PERSPECTIVES

PERSPECTIVES Lederberg’s one-signal model (1959) TIMELINE

Activation rules: the two-signal theories of immune activation Alan G. Baxter and Philip D. Hodgkin Two-signal theories of lymphocyte activation have evolved considerably over the past 35 years. In this article, we examine the contemporary experimental observations and theoretical concerns that have helped to forge the most influential variants of the theory. We also propose that more-rigorous quantitative methods are required to sustain theoretical development in the future.

By the 1960s, immunology had come of age, and the main components of a cellular recognition system had been assembled.

Lymphocytes recirculated around the body1 and responded to foreign molecules. Their responses could involve the production of specific antibodies2 or direct cellular attack3,4, and the size of the response was, in part, determined by the number of responding cells5. Burnet’s clonal selection theory 6 provided a rationale for tolerance; the lymphocyte populations that were able to respond to the body’s own tissues were depleted during the window of prenatal, actively acquired tolerance that had been described by Billingham, Brent and Medawar7.

Box 1 | Lederberg’s nine postulates • The stereospecific segment of each antibody globulin is determined by a unique sequence of amino acids. • The cell making a given antibody has a correspondingly unique sequence of nucleotides in a segment of its chromosomal DNA — its ‘gene for globulin synthesis’. • The genetic diversity of the precursors of antibody-forming cells arises from a high rate of spontaneous mutation during their lifelong proliferation. • This hypermutability consists of the random assembly of the DNA of the globulin gene during certain stages of cellular proliferation. • Each cell, as it begins to mature, spontaneously produces small amounts of the antibody corresponding to its own genotype. • The immature antibody-forming cell is hypersensitive to an antigen–antibody complex; it will be suppressed if it encounters the homologous antigen at this time. • The mature antibody-forming cell is reactive to an antigen–antibody complex; it will be stimulated if it first encounters the homologous antigen at this time. The stimulation comprises the acceleration of protein synthesis and the cytological maturation which mark the ‘plasma cell’. • Mature cells proliferate extensively under antigenic stimulation but are genetically stable, and therefore generate large clones genotypically pre-adapted to produce the homologous antibody. • These clones tend to persist after the disappearance of the antigen, retaining their capacity to react promptly to its later reintroduction.

NATURE REVIEWS | IMMUNOLOGY

Lederberg’s nine postulates8, which were drafted while visiting Burnet’s laboratory at the Walter and Eliza Hall Institute, adequately summarized the state of immunological theory at that time (BOX 1). Together, these postulates formed the basis of a one-signal model of lymphocyte activation. This model incorporated a temporal switch that governed the outcome of antigenic stimulation — switching from suppression (Burnet suggested deletion) of immature lymphocytes to activation of mature lymphocytes. The model deviated significantly from Burnet’s theory of immunological tolerance9, as modified by his clonal selection theory10, which stated that lymphocytes were susceptible to elimination only “in the late embryonic period with the concomitant development of immune tolerance”. Talmage and Pearlman (1963)

Even at the time of its development, there were experimental data that did not easily fit Lederberg’s one-signal model. Hapten– carrier phenomena had been studied ever since Landsteiner11 originally divided antigens into two classes. The first class,‘carriers’, were themselves immunogenic, and antibodies could be raised against them easily. The second class, ‘haptens’, were not immunogenic unless administered conjugated to a carrier, in which case antibodies that were specific for both the hapten and the carrier parts of the hybrid molecule could be produced. Clearly, the structure of the antigen was contributing to the outcome. Furthermore, Dresser12 subsequently published experiments that indicated that reactive immunocytes could be either tolerized or activated, depending on the physical properties of the antigen. An attempt to accommodate all of these data within a theoretical framework was made by Talmage and Pearlman13. They proposed that although antigen alone could induce the maturation of a lymphoid cell into a nondividing plasma cell, this resulted in minimal

VOLUME 2 | JUNE 2002 | 4 3 9

PERSPECTIVES antibody production and led to tolerance, as the antigen-reactive cell pool was now depleted. By contrast, they suggested that aggregates of antigen — perhaps associated with complement — could deliver an additional nonspecific stimulus, and the combination of antigen-specific and -nonspecific stimuli would trigger substantial clonal expansion and significant production of antibody. “Since it is unlikely that natural proteins possess closely spaced identical determinants,” they wrote, “the fixation of complement probably requires the presence of [pre-existing] antibody directed to two or more different determinants”. By 1968, studies of hybrid antigens had produced other observations that were inconsistent with Burnet’s and Lederberg’s models. In 1967, Rajewsky and Rottlander14, and Mitchison15 applied the concept of hapten–carrier hybrids to a range of naturally occurring molecules, and they reported that tolerance to an antigen could be broken by immunization with that antigen conjugated to an immunogenic epitope. This result provided evidence that cellular collaboration underpins many immune responses. Mitchison16 proposed a model that involved two or more receptors, but confessed,“we are reluctant on general grounds to postulate more than one specificity of receptor in an individual antigen-sensitive cell … therefore cooperation must occur prior to cell stimulation”. A further important problem for one-signal models was also emerging; like other dividing mammalian cells17, lymphocytes presumably continue to mutate at a basal rate, and there was some indication that the generation of high-affinity antibody might involve further mutation. So, the specificity of responding lymphocytes could drift towards self-reactivity, which indicated the need for a tolerogenic mechanism throughout the life of the clone.

would initiate cellular activation. They suggested specifically that the activating configuration of repeated determinants would be the result of a second antibody (the carrier antibody) binding to a second, independent determinant on the antigen, which would produce a defined spatial distribution of bonds. Activation

