( included in the web site ninio )



- General comments

- Biographical notes

- Kinetic modulation (kinetic partitioning)

- The next nucleotide effect

- Kinetic amplification (kinetic proofreading) on sticking times

- Kinetic amplification of a proofreading activity

- Kinetic amplification on a forward rate

- Comment on Hopfield's energy relay scheme

- The accuracy tuner

- Diffusional proofreading

- The fate of the kinetic proofreading / kinetic amplification ideas






- the 1978 EMBO workshop on accuracy (section on accuracy, and its control)

- Experimental work (section on DNA and RNA polymerase fidelity and kinetics )

- Non-canonical (odd) nucleotide pairs (in RNA structure section)

- Codon-anticodon recognition (in codon-anticodon recognition section)

- connections between error-rates (section on accuracy, and its control)

- Mutation strategies (in molecular evolution section)

- Immunological specificity (in section on immunological specificity)



++++++++++++++++ GENERAL COMMENTS +++++++++++++++++++++


Up until now (2005) my most highly praised scientific contribution is the co-discovery, with John Hopfield, of theoretical enzyme mechanisms to enhance the discrimination between cognate and non-cognate susbstrates, which I called "kinetic amplification" and which he called "kinetic proofreading".

In my case, the kinetic amplification schemes were developped while I was working in Leslie Orgel's laboratory at the Salk Institute. Orgel contributed in a decisive manner to the work, although he did not sign the article describing it. I hit upon the kinetic amplification idea after a long maturation period, in which I developped, in a number of steps, a new way of thinking about recognition problems in molecular biology [1-3].

Although John Hopfield's contribution and mine have been acclaimed, I am not sure that they are clearly understood. Prior to our work, some important notions had already been clearly established, experimentally: The existence of "classical" proofreading was known in the case of DNA polymerases, and well on its way of being established in the case of amino-acyl tRNA ligases. The existence of DNA repair mechanisms was also established. So what was new? I propose here one way to explain the novelty. Classical thinking follows an "If ... then" logic. IF the enzyme makes a mistake, THEN the mistake is subsquently corrected by proofreading or repair - or: IF the enzyme binds the correct substrate, THEN it switches to a conformation allowing the reaction to proceed. This reasoning assumes that erroneous binding can be distinguished from correct binding with full confidence. But what if the problem, precisely, is to distinguish between correct and incorrect binding? Let us say that you have a kinetic mechanism of reaction, and there is, at one place or another in the scheme, a difference in the kinetic parameters for correct versus incorrect substrates. How far can THIS difference be exploited? Can discrimination go beyond the ratio of the kinetic parameters? This is the problem that John Hopfield and I solved independently.

The new way of thinking about accuracy allowed one to make sense of experimental observations which were, otherwise, paradoxical. The best example is that of the "anti-mutator" (high-fidelity) DNA polymerase of phage T4. This mutant DNA polymerase makes a very low level of errors. So you may think, in the lock-and-key and induced-fit frameworks, it must deal with the correct substrate very efficiently. Quite to the contrary, as shown by N. Muzyczka, R.L. Poland and Maurice J. Bessman (Journal of Biological Chemistry, 1972, vol. 247, pages 7116-7122) the high-fidelity DNA polymerase does not select better than wildtype at the incorporation level, and its proofreading function destroys 30% of the CORRECT incorporations, while the wildtype DNA polymerase destroys only 4% of the correct substrate! This paradox is easily explained in the general kinetic framework. Unfortunately, the biolgy students are still fed with a mixture of old and new concepts: locks and keys, induced fits, proofreading, and sometimes kinetic proofreading, but all these are given as components of a consensus picture, without explaining when the different explanations collide, and lead to contrasting predictions. This situation reflects the lack of maturity of biology, as compared to physics.

The kinetic amplification paper was extremely beneficial to my carreer, since it rapidly gained popularity among the researchers interested in ribosomal mechanisms, it brought me a decent level of citations, and gave me scientific visibility and respectability. This, in turn, gave me time and freedom to think about other subjects, and move from one field to another according to my latest interests.

The kinetic amplification/proofreading schemes were initially discussed in relation to translation, replication or transcription. Recently, these schemes have invaded other domains, and in particular, immunology. However, these schemes are often taken as "cosmetic tricks", and the more elementary principles which led to the schemes need yet to make their way into the collective consciousness of molecular biologists.

My initial article on accuracy [1] was not on mechanistic details, but on how you could match two broad families of potential partners - codons and anticodons, and how degeneracy could contribute to hide real "underground" ambiguities. The article had little impact, when it was published in part because people were not yet prepared to this type of thinking, in part because I had at that time to rely on poor evidence. In addition, I reasoned in terms of equilibrium constants, instead of kinetic parameters. I consider though that this article is essentially correct (see section on codon-anticodon recognition), and that it came too early. To-day it could be rewritten and submitted in a fashionable area like "complexity" or "network theory".



+++++++++++++++++ BIOGRAPHICAL NOTES +++++++++++++++++++++


Very early, I was dissatisfied with lock-and-key thinking in molecular biology. I remember listening, as a student, to a course by Jacques Monod, in which he talked about the amazing specificity of enzymes, giving as an example the capacity of an enzyme to distinguish between the left and right enantiomers of a sugar. I was not impressed, knowing that the enzyme was not reading a structure on paper, but interacting in 3D. I also remember reading a book by Wolfgang Köhler: "Gestalt Psychology". In this book, Köhler described an experiment in which he trained chickens to take the food presented on dark-grey paper and reject the food presented simultaneously on medium-grey. Afterwards, when the chickens had to choose between food on medium-grey and light-grey paper, they selected the former. Thus, the animals did not recognize in an absolute manner a particular shade of grey, but had developped a criterion for choosing between two competing offers. From this reading, I became used to consider molecular recognition, not as a lock and key matching, but as a discrimination between competing offers.

Later, as part of my thesis work, I fiddled with molecular models of nucleotides, looking at all possible associations between couples, and triples. From this, I developped a rather tolerant view of the potential associations between nucleic acids. I was also aware of articles by Eck and Woese, on the genetic code, implying loose binding specificity at some positions of the codon-anticodon associations, and Crick's wobble hypothesis was also circulating as a memo.

In December1965, I attended a meeting of the British Biophysical Society in London, and during the meeting, I laid down some thoughts on the booklet of abstracts, one of which shows a clear concern for a quantitative understanding of accuracy (note 1, below).

The "missing triplet hypothesis" was published in 1971 (see the section on codon-anticodon recognition). In this framework, the codon-anticodon interaction was characterized by a binding energy, and I was arguing that by decreasing the binding capabilities of an anticodon, you could reduce its capacity to misread significantly, while maintaining enough capacity to bind to a cognate codon. So, I had the clear notion that a cognate codon-anticodon interaction could be strong, as well as being just strong enough to permit polypeptide extension with a reasonable probability.

Later in 1971, Luigi Gorini published in "Nature New Biology" (volume 234, pages 261-264) an article entitled "Ribosomal discrimination of tRNAs" which, I immediately felt, was a clear step backwards. He had data on low fidelity and high fidelity ribosomal mutants, which he interpreted by postulating a capacity of the ribosome to read the codon before the binding of the tRNA, and switch to a conformation that would favour the wildtype tRNAs corresponding to the codon. This screening procedure would be enhanced in high fidelity ribosomes, and diminished in low fidelity ribosomes. Nevertheless, Gorini's data were extremely interesting, and offered me a unique opportunity to sharpen my ideas.

Prior to my exposure to Gorini's paper, I was reasoning in terms of energy levels and energy differences. Now, with his detailed results in hand, I needed to understand quantitatively the relationship between a pattern of binding energies, and a pattern of probabilities of incorporation of amino acids into proteins, and find a decently simple explanation for how these patterns changed from one mutant ribosome to the other.

In 1972, I was writing a review article on codon-anticodon recognition, and decided to clear up the matter at the start of the article. After discussing with Pierre Claverie, I wrote a section entitled "Recognition from the point of view of thermochemistry" in which I groped my way from equilibrium to kinetic thinking. After having expounded the equilibrium viewpoint, I noted that the current sequential desccription of the elongation process was more like "initial kinetics". Then, in one paragraph ([2], page 305) I proposed that ribosomal accuracy could vary in-between the two limiting-cases. In practice, ribsosmal accuracy could be governed by the relative magnitudes of the period of some ribosomal oscillator and the magnitude of the lifetimes of the codon-anticodon associations (see Note 2).

On page 328, I made the link with Gorini's work and refuted his interpretations. I sent my manuscript to Leslie Orgel, and he replied (May 8th, 1972) "I, too, got interested in Gorini's recent paper on the Ram mutation. Like you, I concluded that his final comments on the specificity of recognition are incorrect, but the kinetic model that I formulated is somewhat different from yours in detail". And in a subsequent letter (May 23rd), he wrote:"The kinetic scheme that I considered for protein elongation is very similar to your own, except that I treat the elongation step conventionally. (See attached page)". In fact, he had considered a Michaelis scheme, and also raised the question: what happens if the second step is made reversible ?. Being still very ignorant in fundamental chemical kinetics, I did not realize that what he had written was a perfect transcription of what I had in mind.

