Crick's Adaptor Hypothesis and the Discovery of Transfer RNA: Experiment Surpassing Theoretical Prediction

1.1 Predictive versus Accommodation Hypotheses and Theories

Philosophers of science distinguish between two types of hypotheses and theories. Hypotheses of one category predict novel entities or facts prior to their empirical discovery, whereas theories of a different type accommodate and explain previously gathered data. It is generally agreed that the capacity of predictive hypotheses to forecast novel facts makes them aesthetically more pleasing than accommodation hypotheses that explicate extant knowledge. However, thinkers disagree on the validity of a broader predictivist thesis (also dubbed predictivism; Maher 1988), which argues that prediction of novel facts provides greater epistemic certitude than does accommodation of known data. Whereas many thinkers have advanced various versions of the predictivist idea, a sizable school of thought has also rejected the notion that predictions have epistemic advantage over accommodation of data.

First raised by Renaissance astronomers, predictivist ideas were articulated in the seventeenth century by Descartes, Huygens, and Leibnitz (Giere 1983) and expanded and refined by thinkers of the nineteenth to twenty-first centuries (for selected writings in support of predictivism see: Achinstein 1990, 1994; Lipton 2004 2005; White 2003; Douglas and Magnus 2013; Kahn et al. 1992; Lange 2001; Worrall 2014; Maher 1993; Barnes 1996; McIntyre 2001; Hitchcock and Sober 2004; Gardner 1982; Whewell 1860; Popper 1963; Lakatos 1978; Hempel 1966; Duhem 1991). On the other hand, opposing scholars have raised substantial arguments against the predictivist thesis (see for instance: Collins 1994; Harker 2008; Howson 1988; Howson and Franklin 1991; Achinstein 1994; Scheffler 1957; Keynes 1921; Mill 1843).

1.2 Putting Predictivism to Test: Cases from the Histories of Physics and Chemistry

A historical approach to the philosophy of science tests the legitimacy of philosophical theses by examining their validity in actual cases from the history of science (Laudan et al. 1986; Brush 2015). Employing this method, historians of physics and chemistry put the predictivist thesis to test by comparing the relative impacts that novel predictions and accommodation hypotheses had on practicing scientists¹ (Barnes 2005; Brush 1989, 1990, 1993, 1994, 1996, 2015; Earman and Glymour 1978; Gardner 1982; Howson and Franklin 1991; Lipton 2004; Maher 1988; Scerri 2005; Scerri and Worrall 2001; Worrall 2005; Zahar 1973). In appraising most of these studies, Brush concluded that despite their intuitive appeal, predictions of novel facts did not impact physicists or chemists to greater extents than did explanatory theories:

A successful prediction may yield much favorable publicity for a theory (including statements that call attention to the novelty of the phenomenon predicted) and thereby force other scientists to give it serious consideration. But subsequent evaluation of the theory in the technical literature do not seem to give greater weight to the prediction of novel facts than to persuasive deductions of known facts. (Brush 1989, 1127)

1.3 Complexity of Biological Systems Thwarted Predictions of Their Workings

Successful predictions in biology are much less frequent than in the physical sciences. Predictive hypotheses on the workings of complex biological systems have generally failed because of their inevitable blindness to yet unidentified constituents and effectors of the systems. Evolutionary biologists (Scriven 1959; Mayr 1985, 1988) pointed out that because biological systems are shaped by multiple and largely unknown causes and effects, their behavior is hard or impossible to predict. This point was cogently articulated by Ernst Mayr (1988):

A belief in universal, deterministic laws implies a belief in absolute prediction. The ability to predict was therefore the classical test of the goodness of explanation in physics. In biology, the pluralism of causations and solutions makes predictions probabilistic, if it is possible at all. Prediction in the vernacular sense, that is, the foretelling of future events, is as precarious in biology as it is in meteorology and other physical sciences dealing with complex systems. […] As Scriven (1959) pointed out, the ability to predict is not a requirement for the validity of a biological theory. (Mayr 1988, 19–20)

Historians of science have assessed the relative contribution of novel predictions in forming knowledge in selected areas of biology. One carefully studied case was Morgan’s chromosome theory of heredity (CTH) (Morgan et al. 1915). Analysis showed that American and British geneticists adopted this theory mostly because of its ability to explain empirical data on Mendelian and sex-linked inheritance, on disjunction, and on linkage groups. These accommodations of empirical data impacted scientists to a greater extent than the confirmation of the CTH predictions (Brush 2002, 2015, 397–412).

Similar assessments revealed that it was empirical evidence and not predictions that convinced evolutionists in the 1950s that natural selection drives evolutionary change (Brush 2009, 2015, 442–464). Yet, because evolution and natural selection are multidimensional processes, they are unlikely to yield a definitive picture of the relative contributions of predictive versus accommodation hypotheses.

Population geneticists weighed the success of hypotheses that predicted fluxes in populations of Drosophila. It was initially claimed that measurements of past population dynamics of three species of Drosophila produced statistical estimators that predicted with high confidence future fluctuations in these populations (Poole 1976, 1978). It was pointed out, however, that such predictions can succeed only if past and future conditions remain the same. More importantly, even if statistical models could correctly predict future population fluxes, they would not provide any insight into the fundamental biological basis of either past or future population changes (Anderson 1978b, 1978a). A more recent attempt was made to forecast future morphological evolution of stick insects on the basis of long-term observations and experiments. Results indicated, however, that predicting future morphological changes on the basis of past data required a yet undiscovered full understanding of the biology of the organism (Nosil et al. 2018).

