The following papers provide a mechanism, operating at the protein level, by which an organism might distinguish self from not-self intracellularly, thus leading to the presentation of peptides in association with MHC proteins, so recording a not-self presence in the cell. 

     The intracellular alarm signal which triggers such display would be the presence of a virus protein, or a mutated "self" protein. These would form homoaggregates (and also heteroaggregates with unmutated "self" proteins) which would then trigger protein degradation and peptide formation for MHC. Double-stranded RNA is also implicated. It is not until (and if) this alarm rings that an intracellular protein is labelled as "not-self" and peptides presented. Thus, until this moment, the body distinguishes mainly extracellular self from not-self. 

    This idea has become of increasing interest to cellular immunologists (see Matzinger's "danger" hypothesis 1994. Annu. Rev. Immunol. 12, 991-1045). But a word of warning to researchers. Although you may find the ideas appealing, be cautioned. If you mention them in your next grant application the reviewers may not share your enthusiasm (see web page on Peer Review; Click here).

 Hint: the first time through you might find it easier to go to the 1995 paper, or read "A Digression on Prions". (Click Here)

 In the early 1980s, prior to the discovery of antigen presentation to T cells as MHC-peptide complexes, it was difficult to link the various phenomena associated with heat shock proteins to immunity. The ideas outlined below were in place by 1985 as the MHC-peptide story began to unfold (Townsend et al. 1985; Babbitt et al. 1985), and a more satisfying (though not necessarily correct) linkage began to emerge (Forsdyke, 1991; Click Here) 

When cells of a variety of species are subjected to certain metabolic stresses, such as an increase of temperature, there is a decrease in normal protein synthesis and the induction of a set of several specific proteins which accumulate intracellularly (Schlesinger, Ashburner & Tissieres, 1982). The adaptive value of this "heat shock" response is unknown. 

      In a previous communication it was suggested that a set of extracellular proteins which accumulate in certain clinical conditions, act to cause the aggregation of extracellular pathogens in regions of stagnant blood flow (e.g. the sinusoids of the spleen; Forsdyke & Ford, 1983a). By analogy, it was suggested that heat shock proteins might cause a similar aggregation of intracellular pathogens.

    There have been reports that heat shock proteins are indeed induced when cells are infected with viruses (Nevins, 1982; Collins & Hightower, 1982; Khandjian & Turler, 1983; LaThangue et al., 1984). I here outline more fully the hypothesis that heat shock proteins create conditions which promote the aggregation ("phase separation") of particles of an intracellular pathogen when particle concentration exceeds a critical value.

      To eliminate a pathogen, the pathogen must first be recognized as different from the host. Pathogens differ from their hosts in one or more of the following respects.

• (1) Pathogens have a different genotype which may be expressed as antigenic determinants at the surface of the pathogen.
  
  
• (2) Pathogens are at a concentration within the host which is likely to be zero during the early stages of host development. The host's own tissue components (self) are likely to be constantly present at a concentration higher than zero.
  
  
• (3) As replication progresses, pathogens are likely to attain a higher local concentration  within the host than the host's own tissue components.

A mechanism by which a host could exploit the first two of these differences to defend itself against a pathogen has been discussed elsewhere (Forsdyke, 1969, 1975). I here draw attention to a mechanism by which a host could exploit the third difference.

        When molecules in solution reach a critical concentration, it becomes energetically more favourable for them to aggregate (form networks, crystallize, precipitate) than to remain in simple solution. When certain lipids do this ordered structures (membranes) result: the process is referred to as a lyotropic (concentration-dependent) phase transition.

        Without any obvious change in solution appearance, some high molecular weight polymers (e.g. Ficoll, polyvinylpyrrolidone, dextran sulphate, heated albumin) form networks in solution at high concentrations (Ogston & Woods, 1954; Comper & Laurent, 1978; Wahl, Stern & Stark, 1979; Minton, 1981; Fulton, 1982). This becomes apparent when other polymers or particles (e.g. erythrocytes) are added. The added polymers or particles are excluded from the phase of the original polymer molecules ("excluded volume effect") and become locally concentrated. This increases the chemical activity of the added polymers or particles, which may aggregate and form ordered structures. 

• In the case of an added nucleic acid polymer, specific hybridization between complementary strands is accelerated (Denhardt, 1966; Wetmur, 1976). 
  
  
• In the case of disc-shaped erythrocytes, species-specific "piles of coins" or "rouleaux" are formed (Forsdyke, Palfree & Takeda, 1982; Forsdyke & Ford, 1983a,b). In the absence of the excluded volume effect, nucleic acid hybridization and rouleaux formation would be considerably delayed or would not occur.

    In this light, I propose that the production of heat-shock proteins following infection with an intracellular pathogen (Cosgrove & Brown, 1983) would create intracellular conditions which would promote the phase-separation and aggregation of particles of the pathogen. This aggregation would occur at a far lower concentration of the particles than the concentration required for aggregation in the absence of the heat-shock proteins. The hypothesis predicts that some heat shock proteins should form networks which would exclude other polymers and particles from the solution phase.

Work in the author's laboratory was supported by the Physicians' Services Incorporated Foundation, the Leukaemia Research Fund of Toronto, and the Canadian Medical Research Council.