I moved to Orgel's laboratory in September 1972, and this was the beginning of an exceptional period in my life. I had as main project experimental work on non-enzymatic replication of nucleic acids (see Section on the Origins of Life) but in parallel, we were discussing almost every day of all matters related to the accuracy of molecular processes.

We were agreeing on the interpretation of Gorini's data and I soon made my mind to undertake a thorough quantitative analysis of the data, and see whether or not I could explain the behaviour of high fidelity and low fidelity ribosomal mutants by changes of the standard value of a single kinetic parameter. At that time, I was reasoning already in terms of probabilities, and was ignorant of the fact that the kinetic constants in an enzyme kinetic scheme were all, without exception, average values, taken from zero to infinity, on a probabilistic process. So, I started working on Gorini's data, and ultimately produced a complete kinetic interpretation of his data, involving one characteristic sticking time theta for each codon-anticodon interaction, independent of the particular ribosome mutatnt, and one particular decision time tau for each ribosomal mutant, independent of the codon-anticodon interaction. In addition, I needed to introduce a concentration factor for each competitor.

This work was one of the works which gave me the most intense pleasure in all my scientific carreer. It was not for the fact of having refuted the ribosomal screen hypothesis, and replaced it with a kinetic model. It was due to the fact that I had produced a set of parameters which were predictive of the in vivo situation. I had derived from Gorini's data estimates of the sticking times between codons and suppressor tRNA's in vivo! It was almost as having produced experimental results on the genetic code with my own hands.

I wrote a MS describing the work and proposed to Leslie Orgel to be a co-author but he firmly declined the invitation. The article was accepted without serious difficulty (Note 3).

During the early days at the Salk Institute, I was also scrutinizing the articles on DNA polymerases, looking for clues on nucleotide competition. I spent a considerable time on two articles from Arthur Kornberg's laboratory, one on the affinities of his DNA polymerase for nucleotides in the absence of template, the other on nearest-neighbour statistics. This led to nowhere.

I learnt from Orgel about the recent discovery of proofreading functions in DNA polymerases. Among the papers I read at that time, there was one by Z.W. Hall and I.R. Lehman (J. Mol. Biol. volume 36, pages 321-333, year 1968). Understanding what was going on led me to "the next nucleotide effect" (see the section below on this effect).

I was also reading the papers on the accuracy of amino acyl tRNA synthetases. Eldred and Schimmel on one side, Yarus on the other had described hydrolytic activities exerted by the amino acyl tRNA synthetases, thanks to which they could destroy incorrectly acylated tRNAs. Two French groups were also studying this hydrolytic activity, and came to the conclusion that it did not contribute to specificty. What had not been clearly realized, at that time, was that in order to be efficicient, the proofreading activity of a synthetase had to be exerted on its own errors, right after the error was made. If correction is not made "on the spot", the misacylated tRNA may leave the enzyme, and go to the ribosome before having a chance to encounter a synthetase that could hydrolyze it. I included later this remark in [6]. Prior to this, the researchers in the field were mixing a synthetase preparation with misacylated tRNA, and were showing that the misacylated tRNA was hydrolyzed. When the kinetic ideas began to gain acceptance, the experimental paradigms changed: A synthetase was then mixed with unacylated tRNA, and the incorrect amino acid. The net result was a hydrolysis of ATP, a testimony of the fact that the synthetatse had made (or attempted to make) a misacylated tRNA, thus hydrolyzing ATP, but did not let the tRNA leave with the incorrect amino acid attached to it. I started a correspondance on various points of enzyme kinetics and accuracy with Jacques Bonnet who was working in Strasbourg on amino acyl tRNA synthetase kinetics, but this did not deviate him from the course of his work.

While writing the paper on ribosome kinetics for J. Mol. Biol. [3], I was aware of the fact that my kinetic scheme, being a minimal one, could have been an oversimplification. In particular, I was a bit worried with the fact that within the formalism I was using, there were strong limits to ribosomal accuracy. I included a sentence in the manuscript showing my concern for the problem: "For instance, non-linear features and threshold devices can be directly built in the function p(theta, tau)." ([2], top of page 311). This sentence was already on the first submission, but I had not yet developped any scheme for creating thresholds. It would have been disturbing to think about this, while I was developping the linear model of Gorini's data!

After completing the ribosome kinetics paper, and having also analyzed Hall and Lehman's data on misoncorporation by T4 DNA polymerases, I started writing a note putting my kinetic analyses in a more general context of enzyme specificity. I used the expression "the timing effect" to describe both the interplay between kinetic constants used to interpret ribosomal mutants, and the interplay between incorporation and excision to explain the DNA polymerase work. In the particular case of a misincorporation error due to a tautomery, the pair looks correct to the proofreading function, so it is better to wait a certain time before exerting proofreading, in order to give a chance to the incorrect nucleotide to return to its most natural state. I scrutinized papers on base-analog mutagenesis by John Drake, and by Howard and Tessman, and this strengthened my conviction. This is how I came to think practically about time-delays and non-linearities (Note 3a).

In all my treatments of ribosomal accuracy so far, the probability of incorporation (or misincorporation) was, at low sticking times theta, proportional to theta. Then, the correct/incorrect incorporation ratio was at best proportional to the ratio of the sticking times for the correct and incorrect interactions. In a sense, it was as though the ribosome was comparing the sticking times of the two substrates. In order to achieve a better discrimination, the obvious idea was to introduce a time-delay in the mechanism of the reaction. If you compare, not the sticking times, but the sticking times minus some "dead time" you can make the ratio of the residual times as large as you wish. The question was how to implement this abstract idea into a valid kinetic mechanism. I first thought that in order to have a time-delay, it was just enough to add one step in the mechanism of the reaction. For instance, the incorporation of the amino acid into a peptide would require the prior occurence of a second event, occurring after binding. The departure of the tRNA would be allowed during the first two stages of the reaction, but elongation could proceed only upon entry into the second stage. I submitted the scheme for discrimination in two steps to Leslie Orgel, and he immediately saw that it could not be correct because if the tRNA was allowed to leave at the second stage, it should also be allowed to enter at that stage, according to the principle of microscopic reversibility. I probably left the matter settle down a certain time, and was probably busy revising the MS on Gorini's mutants for J. Mol. Biol.

The revised MS being sent, I came to Leslie with a modified scheme, in which two substrates S1 and S2 were binding in random order, S1 being the substrate on which good discrmination needed to be achieved, and S2 being an auxilliary substrate. The "S1 first" pathway would be a high accuracy pathway, because S1 could leave at two stages (before and after S2 binding), and the "S2 first" pathway would be a low accuracy one, because S1 would leave only at the second stage. I then postulated that the kinetic constants were such that "S1 first" was much more frequent than "S2 first" (Note 4). It was one of the last days of December 73. Leslie said again that the scheme could not work, because of microscopic reversibility. I replied something like "so let us add an ATPase activity to prevent tRNA from entering at stage 2". He agreed, and that was it.

In January 74, I had a kinetic amplification scheme which appeared to hold against all objections.

I wrote a short note describing the idea, and proposed to Leslie to be a co-authour, but magnanimously again, he declined the invitation. I was not as proud of this work as I was of the previous one. I considered the previous one as a conceptual breakthrough, while the second one looked to me more like the cherry on the pie. We had decided that the kinetic amplification paper would be submitted to "Nature". The paper went through many versions, and Orgel, made style contributions, trying to make it make it more comprehensible. (I have kept nine versions, and Orgel's handwritten corrections). At the end of April, 1974, my period in the Salk Institute was coming to an end. The final version of the manuscript was confided to Orgel's technician in charge of the illustrations, Paul Fuhrman. At that time, the figures were drawn manually with China ink, and Paul had also to calligraph all equations with China ink.

I departed from San Diego on April 29th, and made a stop in Boston. I was received by Alexander Rich at the MIT. We had dinner in the evening at the MIT restaurant. Paul Schimmel was seated at a table close to ours. I told him that I had a scheme for amplifying discrimination, he said he would be interested to look at the work if I had something written, but I did not send him the MS. Then I made a stop in New Haven. I gave a seminar in Dieter Söll's laboratory on ribosome kinetics and introduced the kinetic amplification model. Last, I think I paid a visit to Freeman Dyson in Princeton. We discussed, I think, error-propagation and origins of life, but I have no written trace of this visit. I left the USA from New-York on May 3rd.

Upon my return in France, I was expecting to receive a letter from Nature acknowledging the receipt of the manuscript, but received nothing. Fuhrman had in fact left the MS on his desk. The MS was submitted in June.

In May 74 (13th-15th), I attended a meeting in Roscoff (France) on the evolution of biological macromolecules. I gave a conference on the accuracy of translation, and introduced my work on ribosome kinetics, and on kinetic amplification. Just after this meeting, I attended another meeting on automata theory and theoretical biology, organized by the mathematician Jacques Riguet. There I made acquaintance with Françoise Bernardi, and she showed an interest in collaborating with me in Chapeville's laboratory.

Around the 25th of July 74, in the morning, I received a letter from Nature, reproduced in Note 5 at the end, announcing that my MS had been rejected because of "lack of sufficient novelty in concept". Half-an-hour later, perhaps, Leslie Orgel, who happened to be in Paris, dropped by in the lab. I showed him Nature's letter, and he shrugged his shoulders. I was not too much affected. I did not suspect that someone else (John Hopfield) was already on the track. Furthermore, I felt secure because the contributions to the Roscoff meeting were to be published in a book. I delivered the Roscoff MS [4] in Sadron's hands, in Orleans, on Sept. 19th.

The chronology is becoming very precise because I happened to hold a handwritten, intimate diary from August 16th, 2004, to February 10th, 2005 (Note 6).

In the Autumn, I was busy writing several papers: a congress article on the problem of the co-evolution of cytoplasmic and mitochondrial translation apparatuses ([15], see also the section on codon-anticodon recognition), a short commentary article on DNA polymerase accuracy, to appear in the information pages of "Biochimie" [16], and a popular science article on the origins of life to be published in "La Recherche" [17]. I was also initiating two lines of experiments with Françoise Bernardi. One was the search for catalytic activities in randomly generated peptides (see section on the origins of life) and another was on the enzymology of DNA polymerases - measuring the influence of nucleotide concentrations on error-rates and proofreading.

On October 16th, I found in "Nature" an article by an MIT Professor (volume 251, pages 385-388) in which, without quoting my work on ribosome kinetics, he was applying similar ideas to the problem of the competition of two messenger RNA species on the ribosome. Fortunately for me, his mathematics were obviously erroneous. I pointed it out in a commentary [5], to which he replied arrogantly, writing at the end "But I am afraid Ninio's equations are incorrect, and do not really help matters". This episode was traumatic for me, not because of the reaction of the MIT Professor, but because of the Pilate's attitude of the editors of Nature. This attitude would have been unthinkable in physics or mathematics, and it amounted to an editorial scientific fraud (see forthcoming section on error and fraud in science).

November 1974 was characterized, in France, by huge postal strikes, and we were receiving nothing from the USA. On November 5th, I gave a telephone call to Leslie Orgel, to discuss various matters - mainly my future trip to the Salk Institute. He told me that a person from the Bell laboratories had reached the same conclusion as I on the specificity of enzyme reactions, and had a paper in press. Feeling the danger, I started writing an expanded version of the article previously submitted to Nature, with more discussions of the current litterature pertinent to the topic.

On November 6th, I went to Grunberg-Manago's institute to discuss catalysis by metal-peptide complexes with Antoine Danchin. In the library, I found an issue of the Journal of Molecular Biology, bearing the date of Sept. 14th, which contained a mathematical treatment of proofreading by wildtype and mutant T4 DNA polymerases. I realized rapidly that the treatment was a complete aberration (see section on mathematical biology). Later, I had a correspondance with David Galas. He had made exactly the same diagnosis [Note 8]. By November 21st, I had derived the correct "peelback equation", which was missed by the authors of the JMB articles. And on Nov. 22nd, Françoise Bernardi and I made our first experiment on the accuracy of a DNA polymerase.

On dec. 11th, in the morning, I finalized the new manuscript on kinetic amplification, and just as a I finished stapplling the pages together, I received from John Hopfield a preprint of his article. This was the beginning of a correspondance between us (Note 7). John Hopfield acted in a very ethical and chivalrous manner, offering his help to have my manuscript published in "Biochimie". Our two names are now associated in the international scene for the invention of the kinetic proofreading concept, and I wish to thank here John Hopfield for never making a quarel of priority. If I insist on this point, it is because I have been in France the target of two stinking articles, damaging to my reputation, by the French popular science magazine "La Recherche", written at twenty years interval by the same person (Note 10). In any event, the history of how I developped my ideas, could, I think, be interesting in itself. John Hopfield's path was different. He told me that the kinetic proofreading idea occurred to him while he was writing a document nominating Overhauser for the Nobel prize in physics. He realized that the Overhauser effect could be adapted to molecular biology, and provide a solution to the molecular accuracy problem.

After the acceptance of the kinetic amplfication paper, which took quite a time (Note 9), I was feeling that it was now time for the experimentalists to catch up with theory. While awaiting the next crop of results, that could stimulate new ideas, the best I could do, as a theoretician, was to move to another field. This is how I came to attend, in May 1975, a school on neurobiology in Arcachon. And in June 1976, I already had a manuscrpt ready on visual illusions! (see Sections on cognitive sciences and on visual illusions).

I maintained though a close contact with the accuracy work. In August 1976, at the suggestion of Henri Grosjean, I was invited to an EMBO meeting on ribosomes, in Brussels. I was scheduled to take part in a "round table" in the last session of the meeting. (An account of the meeting, by Richard Brimacombe appeared in Nature, volume 262, pages 740-742). The inaugural talk was given by Charles Kurland. I was seated anonymously among the participants in a big conference room with, on my right side, my future wife Ewa. Kurland said that two important events had marked the recent history of ribosome research: First the progress in understanding the ribosome's 3D structure, initiated with James Albert Lake's electron microscopy studies, then the introduction of kinetic thinking to explain ribosomal accuracy. Very kind to my work, Kurland distinguished the first stage which was the initiation of the kinetic paradigm (my analysis of Gorini's mutants) from the second stage which was the introduction of the kinetic proofreading mechanisms. I was aged 34, and for the first time in my life, I received a public praise for my scientific contributions. It was fun for me to be associated with J. -A. Lake, whom I knew from the tRNA structure period (see section on RNA structure), and I spent a pleasant time with him. At the end of the meeting, I presented my viewpoint. Very kindly, Alexander Spirin proposed to take on his time to give me more time to develop my ideas. In the audience, David Apirion raised his hand and pointed out, that independently of Gorini, he had also contributed to the genetics of ribosome accuracy. Later, I read his papers willing, if possible to repair my omission. The problem, however, was that his data were not suitable for the type of analysis I had applied to Gorini's data.

In the period that followed, I was retrieving the benefits of my contributions to accuracy, in large measure thanks to the strong and active support of Charles Kurland. Kurland showed an interest to have me spend some time in Sweden, and arranged for me a stay in Rudolf Rigler's laboratory in Stockholm, on a two months EMBO fellowship. There, I shared an office with Mans Ehrenberg, and besides discussing with him, I did many many things at the same time: I was writing a paper on stereoscopic vision, and starting my book on molecular evolution. I even performed small-angle X-ray scattering calculations. Mans Ehrenberg believed that kinetic amplification schemes could be transposed to discrimination on forward rates, I believed that they could not. Finally, I found some kind of formal proof suggesting that such schemes, if they existed, required that the enzyme would do something, undo it, and do it again (see [7] and section below on forward rates).

There, I also did experimental work on proofreading by an aminoacyl tRNA synthetase, specific for phenylalanine. I transposed the technologies Françoise Bernardi was using for DNA polymerases: making an incubation, spotting, on a sheet for thin-layer chromatography, the extracts taken at various times, elute, and follow the radioactivity in the various fractions (the polymerized material, ATP, and its decomposition products, ADP and AMP). My aim was to detect the abortive incorporation of an incorrect amino acid as an excess of ATP degradation. The experimental principle was a valid one, and it had not been, as far as I remember, used at that time by other workers in the field. Tyrosine emerged as the amino acid giving the highest level of ATPase activity (one per mil that of phenylalanine), but this ATPase was not tRNA-dependent. The work possibly stopped for this reason, or because the amino acyl tRNA synthetase was in short supply.

In thinking about possible extensions of the kinetic amplification ideas to other fields, I became attentive to protein phosphorylation. I attended, in May 78, a meeting on protein phosphorylation in Cary-le-Rouet. The dominating feeling was that phosphatases and kinases had very low specificity, and this oriented me towards thinking about "network specificity": how you could design a system with a large number of proteins, adding or removing phosphate groups in a very degenerate manner, and get at the end very sharp responses [13]. More interestingly, I proposed that with such a system, one could have an ADAPTIVE regulation network, a network that would be custom-fit (a particlularly important feature, for diploid organisms). I speculated that this was conceivable, and expounded the idea, without any mathematical development, in [14]. Due to my biochemical illiteracy, I used the word "phosphorylase" instead of "phosphatase". The idea of an adaptive phosphorylation network was also developped by Tony Hunter in 1987 in an article published in "Cell".

The EMBO workshop on accuracy, which I organized in September1978 in Grignon, was one of the most rewarding events in my scientific life, and it was followed by two similar workshops, one organized by Robin Holliday in 1981, and another by Charles Kurland in 1985 (see section on accuracy and its control).

