J. theor. Biol. (1991) 151, 169-192
A Unified Model for T Cell Antigen Recognition and Thymic Selection of the T Cell Repertoire M A R K D. M A N N I E t
Department of Microbiology and Immunology, University of Michigan Medical School, Ann Arbor, MI 48109-0620, U.S.A. (Received on 30 October 1989, accepted on 28 January 1991) Positive selection of T cells during thymic differentiation predisposes mature T cells to recognize glycoproteins encoded by "self" alleles of the major histocompatibility complex (MHC) as "restricting elements" for antigen presentation. Yet, negative selection also occurs during thymic differentiation resulting in the clonal deletion of T cells reactive with "self" MHC glycoproteins. Thymic processes of positive and negative selection represent a paradox because the "altered self" view of T cell receptor (TcR) recognition indicates that the same type of TcR-MHC glycoprotein binding interaction mediates both positive and negative selection of the T cell repertoire. Most contemporary models of thymic selection attempt to explain these paradoxical observations by quantitative differences of affinity. That is, TcR interactions with MHC ligands that are of medium affinity lead to positive selection whereas those that are of high affinity lead to negative selection. The purpose of this manuscript is to provide an alternative model of thymic selection based on the efficacy (the ability of a ligand to catalyze receptor mediated biological activity) of T c R - M H C ligand interactions. The "'efficacy" model predicts that among those thymocytes exhibiting affinity for self MHC ligands, some clones bind self MHC ligands without efficacy whereas others bind these ligands with efficacy. Immature T cells that bind MHC ligands without efficacy do not undergo TcR mediated activation and thereby escape clonal deletion. Instead, these T cells compete for growth--promoting sites on thymic antigen presenting cells (APC) based upon their clonotypic TcR affinity for self MHC ligands. These T cells experience positive thymic selection and eventually dominate a repertoire of mature T cells predisposed to exhibit non-efficacious binding to "self" MHC ligands. In contrast, immature T cells that exhibit efficacious binding to self MHC ligands are deleted from the T cell repertoire during thymic maturation. By this mechanism, the mature T cell repertoire is selected so that clonotypic T cells are predisposed to bind the very sites on MHC glycoproteins responsible for antigen presentation without risk of autoimmunity. Given the clonotypic diversity of the mature repertoire, complexes of foreign peptides and self MHC glycoproteins would be recognized by clones of the appropriate specificity as highly efficacious Iigands. In summary, the "'efficacy" model is entirely consistent with the "altered self" concept of T cell antigen recognition and readily accounts for both positive and negative selection of the T cell repertoire. I. Introduction I m m u n o l o g i c a l discrimination o f self and n o n - s e l f entities is mediated by specific antigen receptors expressed on clones o f mature T l y m p h o c y t e s (reviewed by t Present address: Department of Microbiology and Immunology, East Carolina University School of Medicine, Greenville, NC 27858-4354, U.S.A. 169 0022-5193/91/140169+ 24 $03.00/0 O 1991 Academic Press Limited
M. D, M A N N 1 E
Marrack & Kappler, 1987). Unlike B cell antibody, T cell receptors (TcR) do not directly bind soluble antigen. Rather, T cells recognize only those antigens which have been bound by a separate class of proteins from the major histocompatibility complex (MHC) and expressed at the cell surface of antigen presenting cells (APC) (Babbitt et al., t985; Buus et al., 1987; Bjorkman et al., 1987a; Townsend et al., 1989). T cells exhibit a dual specificity because their TcR are specific for both foreign antigen and for "self" glycoproteins of the highly polymorphic MHC (Shevach & Rosenthal, 1973; Zinkernagel & Doherty, 1974). Thus, TcR are referred to as "'MHC restricted" because recognition of antigen is "restricted" in time and space to the co-recognition of self MHC glycoproteins. The "altered self" view of T cell antigen recognition predicts that clonotypic T cells become activated when their TcR binds a pre-existing complex of foreign antigen-MHC glycoprotein on APC to form a trimolecular complex (reviewed by Zinkernagel & Doherty, 1977; Schwartz, 1985). According to this hypothesis, foreign antigen binds MHC glycoproteins to form an extremely stable complex. Clonotypically specific TcR then bind complexes of MHC glycoprotein and antigen to trigger T cell activation. According to the most accepted view of this model, dual specificity of T cells derives from the structure of the TcR binding pocket which is designed to make high affinity contacts with residues of the antigen as well as with adjacent residues of the MHC glycoprotein. The sites on MHC glycoproteins responsible for binding TcR and antigen are comprised of highly polymorphic residues clustered on the top and inward surfaces of a large, outwardly facing "peptide-binding" groove (Bjorkman et al., 1987b; Brown et al., 1988). The present understanding of T cell antigen recognition forms the basis of hypotheses addressing a question central to immune system function. That is, what processes of thymic differentiation and selection are responsible for shaping the T cell repertoire so that mature T cells are able to so accurately distinguish self from non-self entities? During differentiation in the thymus, immature T cells rearrange a multitude of TcR gene segments to produce a large diversity of antigen receptors (>109 TcR) (reviewed by Davis, 1988). Each clone of this highly diverse repertoire expresses a unique TcR "variable" region which engenders a unique dual specificity for antigen and self MHC glycoprotein. During maturation in the thymus, this diverse population of cionotypic T cells undergoes both positive selection and negative selection to produce the immunocompetent repertoire of T cells. During positive selection, T cells able to recognize self MHC "restriction elements" are stimulated to differentiate into mature immunocompetent T cells (Bevan, 1977; Zinkernagel et al., 1978; Singer et al., 1982; Lo & Sprent, 1986, Marrack et al., 1988a; Marusic-Galesic et al., 1989; Teh et at., 1988; Sha et aL, 1988b; Berg et al., 1989; Benoist & Mathis, 1989]. Positive selection insures that the T cell repertoire is primarily comprised of clones able to recognize foreign antigens in the context of self MHC glycoproteins. During negative selection, T cells able to recognize self MHC glycoproteins complexed with "self" peptides (derived from the body in general) are deleted from the repertoire (Kappler et al., 1987a; 1988; MacDonald et al., 1988c; Kieslow et al., 1988; Pullen et al., 1988; Sha et al., 1988b; White et al., 1989). Negative selection of autoreactive TcR insures that the T cell
THE EFFICACY MODEL OF T CELL SELECTION
repertoire is purged of cells capable of inducing autoimmune disease (Glazier et al., 1983; Jenkins et al., 1988; Smith et al., 1989). 2. The Double Paradox of Positive Selection
A theoretical basis for positive selection of thymic T cells is enigmatic because positive selection is not easily reconciled with the "'altered self" model of T cell antigen recognition or with the process of negative selection of autoreactive T cells. For example, how can T cells be positively selected to recognize "restriction sites" on self MHC glycoproteins? According to the "altered self" model, recognition of self MHC glycoproteins cannot occur in the absence of foreign antigen. Furthermore, the MHC restriction sites that stimulate positive selection have not been distinguished theoretically from the MHC iigands that induce negative selection. Hence, how can the opposite processes of positive and negative selection be mediated by theoretically equivalent interactions of TcR with MHC ligands? The first paradox is based on the observation that positive selection of thymic T cells produces a mature T cell repertoire that preferentially recognizes self MHC glycoproteins during MHC-restricted antigen recognition. That is, (A x B) T cells that mature in a chimeric thymus of an "A" haplotype predominantly recognize antigens restricted by APC having "A" but not " B " haplotypes, and vice versa (Bevan, 1977; Zinkernagel et aL, 1978; Singer et aL, 1982; Lo & Sprent, 1986). Furthermore, thymocytes maturing in the presence of non-cytotoxic antibodies against a particular type of MHC glycoprotein yield a mature repertoire having a selective defect in the use of that MHC glycoprotein as a restriction element while exhibiting normal restriction patterns to other types of endogenous MHC glycoproteins (Kruisbeek et at., 1985; Marrack et aL, 1988a; Marusic-Galesic et at., 1988, 1989; Zuniga-Pflucker et al., 1989). Together, these two lines of investigation indicate that MHC glycoproteins on thymic APC are critical selecting elements for the maturing T cell repertoire. However, it is a mystery how maturing T cells detect thymic MHC ligands. The observations of positive selection indicate that maturing T cells must recognize self MHC ligands in the absence of foreign antigen whereas the widely accepted "altered self" view of T cell antigen recognition holds that their mature counterparts can only recognize self MHC glycoproteins in the presence of foreign antigen. The second paradox is derived from the contrasting experimental implications that thymocytes may be either positively selected or negatively selected by the same types of TcR interactions with self MHC ligands. For example, it has been shown that immature thymocytes having TcR specific for self MHC ligands are deleted from the maturing T cell repertoire (reviewed by Schwartz, 1989). According to immunological theory, "'self" peptides which are derived from intrathymic as well as from extrathymic tissues can be bound by thymic MHC glycoproteins and presented to T cells having self-reactive TcR (Lorenz & Allen, 1989; Marrack et al., 1989). In the thymic tissue of a neonate, MHC glycoproteins almost entirely have "self" peptides in their peptide binding grooves because foreign antigens would generally be absent in this environment. T cells that exhibit autoreactivity with self
M H C glycoproteins or a vast array of " s e l f " peptides complexed with M H C glycoproteins are deleted from the T cell repertoire. Negative selection o f thymocytes provides an important basis o f immunological self-non-self discrimination because the T cell repertoire is purged o f autoreactivity while retaining reactivity to a wide range of potentially foreign compounds. However, maturing thymocytes would be exposed to a wide array o f " s e l f " antigens which would include the M H C "restriction" sites that promote positive selection. Because T cells recognize M H C restriction sites with the same TcR binding pocket used for recognition of other self antigens, why is the TcR repertoire not also purged of reactivity to M H C restriction sites?