The involvement of ‘carrier antibody’ in the initiation of the second form of signal explained the apparent requirement for at least two antigenic epitopes for immunogenicity. Note that, in this model, the two signals are different, mutually exclusive signals that achieve two different outcomes; there is Paralysis

a Lederberg, 1959

h c

1

1

Mature

h c

Young

b Bretscher and Cohn, 1968

h c

2

1

h c

1

h c

1

h c

h c

c Bretscher and Cohn, 1970 ('old' model)

1

h c

2

h c

d Bretscher and Cohn, 1970 (modern synthesis)

Bretscher and Cohn (1968)

Bretscher and Cohn attempted to accommodate hapten–carrier phenomena within the framework of self–non-self discrimination, taking into account the problem that was presented by the constant threat of a clone mutating towards self-reactivity. Their first model, published in 1968 (REF. 18), proposed that antigen receptors on the surface of immunocytes were able to transmit two qualitatively different signals. Should a single free determinant bind to a receptor, a specific signal would be generated that would induce cellular paralysis (or death) of the lymphocyte. By contrast, binding to an appropriate aggregation of two or more determinants would induce a second type of signal that

440

| JUNE 2002 | VOLUME 2

c h

1

B cell

T cell 2

c

Figure 1 | The evolution of Bretscher and Cohn’s associative recognition model. An immuneresponsive cell carries receptors (antibodies) on its surface (green). a | In Lederberg’s model8, the outcome of antigenic stimulation (activation or paralysis of the immune-responsive cell) depended on the timing of stimulation (mature or young immunocytes, respectively). By contrast, in Bretscher and Cohn’s models, the outcome of stimulation depended on associative antigen recognition. b | In their original model18, activation required that a carrier antibody bound the carrier portion (c) of a hybrid antigen, and that the responding immunocyte bound the hapten portion (h). If carrier antibody was not bound, then an inhibitory signal was generated. c | Two years later, Bretscher and Cohn20 replaced this system of two mutually exclusive signals with a form of cellular calculus, in which a second signal was used to interpret the appropriate response to antigenic stimulation. d | This model is now interpreted in terms of T-cell-dependent B-cell stimulation; a carrier peptide (c) that is presented by an MHC molecule (blue) on the surface of a B cell triggers a T cell through its antigen receptor (pink) to provide a second signal to the B cell.

www.nature.com/reviews/immunol

PERSPECTIVES no logical summation of multiple signals. Furthermore, tolerance induction is effective throughout the life of the cell — which ensures that clones that mutate to being selfreactive are deleted — because of the low probability that a second self-antigen-specific clone will be generated simultaneously to provide the self-reactive carrier antibody. Bretscher and Cohn briefly considered the possibility that the carrier antibody could be present on the surface of another antigenspecific cell, but did not think this probable because,“the induction of an antigen-sensitive cell would then require an antigen molecule to interact simultaneously with two cells which are presumably rare”. The main problem with their theory was the origin of the carrier antibody. Although they proposed that the carrier antibody be of a special class, there was no qualitative difference described in the model between the haptenspecific antibody and the carrier antibody. The lymphocyte that produced the carrier antibody would, therefore, presumably have the same activation requirements as that producing the hapten-specific antibody. It would also need its own carrier antibody to ensure stimulation instead of paralysis — forming a circular paradox that is now known as the ‘primer problem’19. The only solution offered was that “such carrier antibody may be acquired early in life from maternal sources”. Associative recognition (1970)

When Bretscher and Cohn redrafted their two-signal model in 1970 (REF. 20), they began the manuscript by summarizing their 1968 position — but with some interesting and important changes. In 1968, the two signals (either stimulatory or inhibitory) were mutually exclusive and both were mediated through the surface-bound receptor of the lymphocyte (FIG. 1). By 1970, the carrier antibody mediated its own ‘second’ signal. This second signal was used to interpret the appropriate response to signal one, the haptenspecific signal (FIG. 1). The two signals were no longer mutually exclusive, and the lymphocyte was now responsible for integrating them. A form of cellular calculus was implied. Another important modification was influenced by the recent realization that there were at least two types of lymphocyte; the production of antibody had been associated with the bursa of Fabricius in chickens21 and the recognition of histocompatibility antigens with the thymus22. It was known that the lymphocytes that were produced by each of these organs — with the bone marrow taking the role of the bursa in mammals — were synergistic for the production of antibody23,24.


Figure 2 | Kevin J. Lafferty. Photograph courtesy of Marc Fenning, Photography, John Curtin School of Medical Research.

“The simplest interpretation of these results,” wrote Bretscher and Cohn20,“[is that] the formation of carrier-antigen-sensitive cells is thymus dependent, whereas humoral-antigensensitive cells are derived from the bone marrow”. Mitchison subsequently confirmed that the helper activity of the carrier-antigensensitive cells was concentrated in lymphocytes of thymic origin25,26, which showed that T-cell-dependent B-cell activation involved the associative recognition of two distinct antigenic determinants. Bretscher and Cohn20 maintained their earlier position that the induction of carrier antibody required the same activation conditions as the induction of hapten-specific antibody, so invoking the ‘primer problem’. As each antigen-specific receptor was associated with a different type of lymphocyte, however, this was no longer necessarily a component of the model19,27, because different types of lymphocyte could reasonably be expected to have different activation conditions. Lafferty and Jones (1969)

