A Discussion with
Recorded and edited by
April 21, 2000
Harvard Medical School
SS Jim, let's start with where you grew up and your first interests in science.
JH I was born in Detroit. When I was 9 years old my family moved. My father was transferred to Minneapolis, so I regard Minneapolis as my home. I guess because it shaped a lot of the things that went on in my life in later times. My interest in science started when I was really young. I had no real concept of science or what it was all about, but I loved tinkering with things, constantly taking things apart and putting them together, trying to figure out how things work. I think in common with a lot of people in science, I was also a bit of a pyromaniac as a child. Basically from an early age, I liked science and when I went to the University of Minnesota got into a lab as quickly as possible, making my way in as a glassware washer, working my way to having my own experiments to do.
SS Was it a biology lab?
SS What years are we talking about now?
JH As an undergraduate, '68-'72
SS And then you went to graduate school?
JH Graduate school at the University of Wisconsin. I was there from '72 to '78. It really wasn't clear what I wanted to do when I went there. There was a real paradox. I wasn't sure what I wanted to do, but I was very visually oriented. I was accepted to both the oncology program and the biochemistry program there. I decided against oncology because I was afraid it might be too specialized and then ended up working in a crystallography lab.
SS Let's talk about how you decided which lab to go into.
JH I talked to a number of people and started the summer before my official program began. One of the people I was really impressed with at the time I was interviewing was Roland Rueckert. I was really interested in the things he was doing. He was not available the summer that I started so I decided to do a summer rotation in Sundaralingam's lab, because I wanted to find out what crystallography was all about. I was immediately hooked on crystallography. I went on to do a final rotation in Julian Davies' lab, again a project that was intended to have structural implications: I was working on one of the gentamycin adenylases, one of the proteins that inactivates antibiotics. Then I decided to commit to Sundaralingam's lab to do the structural work. One of the projects early on that I was interested in was ribosomes. That would have been a great project for a graduate student in 1972!
SS You probably would still be a graduate student.
JH Masayasu Nomura was at Wisconsin at the time and had shown that you could reconstitute the bacterial ribosome from the small and large subunits. Actually most of the proteins from the small subunit had been purified. We started looking at trying to purify some of the individual proteins. But Nomura's work really clicked with me because I had been interested in protein folding and the idea of being able to self-assemble something really caught my attention. And that was essentially also the "in" back into viruses. I was fascinated by the idea that the program for assembly was in the sequence of the protein and that put together a three-dimensional structure. I worked on a number of projects in Sundaralingam's lab. My initial thesis project was supposed to be thymidylate synthetase. The protein came from an amethopterin resistant mutant of Lactobacillus casei, a strain that developed multiple copies of the gene and produced lots of the protein. We made ourselves very unpopular on several occasions by growing large batches of lactobacillus. The smell of the culture when harvesting cleared the building. We had crystals and a number of other groups did as well, but there turned out to be an intractable problem with the crystals. As a matter of fact, it was another project that a graduate student could have worked on for a long time. Bob Stroud at UCSF was working with Dan Santi and they didn't solve that structure until the late 80's. This would have been another long-term commitment for a graduate student.
The deciding factor for that was when Charlie Heidelberger died - he was our collaborator at Wisconsin and when he died, it became clear that it was time to find a project to work on. My thesis project was what you might call a parachute project. It was to look at a monoclinic form of egg white lysozyme. Lysozyme crystallizes in a number of different crystal forms which differ in the way that the lysozyme molecule packs in the crystal. At the time there had been a structure solved from the tetragonal form by Philips' group at Oxford, and the structure from the triclinic form was being solved in Israel and at the University of Washington. The monoclinic form had two molecules in the asymmetric unit. My interest in the structure was to see what the effect of crystal packing was on the structure. A real question at the time was how great was the deformation in the solid state compared to solution. I hypothesized that you would see deformations but they would be minor and essentially give you some idea of what the real flexible parts of the protein were, based on the assumption that you can only induce distortions that can be induced at a very low cost in energy. This project marked a gradual transition into crystallography for me. I originally went into a crystallography lab to be the resident protein chemist, not necessarily to be in crystallography, and ended up in a hard core crystallography project and, in fact, one where the final readout would require refinement techniques that did not really become accessible with the computers available until well into the time I was in Steve's lab.
In the meantime I maintained an interest in virology. All the students were required to select a seminar series to attend during their graduate studies. I went to the virology seminar series and presented a couple of seminars at that series including one on image reconstruction techniques from electron microscopy. There was an amusing incident while I was preparing for that seminar. Sundaralingam frequently had visiting speakers and when he was busy, they would wander around the lab, but we weren't always introduced to them. Visitors would go around the lab and talk to people, and this guy came up and asked what was I doing. I said that I was preparing a seminar on this really interesting topic of image reconstruction technique by Aaron Klug and this other person whose name I forget. That person was David DeRosier, who was standing behind me. He doesn't remember that, but I reminded him of that when I first came back to Boston in 1991. I was very interested in the image reconstruction from electron micrographs and also was aware of the work published in the Cold Spring Harbor book in 1971 that was mentioned in Michael Rossmann's oral history. Steve had an image reconstruction of TBSV, but the most striking thing for me in that volume was Bror Strandberg's precession photograph from crystals of STNV at 3 Ångstroms. This one photo really said that one could, in principle, collect data on a virus at high resolution.
To me the idea that you could collect high resolution crystallographic data on a virus was very exciting, and became even more exciting when it became clear that Michael's group and Steve's group were making real progress with their crystallographic studies of icosahedral viruses. It was clear that these structures were going to be solved in the near future. As Roland discusses he and I started talking about polio at that time. He had crystallized it accidentally and in fact earlier it had been crystallized in Berkeley. That crystalline virus was sent to England. Rosalind Franklin, before she died, had looked at crystals and showed that they diffracted, and John Finch and Aaron Klug followed up on her work and actually published a paper that described crystals of type 1 Mahoney virus. The original crystals diffracted to very high resolution but were very X-ray sensitive. But at the time of the publication, 1959-1960, there was nothing you could really do. Even if you could collect data there was no way to do a structure of any reasonable resolution.
The second seminar I presented in this series was the structure of TBSV (tomato bushy stunt virus). Just before I left Wisconsin, the 5.5 Ångstrom structure of TBSV had come out (from Steve Harrison's lab). By that time I was pretty sure I was going to work for Steve as a postdoc, although along the way to that decision some problems came up that almost prevented me working with Steve. I had put out several job applications, including an application to Steve. Initially Steve had politely declined, as he had just hired a new postdoc and was reluctant to increase his group any further. Fortunately for me, the postdoc subsequently decided not to come to Steve's lab, and Steve called me back and asked if I was still interested. I was thrilled, and might have actually joined the lab in time to be involved in determining the structure at 2.9Å, but because of a delay in refining my structure and writing my thesis, I arrived just a bit too late.
SS So when you were talking about going to Steve's lab, you weren't thinking about doing polio?
JH I had talked to Roland about doing polio as an independent project later on, but it was clear that it was going to be a huge undertaking. There was a real advantage to working with plant viruses that made them much more tractable projects given the resources available at the time. The plant viruses such as TBSV tended to be more stable physically than polio and could be made in much larger amounts. In fact, several of the years I was in Steve's lab, in the Spring and Fall, I could grow 5 grams of both TBSV and TCV. With 5 grams you had a visible pellet - that was huge. Going to animal viruses, if you had a milligram or two, that was a large prep. So we're talking about 3-orders of magnitude difference in amounts that can be produced conveniently. It was just a great way to get into virus structure work at a time when the technology required to solve structures was just being invented.
However, in terms of the biological implications working on plant viruses was frustrating. Although, there was a lot of information about the assembly of the plant viruses at the time, there was so much less information available about how infection by plant viruses worked compared to the better characterized animal viruses (such as polio) or even phages. That is not to say that all plant viruses were poorly studied in terms of things other than assembly, but very little was known about the viruses that were being worked on as the structural models. Very little was known about TBSV and there were only a few reports at that time about TCV. So there was an obvious attraction of the animal viruses, especially polio. Because of the interest in vaccine development in the '30's, '40's and '50's, it had been widely studied. As a result it became one of the models of choice during the '60's and early '70's for studying the molecular biology of animal viruses. With this wealth of information, it was clear that structural information was going to be immediately relevant. It was also really clear how valuable working in the Harrison-Wiley lab was at the time. Watching Don and Ian's structure of the influenza hemagglutinin come out and just immediately seeing what knowing the structure of the hemagglutinin told you about the biology of the virus. There was already lots of information known about antigenic variation, about fusion and cell entry. It was clear that there was going to be a context for the structure of the animal viruses that was going to be very exciting.
SS It was going to have a much bigger impact immediately than the structure of TBSV
JH TBSV in itself was a stunning achievement from the structural point of view and from just defining how the rules of quasiequivalence can work or not, depending on how you look at it. And starting with some of the ideas about how assembly is keyed when it's not a fully T=1 virus, that was a really key development from TBSV, and obviously a remarkable accomplishment given that for Steve it was begun when he was a graduate student, pursued while he was a post doc, and was his first project as a starting faculty member. It was essentially a 1 to 2-man show for quite awhile - quite a remarkable thing itself and clearly there were a number of things that were developed as a result of that work that were key to being able to think about working on poliovirus at all. For example, one of the important developments was the introduction of new methods for collecting and processing oscillation data from crystals with very large unit cells. You couldn't collect many pictures from a single crystal, you had to come up with a way of recovering the intensities only partially measured on one photograph. That was the work of Winkler and Shutt in Steve's lab. Obviously parallel efforts were going on at Purdue. Jack Johnson had developed a similar method in Michael's lab. These contributions were key to pushing the limits of processing oscillation data, and making it possible to acquire high-resolution data from virus crystals.
SS I suspect people don't appreciate that today.
JH You have to remember that we were working with computers that, even with the largest of them available at the time, were primitive by today's standards. It seems strange to say that now, because they didn't seem so then. When I was a graduate student it took an act of God to get more than 32K of memory because of limitations imposed by the address space of the available computers at that time. When I first came to Steve's lab, we were doing most of our computations at a remote site at Columbia (in New York City) on an IBM computer over a very flaky phone line communicating with it via punched cards. When the first VAX computer came in to the Chemistry Department at Harvard, it was a real godsend. We were sharing even then, with Don's group, with Lipscomb's group, with Karplus' group, and with several other computational chemists at Harvard. So it was a heavily used resource.
