Immunology (1955-1975): The Natural Selection Theory, the Two Signal Hypothesis and Positive Repertoire Selection

Introduction

The clonal selection theory transformed immunology (Burnet 1959; Forsdyke 1995a).¹ Analysis of this paradigm shift, comparable to the emergence of Mendelism a half century earlier, is an on-going task (Soderqvist and Stillwell 1999). As with Mendelism (Barber 1961; Cock and Forsdyke 2008), the transition involved colorful figures and spirited defenses of old ways. Much of the transformation occurred in the decade that began with the natural selection theory of Jerne (1955). In the following decade clonal selection was consolidated and extended. The extensions included the two signal hypothesis and the affinity/avidity model for the shaping of lymphocyte repertoires.

     In a deeply perceptive biography, Soderqvist (2003) related the various roles of Niels Jerne in these 'decades of major and radical transition in the field' (Soderqvist and Stillwell 1999). As a laboratory scientist and theoretician, I was, in a small way, also a participant in the transition. Thus, I am able to contribute elements to the story that may be of historical interest. Apart from providing primary source material, the present paper draws attention to problems of the young theoretician, and the need to close the gap between first discoveries and discoveries later deemed to have been foundational in the sense that they have positively contributed to scientific advance.

Natural Antibody

Historians have suggested that the first two signal hypothesis was that which I presented in 1968 (Podolsky and Tauber 1997, p. 152; Doherty and Robertson 2004). There is a parallel between the genesis of this hypothesis and the earlier genesis of the natural selection theory of antibody formation (Jerne 1955); both were inspired by the idea of 'natural antibody.' There is also a link with later emerging ideas on the shaping of lymphocyte repertoires prior to an organism's encounter with foreign antigens (Jerne 1971).²

    The natural selection theory came to the 42 year old Niels Jerne while walking near the Knippel Bridge in Copenhagen in the summer of 1954 (Soderqvist 2003, p. 247). By this date he was an established figure in his field - a man of considerable savoir faire with friends who could smooth the path to publication, achieved in November 1955. On the other hand, the two signal hypothesis came one spring day in 1966 to a 27 year old, relatively naive, doctoral student in Cambridge, England. Jerne later gave an account of the genesis of his theory (Jerne 1966); but Soderqvist (1994), from a study of Jerne's archived papers, concluded that his autobiographical story 'is not satisfactory as a source document for the reconstruction of conceptual origins.' The same caveat may apply to the story the Cambridge student will here relate. And it should also be noted that others were thinking along two signal lines. For example, in the USA Herman Eisen (1966), referring to bivalent cell surface receptors as 'sentinel' antibodies, wrote: 'If 1 of the sentinel antibody's 2 sites is occupied by the antigenic determinant, the cell is stimulated - - . But if both of the sentinel's sites are occupied, the cell is not stimulated - i.e. it is suppressed or remains dormant.'³

    For both Jerne and myself, ideas were triggered by experimental observations relating to 'natural antibody' - antibody present in an organism prior to that organism's exposure to the specific antigen to which the antibody could bind. When studying the reactivity of viruses with blood serum in the spring of 1954, Jerne unexpectedly discovered an activity - dubbed 'P-star' - which he interpreted as due to natural antibody. This set off a train of thought leading to a theory of antibody formation that seemed to unify disparate observations better than some alternatives (Jerne 1955). He postulated a process by which an organism would, somehow, randomly generate a repertoire of antibodies covering a wide range of specificities. An incoming antigen (e.g. virus) would select the antibody that best complemented it, and would then conduct it to a cell where the antibody would, somehow, be replicated and secreted in large quantities (thus increasing protection against the virus). At the time of their first 'natural' generation, antibodies that reacted with 'self' antigens would have been absorbed by 'structures in the body of the animal itself' so that auto-immune reactivity would not develop.

    In the decade that followed there were many outstanding discoveries that reoriented immunology within a clonal selection framework. From a state of relative impoverishment, immunological knowledge suddenly expanded. The randomization process postulated by Jerne would first generate a repertoire, not of antibodies, but of lymphocytes bearing antibody-like receptors. Each lymphocyte, on being selected by specific antigen by virtue of its complementary receptor, would clonally expand and secrete large quantities of the corresponding antibody. It was found that antibodies had constant and variable regions (Porter 1958), which implied that constancy and variability could be aspects of the corresponding DNA sequences (Brenner and Milstein 1966). There were central lymphoid organs, such as the thymus, where immunological competence was conferred on naive stem cells (Miller 1961). The small lymphocytes emerging from such 'factories' were long-lived and, assisted by their smallness, were able to circulate around the body, like police ready to spring into action should an intruder appear (Gowans and Uhr 1966). When cultured in a synthetic medium containing blood serum, 'resting' small lymphocytes could be artificially activated by lectins (plant proteins that bound to cell surfaces). The lymphocytes then enlarged and proliferated, displaying features similar to those seen within peripheral lymphoid organs (e.g. lymph-nodes) in the course of immune responses (Nowell 1960). Thus, by the mid-1960s the framework within which immunological ideas could be set was better defined. The demands on those who presumed to offer unifying syntheses had become more restrictive. The loose hand-waving that might satisfy critics in the 1950s would less likely be tolerated.

    In the course of doctoral studies (1964-1967) in the Department of Biochemistry, Cambridge, I used lectins as 'antigen analogues' to stimulate cultured lymphocytes. Immune responses in vivo were known to be antigen dose-dependent, with inhibition of responses at high antigen doses (Mitchison 1964). Thus an antigen seemed capable of giving at least two, concentration-dependent, signals to an organism, one leading to an immune response, and one leading to response inhibition. Could this play out at the level of individual cells? Could inhibition by high antigen concentrations relate to the way an organism's own antigens ('self') prevented autoimmune reactivity? With these thoughts, I paid particular attention to lectin dose. If lectin bound to lymphocytes, then an increase in lymphocyte concentration should increase lectin requirement. But this was not found. The lymphocyte response was dependent on the ratio of the quantity of lectin to the quantity of serum present in the culture medium. This observation (in January 1966) was briefly reported (Forsdyke 1966a). A simple explanation, supported by subsequent studies, was that lectin-binding sites were present both in serum and on cells, but since serum sites were in great excess, they served to buffer cell-borne sites against reaction with lectin. Only by virtue of the cell-lectin interaction would peripheral lymphocytes be activated to enlarge and proliferate (signal one), or die (signal two; Milthorp and Forsdyke 1970; Forsdyke 1980). Thymus lymphocytes were not activatable (Forsdyke 1969a).