3. Contemporary Models of Thymic Selection Without doubt, the repertoire of TcR undergoing positive selection must be different from that for TcR undergoing negative selection, or all maturing T cells would be deleted. Although the basis of this difference is not known, strict compliance to the altered self hypothesis suggests that quantitative differences rather than qualitative differences in TcR recognition of M H C ligands account for the distinct processes o f positive and negative selection. Many models proposed to reconcile the opposite processes o f thymic selection are based upon a c o m m o n theme. That is, immature thymocytes that experience low to medium affinity T c R - M H C ligand interactions are positively selected whereas those experiencing high affinity T c R M H C ligand interactions are deleted by mechanisms of negative selection (Lo et al., 1986; Sprent et al., 1988; Singer et al., 1986; Fowlkes & Pardoll, 1989). This model is presented in Fig. 1. According to one version of this model (Lo et al., 1986; Sprent et al., 1988), thymocytes that experience low to high affinity interactions with self M H C ligands on thymic epithelium undergo positive selection. Then, in a subsequent bout o f negative selection, thymocytes that experience high affinity interactions with self M H C ligands on bone marrow derived thymic APC undergo clonal deletion. By this mechanism, T cells expressing low to medium affinity for self M H C ligands would constitute the mature T cell repertoire. Given the surviving clonotypic diversity, a T cell repertoire selected to inefficiently recognize self M H C ligands would almost invariably contain clones able to bind complexes of M H C and foreign antigen with high affinity.
FiG. 1. The "'affinity" model of thymic selection. This represents a contemporary model of thymic selection based upon quantitative recognition of self MHC glycoproteins. (a) The immature repertoire of T cells express a range of affinities for self MHC ligands (represented by arbitrary units on the left y-axis). These T cells are arranged along the x-axis in order of increasing affinity for self MHC ligands. Initially, different T cell clonotypes are present at similar frequencies (clonal abundance represented by arbitrary units on the right y-axis). During thymic selection, T cells expressing either medium affinity TcR or high affinityTcR for MHC ligands undergo either positive selection or negative selection (clonal deletion), respectively. (b) The mature T cell repertoire is thereby primarily comprised of T cell clones exhibiting medium affinity interactions with self MHC ligands. (c) Within the mature T cell repertoire, a small number of clones having the appropriate clonotypic specificitywill recognize complexes of MHC glycoproteins and foreign antigen "'X'" as high affinity ligands and will mediate a protective immune response.
( o ) The immature Tcell repertoire: T cells will experience 140 I I .... I 120 tL
. . . . . . . . . . . . . . . . . . . . . . . . . . [I',I I ........................
Individual clones arranged in order of increasing affinity for self MHC proteins
The mature T c e l l r e p e r t o i r e 100
~=+I o I "6
individual clones arranged in order of increasing affinity for self MHC proteins
c ) The response to foreign antigen
120 I00 80
_~ 4o u
Individual clones arranged in order of increasing affinity for self MHC proteins plus foreign antigen "X'"
The central question concerning these "affinity" models is whether a mature T cell repertoire expressing intermediate affinity for self MHC ligands would provide fail-safe protection against autoimmunity. According to the contemporary view of T cell activation, TcR occupancy is one of the primary determining factors of T cell selection and T cell responses. The affinity model is based on the assumption that TcR affinity for MHC ligands would be proportional to the occupancy of TcR binding sites by MHC ligands. However, TcR occupancy is also critically dependent upon the concentration of the respective MHC ligand. Even if the TcR affinity for certain self MHC ligands is intermediate as would be the case for positively selected T cells, threshold levels of TcR occupancy would occur at high ligand concentrations. Thus, the affinity model would not provide protection against autoimmune disease when the peripheral concentration of a self MHC ligand substantially exceeds its corresponding concentration in the thymus. By this model, quantitative differences in affinity represent the critical variable distinguishing normal homeostatic TcR-MHC-self peptide interactions from TcRMHC-foreign antigen interactions. The question then is whether quantitative differences in affinity can represent the relevant parameter by which the emerging T cell repertoire is selected given the large variation in the concentration of self MHC ligands from tissue to tissue. Nevertheless, T cells adequately discriminate self from non-self in tissues expressing low levels of Class II MHC glycoproteins as well as in sites of inflammation where clonal expansion leads to a high density of clonotypic TcR and where gamma-interferon acts to greatly amplify expression of MHC gene products. Thus, by itself, the affinity model does not appear to adequately explain self/non-self discrimination unless it is complemented by independent mechanisms such as peripheral tolerance induction. The thesis herein concerns whether processes of thymic selection based on parameters other than affinity can theoretically account for the virtually flawless form of self-non-self discrimination mediated by the immune system. Alternative models have been proposed suggesting that qualitative differences rather than quantitative differences in TcR-MHC ligand interactions account for immunological learning of self-non-self discrimination. For example, one model proposes that positive and negative selection are respectively mediated by thymic epithelium and thymic macrophages with each cell type presenting a distinct set of "self" peptides to maturing T cells (Marrack & Kappler, 1988; Marrack et aL, 1988b). However, it is ditficult to envision how such a mechanism would adaptively select the maturing T cell repertoire. An alternative version of this hypothesis is that thymic epithelium produces a broad range of "self" peptides to represent the universe of foreign antigens (Kourilsky & Claverie, 1989). These mutated peptides would be presented to immature T cells by thymic epithelium to serve as critical selecting elements of positive selection. In contrast, "'self" peptides presented to maturing T cells by thymic macrophages would serve as critical elements of negative selection. However, it is not clear how compartmentalization of "mutated foreign" peptides and "self" peptides could be maintained in adjacent cellular pools without extensive "cross-presentation" of peptides. Furthermore, peptides that are randomly generated by mutational mechanisms could not selectively represent just the universe
THE EFFICACY MODEL OF T CELL SELECTION
of foreign antigens because peptides mimicking self structures would also be generated. The above models of T cell antigen recognition and thymic selection are based on the assumption that threshold levels TcR occupancy by MHC proteins or by MHC-antigen complexes invariably result in TcR mediated activation. Yet this assumption has not been verified experimentally for TcR-MHC ligand interactions and has proven not to be a valid description of hormone-receptor interactions. 4. The "Efficacy" Model of Thymic Selection
The "efficacy" model presented herein reconciles the conflicting observations of positive and negative selection of thymocytes based on a modified "altered self" view of TcR recognition. The basis for this model stems from pharmacological analysis of hormone-receptor interactions (reviewed by Limbird, 1986). Early studies of the quantitative relationship between receptor occupancy and receptor-induced responses revealed many instances of non-proportionality. When concentrations of different ligands were controlled to achieve the same level of receptor occupancy in a given bioassay, these ligands were found to differ drastically in their capacity to elicit receptor mediated activity. In fact, many ligands exhibiting high affinity binding to physiological receptors did not elicit any receptor mediated activity. Thus, receptor ligands varied in a property referred to as "efficacy" which can be defined as a ligand's capacity to catalyze receptor mediated biological activity (Stephenson, 1956; Franklin, 1980). Ligands lacking efficacy were referred to as competitive antagonists whereas other ligands having full efficacy were referred to as agonists. Efficacy and affinity are independent parameters of ligand-receptor interactions and are conferred by physically different regions of the ligand. That is, competitive antagonists bind their receptors with high affinity but lack structural domains necessary for catalysis of biological activity. Efficacy is a concept that can be applied generally to describe most receptor-ligand interactions because high affinity antagonists have been described for many receptor systems. Given that T cells almost exclusively recognize peptidic compounds as antigens, it is of interest to note that structure-function analysis of physiological peptides reveals these ligands are comprised of two separable domains. According to the "address-message" model (Schwyzer, 1977; Chavkin & Goldstein, 1981; Portoghese et at., 1988), an "'address" domain mediates the specificity and affinity of receptor binding whereas a distinct "message" domain confers efficacy and thereby induces receptor mediated biological activity. By this model, peptide residues which confer high affinity binding are often separate from those residues which confer efficacy. This principle has led to the synthesis of a wide variety of antagonists for physiological peptides, including angiotensin II, bombesin, bradykinin, calcitonin gene related peptide, corticotropin releasing factor, cholecystokinin, dynorphin, gastrin, gastrin releasing peptide, glucagon, interleukin 1, leutinizing hormone releasing hormone, neutrophii chemotactic peptides, parathyroid hormone, somatostatin, substance P, vasoactive intestinal peptide, and vasopressin (Douglas et al., 1985; Coy et aL, 1988; Vavrek & Stewart, 1985; Chiba et al., 1989; Rivier