Thanks to the EMBO workshops, I developped good personal contacts with most workers in the field of accuracy. Jonathan Gallant was quite kind to me, and I began to think seriously about the stringent response. The paradox, following Gallant's views, was that upon amino acid starvation for a particular amino acid, the ribosome could translate the codons for this amino acid at the standard error-rate level, despite its under-representation, without changing anything about its translation at the codons for the other amino acids. I designed a scheme which could do that, the "accuracy tuner", published in FEBS Letters in 1986 [9]. The mechanism could buffer accuracy against variations in the substrate concentrations. I have not followed the recent developments on the stringent response, so I do not know how this mechanism fares on the ribosome. It might play a role in immunology, because it would be useful there to make the immune responses not too dependent upon the precise concentration of foreign antigens.

Finally, thanks to visits made in my lab. by Eduardo Mizraji, from Uruguay, I became sensitive to problems of macromolecular diffusion. So, I thought about the mechanism thanks to which a repair system locates a mismatch, and how to make this mechanism as precise as possible. I developped a "diffusional proofreading scheme", but without mathematics. I had reached my limits. I gave a phone call to John Hopfield, asking him if he had worked on the question, but it was not the case. I put my speculation in a Cold Spring Harbour review (see [11] and section on diffusional proofreading below). This was my ultimate contribution to the kinetic theory of accuracy. All I have to say is encapsulated in two review articles, one technical, on concepts and calculation methods [10], and one more empirical, discussing the experimental evidence [11].





Having isolated a collection of low fidelity and high fidelity ribosomal mutants, Gorini and co-workers had the brilliant idea of studying the behaviour of suppressor tRNAs in these mutants. The "nonsense suppressor" transfer RNAs known at that time, and which they used, were mutant tRNAs, not normally present in the cell, which could bind to termination codons through orthodox pairings, and donate their amino acids to the groing polypeptide chain. Similarly, there were "missense suppressor" tRNAs which were mutant tRNAs, thought to bind to codons through orthodox codon-anticodon pairings, but carrying a "wrong" amino acid.

Under the conceptual framework of the time, the nonsense suppressor and missense suppressor tRNAs studied by Gorini were making correct pairings with the codons, they followed the wobble rules and were expected to follow the same fate as the correct tRNAs. What Gorini and his associates found - to their great surprise, it seems - was that the reading of the cognate codons by the mutant tRNAs decreased in high-fidelity ribosomes, and increased in low fidelity ribosomes - they behaved as though they were using non-cognate interactions.

For me, the result had a very simple, natural explanation; Although these mutants were making orthodox pairings with the codons, they were in fact weak binders. On a binding energy scale, they would be intermediate between wildtype tRNAs making cognate interactions, and wild-type tRNAs making mistakes.

With his data in hands, Gorini might have reacted by reexamining his starting assumption (recognition as a purely geometrical lock-and-key matching). Instead, he proposed a baroque theoretical construct which he called the "ribosomal screen". In this construct, the wild type tRNAs responding to a particular set of codons had, outside the anticodon region, a special distinctive feature. A ribosome, having "read" a codon in the absence of tRNA would undergo a conformational change which would favor the binding of the tRNAs having the special feature associated with the codon. At the end of his article, he wrote that "mistranslation is kept under control by the wild-type ribosome without interfering with normal translation". His way of thinking appeared to me archaic.

In my review on codon-anticodon recognition [2], I explained why translation errors and missense or nonsense suppression could increase or decrease simultaneously. Both could be aspects of a same phenomenon, "an increased or decreased capacity to discriminate between two competitors" (page 328) and proposed that ribosomal accuracy could be goverened by the relative magnitude of sticking times "tau's", and times reflecting the speed of a decision-making process, "theta's". In conformity with the notations of subsequent articles, I now designate the sticking times by theta's, and the decision times by tau's.

I visualized the ribosomal events occurring on a time axis. After the landing of a tRNA molecule, there was a time-segment corresponding to the sticking time theta of the codon-anticodon association . At intervals of time tau, some ribosomal oscillator would probe the occupancy of the A site. In case the site was occupied, the ribsome would go one step further in the elongation process. But a tRNA might land and leave in-between two occupancy checks, so there would be abortive binding events.

The crucial element in the picture was to view the tRNA binding and departure events on one hand, and the ribosome's attempts at going to the next stage of elongation as completely independent events. This goes against common sense if one thinks of an enzyme (or a ribosome) as a motor, and I suspect that even now, this image is counter-intuitive to most biologists. Yet it is just the natural physicist's view of nature, since the time of Boltzmann.

Technically, I considered for the theta's exponential decays of average time theta. Initially I took the tau's constant (as would be the case for some kind of pendulum) then, instructed by Orgel, I took an exponential decay of average tau. I was considering the complete probability distributions, and integrating from zero to infinity, I derived the (correct) result that the elongation probabilities were just equal to theta/(theta + tau). Orgel then pointed out that this was just the k2/(k_1 + k2) ratio in a Michelis scheme.

Most of the subtleties of ribosomal accuracy mutants could be encapsulated in this simple formula:

probability of success, after a binding event = theta/ (theta + tau)= k2/(k_1 + k2)

On a ribosome working extremely fast (tau close to zero, very large k2) the probability of success would always be close to one. There would be no discrimination between weak binders (small theta's) and strong binders (large theta's). On a ribosome working very slowly (very large value of tau, very small k2) the probability of success would be proportional to theta, and it could be significantly smaller than one even for cognate interactions. Thus, I proposed, for the most extreme high fidelity ribsomal mutant, that in the case of cognate interactions, the probability of elongation would decrease to about one half [3].There would be on average two cognate binding events for one successful incorporation. In terms of efficiency, accurate ribosomes would be less efficient than wildtype. This antagonism between accuracy and efficiency is again at variance with the "mechanical engine description", and with the "induced-fit" models.

I realized that this new understanding of how accuracy in a molecular recognition process could be controlled by a kinetic parameter could have wide applications, outside the initial field of ribosome mechanisms. Later, I coined the expression "kinetic modulation" to describe the effect. Others tend to use now the expression "kinetic partitioning" to describe the same phenomena.

In 1974, Schwartz and Lysikov published a short article in which they discussed ribosomal accuracy in terms of the interplay between two kinetic constants, but not going into a detailed analysis of Gorini's data (see Note 11). It seems to me that their article stood at an intermediate stage between [2] and [3].



+++++++++++++ THE NEXT NUCLEOTIDE EFFECT +++++++++++++++++


The principle which helped me see through the ribsomal mutants, could be easily transposed to the data on DNA polymerase accuracy. My starting point was, as mentionned in the biographical notes, a paper by Hall and Lehman, published before proofreading was known.

Hall and Lehman studied nucleotide misincorporation errors by phage T4 DNA polymerase, and by a low-fidelity mutant of this polymerase. The chosen tempate was Poly(dC), the correct nucleotide was thus dGTP, and its incorrect competitor was TTP. They studied the T/G misincorporation rate as a function of the dGTP concentration, at constant TTP concentration. They expected that by increasing the concentration of dGTP, they would dilute the errors. They found instead that the T/G ratio reached a plateau at high G concentration as though dGTP could not dilute the errors.

I immediately saw how the kinetic thinking I had developped for ribosomes could be transposed to this situation, and understood that the observed behaviour was a direct consequence of the existence of the proofreading function: after the misincorporation of a T residue the period of time during which proofreading may occur comes to an end when a nucleotide is incorporated at the next position. Remembering that several of my colleagues in my laboratory of origin (Chapeville's) were working on DNA polymerases, I wrote a note explaining the effect, making a complete quantitative analysis of Hall and Lehman's data, and urging them to do the experiments, but in vain. At the end of the note, I considered the case in which misincorporation was due to a tautomery, making the mispair hard to detect, for a certain time.

One immediate consequence of the next nucleotide effect is that accurate replication with a DNA polymerase having a proofreading function is achieved at rather low concentrations of substrates. This went against the enzymologist's intuition. At that time, and for reasons incomprehensible to me, the enzymologists took as an article of faith that enzymes had to be studied at high substrate concentrations (at least, above the KM). Any experiment performed below the KM was considered to be unphysiological rubbish. Therefore, understanding the next nucleotide effect had an important practical value in guiding the design of high accuracy DNA replication systems for the needs of DNA sequencing. So, it is fun to think that I have contributed in a very practical manner to the rise of modern genomics, first with the invention of the next nucleotide effect, second with the invention of the fast sequence comparison algorithm, now famous under the name of "BLAST" [18] (see Section on "contributions to bio-informatics").

The next nucleotide effect was mentioned in one paragraph in [6]. Having a co-worker in the person of Françoise Bernardi, we started quantitative studies on incorporation and excision by DNA polymerases (see the website section on "contributions to DNA and RNA polymerase kinetics"). We had results consistent with the next nucleotide effect ([8]) but since these results were obtained with homopolymers, they were not as decisive as the results later obtained in 1979 by Alan Fersht (Proc. Nat. Acad. Sci. USA, vol. 76, pp. 4946-4950) and in 1981 by Tom Kunkel et al., (Journal of Biological Chemistry volume 256, pp. 9883-9889) with heteropolymers.



The genesis of my kinetic amplification mechanism [6] was given in the "biographical notes" section. This mechanism is different from Hopfield's mechanism (John J. Hopfield: Proc. Nat. Acad. Sci. USA, 1974, vol. 71, pages 4135-4139) in some details, but both mechanisms work on the principle of time-delays, and there does not seem to be a difference in energetical effciency between the two - at least, I am not aware of a comparative work bearing on this topic.

Before discussing the very delicate problem of what is a valid biochemical implementation of time delays, I give the much simpler abstract idea behind it. Suppose you have two alternative substrates, one characterized by a long sticking time, say theta1 = 100 msec, and another characterized by a shorter sticking time, say theta2 = 25 msec. If, after binding, the initiation of the next step in the processing of the substrate occurs at random, with uniform probabiiity in time, the probabilities of initiating the next step, p(1) and p(2) for the first and second substrates respectively are in a ratio which cannot be larger than theta1/theta2. This can be shown rigorously. Now assume that, after binding, there is some "dead time" td, during which the next step cannot be initiated. So, the part of the sticking time remaining available for the intiation of the next step is theta1-td for the first substrate, and theta2-td for the second substrate. In the above example, we had theta1/theta2 = 100/25 = 4. Now if we take td = 20, we would have (theta1-td)/(theta2-td) = 80/5 = 16. When the calculation is made rigorously, one finds that the ratio of the probabilities may now approach the ratio of the square of the sticking times.

In order to go further, we have to use a physical reasoning which is at variance with the classical enzymological tradition. Saying that there is a certain sticking time theta amounts to say that, after binding, the substrate is free to leave, with a probability dt/theta per interval of time dt, which also amounts to say that departure is governed by a kinetic constant k_1 = 1/theta. Now, if you describe the enzyme mechnaism with a kinetic scheme, the departure of the substrate will appear at two places in the scheme: First, during the "dead time" then after this time. Therefore, the k_1 constant will be seen at two places in the scheme. If you are not aware of the reasoning behind the scheme, and look at the scheme with a classical enzymologist's turn of mind you might say, "so what's original? There are two places in the scheme in which the enzyme makes a difference between the two substrates, and global discrimination is the product of the discriminations achieved in each of the two steps". Then you would miss the point, because the scheme is not the real thing. It is a symbolic representation, which captures everything needed for the calculations, but does not necessarily capture everything needed for comprehension. From the point of view of susbtrate binding and departure there is a single physical process, although, from the point of view of the symbolic transcription into a scheme, there are two steps. For a more comprehensive discussion of kinetic formalism versus physical reality, see [10].

Once the idea of time-delay is understood, and it is a rather easy mathematical idea, we need to go to the second point, which is a deep physical one. If the substrate is allowed to leave, after the occurence of the event which terminates the dead time, as required by the scheme, what prevents it from entering after the occurence of this event, thus bypassing the time-delay ? Hopfield and I proposed two, slightly different ways of preventing substrate entry at the second stage. I recommend to the readers interested in the topic to read our original contributions, because subsequent reviews have often been misleading.

People have sometimes assumed that the schemes were identical, and did not bother to read my articles. For instance, Sean M. Burgess and Christine Guthrie proposed (Trends in Biochemical Science, 1993, vol. 18, pp. 381-384) a "new" scheme, which was in fact exactly the scheme I had described in [6].



++++++++++++ DELAYED ESCAPE OF THE PRODUCT, ++++++++++++++++++



So far, we reasoned on the case in which enzyme discrimination was based on differences in sticking times between two substrates. Now let us try to extend this type of thinking to other discriminative steps. The case of discrimination on a forward rate is particularly difficult, I discuss it in the next section. Amplification OF a classical proofreading activity is rather easy.

Let us go back to the next mucleotide effect. Proofreading is exerted during a certain time theta, which is the time left until the next incorporation begins. Now, we assume that the enzyme has the capacity to destroy the incorrect substrate faster than it can destroy the correct one. The subtlety here, is that you do not learn much by looking at the rates of hydrolysis, for correct and incorrect substrates, and comparing them. The thing which matters most, from the point of view of accuracy, is the probability of SURVIVAL of the erroneous product. So what is important is to have a proofreading activity working as efficiently as possible. The solution again is in having a time-delay, but it is a delay in the departure of the product. I proposed a scheme of "delayed departure of the product" which was presented in [6]. John Hopfield hit upon a similar conclusion, and presented his idea in the context of the origin of life (Proc. Nat. acad. Sci. USA, 1978, vol. 75, pp. 4334-4338).

The scheme has attracted very little attention, so far. Looking at the kinetic studies on aminoacyl-tRNA ligases, I believe there is room, at least with some of the ligases, for such a mechanism. There is also room for such a mechanisms in the case of DNA polymerases, but outisde the work carried out by my co-workers (see section on DNA polymerase kinetics) and by Robert A. Beckman and Lawrence A. Loeb (Quarterly review of Biophysics, 1993, vol. 26, pp. 225-331) I am not aware of any strong interest in the scheme.




Let us now take the case in which, after substrate binding, the enzyme may engage into one of the subsequent steps faster with one substrate than with another. If there is a certain ratio of the forward kinetic constants for the two substrates, what is the ratio of the corresponding probabilities of completing the reactions ? Can it be higher than the ratio of the kinetic constatnts ? After analyzing the problem, I reached a pessimistic conclusion [7]. Kinetic amplification on a forward rate is impossible, within "the conceptual limitations of the study" unless, it seems, one uses "a mechanism in which the substrate has to go more than once through a same step, and would go more often through that step when it is the correct one". I did not propose any scheme that could do it. Remember that in my discussion of the delayed reaction, I insisted on the point that although a kinetic constant appeared twice in a scheme, the departure of the substrate was a single physical process. The formal proof I gave concerning forward rates was about amplification on a single pprocess taking place in a continuous stretch of time.

Recently, Jie Yan, Marcelo O. Magnasco and John F. Marko published a scheme for kinetic amplfication on a forward rate, applicable to DNA topoisomerases (Nature, 1999, vol. 401, pp. 932-935) and this scheme is consistent with the analysis given above. The enzyme in this scheme does something, goes back; and does it again. The mathematics are correct, but I have two difficulties with the scheme:

- in the topoisomerase reactions, the product of a successful reaction may be the substrate for the next reaction. So, what counts as "correct" at one stage (the product), counts as "incorrect" at the next stage (the substrate). So the whole chain of events counts, and I have not seen an analysis of this.

- I do not see why the scheme forbids (see their Fig. 4) the 2 --> 2' or the 2' --> 2 direct transitions.

In addition, I do not agree with the author's claim that the scheme solves an entropic or an energetic problem, but this is a general position I have concerning all kinetic mechanisms. A kinetic mechanism cannot solve an equilibrium problem.





In 1980, John Hopfield published one more kinetic porofreading / kinetic amplicfication scheme, acting on sticking times (Proc. Nat. Acad. Sci. USA vol. 77, pp. 5248-5252). Here, the enzyme is capable of accumulating energy derived from the reaction with the correct substrate, and use this energy to create a time-delay during the next reaction cycle. The principle looks to me entirely valid. The enzyme would have mnemonic (hysteretic) properties, since the state in which it initiates a reaction depends upon the events which occurred at the preceding cycle of interactions. It must be rather dificult, to prove exeprimentally, that an enzyme follows an energy relay scheme, so this variant has not yet attracted much attention (but see W. Radding and G.N. Phillips, Jr, BioEssays, 2004, volume 26, pages 422-433).


+++++++++++++++++ THE ACCURACY TUNER +++++++++++++++++++++++


So far, we reasoned in terms of the probability of completing a reaction, after an initial binding of the substrate. To make the picture complete, we must also take into account the conccentrations of the competing substrates. Usually, the error-rate will rise proportionally to the ratio of the incorrect to the correct substrate concentrations. There are departures from proportionality at low substrate concentrations, due to the "sequestration effect" [10]. In protein synthesis, when there is an accidental shortage of an amino acid, there is a shortage of the tRNAs carrying this amino acid, and therefore an enhanced risk of errors in the reading of the "hungry codons", those that correspond to these amino acids. Wildtype bacteria have a strategy, called the "stringent response" which allow them to compensate for the shortage of cognate tRNAs at the hungry codons. Interested in the problem, I designed a mechanism which could, in principle, "buffer" accuracy against changes in substrate concentrations. I called it the "accuracy tuner" [9].

The accuracy tuner is a mnemonic (hysteretic) scheme. In this scheme, the ribosome is initially in a standard accuracy state, and as it waits and waits for a stable interaction with a tRNA, it switches to a high accuracy state, in which the non-cognate tRNAs have an increased probability of being rejected.