     It was a small step to go from this, to the idea that antibody-like receptors at the surface of a lymphocyte could be buffered against reaction with specific antigen by the corresponding free natural antibodies in the body fluids around it. The natural antibodies that had so excited Jerne would serve not only to bind an antigen, hence facilitating its disposal, but would also prevent its interaction with specific lymphocytes. Whether a lymphocyte was activated by specific antigen was determined by the ratio of the concentration of antigen to the concentration of the corresponding natural antibody. Thus, the signal given to a lymphocyte could be modified either by variation in antigen concentration (numerator), or by variation in antibody concentration (denominator). Within this framework, there emerged explanations for various immunological phenomena in terms of the delivery of one of two signals to a lymphocyte - one stimulatory and one inhibitory. This dispensed with the postulate that whether antigen stimulated or inhibited lymphocytes depended on their stage of development (Burnet 1959), or that stimulation required a cellular cofactor (Medawar 1963).

A Few New Things

First mention of the new hypothesis is found in my diary entry for 23rd April 1966.⁴ Following an examination of the relevant literature, a later entry (19th June 1966) declared that 'many main features of my ideas have already been presented by Jerne 1955, and Lederberg 1959.' However, I added that there were 'a few new things to say.' A paper setting out the 'few new things' was kindly read by two Cambridge biochemists well versed in immunology (Cesar Milstein and Alan Munro). They advised that theories were for old guys in their dotage and that at my stage of career it would be better to do experiments. Apart from this, while displaying no enthusiasm, they had no major criticisms. I formally submitted the paper from the Department of Biochemistry to Nature on the 16th of September, and the hypothesis was outlined at a scientific meeting at Oswestry a week later (Forsdyke 1966b).

     On my return from Oswestry, I learned of the policy that papers submitted from the Biochemistry Department had to be approved, prior to submission, by the Head, Frank G. Young. As a matter of courtesy, I had dropped off a copy of the paper to his office, together with a paper on the lectin-serum interaction that I had submitted to the Biochemical Journal. Young allowed my submission to the Biochemical Journal to stand (Forsdyke 1967). However, after consulting an immunologist - Robin Coombs of 'Coombs test' fame - Young insisted (28th September) that I withdraw the Nature paper. At my request the Editor returned it.⁵

    At a subsequent meeting Coombs told me that the paper was not clear and advised that I leave it a few months and then reconsider. My diary records (29th October 1966) that I sent copies of the paper to various immunologists - Frank McFarlane Burnet, Peter B. Gell, James Gowans, John Humphrey, Niels Jerne and Joshua Lederberg. Replies were received from all except the last two. Burnet replied promptly with encouragement. Humphrey invited me to Mill Hill for a discussion (17th November).   

    The theory formed a major part of my Ph.D. thesis, submitted in November 1966 and examined (5th May 1967) by Humphrey (external examiner) and Kenneth McQuillen (internal examiner). They required extensive revisions and less emphasis on theoretical aspects. The revised thesis was submitted on 2nd September, by which time I had relocated to the Institute of Animal Physiology, Babraham, Cambridge.

A Short Version

Meanwhile, prompted by Coomb's advice, I had written a short version of the paper citing my doctoral thesis where 'a more extensive hypothesis' would be found.  My diary notes (24th September 1967): 'I have written - - a very simple, and rather popular theory paper - taking the key point of my first paper and adapting it to the analogy of the distinction between "self" and "not-self" in a liquid scintillation counter.'⁶ On learning (2nd December 1967) that my Ph.D. thesis had been accepted, I submitted the paper to the Lancet, giving my home address, not an institutional address. The paper was accepted without demur (13th January 1968). I received the page proofs shortly thereafter (23rd January), and the paper was published on the 10th February (Forsdyke 1968).

    Later in the year, after relocation to Canada, my diary records (24th November 1968) that, prompted by the appearance of a different two signal hypothesis in Nature (Bretcher and Cohn 1968), I decided to publishing the 'more extensive hypothesis.' The appendix of my thesis, with minor modifications, was duly submitted to the Journal of Theoretical Biology (13th December 1968). Six months later (8th June), I was informed that, while acceptable to one reviewer, the other had criticisms. I responded to the criticisms and the paper was accepted (Forsdyke 1969b). The response was somewhat underwhelming. However, Podolsky and Tauber (1997, p. 400) later noted that one of the authors of the Nature two signal paper had retrospectively credited my Lancet paper as having 'put us on the right track,' but had added that 'it provided us with only a mere glimmer of the principle' (Cohn 1989; 1994).⁷ To date (2010) the 1968 paper has received only six independent citations, and the 1969 'more extensive hypothesis' only two. Nevertheless, as will be set out here, it was my two signal hypothesis that launched the affinity/avidity model with the now widely accepted view that the repertoire of lymphocyte receptor specificities is shaped by processes of both negative and positive selection.

Affinity/Avidity Model

While not documented, I recollect that it was during the correction of the page proofs of the Journal of Theoretical Biology paper (perhaps in October 1969) that I began to develop what is now known as the affinity/avidity model⁸ for the selection of lymphocyte repertoires. The basic idea was that the antibody-like receptors borne by lymphocytes could vary in their affinity for antigen. A low antigen concentration would mainly stimulate lymphocytes with high affinity receptors. But a high antigen concentration could stimulate lymphocytes with low affinity receptors (first signal), while giving a second signal (inhibition) to lymphocytes with high affinity receptors. My two signal hypothesis had suggested a way (by removing the natural antibody buffer) that those with high affinity for self antigens could just receive the second signal (negative selection), even at low antigen concentrations, while those with low affinity for self antigens could still receive the first signal (positive selection).

    Some notes dated 11^th July 1971 give the key elements of the new model.⁹ But, having left Cambridge with the admonition ringing in my ears - stop thinking and do experiments, and being required for the first time to coax funds from granting agencies with plausible stories that did not stretch the imagination, it made sense to follow the Cambridge advice. In that period my publications became solidly grounded on the experiments in my laboratory.