Because biological systems appeared to be less complex at the molecular level than at the organismal or population levels, it could be thought that their working would be easier to predict. Indeed, some major theories effectively delineated fundamental molecular-level principles of biology. Such was the Watson and Crick (1953) proposition that complementarity of replicated DNA strands was the key to faithful gene inheritance. Other grand theories were Crick’s (1958) sequence hypothesis and his central dogma, and the operon model of regulation of gene expression (Jacob and Monod 1961). However, in contrast to these productive global theories, hypotheses that attempted to envisage physical and functional details of particular multi-component bio-molecular systems were consistently unsuccessful. Thus, for instance, theories were wide off the mark in picturing the makeup and modes of function of the machineries of protein synthesis or of non-lysosomal degradation. Similarly, theories did not contribute to the elucidation of the biosynthesis of DNA or RNA or to the discovery and dissection of the RNA splicing apparatus. Failures of predictive conjectures in these and similar cases were primarily due to their inevitable ignorance of yet undiscovered constituents of the studied systems. It was ultimately left, therefore, for stepwise experimentation to identify components of such systems and to elucidate their workings. Thus, for instance, because messenger RNA was not envisaged and searched for in the 1950s, the dominant theory of the time erroneously maintained that it was ribosomal RNA which mediated transmission of genetic information from DNA to the protein synthesis machinery (Fry 2016, 365–370). Similarly, theories made wrong predictions on the mode of non-lysosomal degradation of proteins (Fry 2018), the mechanism of translation of nucleic acid into protein, or the manner of antibody formation (Fry 2020). Ultimately, these and other complex bio-molecular systems were elucidated by experiment and organization of the gathered data into accommodation theories. In failing to provide effective forecasts of the workings of complex biological systems, predictive hypotheses therefore held no epistemic value.

1.4 The Exceptional Case of Crick’s Prescient Adaptor Hypothesis

Francis Crick’s so-called ‘adaptor hypothesis’ presents a rare case of a confirmed abstract conjecture in biology. This astute hypothesis stands in stark contrast to previously reviewed theories whose conceptions of the constitution and mode of operation of diverse complex bio-molecular systems were incorrect (Fry 2016, 365–370, 2018, 2020). Crick’s hypothesis warrants the special analysis of this paper because it is singled out by its exceptional success from the habitual failures of other theoretical predictions about the working of other systems. In his adaptor hypothesis Crick prefigured the existence of amino acid-specific RNA adaptors and of specialized enzymes that charge them with particular amino acids. He also conjectured that the amino acid carrying adaptors bind to complementary protein-coding RNA template and their liberated amino acids are incorporated into a growing polypeptide chain (Crick 1955). When Crick introduced and disseminated his hypothesis in 1955–57, he was entirely unaware of concomitant, closely pertinent, experimental findings of the Harvard biochemists Mahlon Hoagland and Paul Zamecnik (Crick 1988, 94). These investigators discovered the actual adaptors—transfer RNA (tRNA) molecules—and identified enzymes that activate and charge the amino acids onto cognate tRNA. Based on these empirical data, they also formulated a skeleton theory of protein synthesis that shared many similarities with Crick’s hypothesis. Curiously, while Crick was unaware of their findings, Hoagland and Zamecnik also had no knowledge of his adaptor hypothesis as they were carrying out their experiments (Hoagland 1959, 93–94; Zamecnik 1960).

The historically rare case of Crick’s abstract hypothesis is especially instructive because of Hoagland and Zamecnik’s unknowing, concomitant, affirming experiments. Intriguingly, historical evidence shows that despite its striking brilliance, Crick’s hypothesis was mostly ignored or distrusted by contemporaneous researchers. By contrast, the experimental findings of Hoagland and Zamecnik had immediate and enduring impact on practicing scientists. It was, therefore, the empirical discoveries of Hoagland and Zamecnik and not Crick’s celebrated hypothesis that exposed the scientific community to the mechanism of translation of protein-coding nucleic acid.

2.1 Birth of the Hypothesis: Historical Background

Crick’s interests in the mid-1950s encompassed protein structure (Crick and Rich 1955; Rich and Crick 1961, 1955) and the spatial organization of coat proteins of RNA viruses (Crick and Watson 1956; Crick and Watson 1957). A third, and arguably his most important, area of inquiry was the nature and properties of the nucleic acid genetic code and how it encodes proteins. Crick’s earliest direct contribution to this field was his critical evaluation of an abstract model of the structure and translation of the code that had been proposed by the theoretical physicist George Gamow. Crick dismantled most of Gamow’s propositions in an informal paper that he circulated in 1955 among a few interested scientists. In its place he offered his so-called adaptor hypothesis that was independently confirmed by the parallel experimental work of Hoagland and Zamecnik. In hindsight, Gamow’s incorrect model served, therefore, as a useful, disposed scaffolding for Crick’s more accurate model.