REFERENCES

Collins, P. L. & Hightower, L. E. (1982) J. Virol. 44, 703.
Comper, W. D. & Laurent, T. C. (1978) Physiol. Rev. 58, 253.
Cosgrove, J. W. & Brown, I. R. (1983) Proc. Natn. Acad. Sci. USA 80, 569.
Denhardt, D. T. (1966) Biochem. Biophys. Res. Comm. 23, 641.
Forsdyke, D. R. (1969) J. Theor. Biol. 25, 173.
Forsdyke, D. R. (1975) J. Theor. Biol. 52, 187.
Forsdyke, D. R. & Ford, P. M. (1983a) J. Theor. Biol. 103, 467.
Forsdyke, D. R. & Ford, P. M. (1983b) Biochem. J. 214, 257.
Forsdyke, D. R., Palfree, R. G. E. & Takeda, A. (1982) Can. J. Biochem. 60, 705.
Fulton, A. B. (1982) Cell 30, 345.
Kandjian, E. W. & Turler, H. (1983) Mol. Cell. Biol. 3, 1.
LaThangue, N. B., Schriver,K., Dawson, C. & Chan, W. L. (1984). EMBOJ 3, 267.
Minton, A. P. (1981) Biopolymers 20, 2093.
Nevins, J. R. (1982) Cell 29, 913.
Ogston, A. G. & Woods, E. F. (1954) Trans. Faraday Soc. 50, 635.
Schlesinger, M. J., Ashburner, M. & Tissieres, A. (1982). Cold Spring Harbor Symp. Heat Shock from Bacteria to Man. New York: Cold Spring Harbor Laboratory.
Wahl, G. M., Stern, M. & Stark, G. R. (1979) Proc. Natl. Acad. Sci. USA 76, 3683.
Wetmur, J. G. (1976). Ann. Rev. Biophys. Bioeng. 5, 337.

Abstract. There is unwarranted satisfaction with the view that MHC polymorphism evolved because there was a selective advantage in having a variety of MHC proteins to bind a variety of peptide subsets for presentation to T cells. While this may, in part, explain its maintainance, polymorphism may have evolved initially to reject foreign virus "grafts".

        The possession of similar membranes promotes aggregation between "like" cells, but it also promotes aggregation between the cells and viruses which retain membrane components of their previous host. The selection pressure afforded by hostile virus "grafts" would favour cells which developed polymorphic membrane components (since "like" will not aggregate with "not-like").

        This polymorphism would have evolved before the appearance of multicellular organisms. Thus, the evolution of modern immune systems would have been imposed upon pre-existing polymorphic systems. A path this evolution may have taken involves the development of mechanisms for intracellular distinction between self and not-self.

1. Introduction

The hypotheses that the education of potential immunologically competent cells requires both negative selection (Burnet, 1959) and positive selection (Forsdyke, 1975) have found considerable experimental support (reviewed in Schwartz, 1989). However, the relationship of these processes to the phenomenon of MHC polymorphism is poorly understood. It is argued that, although histocompatibility antigens were first recognized through tissue grafting experiments, such grafting does not normally occur in nature. Thus, MHC polymorphism could not have evolved in response to a selection pressure generated by the need to reject foreign grafts. So we must look elsewhere for an explanation for the origin of MHC polymorphism. This view is widely disseminated in textbooks and popular science journals (Marrack & Kappler, 1986; Hopkins, 1987; Grey et al, 1989; Raff, 1989).

    The discovery of MHC-restricted immune responsiveness (Zinkernagel & Doherty, 1974) and the role of MHC molecules in antigen presentation (Grey et al, 1989) provided a convenient explanation for MHC polymorphism. Extensive MHC polymorphism means that most individuals are heterozygotes. Since all potentially immunogenic peptides cannot bind to one MHC gene product (Grey et al, 1989), having two MHC gene products doubles the chance that an individual will be able to assemble a functional MHC-peptide complex ("heterozygote advantage"; "overdominant selection"; Hughes & Nei, 1989). 

    However, this factor-of-two advantage does not seem particularly compelling. A doubling of the number of MHC loci would achieve the same result. Those of the "group selectionist" school would argue for a much greater advantage at the population level. Thus, a group which had MHC polymorphism would be likely to have at least some members who would have a MHC protein able to form a functional complex with a peptide from a virulent pathogen. However, the general validity of group selectionist arguments has been questioned (Dawkins, 1976).

2. Viruses are Grafts

    There would be no need to look to explanations such as the above if graft transplantation were an "experiment of nature". Actually, such experiments have probably been going on since life first evolved. When a virus passes from one host cell to another it may transfer some of the cell membrane components of its former host. These may include MHC proteins (Hecht & Summers, 1976; Bubbers & Lilly, 1977; Gelderblom et al., 1987). Such a virus is, in essence, a graft. Individuals highly polymorphic for a membrane component transferred by the virus are likely to reject the virus "graft" as foreign, just as they would a cellular allograft.

    This rejection response could have evolved before the evolution of the immune response as we know it today. The membranes of the "survival machines" within which the early "replicators" (RNA, DNA) encased themselves (Dawkins, 1976), are initially likely to have been quite similar to each other. We and others have shown that, in an appropriate environment, it can be thermodynamically favourable for cells with similar surface membranes to aggregate preferentially, "like with like" (Forsdyke & Ford, 1983a,b; Armstrong, 1989). Similarly, a nucleic acid fragment (selfish gene/virus) could readily transfer by budding from its cell of origin and, enveloped in the cell membrane of its former host, aggregating with a "like" target cell (Fig. 1a).