    A paper reporting antigen dose-response relationships in lymphocyte cultures was submitted to the Journal of Experimental Medicine in April 1971. It was returned without having been sent for review. A succession of rejections followed (Journal of Immunology, Cellular Immunology, Biochemical Journal). Experiments continued and the data soon outgrew the original paper. In 1973 three interrelated papers were submitted to Immunology (3rd February) and were collectively accepted (28th March; Forsdyke 1973a, b, c). In these papers positive selection of lymphocyte repertoires was briefly referred to, but it was becoming apparent that a regression to bold public theorizing was in order.

    Just as factual restrictions on the scope of theorizing had increased dramatically in the decade between Jerne's natural selection theory and my two signal hypothesis, so between the mid 60s and the mid 70s there were even greater advances. These included, the sequences and structures of antibodies (Hilschmann and Craig 1965), the sub-division of small lymphocytes into cooperating populations (dubbed 'T' to indicate an origin in the thymus and 'B' to indicate origin in the bone marrow; Claman et al. 1966), and the finding that immunological responsiveness was influenced by cell surface components that had a major influence on whether grafts would be accepted between members of the same species (major histocompatibility complex antigens, MHC antigens; Benecerraf and McDevitt 1972; Bodmer 1972).¹⁰

    Furthermore, studies of the reactivity ('alloreactivity') of lymphocytes from one organism, that had been injected into another member of the same species ('graft-versus-host reaction;' Cock and Simonsen 1957),¹¹ had led Morten Simonsen to challenge the clonal selection theory. The frequency of alloreactive cells was 'too high to be compatible with the orthodox version of clonal selection' (Nisbet et al. 1969). In other words, organisms seemed to have put all their eggs into the alloreactivity basket. Their repertoires were so misshapen that insufficient reactivity was left for the universe of other antigens that might be encountered. Thus, as later expressed, 'the thymic education of T cells provides them with blinkers' (Simonsen 1990).

Jerne's Defense of Clonal Selection

Now approaching his sixtieth birthday and director of the newly founded Basel Institute for Immunology, Jerne met his fellow Dane's challenge with a new theory: 'One of my basic intentions in formulating the present theory was to save the clonal selection theory from the consequences of Simonsen's argument' (Jerne 1970, p. 359). Since skin grafts between parents are always rejected, it seemed likely that their child, containing MHC antigens from both parents, would need to define 'self' afresh. Accordingly, the orthodox version of the clonal selection theory required diverse lymphocytes, each with a distinct antigen specificity, to be randomly generated during the life of an individual organism. At the outset, those reactive with an organism's own antigens would be eliminated. This orthodox version placed no onus on the variable genes themselves, save that their sequences and structures should not have mutated over evolutionary time in a way that would impede subsequent diversification, on a cell by cell basis, over somatic time. This would then permit the generation of a spectrum of specificities, one for each lymphocyte. Lymphocytes that emerged from this self/not-self discrimination filter would suffice to confront the universe of foreign antigens.

    Jerne begged to differ. His speculations were first released as a privately circulated 'samizdat' document,¹² then as a World Health Organization (WHO) document (Jerne, 1969), and then as part of the proceedings of a conference held at Brooke Lodge (May 1970), where there was extensive discussion by Simonsen and others (Jerne, 1970). A submission to the Proceedings of the National Academy of Sciences was declined because it was too long (Soderqvist 2003, p. 265). The final version came to immediate prominence in 1971 as the first paper in the first issue of the European Journal of Immunology (Jerne 1971). This was the year of the First International Congress of Immunology, which can be seen as culminating Jerne's efforts, through WHO, to further international cooperation in immunology (Soderqvist, 2003, p. 229).

     Jerne proposed that the generation of diversity in a 'mutant breeding organ' (thymus or bone marrow) would have been driven by self antigens - in particular, by self MHC antigens: 'I don't really want to define too precisely the set of cell-surface antigens against which antibodies are determined by the v-genes in my theory. But let us try the assumption that the surface antigens involved are a certain set of major histocompatibility antigens' (Jerne 1970, p. 349). Thus, the DNA encoding antibody variable regions (v-genes) would have been selected over evolutionary time to determine reactivity with MHC antigens. Organisms that did not display sufficient reactivity would have been eliminated by natural selection. Germ line DNA was not a blank slate upon which somatic mutational processes would subsequently act. DNA already encoded MHC reactivity. The repertoire was biased towards reactivity with MHC antigens.

    Jerne acknowledged: 'It is easiest to understand the presence of a set of germ-line v-genes if its maintenance were subject to a strong evolutionary selection pressure' (Jerne 1971). Yet, what could that selection pressure be? 'I assume that antibodies directed against self components on cell surfaces have some important function in ontogeny, namely in the cell to cell recognition that is needed for the formation of specialized tissues and for morphogenesis.' This was more simply stated in an earlier version (Jerne 1970, p. 349): 'I assume than an embryo cannot even develop, let alone an individual survive, unless its cells can make antibodies that fit to the surface antigens of the individual.' That Jerne was somewhat out of his depth here, is indicated by an allusion in the final paper (Jerne 1971) to the offspring of mules (which, being sterile, do not have offspring). If we forgive his sweeping assumptions (and many were so inclined),¹³ it seemed that, in one fell swoop, Jerne had explained the 'Simonsen phenomenon.' Since the distinction between B cells and T cells had only recently emerged, it was easy to suppose that the theory applied 'equally to cells that make antibodies for secretion [B cell lineage], and for those cells that incorporate the antibodies only on their surfaces as receptors [T cell lineage]'¹⁴ (Jerne 1970, p. 345).  

    There were added subtleties. For any individual, two lymphocyte subsets were postulated. The Simonsen phenomenon (alloreactivity) was due to a subset with receptors directed against the MHC antigens of the species that the individual did not possess (allogeneic histocompatibility antigens; subset A).¹⁵ These receptors were directed against all the MHC 'cards' that nature had available to deal out, less those that had been dealt out to that particular individual. Thus, the A subset genes, by definition, did not encode reactivity with 'self' MHC antigens. But the individual, while not reacting against its own antigens, had to develop reactivity against the universe of other possible antigens (not necessarily MHC antigens) that might be encountered. This was where the second subset came in. Lymphocytes of the second subset started out with receptors directed against the MHC antigens that the individual happened to have inherited from its parents (subset S). Jerne postulated that the corresponding v-genes would be subject to mutation in the mutant breeding organ, so that the affinity of lymphocyte receptors for self MHC would erode and new affinities for foreign antigens that might later confront the organism, would randomly arise. If a cell failed to mutate sufficiently (i.e. it retained reactivity with self-MHC) it would be destroyed (negative selection).