The Gamow model. Starting in the 1940s, early speculations linked protein synthesis to nucleic acids (Fry 2016, 277–280). This insinuated linkage gained a sharper focus when Watson and Crick opined in 1953 that: “It seems likely that the precise sequences of bases [in DNA] carries the genetical information” (Watson and Crick 1953). This terse statement provoked George Gamow, a physicist with no direct experience in biology, to consider ways in which base sequences in nucleic acid encode ordered arrangements of amino acids in proteins. Starting a correspondence with Watson and Crick in the mid-1953, Gamow aired ideas on the possible nature of the code and its probable mode of translation into proteins.² The core question with which Gamow grappled in the first half of the 1950s was later enunciated by Crick and associates (Crick et al. 1957): “The problem of how, in protein synthesis, a sequence of four things (nucleotides) determines a sequence of many more things (amino acids) is known as the coding problem.” In his first attempt to answer this question Gamow constructed a theoretical model of a code based on stereo-chemical fit between amino acids and inter-nucleotide spaces in a DNA template. He first described this model in an October 22, 1953 letter to Linus Pauling (Gamow 1953) and then delineated his ideas in two formal publications (Gamow 1954a, 1954b). In this model DNA served as direct protein-encoding template that accommodated stereochemically fitting amino acids in inter-nucleotide clefts. Specifically, Gamow postulated that amino acids inserted into hypothetical, diamond-shaped niches between two adjacent base pairs in the DNA double helix. Allegedly, specific combinations of three out of the four nucleotides in two adjacent base pairs dictated specific contours of each particular cavity.³ Combinatorially, there were twenty possible different base triads and twenty different contours, a number that conveniently matched the twenty different amino acids. Gamow’s model further purported that after becoming embedded within their specific cavities in the DNA, the amino acids bonded to one another to form a polypeptide chain. (For more detailed descriptions of the model, see Olby 2009, 223–228; Fry 2016, 280–284.)

Gamow’s theory endowed the code with several attributes. First, because every base pair in the DNA was a part of two bordering cavities, it was also a part of two code words (later named codons). Each base pair in such an overlapping code dictated the identity of two adjacent amino acids. This feature of the code limited the number of possible permutations of two contiguous amino acids. Another feature of the Gamow code was its lack of polarity such that a string of code words would be translated at equal probability to the polypeptide chain A-B-C-D- and its reverse D-C-B-A-. Last, this model raised a prospect of a degenerate code in which several different code words encode the same amino acid.

Heeding some of Crick’s criticisms of his model (section 2.2), Gamow later replaced DNA with RNA as the direct template, professing that amino acids lodged in presumed clefts in the secondary structure of single-stranded RNA. He also conducted computer-aided calculations in an attempt to define code words. Although his efforts to solve the code problem were unsuccessful, Gamow made a historically notable contribution by creating the so-called RNA Tie Club: a virtual think tank dedicated to elucidation of the genetic code. Twenty members of the Club—biochemists, physicists, and mathematicians—stood for the twenty amino acids, while four ‘honorary’ members represented the 4 nucleotides. The habitually traveling Gamow kept the geographically scattered club members abreast of new, real or imagined, developments in the pursuit of the code. In addition, he encouraged members to circulate informal papers with ideas on the code. Crick first introduced his adaptor hypothesis in one such internal, never-published manuscript (Crick 1955), which is considered now to be the most famous unpublished paper in the annals of molecular biology.

2.2 Emergence of the Adaptor Hypothesis

In December of 1953 Gamow met for the first time with Crick who was then working at Brooklyn Polytechnic.⁴ Their two-hour animated discussion revitalized Crick’s enthusiasm for the code problem (Olby 2009, 221). Thinking about the code problem for the rest of his sojourn in the United States, Crick became increasingly critical of the idea of stereo-specific fit between amino acids and cavities in a DNA template. Instead, he came up with an alternative model of adaptor molecules that specifically bind amino acids and direct them to particular code words in the nucleic acid template. The adaptor hypothesis was hatched in August 1954 while Crick was still in the United States (Judson 1996, 284), and it was fully articulated in an informal manuscript “On Degenerate Templates and the Adaptor Hypothesis”⁵ that Crick wrote after his return to Cambridge, in the winter of 1954–55. Circulated among members of the RNA Tie Club in early 1955,⁶ the paper critiqued the Gamow model, raised ideas on some principal features of the code, and most notably introduced the adaptor hypothesis. A timeline of the launch and subsequent dissemination of the hypothesis is charted in figure 1 (top line).

Figure 1:

Timelines of the introduction and dissemination of the adaptor hypothesis (upper line) and the parallel development of a cell-free protein synthesis system, and discoveries of amino acid activation, activating and charging enzymes, and tRNA (lower line). Marked along the axes are dates of conference presentations, submission of key papers, and significant meetings. Precise dates of these events are detailed in the body of the text and in footnotes.

Crick opened his RNA Tie Club communication with epigraph from Qabus-Nama (“A Mirror of Princes” in Persian): “Is there anyone as utterly lost as he that seeks a way where there is no way?” This quote from the rather obscure ruler of the Ziyarid dynasty and poet Kai Kāvus ibn Iskandar (1021–1099), seemed to reflect a pessimistic sense of dead end in the pursuit of the code. In point of fact, however, the paper presented ideas that significantly advanced the definition of the nucleic acid code and its mode of translation to proteins. The following sections describe Crick’s critique of basic tenets of the Gamow model and his alternative adaptor hypothesis. Not covered are parts of the paper that dealt with overlap, degeneracy, and directionality of the code.

2.2.2 The adaptor hypothesis as an alternative to the Gamow model

Crick’s fifth, and most central objection to the Gamow model concerned its underlying structural assumptions. Finding the idea of direct stereo-chemical fit between DNA template and amino acids to be groundless, he replaced it with a conjecture of adaptors.