et al., 1984; Lignon et al., 1987; Chavkin & Goldstein, 1981; Martinez et aL, 1984; H e i m b r o o k et aL, 1989; Unson et aL, 1987; Palaszynski, 1987; Coy, 1981; Day et al., 1980; Rosenblatt, 1986; Coy et al., 1985; Caranikas et aL, 1982; Pandol et al., 1986; Manning et aL, 1984; respectively). The concept of efficacy has not previously been associated with TcR transduction mechanisms, probably in part because cross-linking studies with bivalent anti-TcR antibodies have indicated that receptor aggregation induces T cell activation (Meuer et al., 1983; Kaye et aL, 1983; Kappler et al., 1983). However, efficacy and receptor aggregation do not necessarily represent exclusive mechanisms of signal transduction. For example, structure-function studies of L H R H indicate that the concept of efficacy accurately represents L H R H - r e c e p t o r interactions. That is, residues at positions 1, 6, 10 are crucial for high affinity binding whereas residues at positions 2 and 3 are largely responsible for transducing receptor activation (Coy, 1981). Nevertheless, artificially induced receptor aggregation also induces receptor mediated biological activity (Conn et al., 1982; Gregory et al., 1982). That is, L H R H antagonists which are chemically dimerized retain their antagonistic properties yet when these bivalent L H R H antagonists are cross-linked with a n t i - L H R H antibody, the resulting complex induces receptor aggregation and behaves as an agonist. One interpretation is that efficacious receptor-ligand interactions induce conformational changes that in turn induce receptor aggregation and biological activity. The purpose of this manuscript is to provide a theoretical alternative to the c o n t e m p o r a r y "'affinity" models o f t h y m i c selection. In discussion o f this new model, the term " M H C ligand" will be used extensively to incorporate the possibility that the T cell repertoire may be selected by both unoccupied M H C glycoproteins and complexes of self peptide and M H C glycoprotein. As summarized in Fig. 2, this model is based on the concept that efficacy is a central p a r a m e t e r controlling acquisition of immunological self-non-self discrimination during thymic selection. The following six tenets comprise the "efficacy" model. (a) The germline TcR binds self M H C ligands with a range of affinities. Among those clones able to bind self M H C ligands, some clones bind with high efficacy whereas other clones exhibit non-efficacious interactions with these M H C ligands. (b) The consequences o f TcR mediated cellular activation for immature T cells are fundamentally different from those for mature T cells. I m m a t u r e T cells that experience TcR mediated activation undergo cell death and clonal FIG. 2. The "efficacy" model of thymic selection. This represents a novel hypothesis in which the concept of efficacyprovides the basis of immunological self-non-self discrimination and thymic selection. (a) The immature repertoire of T cells express a range of affinities for self MHC ligands (represented by arbitrary units on the y-axis). T cells exhibiting affinity for self MHC ligands may bind these ligands as either agonists (with efficacy) or antagonists (without efficacy). Initially, different T cell clonotypes are present at similar frequencies. (b) T cells that recognize MHC ligands as agonists undergo negative selection and clonal deletion. T cells that recognize MHC ligands as antagonists undergo positive selection to the extent that their clonal frequency is proportional to their affinity for self MHC. (c) Given the diversity of the mature T cell repertoire, it is probable that at least a small number of clones will bind complexes of MHC glycoproteins and foreign antigen "'X" as agonistic ligands. These T cells will then clonally expand and mediate an adaptive immune response.
( o ) The immature Tcell repertoire: ---.I-- Antagonistic binding t'o self MHC ligonds ----0-- Agonistic binding to self MNC ligands
o u 40 "6 >, .~- 20
individual clones arranged in order of increasing affinity for self MHC proteins
b ) The mature T cell r e p o r t o i r e
! Bindin~ to self MHC figands lacks efficacy
§ 6o o
4o L~ 20
Individual clones arranged in order of mcreosm( affinity for self MHC proteins
c ) The response to foreign antigen
I00 80 c
~ Binds MHC plus foreign antigen "X • as on agonist
o: ........... ,,,i,,, I,!, ..... !, ..................................... Individual clones arranged in order of increasing affinity for self MHC proteins
deletion. Mature T cells that experience TcR mediated activation exhibit vigorous proliferation and express specific immunological activities. Immature T cells that exhibit non-efficacious binding to thymic MHC ligands are positively selected. Based on their respective TcR affinity for MHC ligands, T cells compete for attachment sites on thymic APC and thereby compete for preferential access to differentiation/growth-promoting influences of thymic APC. Positive selection of T cells may thereby be mediated by a paracrine exchange of tropic influences during thymocyte-APC adhesion. The paracrine influences would be transduced through non-TcR mediated pathways to promote positive selection by facilitating T cell maturation. Immature T cells that exhibit efficacious interactions with self MHC ligands undergo negative selection and clonal deletion. An immature T cell clone that exhibits efficacious interactions with a self peptide " A " - M H C complex as well as non-efficacious interactions with a crossreactive self peptide "'B"MHC complex would also be deleted from the repertoire. By this mechanism, the T cell repertoire is purged of harmful autoreactivity. Given the clonotypic diversity of the mature T cell repertoire, at least a few clones of the appropriate specificity would bind complexes of MHC-foreign antigen as highly efficacious ligands. For those clones, previously selected TcR-MHC ligand interactions and prior antigen-MHC glycoprotein interactions would provide much of the affinity stabilizing the trimolecular complex. Additional TcR-foreign antigen interactions would provide efficacy to induce biological activity. CD4 and CD8 accessory molecules on immature Tcells potentiate interactions of clonotypic TcR with their respective class II and class I MHC ligands during thymic differentiation, respectively. In the thymus, CD4- or CD8mediated potentiation of non-efficacious interactions facilitates positive selection whereas CD4- or CD8-mediated potentiation of efficacious interactions enhances negative selection.
4+I+ T H E T C R
The "efficacy" model adopts previous suggestions (Jerne, 1971; Marrack & Kappler, 1987) that the germline antigen receptor repertoire has inherent affinity for MHC glycoproteins. That is, TcR molecules may be structurally designed to interact with MHC gene products. This inherent binding tendency may explain why MHC molecules rather than other types of cell surface molecules are able to restrict Tcell responses. This type of interaction may involve framework interactions between particular V regions and particular class I or class II gene products and may contribute to known biases in antigen reactivity of particular TcR V region genes with particular MHC glycoproteins (Kappler et aL, 1987b). In such cases, biases in antigen reactivity may also reflect the tendency of certain antigenic peptides to preferentially bind polymorphic domains of allelic MHC molecules such that the complex is recognized by certain TcR V region gene products.
In addition to genetically predisposed TcR recognition of MHC molecules, a second mechanism of positive thymic selection is required to insure that the TcR repertoire optimally recognizes polymorphic "restriction sites" on an individual's MHC glycoproteins. During the initial phases of thymic selection, a given clonotypic TcR may be exposed to a myriad of different "self" peptide-MHC complexes. This TcR may exhibit substantial affinity for one or a number of these MHC ligands. The basic tenet of the "efficacy" model is that TcR interactions with MHC ligands may vary in efficacy. Some clonotypic TcR would interact with these MHC ligands as agonists. In all probability, certain TcR would interact with more than one MHC ligand and recognize some as agonists and others as antagonists. Other TcR would recognize their respective MHC ligands exclusively as antagonists. 4.2. A U N I F I E D
Lederberg (1959) originally suggested that immature lymphocytes exhibit an adverse sensitivity to antigens whereas mature lymphocytes exhibit reactivity to antigens. This postulate has since been confirmed as an accurate description of T cell maturation (e.g. see Matzinger & Guerder, 1989) and has therefore been adopted as a central tenet of the "efficacy" model. That is, immature T cells that bind agonistic MHC ligands undergo cell death. In contrast, their mature counterparts that bind the same agonistic MHC ligands are induced to proliferate and to mediate specific immune functions. According to the "efficacy" model of thymic selection, an immature thymocyte may experience one of three potential fates that are largely determined by the expresssion of a clonotypic TcR and the haplotype of "self" MHC. As illustrated in Fig. 3, T cells that bind self MHC ligands as non-efficacious ligands are positively selected whereas those clones that bind self MHC ligands as efficacious ligands are negatively selected. Clones not possessing affinity for self MHC ligands are not subject to thymic selection. 4.2.1. Positive thymic selection According to the "efficacy" model (Figs 2 and 3), certain clones of immature T cells exhibit non-efficacious interactions with self MHC ligands. Because these binding events lack efficacy, these clones do not experience TcR mediated activation during thymic selection and therefore escape clonal deletion. By a mechanism of intercellular adhesion initiated by TcR-MHC ligand conjugation, these immature T cells compete for MHC ligands expressed at the surfaces of thymic APC and thereby compete for privileged access to differentiation and growth promoting factors. These thymic growth promoting factors stimulate positive selection of adherent thymocytes through non-cytolytic pathways that do not involve TcR mediated transduction signals. In fact, several studies support the existence of alternative, non-TcR mediated pathways of thymocyte stimulation which may involve interleukin 1 as well as CD2 adhesion molecules (Rock & Benacerraf, 1984; Reem et al., 1987). By this mechanism, thymocytes exhibiting non-efficacious interactions with thymic MHC ligands experience positive selection in proportion to their affinity for these ligands.