By the late 1960s, the contribution of MHC genes to tissue incompatibility — in particular, transplant rejection and graft-versus-host disease — had emerged as an intriguing phenomenon. A particular focus of discussion was the high frequency of lymphocytes that responded to foreign tissues and, consequently, the surprising vigour of the allogeneic response. For practical reasons, Lafferty (FIG. 2) and Jones28 chose to study graft-versus-host disease by inoculating the allantoic membrane of

viable chicken eggs with mature lymphocyte preparations — either allogeneic or xenogeneic. This model had been validated previously by Simonsen29, and the resulting growth of lymphocyte colonies was regarded widely as being consistent with Burnet’s theory of clonal selection. To their surprise, Lafferty and Jones28 found that “as the genetic relationship between donor and recipient becomes more distinct, the degree of reactivity falls to an undetectable level”. For example, chicken lymphocytes injected into a genetically distinct chicken egg vigorously attacked the host cells. By contrast, grafted lymphocytes from a pigeon gave a much diminished graft-versushost reaction, and no response was initiated by the injection of sheep or mouse lymphocytes28. Lafferty and Jones concluded that the reactivity of xenogeneic cells was always less than that of allogeneic cells, a paradox that had been reported previously, but not explained adequately (reviewed in REF. 29). Over the course of a series of experimental papers with different collaborators, Lafferty developed a new theory to explain allogeneic interactions. He proposed that something more than antigen was required to stimulate an allograft response and that the interaction of a lymphocyte with a histoincompatible target triggered some of the foreign antigenic cells (the ‘stimulator’ cells) to produce a strong proliferative stimulus. It was the need for lymphocytes to recognize this second signal in addition to the antigenic difference, Lafferty argued, that accounted for the requirement for species compatibility and, therefore, explained the paradox of alloreactivity (FIG. 3). Lafferty called this second signal the ‘allogeneic stimulus’30. The many unanswered questions that were associated with this theory were a significant spur to further experiments. In the early 1970s, Lafferty and co-workers determined that the cells that produced the allogeneic stimulus were derived from the haematopoietic system and that they had to be metabolically active to stimulate lymphocytes28,30. Also, they found that, once generated, activated cytotoxic T lymphocytes were able to kill any cell that expressed the foreign antigen — that is, once activated, the requirement for the allogeneic stimulus was lost30. On the basis of these results, Lafferty proposed that the donor haematopoietic cells that are carried within grafted tissue provided a potent activation signal for the host immunocytes, which then attacked the main body of the graft. These ‘passenger leukocytes’ were, therefore, the main barrier to graft acceptance. He found that these cells could be removed from a graft by briefly culturing the tissues that

VOLUME 2 | JUNE 2002 | 4 4 1

PERSPECTIVES were to be transplanted, and that thyroid tissue treated in this way could be grafted indefinitely across a histoincompatible barrier31. In remarkable agreement with his theory, Lafferty found that the injection of a small number of spleen or peritoneal-exudate cells from the donor could induce the rapid rejection of a previously tolerated graft32. This sequence of experiments showed unequivocally that antigen-bearing stimulator cells,

and not antigen itself, were the main barrier to allograft acceptance, and that activation of the T-cell response in this case was the unique property of these cells.

paper proposed that the concept of the stimulator cell should be extended from the consideration of allogeneic reactions to that of all cellular immune responses.“We would modify [Bretscher and Cohn’s] theory somewhat to suggest that ‘signal two’ is provided by a stimulator cell … which is bound to the responsive cell … by means of an antigen bridge,” they wrote. “Normal antigen induction now has the same general form as an allogeneic

Lafferty and Cunningham (1975)

In 1975, Lafferty, together with a new collaborator, Cunningham, produced a remarkably comprehensive manuscript entitled “A new analysis of allogeneic interactions”33. This

a Theory of allogeneic interactions

b Preventing graft rejection

? L

R

Activation

S

Allogeneic stimulus ? NL

R

No activation ? or signal (1) Graft rejection S

R

Allogeneic stimulus or signal (2)

c

Antigen

d

Ab

Ab

(1)

S

(1)

R

S

R'

(2) Stimulator cell

(2) Responder cell

e

R

Stimulator cell

Activated responder cell

Responder cell

MHC forms compound with antigen (1) T cell

Activated T cell

S MHC triggers second signal

(2)

All activated T cells are MHC restricted

Figure 3 | The development of the two-signal theory in Kevin Lafferty’s laboratory. a | The theory of allogeneic interactions. The immune-responsive cell (R) interacts with a foreign lymphoid cell (L). This interaction leads to the provision of a potent allogeneic stimulus; the stimulus was proposed to be species specific, thereby accounting for the fact that allogeneic interactions lead to more vigorous responses than xenogeneic ones. Whether the responsive cell had to receive an antigenic signal was not known (as indicated by the question mark). However, it was clear that some foreign cells could be antigenic, but without providing an allogeneic stimulus (non-lymphoid cells; NL). b | Lafferty proposed a model to explain the initiation of graft rejection. The panel shows graft tissue containing resident donor cells (stimulatory cells; S) that are able to provide an allogeneic stimulus to the host immune system (responsive cells; R). Once activated, host T cells could attack the main body of the graft, leading to graft rejection. This theory was consistent with all of the versions of Lafferty’s two-signal model. c | In 1975, Lafferty and Cunningham identified a basic similarity between allogeneic interactions and normal lymphocyte activation. Their model of normal lymphocyte activation used Bretscher and Cohn’s ‘signal one’ and ‘signal two’ terminology and borrowed the idea that the cell operated as a logical device, requiring both signals before becoming activated. Signal two in this model was provided by the stimulatory cell after an antigen-dependent interaction with the responsive cell. d | Lafferty and Cunningham’s equivalent scheme for allogeneic interaction, in which the stimulatory cell passively presented the antigen and also responded to cellular interactions with a responsive cell by producing signal two. e | In 1977, Lafferty and Cunningham presented a version of their scheme that accounted for the MHC restriction of T cells by placing control of the production of the second signal with MHC engagement on the stimulating-cell surface.