FIG. 1a. Early evolution of virus-cell interactions. (a) A virus enveloped in the membrane of its former host is able to bind to a "like" target cell possessing similar membrane characteristics. Host and virus genomes are represented by wavy lines. The rectangle in the host genome represents a gene encoding a minor host surface component. The latter is represented by open red boxes embedded in the membrane with "caps" pointing to the exterior. 

    In this circumstance, a cell which had evolved prototypic MHC proteins with some degree of polymorphism, would be better prepared to maintain the integrity of its own nucleic acid by preventing invasion by a foreign nucleic acid (since "like" does not aggregate with "not-like"). MHC proteins could have evolved as a defence against viruses before the evolution of multicellular organisms. The evolutionary pressure to generate extensive polymorphism would have been high, since the greater the polymorphism, the less likely would be a virus to encounter a "like" cell.

3. Multicellularity Provokes Acceleration of the Arms Race

   However, multicellular organisms, each cell of which would bear "like" polymorphic antigens, would provide a fertile soil for viruses to spread between cells. Thus, the evolution of the multicellular state should be accompanied by a strong pressure for the evolution of immune systems. This evolution would be imposed upon, and might incorporate and adapt, a pre-existing polymorphic MHC system.

FIG 1b. Polymorphism of membrane components (represented on the left by closed black boxes and on the right by closed black boxes with round caps) prevents viral attachment by the low affinity like-with-like mechanism.

   As long as they depended upon the like-with-like mechanism to associate with their target cell membrane, virus envelopes would have been indistinguishable from self. However, at an early stage of their evolution, balked by the development of MHC-polymorphism, viruses would have developed specific coat proteins capable of recognizing non-polymorphic host cell surface proteins as receptors for virus attachment. Steps in developing this capacity might have included:

• (i) incorporation of a copy of the host gene encoding the non-polymorphic protein into the virus genome, and

• (ii) mutation of the latter to increase the affinity of binding of the gene product with the extracellular domain of the host-cell surface determinant (Fig. 1c,d).

FIG 1d. Mutation of the viral copy of the host gene (shaded rectangle) to generate a surface ligand with high affinity for the target cell receptor.

Thus, the pre-existing like-with-like affinity of host molecules presenting non-polymorphic surface determinants would have been enhanced by mutations of the viral copy of the same gene. The virus coat protein ligand and its host cell receptor would have evolved from a common source, a prototypic non-polymorphic host cell receptor. In entering an "arms race" with its host (Dawkins, 1976), the virus would have "shown its colours" (i.e. acquired a virus-specific surface protein), thus paving the way for specific recognition by a prototypic host immune system.

     Initially this might have taken the form of the release of soluble decoy receptors from target cells (Ehrlich, 1900). With tissue differentiation, certain cells would specialize in generating decoy receptors. A local tendency to hypermutate would generate a decoy receptor (prototypic antibody) repertoire which might protect organisms against a range of viruses. This would require the coevolution of extracellular mechanisms for distinguishing self from not-self, as discussed previously (Forsdyke, 1968, 1969, 1975). In such a frame-work the immune system as we know it today could have evolved.

4. Simplicity Precedes Complexity. Positive Selection

    It is possible that one prototypic antibody locus (with multiple V region genes) would have had the opportunity to develop evolutionarily before duplication to generate other loci. A simple model has been considered elsewhere, perhaps corresponding to an organism which once existed (or may even still exist). The model organism had a single hypervariable locus and could have mounted both cell-mediated and humoral immune responses (Forsdyke, 1968, 1969, 1975). 

    The model proposed that enhanced immunological reactivity against "near-self " MHC antigens ("alloaggression") is not encoded in the germ line as suggested by Jerne (1971), but requires a phase of positive selection of the immunological repertoire (Forsdyke, 1975). MHC proteins and other membrane components (Bluestone et al, 1988; Porcelli et al, 1989; Vidovik et al, 1989) would play a major role in the shaping of this repertoire. It was also pointed out that expanded clones of anti-near-self immunologically competent cells would present a barrier opposing the progressive mutation of molecules of a pathogen towards forms indistinguishable from host molecules ("molecular mimicry").

5. Intracellular Distinction between Self and Not-self

It is argued (Bevan, 1989) that:

"there are no obvious cues to distinguish self from foreign proteins inside the cell".

Recently it was stated (Kourilsky & Claverie, 1989) that the:

"finding ... that the transfected gene encoding influenza nucleoprotein yields MHC class-l-associated peptides on the cell surface has suggested that the transfected cell, not being an immune system, cannot know that the influenza gene is foreign".

Thus, it is envisaged (Bevan, 1989) that:

"foreign proteins have to compete with more than 10,000 species of self proteins for presentation at the cell surface" .

However, absence of evidence is not evidence of absence. Absence of a known mechanism for intracellular self/non-self distinction does not mean that such a mechanism does not exist. Indeed, the principle of intracellular self/non-self distinction was firmly established with the discovery of bacterial DNA restriction- modification systems (Arber & Linn, 1969). 