     It should be noted that there was no suggestion of selection for self reactivity (positive selection; Forsdyke 1995b). For Jerne, a cell of subset S: 'proliferates, perhaps because a hormone stimulates the proliferation of all stem cells entering the thymus. Possibly, the histocompatibility antigens in the thymus which fit to cell receptors provide additional stimulation. I assume, however, that for this same reason none of the cells of this "forbidden" clone will be permitted to leave the thymus as antigen-sensitive cells, but that they will eventually all die out' (Jerne 1971). Only cells of subset S that could mutate would survive, so that: 'finally, cells will arise that have entirely lost their fit to the histocompatibility antigens of the individual itself.' Thus, there was ongoing proliferation in the thymus and it was the absence of negative selection that, in Jerne's view, allowed cells with v-gene mutations to survive. The role of intrathymic self-histocompatibility antigens in driving the proliferation of the S subset was an afterthought - a way of getting the mutations that would not be seen in subset A, even though these cells would also be proliferating. There was no implication of the S subset being positively selected somatically for the ability to react with not-self histocompatibility antigens (the role of the A subset as determined in the germ-line).

     Jerne (1971) concluded: 'The restriction of ontogenic selection to random mutants of cells expressing v-genes of subset S thus determines both the responsiveness to certain types of antigen and the range of antibody specificities that an individual animal can produce, so that, indirectly, these properties are under dominant control of histocompatibility genes.' The possibility that development from an initially biased state (anti-self MHC) might so restrict the range of final specificities that responses to some pathogens might be compromised, was not entertained. Nor was there further clarification of 'indirectly.' I met Jerne briefly at a Woods Hole conference in September 1973, but I cannot recall that we discussed science.

Positive Repertoire Selection

There is no record as to when I began writing a new theoretical paper. Not the least of many distractions was Watergate, famous for: 'What did he know, and when did he know it?' The Journal of Theoretical Biology records that my paper was received 28th May 1974. My diary records (4th January 1975) that the paper was accepted 'after 5 months delay without explanation.' It was published in July (Forsdyke 1975).

     Like Jerne's, my paper treated lymphoid cells generically without a distinction between T and B lymphocytes, but there was no division into subsets. My model was derived from the first principles set out in my two signal papers, and from consideration of the probability that a pathogenic microbe (e.g. virus) would usually adapt by mutation quicker than its host. A microbe that could, in one step, mutate one of its surface antigens from a form that was not-self with respect to its host, to a form that was self with respect to its host, would have largely overcome the host's immune defences with respect to that antigen. It could then exploit the 'holes' in the repertoire that had been created by the elimination of self-reacting lymphocytes. However, mutation is generally a stepwise process. If a microbe (not-self), by mutating a step towards self along the path from not-self to self, could secure a selective advantage, then the mutant form would come to dominate the microbe population. If a microbe from this mutant population, by mutating a further step along the path, secured a further advantage, then this new mutant form would, in turn, come to dominate the population. Thus, an average member of the microbe population would progressively become better adapted, to the detriment of the host.

    This supposes that progressive mutation along the not-self-to-self path would be increasingly advantageous to the microbe. However, the advantage would be lost if, as it mutated closer to host-self, the microbe encountered progressively stiffer host defences. Thus, positive selection of lymphocytes for specificities that were very close to, but not quite, anti-self, could be an important host adaptation providing 'a barrier opposing the progressive evolution of the surface determinants of a pathogen into forms identical with the surface determinants of its host' (Forsdyke 1975). To emphasize this proximity to self, positive selection was described as the somatic generation of repertoires that would have been preselected to respond against 'near-self' antigenic determinants. Thus, MHC involvement would have been an automatic consequence of the positive selection of 'anti-near-self' cells during their initial maturation. The ideas of positive selection and of its mechanism came as one conceptual package. By virtue of their priming by self, some of the anti-near-self cells would cross-react with foreign MHC antigens, so would correspond to Jerne's A subset. The need to take into account total population-size homeostasis was also emphasized, so that an increase in a population of one affinity would require proportionate decreases in other populations.

'Near Self' or 'Altered Self'?

Shortly before the submission of my paper, there appeared in Nature a paper by Rolf Zinkernagel and Peter Doherty (1974). Extending earlier work on MHC-restriction of B and T cell cooperation, they had shown that T cell-mediated cytotoxicity was also MHC-restricted. Their paper, which to date (2010) has received 1515 citations, was the basis of their Nobel prize (Zinkernagel and Doherty 1997). Instead of my term 'near self,' which emphasized how close an antigen might be to self, they had used the term 'altered self,' which emphasized the fact of a difference from self. The relative merits of the two terms are discussed elsewhere (Forsdyke 2005).

     The paper of Zinkernagel and Doherty (1974) does not appear among my cited references (Forsdyke 1975). While not documented, I recollect that I read it some time after its publication date (19^th April), and prior to submission of my paper (about 20^th May). Since it supported my viewpoint, I considered adding it to my cited references, but did not want to imply that my paper was in any way derivative of theirs. While treating host cell surface antigens generically, I had already indicated the possibility of MHC involvement both by citing two MHC papers (Benecerraf and McDevitt 1972; Bodmer 1972), and by mentioning, not in a way that can be deemed prescient, a possible association of pathogen surface antigenic determinants with host cell surface components - a theme later developed more extensively (Forsdyke 1991). The term 'near-self' is in my notes dated July 1^st 1971, and the concept was touched upon in 1973: when discussing '"self" antigens capable of stimulating low specificity anti-self cells,' I remarked that: 'A high concentration of such low specificity cells would be predicted from theoretical considerations previously advanced'(Forsdyke 1973a).

      A decade later the differential affinity/avidity model was independently introduced for T lymphocytes by Jonathan Sprent (Lo and Sprent 1986; Sprent and Webb 1987). Although for some time opposed by the 'peptide' model (Claverie and Kourilsky 1986; Marrack and Kappler 1987), it eventually became widely accepted as applying generically to both T and B lymphocytes (Gaudin et al. 2004; Wen et al. 2005). Thus, Janeway (2001) concluded: 'Both the mature, naive T cell repertoire and the mature, naive B cell repertoire are generated by interaction with self-ligands rather than non-self ligands. These self ligands can signal B and T lymphocytes to mature and to survive.' Cancro and Kierney (2004) pondered 'the underlying biological rationale' of the positive selection of B and T lymphocytes, and the role of 'unaccounted structural space' (i.e. holes in lymphocyte repertoires). They envisaged the development of 'subthreshold self reactivity" (i.e. near-self), which would minimize 'potential evasion by rapid microbial evolution.' This came close to the case made in my 1975 paper that, to date (2010), has received only 4 independent citations.  