Another key component was the development of methods that allowed structures to be determined using noncrystallographic symmetry. Spherical viruses are highly symmetric. Although they are very large the high symmetry of the structures can be used at many stages of the structure determination. However, until the mid '70's it had not been possible, given the computational resources of the time, to take advantage of the constraints provided by the noncrystallographic symmetry. Rossmann and others (including Peter Main and Tony Crowther) had much earlier demonstrated the utility of noncrystallographic symmetry constraints applied in reciprocal space (also called diffraction space) and had utilized real space averaging as a method for improving maps. The problem was that, as originally formulated, the reciprocal space constraints became intractable as the resolution increased. One of the key developments in the mid '70's was Gerard Bricogne's demonstration that iterative refinement in real space (averaging maps and then doing Fourier transform of the average maps to obtain improved phases) was equivalent to real space averaging and computationally tractable. In addition to establishing the validity of iterative real space averaging, Bricogne then developed a very "cute" technique that would allow a super computer, a big IBM, to handle a problem like TBSV to 2.9 Ångstrom resolution. So those developments all came in the context of the TBSV structure and the TMV disc structure.
SS Was this being done while you still were at Wisconsin?
JH That work was being done in Steve's lab at Harvard and by Gerard in Cambridge while I was still at Wisconsin. Parallel work with southern bean mosaic virus in Michael Rossmann's lab at Purdue was pushing the development of some of these techniques as well. So all this key methodology was essentially being developed as the structures were being solved. Each step on the way you could almost punctuate by: Here's something that actually allowed this part to be pushed to 5.5 Ångstroms with TBSV and then to 2.9Å. Bricogne's averaging programs and the post-refinement programs of Winkler and Schutt with TBSV structure; those were among the key developments that essentially allowed you to make the next leap. The tools were there and available for us when we started working on polio.
SS We're going to polio a little too soon, we're still on turnip crinkle virus.
JH The dreaded killer turnip crinkle virus. Just before I left Steve's lab I ran into a student at an ASV meeting who also worked on TCV. She was working on a satellite viroid that altered the pathogenicity of the virus itself, and had decided to make her project more interesting by renaming TCV, the dreaded killer turnip crinkle virus. To much of the world, it is a decidedly obscure virus. The rationale behind working on it was that unlike TBSV which cannot be disassembled in vitro unless you denature the virus, TCV can be dissociated readily to obtain isolated viral proteins and viral RNA. Both TBSV and TCV expand when you deplete divalent cations and raise the pH. However, TCV will then go on to dissociate into capsid protein dimers and RNA if you alter the ionic strength. As an aside, the expansion process itself is a interesting process that was first described in plant viruses (by Paul Kaesberg). A beautiful, and in my opinion, grossly under-recognized structure of the expanded form of TBSV was done by Ian Robinson, in Steve's lab, at the same time I was working with TCV. This structure has contributed strongly to the ways I think about structural transitions in the virus.
SS Was the ability to dissociate TCV known before you started?
JH That was known before I started. Steve had worked on TCV with Klug at one point and obtained crystals and found that you could make better crystals by making the mercury derivative with methyl mercury nitrate.
SS When the structure of tomato bushy stunt came out, do you remember anything surprising about the structure? I'll ask the same question about TCV so perhaps we can discuss them together.
JH Aside from the novelty of just knowing a virus structure, one of the most stunning things about the TBSV structure, I think, was the idea that on the inside of the virus, the C subunits have this lovely interaction where the arm of the N-terminal portion that linked the body of the subunit with the N-terminal domain wrapped across the bottom of the subunit, and then interacted with two of its mates around a 3-fold axis, and made what Steve described as a lattice that described a T=1 lattice. It immediately showed that a density feature that had been attributed to RNA in the 5.5 Ångstrom structure was protein and that that protein being ordered in the C subunits, in order to go around the 3-fold axis, caused the angle of the interaction between the surface domains of the C subunits to be rather flat for the C subunits. Whereas the A and B subunits where the arm did not intervene in the site, the faces of the subunits interacted with each other directly, the interaction was more angled. As a matter of fact, if you take the A and B subunits and take that interaction, you could model a T=1 structure. It had been shown earlier with TCV, that if in trying to crystallize the capsid protein you accidentally introduced a protease that cleaved off the N-terminal arm that you made a T=1 structure.
SS That was all known before?
JH That was all known before. The idea of the control of a T=3 structure by simply putting an arm up or an arm down, and that once you made an interaction of several dimers of a C subunit that you could now only add things in on AB conformations essentially gave you an idea of the metric of how the structure could be put together. No one truly believes that there's that full network that's formed, but it still gives you an idea of the metric that determines the unique size of the capsid. I think that was one of the real key concepts to emerge from the structure. Also there were some unusual protein folds. I think the biggest disappointment about the structure was that there was no ordered RNA.
SS Steve made two points; he made the point that the concept of proteins having arms that were flexible was one that not only wasn't known but that Roger Kornberg had argued against it.
JH As a matter of fact, it's easy to forget that we were all rather naive about how rigid protein structures were in the first place. We tended to think about all protein structures in terms of the structures that were available at the time, which were generally small well-folded proteins that were exported from the cell and functioned as monomers. We now know that proteins are far more flexible then we thought then, but even so the idea of having large segments of protein chain be highly flexible was surprising. But, you have to remember, what's different about the virus assembly (and to some extent the assembly of other multimeric proteins) is that you can have what we call "dangley" bits, as long as you hide them early in assembly. They would not be subject to the editing that would normally go on for a poorly folded portion of a protein in a cell. Normally, poorly folded portions of proteins are targeted into degradation pathways. However, if you hide the poorly folded bits in a shell that protects it from the proteases, especially the large intracellular proteases, you can have unfolded portions and not have them edited out. As a matter of fact, one of the other important concepts that came out of the observation of the disordered or partially ordered N-terminal extensions of the capsid proteins was that this relaxed the need for forcing symmetry on the packed RNA genome. Thus, the reason that the RNA wasn't ordered was because, conceptually and teleologically, instead of inducing order on a linear molecule, you allow a flexibility in an arm that connects a domain that is rich in positive charge to the body of the capsid protein that makes a shell. The bits that interact with the RNA are rather flexible and can accommodate an asymmetric structure. Although this also appears to be true in a number of other viruses, the overall generality of this concept has since been questioned because there are some viruses that have since been shown to have significant pieces of the RNA that adopt a quasi-icosahedral ordering.
SS I thought it was only the one that Jack (Johnson) was looking at. Are there others?
JH One of the DNA viruses, the parvoviruses that Michael (Rossmann) has studied has an ordered RNA fragment. Some of the viruses that Jack (Johnson) has studied have ordered RNA, and something like 20% of the RNA can be ordered to the extent of having well defined helical structures.
SS The other point that Steve made that changed his thinking was that in terms of packaging, that the major reason for having a packaging signal was so you didn't get more than one RNA molecule into the subunits rather than that the proteins would package only the viral RNA.
JH You really want to have some level of discrimination although the viruses occasionally help with that. For example, polio not only turns off host protein synthesis but also RNA synthesis and cellular messages are degraded so most of the RNA that is around is polio RNA. Then, of course, it has to discriminate between plus and minus strands, but there is a 100 fold excess of plus strands. Still it discriminates at a level better than that. It is clear that it picks its own RNA preferentially and that it discriminates between + and - strands.
SS Before we go to picornavirus, let's just say a little bit more about TCV. I guess the question is: did anything come out of the structural work that changed your thinking?
JH I don't think that TCV had the impact that TBSV did. It turns out that sequence-wise they are far less similar than we had thought based on the similarities of their structures in the electron micrographs. It was obvious that TCV was going to be very closely related to TBSV in structure and in fact, it was. There were some differences, but I think the differences were less important than the similarities. The real key with TCV was that it gave you a tractable system for doing assembly. The paper by Peter Sorger (Sorger, PK, Stockley, PG and Harrison, SC. J. Mol. Biol. 191:639-658. 1986) is really a beautiful piece of work.
Realistically, TCV meant more to me in the long run, than it has meant to the community as a whole. For the real value gained from solving this structure, was that it provided an opportunity to learn a tremendous amount from Steve. His way of thinking about switches in assembly and ability to interpret the biological significance of a structure were really valuable lessons for me as were his insights into crystallography. I learned a tremendous amount from the time working with Steve. We were a very small group at that time. There was Steve and Craig Steele who was a technician who was increasingly distracted by running the computer, especially when the VAX came. He, Ian Robinson and I, that was the Harrison lab at the time. Don Wiley's lab was correspondingly small (with Ian Wilson, Judy White and a technician Ed Gordon), incredibly small, unbelievable today, but we were very small. I had to do something I had not done a lot of before, which was to program. I had programming skills, but I am not a programmer. In my life after going to my own position at Scripps, I was fortunate to have Dave Filman come as a postdoc. Immediately after he arrived, he demonstrated that he is such a superior programmer, that I simply gave it up. At the time that I was working on TCV in Steve's lab there were a substantial number of programs that needed to be expanded in order to work on TCV which was technically a larger structure than TBSV. We needed to modify the programs for processing the data, and the programs for applying the noncrystallographic symmetry constraints. Don Wiley had already begun the process of modifying the programs, but there were still changes that needed to be made for the TCV structure, and no one else that could spare the time. In fairness, I got a great deal of help from several people, especially from Craig Steele.
SS Would it have been harder to do TCV before TBSV?
JH Yes. Computationally it would have been that much harder considering the fact that when TBSV was done they had a dedicated IBM supercomputer at the facility at Orsay. They had it for several days in order to do the calculations that they were doing. We did them on a smaller computer but were forced to do it in a very, very awkward way where scratch files were being written out onto low density tapes. Each calculation had to have a file that could occupy 8 full tapes. I was running these calculations at some point night and day and I had friends who were there who would mount tapes as my job requested them when I wasn't around. Again, computers have grown so rapidly that the difference in the couple of years between the TBSV structure and the TCV structure meant that instead of having a dedicated supercomputer facility we were doing it on a midsize VAX. We still found that we had to add some memory to the VAX and disks in order to get the VAX to do the TCV calculations at all. And so it would have been difficult because the crystallographic problem with TCV was larger than TBSV. Polio would have been even harder because you could only produce milligrams of virus, and the crystallographic problem with polio was even larger than TCV.