442



PERSPECTIVES interaction. In both cases, induction depends not on a simple antigen–cell contact (signal one), but on the coming together of a stimulator and a responder cell.”When the two cells came together, the release of a second signal (later termed a ‘co-stimulator’34) by the stimulator cell would complete the requirements for activation (FIG. 3). In attempting to use their general scheme to explain all types of immune activation, Lafferty and Cunningham did not distinguish between T and B cells. For this reason, the requirement of Bretscher and Cohn’s theory for associative recognition was not always satisfied. For example, although their scheme for ‘normal’ antigen-mediated immune induction involved antigen being presented to the reactive lymphocyte by carrier antibody, the constraint of associative recognition was relaxed for allogeneic stimulation, for which there was clearly only one antigen specificity involved. This reintroduced the possibility of peripheral tolerance being broken if an immunocyte mutated towards self-reactivity. “There is a risk of autoimmunity associated with this mechanism of activation”, Lafferty and Cunningham wrote.“If an immunocyte were to spontaneously arise with receptors for self-antigens, these receptors could mediate the combination of this cell with a stimulator cell. The result would be the induction of an autoimmune clone.” Their solution was to retain from Bretscher and Cohn the idea that stimulation with signal one in the absence of signal two would favour tolerance induction. So, they argued,“in the self-environment, this risk would be low because soluble antigen present in the tissue fluids would favour tolerance induction by delivering signal one alone to the potentially responsive cell”. As stimulator cells are rare and do not present self-antigens exclusively, the large number of non-stimulating cells should ensure a tolerogenic environment. In fact, Lafferty and Cunningham borrowed a lot more from Bretscher and Cohn than just the idea of two signals and a default for tolerance. Their model had a marked formulaic similarity to Bretscher and Cohn’s 1970 model — the allogeneic stimulus was converted from being a nonspecific inducer to a participant in a logical operation, in which the responding lymphocyte adapted its response to signal one according to the presence or absence of signal two.“Our model was derived from Bretscher and Cohn’s”, Lafferty told us,“there’s no doubt about that”35. MHC as a second signal (1977)

The discovery of MHC restriction by Zinkernagel and Doherty36 was made just down the corridor from Lafferty’s laboratory


at the John Curtin School in Canberra, and it further linked alloreactivity with normal T-cell behaviour in his thoughts. Lafferty and Cunningham37 suggested a new version of their general scheme, which was specific for T cells and in which the MHC-encoded molecule was the trigger for the stimulator cell to provide a co-stimulatory signal. As a consequence of this, only T cells that saw antigen in the context of a self-MHC product (they referred to it as a ‘compound’ or ‘interaction’ antigen) could ever induce the production of a second signal. T cells that recognized antigenic differences, but not the framework of the MHC product, would receive the first signal, but not the essential second signal. So, all T cells that are capable of activation must be restricted to a particular allelic MHC product. This very elegant version of the two-signal theory returned the site of control to the responding lymphocyte and economically explained allogeneic interactions, MHC restriction, normal T-cell activation and the need for co-stimulation in one theory37,38. Jenkins and Schwartz (1987)

The discovery of T-cell growth factor (interleukin-2; IL-2) allowed the prolonged culture of T-cell clones and the detailed exploration of their requirements for activation. Although this technique produced T cells with uniform specificities, these cells were ‘antigen experienced’39 and were, therefore, likely to have different activation requirements from naive T cells30. Despite these concerns, a resting state could be induced after a rapid phase of growth, and Jenkins and Schwartz40 chose to examine the proliferative responses of several clones after restimulation with defined antigens that were presented in different ways. They found that an unresponsive state could be induced when antigen was presented by killed or fixed antigen-presenting cells (APCs), and that viable APCs were required to provide an essential second signal for full activation40. Antigen alone seemed to lead to an inert state, which was later termed ‘anergy’, in analogy to a similar state that had been reported in B cells41. These experiments provided important confirmation that an antigenic stimulus in the absence of co-stimulation could inactivate a mature T cell. Although the demonstration that this effect could be overcome by a nonantigen-dependent stimulus from a viable APC was operationally identical to Lafferty and Cunningham’s33 two-signal model, Jenkins and Schwartz were more taken with the similarities to Bretscher and Cohn’s 1970 model20. They wrote, “The model that we have at the present time which, we feel, best

explains the results is a modification of the model originally proposed by Bretscher and Cohn”42. Cohn was unhappy with this interpretation of Jenkins and Schwartz’s model, because it did not involve associative recognition — the crucial feature of his model. “I mistook his concerns at the time for a reluctance to abandon the T-cell receptor model”, Schwartz wrote, “It was not until several years later — after much harping by Polly Matzinger — that I appreciated his real reluctance stemmed from his difficulty in accepting the non-antigen specificity of the second signal”43. Jenkins and Schwartz helped to popularize a hybrid form of the Bretscher and Cohn, and Lafferty and Cunningham two-signal models. Accepting that T-helper activation did not require associative recognition solved the primer problem, but at a terrible cost — it was no longer clear how immunological tolerance could be maintained. The hybrid model had transferred the decision about which cells to activate and, therefore, the preservation of self-tolerance to the stimulatory cell. This solution brought its own problems, as Matzinger noted: “The upshot is that the presence or absence of co-stimulation cannot account for self tolerance because the APC does not distinguish self from non-self”39. Janeway’s infectious non-self (1989)

Charles Janeway introduced the published proceedings of the 1989 Cold Spring Harbor Symposium on Immune Recognition44 with a manuscript that denounced as a fallacy Landsteiner’s work on the specificity of immunological reactions.“The Landsteinerian fallacy”, he wrote,“is the idea that the immune system has evolved to recognise equally all nonself substances … I contend that [it] has evolved specifically to recognise and respond to infectious organisms, and that this involves recognition not only of specific antigenic determinants, but also of certain characteristics or patterns common on infectious organisms but absent from the host”. He pointed out that for Landsteiner and others who were working on hapten–carrier systems to raise antibodies against innocuous antigens, these antigens usually had to be mixed with adjuvants that contained killed bacteria, such as Mycobacterium tuberculosis or Bordetella pertussis.“I call this the immunologists’ dirty little secret”, he wrote. Although the innate immune system and its pattern recognition of pathogens by germline-encoded receptors had already been recognized as being important for host defence, Janeway 45 proposed that these receptors had the further role of being gatekeepers

VOLUME 2 | JUNE 2002 | 4 4 3

PERSPECTIVES

Timeline | The evolution of two-signal theories of immune activation

Landsteiner distinguished between carriers and haptens.