Discussion

A published theory may not necessarily be first, but it can be foundational (in a constructive sense). Sometimes it can be both - the classical example being Gregor Mendel's discovery of what we now call genes. Mendel's paper (1865) later became foundational even though it only began to influence others in 1900. In the interim no alternative foundational challenger had emerged (Cock and Forsdyke 2008). On the other hand, Paul Ehrlich (1900), although first in many respects, seems not to have been foundational in terms of influencing those who opened up immunology in the 1950s (Forsdyke 1995a). Yet surely science would better advance if all firsts were foundational? Rather than dismissing non-foundational firsts as premature (Barber 1961; Zuckerman and Lederberg 1986), an important task for historians of science would seem to be (i) collecting cases of temporal gaps between first and foundational, (ii) for each case determining the reasons for the gap, (iii) comparing cases in search of common elements and, if such are found, (iv) making conclusions available to science policy advisors (Forsdyke, 2000). In this way, taking immunology as an example, we might hope to accelerate progress towards effective treatments of autoimmune diseases (e.g. rheumatoid arthritis), and infections (e.g. AIDS).

     From a 'presentist' perspective (Harrison 1987), Jerne's first theoretical immunology paper (the natural selection theory; Jerne, 1955) later appeared primitive relative to the sophisticated cellular hypothesis advanced by Ehrlich a half century earlier (Forsdyke 1995a). Indeed, so rapid were developments in molecular biology that, by the time of publication Jerne's proposed mechanisms were already obsolete (Cohn 1994). Protein directly begetting protein (more antibody) was inconsistent with much of genetics and molecular biology. If a cell could somehow 'read' a protein and turn that information into nucleic acid, Jerne's idea might have worked, because nucleic acid can beget nucleic acid, and that nucleic acid can then beget more of the protein. But no mechanism for directly 'reading' a protein into nucleic acid was known then, or has since emerged.

     Yet Jerne's natural selection theory has proved both first and foundational in two important respects - he overcame 'Ehrlich's dilemma,' and he pointed out the need for a randomization process to generate potential reactivity to a wide range of antigens. Influential contemporaries (e.g. Linus Pauling) were then wedded to the idea that a foreign antigen 'instructs' a cell to make complementary antibody. To flatten any opposition, they needed only to point to Ehrlich's dilemma - it was inconceivable that an organism could be prepared in advance for the infinite range of antigens that might possibly present themselves (Ehrlich 1900). Indeed, it was not until Burnet brought Jerne's paper to his attention, that Lederberg saw a way round this problem (Lederberg 1994).

    Jerne's studies of the avidity of serum antibody activity after immunization had revealed to him that 'among the population of circulating globulin molecules' there could be some 'better-fitting' an antigen than others. There could be a '"low grade" antibody of low combining capacity' and a '"more avid" antibody of high combining power.' So the affinity of an antibody for an antigen was a matter of degree. 'The number of specific configurations which a globulin molecule can exhibit is large' and 'since normal mammalian serum contains more than 10¹⁷ globulin molecules per milliliter, these may include a million 10¹¹ fractions of different specificity. This would seem an amply sufficient number.' So 'there will - - be fractions possessing affinity toward any antigen to which the animal can respond.' Ehrlich's dilemma was false (Jerne, 1955). And how might an organism arrive at an 'amply sufficient number'? Jerne further proposed that: 'Somewhere in the beginning - - we have to postulate a spontaneous production of globulin molecules of a great variety of random specificities in order to start the process.' This might occur in 'a specialized lymphoid tissue, such as the thymus.'

    What remained puzzling is why Jerne did not go further. Jerne had invoked 'multiplication of the cells' that made antibody, and in 1995 I assumed  that, had he known of Ehrlich's side chain theory, he would have quickly realized that his proposed randomization process could have occurred at the cellular level (Forsdyke 1995a). Others were already toying with the idea. It was in the air. Indeed, in his biography Soderqvist (2003, p. 221) related that a few months after publication of the natural selection theory a young colleague, Jorgen Sparck, had suggested that cells were the selection unit, but Jerne 'was evasive and did not accept it.'

     It is now apparent (Soderqvist 2003, p. 179) that, in an early draft of the 1955 paper, Jerne had indeed discussed Ehrlich's theory, but had omitted this from the final version. Jerne had conceded to Soderqvist that he might have been unconsciously influenced by his prior reading of Ehrlich's papers but, as far as he was aware, the key idea came out of the blue. Soderqvist speculated that Jerne 'wanted to be unique, that he opted for originality rather than displaying his connectedness with tradition.'

    The fact that Jerne did know of Ehrlich's work is surprising since it seems such a short step to go from the selection of antibodies to the selection of cells. Jerne did not take that step, perhaps because he suspected that his solution of Ehrlich's dilemma would seem less cogent if the diversifying unit were cellular. Within an organism there are far fewer lymphocytes than globulin molecules. If each cell corresponded to one distinctive antibody specificity (one cell one antibody) then, although the antibody combining region might adopt a large 'number of specific configurations,' this might not be sufficient for the organism to confront the universe of potential antigens? There would be more holes in the repertoire that a pathogen could exploit, and perhaps Jerne was not prepared to consider this possibility.

    Jerne's second major theoretical immunology paper was his defense of clonal selection (Jerne 1971). It was a first in suggesting that germ-line variable region genes had been preselected over evolutionary time to encode reactivity with the MHC antigens of the species, but it was not foundational in two respects: his explanation of why there should have been this preselection was vague, and later investigations of antibody receptor structures gave only a hint of preadaptation for MHC reactivity (Huseby et al. 2004; Scott-Browne et al. 2009; Roomp and Domingues 2011). Thus, germ-line v-genes, at least in the antigen-combining regions, are essentially blank slates. It is mainly after somatic variation and selection for MHC-reactivity, that such reactivity becomes evident. Jerne's assertion that 'germ-line v-genes code very precisely for antibodies that fit to a certain set of histocompatibility antigens, i. e. that these postulated genes determine antibodies that fit to the surface antigens of cells of the same animal species,' was incorrect. However, in the years that followed, his 1971 paper received many citations. It might be thought that it was foundational because it had focused attention on the Simonsen phenomenon and MHC antigens, but these were already subjects of high immunological interest.

