What Mad Pursuit

by Francis Crick

  One of the oddities of the whole episode is that neither Jim nor I were officially working on DNA at all. I was trying to write a thesis on the X-ray diffraction of polypeptides and proteins, while Jim had ostensibly come to Cambridge to help John Kendrew crystallize myoglobin. As a friend of Maurice Wilkins I had learned a lot about their work on DNA—which was officially recognized—while Jim had become intrigued by the diffraction problem after hearing Maurice talk in Naples.

  People often ask how long Jim and I worked on DNA. This rather depends on what one means by work. Over a period of almost two years we often discussed the problem, either in the laboratory or in our daily lunchtime walk around the Backs (the college gardens that border the river) or at home, since Jim occasionally dropped in near dinnertime, with a hungry look in his eye. Sometimes, when the summer weather was particularly tempting, we would take the afternoon off and punt up the river toward Grantchester. We both believed that DNA was important though I don’t think we realized just how important it would turn out to be. Originally my view was that solving the X-ray diffraction patterns of the DNA fibers was a job for Maurice and Rosalind and their colleagues at King’s College, London, but as time went on both Jim and I became impatient with their slow progress and their pedestrian methods. The coolness between Rosalind and Maurice did not help matters.

  The main difference of approach was that Jim and I had an intimate knowledge of the way the α helix was discovered. We appreciated what a strong set of constraints the known interatomic distances and angles provided and how postulating that the structure was a regular helix reduced the number of free parameters drastically. The King’s workers were reluctant to be converted to such an approach. Rosalind, in particular, wanted to use her experimental data as fully as possible. I think she thought that to guess the structure by trying various models, using a minimum of experimental facts, was too flashy.

  People have discussed the handicap that Rosalind suffered in being both a scientist and a woman. Undoubtedly there were irritating restrictions—she was not allowed to have coffee in one of the faculty rooms reserved for men only—but these were mainly trivial, or so it seemed to me at the time. As far as I could see her colleagues treated men and women scientists alike. There were other women in Randall’s group—Pauline Cowan (now Harrison), for example—and moreover, their scientific advisor was Honor B. Fell, a distinguished tissue culturist. The only opposition I ever heard about was that of Rosalind’s family. She came from a solid banking family who felt that a nice Jewish girl should get married and have babies, rather than devote her life to scientific research, but even they did not provide really active opposition to her choice of a career.

  But, in spite of her freedom to pursue research as she wished, I think there were more subtle handicaps. Part of the problem Rosalind had with Maurice was her suspicions that he really wanted her as an assistant rather than as an independent worker. Rosalind did not herself choose to work on DNA because she thought it to be biologically important. When John Randall first offered her a job, it was so that she could study the X-ray diffraction of proteins in solution. Rosalind’s previous work on the X-ray diffraction of coal was well suited as an introduction to such a study. Then Randall changed his mind and suggested that, as the DNA fiber work (which Maurice had been doing) had become interesting, it might be better if she worked on that. I doubt if Rosalind knew very much about DNA before Randall suggested that she work on it.

  Feminists have sometimes tried to make out that Rosalind was an early martyr to their cause, but I do not believe the facts support this interpretation. Aaron Klug, who knew Rosalind well, once remarked to me, with reference to a book by a feminist, that “Rosalind would have hated it.” I don’t think Rosalind saw herself as a crusader or a pioneer. I think she just wanted to be treated as a serious scientist.

  In any event, Rosalind’s experimental work was first class. It is difficult to see how it could be bettered. She was less at home, however, in the detailed interpretation of the X-ray photographs. Everything she did was sound enough—almost too sound. She lacked Pauling’s panache. And I believe that one reason for this, apart from the marked difference in temperament, was because she felt that a woman must show herself to be fully professional. Jim had no such anxieties about his abilities. He just wanted the answer, and whether he got it by sound methods or flashy ones did not bother him one bit. All he wanted was to get it as quickly as possible. People have argued that this was because we were over-competitive, but the facts hardly support this. In our enthusiasm for the model-building approach we not only lectured Maurice on how to go about it but even lent him our jigs for making the necessary parts of the model. In some ways I can see that we behaved insufferably (they never did use our jigs), but it was not all due to competitiveness. It was because we passionately wanted to know the details of the structure.