    A mechanism for distinction at the protein level, based on concentration-dependent (lyotropic) phase transitions, has been proposed (Forsdyke, 1985). Proteins which exceed their solubility limits in the crowded intracellular environment will aggregate and mark themselves as different from other proteins (Marston, 1986). The synthesis and breakdown of viral proteins would not be so tightly regulated as in the case of host cell proteins. After all, the raison d'tre of a virus is to increase in number. Changes in environmental conditions (e.g. heat shock), which might favour the aggregation of self proteins, would be dealt with by special mechanisms, such as the activation of chaperonin-like proteins which would reverse aggregation (Fischer & Schmid, 1990).

    Having been marked as different, foreign aggregates would be degraded, first to peptides and then to amino acids. This level of coping with foreign material could be achieved by unicellular organisms. Association of foreign peptide fragments with prototypic polymorphic MHC proteins en route to the cell surface would not seem advantageous at the unicellular stage of evolution. However, in multicellular organisms, the modified, "near-self ", MHC antigenic determinants so created would be recognized by host immunologically competent cells which would already have been educated, by mechanisms such as outlined previously (Forsdyke, 1975), to respond against near-self.

    By associating peptide fragments with MHC determinants, the organism would merely be taking advantage of the pre-existence of a large population of immunologically competent cells reactive with near-self. A cell displaying modified MHC proteins would have become labelled for destruction by immunologically competent cells or their products, a feature of considerable selective advantage for the organism. The pre-existing polymorphism of MHC proteins would increase the chance that a successful MHC-peptide complex would be formed and, in the continuing evolutionary arms race with pathogens, modifications of MHC proteins might have occurred to enhance their recognition of sets of structural features shared by many peptides.

    Since the critical self-vs.-non-self decision had already been made intracellularly, it would be sufficient for the peptide-MHC protein complex to be recognizable as near-self. This might involve recognition by T cell receptors both of conformational changes in MHC molecules and of features of the associated peptide. There would be no necessity for a stage in the education of immunologically competent cells in which cells reactive with complexes of MHC proteins and self-peptides derived from the organism's intracellular proteins would be eliminated (Berg et al. 1989). This would then leave open to the organism the option of destroying cells in which the regulation of the intracellular concentration of one of its own proteins had become disordered.

6. Positive Selection in Different Environments. Cell Lineages

   Duplication of hypervariable loci and segregation of their expression to different tissue compartments would provide opportunity for exposure of different populations of potential immunologically competent cells to different spectra of antigens during the phase of positive selection. Thus, cell lineages (e.g. prototypic T and B lineages) might have arisen. This would provide the opportunity for the evolution of a division of labour. One lineage would be concerned primarily with cell-mediated responses and another lineage with humoral responses.[For further discussion of lineages, please see Forsdyke (1992) J. theor. Biol. 154, 109-118.]

From a consideration of the evolution of the first immune systems in simple organisms, it is proposed that mechanisms for self/not-self discrimination evolved at an early stage of evolution in prototypic B cells, prior to separation of the B cell and T cell lineages. After passing an MHC protein-limited self/not-self discrimination gate, T cells depend for fine self/not-self discrimination either on B cells (extracellular discrimination) or on cells harbouring pathogens (intracellular discrimination). Full activation of B cells needs interaction with T cells, but this evolved late and is not part of the self/not-self decision.

1. Introduction

Immunologically competent cells (ICCs) are lymphocytes of either the B or T cell lineage. Burnet (1959) and Lederberg (1959) originally proposed that generation of ICCs involves an early time-window during which cells bearing antibody-like receptors are destroyed on encountering specific self antigenic determinants which react with these receptors. Cells surviving this encounter pass through the window and respond immunologically if they then encounter specific not-self antigenic determinants. Thus cells receive one signal from antigen, but respond differentially depending on their physiological state.

    These ideas were subsequently elaborated to accommodate the ability of high antigen concentrations to inhibit the immune response in adults (Mitchison, 1964), and the possibility that cells which have passed the window might mutate to react to self antigenic determinants. Extending earlier work of Medawar (1963) and Nossal (1965), I proposed that antigenic determinants are capable of giving either of two signals to specific lymphocytes at any stage of their development [Forsdyke, 1966, 1968, 1969; Fig. l(a)].

FIG. 1. Two signal models for lymphocyte activation and inactivation. Elements shown are cells with antibody-like receptors, free antibody and antigen. The latter may either be self (S) or not-self (NS). 

• In model (a) (Forsdyke, 1966, 1968, 1969), natural antibody secreted by a prototypic B cell buffers the cell against excess not-self antigen. Only one receptor (or a limited number of widely dispersed receptors) is occupied (Signal 1; activation). In contrast, a newborn prototypic B cell has just displayed receptors at the cell surface, but has not yet initiated the secretion of buffering antibody. At this unique time-point self antigens, present even at low concentrations, tend to saturate receptors and deliver signal 2 (inactivation). 

•  Model (b) (Bretcher & Cohn, 1968, 1970) places no special emphasis on antigen concentration, but does require that the antigen possess 1 antigenic determinant. Reaction with self antigen delivers an inactivation signal. Activation requires "associative recognition" of not-self antigen both by receptors on the lymphocyte and by another receptor-bearing activity which delivers a positive signal (+).