M.D. MANNIE Paracrine growth faclors
Non-efficacious binding ~o self MHC ligands
posdive selection <
Clonal deletion, negative selection <
No thymic selection <
Efficacious binding to self MFIC ligands
No affinity for self MHC ligands ~ X
FIG. 3. The "'efficacy" model o f thymic selection: a diagrammatic view. According to the "'efficacy'" model, maturing thymocytes may experience one of three fates based on the affinity and efficacy of clonotypic TcR interactions with self MHC ligands. (a) T cells expressing TcR that bind self MHC ligands as non-efficacious ligands are positively selected during thymic development and eventually dominate the mature T cell repertoire. (b) T cells expressing TcR that bind self MHC as efficacious ligands are deleted from the repertoire. (c) T cells expressing TcR that do not bind self MHC ligands do not experience either positive or negative selection, This model thereby predicts that the affinity of the T c R - M H C interaction determines whether a given T cell will be subject to thymic selection. This model also predicts that among those T cells expressing affinity for self MHC ligands, the efficacy o f the T c R - M H C interaction determines whether a given T cell will be positively or negatively selected during development.
One possibility is that intercellular associations stabilized solely by non-efficacious TcR-MHC iigand interactions are of sufficient tenacity to promote positive selection. Alternatively, these non-efficacious TcR-MHC may merely initiate cell-cell interactions which are subsequently strengthened through independent mechanisms of adhesion. The signal for upregulated adhesion would not stem from the TcR p e r s e but rather from the time dependent activation of cell surface receptors triggered by cell-cell apposition. Thus, non-efficacious TcR binding to MHC ligands may catalyze adhesive responses merely by initiating cell-cell interaction.
THE EFFICACY MODEL OF T CELL SELECTION
4.2.2. Negative thymic selection Other immature T cells exhibit highly efficacious interactions with thymic MHC ligands. In a manner proportional to the affinity of the TcR for thymic MHC ligands, these T cells undergo cellular activation and are proportionally deleted from the repertoire. An immature T cell clone exhibiting both highly efficacious interactions with a complex of MHC-self peptide "A" and non-efficacious interactions with a crossreactive complex of MHC-self peptide "B" would also be deleted. By this mechanism, T cell clones that could potentially induce autoimmune disease are purged from the repertoire. 4.2.3. No thyrnic selection Thymocytes not having sufficient affinity for thymic MHC ligands are neither positively nor negatively selected. These cells may undergo cell death due to insufficient access to thymic differentiation/growth-promotingfactors. Alternatively, these cells may survive and mature but may never comprise more than a minor T cell population. Evidence supporting the latter possibility stems from the observation that peripheral T cells purged of ailoreactivity can be primed to respond to specific antigens presented by ailogeneic MHC gene products (Thomas & Shevach, 1977).
4.3. R O L E O F C D 4 A N D
IN T H Y M I C
Any credible model of thymic selection must be able to fully account for pivotal roles of CD4 and CD8 in both positive and negative thymic selection. Mature progeny of "double positive" (CD4 ÷, CD8 ÷) thymocytes exclusively express either CD4 or CD8 and recognize antigens restricted by either class II or class I MHC glycoproteins, respectively (Wilde et aL, 1983; Marrack et aL, 1983; Swain, 1981). This relationship stems from the respective capacities of CD4 and CD8 to bind non-polymorphic regions of class II and class I MHC glycoproteins and thereby to potentiate TcR recognition of class II and class I MHC restricted antigens (Doyle & Strominger, 1987; Gay etaL, 1987; Norment etaL, 1988; Gabert etal., 1987). Several lines of evidence indicate that these accessory molecules are implicated not only in T cell antigen recognition but that they are also pivotal in thymic selection of the T cell repertoire (MacDonald et al., 1988b; Zuniga-Pflucker et al., 1989). For example, a and/3 transgenes encoding a TcR specific for class I MHC of haplotype " A " and foreign antigen " X " were expressed in either of the three following types of mice; (a) in type '~A" mice that possessed endogenous antigen "'X", (b) in type " A " mice that did not express antigen " X " , and (c) in type "B'" mice (Sha et al., 1988a, b; Teh et aL, 1988; Kisielow et al., 1988). In the first group of mice, only those T cells bearing TcR transgenes and high levels of CD8 were deleted in the thymus. Minor populations of T cells bearing the TcR transgenes together with low or negligible levels of CD8 were spared from negative selection. In the second group of mice, the TcR transgenes were selectively expressed in the CD8 ÷ lineage of mature T cells. In the third group of mice, the TcR transgenes were largely not detected in mature T cells, reflecting the lack of positive selection of those transgene
bearing T cells. Additional experiments have shown that mice chronically treated from birth with either anti-class II or anti-class I MHC antibodies failed to develop either CD4 ÷ or CD8 + cell lineages, respectively (Kruisbeek et al., 1985; MarusicGalesic et al., 1988). Conversely, in vivo treatment with non-depleting anti-CD4 monoclonal antibodies prevented clonal deletion of autoreactive class II M H C - restricted T cells (MacDonald et al., 1988a; Lynch et al., 1985). These observations indicate that interaction of TcR and either CD4 or CD8 with class II or class I MHC molecules is critical for thymic selection of the CD4 ÷ or CD8 + cell lineages, respectively. The "efficacy" model can readily account for the selective expression of CD4 and CD8 molecules on T cells that are restricted by class II and class I MHC molecules, respectively. According to this model, CD4 and CD8 augment interactions of TcR with self MHC ligands during thymic differentiation in the same manner as they augment MHC--restricted recognition of foreign antigens by mature T cells. As discussed previously (section 4.1 ), clonotypic TcR expressed on "'double positive'" (CD4 ÷, CD8 ÷) thymocytes may have an inherent bias to bind particular class I or class II MHC-restricted ligands. As diagrammed in Fig. 4, the CD4/CD8 accessory molecules would potentiate TcR interactions with self peptides presented by these MHC molecules without altering the intrinsic efficacy of this interaction. For example, CD4 on "'double positive" thymocytes would potentiate non-efficacious interactions of certain TcR with self peptide-class II MHC complexes and would thereby facilitate positive selection. Conversely, CD4-mediated potentiation of efficacious interactions between other TcR and their respective MHC ligands would facilitate negative selection. One possibility is that physical conjugation of either CD4 or CD8 during stabilization of TcR-MHC antigen complexes may trigger feedback signals which promote differentiation of the CD4 ÷ or CD8 ÷ lineages, respectively. For example, CD4 engagement and stabilization of non-efficacious interactions between clonotypic TcR and a given class II MHC ligand may not only promote the clone's positive selection but also may trigger the CD4 ÷ differentiation pathway (Fig. 4). 4.4. A R E V I S E D
SELF'* VIEW OF T CELL
The "efficacy" model provides a novel view of TcR recognition of antigen. The central prediction is that TcR of mature T cells bind "restriction sites" in self MHC glycoproteins as if they were antagonistic ligands. Additional interactions between clonotypic TcR and self peptides complexed with MHC glycoproteins would further strengthen these non-efficacious binding interactions. A central question concerns whether the concept of efficacy can account for how the binding of highly variable signals (i.e. antigens presented by polymorphic MHC molecules) to highly variable binding pockets (i.e. CDR3-1ike or junctional regions of the TcR heterodimer) can be transduced into a limited set of active or inactive receptor conformations. One possibility is that conserved or framework residues which are located either within or in close proximity to the TcR binding pocket may represent pivot points which control the global conformation of the domain. Depending upon the unique
OF T CELL
Positiveselectionof CD4 bearingT ceils
Negative seleclionof CO4bearingTcells
,.Week affinity Non-efficaciousr
.Weak affinity / Efficacious
l Class MHCIgand
~o c) J ClassI MHCIgond J L,
( b ) ~ Highaffinity / I TcRI I Non-effcacous /
,,.High afflniiy I MHC i lassigand ~ Efficacious~ fh
c ~oo rr
JMHC Class [ igand
Differentiation info singie posdive C04 bearing T celt
FIG. 4. Hypothetical role o f CD4 and CD8 accessory molecules in thymic selection. According to this model, CD4 and CD8 accessory molecules enhance the affinity of naturally occurring T c R - M H C ligand interactions without alternating the efficacy of this interaction. (a) For example, a weak non-efficacious interaction between a clonotypic TcRI and a particular class II MHC ligand would be substantially enhanced by a co-operative C D 4 - M H C glycoprotein interaction. (b) CD4 enhancement of this TcR1MHC glycoprotein interaction would facilitate the clone's ability to compete for self MHC ligands and would thereby facilitate positive selection of that clone. Differentiation signals originating from these complexed CD4 or class II MHC molecules would commit the T cell to the CD4 single positive lineage. By this mechanism, the tendency o f TcR variable regions to bind a particular class I or class II MHC molecule would ultimately determine committment to either the CD8 or CD4 lineage, respectively. (c) and (d) Conversely, a weak efficacious interaction between clonotypic TcR2 and a class II MHC tigand would also be enhanced by CD4. tn this case however, CD4 enhancement of TcR2-MHC interactions would facilitate clonal deletion.