Crick discounted Gamow’s structural assumptions on two accounts. First, an amino acid in a polypeptide chain matches by size only a single base pair in the common B conformation of DNA and not two as Gamow posited. Crick’s (1955) second and more important point was that there was no structural basis for the presumed stereo-chemical fit between amino acids and differently shaped inter-nucleotide cavities in the DNA template:

I cannot conceive of any structure (for RNA or DNA) acting as a direct template for amino acids, or at least as a specific template. In other words, if one considers the physical-chemical nature of the amino acid side chains we do not find complimentary [sic] features on the nucleic acid. Where are the knobby hydrophobic surfaces to distinguish valine from leucine and isoleucine? Where are the charged groups, in specific positions, to go with the acidic and basic amino acids? (7)

Dismissing the idea of direct stereo-specific interaction of amino acids with a nucleic acid template, Crick presented instead an alternative adaptor hypothesis. Under this hypothesis adaptor molecules mediated indirect interaction between the amino acids and template (fig. 2). In essence, Crick proposed that specialized enzymes charge amino acids onto cognate adaptor molecules. The charged adaptors next form complementary hydrogen bonds with corresponding code words in the nucleic acid template and release their amino acids to be joined to a growing protein chain. Thus, Gamow’s notion of specific direct spatial fit between amino acids and nucleic acid was replaced by an idea of specific hydrogen bonding of adaptor to template:

What the DNA structure does show (and probably RNA will do the same) is a specific pattern of hydrogen bonds, and very little else. It seems to me, therefore, that we should widen our thinking to embrace this obvious fact. (Crick 1955, 7)

Figure 2:

Illustration of Crick’s original adaptor hypothesis (Crick 1955). Shown are three out of at least 20 different hypothetical RNA adaptors, each one comprising three and a few additional nucleotides. Specialized enzymes were proposed to ‘chemically join’ each amino acid onto its particular adaptor that then hydrogen bonded to complementary code word in an RNA template. Crick’s original hypothesis did not specify how the amino acids detached from the adaptors and joined growing polypeptide chain.

Expounding in full his adaptor idea, Crick further wrote:

each amino acid would combine chemically, at a special enzyme, with a small molecule which, having a specific hydrogen-bonding surface, would combine specifically with the nucleic acid template. […] In its simplest form there would be 20 different kinds of adaptor molecule, one for each amino acid, and 20 different enzymes to join the amino acid to their adaptors. Sydney Brenner, with whom I have discussed this idea, calls this the “adaptor hypothesis,”⁸ since each amino acid is fitted with an adaptor to go on to the template. (1955, 8)

Interestingly, Crick did not stipulate in his 1955 hypothesis how the adaptors liberate their amino acids to be linked to growing polypeptide chain. However, he did insightfully speculate that the number of different adaptor molecules may be greater, or possibly lower, than twenty:

It is also conceivable that there is more than one adaptor molecule for one amino acid, and the number 20 may be simply an accident (in any case we need a code for “end chain”, so perhaps 21 would be more reasonable). Alternatively the same adaptor molecule might fit on in more than one way (related, say, by a rotation of ${\mathrm{\Theta }}^{\circ }$). (1955, 9)

Countering potential criticism indicating absence of evidence for the existence of adaptors or their complexes with amino acids, Crick argued that such molecules were “in short supply” relative to a great excess of free amino acids and were thus hard to detect.

To which nucleic acid template did the adaptors bond? At first Crick seemed to think of DNA as the direct template for protein synthesis:

If we accept the idea that what matters in DNA are the hydrogen-bonding sites, it seems plausible to assume that each “site” will combine with one adaptor and one adaptor only. […] This requirement is not essential but it is likely if adjacent adaptors have to be combined with the DNA at the same time for polymerization to occur. (1955, 14)

Notably, already at that early time he had the insight to raise the possibility that adaptors hydrogen-bond to an RNA and not to a DNA template:

I have tacitly dealt with DNA throughout, but the arguments would carry over to some types of RNA structure. If it turns out that DNA, in the double-helix form, does not act directly as a template for protein synthesis, but that RNA does, many more families of codes are of course possible. (Crick 1955, 16)

Yet, lacking at the time knowledge of the structure and properties of RNA, Crick could not ascertain that it was the actual adaptor-binding template:

Incidentally the protein sequences we use to test out theories—insulin, for example—are probably RNA-made proteins. Perhaps a special class of DNA-made protein exists, almost in small quantities (and thus normally overlooked).⁹ Except perhaps where there are giant chromosomes. In particular base pairing may be absent in RNA or take a radically different form, and there may be more than one base to the asymmetric unit. Without a structure for RNA one can only guess. (Crick 1955, 16–17)

2.3 Known to Relatively Few, The Adaptor Hypothesis was Doubted or Dismissed

Because Crick’s paper was circulated only among the few members of the RNA Tie Club, it had minimal immediate impact. One unconvinced reader was Jim Watson, then at California Institute of Technology. In a February 10, 1955 letter to Crick¹⁰ he tersely remarked: “Your TIECLUB note arrived during visit.¹¹ Am not so pessimistic.¹² Dislike adapters.”¹³ Some of the leading students of the code and protein synthesis, including Gamow, Alexander Dounce, Paul Zamecnik, and Henry Borsook, met on April 4–6, 1955 at the Oak Ridge National Laboratories for a symposium on enzyme and protein structure. Discussing state of the art developments in protein biosynthesis, none of the speakers at the meeting mentioned Crick or his adaptor hypothesis (i.e., Zamecnik et al. 1956; Borsook 1956). Most pointedly, Dounce and Gamow, both members of the RNA Tie Club and likely readers of Crick’s paper, evaluated code word composition, overlap, and degeneracy without mentioning Crick’s hypothesis (Dounce et al. 1956).