    In contrast, my paper presenting the affinity/avidity model (Forsdyke 1975) has, to date (2010), received only 4 independent citations. Since it cited Jerne's 1971 paper, a naive future historian might regard his paper as foundational in this respect. However, as set out here, my model was independently developed from the premises set out in my earlier two signal papers, which might accordingly be regarded as foundational. Yet, although it would now appear that my affinity/avidity paper (Forsdyke 1975) was a first, it was not foundational in that others (i.e. Sprent and his coauthors) developed similar models while being unaware of mine. To date (2010) the paper of Lo and Sprent (1986) has received 305 citations, Sprent and Webb (1987) 210 citations, and a later review (Sprent et al. 1988) 408 citations. Podolsky and Tauber (1997, pp. 331-332) imply that positive selection became apparent in 1986.

    Jerne's third major theoretical immunology paper (Jerne 1974) presented his 'network' theory (Soderqvist 2003, pp. 273-277). The details of this and of its apparent downfall need not concern us here, save for a recent historical comment (Tauber 2010): 'This theory enjoyed a great vogue from its presentation in 1974 through the 1980s - - . That major textbooks completely omitted any mention of Jerne's network by the late 1990s reflects how other theoretical concerns vying for dominance signify a key conceptual struggle within the discipline that dates to its origins.' Thus, like his 1971 paper, it now seems to have been first, incorrect, and non-foundational.

Concluding Remarks 

References

Barber, B. 1961. Resistance by scientists to scientific discovery. Science 134: 596-602.

Benecerraf, B. and McDevitt, H. 0. 1972. Histocompatibility-linked immune response genes. Science 175: 273-279.

Bodmer, W. F. 1972. Evolutionary significance of the HLA system. Nature 237: 139-145.

Brenner, S. and Milstein, C. 1966. Origin of antibody variation. Nature 211: 242-243.

Brent, L. 1997. A History of Transplantation Immunology. San Diego: Academic Press.

Bretscher, P. 2004. Living with the ups and downs of the two signal model. Immunology and Cell Biology 82: 141-148.

Bretscher, P. and Cohn, M. 1968. Minimal model for the mechanism of antibody induction and paralysis by antigen. Nature 220: 444-448.

Burnet, F. M. 1959. The Clonal Selection Theory of Acquired Immunity. Cambridge: Cambridge University Press, pp. 64-66.

Cancro, M. P. and Kearney, J. F. 2004. B cell positive selection: road map to the primary repertoire? Journal of Immunology 173: 15-19.

Claman, H. N., Chaperon, E. A. and Triplett, R. F. 1966. Thymus-marrow cell combinations: synergism in antibody production. Proceedings of the Society for Experimental Biology and Medicine 122: 1167-1171.

Claverie, J. M. and Kourilsky, P. 1986. The peptidic self model: a reassessment of the role of the major histocompatibility complex molecules in the restriction of the T cell response. Annales de l'Institute Pasteur-Immunologie 137D, 425-442.

Cock, A. G. and Forsdyke, D. R. 2008. Treasure Your Exceptions. The Science and Life of William Bateson. Cambridge: Cambridge University Press.

Cock, A. G. and Simonsen, M. 1958. Immunological attack on newborn chickens by injected adult cells. Immunology 1, 103-110.

Cohn, M. 1989. History of associative recognition theory. In: R. E. Langman (au.) The Immune System: Evolutionary Principles Guide Our Understanding of this Complex Biological Defense System. San Diego, Academic Press, pp. xvii-xxvi.

Cohn, M. 1994. The wisdom of hindsight. Annual Reviews of Immunology 12: 1-62.

Doherty, M. and Robinson, M. J. 2004. Some early trends in immunology. Trends in Immunology 25: 623-631.

Ehrlich, P. 1900. On immunity with special reference to cell life. Proceedings of the Royal Society (London) 66: 424-448.

Eisen, H. N. 1966. Learning and memory in the immune response. Cancer Research 26: 2005-2011.

Forsdyke, D. R. 1966a. Response to phytohaemagglutinin. Lancet 287, 713.

Forsdyke, D. R. 1966b. Nucleic acid changes during PHA-induced lymphocyte transformation in vitro and the possible relationship of this phenomenon to immune processes in vivo. In: M. W. Elves (ed.) The biological effects of phytohaemagglutinin. Oswestry: The R. Jones and A. Hunt Orthopaedic Hospital, pp. ll5-l23.

Forsdyke, D. R. 1967. Quantitative nucleic acid changes during PHA-induced lymphocyte transformation in vitro: dependence of the response on the PHA/serum ratio. Biochemical Journal l05: 679-684.

Forsdyke, D. R. 1968. The liquid scintillation counter as an analogy for the distinction between 'self' and 'not-self' in immunological systems. Lancet 291, 28l-283.

Forsdyke, D. R. 1969a. Impaired activation of thymus lymphocytes by PHA. Journal of Immunology l03: 8l8-823.

Forsdyke, D. R. 1969b. A theory of immunity. Journal of Theoretical Biology 25: l73-l85.

Forsdyke, D. R. 1973a. Serum factors affecting the incorporation of [³H]thymidine by lymphocytes stimulated by antigen:  I. Serum concentration. Immunology 25: 583-596.

Forsdyke, D. R. 1973b. Serum factors affecting the incorporation of [³H]thymidine by lymphocytes stimulated by antigen:  II.  Evidence for a role of complement from studies with heated serum. Immunology 25: 597-6l2.

Forsdyke, D. R. 1973c. Serum factors affecting the incorporation of [³H]thymidine by lymphocytes stimulated by antigen:  III. Evidence for a role of complement from studies with specific complement inhibitors. Immunology 25: 6l3-6l9.

Forsdyke, D. R. 1975. Further implications of a theory of immunity. Journal of Theoretical Biology 52: l87-l98.

Forsdyke, D. R. 1980. Lectin pulses as determinants of lymphocyte activation and inactivation during the first six hours of culture: sequential action of concanavalin-A and complement cause cell lysis. Canadian Journal of Biochemistry 58: l387-l396.