SS At this point are you still in Steve's lab?
JH I was still in Steve's lab. One of the interesting developments that did come out of the TCV structure was that it was the first, albeit very crude, application of phase extension from low to high resolution using noncrystallographic symmetry. When we were planning a strategy for solving the structure, we decided that the structure was similar enough to TBSV, that we could use the TBSV structure as the starting point for the structure determination, rather than go through the traditional heavy atom methods. We didn't have any information that said it was similar beyond about 12 angstroms, which is where I had some indication that one could recognize similarities in the diffraction pattern in precession photographs. So we started the phasing at low resolution using TBSV as a model, then extended the phases to higher resolution in fairly big chunks. So we had a kind of a hybrid phase extension where in each cycle of the process the refined nominal TCV phases at the current resolution were kept, the resolution was extended in large steps by appending TBSV phases to the next resolution shell, and these hybrid were refined by applying the noncrystallographic symmetry constraints. This process was repeated until we reached the limit of the resolution of the crystals (about 3.2Å).
In later work with rhinovirus 14, Michael Rossmann demonstrated that the phase extension worked better (and without a model providing initial phases for the extended resolution shell) when the resolution was increased in very small steps. However, with the resources that were available to us at the time, which was 1979-80, we could not afford to be doing the calculations with very small resolution increments and could not afford to do a large number of iterations of the calculations. We decided to extend in rather large resolution shells to reduce the number of iterations and to use TBSV phases as starting points for the next resolution shell. This approach worked, and the structure was essentially completed by the time I left Steve's lab in 1982. Unfortunately, the paper was not published for several years, as the sequence of the coat protein had yet to be determined. This was critical because the sequence provides a critical confirmation that the structure is correct. The sequence information is especially important when a structure is solved by molecular replacement.
So, again, I built up a real familiarity with all the computations involved and the programs. Don (Wiley) had started to port Gerard's (Bricogne) package for applying noncrystallographic symmetry constraints to the VAX. Gerard's programs were written for the IBM. Gerard writes gorgeous Fortran, but there were some things in there that were very specific for IBM and were incompatible for the VAX computer. Don had done a partial conversion of the program so it would run on the VAX for the hemagglutinin structure and I finished up the other portions that were necessary to do for the virus structure. You become very, very familiar with the program when you are going in and finding out because every time you change something, you find out you break something else and you have to be able to trace through the program. Fortunately Gerard's programs were lovely, and it was easier than it might have been, but there were some IBM specific things that were tricky.
As I was wrapping up the TCV structure, I started renewing the conversation with Roland about working on polio again with Steve's permission. This was going to be my leaving project. I would start the project and take it with me when I left. Steve was extremely generous. Steve himself probably would have loved to work on polio. Roland was more comfortable with making large amounts and shipping the Sabin (attenuated vaccine) strain of Type 1, and initially I had samples of the Sabin strain type 1 and some clues that he had provided me. Earlier work had demonstrated that the virus formed microcrystals when you pelleted it. This was problematic, as it was very difficult to redissolve these crystals. Roland had found that if you pellet the virus through a cushion of 1M NaCl in 30% sucrose, the pellet is nice and clear and easily dissolved and that said to me that the virus didn't like to be concentrated at low ionic strength. It immediately suggested two approaches to producing crystals in a more controlled fashion. One of these approaches was to begin with a concentrated stock of virus at fairly low ionic strength and then do a vapor diffusion experiment in which the concentration of the virus and the concentration of something like polyethylene glycol are increased. A second approach would be to make up a concentrated solution of the virus at moderately high ionic strength and then slowly or stepwise dialyze the salt away so you lower the salt concentration. Both worked, so I had crystals of the Sabin 1 strain within days of getting the samples from Roland. And then started a frustrating year, literally a year, of trying really everything to get those crystals to be stable to take a picture in the beam.
SS Were they just so X-ray sensitive?
JH I think that was the problem. I don't know if the structure has ever been looked at using the crystal form I first produced. David Stuart's lab has done some work with Sabin 1, but it may have been with a second crystal form that Carl Fricks and I obtained after we had gone to Scripps. This form seemed more stable, but had a much larger unit cell. With the crystal form that I originally produced, you could literally see the crystal being destroyed in the beam. If you had a big crystal, some crystals were quite large, where the beam was smaller than the crystal, you could see the beam track in the crystal and it would appear within minutes after opening the shutter. You could then move the crystal in the beam and see the beam track in a second part of the crystal. It was exceptionally X-ray sensitive, but wasn't clear initially if the failure to get any data was strictly due to the radiation sensitivity alone, or if it was due to the problems with manipulating the crystals during mounting. To check out this possibility, I grew the crystals in the capillary, growing them by dialysis. I plugged the capillary with polyacrylamide, and then did the dialysis through the polyacrylamide. The crystals essentially grew already mounted and ready for diffraction studies. You don't ever harvest them or even remove the liquid around them. It was a good approach to make sure the problem was the crystals and not the mounting procedure, but it would never be practical for data collection as the pictures were very noisy because you are shooting the x-rays through all the buffer.
SS Wouldn't there be lots of crystals?
JH They behaved very nicely. In most experiments I got a few moderate size crystals and several smaller ones. Steve's x-ray camera at that time had a very nice small beam and the crystals were big enough and separate enough so you could irradiate one crystal and see that crystal destroyed and the other crystal sitting next to it was still fine, and then move the capillary so you irradiated another crystal and that crystal would be destroyed and so on. Even though we were attempting to cool the crystals at the time, it was clear that these crystals were not going to be amenable for data collection. I think now you could possibly overcome the extreme radiation sensitivity by freezing the crystals and collecting diffraction data at -170C, but freezing protein crystals was not really in widespread use until much later. Freezing virus crystals without seriously damaging their ability to diffract x-rays was first done by Michelle Wien in my lab much, much later (1995 or '96). Up until that time no one had frozen a virus crystal in a way that preserved the order. I was a long way away from getting to that point.
I know you are familiar with getting crystals and not getting any diffraction. (SS had spent two years trying to get crystals of the enveloped virus, Sindbis virus, that would give a reasonable diffraction pattern.) After my experience with the Sabin strain, we decided it was really going to be necessary to try Mahoney. Roland did provide me with one small batch of Mahoney virus.
SS Were you at all worried about that?
JH At first I was a little nervous even with the Sabin. I had worked with tons of virus, but that was plant viruses. I had not ever worked with a significant human pathogen. We had to work to get permission to work on polio, even the Sabin strain and that was in spite of the fact that there was essentially a P3 facility in the old Gibbs laboratory where Steve and Don's labs were before they moved to Fairchild. The facility even had a pass-through autoclave. It didn't have a shower, but except for that it was a P3 facility. There was a centrifuge, a cold room and a laminar flow hood, essentially everything you needed to work with in one small lab. We had established protocols for working with the virus and gotten them approved, but I was still a little nervous when I first got started. Here I was with my first sample of Sabin 1 virus, aliquoting it out into small batches for crystallization experiments. I had a nasty solution of SDS, ethanol and bleach altogether (a decontamination solution that Judy White had worked with earlier) for decontaminating pipettes. I was working with a Pasteur pipette and was supposed to take a little of the virus stock and then wash the pipette in that solution and throw it away. I got the cycle wrong! I washed the pipette in the decontamination solution and then put it into the stock. I watched in dismay as the entire stock turned milky white and precipitated. Basically, there are times that you make yourself nervous enough working in a clean laboratory that you make mistakes. The protocols were good, I just had to learn to be relaxed about them.
By the time I started working with Mahoney, I was fairly comfortable. I worried far more about it in the early days at Scripps because my daughter, Michelle, was born shortly after we went to Scripps and she was not yet vaccinated. That was a bit scary. While we were waiting for approval to work with the virus in my lab, I was working in a true P3 facility at Scripps, and I was changing in and changing out each time I worked with the virus, but I still always made a point of changing clothes and showering again when I got home before I handled the baby. I was concerned about the Mahoney strain for that.
SS Had anybody checked your antibody titers?
JH I had my titer check several times. When I first went to Scripps we were looking at people in the lab in general and also looking at immune responses. My titers ran like a fresh vaccinee with the live vaccine, somewhere around 1:200 neutralizing titers, very common for someone who recently had the oral vaccine, but was never infected with wild type virus. But that was nothing like what many of the people who had worked with poliovirus for many years. Ellie Ehernfeld, for example, had sufficiently high titers that her antisera were used at one point for a classroom laboratory exercise.
SS You know the famous story that Bernie Roizman tells that when he was at Hopkins; they wouldn't let him work on polio so he switched to herpes.
JH And I believe herpes is probably far more dangerous in many ways. It gives you pause for thought, and Roland's concerns about growing the virus in large amounts were very real. At the time they were growing the cells in flasks on shakers with the danger of aerosols. The virus is mainly cell-associated, but you grow the virus until the cells begin to lyse and there are appreciable titers in the medium. He was concerned about growing the virus on a large scale, but was willing to send me a small sample to see if I could reproduce the original crystals that Rosalind (Franklin), Aaron (Klug), and John (Finch) had characterized. So Roland sent me a sample containing about a quarter of a milligram of Mahoney virus. I had crystals within days and reproduced the information that had been shown by Finch and Klug that the crystals diffracted to very high resolution and confirmed they were X-ray sensitive though not nearly as sensitive as the Sabin strain. We could get photographs of them. We could get stills, but in order to get a usable high-resolution oscillation photograph it would take exposures on the order of 10-12 hours, and the life-time in the beam was more like 4-6 hours. I knew that I was going to have to slow down the damage, and I spent a lot of time initially figuring out how to cool the crystals long enough to get one or two pictures.
SS What temperatures were you cooling to?
JH Initially to 4 degrees Centigrade in Steve's lab and while still in Steve's lab I began to push the temperature down even more, transferring the crystals to ethylene glycol to prevent freezing. In the end I found that the crystals would tolerate ethylene glycol concentrations up to 25% which allowed the crystals to be cooled down to -12 degrees. It was still not trivial to do this and it took some significant experimentation to get the entire process to be sufficiently robust to get a good picture. A lot of people worked with cooling samples before, but one of the problems with cooling is that the crystals are mounted in capillaries. If you have one end of the capillary cooling more than the other, then you begin to get transfer of water either onto the crystal which stresses and ultimately dissolves the crystal, or away from the crystal which dries it out. The polio crystals were not only exceptionally sensitive to x-rays, they were also exceptionally sensitive to desiccation. In order to control the water transfer, you had to make sure the temperature was as even as possible across the capillary. At one point I worked with a mounting technique in which I placed a strip of capillary paper soaked in buffer next to the crystal, all along the length of the crystal, so that there was an equilibration of the vapor with the saturated filter paper strip. We eventually got around that by making better nozzles and doing better temperature controls.
SS This is all before you went to Scripps?
JH Yes, although the set up was further refined after I got to Scripps, much of this was work that went on during my last year in Steve's lab, and I was able to get some very nice diffraction photos before I left Harvard. The amount of material we got directly from Roland Rueckert was sufficient to demonstrate we could get diffraction and that we had problems with the lifetimes of the crystals. It was the beginning of a trial and error period to get a cooling system working. Roland said he would be reluctant and maybe even unable to provide virus for a long-term effort and suggested that we talk to Baltimore's lab. This was very generous of Roland. I had tried growing some virus at the tissue culture center at MIT, to see if I could scale up to very large preps. I grew some virus, but it was clear that scaling up was not going to be simple. I don't know anyone that goes to 10 times the volume and gets 10 times the yield of virus. It really looked like the way to go was to work with medium size preps, about 6 to12 liters, where the yields would be semi-reproducibly large. Given my limited experience (and skills) with tissue culture, it was also clear that it would be very desirable to work with someone who was actually good at growing cells and producing virus.
It was about then that I started initially with a technician in David Baltimore's lab and then later with Marie Chow. Marie was in the lab as a postdoc and stayed in David's lab for a couple of years and then eventually moved into her own lab at MIT. Most of the virus that was ultimately used in the structure determination was actually supplied by Marie after she started out in her own lab at MIT. So we switched the source of virus from Roland to David and then to Marie's lab. The switch was sad in the sense that it would have been fun to work with Roland. I respected Roland a great deal. As far as I was concerned, the review that he published on the morphogenesis of picornaviruses, in Comprehensive Virology in 1969, was one of the most comprehensive and beautifully written reviews I had read. It shaped my thoughts about working on assembly intermediates and cell entry. My copy of that paper has handwritten notes all over it. On the other hand the collaboration with Marie has been very, very fruitful. In addition to the crystallography, she was interested in the immunology and pathogenesis of the virus and in vaccines, and the collaboration contributed significantly to the direction of our structural studies. One of her interests at the time that we first started working together was in exploring synthetic peptides as potential vaccines. This was a natural bridge with Scripps. I was considering an offer from Scripps at this time. The peptide work ended up giving us important clues about what the virus does when it comes into contact with the cell.
SS What were the peptides?
JH A year of so before that time Richard Lerner (at Scripps) collaborating with Dave Rowlands and Fred Brown at Pirbright, showed that animals immunized with synthetic peptides corresponding to an antigenic site in foot-and-mouth disease virus, raised antibodies that neutralized virus, and in several instances were able to show that susceptible animals immunized with the peptides were protected against challenge.
SS You mean using peptides from virus sequences to make antibodies.
JH Yes. So we looked at the polio sequence that had just come out. This was in '81 and Vince(Racaniello) and David (Baltimore) had completed the sequence of the Mahoney strain. Richard Lerner arranged for us to have Richard Houghton (who was then at Scripps) make peptides for us. We were limited by the number that could be made at the time and couldn't do a complete scan of the capsid, so we went through VP1 and picked a handful of peptides. Everybody thought at the time that VP1 was the major antigenic determinant for all these viruses. We (Marie and I) went through the VP1 sequence and picked any real stretch of hydrophilic residues that was punctuated by hydrophobic residues for screening criteria for making these peptides. Then Marie, and simultaneously Jim Bittle's group at Scripps, started making antibodies against these peptides. That could have been very awkward but it turned out that Jim Bittle - who is a really great guy - and Marie decided to pool all the data and write one paper.
SS Let's go back. So you're now at the stage where you have interacted with Marie to get material, but you're still in Steve's lab.
JH I'm still in Steve's lab and Marie and I started talking about long term collaborations: what we would like to do with the structure and the peptide work, antigenic sites, cell entry, assembly. She had pretty broad interests at the time. And it was about then that Richard Lerner (at Scripps Institute in La Jolla) became very interested in establishing structural biology at Scripps, in part as a result of the hemagglutinin structure that was published by Don Wiley, Ian Wilson and John Skehel in '81. He made a very serious effort to recruit Ian Wilson. Richard had been told by his consultants that it was be best to have a critical mass, and that you should not just have one structure person but a group that could talk to each other. Richard asked Ian for some suggestions of other people he could contact, and Ian suggested that Richard might want to talk to me.
SS Ian was coming from Don's (Wiley) lab?
JH Ian was coming out of Don's lab. Ian was already on the job market, I was not yet. I hadn't started seriously looking for a position, and I can't honestly say that I felt I did a great job in my initial interview at Scripps. To my surprise, I got an offer from Richard. It was a joint offer with Ian, Art Olson and I. Art was still in Steve's lab when I went there, but moved to Berkeley about a year later. When the first graphics unit came into Steve's lab, a black and white graphics unit, Art's destiny was clear. He was a bit of an artist himself and graphics was something that thrilled him. He went to work at a unit in Berkeley but that closed down about the time Lerner contacted Ian. So the three of us, who had known each other at Harvard, came to start the structure group at Scripps. We moved into temporary facilities and had just gotten the X-ray generators and cameras working when we moved again into our new building.
SS Did Dave Filman come with you?
JH Dave Filman joined me right after I went to Scripps. This is another interesting story because again it was the generosity of a colleague that made it possible. Mike Oldstone offered a slot on a postdoctoral training grant to me to hire a postdoc. The caveat was that I had to do it pretty quickly, so I was scrambling trying to get all the people I might be able to bring in for interviews to find someone who might be interested in coming. Joe Kraut at UCSD suggested I talk to Dave Filman and Dave came. Dave is in many ways a very impressive person. He is my age, and really impressed me when he came by to talk with me about the position. During the conversation I was trying to explain some of the problems that we were going to be encountering initially with processing the polio data. Data processing was going to more difficult than either TBSV or TCV had been. I said something regarding a specific point about the method and Dave said: "That's just not correct". I said: "O.K, I'm sorry I have been working on this for the last 6 years and you just heard about this an hour ago. What do you mean, it's not right?" He then proceeded to show why it was not right, and he got the job. I don't think Dave has ever hesitated to tell me I'm wrong. In a way it was really great to have a postdoc, my age, with tremendous talents, who was not afraid to voice his own opinion.
SS Was he trained in biology?
JH He actually had been trained in biology. He had worked in an immunology lab. At one point he had briefly been a medical student and decided to pursue a Ph.D. instead. In Joe Kraut's lab he had been a hard core crystallographer working on dihydrofolate reductase and did some really nice work on that enzyme. The other person that came at about the same time was Joe Icenogle, who came from John Andregg and Roland Rueckert's lab. Joe brought in the expertise for tissue culture, for growing virus, and for experimental virology. He was also a real find. He was also about my age, and he was quite independent and a great experimentalist. Dave and Joe were the second and third persons to work for me. The first person was Carl Fricks, who actually preceded me at Scripps. Carl was a graduate student at UCSD looking at Scripps for a place where he might do his work. He heard I was coming, contacted me and actually worked in Marie's lab the summer before I left for Scripps in October of '82. Carl came with me, Dave Filman joined the lab shortly after I arrived, and Joe came early in 1983. Joe trained Carl in tissue culture. Joe was working primarily on assembly and immunology and Carl's experiments evolved into studying cell entry. Neither Joe nor Carl were doing any crystallography. The idea behind that was that what Dave and I were doing was pretty boring. I had already been on the lecture tour: here's the crystal, here's the diffraction picture, we'll have the structure someday. I was in the process of trying to push the data collection, and if we were going to survive as a laboratory we had to have some results coming out. Joe and Carl were doing virology. They were doing protein chemistry and virology. They provided the lab with its first publications, and essentially set the stage for projects that we are working on still today. It was a great opportunity to be able to do that.
SS So before you went to Scripps, you knew you had a diffraction pattern that could be solved.
JH We hadn't yet completely fixed the cooling problem, and we hadn't yet worked out how to process the data. We knew the crystals diffracted and we knew how to grow the crystals reproducibly. Marie was shipping us material for the work with the Mahoney strain. Joe was doing his work mainly with the Sabin strain. Joe worked with the Sabin strain in Roland's lab and we were using that as our primary virus for the studies other than the crystallography. I was taking material from Marie, concentrating it, setting up crystals constantly, mounting a crystal every day, taking as many pictures as I could get from it, some times mounting and getting photographs from two crystals a day (the exposure times were about 12 hours). This was in the days before the "American method" (See Michael Rossmann's oral history). I was manually aligning the crystals; very, very tough work but boring. The only way to get through this was to consider yourself an artist and take pride that you did it well. As Roland said in his section, we could not get the project in on the synchrotron so we were going to be forced to collect the data with laboratory rotating anode sources. Dave was working on the processing programs to get them so we would be able to process the data and it took some time to get the data processing programs together.
Dave was nominally being paid on Mike Oldstone's training grant. Dave was not necessarily a great match for Mike Oldstone's group meetings, but he stipulated that either Dave or I were going to have to attend the group meetings. After a couple of weeks it became clear that it was better if I went to the group meetings, so I started attending Mike's meetings, like one of his postdocs. I learned a tremendous amount and my interests expanded from primarily assembly and structure into pathogenesis and immunology and many other areas. It was a great opportunity for me, but required getting up at some ungodly hour on Monday mornings because Mike's group meetings were held the first thing in the morning on Mondays. For a structural biologist this was horrible.
SS Only topped by the virologists at the University of Zurich who met on Saturday mornings.