Shortly after circulating his paper, Crick made two pubic presentations of his hypothesis. First, he spoke in February 18, 1956 in a London Biochemical Society symposium on the structure of nucleic acids and their role in protein synthesis.¹⁴ As Brenner later reminisced the audience contested the hypothesis¹⁵ (Brenner 2014; see also McElheny 2004, 89):

So he [Crick] gave the lecture and biochemists stood up in the audience and said this is completely ridiculous, because. […]¹⁶ What people don’t realise [sic] is that at the beginning, it was just a handful of people who saw the light, if there were twenty enzymes, we biochemists would have already discovered them. To them, the fact that they still hadn’t, went to show that this was nonsense I can put it that way. So it was like belonging to an evangelical sect, because there were so few of us, and all the others sort of thought that there was something wrong with us.

Crick also spoke about the adaptor hypothesis in a June 25–29, 1956 Gordon Conference on proteins and nucleic acids. Among those attending were key players in the areas of the code and protein synthesis: Gamow, Alex Rich, Dounce, and Zamecnik.¹⁷ As a rule, proceedings of Gordon conferences have no written records, and I could not find any account of Crick’s talk at the 1956 conference. Interestingly, however, this event illustrates the disconnection that existed between Crick’s theory and the contemporaneous experimental discovery of tRNA. The conference could have provided an opportunity for Crick and Zamecnik to compare notes on their respective hypothesis and its experimental corroboration. However, according to Hoagland, Zamecnik had missed Crick’s presentation:

In the summer of 1956 Francis gave an informal talk on ribosome structure and the adaptor idea at a Gordon conference in New Hampshire, not far from Boston, in the evening at the end of a five day meeting. Paul Zamecnik had attended the meeting but had left early and had not heard Francis’s paper. (1990, 96)

Yet, Crick was present at the conference when Zamecnik gave a talk on its third day titled “Studies on the Role of RNA and Nucleotides in Protein Synthesis.” By the time of that talk enzyme-catalyzed activation of amino acids has already been discovered (Hoagland et al. 1956) and five months later Hoagland and Zamecnik submitted for publication their first report of the discovery of tRNA (Hoagland et al. 1957). Considering this timeline and the fact that RNA was mentioned in the title of Zamecnik’s talk, it is likely that it included first hints of tRNA. Because the historical record shows that Crick learned about the discovery of tRNA and of amino acid charging enzymes only in January 1957 (section 4.2), it must be deduced that he either missed the talk or that its implications escaped him.

Crick also presented his adaptor hypothesis in a Society for Experimental Biology symposium on biological replication of macromolecules that took place in London in September 19, 1957. In this talk titled “On Protein Synthesis,” which was published one year later (Crick 1958), he addressed cardinal questions of the hour: the sequence hypothesis and the central dogma, protein synthesis, and the adaptor hypothesis.¹⁸ At the time of this presentation Crick was already aware of the discovery of amino acid activating and charging enzymes and of tRNA (see section 3). Yet, because the empirically estimated size of tRNA was larger than the size that he imagined for his adaptor, Crick did not equate tRNA with the adaptor. The 1957 talk included, therefore, repeated exposition of the adaptor hypothesis, a mention of the experimental detection of tRNA and amino acid charging enzymes and some speculations on the relationships of the adaptor to tRNA. Crick’s seminal talk is discussed in greater details in a following section (4.2.1).

3.1 Establishment of A Cell-Free Protein Synthesis System

3.2 Discoveries of Amino Acid Activation, Activating and Charging Enzymes, and tRNA

Using the cell-free protein synthesis system, Hoagland, Zamecnik, and associates initially discovered that prior to being incorporated into proteins, amino acids were activated by specialized activating enzymes. Subsequent experiments identified transfer RNA (tRNA)²¹ and showed that it was charged with cognate amino acids by the activating enzymes, currently named amino-acyl-tRNA synthetases. Based on these landmark discoveries Zamecnik and Hoagland proffered a tentative evidence-based skeleton scheme of protein synthesis (Hoagland et al. 1957).

3.3 Based on Their Experimental Results, Hoagland and Zamecnik Framed a Skeleton Model of Protein Biosynthesis

Based on their discoveries, Hoagland et al. offered in 1957 a tentative explanatory model of the passage from free to activated amino acids, their bonding to tRNA, and finally incorporation into proteins:

It is now further postulated that this initial activation of amino acids is followed by transfer of activated amino acids to S-RNA. This latter reaction is ribonuclease sensitive, while the former is not. GTP mediates the transfer of this activated amino acid to peptide linkage via the microsomes by a mechanism as yet unknown. (Hoagland et al. 1957, 216)

A year later they added more details to their model:

We have suggested elsewhere [Hoagland, Zamecnik, and Stephenson 1957] a hypothetical reaction sequence for protein synthesis which accounts for the findings presented in this paper. Its central idea is that pH 5 RNA molecules, [meaning tRNA] each charged with amino acids in characteristic sequence, polymerize in microsomes (in specific order determined by the complementary structure of microsomal RNA²³ to higher molecular weight units with resultant configurational changes which permit peptide condensation between contiguous amino acids. (Hoagland et al. 1958, 256)

The model was graphically presented in Zamecnik’s 1958–1959 Harvey Lecture (Zamecnik 1960; figure 3).