stereochemistry of a given binding interaction, these pivotal residues either would or would not be triggered to transduce conformational shifts. Thus, within the immature T cell repertoire, a large degree of randomness would be inherent in whether a given clonotype recognizes a self MHC ligand as an agonist or an antagonist. However, after thymic selection as described herein, T cells would recognize self M HC ligands exclusively as antagonists and would recognize foreign MHC ligands as either agonists or antagonists. Alternatively, one may envision that the TcR is structurally designed such that the periphery of the binding pocket is conformationally rigid and is positioned to interact with outlying residues of the MHC peptide binding groove to mediate high affinity non-efficacious binding. Central, more conformationally sensitive regions
of the TcR binding pocket may be positioned to interact with MHC bound antigen to engender efficacy. In a similar manner, the MHC-antigen ligand may be structurally designed such that the residues of the MHC glycoprotein which are oriented upward from the rim of the "peptide-binding" groove may be positioned to bind conformationally rigid regions of the TcR and thereby engender non-efficacious interactions. Residues of the antigen may be positioned to encounter more flexible regions of the TcR and thereby would affect the efficacy of the interaction. Thus, the "efficacy" model of T cell antigen recognition predicts that the attributes of affinity and efficacy are manifest in physically separate domains of the TcR binding pocket.
4.5. A N O V E L V I E W O F A L L O G E N E I C
T CELL REACTIVITY
The "efficacy" model also provides an unique explanation of T cell alloreactivity. That is, T cells maturing in the presence of haplotype " A " MHC glycoproteins would be positively selected to recognize these MHC glycoproteins and a vast array of self peptide-MHC " A " complexes as antagonists. As discussed in section 4.1, germline TCR may be inherently predisposed to bind MHC molecules. Just like an immature repertoire, these mature "A"-restricted T cells would interact with haplotype " B " MHC glycoproteins and an unique set of peptide-MHC "B'" complexes. But because these " A " restricted T cells have not been negatively selected by haplotype " B " MHC ligands, many of these T cells would recognize haplotype " B " MHC ligands as agonists. Hence, a large proportion of these T cells would be activated and would thereby exhibit alloreactivity when mixed in culture with APC expressing haplotype "B" MHC glycoproteins. This concept of alloreactivity differs qualitatively from the contemporary views of alloreactivity supported by the "affinity" model. That is, the "efficacy" model attributes alloreactivity to highly efficacious TcR interactions with allogeneic MHC ligands whereas the "affinity" model attributes alloreactivity solely to high affinity interactions of TcR with allogeneic MHC ligands.
4.6. T H E
OF THE "'EFFICACY"
The "efficacy" model of thymic selection describes MHC-restricted antigen recognition as a qualitative distinction between non-self and self entities based respectively upon the presence or absence of efficacy in TcR-MHC ligand interactions. This is in marked contrast to the "affinity" model which proposes that self-non-self discrimination is based solely upon differences in TcR affinity for MHC ligands. Furthermore, the "efficacy" model describes a process of thymic selection that would produce a mature repertoire of T cells able to bind with high affinity to self MHC glycoproteins without experiencing cellular activation or autoimmunity. By this mechanism, the T cell repertoire is selected to optimally find foreign antigens where they are most likely to appear, i.e. in the peptide binding groove of a self MHC glycoprotein.
THE EFFICACY MODEL OF T CELL SELECTION
One o f the more important rationales for this model is that MHC-restricted recognition based upon an "efficacy" mechanism of thymic selection would provide a highly efficient means for self-non-self discrimination. According to the "'efficacy" model, immunological self-non-self discrimination is based entirely on those residues in antigenic peptides that confer efficacy to T c R - M H C interaction. That is, for mature T cells, any entity that confers efficacy to homeostatic T c R - M H C ligand interactions represents a foreign antigen. Efficacy of peptide hormones is often attributed to a single or a few residues. By this model, T cells are selected to recognize self peptides complexed with MHC glycoproteins and would thereby be extremely sensitive to any non-self structure that closely mimicked a self structure. As with foreign proteins, T cell recognition o f self proteins may be focused upon " i m m u n o d o m i n a n t " sequences (i.e. peptides that form stable complexes with M H C glycoproteins) and be mediated by multiple clones o f T cells that recognize a cluster of overlapping determinants. In all probability, peptides that differ by even minor modifications from a self sequence will confer efficacy to at least one of these clonotypes and would thereby be recognized as a foreign entity. This model thereby accounts for one of the hallmarks o f immunological specificity, that T cells often exhibit all-or-none discrimination between self peptides and very closely related foreign peptides. By this mechanism, the immune response may be very well-adapted to provide effective immunity, especially in cases wherein pathogens have evolved to mimic self proteins and in cases wherein somatic mutation of a self protein promotes a pathogenesis such as cancer. Implicit to the "efficacy" model is a mechanism which would greatly facilitate T cell recognition of foreign antigens. That is, non-efficacious interactions between TcR and M HC glycoproteins would catalyze antigen recognition because a multitude of normal homeostatic T c R - M H C ligand interactions would facilitate adherence of the T cell and APC cell surfaces, thereby promoting yet additional T c R - M H C ligand interactions. Such a mechanism would enhance the probability that a TcR would interact with a potentially small pool of MHC-foreign antigen complexes. Thus, homeostatic T c R - M H C ligand interactions would comprise a mechanism of immune surveillance because these interactions would direct the T cell "search" for antigens to the surface of APC. Note that such a mechanism is not predicted by the "affinity" model of thymic selection because high affinity T c R - M H C ligand interactions would uniformly lead to negative selection and clonal deletion. Conversely, one may argue that high levels of MHC "self antagonists" may compete with and block the relevant interactions of TcR with MHC-foreign antigen complexes. Yet competitive antagonism may not be an important factor because only a small percentage of TcR occupancy by agonistic MHC ligands may be required for T cell activation. That is, T cells may have "'spare receptor" capacity in that the actual surface concentration o f TcR may far exceed that required for maximal biological responses. Spare receptor capacity appears to be an adaptation designed to confer high sensitivity to low concentrations o f agonist (Nickerson, 1959; reviewed by Limbird, 1986). Competitive antagonism may be even less of a factor if TcR concentration were maintained in excess of MHC or if TcR were able to sequentially bind many different MHC ligands during cell-cell contact with the
APC. Any mechanism enabling TcR to "sample" a substantial proportion of the MHC ligand on the APC cell surface while integrating agonistic TcR-MHC ligand signals would further enhance efficiency of foreign antigen recognition. 4.7. U N I Q U E E X P E R I M E N T A L P R E D I C T I O N S OF TH E " ' E F F I C A C Y " M O D E L
