch. I assume that meant the White House. He really couldn't tell me very much but to send him the details and he would see what he could do. All I know was two months after I called, the Consul called the kids in China and said come get your visa.

SS You suppressed the other Chinese postdoctoral fellow, the one who spent all his time when we were in the lab working on politics

SCH Oh, Hai ching. I'll tell you about him in a minute. Two people came to my lab as a result of my visit to China. Yicheng Dong who came from Jiahuai's Institute, and Yaiwei Yan. Yicheng Dong stayed only a year. He was a good crystallographer, but the English communication problem for him was quite severe. He worked with Cynthia Wolberger on 434 Cro, and his name is on the paper. Yawei was very successful. He made important contributions to the SV40 project, collaborated with Dan Branton on the structure of spectrin, and ultimately moved to Merck in West Point, PA

Hai ching Zhao had come to the lab I guess early '89, because he was there when Tiananmen occurred, but not a lot before that. He came from the University of Connecticut, where he had done a Ph.D. with Jim Knox as one of the new generation of Chinese students who had come out of China and done an American Ph.D. at one place or another. He was especially interested in organization or administrative things. He was a big leader in the Chinese students' community nationally; the students has all come under a program called CUSBEA, Chinese-U.S. Biological Exchange , and he was involved in the annual national meetings of all the CUSBEA students and stuff like that. Indeed, in the summer of '89 it was already planned to have one of those meetings, and he was organizing it at Harvard in July or August of '89. Of course it was an exceedingly emotional meeting, and he asked me to give a talk. Actually, the meeting was mainly focused on science. It was quite good in that way, but he asked me to give a keynote talk. I can't remember now exactly what I said. He taught me one sentence in Chinese, which I have lost, to say "welcome" or "glad to have you here" or something like "welcome to Harvard", which got a laugh. Basically for the next year he was a full time lobbyist; in fact for the next n years he was a lobbyist. I think he probably is the person most responsible for a law that passed Congress at the very end of the Bush Administration. Bush pocket vetoed it, and it was then overridden. So that would have been - I can't remember if it was '91 or whenever. It basically granted more or less automatically a green card to any visiting Chinese scholar who was in the country at the time of Tiananmen Square - student, postdoc or sabbatical visitor like Jiahuai. I think actually it was an incredible move, because it meant that the U.S. overnight acquired a generation of the very best young scientists from China. It was quite a stroke actually. It meant all these people could settle in without much hassle. They all wanted to try to figure out how to stay, and this made it possible without great hassle. Ten years from now, when they will have made their first major scientific contributions and academic contributions, it will have been seen as a pretty interesting historical event in the history of academic sciences -in the demographic sense.

SCH Not as many as should. I think somebody ought to start doing a Ph.D. thesis on it now. I think that Hai ching is now still in Washington. After all this was successful, he set up some sort of company or organization to provide services for Chinese students now trying to settle in the United States. They couldn't get credit easily, because they didn't have a credit record, so there was some sort of credit union to provide loans to buy houses or whatever. I don't know what he's up to now, but for a long time when I would be in the presence of some Chinese students at some of these affairs and somebody would say "Hai ching was in his lab", they would say "Oh, Hai ching Zhao." He was much more famous than I was - and a very engaging guy and very skillful. I think Youwei Yan whom I just saw at Merck, said that he was not in touch with him exactly, but had made some contact a year or two ago. Actually I would like to track him down again and find out what he's up to.

So to add one other little coda to the Jiahuai Wang childrens' visa issue - their son, Jing, graduated from Harvard four years ago and is now an M.D./Ph.D. student at Columbia Medical School where he is doing a Ph.D. with one of the people in the neuroscience programs. Their daughter, Miao, just graduated from the University of Chicago in economics and is working in consulting.

SCH It's a neat story, actually. Those were the kids that we struggled to get visas for, and I think they are paying back our confidence in them in a very nice way.

SCH I understand. One of the things that the Web exposure of this might do is perhaps that people interested in the history of science will look at it, and it would raise the issue of that generation of young Chinese scientists whom we so successfully captured. It could have people start to look in a prospective way of their role in biology and other areas, physics as well, in the U.S. Since they are now getting assistant professorships, it would be very interesting to see what kind of impact they are making. Two of them, I happen to know, have been nominated as young Hughes investigators or at least are in the Harvard competition this year. That by itself is an interesting statement.