1930

Burnet proposed that discrimination between self and non-self is established in early development.

1944

1944–1945 Medawar characterized allograft rejection.

1949

1954

Basal rates of somatic-cell mutation were measured by Russell.

1956

Billingham, Brent and Medawar supported Burnet’s prediction of prenatally acquired tolerance.

of the adaptive immune system. His model incorporated Bretscher and Cohn’s associative recognition model for T-cell-dependent B-cell activation, but rejected it for CD4+ T-cell activation because of the primer problem. Accepting the need for a signal in addition to antigenic stimulation, he adapted the Lafferty and Cunningham model. Janeway proposed that immunogenicity required both signalling through the antigen receptor and a second signal that was induced on host APCs by infectious agents as a result of pattern recognition of the constituents of microorganisms. So, in Janeway’s model, Lafferty’s second signal was not triggered by the T cell engaging the APC, but was instead triggered by common microbial products binding germline-encoded receptors on (or within) the APC. He suggested that the antigen-nonspecific triggers of APC activation could even operate before the development of specific antigen recognition and might, therefore, be “viewed more as positive initiators of immunity than as late adaptations to avoid autoimmunity”. This proposal accommodated the facts that bacterial products do act as adjuvants and that this effect is mediated by germlineencoded receptors. Furthermore, Janeway’s hypothesis has been productive; it has had an important role in the characterization of Tolllike molecules as pathogen receptors. Toll, a transmembrane protein that was identified originally as being required for dorsal–ventral polarity in the Drosophila embryo, can also activate the immune responses of Drosophila haemocytes. Medzhitov, working in Janeway’s laboratory, showed that homologues of Toll are also important in vertebrate immune responses to infection46. He showed that a constitutively active mutant of Toll that was

444


Gowans reported the recirculation of lymphocytes.

1957

1959

Burnet published his clonal selection theory.

1969–1974 Lafferty’s theory of alloreactivity accounted for the low immunogenicity of xenogeneic grafts.

Talmage and Pearlman suggested that antigen aggregates produce a second, nonspecific stimulus.

1961

Bretscher and Cohn’s two-signal model proposed that there is cooperative associative recognition between T cells and B cells.

Hapten–carrier systems revealed the dual specificity of some immune responses.

Lederberg proposed that tolerance could be induced in immature lymphocytes at any time.

1963

Evidence accumulated that tolerance requires continuous exposure to (self-)antigen.

1966

1966–1968 T-cell–B-cell collaboration was reported.

transfected into human cell lines induced the activation of nuclear factor-κB (NF-κB) and the expression of the inflammatory cytokines IL-1, IL-6 and IL-8, as well as the expression of the co-stimulatory molecule B7.1 (CD80). So far, ten mammalian Toll-like receptors have been identified, and their known bacterial ligands include lipopolysaccharide47, peptidoglycan48 and flagellin49. Paradoxically, although Bretscher and Cohn’s associative recognition model required two recognition signals of the same type (antigenic), their model described a very different outcome if only one of the two signals was received rather than both. Janeway’s model required two very different signals — one antigenic, one pattern recognition — yet the outcome of failed pattern recognition was the same in a host who possessed an adaptive immune system as one who did not. One might, therefore, argue that such a dual system of activation is needlessly redundant. Janeway 45 recognized this problem, writing, “A successful pathogen (could) simply avoid these receptors and … thus induce tolerance rather than immunity”. He proposed that mature dendritic cells, which are characterized by poor antigen uptake but constitutive co-stimulatory activity, could have evolved specifically to deal with such pathogens. Despite this, it remains the case that it is frequently possible for an effective immune response to be generated against foreign cells (for example allogeneic cells) without either signalling from innate receptors or the involvement of dendritic cells. Matzinger’s danger hypothesis (1994)

Like Janeway, Matzinger39 accepted Bretscher and Cohn’s associative recognition model for

1967

1968

1969

1970

Bretscher and Cohn’s carrier-antibody theory linked detection of the number of foreign determinants to activation or tolerance.

T-cell-dependent B-cell activation, as well as the Lafferty and Cunningham two-signal model for CD4+ T-cell activation. She also placed the control of immune activation with the APCs, which administered the second signal under appropriate circumstances. Where she differed from Janeway was in terms of the nature of the stimulus for a second signal. Matzinger proposed that APC activation was induced by various triggers that are associated with host-cell damage, termed ‘danger signals’. These stimuli were divided into two groups; the first group encompassed exclusively intracellular components that are released when cells are damaged (such as DNA, RNA and mitochondria), whereas the second group contained ‘inducible alarm signals’, such as heatshock proteins and interferons. As pathogens, by definition, induce tissue damage, the recognition of that damage could provide a crucial validation for any immune response. Another important difference between the ‘danger hypothesis’ and Janeway’s model (which is sometimes known as the ‘stranger hypothesis’) is that Matzinger strongly emphasized the role of tolerance. Her model predicted that any T cell that is stimulated through its antigen receptors in the absence of a second signal would be inactivated. So, selfreactive T cells that emigrate from the thymus would be rendered harmless as soon as they met healthy tissues. The danger hypothesis neatly circumvents the main problems that were identified in Janeway’s model. For example, a successful pathogen that had evolved to avoid triggering damage receptors would probably have ceased to damage tissues, and would no longer be a pathogen, but a commensal. Furthermore, the hypothesis is consistent with the observation


PERSPECTIVES

Lafferty and Cunningham’s twosignal model generalized the role of a stimulating-cell-derived second signal in immune activation. Confirmation of associative recognition in T-cell-dependent B-cell responses.