Theoretical frameworks that best facilitate advancement of science are comprised of alternative hypotheses which are broadly enough defined so that together they account for all possible possibilities and yet are narrowly enough defined so that each is testable and there by amenable to experimental falsification. The "affinity" model was originally proposed to account for paradoxical observations of positive thymic selection. The "efficacy" model has been proposed to provide an alternative explanation of positive thymic selection. Both models are broadly defined in that either one or the other will probably provide an accurate representative of thymic selection. Because both models are broadly defined, each model has many experimental predictions. Direct comparison of the two models reveals instances of contrasting experimental predictions, thereby revealing relevant areas of experimentation. One potential means to distinguish the "affinity" and "efficacy" models derives from their differing predictions concerning the mechanism of antigenic competition (see Fig. 5). According to both models, one focus of antigenic competition is the peptide binding groove of MHC glycoproteins. That is, antagonistic peptides competitively inhibit antigen presentation by binding MHC glycoproteins and thereby effectively reducing the available pool of MHC glycoproteins. However, the two models have contrasting predictions concerning whether peptide/MHC complexes can compete for binding sites on TcR. In contrast to the efficacy model, the affinity model predicts that antagonist/MHC molecules do not interact with TcR of the responding T cell. According to the "efficacy" model, some MHC bound antagonists act upon the TcR much the same as a hormone antagonist would act upon a hormone receptor. This mode of competitive inhibition is mediated by an equilibrium of MHC-antagonist complexes with TcR, thereby reducing the pool of TcR available for agonistic interactions. One means to distinguish these hypotheses is to arrange experimental conditions such that there is no competition between agonist and antagonist peptides for the same set of MHC molecules. In other words, agonist/MHC complexes should be formed separate from the antagonist/MHC complexes and then the two types of complexes should be mixed to assess whether there is antigenic competition for TcR. The contrasting predictions of the efficacy model as compared to the affinity model is that these complexes either should or should not be potent inhibitors of T cell responses. In fact, any evidence indicating that isolated MHC-antagonist complexes can physically interact with TcR would strongly favor the "efficacy" model. A second potential means to distinguish the two models is that the "efficacy" model but not the "affinity" model predicts that some antigens will behave as partial agonists. Partial agonists are ligands that have mixed agonistic/antagonistic properties. In other words, mixed agonists stimulate maximal biological activity at levels
MODEL OF T CELL SELECTION
The "efficacy" model of antigenic competition MHC glycopre,teinagoaist complex
Agonist (enfigen) + MHC gtycoprofein + self peptide "~ + antagonist
MHC gtycoprotein.....~ s e l f peptide complex
MHC glycoprotem- .. antagonist complex
MHC glycoproleinTcR-Sogonist complex
T ceil antigen receptor (TcR)
MHC glycopmteinTcR-entegonist complex
The "affindy" model of antigenic compeirhon MHC glycoprotein-,agonlst complex
Agonist (antigen) + MHC g~ycoprotein + self peptide "" + onlogomst
MHC glycoprotelns self peptide complex
MHC glycoprotein....." ~ TcR-agonisl complex
T cell > oc|ivotion
T cell antigen receptor (TcR)
MHC glycoprotelnantagonist complex
FIG. 5. Differential experimental predictions of the "'efficacy'" and "'affinity'" models of thymic selection with regard to the mechanism of antigenic competition. (a) Common predictions; MHC glycoproteins would exist in equilibrium with antigen (agonist), the competitive inhibitor (antagonist), and wide array of self peptides. This equilibrium would strongly favor formation of M HC-peptide complexes. Competitive inhibitors would act by reducing the available concentration of M HC glycoproteins, thereby reducing the concentration of both MHC-antigen complexes and trimolecular MHC-antigen-TcR complexes. (b) The "efficacy" model predicts that some MHC-antagonist complexes interact with TcR whereas the "'affinity" model predicts that MHC-antagonist complexes should not interact with TcR. According to the "*efficacy" model, some competitive antagonists act by not only reducing the available pool of MHC glycoproteins but also by reducing the available pool of TcR.
below that elicited by full agonists. The main question involves whether optimal concentrations of different antigenic analogs in complexes with the same type of MHC molecules would elicit similar of different optimal levels of T cell response. For this approach to be meaningful, the experimental conditions would have to be arranged such that plateaus in concentration-response curves represent maximal TcR activation by their respective MHC ligands rather than saturation/limitations of the available MHC binding sites. Because partial agonists are also partial antagonists, a second prediction is that antigens behaving as partial agonists would partially inhibit the antigenic activity of a full agonist. Identification of antigens that behave as partial agonists/antagonists would strongly support the "efficacy" model. A third potential means to distinguish the two models stems from the prediction of the "affinity" model that T cells exposed to artifically high concentrations of the MHC restricting element (perhaps in planar membranes) should exhibit T cell
activation in the complete absence of a specific foreign antigen. Furthermore, most positively selected T cells should exhibit this type of "autologous" responsiveness. In contrast, the "efficacy" model predicts that even extremely high concentrations of the restricting M H C glycoprotein should not induce T cell activation in the absence of specific antigen. In such an experiment, it would be essential to determine that the "'autologous" activation signal was not provided non-specifically through adhesion molecules, co-stimulatory factors, or lymphokines.
The purpose of this manuscript is to explore theoretical implications of applying the pharmacological concept of efficacy to T cell antigen receptor function. This analysis provided a unified model for T cell antigen recognition and thymic selection of the T cell repertoire. In summation, this model predicts that immature thymocytes exhibiting efficacious interactions with self M H C glycoproteins are deleted from the repertoire. In contrast, immature thymocytes exhibiting non-efficacious interactions with self M H C glycoprotein are positively selected and eventually dominate the T cell repertoire. By this mechanism, mature T cells would be predisposed to bind the very sites responsible for antigen presentation without undergoing cellular activation. For mature T cell clones of the appropriate specificity, foreign antigens bound within the M H C glycoprotein peptide-binding groove would confer efficacy to the interaction with the respective TcR. As opposed to the "affinity" models which explain self-non-self discrimination based on gradations in affinity, the "'efficacy" model describes self-non-self discrimination based upon qualitative differences in the efficacy of T c R - M H C ligand interactions. Dr Mark D. Mannie is a Postdoctoral Fellow of the National Multiple Sclerosis Society. Additional salary support was provided by NIH Grant A1-19273. The author thanks Dr Craig Hammerberg, Dr Roderick Nairn, Dr David Thomas and Dr Michael Savageau for insightful discussions and careful critiquing of this manuscript. The author also thanks Peggy Shirley for her expert typing of the manuscript. REFERENCES BABBITT, B. P., ALLEN, P. M., MATSUEDA,G., HABER, E. & UNANUE, E. R. (1985). Binding of immunogenic peptides to la histocompatibility molecules. Nature, Lond. 317, 395-361. BENOIST, C. & MATHIS,D. (1989). Positive selection of the T cell repertoire: where and when does it occur? Cell 58, 1027-1033. BERG, L. J., PULLEN,A. M., DE ST. GROTH, B. F., MATHIS,D,, BENOIST,C. &DAVlS, M. M. (1989). Antigen/MHC-specific T cells are preferentially exported from the thymus in the presence of their MHC ligand. Cell 58, 1035-1046. BEVAN, M. J. (1977). In a radiation chimaera, host H-2 antigens determine immune responsiveness of donor cytotoxic cells. Nature, Lond. 269, 417-418. BJORKMAN, P. J., SAPER, M. A., SAMRAOU1,B., BENNE'I-F,W. S., STROMINGER,J. L. • WILEY,D, C. (1987a), Structure of the human class I histocompatibility antigen, HLA-A2. Nature, Land. 329, 506-512. BJORKMAN, P. J., SAPER, M. m., SAMRAOUI,B., BENNETT,W. S., STROMINGER,J. L. ~¢.WILEY,D. C. (1987b). The foreign antigen binding site and T cell recognition regions of class I histocompatibility antigens. Nature, Lond. 329, 512-518.