Figure 3:

Zamecnik’s evidence-based model of the path from free amino acids to their polymerization into protein. Original scheme in Zamecnik (1960) modified by Fry (2016, 346). Note that this model adhered to the time’s idea that ribosomal RNA, and not the yet undiscovered mRNA, was the template for proteins.

4.1 Hoagland was alerted to the Adaptor Hypothesis

Hoagland first heard about Crick’s adaptor hypothesis in late 1956. Decades later he described the circumstances of this revelation:

At the end of 1956 I had my first visit from a card-carrying molecular biologist. Jim Watson had just become a professor of biology at Harvard and was probing the structure of the ribosome. He had heard rumors of my discovery of transfer RNA, and I eagerly told him of our findings. He was restlessly attentive and when I had finished, told me that Francis Crick had forecast the existence of transfer RNA-like molecules a year or so earlier. He wondered whether I had heard of the adaptor hypothesis. I was astonished and a bit miffed that I had not! […] I was bowled over by the ingenuity and beauty of Francis’s idea and sensed that it had to be the explanation of our experimental findings.²⁴ An image arose in my mind: we biochemical explorers were hacking our way through a dense jungle to discover a beautiful long-lost temple, while Francis Crick flying gracefully overhead on gossamer wings of theory waited for us to see the goals he clearly was gazing upon. (Hoagland 1990, 93; also see Hoagland 1989, 2004, 2006)

In an undated conversation with Judson, Hoagland recalled his somewhat more aggrieved reaction to the news of the adaptor hypothesis:

In fact, I can remember vividly leaning over a centrifuge in the particular laboratory and talking with Jim about it, and his saying, ‘This is the interpretation of your results’, Hoagland said. ‘And, I can sense, palpably, my feeling of resentment, at the time, that Jim would be telling me how to interpret my result—but also the feeling that, God damn it, he is right. You know. It was just—it was just right.’ (Judson 1996, 324)

Despite their different tenor, both versions similarly show that Hoagland subscribed almost instantly to the idea that tRNA and Crick’s adaptor were synonymous. Importantly, however, Hoagland and Zamecnik did not need the adaptor hypothesis to explain their results since by that time they had already defined tRNA as “a link in the chain of intermediate reactions in protein synthesis” (Rheinberger 1997, 157 and sections 3.2.2 and 3.3). Yet, whereas Crick explicitly pointed to the crucial role of hydrogen bonding of his hypothetical adaptor and to the RNA template, Hoagland and Zamecnik more vaguely invoked complementarity between tRNA and an RNA template.

4.2 Learning of tRNA, Crick Posited That It Was Precursor to Smaller Adaptors

Crick first learned about tRNA and activating enzymes during his visit to the laboratory of Hoagland and Zamecnik in early 1957.²⁵ According to Hoagland (1990) he accepted at once that tRNA represented his putative adaptor:

In the spring of 1957, Francis Crick came to visit us. He was elated. The discovery of an RNA species that apparently performed the task of his postulated adapters prompted him to enter the experimental fray.²⁶ (97)

The historical record contradicts, however, this description of Crick’s prompt adoption of tRNA as the adaptor. In point of fact, Crick thought that tRNA molecules were precursors to much smaller adaptors. This idea grew from the contrast between Crick’s original notion of an adaptor of just three or very few nucleotides and an empirically measured, larger size of tRNA. In 1956, Crick, John Griffith, and Leslie Orgel circulated among members of the RNA Tie Club an informal communication on postulated properties of the genetic code.²⁷ This and a subsequent formal paper (Crick et al. 1957), contended that the code consisted of non-overlapping trinucleotide code words that were complemented by amino acid carrying adaptors of three or very few nucleotides. Zamecnik’s concomitant unpublished experimental measurements indicated, however, that tRNA was about sixty bases long (Rheinberger 1997, 159). Confronted with these results, Crick argued that tRNA molecules could not serve as adaptors because their large size would critically slow their diffusion and impede their entree into the microsomal site of protein synthesis (Olby 2009, 265). Writing to Hoagland in January 1957, Crick opined that tRNA was too large to go “intact into the microsomal particle.” Instead he conjectured that the activating and charging enzymes first acted as nucleases that cleaved tRNA precursors into trinucleotide adaptors that were then charged with the activated cognate amino acids²⁸:

How can we fit your results into this scheme? The idea is that your ‘activating enzymes’ cut up this new RNA into trinucleotides at the same time attaching an amino acid to each one.

The charged trinucleotide adaptors were next speculated to diffuse to the microsomes where they base-paired with an RNA template, liberated their carried amino acids which were then incorporated into growing polypeptide chains. Whereas Hoagland was open to the idea that tRNA molecules were precursors to smaller adaptors, Zamecnik was more cautious (Rheinberger 1997, 160). Thus, while Hoagland joined Crick in Cambridge to experimentally pursue this notion²⁴, Zamecnik published in 1958 and 1959 schemes of intermediate reactions in protein biosynthesis in which intact tRNA was depicted as the very carrier of amino acids with no mention of cleaved smaller adaptors (Zamecnik et al. 1958; Zamecnik 1960; fig. 3). Zamecnik also reasoned in his 1958–1959 Harvey Lecture that in order to be recognized by specific activating enzymes, adaptor molecules had to be bulkier than mere trinucleotides (Zamecnik 1960).