1971

Allogeneic stimulating cells were shown to activate graft rejection.

1975

1977

Lafferty and Cunningham’s MHC triggering model explained allogeneic interactions, MHC restriction, normal T-cell activation and the need for co-stimulation.

Jenkins and Schwartz experimentally induced T-cell unresponsiveness with signal one alone.

Matzinger described her danger hypothesis, in which the second signal was triggered by the detection of tissue damage.

1987

1994

Janeway described his theory of infectious non-self as a second signal, in which the second signal was triggered by microbial products.

that foreign cells can generate effective immune responses, because, although it is relatively easy to avoid bacterial contamination, it is virtually impossible to avoid any cellular damage during the preparation and systemic administration of allogeneic cells. Future developments

The most surprising recurring idea throughout these models is the concept that stimulation requires something else together with an antigen signal. Matzinger called it her “First law of lymphotics” (without any apparent apology to Asimov): “Die if you receive signal one in the absence of signal two”39. Janeway wrote that “Co-stimulatory molecules are absolutely essential to the activation of naive T cells…”50. The problem with this idea is that it isn’t quite true. The tools of the molecular revolution have greatly facilitated the identification of costimulators. An important second signal for B cells has been identified as being provided by CD40L, which is expressed on the helper T-cell surface after activation. At this point, the T-cell surface alone is strongly stimulatory and will activate even naive B cells. So, B cells can be activated by ‘signal two’ alone, although the final outcome is determined by the net contribution of several positive and negative ligand interactions51. Clearly, although the activation of T-cell-dependent B cells operationally conforms to associative recognition, it does not fit a simple binary system whereby both signals must be received simultaneously. If anything, the status of two-signal theories in our understanding of T-cell behaviour is even less clear. A series of co-stimulators has been identified that includes cytokines such as IL-1, IL-6 and IL-4, as well as cell-surface


1989

1997

Medzhitov reported mammalian homologues of Toll, which supported Janeway’s theory of infectious non-self as a second signal.

molecules that are expressed on APCs, such as B7.1, B7.2 (CD86) and CD40. The term ‘costimulator’ has now acquired the pragmatic definition of anything that can enhance proliferation in a T-cell-stimulation assay in which signal one is relatively weak — for example, in the presence of low concentrations of anti-CD3 antibody. In vivo, however, it is clear from gene-knockout studies that none of these molecules is obligatory as a T-cell second signal. We believe that two mindsets contribute to these difficulties. The first is the level at which these theories operate. Since the theories of Burnet, immunological models have considered interactions primarily at the cellular level and have placed the decision-making unit within a cell or a pair of cells. With the exception of certain phenomena, such as the role of immune deviation in determining the class of immune responses, there has been little discussion of population distributions of cellular characteristics or of features of immune responses that emerge only at the level of the population, organ or whole organism. For example, T-cell growth is largely dependent on autocrine factors and is, therefore, affected profoundly by the precursor frequency of reactive cells. So, the strength of the response — even the distinction between self and non-self — can differ without there being any significant difference in the proliferative signals that are received by individual T cells52,53. The second problem is that these theories are qualitative. As a consequence, the experimental methods that are used to test them, such as co-stimulation assays, are usually interpreted in a black-and-white manner — in terms of whether the particular molecule

was or was not a co-stimulator. The field is in a cycle — the theories are qualitative, so the experiments are. We maintain that predictive power is greater for quantitative models, and that progress will depend on new systems for the quantitative description of cellular signal integration and immunological outcomes. For example, after an antigenic signal, costimulators decrease T-cell cycle time in an additive fashion and, therefore, relatively small changes in conditions contribute to remarkably large differences in cell number53. A theory of signal integration developed in this way can account for the fact that no single component is obligatory, as well as explain why the many co-stimulation assays give such dramatic differences in response. We expect that by following a more quantitative course, our understanding of the immune system can proceed to a richer level that will accommodate a theory of immuneresponse class discrimination, as well as incorporate the many molecular contributors to decision making. Although we should pay due deference to the past attempts to formulate general principles from immensely complex data (TIMELINE), it is time to move on to a new era. Alan G. Baxter is at the Centenary Institute of Cancer Medicine and Cell Biology, Locked bag #6, Newtown, New South Wales 2042, Australia. Philip D. Hodgkin is at the Walter and Eliza Hall Institute, c/o Post Office, Royal Melbourne Hospital, Parkville, Victoria 3050, Australia. Correspondence to A.G.B. e-mail: [email protected] doi:10.1038/nri823 1.

Gowans, J. L. The recirculation of lymphocytes from blood to lymph in the rat. J. Physiol. (Lond.) 146, 54–69 (1959). 2. Fagraeus, A. Antibody production in relation to development of plasma cells: in vivo and in vitro experiments. Acta Med. Scand. 130, 3–116 (1948). 3. Medawar, P. B. The behaviour and fate of skin autografts and skin homografts in rabbits. J. Anat. 78, 176–199 (1944). 4. Medawar, P. B. A second study of the behaviour and fate of skin homografts in rabbits. J. Anat. 79, 157–176 (1945). 5. Billingham, R. E., Brent, L., Medawar, P. B. & Sparrow, E. M. Quantitative studies on tissue transplantation immunity. II. The origin, strength and duration of actively and adoptively acquired immunity. Proc. R. Soc. Lond. B 143, 58–80 (1954). 6. Burnet, F. M. The Clonal Selection Theory of Acquired Immunity (Vanderbilt Univ. Press, Nashville, Tennessee, 1959). 7. Billingham, R. E., Brent, L. & Medawar, P. B. Actively acquired tolerance of foreign cells. Nature 172, 603–606 (1953). 8. Lederberg, J. Genes and antibodies. Science 129, 1649–1653 (1959). 9. Burnet, F. M. & Fenner, F. The Production of Antibodies (Macmillan, London, 1949). 10. Burnet, F. M. A modification of Jerne’s theory of antibody production using the concept of clonal selection. Aust. J. Sci. 20, 67–69 (1957). 11. Landsteiner, K. The Specificity of Serological Reactions (Harvard Univ. Press, Cambridge, Massachusetts, 1945). 12. Dresser, D. W. Specific inhibition of antibody production. II. Paralysis induced in adult mice by small quantities of protein antigen. Immunology 5, 378–388 (1962).