++++++++++++++++ DIFFUSIONAL PROOFREADING ++++++++++++++++++


I also gave attention to the problem, faced with the mismatch repair system, of efficiently locating a mismatch. I came with a scheme in which a mismatch repair complex slides on a section of DNA, making some rapid but loose scanning. When encountering a mismatch, there would be a conformational change in which the complex would scan DNA with steps of smaller amplitude, and achieve a tighter sensing of normal and abnormal base-pairs [11]. I did not perform a mathematical treatment of the scheme. The related problem of single-stranded DNA detection by the recA proteins has been addressed by T. Tiusty et al. (Phys. Rev. Lett. 2004, vol. 93, 258103).





After the burst of the theoretical papers in the 1970's and 1980's, there have been a large number of studies linking accuracy to kinetics in protein synthesis (ribosomes, amicoacyl tRNA synthetases) and in DNA and RNA replication. After 1985, the crystallographic studies gained momentum, and mechanisms of reaction were proposed on the basis of crystallographic studies.

In the case of DNA polymerase no one, it seems, had anticipated the fact that the proofreading function is ahead of the polymerization function, not behind it (Tom Steitz and co-workers). Features derived from the crystallographic studies then imposed strong constraints on kinetic mechanisms. On the other hand, the crystallographers are still more attached to the old concept of "lock and key", and this makes the dialog with the kineticists difficult. Furthermore, when there are alternative pathways in a mechanism of reaction, the crystallographers tend to see only the most probable pathway, and are therefore led to propose strictly sequential schemes, as opposed to branched schemes. Hence, there has been a revival of "induce-fit" ideas, despite the fact that induced-fit schemes make predictions which are flatly contradicted by known experimental results (see the section above on the next nucleotide effect).

The same can be said of other fields. Crystallographic studies are bringing insights which could not have been obtained by pure kinetic considerations. The recent progress in understanding ribosome structure and looking into some aspects of ribosome mechanisms is amazing. On the other hand, one looses sight of the still valid paradoxical experimental results which were at the start of the kinetic ideas.

On top of that, it is regrettable that the rather elementary mathematics needed to deal with error-rates are not taught, so that papers on accuracy are often fraught with mathematical aberrations. People use formulas that are adequate for Michaelis kinetics, but not applicable to their particular case. (General conditions under which a complex kinetic mechanism conforms to a Michaelis scheme are given in [12]).

Kinetic thinking has migrated to new fields, outside the traditional domains of protein and DNA synthesis. The most interesting event was the transplantation of these ideas to immunology - more specifically, the mechanism of antigen presentation to the T cell receptor. Timothy McKeithan (Proc. Nat. Acad. Sci. USA, 1995, pp. 5042-5046) proposed a variant of the kinetic proofreading/amplification schemes in which the event terminating the dead time was a protein phosphorylation. His article has had an enormous success, among immunologists and furthermore, it contributed to disseminate the kinetic proofreading / kinetic amplification ideas to signal transduction, metabolic networks, or cellular receptors specificity.

Kinetic ideas are even spreading to still more improbable domains, the most surprizing extension being the discussion of oxygen photolysis by myoglobin in terms of the energy relay mechanism (W. Radding and G.N. Phillips, Jr, BioEssays, 2004, volume 26, pages 422-433). More interestingly, some physicists are beginning to investigate situations in which the original schemes cannot apply as such, and require fresh thinking (e.g., T. Tlusty, R. Bar-Ziv and A. Libchaber, 2004, Physical Review Letters vol. 93, 258103-1 to 258103-4).




++++++++++++++++++++ Note 1:++++++++++++++++++++++


Handwritten note on a 1965 conference abstract booklet, first as written in French, then translated here into English.

"Si la charge du sRNA a sa spécificité due aux enzymes, cela entraîne des conclusions précises concernant l'équilibre sRNA -- aa en présence ou en l'absence d'enzyme. En général, l' "orientation" des réactions par les enzymes signifie que l'on est dans un système hors d'équilibre, en général un "steady-state". Si tel est le cas, la charge in vitro non suivie d'incorporation dans les protéines n'a pas grand sens."

"If the loading specificity of tRNA is due to enzymes, this will have precise implications concerning the tRNA -- amino acid equilibrium in the presence or absence of enzyme. In general, the "channeling" of reactions by enzymes means that we are in an out-of-equilibrium system, in general a "steady-state". If this is the case, in vitro loading which would not be followed by incorporation into protein would not have much meaning."

+++++++++++++++++++ Note 2:+++++++++++++++++++++++

First model of ribosomal accuracy ([2], page 305). The greek letters in the original text have been replaced here by their spellings, "tau" for the sticking time of the codon-anticodon association, and "theta" for the period of a ribosomal oscillator.

"In general, one thinks of polypeptide elongation in terms of Eq. (4). As soon as a codon-anticodon association is formed, elongation proceeds. However, it is far from obvious that the ribosome should function in this way. One can imagine instead an oscillatory mechanism by which the ribosome checks at intervals of time theta the occupancy of the A site. Elongation proceeds if the check is positive. In this case, the important parameter controlling reliability in translation is theta/tau. When this parameter is large, the discrimination is described by Eq. (2), and the rate of polypeptide elongation is related to the equilibrium constants. When the parameter is small, the rate of polypeptide elongation is related to the rate constant of Eq. (4), and the fidelity in translation is described by Eq. (6). Intermediate situations are conceivable. A factor that would modify ribosomal kinetics by changing the ratio theta/tau would increase or decrease accordingly all the errors of translation".

+++++++++++++++++++ Note 3:++++++++++++++++++++++++++

The manuscript on ribosomal kinetics [3] was handled by Georges Cohen. Luigi Gorini was one of the reviewers, and the other one was, most probably, Paul Berg.

+++++++++++++++++++ Note 3a ++++++++++++++++++++++++

At the end of the manuscript on the 'timing effect", I discussed the case of proofreading by DNA polymerases, when the error was due to a tautomery. This is what I wrote:

"In the case where a replication error occurs through the insertion of the base analogue 5-bromo deoxyuridine, one may expect spectacular timing effects. The mutagenic effect of 5-BU is related to its tautomeric properties. Most of the time, 5-BU behaves like T, and some of the time it behaves like C. Now, consider that 5-BU has just been incorporated while in the C configuration, opposite to a G. It will remain in that configuration a time t1, and then shift to the T-like configuration. If it takes a time t to add the next base, the nuclease function will be really efficient during the time t2 = t-t1. If the various times are in suitable proportions, it is clear that a doubling in t may produce much more than a doubling in t2. With this in mind, we can interpret a very general observation which is apparent in the data of Drake et al. Suppose there is a point mutation in the T4 genome which reverses to wildtype with a frequency f1, when T4 has its normal DNA polymerase, and with a frequency f2 (f2 << f1) when T4 has a DNA polymerase of the antimutator type. Now, if the reversion frequencies are measured with cells grown in the presence of 5-BU, the f1/f2 is always much higher, by a factor of 10 at least.

"Now, consider another situation, of a kind which was studied by Howard and Tessman. There is a point mutation which one tries to reverse by 5-BU. In a series of experiments, one tries also to chase 5-BU by adding huge excesses of each of the four dNTP. Normally, one expects that if 5-BU is diluted with dT, the reversion rates will be reduced, and this conforms to the observations. However, there must also be a base which will essentially increase the reversion rate. This is the base whose incorporation determines the time during which the nuclease function is allowed to detect the micincorporation. From the fact that there are two pathways for 5-BU mutagenesis (although not equally used), there may be two different bases which presence in excess would help increase the reversion rates. This is again in conformity with the observations of Howard and Tessman. From a close examination of the situation one can then realize that this type of experiment may potentially permit to determine the exact nature of the two nearest neighbouring bases to the point mutation. That type of analysis would be extremely helpful in the studies on suppressor tRNA's."

+++++++++++++++++++ Note 4:+++++++++++++++++++++++++

The scheme with random binding of two substrates, and a strongly preferred pathway was inspired to me by an experimental result on an amino acyl tRNA ligase specific for valine. In the normal course of events, a binding of valine is required to expell from the enzyme the tRNA which was acylated in the previous cycle. So, there is an overlapping of two acylation cycles. But this, for obvious reasons, cannot be a strict requirement. Later, I interpreted this feature in terms of the "delayed escape of the product" scheme.

++++++++++++++++++++ Note 5:++++++++++++++++++++++++



PN/jl 23rd July 1974

Mr. Jacques Ninio

The Salk Institute for Biological Studies

San Diego

California 92112


Dear Mr. Ninio,

With regret I am returning the manuscript you submitted
to 'Nature.' In the opinion of our referee it was not of
sufficient novelty in concept to warrant our finding the room
to publish it. As you will appreciate, we have tremendous
pressures on our publication space and often have to make
this kind of decision. It does not imply any criticism of
your work. Thank you for submitting it to us in the first

Sincerely yours



Peter Newmark

Biology Manuscript Editor

In reply please quote: N0607 PN/jl


++++++++++++++++++++ Note 6:++++++++++++++++++++++++


November 5th, 1974. (--) "A 17h30 j'essaye d'appeler Orgel. Je l'obtiens à 19h. On évoque plusieurs choses: mon retour au Salk (il propose Août - septembre) - un gars de chez Bell est arrivé aux mêmes conclusions que moi sur la spécificité enzymatique, et a un papier sous presse. Son idée sur les feuillets béta se confirme. Enfin il me donne le nom d'un produit utilisé pour faire une réaction sur papier sur les fingerprints d'hémoglobine."