4.3 tRNA Was Finally Identified as the Functioning Adaptor

It is hard to pinpoint the exact time at which Crick accepted that intact tRNA was indeed his adaptor. Decades later he recalled his evolving perception of tRNA:

As every molecular biologist now knows, the job is done by a family of molecules now called transfer RNA. Ironically, I did not immediately recognize that these transfer RNA molecules were the predicted adaptor because they were bigger than I had expected, but I soon saw that there were no grounds for my objection. A little later Mahlon [Hoagland] came to Cambridge …. (Crick 1988, 96)

According to this rendering Crick abandoned the idea of tRNA as precursor to smaller adaptor molecules before the summer of 1957, the time of Hoagland’s arrival in England. This is unlikely, since in his September 19, 1957 talk Crick was still promoting the concept of tRNA as precursor (Crick 1958). Also, under the spell of Crick’s scientific and personal charisma, Zamecnik and Hoagland continued to pursue the idea of tRNA as precursor to shorter adaptors. This is indicated by what Zamecnik said in his 1959 Harvey Lecture (emphasis mine): “According to Crick’s hypothesis, and to our views, the soluble RNA molecule appears to be too long and complex to serve in its entirety as a suitable transfer agent” (Zamecnik 1960). Thus, from 1957 till 1959 the Boston laboratory conducted futile search for small adaptor products of tRNA cleavage (Rheinberger 1997, 192). It was only in late 1960 that Hoagland and Lucy Comly announced their failure to find evidence for breakup of tRNA into “smaller utilizable pieces” (Hoagland and Comly 1960). It thus appears that by the end of 1960 the weight of accumulated experimental data convinced Crick, Hoagland, Zamecnik, and the broader community of biochemists, that Crick’s adaptor and intact tRNA were synonymous.

4.4 Why Were Crick and Hoagland & Zamecnik Unaware of Each Other’s Hypothesis and Experimental Discoveries?

What could be the reasons for the disconnection between the hypothesis and the independent experimental discovery of the same entities? In his comprehensive analysis Rheinberger convincingly argued that a decisive factor was the cultural and epistemic divide that separated the nascent group of molecular biologists of the 1950s and the established larger community of biochemists. Whereas molecular biologists were interested in biological code and information, biochemists were engaged in isolation and characterization of new molecules and their interactions (Rheinberger 1997, 156–164). Another possible factor was the slow and limited scientific communication of the 1950s (no e-mails, video calls, or even inexpensive, transatlantic telephone calls). This likely minimized potential connection between Crick at Cambridge and the Harvard laboratory. Lastly, there was also the unfortunate, missed opportunity of interaction between Crick and Zamecnik when both attended the June 1956 Gordon conference.

5.1 Impacts of Crick’s Hypothesis and of the Experimental Discovery of tRNA: Historical Record

One way to assess the influences of Crick’s hypothesis and of the experimental discovery of tRNA is by looking at the reactions of contemporaneous scientists to these autonomous developments. As detailed in section 2.2.1, the original text of the adaptor hypothesis (Crick 1955) and its subsequent dissemination in scientific meetings (fig. 1) were met with indifference or distrust. This is also illustrated by the fact that Crick’s 1955 manuscript was cited for the first time only in 1960 (table 1) when the adaptor has already been recognized to be synonymous with tRNA (section 4.3). Most significantly, there is no detectable record of attempt to put the hypothesis to experimental test. Because Crick’s adaptor hypothesis did not motivate experimentalists to test its veracity, it remained an infertile theoretical model.

In contrast to the unresponsive reception of Crick’s hypothesis, the work of Hoagland and Zamecnik gained immediate and substantial recognition. Biochemists cited, replicated, and extended their findings on cell-free protein synthesis, enzyme-catalyzed amino acids activation and bonding to tRNA, and on discharge of the amino acids from tRNA and their incorporation into growing polypeptide chains. Promptly citing Hoagland and Zamecnik, numerous laboratories confirmed and expanded their findings (see for instance: Berg 1956; Davie et al. 1956; Demoss and Novelli 1955; Lipmann et al. 1959; Zachau et al. 1958).

5.2 Contrasting Impacts of Crick’s Hypothesis and of the Experimental Discovery of tRNA: Bibliometric Evidence

Bibliometric analysis provides another way to assess the relative impacts of Crick’s hypothesis and of the experimental discovery of tRNA. Low number of citations of Crick’s relevant key papers (tables 1 and 2) indicates that his hypothesis elicited weak reaction, both when it was still a stand-alone theory (1955–1957) and after the adaptor had already been equated with tRNA. By contrast, high numbers of citations of relevant key reports indicate that the discoveries of aminoacyl tRNA synthetases, tRNA and their modes of operation had considerable impact.

Table 1 shows how many times the term ‘Adaptor Hypothesis’ was mentioned between 1955 and early 2020 in articles in leading professional journals. Quite surprisingly, this phrase appeared only sparingly in the listed journals. Moreover, ‘Adaptor Hypothesis’ was mentioned for the first time (by Crick himself), only in 1958, nearly two years after the adaptor has already been associated with the empirically discovered tRNA. As to the sources of the term, 70% of the articles (24 out of 34), cited Crick’s 1957 lecture “On Protein Synthesis” (Crick 1958) which already spoke of the relationship between his adaptor and the concomitantly discovered tRNA (section 4.2.1). Only three articles cited Crick’s original 1955 RNA Tie Club communication (Crick 1955), two papers referred to his 1963 retrospective review in Progress in Nucleic Acid Research and Molecular Biology (Crick 1963), a single article cited his 1956 talk at the Biochemical Society Symposium (Crick 1957), and three papers mentioned the hypothesis without ascribed reference.