The effective concentration of an antigenic determinant depends, not so much on its absolute concentration within the body, but on the ratio of the concentration of antigen to the concentration of available antigen-binding activity. In an unimmunized organism the latter is largely provided by natural antibody. At an early stage of immunological maturation there is little specific extracellular antigen-binding activity. Unbuffered by this activity, a specific lymphocyte encounters a high effective antigen concentration and is destroyed (signal 2). 

    Later the concentration of specific antibody increases so that effective antigen concentrations tend to be lower. The signal given to a specific lymphocyte then induces an immune response (signal 1). Thus, the outcome of reaction of a lymphocyte with antigenic determinants is critically dependent on a second antigen-recognizing activity, natural antibody, the concentration of which is time-dependent [Fig. l(a)]. The model is "two signal" in another sense. Signal I requires limited receptor occupancy (minimum one signal), whereas signal 2 requires reaction with a second receptor in close proximity to the first (two signals).

    An alternative formulation, closer to earlier ideas (Medawar, 1963; Nossal, 1965) was offered by Bretcher & Cohn (1968, 1970; reviewed by Cohn in Langman, 1989). They proposed that the second antigen-recognizing activity actually delivers a positive signal to a specific lymphocyte [Fig. 1(b)]. This "helper" activity, initially envisioned as free antibody and later as antibody synthesized and borne by another lymphocyte, is essential for an immune response. In the absence of the signal from the second antigen-recognizing activity there is cell destruction. Focusing of the latter activity to a lymphocyte requires that an antigen present at least two antigenic determinants.

     Under this model inactivation occurs slowly and is by default (failure to receive a help signal). Under my two signal model inactivation is an active process, which in some circumstances may involve complement (for references see Forsdyke, 1975, 1977, 1980). Cohn and coworkers have recently argued that their model is consistent with modern concepts in immunology (Langman, 1989; Cohn & Langman, 1990a, b, c). Here a similar case is made for my model. Some aspects of this work are discussed more fully elsewhere (Forsdyke, 1985, 1991).

2. Some Principles

     The result of a long evolutionary process, modern vertebrate immune systems appear very complex (Burnet, 1978; Marchalonis & Schluter, 1989). However, all immune systems have basic features which would have evolved at an early stage. It seems likely that organisms with one set of antibody variable (V) region genes would have evolved and have achieved some degree of immunological sophistication before being superseded by organisms with two sets of V region genes. 

    A prototypic immune system with one set of V region genes might have displayed all basic system features, such as discrimination between self and not-self and the ability to mount both cellmediated and humoral immune responses (Forsdyke, 1969). Duplication of the set of V-region genes would have produced an initial redundancy followed by a division of labour so that one set of V-regions would have become specialized for cell-mediated responses and another for humoral responses. 

    There would have been an interplay between the cell-mediated and humoral response systems, so that each would have evolved in the context of the other. Similarly there would have been an evolutionary interplay with other coevolving systems, such as those concerned with metabolic regulation and with maintaining cell population-size homeostasis (Rocha et al., 1989).

     Along similar lines, it seems likely that one MHC protein system would have evolved before a second (Lawlor et al., 1990). The second system would have evolved in an organism already possessing the first system. There would have been an evolutionary interplay between the two systems.

3. Prototypic B Cells First

    Prototypic ICCs are held to have properties of both B cells and T cells. They are generated in a "mutant breeding organ" and undergo a subsequent "education" such that high specificity anti-self cells are destroyed (negative selection) and low specificity anti-self cells accumulate (positive selection; Forsdyke, 1975). 

    Duplication of the prototype V-region set would facilitate the development of independent systems of B cells and T cells. The steps required to develop a mature B cell would appear less complex than those required to develop a mature T cell. Whereas progenitors of both B cells and T cells arise in the bone marrow, T cell progenitors must migrate to the thymus to become ICCs. Thus, it is postulated that a prototypic mutant breeding organ was bone marrow-like and, in this respect, prototypic ICCs were more like B than T cells.

    In the bone marrow positive selection could be influenced by a wide range of self antigens and hence might not be well-developed for any particular antigen. However, sequestration of potential ICCs in a compartment where they could be selectively exposed only to a narrow range of self antigens would result in the positive selection of a restricted set of low specificity anti-self cells. The selective advantage of accumulating ICCs of low specificity for self MHC antigens has been discussed previously (Forsdyke, 1975, 1991). Thus, the evolution of a compartment where potential ICCs could be exposed selectively to self MHC antigens would be favoured. The result of this process would be the thymus of modern vertebrates.

    The process would involve evolution of a germ line V region gene set with an enhanced probability of efficiently generating variable regions capable of low specificity reaction with self MHC. Duplication of the prototypic V region set would have released a set of V region genes from prototypic B cell constraints (marrow-based) and facilitated the evolution of this set of V region genes dedicated to prototypic T cells (thymus-based).

4. Education of Prototypic B Cells

    After association of a particular prototypic variable region gene with a prototypic constant region gene, the corresponding IgM-like immunoglobulin would be assembled in prototypic B cells and first displayed as a receptor at the cell surface. At this unique time-point and in the absence of the corresponding free antibody, the effective concentration of self antigenic determinants (even if present at low absolute concentrations) would be high. By mechanisms such as previously outlined (Forsdyke, 1966, 1968, 1969), reaction of a potential ICC with high concentrations of such determinants would result in cell destruction. Failing this, and following positive selection by a wide range of self antigenic determinants (Forsdyke, 1975), immunoglobulin would be secreted generating a pool of natural antibody capable of buffering the cell against future reaction with excess antigen. In these circumstances, the effective concentration of most foreign antigens would be low and reaction with the cell would trigger enhanced secretion of the appropriate antibody.