VOLUME 2 | JUNE 2002 | 4 4 5

PERSPECTIVES 13. Talmage, D. W. & Pearlman, D. S. The antibody response: a model based on antagonistic actions of antigen. J. Theor. Biol. 5, 321–339 (1963). 14. Rajewsky, K. & Rottlander, E. Tolerance specificity and the immune response to lactic dehydrogenase isoenzymes. Cold Spring Harb. Symp. Quant. Biol. 32, 547–554 (1967). 15. Mitchison, N. A. Antigen recognition responsible for the induction in vitro of the secondary response. Cold Spring Harb. Symp. Quant. Biol. 32, 431–439 (1967). 16. Mitchison, N. A. Recognition of antigen by cells. Prog. Biophys. Mol. Biol. 16, 3–14 (1966). 17. Russell, W. L. A high rate of somatic reversion in the mouse. Genetics 41, 658 (1956). 18. Bretscher, P. A. & Cohn, M. Minimal model for the mechanism of antibody induction and paralysis by antigen. Nature 220, 444–448 (1968). 19. Bretscher, P. The control of humoral and associative antibody synthesis. Transplant. Rev. 11, 217–267 (1972). 20. Bretscher, P. & Cohn, M. A theory of self–nonself discrimination. Science 169, 1042–1049 (1970). 21. Glick, B., Chang, T. S. & Jaap, R. G. The bursa of Fabricius and antibody production. Poult. Sci. 35, 224–226 (1956). 22. Szenberg, A. & Warner, N. L. Dissociation of immunological responsiveness in fowls with a hormonally arrested development of lymphoid tissue. Nature 194, 146–147 (1962). 23. Claman, H. N., Chaperon, E. A. & Triplett, R. F. Thymus–marrow-cell combinations — synergism in antibody production. Proc. Soc. Exp. Biol. Med. 122, 1167–1171 (1966). 24. Mitchell, G. F. & Miller, J. F. A. P. Immunological activity of the thymus and thoracic-duct lymphocytes. Proc. Natl Acad. Sci. USA 59, 296–303 (1968). 25. Mitchison, N. A. The carrier effect in the secondary response to hapten–protein conjugates. II. Cellular cooperation. Eur. J. Immunol. 1, 18–27 (1971). 26. Boak, J. L., Mitchison, N. A. & Pattison, P. H. The carrier effect in the secondary response to hapten–protein conjugates. III. The anatomical distribution of helper cells and antibody-forming cell precursors. Eur. J. Immunol. 1, 63–65 (1971). 27. Bretscher, P. A. A two-step, two-signal model for the primary activation of precursor helper T cells. Proc. Natl Acad. Sci. USA 96, 185–190 (1999). 28. Lafferty, K. J. & Jones, M. A. S. Reactions of the graftversus-host (GVH) type. Aust. J. Exp. Biol. Med. Sci. 47, 17–54 (1969). 29. Simonsen, M. Graft-versus-host reactions. Their natural history and applicability as tools of research. Prog. Allergy 6, 349–360 (1962). 30. Lafferty, K. J., Misko, I. S. & Cooley, M. A. Allogeneic stimulation modulates the in vitro response of T cells to transplantation antigen. Nature 249, 275–276 (1974). 31. Lafferty, K. J., Cooley, M. A., Woolnough, J. A. & Walker, K. Z. Thyroid allograft immunogenicity is reduced after a period in organ culture. Science 18, 259–261 (1975). 32. Talmage, D. W., Dart, G., Radovich, J. & Lafferty, K. J. Activation of transplant immunity: effect of donor leukocytes on thyroid allograft rejection. Science 191, 385–388 (1976). 33. Lafferty, K. J. & Cunningham, A. J. A new analysis of allogeneic interactions. Aust. J. Exp. Biol. Med. Sci. 53, 27–42 (1975). 34. Lafferty, K. J., Warren, H. S., Woolnough, J. A. & Talmage, D. W. Immunological induction of T lymphocytes: role of antigen and the lymphocyte costimulator. Blood Cells 4, 395–404 (1978). 35. Baxter, A. G. Germ Warfare: Breakthroughs in Immunology (Allen and Unwin, Sydney, 2000). 36. Zinkernagel, R. M. & Doherty, P. C. Restriction of in vitro T-cell-mediated cytotoxicity in lymphocytic choriomeningitis within a syngeneic or semiallogeneic system. Nature 248, 701–702 (1974). 37. Cunningham, A. J. & Lafferty, K. J. A simple conservative explanation of the H-2 restriction of interactions between lymphocytes. Scand. J. Immunol. 6, 1–6 (1977). 38. Lafferty, K. J., Andrus, L. & Prowse, S. J. Role of lymphokine and antigen in the control of specific T-cell responses. Immunol. Rev. 51, 280–314 (1980). 39. Matzinger, P. Tolerance, danger and the extended family. Annu. Rev. Immunol. 12, 991–1045 (1994). 40. Jenkins, M. K. & Schwartz, R. H. Antigen presentation by chemically modified splenocytes induces antigen-specific T-cell unresponsiveness in vitro and in vivo. J. Exp. Med. 165, 302–319 (1987). 41. Nossal, G. J. & Pike, B. L. Clonal anergy: persistence in tolerant mice of antigen-binding B lymphocytes incapable of responding to antigen or mitogen. Proc. Natl Acad. Sci. USA 77, 1602–1606 (1980).