Forsdyke, D. R. 1991. Early evolution of MHC polymorphism. Journal of Theoretical Biology 150: 451-456.

Forsdyke, D. R. 1995a. The origins of the clonal selection theory of immunity as a case study for evaluation in science. Federation of American Societies for Experimental Biology Journal 9: 164-166.

Forsdyke, D. R. 1995b. Jerne and positive selection. Immunology Today 16: 105.

Forsdyke, D. R. 2000. Tomorrow's Cures Today? How to Reform the Health Research System. Amsterdam: Harwood Academic.

Forsdyke, D. R. 2005. 'Altered-self' or 'near-self' in the positive selection of lymphocyte repertoires? Immunology Letters 100: 103-106.

Gaudin, E., Hao, Y., Rosado, M. M., Chaby, R., Girard, R. and Freitas, A. A. 2004. Positive selection of B cells expressing low densities of self-reactive BCRs. Journal of Experimental Medicine 199: 843-853.

Gowans, J. L. and Uhr, J. W. 1966. The carriage of immunological memory by small lymphocytes in the rat. Journal of Experimental Medicine 124: 1017-1030.

Harrison, E. (1987) Whigs, prigs and historians of science. Nature 329, 213-214.

Hilschmann, N. and Craig, L. C. 1965. Amino acid sequence studies with Bence-Jones proteins. Proceedings of the National Academy of Sciences USA 53: 1403-1409.

Huseby, E., Kappler, J. and Marrack, P. 2004. TCR-MHC/peptide interactions: kissing cousins or a shotgun wedding? European Journal of Immunology 34: 1243-1250.

Janeway, C. A. 2001. How the immune system works to protect the host from infection: a personal view. Proceedings of the National Academy of Sciences USA 98: 7461-7468.

Jerne, N. K. 1955. The natural selection theory of antibody formation. Proceedings of the National Academy of Sciences USA 41: 849-857.

Jerne, N. K. 1966. The natural selection theory of antibody formation; ten years later. In: J. Cairns, G. S. Stent, J. D. Watson (eds.) Phage and the Origins of Molecular Biology. Cold Spring Harbor, NY: Cold Spring Harbor Laboratory Press, pp. 301-312.

Jerne, N. K. 1969. The generation of self-tolerance and of antibody diversity. In: Report of Scientific Group on Factors Regulating the Immune Response. Geneva: World Health Organization, 1-6 September, pp. 1-14.

Jerne, N. K. 1970. Generation of antibody diversity and self tolerance - a new theory. In: R. T. Smith and M. Landy (eds.) Immune surveillance. New York: Academic Press, pp. 343-363.

Jerne, N. K. 1971. The somatic generation of immune recognition. European Journal of Immunology 1: 1-9.

Jerne, N. K. 1974. Towards a network theory of the immune system. Annales d'Immunologie 125C: 373-389.

Lederberg, J. 1959. Genes and antibodies. Do antigens bear instructions for antibody synthesis of do they select cell lines that arise by mutations? Science 129: 1649-1653.

Lederberg, J. 1994. The landmark interviews. The genetic control of antibody synthesis. Journal of NIH Research 6: 72-80.

Lo, D. and Sprent, J. 1986. Identity of cells that imprint H-2-restricted T-cell specificity in the thymus. Nature 319: 672-675.

Marrack, P. and Kappler, J. 1987. The T cell receptor. Science 238: 1073-1079.

Matzinger, P. 1993. Why positive selection? Immunological Reviews 135: 81-117.

Medawar, P. B. 1963. Concluding remarks. In: A. Bussard (ed.) Acquired or Natural Immune Tolerance towards Defined Protein Antigens. New York: Academic Press, p. 475.

Mendel, G. J. 1865. Versuche uber Pflanzenhybriden. Verhandlung des naturforschenden Vereines in Brunn 4: 3-47.

Miller, J. F. A. P. 1961. Immunological function of the thymus. The Lancet 278: 748-749.

Milthorp, P. and Forsdyke, D. R. 1970. Inhibition of lymphocyte activation at high ratios of Con A to serum depends on complement. Nature 227: 1351-1352.

Mitchison, N. A. 1964. Induction of immunological paralysis in two zones of dosage. Proceedings of the Royal Society, Series B, Biological Sciences 161: 275-292.

Nisbet, N. W., Simonsen, M. and Zalenski, M. 1969. The frequency of antigen-sensitive cells in tissue transplantation. A commentary on clonal selection. Journal of Experimental Medicine 129: 459-467.

Nowell, P. C. 1960. Phytohaemagglutinin: an initiator of mitosis in cultures of normal human leukocytes. Cancer Research 20: 462-466.

Podolsky, S. H. and Tauber, A. I. 1997. The Generation of Diversity. Clonal Selection Theory and the Rise of Modern Immunology. Cambridge, MA: Harvard University Press.

Porter, R. R. 1958. Separation and isolation of fractions of rabbit gamma-globulin containing the antibody and antigenic combining sites. Nature 182: 670-671.

Radick, G. 2008. Focus: Counterfactuals and the historian of science. Introduction: Why what if? Isis 99, 547-551.

Roomp, K. and Domingues, F. S. 2011. Predicting interactions between T cell receptors and MHC-peptide complexes. Molecular Immunology 48: 553-562.

Scott-Browne, J. P., White, J., Kappler, J. W., Gapin, L. and Marrack, P. 2009. Germline-encoded amino acids in the αβ T-cell receptor control thymic selection. Nature 458: 1043-1046.

Simonsen, M. 1990. Alloreactive T cells. Scandinavian Journal of Immunology 32: 565-575.

Soderqvist, T. 1994. Darwinian overtones: Niels K. Jerne and the origin of the selection theory of antibody formation. Journal of the History of Biology 27: 481-529.

Soderqvist, T. and Stillwell, C. 1999. Essay review: the historiography of immunology is still in its infancy. Journal of the History of Biology 32: 205-215.

Soderqvist, T. 2003. Science as Autobiography. The Troubled Life of Niels Jerne. New Haven: Yale University Press.

Sprent, J. and Webb, S. R. 1987. Function and specificity of T cell subsets in the mouse. Advances in Immunology 41: 39-133.

Sprent, J., Lo, D., Gao, E. K. and Ron, Y. 1988. T-cell selection in the thymus. Immunological Reviews 101: 173-190.