  This, then, was a powerful force in our favor. I believe there were at least two others. Neither Jim nor I felt any external pressure to get on with the problem. This meant that we could approach it intensively for a period and then leave it alone for a bit. Our other advantage was that we had evolved unstated but fruitful methods of collaboration, something that was quite missing in the London group. If either of us suggested a new idea the other, while taking it seriously, would attempt to demolish it in a candid but nonhostile manner. This turned out to be quite crucial.

  In solving scientific problems of this type, it is almost impossible to avoid falling into error. I’ve already listed some of my mistaken ideas. Now, to obtain the correct solution of a problem, unless it is transparently easy, usually requires a sequence of logical steps. If one of these is a mistake, the answer is often hidden, since the error usually puts one on completely the wrong track. It is therefore extremely important not to be trapped by one’s mistaken ideas. The advantage of intellectual collaboration is that it helps jolt one out of false assumptions. A typical example was Jim’s initial insistence that the phosphates must be on the inside of the structure. His argument was that the long basic amino acids of the histones and protamines (proteins associated with DNA) could then reach into the structure to contact the acidic phosphate groups. I argued at length that this was a very feeble reason and that we should ignore it. “Why not,” I said to Jim one evening, “build models with the phosphates on the outside?” “Because,” he said, “that would be too easy” (meaning that there were too many models he could build in this way). “Then why not try it?” I said, as Jim went up the steps into the night. Meaning that so far we had not been able to build even one satisfactory model, so that even one acceptable model would be an advance, even if it turned out not to be unique.

  This argument had the important effect of directing our attention to the bases. While the phosphates were inside the structure, with the bases on the outside, we could afford to ignore the shape and position of the bases. As soon as we wanted to put them inside, we were forced to look at them more closely. I was amused to discover, when we finally built the bases to scale, that they differed in size from my previous mental picture of them—they were distinctly bigger—though their shape was close to the pictures in my mind.

  There is thus no straightforward answer to the question of how long it took us. We had one intensive period of model building toward the end of 1951 but after that I myself was forbidden, for a period, to do anything further, as I was still a graduate student. For a week or so in the summer of 1952 I had experimented to see if I could find evidence for bases pairing in solution, but the necessity of working on my thesis made me abandon this approach too soon. The final attack, including the measurements of our model’s coordinates, only took a few weeks. Hardly more than a month or so after that our papers appeared in Nature. It seems a ridiculously short period of work but all the hours and hours of reading and discussion that led up to the final model really should be included.

  It soon transpired that our model was not even correct in detail. We had only two hydrogen bonds in o
ur G=C pair, though we recognized that there might be three. Pauling subsequently made a decisive argument for three and was rather cross when the illustration in my Scientific American article showed only two. This, as it happened, was not really my fault, as the editor was in such a hurry (as is usually the case) that I never saw the proofs of the diagrams. We had also put the bases too far from the axis of the structure, but these errors did not alter the fact that our model captured all the essential aspects of the double helix. The two helical chains, running antiparallel, a feature I had deduced from Rosalind’s own data; the backbone on the outside, with the bases stacked on the inside; and, above all, the key feature of the structure, the specific pairing of the bases.

  Certain points are sometimes overlooked. It took courage (or rashness, according to your point of view) and a degree of technical expertise to put firmly to one side the difficult problem of unwinding the double helix and to reject a side-by-side structure. Such a model was suggested by the cosmologist George Gamow not long after ours was published, and it has been suggested again more recently by two other groups of authors. Let me skip forward in time to discuss these two models. In both of them the two DNA chains were not intertwined, as in ours, but lay side by side. This, they argued, would make it easier for the chains to separate during replication. Each chain did a sort of shimmy so that, at a first quick glance, the proposed configurations didn’t look unlike our own. They claimed that these new models fit the X-ray data at least as well as ours did, if not better.

  I didn’t believe a word of this. I doubted very much the claims about the diffraction pattern, since such models would be expected to produce at least a few spots in those characteristic empty spaces in the X-ray fiber diagrams that a true helix produces. Moreover the models were ugly in that the shapes they took were forced on them by the model builders and seemed to exist for no obvious structural reason.

  Such arguments, however, are not decisive and could easily be attributed to mere prejudice on my part. The two groups of innovators felt rather acutely that they were on the fringes of the scientific world. They feared that The Establishment would not listen to them. Quite the contrary was the case since everyone, including the editor of Nature, was bending over backward to give them a fair hearing.

  At about this time Bill Pohl, a pure mathematician, got into the act. He pointed out, quite correctly, that unless something very special happened, the most likely result of replicating a piece of circular DNA would be two interlocked daughter circles rather than two separate ones. From this he deduced that the DNA chains could not be intertwined, as we had suggested, but had to lie side by side.

  I corresponded at some length with him as well as talking to him on the phone. Later on he paid me a visit. He had become very well informed about experimental details and persisted strongly in his view. I told him in a letter that if nature did occasionally produce two interlinked circles, a special mechanism would have been evolved to unlink them. I believe he thought this an outrageous example of special pleading and was not at all convinced by it. It turned out, some years later, that this is exactly what does happen. Nick Cozzarelli and his co-workers showed that a special enzyme, called topoisomerase II, can cut both strands of a piece of DNA, pass another piece of DNA between the two ends, and then join the broken ends together again. It can thus unlink two linked DNA circles, and can even, at high enough concentrations of DNA, produce linked circles from separate ones.

  Fortunately some brilliant work by Walter Keller and by Jim Wang on the “linking number” of circular DNA molecules proved that all these side-by-side models must be wrong. The two DNA chains in circular DNA were shown to wind around each other about the number of times our model predicted. I had spent so much time on this problem that in 1979 Jim Wang, Bill Bauer, and I wrote a review article “Is DNA Really a Double Helix?” setting out all the relevant arguments in some detail.

  I doubt if even this, by itself, would convince a hardened skeptic, though at about that time Bill Pohl threw in the towel. Fortunately there was a new development. The reason a decisive argument could not be made from the previous X-ray data alone was partly that the X-ray photos didn’t contain enough information and also because one had to assume a tentative model and then test it against the rather sparse data.

  By the late 1970s the chemists had found an efficient way of synthesizing reasonable amounts of short stretches of DNA with any required base sequence. With luck, such a short stretch could be crystallized. Its structure could then be determined by X-ray diffraction, using unambiguous methods such as the isomorphous replacement method, which involved no prior assumptions about the result. Moreover the X-ray spots from such crystals extended to a much higher resolution than the old fiber diagrams did, partly because the fiber was produced from DNA that had all sorts of different sequences mixed together. Not surprisingly, fibers gave a more blurred picture of the molecule, since what the X rays see is the average structure of all the molecules.

  The first result (around 1980) on these small bits of DNA, by Alex Rich and his group at M.I.T. and also by Dick Dickerson and his colleagues at Cal Tech, produced another surprise. The X rays showed a left-handed structure, never seen before, with a zigzag appearance. It was christened Z-DNA. Its X-ray pattern was quite unlike the classical DNA patterns, so that it was clearly a new form of DNA. It turns out that Z-DNA forms most easily only with a special type of base sequence (alternating purines and pyrimidines). Exactly what nature uses Z-DNA for is still a hot topic of research; it may well be used in control sequences.

  More ordinary DNA sequences were soon crystallized. This time the resulting structures looked very like those predicted by the X-ray fiber data, though there were small modifications and the helix varied somewhat depending on the local sequence of the bases. This also is still being actively studied.