446


42. Jenkins, M. K., Pardoll, D. M., Mizuguchi, J., Quill, H. & Schwartz, R. H. T-cell unresponsiveness in vivo and in vitro: fine specificity of induction and molecular characterisation of the unresponsive state. Immunol. Rev. 95, 113–135 (1987). 43. Schwartz, R. H. In Immunology: the Making of a Modern Science (eds Gallagher, R. B., Gilder, J., Nossal, G. J. V. & Salvatore, G.) 95–101 (Academic, London, 1995). 44. Janeway, C. A. Approaching the asymptote? Evolution and revolution in immunology. Cold Spring Harb. Symp. Quant. Biol. 54, 1–13 (1989). 45. Janeway, C. A. The immune system evolved to discriminate infectious nonself from noninfectious self. Immunol. Today 13, 11–16 (1992). 46. Medzhitov, R., Preton-Hurlburt, P. & Janeway, C. A. A human homologue of the Drosophila Toll protein signals activation of adaptive immunity. Nature 388, 394–397 (1997). 47. Yang, R. B. et al. Toll-like receptor-2 mediates lipopolysaccharide-induced cellular signalling. Nature 395, 284–288 (1998). 48. Yoshimura, A. et al. Cutting edge: recognition of Gram-positive bacterial cell wall components by the innate immune system occurs via Toll-like receptor 2. J. Immunol. 163, 1–5 (1999). 49. Hayashi, F. et al. The innate immune response to bacterial flagellin is mediated by Toll-like receptor 5. Nature 410, 1099–1102 (2001). 50. Janeway, C. A. Presidential address to the American Association of Immunologists. J. Immunol. 161, 539–544 (1998). 51. Rathmell, J. C., Townsend, S. E., Xu, J. C., Flavell, R. A. & Goodnow, C. C. Expansion or elimination of B cells in vivo: dual roles for CD40- and Fas (CD95)-ligands modulated by the B-cell antigen receptor. Cell 87, 319–329 (1996). 52. Waldmann, H., Qin, S. & Cobbold, S. Monoclonal antibodies as agents to reinduce tolerance in immunity. J. Autoimmun. 5, 93–102 (1992).

53. Gett, A. V. & Hodgkin, P. D. A cellular calculus for signal integration by T cells. Nature Immunol. 1, 239–244 (2000).

Acknowledgements We are grateful to the following people for their recollections, suggestions, input and advice: A. Basten, J. F. A. P. Miller, K. Lafferty (deceased), C. Simeonovic, B. Fazekas and C. Jolly. A.G.B. is the recipient of an interim fellowship from the Australian National Health and Medical Research Council (NHMRC). P.D.H. is the recipient of a senior fellowship from the NHMRC. We dedicate this article to the memory of Kevin Lafferty. His remarkable contributions to the study of T-cell activation and alloreactivity will be long remembered, and his forceful advocacy for reason in scientific endeavour will be sadly missed.

Online links DATABASES The following terms in this article are linked online to: Entrez: http://www.ncbi.nlm.nih.gov/Entrez/ Bordetella pertussis | Mycobacterium tuberculosis LocusLink: http://www.ncbi.nlm.nih.gov/LocusLink/ CD3 | CD40 | CD40L | CD80 | CD86 | IL-1 | IL-2 | IL-4 | IL-6 | IL-8 | NF-κB | Toll | Toll-like receptors FURTHER INFORMATION Bright Sparcs Biography — Frank Macfarlane Burnet: http://www.asap.unimelb.edu.au/bsparcs/biogs/P000279b.htm NOBEL e-MUSEUM — Karl Landsteiner: http://www.nobel.se/medicine/laureates/1930/landsteinerbio.html NOBEL e-MUSEUM — Peter C. Doherty and Rolf M. Zinkernagel (Nobel Prize in Physiology or Medicine 1996): http://www.nobel.se/medicine/laureates/1996/ Access to this interactive links box is free online.

OPINION

The future of antigen-specific immunotherapy of allergy Rudolf Valenta More than 25% of the population in industrialized countries suffers from immunoglobulin-E-mediated allergies. The antigen-specific immunotherapy that is in use at present involves the administration of allergen extracts to patients with the aim to cure allergic symptoms. However, the risk of therapy-induced side effects limits its broad application. Recent work indicates that the epitope complexity of natural allergen extracts can be recreated using recombinant allergens, and hypoallergenic derivatives of these can be engineered to increase treatment safety. It is proposed that these modified molecules will improve the current practice of specific immunotherapy and form a basis for prophylactic vaccination.

Ninety years ago, long before the immunopathological mechanisms that underlie type I

(immunoglobulin-E-dependent) allergy were understood, Leonard Noon immunized patients suffering from pollen-induced hayfever with subcutaneous injections of pollen extracts1. Although this was based on the erroneous belief that seasonal hayfever might be caused by a grass-pollen toxin, successful outcomes were recorded and induced protection was found to last for at least one year after the treatment was discontinued. Since then, many of the underlying immunological and molecular mechanisms, as well as environmental factors, that influence the development of allergy have been analysed2. Type I allergy is a classical IgEmediated disease, which, as shown by a landmark experiment by Prausnitz and Küstner in 1921, requires at least three components: a disease-eliciting antigen (allergen); a transferable serum factor that discriminates allergic patients from healthy individuals (IgE);