MODEL OF T CELL SELECTION
BROWN, J. H., JARDETZKY, T., SAPER, M. A., SAMRAOUI, B., BJORKMAN, P. J. & WILEY, D. C. ( 19883. A hypothetical model of the foreign antigen binding site of Ctass I I histocompatibility molecules. Nature, Lond. 332, 845-850. Buus, S., SE'rq'E, A., COLON, S. M., MILES, C- & GREY, H. M. (19873. The relation between major histocompatibility complex (MHC) restriction and the capacity of la to bind immunogenic peptides. Science 235, 1353-1358. CARANIKAS, S., MIZRAHI, J., D'ORLEANS-JUSTE, P. & REGOLI, D. (19823. Antagonists of substance P. Eur. J. Pharmacol. 77, 205-206. CHAVKIN, C. & GOLDSTEIN, A. (1981). Specific receptor for the opioid peptide dynorphin: structureactivity relationships. Proc. ham. Acad. Sci. U.S.A. 78, 6543-6547. CHIBA, T., YAMAGUCHI, A., YAMATANI, T., NAKAMURA, A., MORtSHITA, T., INUI, T. FUKASE, M., NODA, T. & FUJITA, T. (1989). Calcitonin gene-retated peptide receptor antagonist human CGRP (8-37). Am. J. Physiol. 256, E331-E335. CONN, P. M., ROGERS, D. C., STEWART, J. M., NIEDEL, J. & SHEFFIELD, T. (19823. Conversion of a gonadotropin-releasing hormone antagonist to an agonist. Nature, Lond. 296, 653-655. CoY, D. H. (1981). Chemical structure and biological activity of hormones and related compounds: 1. hypothalamic releasing hormones. In: Amino-acids, Peptides and Proteins Specialist Period (Sheppard, R. C., ed.) II, 386-389. London: Royal Society of Chemistry. CoY, D. H., HEINZ-ERAIN, P., JIANG, N. Y., SASAKI, Y., TAYLOR, J., MOREAU, J.-P., WOLFREY, W. T., GARDNER, J. D. & JENSEN, R. T. (19883. Probing peptide backbone function in bombesin: a reduced peptide bond analogue with potent and specific receptor antagonist activity. J. bioL Chem. 263, 5056-5060. CoY, D. H., MURPHY, W. A., LANCE, V. A., HOCART, S. J., SUEIRAS-DIAz, J. & MEZO, I. (19853. Somatostatin agonists and antagonists-peptide control of growth hormone secretion. In: Advances in Experimental medicine and Biology (Patel, Y. C. & Tannenbaum, G. S., eds) pp. 325-337. New York: Plenum Press. DAVIS, M. M. (19883. T cell antigen receptor genes. In: Molecular Immunology (Hames, B. D. & Glover, D. M., eds) pp. 61-79. Washington, D.C.: IRL Press. DAY, A. R., PINON, D., MUTHUKUMARASWAMY, N. & FREER, R. J. (1980). Synthesis of several chemotactic peptide antagonists. Peptides 1, 289-291. DOUGLAS, J. G., KHOSLA, b/_'..C. & BUMPUS, R. M. ( 19853. Efficacy ofocta- and heptapeptide antagonists of angiotensin II as inhibitors of angiotensin III binding in the rat adrenal glomerutosa. Endocrinology 116, 1598-1602. DOYLE, C. ~. STROMINGER, J. L. ( 19873. Interaction between CD4 and class II MHC molecules mediates cell adhesion. Nature, Lond. 330, 256-259. FOWLKES, B. J. & PARDOLL, D. M. (19893. Molecular and cellular events o f t cell development. Adv. Immunol. 44, 207-264. FRANKLIN, T. J. (1980). Binding energy and the activation of hormone receptors. Biochem. Pharmacol. 29, 853-856. GABERT, J., LANGLET, C., ZAMOYSKA, R., PARNES, J. R., SCHMITT-VERHULST, A.-M. & MALISSEN, B. (19873. Reconstitution of MHC class t specificity by transfer of the T cell receptor and Lyt-2 genes. Cell 50, 545-554. GAY, D., MADDON, P., SEKALY, R., TALLE, M. A., GODFREY, M., LONG, E., GOLDSTEIN, O., CHESS, L., AXEL, R., KAPPLER, J. & MARRACK, P. (1987). Functional interaction between human T-cell protein CD4 and the major histocompatibility complex HLA-DR antigen. Nature, Lond. 328, 626-629. GLAZIER, A., TUTSCHKA, P. J., FARMER, E. R. & SANTOS, G. W. (1983). Graft-versus-host disease in cyclosporin A-treated rats after syngeneic and autologous bone marrow reconstitution. J. expl. Med. 158, 1-8. GREGORY, H., TAYLOR, C. L. & HOPKINS, C. R. (1982). Luteinizing hormone release from dissociated pituitary cells by dimerization of occupied LHRH receptors. Nature, Lond. 300, 269-271. HEtMBROOK, D. C., SAAR1, W. S., BAL1SHIN, N. L., FRIEDMAN~ A., MOORE, K. S., R1EMAN, M. W., K1EFER, D. M., ROTBERG, N. S., WALLEN, J. W. & OL1FF, A. (1989). Carboxy-terminal modification of a gastrin releasing peptide derivative generates potent antagonists, J. biol. Chem. 264, 11258-11262. JENKINS, M. K., SCHWARTZ, R. H. & PARDOLL, D. M. (19883. Effects of cyclosporine A on T cell development and clonal deletion. Science 241, 1655-1658. JERNE, N. K. (19713. The somatic generation of immune recognition. Eur. J. Immunol. !, 1-9. KAPPLER, J., KUBO, R., HASKINS, K., WHtTE, J. & MARRACK, P. (19833. The mouse T cell receptor: comparison of MHC-restricted receptors on two T cell hybridomas. Cell 34, 727-737. KAPPLER, J. W., ROEHM, N. & MARRACK, P. (1987a). T celt tolerance by clonal elimination in the thymus. Cell 49, 273-280.
KAPPLER, J. W., WADE, T., WHITE, J., KUSHNIR, E., BLACKMAN, M., BILL~ J., ROEHM, N. MARRACK, P. (1987b). A T cell receptor V/3 segment that imparts reactivity to a class II major histocompatibility complex product. Cell 49, 263-271. KAPPLER, J. W., STAERZ, U., WHITE, J. & MARRACK, P. C. (1988). Self-tolerance eliminates T cells specific for Mls-modified products of the major histocompatibility complex. Nature, Lond. 332, 35-39. KAYE, J., PORCELLI, S., TITE, J., JONES, B. & JANEWAY, C. A. JR. (1983). Both a monoclonal antibody and antisera specific for determinants unique to individual cloned helper T cell lines can substitute for antigen and antigen-presenting cells in the activation of T cells. J. expl. Med. 158, 836-856. KlSlELOW, P., BLUTHMANN, H., STAERZ, U. D., STE|NMETZ, M. ~¢ VON BOEHMER, H. (t988). Tolerance in T-cell-receptor transgenic mice involves deletion of nonmature CD4+8 ÷ thymocytes. Nature, Lond. 333, 742-746. KOURILSKY, P. & CLAVERIE, J.-M. (1989). MHC restriction, alloreactivity, and thymic education: a common link? Cell 56, 327-329. KRUISBEEK, m. M., MOND, J. J., FOWLKES, B. J., CARMEN, J. m., BRIDGES, S. ,~4 LONGO, D. L. (1985). Absence of the Lyt-2-, L3T4÷ lineage of T cells in mice treated neonatally with anti-l-A correlates with absence of intrathymic l-A-bearing antigen-presenting cell function. J. expl. Med. 161, 1029-1047. LEDERBERG, J. (1959). Genes and antibodies. Science 129, 1649-1653. LIGNON, M.-F., GALAS, M.-C., RODRIGUEZ, M., LAUR, J., AUMELAS, A. & MARTINEZ, J. (1987). A synthetic peptide derivative that is a cholecystokinin receptor antagonist. J. biol. Chem. 262, 7226-7231. LIMBIRD, L. (1986). Cell Surface Receptors: A Short Course on Theory and Methods. Boston: Martinus Nijhoff. Lo, D., RON, Y. & SPRENT, J. (1986). Induction of MHC-restricted specificity and tolerance in the thymus, lmmunol. Res. 5, 221-232. LO, D. & SPRENT, J. (1986). Identity of cells that imprint H-2-restricted T-cell specificity in the thymus. Nature, Lond. 319~ 672-675. LORENZ, R. G. & ALLEN, P. M. (1989). Thymic cortical epithelial cells can present self-antigens in oitro. Nature, Lond. 337, 560-562. LYNCH, D. H., GRESS, R. E., NEEDLEMAN, B. W., ROSENBERG, S. m. & HODES, R. J. (1985). T cell responses to Mls determinants are restricted by cross-reactive MHC determinants..I. Irnmunol. 134, 2071 - 2078. MACDONALD, H. R., HENGARTNER, H. & PEDRAZZINI, T. (1988a). Intrathymic deletion of selfreactive cells prevented by neonatal anti-CD4 antibody treatment. Nature, Lond. 335, t74-176. MACDONALD, H. R., LEES, R. K., SCHNEIDER, R., ZINKERNAGEL, R. M. & HENGARTNER, H. (1988b). Positive selection of CD4 ÷ thymocytes controlled by MHC class II gene products. Nature, Lond. 336, 471-473. MACDONALD, H. R., SCHNEIDER, R., LEES, R. K., HOWE, R. C., ACHA-ORBEA, H., FESTENSTEIN, n., ZINKERNAGEL, R. M. & HENGARTNER, H. (1988c). T-cell receptor V/3 use predicts reactivity and tolerance to Mlsa-encoded antigens. Nature, Lond. 332, 40-45. MANNING, M., OLMA, A., KLIS, W., KOLODZtEJCZYK, A., NAWROCKA, E., MISICKA, A., SETO, J. & SAWYER, W. H. (1984). Carboxy terminus of vasopressin required for activity but not binding. Nature, Lond. 308, 652-653. MARRACK, P., ENDRES, R., SHIMONKEVITZ, R., ZLOTNIK,m., DIALYNAS, D., FITCH, R. & KAPPLER, J. (1983)..L expl. Med. 158, 1077-1091. MARRACK, P. & KAPPLER, J. (1987). The T cell receptor. Science 238, 1073-1079. MARRACK, P. & KAPPLER, J. (1988). The T-cell repertoire for antigen and MHC. Immunol. Today 9, 308-315. MARRACK, P., KUSHNIR, E., BORN, W., McDuFFIE, M. & KAPPLER, J. (1988a). The development of helper T cell precursors in mouse thymus. J. lmmunol. 140, 2508-2513. MARRACK, P., LO, D., BRINSTER, R., PALMITER, R., BURKLY, L., FLAVELL, R. H., & KAPPLER, J. (1988b). The effect of thymus environment on T celt development and tolerance. Cell 53, 627-634. MARRACK, P., McCORMACK, J. & KAPPLER, J. (1989). Presentation of antigen, foreign major histocompatibility complex proteins and self by thymus cortical epithelium. Nature, Lond. 338, 503-505. MARTINEZ, J., MAGOUS, R., L1GNON, M.-F., LAUR, J., CASTRO, B. & BALI, J. P. (1984). Synthesis and biological activity of new peptide segments of gatrin exhibiting gastrin antagonist property. J. reed. Chem. 27, 1597-1601. MARUSIC-GALESIC, S., LONGO, D. L. & KRUISBEEK, A. M. (1989). Preferential differentiation of T cell receptor specificities based on the MHC glycoproteins encountered during development: evidence for positive selection. J. expl. Med. 169, 1619-1630.