5. Education of Prototypic T Cells

    Within the thymic compartment, prototypic T cells would be separated from the antibody secreted by prototypic B cells. The T cells would be exposed to a limited spectrum of self-antigens, primarily MHC antigens. Cells bearing receptors of high specificity for these antigens would be destroyed (negative selection). Cells with receptors of low specificity would accumulate (positive selection). Cells with very low to zero specificity would degenerate (Forsdyke, 1975; Bevan, 1977; Blackman et al., 1990; Boehmer & Kisielow, 1990; Rothenberg.. 1990). Moving through different body compartments (e.g. blood, lymph, lymph-nodes) the selected T cells would be subjected to various degrees of buffering against foreign antigens by antibodies secreted by prototypic B cells.

6. T Cells Permissive for B Cells

    Let us consider the immune response to a virus of an evolutionarily simple organism with one prototypic MHC locus and two sets of variable region genes (one for prototypic B cells and one for prototypic T cells; Fig. 2).

    Various elements shown are a virus antigen (V; coloured red), a prototypic B cell (B), a prototypic T cell (T), a phagocytic cell (M), a host cell (H; bearing a receptor for virus), and free antibody.

      Dashed lines indicate clonal expansion of a particular cell type in response to an appropriate signal. A prototypic B cell (top left), passes the self/not-self discrimination gate (S/NS) and begins synthesis of natural antibody. A foreign virus has three fates:

• (1) It reacts with the B cell triggering the cell to respond immunologically (i.e. clonal expansion to generate memory and enhanced synthesis of antibody). This can only occur, however, if a permissive signal (+) is received from a T cell (likely to be a daughter T cell; see below). The latter recognises complexes of viral peptides with prototypic MHC protein at the B cell surface.

• (2) The virus is processed by a phagocytic cell (M) which also displays complexes of viral peptides with prototypic MHC protein at its surface. Prototypic T cells (right), having passed an MHC protein-limited self/not-self discrimination gate (S/NS), react with the MHC protein peptide complex and are triggered to expand clonally generating daughter T cells.

• (3) The virus enters its target host cell (H) where it may replicate. Viral proteins are recognized as not-self by an intracellular discrimination process (S/NS) and their peptides are displayed at the cell surface in association with MHC protein. Daughter T cells encountering host cells displaying a specific peptide deliver a negative (cytotoxic) signal (-).

Some viruses react with natural antibody and thus are labelled as not-self. Appropriately opsonized, the viruses attach to and are ingested by phagocytic cells. In either a free or attached state, native virus protein epitopes also react with specific B cells. This satisfies one condition necessary for the B cells to respond by proliferation and enhanced secretion of antibody.

     Meanwhile, some viruses, having evaded capture by antibody, penetrate their host cell targets and viral proteins are synthesized. By mechanisms such as discussed elsewhere (Forsdyke, 1985, 1991), the viral proteins are recognized as foreign intracellularly. They are then degraded and peptide fragments displayed at the host cell surface in association with MHC protein (Townsend & Bodmer, 1989). This complex acts as a recognition site stimulating prototypic T cells with specific receptors to proliferate and/or destroy the host cell.

    Within phagocytic cells viral proteins follow a degradation pathway similar to that within host target cells, except that entry into the pathway does not depend upon intracellular self/not-self discrimination. The peptides eventually appear at the cell surface in association with MHC protein (Fig. 2). The MHC protein-peptide complex could, in principle, be recognized by either prototypic B cells or prototypic T cells bearing receptors of the appropriate specificity. However, because of thymic positive selection there would be more potentially reactive T cells. Furthermore, a T cell specific prototypic CD4/CD8-like molecule would evolve to enhance interaction of the T-cell receptors with the MHC protein-peptide complex.

     Prototypic B cells, like host target cells and phagocytic cells, incorporate and process viruses and display peptides at their surfaces complexed with MHC protein (Fig. 2). Complexes are recognized by prototypic T cells which are activated to further proliferate and/or deliver, not a cytotoxic signal, but a specific "help" signal, possibly involving a short-lived cytokine. This satisifes a second and final condition for B cell activation, but is not related to the process of self/not-self discrimination.

     How did this activation of prototypic B cells (capable eventually of secreting small diffusible IgG-like molecules) become dependent on a reaction with prototypic T cells? There are conditions underwhich a humoral response can exclude a cell-mediated response (for references see Forsdyke, 1969). If it were more important to mount a T cell response than a B cell response, then T cell control of the B cell response could evolve. 

    For example, a virus-infected host cell might display native viral epitopes at the cell surface (for references see Forsdyke, 1991). Reaction of antibody with these epitopes might hinder concomitant reaction of T cells with MHC protein-peptide complexes borne by the same host cell. Indeed, to correct for the possibility of B cell activation in the absence of T cell help, there might evolve a mechanism for T cells in some circumstances actually to deliver a negative signal to activated B cells (Shinohara et al., 1988).