Table 2 shows results of scanning of the 1955 to 2020 body of the scientific literature for citations of relevant key publications by Crick and by Hoagland and Zamecnik. As seen, Crick’s papers on the adaptor hypothesis were poorly cited. His original RNA Tie Club manuscript (Crick 1955) was cited only 54 times in 65 years and most of the citations appeared in retrospective review articles rather than in reports of new research. Crick’s (1958) seminal lecture “On Protein Synthesis,” which already acknowledged the discoveries of tRNA and activating and charging enzymes, was cited more the 2600 times. However, only 24 of the citing articles mentioned the term ‘adaptor hypothesis’ (table 1), while the remaining > 2600 citing papers addressed other cardinal elements in the lecture such as the sequence hypothesis, the central dogma, etc. Importantly, the bibliometric data indicate that the impact of the adaptor hypothesis remained low even after it became well-known.

Whereas Crick’s hypothesis prompted belated and weak reaction, the discoveries of Hoagland and Zamecnik had substantial impact as attested by the hundreds of citations that each of their key papers earned (table 2). Of equal importance, the reaction to their discoveries was immediate. For instance, Hoagland’s report on amino acid activation (Hoagland 1955) was cited six times already in the year of its publication and 23 times in the subsequent year. Remarkably, most of the citing articles substantiated and expanded the findings of Hoagland and Zamecnik by reporting isolation of new activating enzymes and/or tRNA species.³⁴ Thus, whereas Crick’s hypothesis was never put to experimental test, the work of Hoagland and Zamecnik inspired robust empirical confirmation and expansion.

5.3 Why Did Crick’s Prescient Hypothesis Fail to Elicit Befitting Response?

It is rather surprising that despite its outstanding brilliance, Crick’s adaptor hypothesis had only minimal impact on practicing scientists. What could explain this incongruity?

Some may argue that the two years that elapsed between the introduction of Crick’s hypothesis in 1955 and its experimental corroboration in 1957–1958 were too short a time for the hypothesis to gain recognition. However, history shows that some hypotheses did gain due recognition, although they were confirmed soon after their introduction. Stephen Brush tallied time intervals that separated introduction of predictive theories in physics, chemistry, and biology from their subsequent corroborations. According to his account the natural selection theory of Mendelian inheritance has been accepted just three years after it was presented and the quantum mechanics theory was adopted one year after its introduction (Brush 2015, 488). Another potential explanation to the low effect of Crick’s hypothesis is its communication to only the small audience of the RNA Tie Club (Crick 1955). This explanation is countered, however, by Crick’s presentations of his hypothesis to hundreds of scientists in two meetings in 1956 (fig. 1). However, as witnessed by Brenner, audiences resisted or were indifferent to the hypothesis (Brenner 2014). Last, the possibility that the adaptor hypothesis was recognized only belatedly is countered by evidence that in the sixty-five years that elapsed since its introduction it had much smaller impact than the empirical work of Hoagland and Zamecnik (tables 1 and 2).

We can only speculate on deeper reasons for the surprisingly low relative impact of Crick’s adaptor hypothesis. One likely element is the resistance of experimental biologists to theorizing. Because complex bio-molecular systems were assembled under indeterminate historical circumstances of selection and evolution, abstract theories usually fail to effectively prognosticate their compositions and modes of operation. Thus, in contrast to its dominant role in physics, theory in the life sciences is only secondary to observation and experiment. Crick’s bent on theorizing in the tradition of physics made him, therefore, an odd man out in the overwhelmingly experimental community of biochemists. It should also be remembered that when he introduced his hypothesis in the mid-1950s, Crick had not reached yet his later prominence. His status at that time of an ex-physicist ‘outsider’ may also have contributed to the indifferent reaction of biochemists to his highly speculative abstract conjecture. By contrast, these scientists readily grasped and promptly followed the findings of Zamecnik and Hoagland which spoke the familiar language of experimental biochemistry. Another possible sociological/cultural reason for the dismissal of the adaptor hypothesis by traditional biochemists could be their alienation from the newly emerging group of molecular biologists who claimed to have a superior, emphatically theoretical, approach to fundamental problems in biology.

All told, evidence presented in this paper indicates that the adaptor hypothesis case did not conform with the predictivist claim of epistemic superiority of predictive theories over theories that accommodate known data. Although the adaptor hypothesis presciently predicted unknowns, its epistemic weight was lower than the Hoagland and Zamecnik’s evidence-based model of protein synthesis (fig. 3). The higher place of experiment over prediction in this particular case was well articulated by Hoagland himself. Decades after Crick introduced his hypothesis and after the discovery of tRNA, he appraised the contributions of theory versus experiment. His conclusion, which this paper espouses, was that despite the brilliance of Crick’s hypothesis, it was the more pedestrian experimental work that ultimately had the greater impact:

Crick’s adaptor idea, while one of the more brilliant insights in molecular biology, stands more as a monument to man’s imagination than as the springboard to a major discovery. In this case, we biochemists could assert with pride that by dint of smaller imaginative jumps, hard labor and considerable luck we had unearthed tRNA and appreciated its significance for determining order without the aid of the hypothesis—even if its promulgation did precede our discovery. In this instance, a grand theory neither substituted for nor guided the successful analytical dissection of the machinery of protein synthesis. (Hoagland 1990, 96; emphasis mine)