JH Structural people, historically when computing was very expensive, worked at night and early mornings were not our forte. Mike's lab has always been very broad in scope. It was tremendously valuable for me to go in and learn about the whole range of viruses that his group was working on including LCMV, mouse hepatitis virus, measles virus, CMV and even Theiler's virus, which we ultimately began working on in my lab. In the meantime, Joe and Carl were doing interesting work on polio assembly and Carl was trying to develop ways to look at conformational alterations using peptide antibodies, chemical modifications, and monoclonal antibodies. Joe was starting a project working on the empty capsids, and on the side pursuing a project trying to unravel a mystery concerning the unusually narrow antigenic specificity of type 3 poliovirus. We were fascinated by the idea that all the monoclonal antibodies against type 3 led to the selection of mutants in a single site in VP1, while panels of monoclonals against type 1 selected mutations that were broadly distributed over VP1, VP2, and VP3. Carl discovered a proteolytic site that was in the middle of the restricted site in VP1 that was present in both the type 1 and type 3 Sabin strains. Joe showed that if you cleaved that site you could raise antibodies to other sites. Joe also demonstrated that most strains of mice, and more importantly, people had a much more broad immune response to type 3 polio.
In the meantime I was trying to play catch up with my background in polio biology. Reading old literature and new papers and talking to people. I started to develop very expensive phone bills talking to Marie and others.
SS This was before e-mail.
JH Well, there was a primitive e-mail called zip mail, but it was nothing like what we call e-mail now. We did have computer contact and Marie was much more computer friendly at the time than she is now. I also started talking to Vincent Racaniello regularly, and about that time I met Phil Minor and added the expense of transatlantic calls.
SS Now we're in the early '80's.
JH We're in the early '80's. All the work is going to be compressed into three years, in less than three years, because I moved to Scripps in October '82 and the structure was effectively solved in late March, early April, '85. So that's the time frame in which all this was going on. The information started coming in on the monoclonal antibody neutralization sites through contact with Phil (Minor), including data that hadn't been published.
SS I want to go back a little bit. When I talked to Steve, we talked about the structures that they solved at 5.5 Ångstroms. Did you also do a structure at that resolution?
JH Yes we did, but the time frame for going from low-resolution to high-resolution was much more compressed, and we did not publish the low-resolution structure. You have to realize that we knew that Michael was working on rhinovirus 14. That was very comfortable until about '84 when we heard about the sequence of rhino 14 and discovered that they were going to be far more similar than had been thought from the physical properties. The two viruses differ greatly in their physical properties (e.g. polio is stable at low pH, whereas rhinovirus is not). Once the rhino sequence came out it was clear that the two viruses were going to be very similar, and it was going to be essential to get the polio structure out as soon as possible.
I had planned a three-pronged approach to solving the polio structure: one was electron microscopy and I had actually done some work with Wah Chiu who was then at Arizona. We wanted to work in ice but we hadn't been able to make any progress in ice. We had done some work with negatively stained samples with B.V. Venkataram Prasad from Wah's lab, but as luck would have it the reconstruction was not completed until after we had solved the structure by other means. The second approach was direct methods. I was talking with Gerard Bricogne about direct methods for determining the structure in the absence of heavy atoms or a model. This was somewhat of a long shot. Although it was clear that the approach might work, given the high noncrystallographic symmetry, it could have taken, and, in retrospect, did take a very long time to figure out how to make it live up to its great promise. The third was the traditional heavy atom method. While I was collecting the native data, I also started screening for potential heavy atom derivatives, and had identified two candidate derivatives, a gold derivative, which turned out to be pretty much useless, and a platinum derivative that worked well. About that time we completed the data collection on the native virus crystals - it's never really complete - but we had all the data, at least an over sampling of the unique data. I started collecting data on the platinum derivative, but was still not sure whether we were going to be able to get the derivative to work. If either the direct methods or the EM had worked, it was my feeling that we were going to start working at very low resolution, at 20-25 Ångstroms and have to phase extend from there. With heavy atoms it wasn't clear whether they would provide phase information at all, and what the resolution limit of that information would be. It turned out that very early on we realized that we were going to be stuck at low or moderate resolution with the platinum derivative because it was not completely isomorphous with the native crystals. In the Mahoney crystals, the particle sits in the cell on a two-fold axis and is free to rotate about the two fold. We have since found out that any significant alteration (including making heavy atom derivatives) results in changes in the rotation of the particle, and therefore a loss of isomorphism. In the platinum derivative crystals the virus actually had rotated slightly, but small rotations are magnified by the large radius of the virus, and the small rotation caused serious nonisomorphism. It became clear that the nonisomorphism limited the utility of the phases, to somewhere between 5 and 8 Ångstroms. We took a very conservative approach, starting at 8 Ångstroms and after a number of cycles of applying noncrystallographic symmetry constraints we got our first map:
it was gorgeous - it was just absolutely gorgeous
When the map first came up, it was completely clear that the structure was solved. It was also immediately clear to us that the subunit structure and even the subunit organization looked an awful like that in TBSV. I had seen TBSV and TCV maps at both low and high resolution and it was very clear that the organization was going to be similar. There were differences. There were some things that looked like they might be tilted a little differently, but it was clear that the subunit organization was going to be very similar. We then pushed the information, initially using some of the heavy atom methods, but then we just dropped the heavy atoms and went straight to the phase extension.
We had obviously heard that Michael had done phase extensions with the rhinovirus structure. In fact Michael had sent us an early draft of their manuscript. I had also talked extensively the year before with Wim Hol at a meeting that I had gone to when I was considering direct methods. Bricogne had had a meeting in Orsay discussing maximum entropy methods for phase determination. Wim Hol was also at the meeting, and Wim and I ate dinner together most of the evenings talking about his experience with phase extension and hemocyanin. He had done a modest phase extension and said that the key was to extend the resolution in very, very narrow steps; essentially extend the resolution by one reciprocal lattice unit at a time. So that meant that the calculations had to be done with many iterations in order to go from low-resolution to high. Although it was clear that the calculations were going to push our available computational resources, we went ahead and started the phase extension. However, it soon became clear that we were at the stage where we had essentially saturated the capacity of our computer (a small VAX 750) and to go with the standard programs that I had used with TBSV was just not going to be feasible with the computers that we had.
By then Dave and I were meeting almost daily in a crisis management situation. We always had worked slightly different shifts. I was then an early morning person (for a crystallographer) because my wife, Doreen, was working and we only had one car. Dave didn't come in until noon and worked late. When Dave would come in we would get together in my office and say: "What was the rate-limiting step today?" We would kick around any ideas we might have how to fix the current bottleneck, and try to flesh out any that seemed promising. Dave would then go off with this list of ideas and come back the next morning with a pad of yellow paper and say: "This isn't going to work, this isn't going to work, this will work, I think, and maybe the next one will work and I think this is the most promising and I have already started to sketch out some algorithms." In an amazingly short time he would have a working modification of the program going and test it and have it in production. Then we would say: "What's the next rate-limiting step", and this was being done literally on a day-to-day basis. In the course of this, Dave came up with several changes in the standard procedures that allowed the calculations to fit into our computer, some of which we and other people are still using to make the computations more tractable.
SS For how long a period was this?
JH Probably the real crisis mode was in February of '85, getting to the maps and pushing through until July. As a matter of fact, I presented the structure at a meeting at a EUROPIC meeting in France in August or September of '85 and by that time I was living on about 4 hours of sleep "max" a night. I don't think I slept at the meeting. We solved the structure and the paper was actually written, submitted and in review at the time. I was in France for that meeting and I was going to come back to France for another meeting and in-between I had to go back to San Diego to check the proofs on the manuscript.
SS So let's talk about the structure now. How did it change your thinking and then what were the responses to it?
JH Well, there were several things that were really stunning in the structure that we tried to talk about in the paper. One was that the theme of interlocking of N-terminal arms of subunits, that had been seen before in TBSV, was expanded on in a gorgeous way in the structure. The entire inner surface of the capsid is decorated by an elaborate interaction of N-terminal extensions of VP1,2 and 3 and by VP4. A couple of things were immediately clear from looking at the network. First of all, the interactions were very important in stabilizing the structure of the virus, but those interactions couldn't be made until the subunits began to assemble together. These were assembly dependent structures that came in, as within the protomer, and were the interactions of bits of VP1 and VP2 that don't make any sense on the outside surface unless they are interacting with each other. All polio proteins are made as a single large precursor that is processed cotranslationally. Cleavage of the capsid protein region of the polyprotein was thought to be linked to assembly. In the structure, the N-termini of all of the proteins are on the inside of the virus and the C-termini of all of the proteins (except VP4) are on the outside. Therefore, it was clear that in the process of cleavage of the polyprotein the termini had to undergo extensive movement in order to form the network. In addition in the uncleaved precursor in which VPO, VP3 and VP1 are all still linked together, the bits that eventually connect the N-termini and C-termini have to be up through the interfaces and interact with each other in the pentamer (the first observed assembly intermediate). This meant that the cleavage in the protomer almost certainly is required for pentamer assembly. Ann Palmenberg has since confirmed this experimentally. From the structure, it was clear that the interfaces that existed in the virus itself couldn't exist in the same form until after cleavage. That was great because it said that proteolysis is not just a convenient way to make a lot of proteins from one gene, it is also a great way to control the timing of assembly.
In subsequent work with Bert Semler, it became clear that the cleavage was also a great way to edit the assembly process. One of Bert's students, Mary Francis Ypma-Wong, showed that proteins have to fold right to get cleaved. They can't assemble until they get cleaved, and misfolded proteins never assemble or enter the assembly pathway. These N-terminal extensions were telling us a lot about the assembly process. Early in assembly, the pentamers had very elaborate interactions, the N-termini of VP3 interacting with each other - that was really exciting to see.
Another striking observation was that the structure immediately provided a three-dimensional context for a growing body of data concerning the antigenic sites. I said earlier that it was clear that the structure and organization of the capsid protein was strikingly similar to that in the plant viruses. However, in the polio capsid proteins the portions of the protein subunits that were exposed on the surface were decorated with more elaborate loops. At the time the structures were solved we already were aware of a great deal of data about residues that were involved in antigentic sites, and these residues were all located in these exposed loops. This was not surprising as Don Wiley and Ian Wilson had previously noted that antigenic sites of the flu hemagglutinin were also all located in highly exposed loops, but the extent of the antigenic surface and the correlation of the three-dimensional organization of residues with clustering determined by cross-neutralization with panels of monoclonal antibodies were stunning and exciting.