  The double-helical structure of DNA was thus finally confirmed only in the early 1980s. It took over twenty-five years for our model of DNA to go from being only rather plausible, to being very plausible (as a result of the detailed work on DNA fibers), and from there to being virtually certainly correct. Even then it was correct only in outline, not in precise detail. Of course the fact that base sequences were complementary (the key to its function) and that the two chains run in opposite directions was firmly established somewhat earlier by the chemical and biochemical work on DNA sequences.

  The establishment of the double helix could serve as a useful case history, showing one example of the complicated way theories become “fact.” I suspect that after about twenty to twenty-five years many human beings have a desire to overturn the old orthodoxy. Each generation needs a new music. In the case of the double helix, the hard bite of scientific facts made the new models unacceptable. In nonscientific subjects it is more difficult to repel the challenge and often the new ideas take over, mainly because of their novelty. Freshness is all. In both cases the new approach tries to preserve some aspects of the older viewpoint, for innovation is most effective when it builds on at least part of the existing tradition.

  What, then, do Jim Watson and I deserve credit for? If we deserve any credit at all, it is for persistence and the willingness to discard ideas when they became untenable. One reviewer thought that we couldn’t have been very clever because we went on so many false trails, but that is the way discoveries are usually made. Most attempts fail not because of lack of brains but because the investigator gets stuck in a cul-de-sac or gives up too soon. We have also been criticized because we had not perfectly mastered all the very diverse fields of knowledge needed to guess the double helix, but at least we were trying to master them all, which is more than can be said for some of our critics.

  However, I don’t believe all this amounts to much. The major credit I think Jim and I deserve, considering how early we were in our research careers, is for selecting the right problem and sticking to it. It’s true that by blundering about we stumbled on gold, but the fact remains that we were looking for gold
. Both of us had decided, quite independently of each other, that the central problem in molecular biology was the chemical structure of the gene. The geneticist Hermann Muller had pointed this out as long ago as the early 1920s, and many others had done so since then. What both Jim and I sensed was that there might be a shortcut to the answer, that things might not be quite as complicated as they seemed. Curiously enough, I believed this partly because of my very detailed grasp of the current knowledge of proteins. We could not at all see what the answer was, but we considered it so important that we were determined to think about it long and hard, from any relevant point of view. Practically nobody else was prepared to make such an intellectual investment, since it involved not only studying genetics, biochemistry, chemistry, and physical chemistry (including X-ray diffraction—and who was prepared to learn that?) but also sorting out the essential gold from the dross. Such discussions, since they tend to go on interminably, are very demanding and sometimes intellectually exhausting. Nobody without an overwhelming interest in the problem could sustain them.

  And yet history of other theoretical discoveries often shows exactly the same pattern. In the broad perspective of the exact sciences we were not thinking very hard, but we were thinking a lot harder than most people in that corner of biology, since in those days, except for geneticists and possibly the people in the Phage Group, most of biology was not thought of as having a highly structured logic.

  Then there is the question of what would have happened if Watson and I had not put forward the DNA structure. This is “iffy” history, which I am told is not in good repute with historians, though if a historian cannot give plausible answers to such questions I do not see what historical analysis is about. If Jim had been killed by a tennis ball, I am reasonably sure I would not have solved the structure alone, but who would? Jim and I always thought that Linus Pauling would be bound to have another shot at the structure once he had seen the King’s College X-ray data, but he has stated that even though he immediately liked our structure it took him a little time to decide finally that his own was wrong. Without our model he might never have done so. Rosalind Franklin was only two steps away from the solution. She needed to realize that the two chains must run in opposite directions and that the bases, in their correct tautomeric forms, were paired together. She was, however, on the point of leaving King’s College and DNA, to work instead on Tobacco Mosaic Virus with Bernal. (She died five years later at the early age of thirty-seven.) Maurice Wilkins had announced to us, just before he knew of our structure, that he was going to work full time on the problem. Our persistent propaganda for model building had also had its effect, and he was proposing to give it a try. Had Jim and I not succeeded, I doubt whether the discovery of the double helix could have been delayed for more than two or three years.