    A prototypic T cell activated by a peptide-MHC complex proliferates thus expanding clone size and creating daughter cells (immunological memory; Fig. 2). Any cell displaying the same MHC protein-peptide complex is a potential target of the daughter cells. Possessing both cytotoxic and helper capacity, these cells have to determine whether the target is a member of the phagocytic and B cell (lymphoreticular) lineages (requiring help) or is a cell hosting a pathogen (requiring destruction).

      The latter cells do not normally respond to the local help signal. Cells of lymphoreticular lineages respond to the local help signal and would have to evolve some mechanism to resist the cytotoxic signal. However, this would leave lymphoreticular cells vulnerable to intracellular pathogens, so there should be a selection pressure favouring the evolution of precise discrimination between help and cytotoxic signals.

7. T Cell Subsets

    Prototypic T cells would have both helper and cytotoxic functions. In the course of evolution duplication of appropriate genes would create an opportunity for T cell subsets to arise with an appropriate division of labour. Thus, T cells recognizing MHC protein peptide complexes on virus-infected host cells would be predominantly cytotoxic. T cells recognizing MHC protein-peptide complexes only on lymphoreticular cell lineages would specialize in helper functions. This division of labour would involve duplication of prototypic MHC loci and duplication of the T cell specific prototypic CD4/CD8-like locus.

    Thus, one set of T cells (prototypic CD4 bearing) would arise with specificity for peptides presented by products of one set of MHC genes (MHC class II prototypes). These cells would be primarily concerned with giving permissive signals to appropriate MHC class II protein-bearing cells.

    Another set of T cells (prototypic CD8 bearing) would arise with specificity for peptides presented by products of another set of MHC genes (MHC class I prototypes). These cells would be primarily concerned with the destruction of appropriate MHC class I protein-bearing cells (Fig. 3). Cells of the lymphoreticular lineage, bearing both classes of MHC molecule, would be potential targets of both types of T cells. The evolution of a dependence of T cells-on further signals (e.g. interleukin-1) will not be considered here.

FIG. 3. Response to an antigen in an evolutionarily more complex host organism in which two distinct prototypic T cell subsets have arisen (T_h,T_c), and the prototypic MHC locus and prototypic CD4/CD8-encoding locus have duplicated. Other elements shown are virus antigen (V), a prototypic B cell (B), a host cell target (H) and free antibody. The role of phagocytic cells is not shown. Dashed lines indicate cell activation to proliferate (clonal expansion). Full self/not-self discrimination (S/NS) operates extracellularly at the level of B cells and intracellularly at the level of host cells. MHC protein-limited self/notself discrimination occurs extracellularly at the level of T cells. The three fates of virus antigen are similar to those shown in Fig. 2 except that separate T cell lineages deliver the permissive signal for the B cell immune response (T_h) and the negative signal for host cell destruction (T_c).

FIG. 4. Education of T cells requires MHC protein-dependent and peptide-dependent gates. 

    Elements shown are T cells (T) bearing receptors of various specificities, thymic stromal cells (S) displaying MHC proteins, and peripheral tissue cells (M; each displaying a peptide from a processed antigen, shown as a filled bullet-like object).

• Gate 1. T cells with receptors of high specificity for antigenic determinants displayed by MHC protein alone are actively deleted. T cells with receptors of low specificity are selected and accumulate peripherally as mature T cells. T cells with receptors of very low to zero specificity degenerate.

•  Gate 2. Peripheral T cells react with MHC proteins complexed with a particular peptide. Only T cells with receptors of high specificity for the complex proliferate. The existence of this gate depends on the peptide, which, in turn, depends on prior "surrogate" self/not-self discrimination elsewhere.

     In essence, T cells pass through two discrimination gates (Fig. 4). The first gate is extracellular and involves cell surface components within the thymus. Cells passing this gate can survive, but proliferation is limited. Clonal expansion is permitted when the T cell receptor encounters an extra-thymic MHC protein-peptide complex. This latter gate can only exist if the antigen from which the peptide was generated has already been recognized as not-self either intracellularly in a cell which has subjected the antigen to processing in a cytosolic compartment (Townsend & Bodmer, 1989), or extracellularly by B cells or their products. 

    The cytosolic and B cell discrimination mechanisms can be regarded as surrogates for a more direct mechanism of ensuring that T cell receptors are of high specificity for their not-self targets. Thus, B cells and host cells "read" various aspects of an antigen (its concentration and specificity for their receptors) and relay this information to T cells. The latter discriminate primarily on the basis of specificity.

I thank R. E. Langman and J. J. Marchalonis for helpful reviews of the manuscript.

    Cytotoxic T cells recognize cell surface complexes of MHC class I proteins with peptides derived from proteins synthesized within the recognized cell. A mechanism permitting some intracellular discrimination between self proteins and not-self proteins (encoded by a foreign species) would allow the preferential loading of MHC proteins with peptides derived from not-self proteins. This would decrease competition with peptides derived from self proteins and decrease gaps in the T cell repertoire.

     A possible mechanism has been derived from studies of the specificity of self-aggregation of erythrocytes and of virus coat protein. It is postulated that genes whose products occupy a common cytosol have co-evolved such that product concentrations are fine-tuned to a maximum consistent with avoiding self-aggregation. Cytosolic proteins collectively generate a pressure tending to drive protein molecules into self-aggregates. Each individual protein species both contributes to, and is influenced by, this pressure. The aggregation involves a liberation of bound water and an increase in entropy.