The monoclonal escape mutation data came from several sources. Phil Minor and Jeff Almond were doing work with Type 3, and eventually also worked with type 1 and type 2. David Diamond, Emilio Emini and Eckard Wimmer were publishing data on antigenic sites on type 1. In addition, there was some information coming out of Pasteur from Radu Crainic and from Bruno Blondel on antigenic sites in type 1. The clustering of the escape mutants to define antigenic sites was based on cross neutralization data using techniques similar to those now used in genomics data for large data sets generated from gene arrays - called cluster analysis. In genomics, this is done computationally. For the viruses this was being done in much smaller data sets essentially by visual inspection. You set up an array of monoclonal antibodies versus positions of mutations or can sort them out and group them into blocks and the blocks tell you where the antigenic sites are. With the partial information that was available for polio, it was absolutely clear that spatially the sites were organized into four separate units on the surface and since have been designated site 1, site 2, site 3A and site 3B. When we began looking at the mutations in the context of the structure we chose to organize them first by spatial characterization and were immediately convinced that the spatial organization was very, very similar to the organization that was coming out at the time from the cluster analysis of monoclonal antibody escape mutations. Both the structural definition of sites and the definitions from escape mutations were also consistent with many identifications we and others had made based on data with antipeptide antibodies.
However, there were a small number of peptides from the N-terminus of VP1 that generated antibodies that immunoprecipitated, and in some cases neutralized, the virus. These were found to be inside the protein shell in the structure. This was puzzling and caused some to question the validity of the peptide data or the structure or both. We were very confident of both the peptide data and the structure, and were convinced that the "clinkers" were trying to tell us something. We wrote in the Science paper that at physiological temperatures the N-terminus must be reversibly exposed on the outer surface and that exposure might also occur during the infection process. There was not a lot more that we could do with that. We already intended to use them as tools to study conformational rearrangements. We thought, initially, to test relative sensitivity to a panel of antibodies to tell us something about a rearrangement - here is something that you know is on the inside that now is on the outside.
SS How did you deal with the possibility that there were some particles that were rearranged or no longer infectious because they had been distorted?
JH Marie had been pretty careful in the immunoprecipitation and neutralization and showed that you could get nearly complete neutralization of the virus with these antibodies and she could completely immunoprecipitate, provided the antibodies and virus were incubated at 37 degrees. Then we also knew, because Carl Fricks had begun using the antibodies, that they didn't immunoprecipitate at 4 degrees or even at room temperature. I postulated there was something different at 37 degrees than at 4 degrees. Neither one of us ran with it for awhile, but Marie eventually picked it up and showed much later that there is reversible breathing that takes place at 37 degrees in which the N-terminus of VP1 and portions of VP4 were transiently and reversibly exposed.
SS Even in the absence of cells?
JH Even in the absence of cells. The virus does all this on its own. Carl's work suggested that the receptors on the cell simply facilitate the exposure and push it past the point where it is irreversible. Simon Tsang has recently shown conclusively that this is the case. Our vision of what was happening with the antibodies at 37 degrees was that the antibody is a kinetic trap. In other words, the internal bits are coming out and can go back inside the virus, but with the antibody "you catch it every time it comes out and don't let go of it so it can't go back in". To get complete immunoprecipitation doesn't mean that every particle has the site exposed at the same time, it just means that over the time you incubate, the antibody is able to interact with enough of the particles to lead to a solid immunoprecipitation.
SS By this time you probably didn't expect to see any RNA.
JH No, we still had hoped to see RNA. In our initial maps at very low resolution there was some density on the inside of the particle, but it disappeared as the resolution was extended. Now many years later, we are doing some experiments where we have collected all of the data that we can. We have gone to as low resolution as we can, out to almost 300 Ångstroms, that's normally well within the beam stop and also extended the Mahoney data (to 2.2Å). When the very low-resolution terms are included, there was definitely density on the inside. Unfortunately, although the density is at significant levels and looks tantalizingly RNA-like, it is not interpretable as any unique RNA structure.
Then there was one really embarrassing bit of density in the structure. We didn't know what to make of some extra density in the core of VP1. Unfortunately, the electron density maps don't come with labels that tell you what each atom is. At this resolution (2.9Å) it is very difficult to assign density in the absence of corroborating chemical information (such as the known amino acid sequence for the density that belongs to the protein). We didn't say anything about it in the initial Science paper because we were at a very early stage of modeling. As the modeling progressed Dave Filman was able to fit an extended hydrocarbon chain into the density, but at this point it had no biological context. A couple of years later it became clear that the site occupied by the mystery density might play a functional role, when it was shown that a family of drugs that had previously been shown to bind virus and neutralize infectivity bind to the same site. Tom Smith, then in Michael's lab, did the first of what are now many studies of complexes of rhinovirus with drugs from Sterling-Winthrop. (The work with these drugs continues at Viropharma.) In our picture of the density originally published in the Science paper we show a portion of the map that includes the extra density, and a couple of years later when the rhinovirus-drug complexes were solved, Tom and Michael immediately recognized that drugs were utilizing the same site.
SS Where does the extra density in the site in polio come from?
JH We don't know. We still haven't been able to figure out what it is by direct chemical experiments. It models in the Mahoney strain and Sabin strain type 3 as sphingosine, in other poliovirus structures we have solved, it models as shorter chain fatty acids. It's probably a mixture of fatty acids, what Roland calls the pocket factor. It was only after we started working with Phil (Minor) on type 3 that it became very clear that pocket factor itself was actually very important. Andrew Macadam was mapping non-ts revertants of Sabin 3 and found a revertant in this site, and subsequently showed that the mutation affected the stability of the virion.
There were several things that the structure immediately told us. However, as David Baltimore stressed in a review of the structure that appeared in the same issue of Science as our report, the structure left many questions unanswered and raised additional questions we had not thought of previously. One of the unanswered questions that David raised was the role of the VP0 cleavage. This cleavage, which occurs late in assembly and is associated with the encapsidation of RNA, converts a precursor protein VP0 into the mature capsid proteins VP4 and VP2. It wasn't clear what VP0 cleavage did because the termini looked pretty close together. Only when we solved the structure of empty capsids (which still contain VP0), did it become clear that the termini move significantly after VP0 cleavage. In fact, the empty capsid structure told us, that the entire network between VP4 and the N-terminal extensions of VP1, VP2, and VP3 doesn't form until VP0 is cleaved. Another question that David raised is what happens to the virus when it interacts with receptor. Although we've made a great deal of progress since, this question remains a major focus of my lab's work on polio to this date.
SS When you published the structure was there a lot of response from virologists?
JH Oh, tremendous. There was this great body of work that provided a context for the structure, and the structure immediately suggested a wide range of new experiments. There was also a great deal of interest from the press, pretty heady stuff for a person three years out of a postdoc. The press interest died down pretty quickly, but the interest from colleagues continued, especially colleagues in the polio field. When you are working on a structure where that much work has been done and continues to be done, there is so much you can think about, provided you can talk to people who are doing related work. In that sense all the attention was wonderful and ultimately led to some long term collaborations and friendships. Obviously, David Baltimore was tremendously generous early when I started as were Geof Schild and Marc Girard. But the real close contacts for me have been with Marie Chow, Vincent Racaniello and Karla Kirkegaard (who were with David when I first met them); with Bert Semler (who was with Eckard Wimmer), with Phil Minor and Jeff Almond (who were with Geof Schild), and with some of the people in Marc Girard's lab including Sylvie van der Werf, Kathy Kean, Annette Martin, Bruno Blondel. To be able to talk to those people was tremendous. It was not only tremendously educational for me, the discussions also often led to experiments. As a result, for a long time, my lab didn't do molecular biology. "I don't do molecular biology, you don't do structure, and we can talk." Although it made for some very large phone bills, I could talk to these people on a regular basis and see them at meetings, including the semi annual EUROPIC meetings sponsored by the World Health Organization.
SS This was before the eradication of polio program began?
JH This was before the eradication program. I was on a committee at WHO, with a number of other people including Phil Minor, Vince Racaniello, and Stan Lemon that was reviewing developments of improved vaccines for polio and a vaccine for hepatitis A. It was a great chance to extend contacts in the community. It was essentially where I met Phil and that's when we started working on Sabin 3 together. That was a lot of fun. My first exposure to the community as a whole went back to 1985, to the meeting of the European picornavirus meeting in Seillac, France. I was not initially invited to the meeting. I invited myself. I called and told them I had the structure, and "would they make room for me on the program?" and they said yes. I did know some of the people, but there were a lot of people in the field I had not met until then.
SS Once you had the structure, where did you think you wanted to go next?
JH That's a good question, and at the time even a scary one. Dave asked me right after the structure came out "What are we going to do next that will have similar impact". I answered that one really has to be grateful for the chance to do one thing of very high impact in a career and that we'd been very lucky. However, it's a bit disconcerting to finish what you thought would be a career defining project before you turn 35. It was clear, however, that we needed to think seriously about future directions. I really enjoyed working with viruses and one approach that I considered was to continue looking at new virus structures. There are viruses that I would really love to work on, and there are still some structures I would like to be involved with if I can be in the future. One of the ones I want to get involved in at the time, but was already taken, was reovirus. I had talked to Bernie (Fields) a lot about reo and thought that would be great but Bernie was working with Steve (Harrison) at the time and it was pointless to duplicate efforts and that project was in good hands.
Instead of pursuing more viruses I decided to work somewhat on a vertical level with polio and its close relatives. I knew I wanted to do assembly intermediates, cell entry and do something on the receptor.
SS In '85, the receptor hadn't been discovered yet.
JH No, but it was shortly thereafter, and we knew Vince was on track. I decided to stick with polio. I was interested in pathogenesis and type 3 provided a great chance for that. We had hoped that Theiler's would be a very closely related virus to follow up on because it was clear that there were determinants on the capsid that would affect whether the virus was going to be encephalitic and rapidly kill mice or move to the gray matter and cause demyelinating disease. I think the structure got ahead of the biology on that problem. We started working on Theiler's, but the main thrust was going to be assembly and cell entry with polio and that's where we are still. I made the commitment to do some of the biology in my own lab and that's been a lot of fun over the years. I have had a series of very gifted postdocs, generally senior postdocs, starting with Joe Icenogle, and Orella Flore, Alicia Gomez Yafal, and Stephen Curry. Both Ornella and Alicia had their own laboratories and came to my lab as senior postdocs. Stephen Curry came on a Wellcome Trust Fellowship and was here for 2 maybe 3 years, made a tremendous contribution understanding cell entry intermediates, including demonstrating that one potential intermediate known as the 135S particle, or sometimes the A particle, was infectious.
