Life's Greatest Secret

by Matthew Cobb

  It seems very likely that one bit of potential structural information will not always transmit the same amount of information; rather, the efficiency of transmission will depend upon the context within which the performance is measured. … This is somewhat analogous to the relative difficulties of determining whether a symbol is 0 or 1, or to determining whether one should get married or not!43

  To understand messages, information theorists needed more than bits; they had to introduce context and meaning into their calculations. Quastler highlighted the problem by using the example of a conversation between two individuals. At first sight, it might appear that during a conversation, information is transmitted by words. This is undoubtedly the case, but many other factors are involved, such as the selection of particular words from various potential synonyms, the tone of voice, the timing of the utterance, speech volume and so on. Identifying all these levels of potential information seemed impossible. What appears to be a simple example of information transmission is in fact extremely complex. Quastler concluded: ‘In such a situation we have obviously no hope ever to obtain a precise, unequivocal, and incontestable measure of information content.’44

  Similar problems were encountered at the molecular level. Leroy Augenstine argued that it would be futile to calculate the amount of information in any chemical communication system, because ‘only a small fraction of the potential information on the surface of the molecule is actively utilized in information transfer.’45 Quastler’s summary of the situation undermined the idea that information was a strictly objective measure: ‘information applications are relative and not absolute; hence, any information measure associated with a given set of biological objects will depend on the set itself and on the scientist who does the estimating.’46 Paradoxically, this excess of potential information made it almost impossible to apply the concept in a biological context. As Quastler put it:

  Every kind of structure and every kind of process has its informational aspect and can be associated with information functions. In this sense, the domain of information theory is universal – that is, information analysis can be applied to absolutely anything. The question is only what applications are useful.47

  As Quastler recognised, concrete examples of such useful applications of theory were turning out to be few and far between. After the symposium had finished, a small group of participants met in the evening to discuss their impressions. A slightly depressed Quastler summed up the general feeling:

  Information theory … has not produced many results so far; it has not led to the discovery of new facts, nor has its application to known facts been tested in critical experiments. To date, a definitive and valid judgement of the value of information theory in biology is not possible.48

  If the leading lights of information theory were uncertain as to its usefulness, the idea of cracking the genetic code by simply applying it to protein sequences, as Gamow suggested, seemed unlikely to succeed.

  Biologists were also expressing doubts about the usefulness of information theory. In 1954, the zoologist J. Z. Young, who had talked extensively about information in his 1950 Reith Lectures, surveyed the field and concluded:

  In spite of much discussion of the application of information theory in biology there is still a considerable uncertainty about the status of the analogy and about its usefulness.49

  In 1955, Joshua Lederberg began a correspondence with von Neumann, in which the pair explored the similarities between von Neumann’s model of self-replicating automata and theories about the origin of life. The two men soon realised that each of them did not understand what the other meant by ‘information’, and Lederberg eventually concluded that this was because they were thinking ‘at very different levels’. For biologists, he argued, the ‘propagation, and evolutionary elaboration, of complexity is self-evident’ – they were interested in the detail of how such a system could work. The logician van Neumann, however, was ‘looking for the foundations of an axiomatic theory of reproduction’ – something much more abstract and not necessarily linked to biology at all. Summarising von Neumann’s views, Lederberg confessed that he ‘could not begin to say whether they would be helpful in genetic analysis.’ Although the two men agreed to meet to discuss the question further, von Neumann’s ideas had no perceptible effect on Lederberg’s biology.50

  Macfarlane Burnet encountered similar difficulties when he was writing Enzyme, Antigen and Virus. As he explained:

  This monograph was originally conceived as an attempt to develop something analogous to a communications theory that would be applicable to the concepts of general biology. However, it has not been found possible to make any serious use of the already extensively developed concepts of information theory in the strict sense.51

  The main reason that Burnet gave for his failure was that ‘only the most generalised sketch of an outline has yet been given of how information theory at the strict level can be applied to biology.’

  J. Z. Young solved the problem by deciding he was not in favour of considering organisms as examples of Shannon’s simple lines of transmission, but instead as something much more like a computer.52 This analogy was also explored by the Bristol psychologist Frank George at a 1959 Society for Experimental Biology symposium on ‘Models in Biology’. George argued that the main contributions of cybernetics were analogies and metaphors, particularly ‘the central importance of feedback – whereby organisms can be likened to, and described in the same manner as, inanimate systems’. But instead of using an approach based on information theory, George thought that ‘the analogy with computers, suitably modified and adapted, is a good one, and one that is essentially testable.’53 This kind of vague metaphor, no matter how supposedly testable, was far removed from the rigours of Wiener and Shannon’s equations.