     The concentrations of proteins encoded by viral genes (not-self) readily exceed the solubility limits imposed by the crowded host cytosol. This leads to their preferential degradation to peptides which associate with MHC proteins. The intracellular and extracellular self/not-self discrimination systems complement each other to ensure that there is no immunological reaction against normal self tissue components.

Keywords: Collective protein function, concentration fine-tuning, heat-shock response, evolution.

1. The Need for Intracellular Self/Not-Self Discrimination

     The maturation of immunologically competent cells involves the deletion or inactivation of cells which react with high specificity with self antigenic determinants. Since the cells first "show their colours" by displaying antigen-specific, outward-facing, receptors at their surfaces, their "education" is regarded as part of an extracellular process of self/not-self discrimination. A cell whose receptors happen to react with some free or cell-borne self-antigenic determinant will either be deleted or inactivated [15, 45].

    The principle of intracellular self/not-self discrimination was firmly established with the discovery of bacterial restriction/modification systems [1]. An advantage to higher multicellular organism of having an intracellular self/not-self discrimination system became particularly apparent with the finding that peptides from both self and foreign intracellular proteins can be displayed at the cell surface in association with major histocompatibility complex (MHC) proteins; here they are recognized by immunologically competent cells of the cytotoxic T cell class [43].

In the absence of a mechanism for distinguishing intracellularly between self peptides or proteins and foreign (e.g., virally-derived) peptides or proteins, cells would have to display all MHC-bound peptides. Hedrick has noted [26] that:

"It is hard to understand how peptides derived from foreign antigens can compete with the tide of self peptides..."

and has suggested that:

"Perhaps there is a mechanism that could help to sort peptides into those originating from self and those originating from foreign peptides".

    In the absence of such a sorting mechanism, an organism would depend solely on prior education of T cells to prevent their being activated by self-peptide-MHC complexes. Several studies indicate that this deletion of potential self-reacting T cells would create numerous gaps in the T cell repertoire, which might adversely affect resistance to infection [42,46,49]. 

    For example, Du Pasquier and coworkers [8] presented evidence that polyploid toad species, while apparently functionally polyploid for most genes, are functionally diploid for MHC genes. That this had occurred over evolutionary time was suggested by the observation that healthy laboratory-made polyploid toads expressed all MHC loci. Since the final T cell repertoire reflects the processes of negative and positive selection [12], every additional functional MHC gene would increase the number of potential anti-self T cells, which would have to be deleted (negative selection). This would decrease the range of foreign proteins to which the surviving positively selected cells could respond. Fitness would be impaired. Individuals who lost excess MHC loci through mutation, would then have a selective advantage so that the modern polyploid survivors would be MHC diploid. As Vidovic and Matzinger succinctly state [49]:

"There will be a selective pressure against the expression of too many MHC genes. Because the number of cells in the immune system is finite, the elimination of cells recognizing self obviously reduces the recognition repertoire of the individual."

    The prior sorting of peptides intracellularly would mean that populations of T cells would exist with the potential to respond against self-peptides should they be inadvertently, or artificially, presented. Thus, Schild and coworkers [46] found that cytotoxic T lymphocytes:

"are only tolerant of those endogenous self peptide sequences that are presented by MHC Class I-positive cells in a physiological manner" (my italics).

2. Distinction Between Specific and Collective Protein Functions

    So how could intracellular peptides be sorted? I here review a proposed mechanism of intracellular self/not-self discrimination derived from studies of the specificity of the aggregation of red blood cells [20-22], and tobacco mosaic virus coat protein [31-33]. Although still unproven, the mechanism appears to throw new light on a number of quite disparate phenomena, including fever, the heat shock response, temperature-sensitive mutations and even X chromosome dosage compensation [16-19]. Three basic postulates are that:

•  (i) Proteins have both specific functions characteristic of each individual protein, and collective functions as proteins per se (e.g., Donnan equilibrium; [28]).

•  (ii) Both functions provide a basis for evolutionary selection.

•  (iii) When sharing a common compartment, there may be incompatibilities between proteins from different species (e.g., a parasite species and its host species), based on a novel collective protein function.

3. Self-Aggregation of Red Blood Cells

    The discovery of this phenomenon has a fascinating early history. The ancient Greeks associated disease with an excess of white "phlegm", one of the four body "humours". To rid the body of phlegm, sick persons were bled. This practice endured for over two millennia.

    Fahraeus in 1921 [10-11], extending earlier observations of Hewson [27], made a convincing case for the origin of therapeutic bleeding. He noted that in ancient times people would have been bled into earthenware vessels. The clot would have contracted expressing yellow serum. On emptying a vessel the Greeks would first have poured off the serum and then have dislodged the clot. This would have been red in its oxygenated upper part and black in its anaerobic base. Thus, three of the humours would have gained their names, choleric, sanguine, melancholic (yellow, red, black). 

    However, in the case of blood from a sick person, an extra white layer (phlegm) would have been seen above the red part of the clot. 

    What the Greeks did not know was that in freshly drawn stationary blood from a sick person the red cells often aggregate to form rouleaux. These sediment rapidly before a clot can form, leaving a cell-free upper plasma layer. The eventual clot involves not only the red cell mass, but also the upper plasma layer. The clot contracts expressing yellow serum, as with blood from a normal subject, but the dislodged clot now shows three layers (Fig. 1).