November 6th, 1974. "A 10 heures, je retrouve Françoise Bernardi à la rue Pierre et Marie Curie. On va à la bibliothèque pour quelques minutes. Je vois un numéro de J. Mol. Biol. en date du 14 sept. où je vois deux articles l'un théorique et l'autre expérimental sur les cinétiques et les erreurs pour les polymérases de T4. Je n'éprouve pas trop de dépit d'avoir été "grillé". (---). "Après examen attentif des deux papiers j'atteins la conclusion provisoire que les auteurs sont passés à côté de l'essentiel. Rentré à Port-Royal, je décortique encore les papiers, et je m'aperçois que les auteurs ont fait une bourde monumentale, qui invalide complètement leur travail. Je m'endors satisfait."

November 21st. (--). "Je travaille d'arrache-pied au traitement de la DNA polymérase, sur un point épineux laissé en suspens la veille, et finis par le résoudre. Cette fois-ci, le terrain semble complètement déblayé."

Dec. 11th 1974. "J'achève les dernières corrections au MS sur l'amplification cinétique." (---). " Alors que je viens d'agrafer les xerox de mon MS, je reçois dans le courrier un preprint ayant pour sujet exactement le même que celui de mon MS: amplification cinétique des discriminations moléculaires au moyen de délais. Orgel m'avait signalé l'existence du travail, et c'est par lui qu'Hopfield a dû avoir mon adresse.".

Dec. 12th, 1974. "Ultimes retouches au MS, et je vais le porter au Collège de France pour la revue Biochimie". (--). "Je pioche le manuscrit de Hopfield, qui commence à révéler ses faiblesses".

Dec. 14th, 1974. (---). "J'analyse encore le MS de Hopfield qui me paraît entièrement correct".

Dec. 16th, 1974."Je pioche encore le MS d'Hopfield, qui m'apparaît de + en + équivalent au mien."

Dec. 30th, 1974. (--). "Dans le courrier, une lettre de Hopfield: son papier est déja paru dans PNAS, ce qui me rend maussade pour le reste de la journée".

Jan. 6th, 1975. "J'expédie le commentaire à Lodish et Nature; j'envoie à Hopfield une copie de mon papier de Roscoff."

Jan. 9th, 1975. (---). "L'après-midi, au courrier, un gros papier d'un Suédois qui lui aussi a pensé à l'amplification cinétque. Il cite Hopfield." (Clas Blomberg)

Jan 14th, 1975. "Le matin, coup de fil au labo: c'est Hopfield qui est à Paris, et propose de paser me voir le lendemain matin. Je dis <<great!>>. Toute la journée j'essaye de joindre de Gennes pour demander des tuyeaux sur Hopfield - en vain.

Jan 15th, 1975. "Hopfield arrive à 10 heures. Tout de suite les nuages se dissipent. Il ne me cherche pas querelle. On discute paisiblement. Karplus descend (il connaît Hopfield). Vers 11h, je conduis Hopfield chez Karplus".

Jan 17th, 1975. "Je reçois deux lettres. Une de Lodish, qui répond à mon commentaire sur son papier. Ou bien il est nul, ou bien il y va au culot. L'autre de Hopfield, du 6 janvier, très sympa, qui discute de mes déboires avec "Nature".

Feb 3rd, 1975. "Une lettre de David Galas avec deux preprints, dont l'un sur un traitement cinétique de la DNA polymérase. Je me mets aussi sec à rédiger une lettre de réponse".


+++++++++++++++++++ Note 7:++++++++++++++++++++++++++


Letter from John Hopfield, dated December 17th, 1974

Dear Dr. Ninio,

I was sorry to see how closely your preprint follows the results presented in my paper in the PNAS which came out in the first week of November, 1974. It is amazing how parallel the lines are which we were following. I presume you will modify your article before publishing it in accomodation to my prior work.

Very truly yours,

John J. Hopfield

Enc. (the PNAS article).


Letter to John Hopfield, Dec. 30th 1974

Dear Dr Hopfield,

Thank you for your letter of Dec. 17th. You were much more fortunate than I from the point of view of publication. My manuscript was sent to "Nature" in June, and rejected for lack of "sufficient novelty in concept". Since April, I gave several talks : in Yale (to Dr Söll's group) in Orleans, in Paris (at a symposium on Automata Theory in Biology), in Roscoff, at a summer-school (held in May) on the Evoluiton of Biological Macrmolecules. The proceedings, in French, of that meeting are in press, and should appear within a year. The manuscript of which I sent you a xerox was submitted to Biochimie in December (the November issue of PNAS has not yet reached us owing to a long strike of the post-office employees). If the manuscript is accepted, I will add a note in order to mention your work.

I think we should not quarrel over questions of priority. I have been convinced, reading the details of your paper that your work was done in total ignorance of mine. On the other hands, I gave enough talks to international audiences and sent reprints to enough people since June to make it clear that I have not been influenced by your work.

In both your paper and mine, there is an oversight. The squaring effect holds for very short sticking times, but not for extremely short ones! How far it extends towards short sticking times depends on the quality of the irreversibility which is artificially introduced. I would be very glad to discuss the subject further with you. Do you have the opportunity to visit France ? I may be in August-Spetember in the U.S.A. and would try to see you then.

With best regards,

Yours sincerely,

Dr. J. Ninio


Letter from John Hopfield, dated January 6th, 1975

Dear Dr Ninio,

Nature! In hinsight I supppose it is not surprising that a short communication would be rejected by a quick referee. I have found that it takes about a quarter of a seminar to convince an audience that there is a problem at all, and half a seminar more to convince them that there is a new idea here. Only after that does an audience becomes at all enthousiastic. I feel badly about the refereeing situation with respect to Nature, and hope that the Biochemistry article meets with appropriate interest. (Actually, one reason I was able to finish my paper quickly was that I was spared the necessity of writing down the results of some earlier considerations by your paper whch had just come out in JMB.) If there is anything I can do to be helpful vis a vis the publication of the Biochemistry article, e.g., write to the editor, I will be happy to do so.

By good fortune, it happens that my oldest daughter is taking her Senior year in High School in Rennes, and the rest of the family leaves for France three days from now to visit her and to spend a few days of vacation. As a result, I shall be in Paris for much of the week of January 13-17, and will call you early in that week to arrange a time for a visit. I look forward to the opportunity to get together,

Sincerely yours,

John J. Hopfield.

++++++++++++++++++++ Note 8:++++++++++++++++++++++



Letter from David Galas


Lawrence Livermore Laboratory

Biochemical Division

January 23, 1975


Dr. Jacques Ninio

Laboratoire de Biochilmie du Développement

Institut de Biologie Moléculaire du C.N.R.S.

Université Paris VII - Touir 43

2 Place Jussieu, 75005



Dear Jacques:

Although I received your letter of December 28, with your discussion of the "evolution of accuracy", more than a week ago, I just received the two manuscripts today. Thank you very much for all this. I am intending to send you a long letter concerning the points you discussed in your letter, but I am prompted to write you now about the manuscript, "Kinetic amplification od Enzyme Discrimination".

I was quite surprised to see that you have given quite a bit of attention to the problem of the accuracy of DNA polymeriszation, in particular, to the very interesting T4 system. By extraordinary coincidence I have, with my colleague Elbert Branscomb, been concerned with a theoretical treatment of the accuracy of the T4 DNA polymerase for several months. We had several ideas for a stochastic model of the enzymatic actions during replication which were evolving into a coherent manuscript in November when the two papers by (---) appeared in JMB. We were immediately struck by two ideas. First, the experimental data they present should be very useful in guiding a theoretical effort to unravel the intricacies of the polymerase functions (it is unfortunate, however, that they did not use homopolymers as templates, though this is not serious). Second, the theoretical treatment attempted by this group (---) is so misguided and full of errors as to constitute a negative contribution to the field. We had arrived at much the same criticism of their treatment as you express in your manuscript (p. 21). In addition to their confusion between the net excision probability, and the single step excision probability, and their attempt to identify an algebraic identity with their particular 'model', they apparently have some serious conceptual difficulties with the idea of "peelback", for their computer simulation is wrong in some way. The results they present in Figure 4 are quite impossible.

In any case, we have since proceeded to formulate and solve the sequential, stochastic (that is, random walk) treatment of the enzyme function, and to propose a simple molecular kinetic model which defines the single step probabilities of this random-walk.

Two weeks ago we completed a brief note disucussing the basic ideas of the model in a simplified fashion. I have enclosed a copy of this paper. The theoretical tretament of the random-walk problem with an analysis of Bessman's data, and the detailed treatments of the kinetic model are the subjects of manuscripts which are nearing completion. I will send you copies of these as soon as they are finished.

Needless to say, I am most interested in the work you have done along similar lines. I am particularly intrigued by the reference in your manuscript to work in progress by yourself and F. Bernardi, which is to include an experimental study. The model we are investigating for enzymatic action of the polymerase leads, of course, to predictions concerning the dependence on substrate concentrations of the various random-walk parameters, and I am not aware of any data yet available which bears on this issue. I am pleased to realize that there seems to be a substantial basis for future cooperation between us in several areas both theoretical and experimental.

In any case, I wanted to inform you of our efforts on the specificity in DNA synthesis problemss, and perhaps elicit further comments from you on this question. I will be writing to you soon concerning accuracy in translation.

I have also enclosed a maunuscript (to be published in Nature) you may be interested in, reporting our initial set of experiments on the possiblity of an error feedback effect.


With best regards,


David J. Galas


Enc. (2)



Letter to David Galas

(Institut de Biologie Moléculaire,


Paris, le 5 February 1975

Dr David Galas

(Lawrence Livermore Laboratory,


Dear David,

I have received your Jan 23rd letter yesterday. Here are some details on my involvement in the DNA polymerase business.

After completing the numerical fitting of Gorini's data (Jan-Feb. 73) I looked for other situations wherein changes in specificity were observed and could be eventually linked to non-specific changes of a kinetic parameter. Leslie Orgel drew my attention towards the T4 system, and I found just the kind of data that was needed in Hall & Lehman's paper (they studied the dependence of the incorporation errors on the concentration of the dNTP's). I then made a mathematical treatment of Hall & Lehman's results and sent it to my original lab. in France where five of my colleagues were doing various experiments on DNA and RNA polymerases. They failed to see the interest of the thing. The point which excited me most in these calculations was a quadratic effect which appeared when the error dealt with was a tautomery. This possibility led me to the notion of delayed reactions (Dec. 73 - Jan. 74). At that time, I did a lot of calculations on the relationship betwwen accuracy and the mechanism of the reaction, and again, my interest was displaced. I thought that I had made a significant progress by finding with Leslie Orgel the "probabilistic method" for treating enzyme kinetics, and I attempted to develop that bit. However, when I came back in Paris, I learnt that the method had been found much earlier by the Russian group of Knorre. The advantage of this method is in some cases fantastic. Considering your kinetic scheme, you can lay down the equation for the elementary excision probaility qi without any intermediate calculation:

qi = (---)

pi = (---)

The manuscript on kinetic amplification (with a qualitative discussion of the T4 system, but without reference to Bessman's latest results) was submitted to Nature in June, and rejected for "lack of sufficient novelty in concept". I then wrote my review on accuracy in translation, in which I included the scheme for kinetic amplification. This review will appear, I hope, within a year.

When I became aware of Bessman's paper in November, I attempted to incite again my colleagues to work on the subject, without any success. However, I obtained a collaborator on the origins of life project and she was interested in the DNA polymerization experiments. I became then motivated to make a further theoretical effort, and solved (I hope) the random-walk problem in the simple case where there are only three independent parameters (q1, q2 and p1 = 1 - p2). The solution is valid for any values of the parameters for which there is, on the whole, a net elongation effect. Since your model, at the level of elementary probabilities is also a three-parameters model, I wonder if you agree with my Eq. (19):

P1 = p1(1 - q1)/[1 - K(1-q1)]

If you find something else, that would be exciting!

In November, I communicated on the phone with Leslie Orgel, and he informed me that someone at the Bell Institute sent him a preprint entitled; "Kinetic proofreading". I hurriedly rewrote my Nature paper and submitted it to "Biochimie". No news of it have reached me yet. I also wrote a note of comments (from a "correspondant en Biochimie évolutive") on the T4 system with special reference to the recent papers by Bessman and Goodman. This note should appear in the February or March issue of Biochimie.

John Hopfield's paper appeared in the October 1974 issue of PNAS. I have been in contact with him, and saw him in Paris; we are in excellent terms.

Experimentally, we are working with Kornberg's polymerase (we dont't have the T4 enzymes). Unfortunately, we have not yet succeeded in obtaining measurable misincorporations. We are limited by the availability of the tempate: poly(dC) - but now we have received enough of it to do the things systematically.

Theoretically, I did not go beyond the mathematical solution of the three-parameters model. I believe that there are various potential sources of complication, and it would be wise to do some experiments in order to know what the major source of troubles will be. For instance, the more I think about it, the more I believe that the result of Mc Clure & Jovin (see xerox) is correct. It makes sense if the strategy of the polymerase is designed to fight against tautomeries, and it explains somehow the dependence of the rate of the reaction on enzyme concentration (Cavalieri's PNAS paper) or, in the T4 system, that mutations in gene 32 may affect the accuracy of DNA polymerization (Drake's review in "Genetics"). But then, if part of the energy coming from the hydrolysis of the dNTP is used for kicking the enzyme away from the template, this introduces a complication in the analysiss of the dependence of rates on substrate concentrations.

I hope to hear from you soon.

With best regards,

Jacques Ninio




+++++++++++++++++++ Note 9:+++++++++++++++++++

About the reviews of the kinetic amplification article, in Biochimie;.

(To be written later).


+++++++++++++++++++ Note 10 :++++++++++++++++++

La Recherche, November 1977 issue, volume 83, pages 989-991.

La Recherche, November 1997 issue, volume 203, page 87

On the other hand, Maurice Guéron made a honest review in American Scientist (1978), volume 66, pages 202-208.

+++++++++++++++++++ Note 11 ++++++++++++++++++

English translation of the article by V.S. Schwartz and V.N. Lysikov (1974)

Doklady Akad Nauk SSSR, volume 217, pp. 1446-1448.

(To be included later).



+++++++++++++++++ REFERENCES ++++++++++++++++++++++



[1] Ninio, J. (1971) Codon-anticodon recognition : the missing triplet hypothesis. J. Mol. Biol. 56, 63-82.

[2] Ninio, J. (1973) Recognition in nucleic acids and the anticodon families. Progress in Nucleic Acids Res. Mol. Biol. 13, 301-337.

[3] .Ninio, J. (1974) A semi-quantitative treatment of missense and nonsense suppression in the strA and ram ribosomal mutants of Escherichia coli . Evaluation of some molecular parameters of translation in vivo. J. Mol. Biol. 84, 297-313.

[4] Ninio, J. (1975) La précision dans la traduction génétique. In Ecole de Roscoff 1974. L'évolution des macromolécules biologiques. (Sadron, C., ed.) pp. 51-68. CNRS, Paris.

[5] Ninio, J. (1975) Specificity in initiation of mRNA translation. Nature 255, 429-430.

[6] Ninio, J. (1975) Kinetic amplification of enzyme discrimination. Biochimie 57, 587-595.

[7] Ninio, J. (1977) Are further kinetic amplification schemes possible? Biochimie 59, 759-760.

[8] Bernardi, F. & Ninio, J. (1978). The accuracy of DNA replication. Biochimie 60, 1083-1095.

[9] Ninio, J. (1986) Fine tuning of ribosomal accuracy. FEBS Lett. 196, 1-4.

[10] Ninio, J. (1986) Kinetic and probabilistic thinking in accuracy. In Accuracy in Molecular Processes (Kirkwood. T.B.L., Rosenberger, R. & Galas, D.J., eds) Chapman & Hall, London, pp. 291-328.

[11] Ninio, J. (1987) Kinetic devices in protein synthesis, DNA replication and mismatch repair. Cold Spring Harb. Symp. Quant. Biol. 52, 639-646.

[12] Ninio, J. (1987) Alternative to the steady-state method : Derivation of reaction rates from first passage times and pathway probabilities. Proc. Nat. Acad. Sci. USA 84, 663-667.

[13] Ninio, J. & Chapeville, F. (1980) Recognition: the kinetic concepts. In Chemical recognition in biology (Chapeville, F. & Haenni, A.-L., eds.) pp. 78-85, Springer-Verlag, Heidelberg.

[14] Ninio, J. & Bokor, V. (1986) Stratégies d'adaptation moléculaire. La Vie des Sciences 3, 121-136.

[15] Ninio, J. (1975) Considerations on the problem of the joint evolution of two different translation apparatuses within the same cell. In Molecular biology of nucleocytoplasmic relationships (Puiseux-Dao, C., ed.) pp. 31-39, Elsevier, Amsterdam.

[16] "De notre correspondant en biochimie évolutive" (1975). La marche aléatoire d'une DNA-polymérase. Regards sur la Biochimie 1, 14-16.

[17] Ninio, J. (1976) Evolutionnisme et origines de la vie. La Recherche 7, 325-334.

[18] Dumas, J.-P. & Ninio, J. (1982) Efficient algorithms for folding and comparing nucleic acid sequences. Nucleic Acids Research 10, 197-206.