MODEL OF T CELL SELECTION
MARUSIC-GALESIC, S., STEPHANY, D. A., LONGO, D. L. & KRUISBEEK, A. M. (1988). Development of CD4-CD8 + cytotoxic T cells requires interactions with class I MHC determinants. Nature, Lond. 333, 180-183. MATZlNGER, P. & GUERDER, S. (1989). Does T-cell tolerance require a dedicated antigen-presenting cell? Nature, Lond. 338, 74-78. MEUER, S. C., HODGDON, J. C., HUSSEY, R. E., PROTENTIS, J. P., SCHLOSSMAN, S. F. • REINHERZ, E. L. (1983). Antigen-like effects of monoclonal antibodies directed at receptors on human T cell clones. J. expL Med. 158, 988-993. NICKERSON, M. (1956). Receptor occupancy and tissue response. Nature, Lond. 178, 697-698. NORMENT, A. M., SALTER, R. D., PARHAM, P., ENGELHARD, V. H. & LITTMAN, D. R. (1988). Cell-cell adhesion mediated by CD8 and MHC class 1 molecules. Nature, Lond. 336, 79-81. PALASZYNSKI, E. W. (1987). Synthetic C-terminal peptide of IL-I functions as a binding domain as well as an antagonist for the IL-I receptor. Biochem. biophys. Res. Commun. 147, 204-211. PANDOL, S. J., DHARMSATHAPHORN, K., SCHOEFF1ELD, M. S., VALE, W. & RIVIER, J. (1986). Vasoactive intestinal peptide receptor antagonist [4C1-D-Phe6, Leu~7]VIP. Am. Z PhysioL 250, G553G557. PORTOGHESE, P. S., SULTANA, M., NAGASE, H. & TAKEMORI, A. E. (1988). Application of the message-address concept in the design of highly potent and selective non-peptide 6 opioid receptor antagonists. J. reed. Chem. 31, 281-282. PULLEN, A. M., MARRACK, P. & KAPPLER, J. W. (1988). The T-cell repertoire is heavily influenced by tolerance to polymorphic self-antigens. Nature, Lond. 335, 796-801. REEM, G. H., CARDING, S. & REINHERZ, E. L (1987). Lymphokine synthesis is induced in human thymocytes by activation of the CD2 (TI 1) pathway. J. Immunol. 139, 130-134. RIVIER, J., RIVIER, C. & VALE, W. (1984). Synthetic competitive antagonists of corticotropin-relasing factor: effect on ACTH secretion in the rat. Science 224, 889-891. ROCK, K. L. & BENACERRAE, B. (1984). Thymic T cells are driven to expand upon interaction with self-class II major histocompatibility complex gene products on accessory cells. Proc. ham. Acad. Sci. U.S.A. 81, 1221-1224. ROSENBLATT, M. (1986). Peptide hormone antagonists that are effective in vivo: lessons from parathyroid hormone. IV. EngL Z Med. 315, 1004-1013. SCHWARTZ, R. H. (1985). T-lymphocyte recognition of antigen in association with gene products of the major histocompatibility complex. Ann. Rev. lmmunoL 3, 237-261. SCHWARTZ, R. H. (1989). Acquisition of immunologic self-tolerance. Cell 57, 1073-1081. SCHWYZER, R. (1977). ACTH: a short introductory review. Ann. N Y Acad. Sci. 297, 3-26. SHA, W. C., NELSON, C. A., NEWBERRY, R. D., KRANZ, D. M., RUSSELL, J. H. &LOH, D. Y. (1988a). Positive and negative selection of an antigen receptor on T cells in transgenic mice. Nature, Lond. 336, 73-76. SHA, W. C., NELSON, C. A., NEWBERRY, R. D., KRANZ, D. M., RUSSELL, J. H. &LOH, D. Y. (1988b). Selective expression of an antigen receptor on CDS-bearing T lymphocytes in transgenic mice. Nature, Lond. 335, 271-274. SHEVACH, E. M. & ROSENTHAL, A. S. (1973). Function of macrophages in antigen recognition by guinea pig T lymphocytes: II. Role of the macrophage in the regulation of genetic control of the immune response. Z expl. Med. 138, 1213-1229. SINGER, A., HATHCOC~'K, K. S. & HODES, R. J. (1982). Self recognition in allogeneic thymic chimeras: self recognition by T helper cells from thymus-engrafted nude mice is restricted to the thymic H-2 haplotype. J, expL Med. 155, 339-344. SINGER, A., MlZUOCHI, T., MUNITZ, T. I. & GRESS, R. E. (1986). Role of self antigen in the selection of the developing T cell repertoire. Prog. ImmunoL VI, 60-66. SMITH, H., CHEN, I.-M., KUBO, R. & TUNG, K. S. K. (1989). Neonatal thymectomy results in a repertoire enriched in T cells deleted in adult thymus. Science 245, 749-752. SPRENT, J., LO, D., GAO, E.-K. & RON, Y. (1988). T cell selection in the thymus. ImmunoL Rev. 101, 173-189. STEPHENSON, R. P. (1956). A modification of receptor theory. Br. Z PharmacoL 11, 379-393. SWAIN, S. L. (1981). Significance of Lyt phenotypes: Lyt2 antibodies block activities of T cells that recognize class 1 major histocompatibility complex antigens regardless of their function. Proc. hath. Acad. Sci. U.S.A. 78, 7101-7105. TEH, H. S., KISIELOW, P., SCOTT, B., KISHI, H., UEMATSU, Y., BLUTHMANN, H. & VON BOEHMER, H. (1988). Thymic major histocompatibility complex antigens and the a g T-cell receptor determine the CD4/CD8 phenotype of T cells. Nature, Lond. 335, 229-233.
THOMAS,D. W. & SHEVACH,E. M. (1977). Nature of the antigenic complex recognized by Tlymphocytes: specific sensitization by antigens associated with allogeneic macrophages. Proc. hath. Acad. Sci. U.S.A. 74, 2104-2108. TOWNSEND, A., OHLEN, C., BASTIN, J., LJUNGGREN, H.-G., FOSTER, L. & KARRE, K. (1989). Association of class 1 major histocompatibility heavy and light chains induced by viral peptides. Nature, Lond. 340, 443-448. UNSON, C. G., ANDREU, D., GURZENDA, E. M. & MERRIFIELD, R. B. (1987). Synthetic peptide antagonists of glucagon. Proc. hath. Acad. Sci. U.S.A. 84, 4083-4087. VAVREK, R. J. & STEWART, J. M. (1985). Competitive antagonists of bradykinin. Peptides 6, 161-164. WHITE, J., HERMAN, A., PULLEN, A. M., KUBO, R., KAPPLER, J. W. & MARRACK, P. (1989). The V-specific superantigen staphylococcal enterotoxin B: stimulation of mature T cells and clonal deletion in neonatal mice. Cell 56, 27-35. WILDE, D. B., MARRACK, P., KAPPLER, J., DIALYNAS, D. P. & FITCH, F. W. (1983). Evidence implicating L3T4 in class I1 MHC antigen reactivity: monoclonal antibody GK1.5 (anti-L3T4a) blocks class I! MHC antigen-specific proliferation, release of lymphokines, and binding by cloned murine helper T lymphocyte lines. Z lmmunoL 131, 2178-2183. ZINKERNAGEL, R. M., CALLAHAN, G. N., ALTHAGE, A., COOPER, S., KLEIN, P. m. • KLEIN, J. (1978). On the thymus in the differentiation of "H-2 self-recognition" by T cells: evidence for dual recognition? 3'. expL IVied. 147, 882-896. ZINKERNAGEL, R. M. & DOHERTY, P. C. (1974). Restriction of in vitro T cell-mediated cytotoxicity in lymphocytic choriomeningitis within a syngeneic or semiallogeneic system. Nature, Lond. 248, 701-702. ZI NKERNAGEL, R. M. & DOHERTY, P. C. (1977). Major transplantation antigens, viruses, and specificity of surveillance T cells. In: Contemporary Topics in lmmunobiology Vol. 7 (Stutman, O., ed.) pp. 179-220. New York: Plenum Press. ZUNIGA-PFLUCKER, J. C., LONGO, D. L. & KRUISBEEK, A. M. (1989). Positive selection ofCD4-CD8 ÷ T cells in the thymus of normal mice. Nature, Lond. 338, 76-78.