  At the same time, the MIT electrical engineer Pete Elias tried to make sense of the multiple ways in which information theory had been transformed in the previous decade, in the realms of coding, communication and cybernetics. Although Elias was convinced of the relevance of Shannon’s ideas to electronic communication, like Shannon he was less certain when it came to the extension of information theory to other fields, including biology. In 1958, Elias made a plea to his fellow-theoreticians not to apply information theory willy-nilly to other disciplines.54 A year later, Elias expressed his doubts about using information theory to understand chemical specificity, arguing that in this field information was used ‘either as a language for the discussion of purely combinatorial problems or as a useful statistic, but … [not] in any coding sense which would imply that the informational treatment was at all necessary or unique.’ Elias was even more dismissive of the application of information theory to the genetic code: ‘Although informational ideas may be useful here,’ he wrote, ‘it seems unlikely that they are essential.’55 A consensus was emerging, without fanfare and without any public declarations: it was difficult, if not impossible, to apply information theory to biology, even to the genetic code.

  There was a lone exception: in 1961, the Japanese theoretical population geneticist Mitoo Kimura published an article entitled ‘Natural selection as the process of accumulating genetic information in adaptive evolution’. Writing when the exact nature of the genetic code was still unclear, Kimura showed how new genetic information arises, by focusing on the consequences of one form of a gene (‘allele’) being replaced by another form (this is the technical definition of evolution), through natural selection. Using some very shaky guesstimates, Kimura calculated that in the animal lineage leading to higher mammals, 108 bits of genetic information had accumulated since the Cambrian explosion around 540 million years ago, which saw the appearance of most groups of animals.

  Intriguingly, Kimura noted that it was probable that the human chromosomes contained around 1010 bits of information; he concluded that either the genetic code was highly redundant or the genetic information we have accumulated is a small fraction of that which we could store in our chromosomes. W
e now know that the human genome contains approximately 3 × 109 bases, or one-third of Kimura’s estimate. Irrespective of the validity of Kimura’s calculations, both of his explanations were correct, and he spent much of the rest of his career trying to understand the evolution of the bulk of the genome, which is apparently not immediately shaped by natural selection.

  Despite the fundamental role of the information metaphor in molecular genetics, and its vital position in Crick’s central dogma, the 1956 Oak Ridge conference marked the swan song of information theory’s influence on biology. By this time, what remained of the cybernetics group had drifted off into studies of psychiatry and human behaviour, which were probably the least likely to produce important results because of their complexity.56 The neurophysiologist Warren McCulloch, who had been present when Wiener gave his paper on ‘Behaviour, purpose and teleology’ back in 1942, turned down an invitation to the 1956 Oak Ridge meeting, arguing that it was not certain where information actually resided in biological systems, including in genes:

  I doubt whether information theory is yet properly attuned to the complexities of biological problems. To apply the theory in its present state except in a most rudimentary fashion we need to crack the code in genetics as surely as we do in the Central Nervous System.57

  Biological codes would not give up their secrets merely by being shown Shannon and Wiener’s fiendish equations. Experimentation, not theory, would be needed. For more than a decade, the twin theoretical approaches of information theory and cybernetics had beguiled researchers and entranced the general public with their promise of a new way of looking at the world, linking biology with the growing wave of electronic devices that were beginning to permeate all aspects of society, from warfare to work. But in the end, the concrete application of both of these theories seemed to have come to nothing. No new scientific disciplines were founded through the development of information theory or cybernetics – there were no new cross-disciplinary journals, no research institutes, no annual conferences, no new funding sources focused on this area.58

  That did not mean that cybernetics and information theory had no influence on the development of biology or on the cracking of the genetic code. They both had an impact, but not in the way in which their partisans might have hoped for. In 1961, Martynas Yčas accepted that ‘no explicit, and especially no quantitative use of information theory has … been made in practice.’59 Instead, as was suggested at the close of the 1956 meeting, Yčas thought that biologists found it ‘preferable to use information theory only in a semi-quantitative fashion’, as a metaphor.60 In this metaphorical form – infuriatingly vague and imprecise for mathematicians and philosophers but extremely powerful for biologists – information came to dominate discussions of the genetic code and its meaning, down to the present day. Although the cybernetics group had fallen apart, its fundamental ideas played an important role in the development of radical new views that changed the way in which scientists understood gene function and the meaning of the genetic code. This came about through the work of Parisian researchers influenced by the cybernetic approach.

  –NINE–

  ENZYME CYBERNETICS

  In the summer of 1947, the French bacteriologist Jacques Monod visited Cold Spring Harbor Laboratory, where Max Delbrück’s training course in phage genetics was in full swing. Monod was good-looking and dynamic, with thick hair combed back from a high forehead; he had fought in the French Resistance and was a charismatic figure.1 Monod later recalled that as he sat in on one of the lectures, there was a ‘short fat man’ in the front row of the audience, who seemed to be asleep. Monod thought that the man, ‘with his round face and fat belly, looked like a petty Italian fruit-merchant, dozing in front of his shop’.2

  The ‘fruit-merchant’ was the physicist Leo Szilárd, who in 1929 had made the link between information and entropy. In 1933, a Jewish refugee from Hitler, he came up with the idea of the nuclear chain reaction, which led him six years later to write a letter to President Roosevelt, co-signed by Einstein, calling for the US to develop atomic weapons. This initiative was at the origin of the Manhattan Project, in which Szilárd was heavily involved. But after the surrender of Germany in May 1945, Szilárd, like so many other Manhattan Project physicists, had a profound change of heart and argued vociferously against using the bomb to attack Japan. After the destruction of Hiroshima and Nagasaki, Szilárd turned away from physics. In middle age, he decided to study biology and soon found that his interests coincided with those of Monod.3