Tauber, A. 2010. Review of Arthur M. Silverstein's A History of Immunology, and Ed Cohen's A Body Worth Defending. Isis 101: 636-637.

Wen, L., Brill-Dashoff, J., Shinton, S. A., Asano, M., Hardy, R. R. and Hayakawa, K. 2005. Evidence of marginal-zone B cell positive selection in spleen. Immunity 23: 297-308.

      Wigzell, H. 1984. Physiology or medicine 1984. In: J. Lindsten (ed.) Nobel lectures, physiology or medicine 1981-1990. Singapore: World Scientific Publishing Co., 1993, pp. 203-208.

Zinkernagel, R. M. and Doherty, P. C. 1974. Restriction of in vitro T cell-mediated cytotoxicity in lymphocytic choriomeningitis with a syngeneic or semiallogeneic system. Nature 248: 701-702.

Zinkernagel, R. M. and Doherty, P. C. 1997. The discovery of MHC restriction. Immunology Today 18: 14-17.

Zuckerman, H. and Lederberg, J. 1986. Post-mature scientific discovery. Nature 324: 629-631.

Footnotes

¹ Typically a virus (antigen) provokes the synthesis of immunoglobulins (antibodies) that bind the virus and destroy it. The clonal selection theory postulated that one cell of the immune system (lymphocyte) corresponds to one antibody specificity (e.g. specificity for virus A, but not for virus B). An antigen would select a particular cell from the pre-existing repertoire of cells. This cell would then multiply into a clone that would secrete its specific antibody into body fluids (e.g. lymph and blood plasma).

² An organism can prepare an antibody repertoire randomly in the hope that, by chance, there will be at least one antibody that can combine with an invading foreign antigen. Alternatively, the organism can, in advance, shape the repertoire taking into account some property common to certain antigens. This shaping can be said to bias, or restrict, the repertoire. If the antigens are MHC, then the repertoire is said to be 'MHC-restricted.' 

³ An organism has its own antigens. There is a need to suppress a cell that can make antibody against these ('self'), and to stimulate a cell that can make antibody against foreign antigens ('not self'). Thus, an antigen has the potential to give one of 'two signals' to a lymphocyte, which hence is selected either positively (clonal expansion) or negatively (clonal deletion or suppression). 

⁴ Photocopies of diary entries are in Queen's University Archives. These may also be viewed at http://www.queensu.ca/academia/forsdyke/theorimm9.htm. 

⁵ A copy of the paper submitted to Nature in 1966 (F9264) is in Queen's University Archives. 

⁶ I learned of the two signal mechanism for distinguishing signals in a liquid scintillation counter from a manufacturer's technical manual, probably read in 1965. In 1967 I saw this as a helpful marketing metaphor, but do not believe it influenced the emergence of my two signal ideas in 1966.

⁷ I later discovered that Bretcher was a contemporary at Cambridge. As far as I am aware, we did not meet. I had no communication with Cohn or his coworkers until the 1990s.

⁸ An antibody may bind an antigen because of the affinity of an individual combining site, but the observed strength of binding (avidity) may reflect binding at multiple sites. Since it is difficult to distinguish these, the model is referred to as the affinity/avidity model. To simplify, this paper refers mainly to affinity. Likewise, often 'antigen' is used where the term 'antigenic determinant,' or 'epitope,' might be more appropriate.

⁹ Photocopies of dated notes are in Queen's University Archives. These may also be viewed at http://www.queensu.ca/academia/forsdyke/theorimm9.htm. There are other notes,  but these are not clearly dated.

¹⁰ For the synthesis of most antigens, the species 'pack' usually contains only two gene 'cards.'  You hold two - one from your father and one from your mother. So these 'cards' can be either the same (e.g. YY or ZZ), or mixed (e.g. YZ). On the other hand, for the synthesis of MHC antigens the species 'pack' contains many 'cards' (e.g. A, B, C, - - X, Y, Z), but you still can hold only two (e.g. BX). So, if genetic shuffling is thorough (unbiased), it is unlikely that the two MHC antigen 'cards' you inherited from your parents will be the same. This means that the differences between graft and host that determine graft rejection are mainly differences in MHC antigens.

¹¹ When skin from one individual (donor) is grafted to another individual (host), it is rejected by host lymphocytes ('host-versus-graft reaction'). When lymphocytes from one individual (donor) are injected ('grafted') into another individual (host), the lymphocytes attempt to reject their host ('graft-versus-host reaction').

¹² Jerne, N. K. 1969. The generation of self-tolerance and of antibody diversity. A copy of this privately circulated document, bearing the stamp 'Dr. Melvin Cohn, Aug 11 1969,' was received by the present author in October 1970 by way of New Zealand immunologist, John Marbrook. A copy has been placed in Queen's University Archives. This is probably the document referred to by Soderqvist (2003, p. 262): 'Jerne completed his manuscript in the beginning of August 1969 and circulated it to his peers.' Thus, there is probably an identical copy in the Archives of the Royal Library, Copenhagen (Soderqvist 2003, p. 329).

¹³ Bodmer (1972) disagreed on population genetic grounds: 'A principle, almost insurmountable difficulty with Jerne's hypothesis - - is that it seems to require strict parallel evolution at the population level of histocompatibility antigen genes and immunoglobulin v genes. Thus, - - as soon as a new histocompatibility antigen arises by mutation or recombination, the corresponding germ line v gene coding for the antibody specificity directed against this new antigen must somehow almost immediately spread through the whole species.'

¹⁴ Cells of the T cell lineage (cytotoxic T cells), by virtue of their antibody-like surface receptors, can react with antigens at the surface of target cells and destroy the cells.

¹⁵ Jerne postulated subsets A an S. Subset S would be 'cards' BX in the example above, and subset A would be the remaining 'cards' in the species pack (A, C, D, - - W, Y, Z).

End Note June 2011 Eichman's Network Collective

End Note Jan 2012 MHC as Blank Slate Scenario

End Note Feb 2013 Mel backs off

End Note Mar 2013 Absense-of-Knowledge Postulate

End Note Oct 2013 Two Signal Support

End Note Oct 2014 Positive Selection Support

End Note Aug 2016 Natural Antibody Blank Slate Scenario

End Note Sept 2017 Close to Self and Near-Self

End Note Dec 2017 Non-Germ-Line Influence on TCR

End Note May 2018 Germinal Centre B Cell Positive Selection