These are all things that were beginning to be built on at the time. The first for us was the assembly intermediates, moving on to the empty capsids. In fact we solved that structure before when I was at Scripps, but I was very uncomfortable with the quality of the data and we essentially solved it again. Here a critical contribution came from a postdoc, Ravi Basavappa, who joined the group right after we moved to Harvard. He developed a better procedure for purifying the empty capsids that led to much better reproducibility of the crystals and much, much better data.
SS When did you leave Scripps?
JH I left Scripps in 1991 and came to Harvard Medical School in July of 1991- 9 years at Scripps and now almost 11 years here. There were some other things that started coming up as well. We were interested in the fact that we had reagents that would tell us if the virus could change conformation. Carl Fricks started working on that, using proteolysis and synthetic peptide antibodies, and showed that the N-terminus of VP1 was exposed when the virus was converted to the 135 S form. That was essentially the central piece of Carl's thesis work. He showed that the exposed N-terminus allowed the virus to attach to liposomes and that if you cleaved off the N terminus it no longer attached. If you attached and then cleaved , the N-terminus went with the liposomes and the rest of the virus stayed at the bottom of the tube in a flotation experiment. Carl was working on that, Joe was working on the assembly intermediates. We were busy trying to refine the structure and beginning our work on Sabin 3.
SS I don't actually know how different the structures of Sabin and Mahoney are.
JH They are very similar. But the development of our work with Sabin 3 is really a fascinating story. Phil and Jeff Almond, initially in Geof Schild's lab, and then Phil took it over, looked at the determinants of attenuation in the Sabin 3 strain and showed that there were 2 important determinants of attenuation: one was a mutation in the 5' end (which turned out to be one of the key points in the discovery of the IRES - the internal ribosome entry site). This mutation was shown to be a key attenuating mutation, and a second one that played a secondary role was in VP3. So we had a capsid mutation in attenuation and that made it exciting. It turned out that it is also a mutation that makes the virus temperature sensitive, temperature sensitive for growth. Andrew Macadam in Phil's lab showed that it was temperature-sensitive and that the defect was probably in assembly. In fact, it turned out that the mutation also altered the stability of the virus itself. Andrew did a number of genetic experiments, and I think the most informative ones were the revertants, second site revertants and that information was becoming available.
SS Were there second site mutations in any of the other proteins?
JH Yes, and some of the mutations are quite remote from the original mutation. Some are in the N-terminal extensions in VP1, and one of the mutations was within the drug-binding site.
SS Even remote in the structure?
JH Even in the structure. The primary mutation destabilizes the virus capsid at some point and you make the secondary mutation that now gives extra stability.
SS If you only have the secondary mutation, how does that change the virus?
JH They were often more stable, but it was frequently difficult to make the revertants on their own. When Andrew selected revertants, he often got multiple mutations. One of the interesting things that came up was that if you move some of the mutations back into the virus, some of them actually stabilized the virus too much. One of the mutants was actually cold sensitive. Other single mutations actually destabilized the virus even more. What you have is a window of stability and probably a window of optimal growth that moves a little bit, but if you make it too stable then you have a virus that can't undergo conformational changes at physiological temperatures and it can't get into the cell. The whole idea is that the virus is going to have to change its structure to get into the cell and again, some of these mutations were in the N-terminal extension of the capsid protein. We reasoned that these regions were important determinants in stability and probably move when the virus is undergoing conformational change.
We knew that Sabin 3 and Mahoney were about 83% identical in the capsid protein region. Now the question is what is the effect of that, and it turns out that there are real changes in the antigenic sites; substantial changes in the antigenic sites, but that's about it. The rest of the structure is virtually identical. But there are substantial changes in the antigenic sites and that was the first time that anybody had looked at what the conformational differences were between two very, very closely related serotypes.
Although our earlier work on the Sabin strain of type 1 was hampered by extreme x-ray sensitivity of the crystals, the crystals of the Sabin strain of type 3 poliovirus were really quite robust, and diffracted even better than our original crystals of the Mahoney strain of type 1. We were able to solve the Sabin 3 structure to about 2.6Å. We determined where the second site mutations mapped and they mapped to the interfaces between subunits and to the N-terminal extensions. I am focusing on these N-terminal extensions and these interfaces because that was the key that led us to come up with our models for what the virus might be doing when it undergoes conformational rearrangements. We were sampling things that were regulating, within allowable limits, the mechanics of the virus. All that information was trying to tell us what the mechanics were. We tried to synthesize that information and tried to come up with ideas for what might be going on during assembly and especially during cell entry.
It was also about that time that we began to think seriously about the bit of extra density (corresponding to the drug binding site) and were puzzled by what it was. The density, the extra density, was clearly in P3/Sabin and looked very similar to what we saw in Mahoney. We thought that it might be important, but in the absence of a chemical determination, we had no way of saying anything other than that it looked like a long chain hydrocarbon with a head group similar to that of sphingosine. Marie might tell you something different, but I thought that the experiment that Marie originally set out to do was to try to determine what was in the site. We figured it looked like sphingosine. A precursor for all these long chain fatty acids is palmitate. We were going to label with radioactive palmitate and see if we would get label in the site and radiolabel the virus. The control was myristate. And the control had the signal and the experimental sample did not!
Even at that time Marie was familiar with very recent literature describing N-terminal myristoylation of proteins. She immediately thought of the possibility that the myristate label in the experiment suggested N-terminal myristoylation, and went on to demonstrate that the myristate group was covalently attached to VP4 via an amide linkage. Then we tied up with Fred Brown who was very interested in this and asked if it was O.K. with us if he went and looked in foot-and-mouth disease virus and a couple of other viruses and extended our work. We pooled our data and published a paper on that. We were in the process of refining the Mahoney structure and we had a major quantum leap in the power of our computing and we could do some things we hadn't been able to do before. Some regions of density that hadn't been particularly clear in the past now became very clear and one of them was that there was something off the N-terminus of VP4. When Dave heard about Marie's myristoylation results, he said to me: "You better come look at this extra density that I have been working on in the new maps. The density extends off the N-terminus of VP4. It doesn't look like protein. It can't be protein because A) there are no more residues in VP4 to account for the density, and B) it just doesn't look like protein. But I do have something that looks like a carbonyl oxygen". Sure enough once we had an idea what to look for the extra density modeled very nicely as an N-terminal myristate modification of VP4.
SS Are these from the synchrotron now?
JH No, we have never collected any data from the synchrotron on polio. It was tough to get permission to go in.
SS Oh, because it was polio.
JH Because of polio. At one time, very early (say 1983) I had one allocation at Lure (the synchrotron in Orsay, France) where they apparently approved a proposal. They were willing to give me beam time if I was able to show up on a week's notice. But then came the big question of what about vaccinating and I said that it should be sufficient to vaccinate the people that are in the immediate vicinity on the floor. They said no, they were going to have to vaccinate everyone. I told them that unless they get a complete history on the people we would have to go with the killed vaccine since I couldn't recommend that adults without a history of vaccination be given the live vaccine, so that fell through. Since then I've received permission to bring polio crystals to CHESS. They have a beam line with a BL2+ facility at Chess, but it's been hard to get polio to the synchrotron.
There are certain advantages to being able to collect the data at home every day. You don't have to make appointments. It's a little harder and it takes longer. Now that you can work with frozen crystals even that is less of a problem. We can collect full-time if we need to and with the frozen crystals, we can bang out a data set in about 10 days, real time. It would take several hours at the synchrotron, it might take us a couple of months to get in. Obviously we would probably be able to collect a little higher resolution data at the synchrotron, but by spending longer collecting data at home we could get very high quality data, and I am not completely convinced that we would learn anything more by a modest increase in resolution. In fact we are defining structure at very high resolution (2.2Å for Mahoney crystals) again using data collected in the lab, but I think the real interesting information was perfectly available at 2.9 Ångstroms from the data we had.
SS But do you see the myristate at 2.9 Ångstroms?
JH We saw the myristate at 2.9 Ångstroms, but not in the maps reported in the original Science manuscript. As I said the particle in the Mahoney crystals sits in a 2-fold axis and can rotate around the 2-fold. We had determined an orientation for the particle and it was just about correct. When we got enough computing to be able to do a full refinement in that orientation, that sharpened things up. It was just a very, very small error (much smaller than the rotation in the platinum derivative) but fixing the small error sharpened things up enough to see a couple of things that were unclear in the original maps, including the myristate. In the Sabin 3 crystals the particle sits in a special position so its position and orientation is known exactly. We were able to see the myristate in the Sabin 3 maps immediately.
SS Let's continue with how the structure influenced your thinking about the biology of the virus, and particularly how the introduction of recombinant DNA techniques affected the directions of your research.
JH Yes, Vince (Racaniello) and David (Baltimore) had shown that the cDNA of polio was infectious and it was obvious that there were things that you were going to be able to do. The structure posed a number of questions, and the ability to design and make mutants was an exciting prospect. However, making mutations in the capsid proteins can be very difficult. Many are simply nonviable, others have pleotrophic phenotypes. For example,. Marie found the myristoylation. We saw where it was in the structure. Even without the structure, an obvious question was: what does the myristate do? Marie generated some mutants in which the N-terminal glycine was modified (glycine is required for myristoylation). She took a look to see if it was necessary for localizing to the membrane during assembly and it turned out not to be. We couldn't do the experiment to find out if it was necessary for cell entry because it wasn't possible to assemble virus. It would start to try to assemble but it could never complete assembly if myristate wasn't present or even if it was underrepresented (e.g. by doing a mixed infection). All the infectious virus that was made apparently had a full complement of myristate.
SS It wasn't possible to make particles or to make infectious particles?
JH Some of the assembly intermediates would be made, but not infectious particles that didn't have a full complement of myristate. But then if you take a look at the structure there are immediate suggestions of what mutations should be made and Marie made some of the mutations. Some of those data suggest that VP4 is important in cell entry. Some of her mutations began to suggest that, but every time Marie asked a question and then designed a mutant, she ended up with pleotropic phenotypes and, by having to figure out what was going on, we ended up learning much more than if we had simply confirmed what we'd predicted. Although it was a lot more work, it was much more informative. If you are able to predict the properties of a mutant from a structure, you have nothing more to learn and I guess at that point you can be very satisfied and quit. I still find that although I am able to rationalize many things with the structure, I am still able to predict only a few. Rationalization is a wonderful thought process, but rationalization just says you can explain a result that you already know and I think you really don't understand something fully until you can make predictions and then confirm them by subsequent experiments.
SS Most likely it's more interesting when you make a prediction and it doesn't come out the way you expect.
JH I think David (Baltimore) actually said that about site-directed mutagenesis. He said in many ways he thought it was a bit of a step backward and he would prefer to do random mutagenesis because you learn more when you had a mutation with an unexpected phenotype than you did when you made a mutation and got the expected result. It's nice to be able to go back and confirm them but surprisingly there are very few of these mutations where we are really able to predict all the phenotypes. Sometimes we get the one we expect, but more often than not it is either masked or in addition to several others.
Despite the structure and despite all of the mutants that have been studied there is still so much that we do not understand about the virus. We are just now beginning to scratch the surface of understanding how the virus gets into the cell, much less details of its pathogenesis. Even in the area of assembly, there are great gaping holes in our understanding. There's so much that we don't understand about packaging: does the RNA get packaged by empties? Does the RNA get packaged by pentamers? We don't even know that yet. There are arguments on both sides. In fact recently there have been mutants selected with a drug from Lilly that seem to cause a partial packaging defect, that suggest that there may be additional intermediates in packaging.
SS What are they?
JH These are mutants that Marie had worked on with Beverly Heinz at Lilly. The drug was originally thought to be a replication RNA-synthesis inhibitor, but it turns out that RNA synthesis is normal but packaging is not. That's fascinating: where does the resistance mutation map? It maps to 2C, a protein that does not yet have a well defined function. Guanidine mutations map to 2C. Vince's cold-adaptive mutants map to 2C. Jing-po Li, in David Baltimore's lab, had a mutation that was a reversion that mapped to 2C that had an entry defect. Now that's strange because everyone says that 2C is not in the particle. So either this is a mutation acting in cis (at the RNA level) or 2C might actually play a role in infection. 2C is a miserable protein to work with. No one has ever been able to express it at high levels in cells without making the cells desperately ill. It may be involved in some way with the reorganization of lipids in addition to all the other things that the virus does to the cell. Polio disrupts all the organized membrane organelles and turns them into these big vacuoles that's where assembly and RNA replication take place, on the surface of those membranes.
SS The last thing I wanted to ask you about is the issue of the eradication of polio.
JH This is a difficult question for people who work on poliovirus. When you have vested interests you have to think very carefully about motivations, because clearly if eradication succeeds (and all of us genuinely hope it does) we will no longer be able to work on the virus. In terms of my own work, I think basically it has given me a real sense of focus because I think there are still real unanswered questions that require the virus and that we need to answer in the years we have left to work with the virus before it becomes, either for one reason or another, impossible. Even when eradication takes place, I don't know when they will stop vaccinating. We will be able to work until they stop vaccinating, unless NIH gives up on funding or unless the universities decide it is not an insurable risk. Either one of those two things could close us down before the law does.
We are moving somewhat into areas that are not depending directly on the virus. We are working with other picornaviruses. We've done some work on echoviruses, and we are working on nonstructural proteins trying to figure out other aspects of the life cycle that you can continue working with in the absence of making capsid protein. That's not something that's going to be shut down. But the wealth of genetic information about this virus exceeds any of the other members in the family and it's just a shame to lose the tools and the mutational and genetic data that are there. So we've been trying to push those, but there is the overall realization that - being able to work on the virus, our time is somewhat limited.
In terms of policy this is where it gets delicate. Essentially in discussion of the decisions about eradication , one is being asked to come in and give an opinion: should we stop vaccinating? And here we have issues where sound science and public health policy may be at odds.
I like and respect Olen Kew a lot. I think Olen has made a very courageous choice.(Olen who is at the Center for Disease Control is one of the leaders in the eradication program.) He has moved a very large component of his work into the service sector and Mark Pallansch (also at Center for Disease Control) has done the same thing. Mark was the person from Roland's lab who actually supplied us with polio initially. I think there are some serious questions that come up and I think they do too. I realize the necessity for stopping vaccination. You cannot convince countries to spend what amounts to a substantial amount of money indefinitely in these crisis-style management polio eradication days (national immunization days) where you vaccinate everybody under the age of 5. You just can't do that forever, and so there needs to be an end to it, but, it's scary! You are doing an experiment on 6 billion people. In voicing my concerns, I don't think I am being motivated by a selfish concern for loss of one of the centers of my research career. There are plenty of other things I can do with my life. I honestly think that the time when the ability to work on polio reaches a crisis is going to be close enough to other decisions in my life that it doesn't matter. I will be old enough to have lots of choices at that point. But we have never had a population that's not immune to polio, a large population. There's always been a substantial portion of the world's population that have been immunized.
SS And you mean even before the vaccine.
JH Yes, before the vaccine essentially, everybody was exposed, and most children became exposed while they were still protected by maternal antibodies. It was only when we separated people from sewage that substantial pockets of people didn't see the virus until they were young children or young adults. That was at the turn of the century (the twentieth century) when the disease switched from an endemic to an epidemic disease. Basically the world population has always been immune to polio with sporadic cases of poliomyelitis in the background. If you eradicate the wild type virus and stop vaccinating, you will now have a population that gradually loses its immunity and ultimately a population that never had immunity and there will be a huge population of people all of whom are sampling huge numbers of enteroviruses, only some of which we know about. Will one of them fill that niche? In other words, will one of them find that it can undergo changes that will allow it to make use of the CD155 poliovirus receptor and have other properties that cause paralytic disease?
SS Some of them already have been able to cause paralytic diseases, but on a small scale.
JH We are talking about 6 billion individuals and untold numbers of different enteroviruses out in the world. Another issue that arises is that once you stop vaccinating you will need to eliminate or at least control all sources of poliovirus. How are you going to get rid of all these viruses, including the ones in the laboratories? Smallpox was fairly easy. There was a very small community working on smallpox.
SS I would say that half of the virologists in the world may have polio in their freezers.
JH Exactly, and most of us have moved stocks from place to place. I think I moved all of my virus out of Scripps, but I can't guarantee it. I think I moved all the virus that temporarily went to MIT when we moved from Scripps to Harvard, or that Marie moved it when she left to go to Arkansas. But that's easy to track. Beyond known stocks of virus, there are all those stool samples that were collected in nutrition studies in the days prior to vaccination or even in the days since vaccination. Those would be sources of the vaccine strains and maybe wild type virus.
It's a tough problem and I am very sympathetic. Part of my work with WHO was in the early stage of the eradication effort and I've always said that I can't imagine a better thing to contribute to even in a minor way. But basically how does a structural biologist contribute to the eradication of polio? Maybe some structural information helps, maybe just being enthusiastic helps. I think right now as we approach it, it is something that is a bit sobering given the stakes that are at hand.
SS And almost impossible to stop, I think.
JH It has to go on. There are countries that just simply can not afford to continue vaccinating.
SS Right, and I think it would be hard to tell them after this that now we have to keep vaccinating.
JH Despite the fact that it is incredibly cheap. The Sabin vaccine costs a couple of pennies a dose to make. It costs about a dollar to deliver and part of that is the cold chain. That's incredibly cheap.
SS But the idea was to stop the Sabin vaccine.
JH And convert to the killed vaccine prior to stopping vaccinating altogether. That has its own problems in that the manufacture of the killed vaccine can be a problem. There was a wonderful effort going on at Lederle that would have been very nice if it had been continued, but it was killed. It was to use the Sabin strain as a stock for the killed vaccine. The approved source of the killed vaccine are wild type neurovirulent strains. You would be working with a very virulent virus and now you would not only have to manufacture with GMP (good manufacturing practice) but also manufacture in level 4 facilities. It is more expensive to produce and deliver and it has to be injected. It will be even more expensive when containment has to be boosted up to a much higher level when the places in the world that are producing it have not been vaccinating for many years. The manufacturing process itself will also ultimately pose a risk. There's at least one case of a floor worker in a vaccine plant bringing the virus home.
SS I think this is the Netherlands case.
JH These are sobering times from a public health point of view. Vaccination has always been a public health issue because any vaccine has side effects and what you are hoping to do is make more people well by several orders of magnitude than you make people sick. Once the wild type virus is gone (or even very rare) you risk harming more people than you help by continuing vaccination. Clearly from a public health perspective you must stop vaccinating. But if you ask how do I feel as a scientist, I need to consider what possible outcomes there could be. From what I know about RNA viruses, I think we need to be cautious and make sure we have in place ideas of what we would do if we saw a re-emergence.
SS It would probably be hard to convince people to have stocks of virus available for vaccination.
JH It's not even clear that the limited stocks of the current vaccine would be useful. Let's suppose that at some time in the future poliomyelitis re-emerges. In the worst case scenario the virus that emerges is a non-polio enterovirus that has switched receptors and now uses CD155. It may not even be cross reactive with existing vaccine strains. Even if the virus that causes the disease is serotypically related to current vaccine stock virus and you catch the outbreak early, you will have a very large number of carriers all of whom are excreting large amounts of virus. Because so few people who get infections have neurological symptoms, catching an outbreak early means when you have 100 cases you probably have at least 10,000 infections, and with a mobile population, the chance for spread are so great. By the time you recognize the problem it may be too late, and you will have to start wide scale vaccination again. Some of my colleagues take the stance that we can't stop vaccinating. I realize that we may have to, but find the prospect frightening.
SS Olen says polio is one of the most highly mutable of all the RNA viruses. The frequency of mutation is greater than for flu and or for any other RNA virus.
JH All RNA viruses are highly mutable. I know that an upper estimate on the fidelity is based on physical principles confirmed by experiments with an RNA polymerase and misincorporations are 1 in every 104 incorporations. I don't think it's an accident that the largest RNA virus genomes are on the order of 104 nucleotides. If they were much bigger there would be too many mistakes in replication, but it's about 1 per round. That makes it really able to sample lots and lots of changes. It's amazing-how many times it has been possible to rescue a dead mutant by multiple blind passages.
| 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 |