  Monod wanted to understand how bacteria responded to being placed on different kinds of food: since 1900 it had been known that bacteria were able to start producing an enzyme necessary to digest a particular sugar, even if they had never previously encountered that type of food. This phenomenon was known at the time as enzymatic adaptation, and no one understood how it worked; the bacteria seemed to be able to sense what was needed to survive in a particular environment. Apart from its intrinsic interest, adaptation was an extremely attractive system for anyone interested in working out exactly what genes did when proteins were synthesised, because it provided a tool for initiating the process under controlled conditions.

  When Szilárd and Monod were introduced at Cold Spring Harbor in 1947, Monod was surprised to learn that Szilárd had read all his papers. Szilárd fired off several ‘unusual, startling, almost incongruous’ questions, leaving the Frenchman bewildered but delighted at his new friend’s knowledge and interest.4 In 1954, Szilárd met Monod’s future collaborator, François Jacob, with exactly the same result. Jacob later recalled:

  At our first encounter, at a colloquium in the United States, he led me over to a corner to ask me about my work. At each response he cut in to reshape my answers to suit his style, to force me to speak his language, to use his words, his expressions. He carefully noted each answer in a notebook. At the end, he said, ‘Sign there!’ Two years later, during another encounter, he asked ‘Is what you told me a few months ago still true?’ And he noted: ‘Still true!’5

  Monod’s study of adaptation focused on how bacteria grown on the sugar lactose were able to synthesise an enzyme called β-galactosidase (the ‘β’ is pronounced ‘beta’), which could break down lactose. Understanding this response proved difficult: Monod and his team discovered that other sugars, which could not be broken down by β-galactosidase, nevertheless acted as β-galactosidase inducers. This meant that bacteria produced β-galactosidase even when there was no lactose present, and the enzyme therefore could have no function.6 Because of this troubling result – bacteria could be induced to produce apparently useless enzymes – the term adaptation was abandoned and replaced by the more neutral word induction.* From 1953 these proteins were therefore called inducible enzymes.7 The situation soon became more confusing: within a few years, Monod had created mutant strains of bacteria in which, just as different inducers could lead to the production of the same enzyme, so, too, different enzymes could be induced by the same inducer – lactose. New data were making induction more complex, not less.

  *

  Crick’s 1957 ‘central dogma’ lecture emphasised that the race to decipher the genetic code was closely intertwined with the attempt to understand protein synthesis. Although neither Monod nor Szilárd was involved in the coding problem, they were both convinced that studies of bacteria would provide insights into how genes function, by understanding the main thing they do, which is to enable the cell to produce proteins. At a conference in 1952 Szilárd and his Chicago colleague, Aaron Novick, described their hypothesis for how protein synthesis was controlled, focusing on how the cell knew when to stop synthesising a particular amino acid: ‘somehow the increased concentration of each amino acid depresses the rate of the individual steps of synthesis leading to the formation of that amino acid.’8 As more of a particular amino acid is produced, the rate at which it is synthesised slows down. Novick and Szilárd thought that protein synthesis involved a negative feedback loop, just like those seen in t
he cybernetic devices studied by Wiener a few years earlier.

  In 1954, Szilárd explained his idea to Monod. The Frenchman later admitted he found it ‘a rather startling assumption’ and did not agree. This was surprising, because a year earlier Monod had shown that the biosynthesis of some enzymes was suppressed by their respective end-products – a negative feedback loop – but had been unable to explain his finding.9 It took several years for Monod to realise the significance of what he had discovered. By the end of the decade, Szilárd’s idea loomed large in Monod’s thinking, eventually influencing how we understand gene function.

  Other scientists were also looking at protein synthesis in bacteria and were groping towards the same kinds of interpretation as Novick and Szilárd. In 1954, Richard Yates and Arthur Pardee, from the University of California at Berkeley, described an investigation into the biosynthesis of the pyrimidines uracil and cytosine in Escherichia coli bacteria. They found that once pyrimidines began to appear in the cell, their presence led to a decrease in the levels of the enzymes involved in the biosynthetic chain. Their 1956 report of their experiment concluded:

  Inhibition by an end-product of its own synthesis appears to be a common control mechanism in the cell.10

  Although Yates and Pardee had clearly described a negative feedback mechanism, they used the term ‘feedback’ only in the title of their article, and beyond asserting that such systems were widespread in the cell, they did not explore the matter any further. At about the same time, Edwin Umbarger of Harvard Medical School used negative feedback to interpret a study of E. coli in which the presence of the end-product of a biosynthetic pathway inhibited its own biosynthesis. In 1956, Umbarger published an article in Science that began with a statement that revealed the influence of Wiener’s cybernetics on the new generation of scientists known as molecular biologists: