Morphogenesis in shape-space. Elementary meta-dynamics in a model of the immune network.

A unique feature of the immune system is that it possesses meta-dynamics: the process governing the removal of certain clones from the active population and the recruitment of new clones from the pool of lymphocytes freshly produced by the bone marrow. In this paper, we present a computer model which focuses on those aspects of the system that characteristically derive from the meta-dynamics as such. We observe that when a region of shape-space is densely populated, there is an emergence of dynamically quasi-stable configurations. Moreover, when the system develops in the presence of permanent self-antigens, the latter are systematically incorporated into such coherent configurations. We conclude that the meta-dynamics of the biological immune system may be such that it gives rise to the emergence of a connected, self-sustaining network that we call the Central Immune System: a coherent self-identity which incorporates the molecules of the somatic self and, more generally, reflects the history of its own development.     

Central immune system,the self and autoimmunity

We model how auto-reactiveB cells are kept under control by an idiotypic network. Autoimmunity occurs when the control is broken by an infection or not achieved through an abnormal ontogenetic evolution. We describe the idiotypic network, viz., the central immune system, by idiotype-anti-idiotype pairs which are coupled to a set of highly connected clones, which interact with each clone of the network. Some clones of the central immune system recognize self-antigen. We find a huge variety of fixed points which can be classified as tolerant, autoimmune, and neutral states according to the concentration of the auto-reactive antibody. Most significant are auto-reactive clones which are a member of an idiotype-anti-idiotype pair. In a healthy individual, an autoimmune disease is induced by an antigen infection which triggers a transition from a tolerant to an autoimmune state. Autoimmunity is induced more readily by an antigen coupling to theanti-idiotype than by one interacting with the auto-reactive clone itself. We indicate a possible therapy which simply reverses the processes that have lead to the autoimmune disease. In the early development of the central immune system its highly connected, core part serves to draw the more specific clones of idiotype-anti-idiotype pairs into the network. In order to avoid autoimmunity in ontogenetic evolution the anti-idiotype of an auto-reactive clone must be formed in advance by a sufficiently long period of time. Thus, a well ordered succession of the appearance of the more specific clones is required.     

Capacity of a model immune network

The capacity of a model immune network in terms of the number of different antigens that can be vaccinated against without any memory lost is computed and tested by numerical simulations. We also investigate memory loss and failure to vaccinate due to overcrowding the network with too many antigens. The computations are done for two different strategies for proliferation, one implying all the antigen specific clones and the second one being more thrifty.     

Pattern formation in one- and two-dimensional shape-space models of the immune system.

A large-scale model of the immune network is analyzed, using the shape-space formalism. In this formalism, it is assumed that the immunoglobulin receptors on B cells can be characterized by their unique portions, or idiotypes, that have shapes that can be represented in a space of a small finite dimension. Two receptors are assumed to interact to the extent that the shapes of their idiotypes are complementary. This is modeled by assuming that shapes interact maximally whenever their coordinates in the space-space are equal and opposite, and that the strength of interaction falls off for less complementary shapes in a manner described by a Gaussian function of the Euclidean "distance" between the pair of interacting shapes. The degree of stimulation of a cell when confronted with complementary idiotypes is modeled using a log bell-shaped interaction function. This leads to three possible equilibrium states for each clone: a virgin, an immune, and a suppressed state. The stability properties of the three possible homogeneous steady states of the network are examined. For the parameters chosen, the homogeneous virgin state is stable to both uniform and sinusoidal perturbations of small amplitude. A sufficiently large perturbation will, however, destabilize the virgin state and lead to an immune reaction. Thus, the virgin system is both stable and responsive to perturbations. The homogeneous immune state is unstable to both uniform and sinusoidal perturbations, whereas the homogeneous suppressed state is stable to uniform, but unstable to sinusoidal, perturbations. The non-uniform patterns that arise from perturbations of the homogeneous states are examined numerically. These patterns represent the actual immune repertoire of an animal, according to the present model. The effect of varying the standard deviation sigma of the Gaussian is numerically analyzed in a one-dimensional model. If sigma is large compared to the size of the shape-space, the system attains a fixed non-uniform equilibrium. Conversely if sigma is small, the system attains one out of many possible non-uniform equilibria, with the final pattern depending on the initial conditions. This demonstrates the plasticity of the immune repertoire in this shape-space model. We describe how the repertoire organizes itself into large clusters of clones having similar behavior. These results are extended by analyzing pattern formation in a two-dimensional (2-D) shape-space.(ABSTRACT TRUNCATED AT 400 WORDS)     

Memory but no suppression in low-dimensional symmetric idiotypic networks

We present a new symmetric model of the idiotypic immune network. The model specifies clones of B-lymphocytes and incorporates: (1) influx and decay of cells; (2) symmetric stimulatory and inhibitory idiotypic interactions; (3) an explicit affinity parameter (matrix); (4) external (i.e. non-idiotypic) antigens. Suppression is the dominant interaction, i.e. strong idiotypic interactions are always suppressive. This precludes reciprocal stimulation of large clones and thus infinite proliferation. Idiotypic interactions first evoke proliferation, this enlarges the clones, and may in turn evoke suppression. We investigate the effect of idiotypic interactions on normal proliferative immune responses to antigens (e.g. viruses). A 2-D, i.e. two clone, network has a maximum of three stable equilibria: the virgin state and two asymmetric immune states. The immune states only exist if the affinity of the idiotypic interaction is high enough. Stimulation with antigen leads to a switch from the virgin state to the corresponding immune state. The network therefore remembers antigens, i.e. it accounts for immunity/memory by switching beteen multiple stable states. 3-D systems have, depending on the affinities, 9 qualitatively different states. Most of these also account for memory by state switching. Our idiotypic network however fails to account for the control of proliferation, e.g. suppression of excessive proliferation. In symmetric networks, the proliferating clones suppress their anti-idiotypic suppressors long before the latter can suppress the former. The absence of proliferation control violates the general assumption that idiotypic interactions play an important role in immune regulation. We therefore test the robustness of these results by abandoning our assumption that proliferation occurs before suppression. We thus define an “escape from suppression” model, i.e. in the “virgin” state idiotypic interactions are now suppressive. This system erratically accounts for memory and never for suppression. We conclude that our “absence of suppression from idiotypic interactions” does not hinge upon our “proliferation before suppression” assumption.     

Thymectomy-induced deterioration of learning and memory.

Age-associated immunodeficiency and cognitive deterioration are two predominant features of the aging process, but the mutual influences between them are not clear yet. Research on the neuroendocrine immunomodulation (NIM) network indicate reciprocal interactions between the neuroendocrine and the immune systems mediated by neurotransmitters, neuropeptides, hormones and cytokines, which form an integrated network to maintain normal physiological functions of the body. An imbalance in the NIM network is believed to accelerate the aging process, in which the thymus plays an important role. We recently discovered that thymectomy in mice not only reduces the immune response, but also deteriorates learning performances. Cytokines such as interleukin-1, interleukin-6 and tumor necrosis factor, and corticosterone affect the induction of hippocampal long-term potentiation, a synaptic model of memory. Clinical studies have demonstrated that Alzheimer's patients show disordered immune function in addition to cognitive deficit, and the brain lesions of Alzheimer's patients may be associated with abnormal immune reactions occurring in the brain. With these findings, it is speculated that the disordered immune function may induce an imbalance in the NIM network, which consequently influences central cognitive function.     

Idiotypic networks incorporating T-B cell co-operation. The conditions for percolation

Previous work was concerned with symmetric immune networks of idiotypic interactions amongst B cell clones. The behaviour of these networks was contrary to expectations. This was caused by an extensive percolation of idiotypic signals. Idiotypic activation was thus expected to affect almost all (greater than 10(7] B cell clones. We here analyse whether the incorporation of helper T cells (Th) into these B cell models could cause a reduction in the percolation. Empirical work on idiotypic interactions between Th and B cells however, would suggest that two different idiotypic Th models should be developed: (1) a Th which recognises native B cell idiotypes, i.e. a non-MHC-restricted "ThId" model, and (2) a "classical" MHC-restricted helper T cell model. In the ThId model, the Th-B cell interaction is symmetric. A 2-D model of a Th and a B cell clone that interact idiotypically with each other accounts for various equilibria (i.e. one virgin and two immune states). Introduction of antigen does indeed lead to a state switch from the virgin to the immune state; such a system is thus able to "remember" its exposure to antigen. Idiotypic signals do however, percolate in ThId models via these "B-Th-B-Th" pathways: proliferating Th and B cell clones that interact idiotypically, will always activate each other reciprocally. In the MHC-restricted Th model, Th-B interactions are asymmetric. Because the B cell idiotypes are processed and subsequently presented by MHC molecules, the Th receptor and the native B cell receptor are not expected to be complementary. Thus the Th and the B cells are unable to activate each other reciprocally, and a 2-D Th-B cell model cannot account for idiotypic memory. In contrast to the ThId model, idiotypic activation cannot percolate via "B-Th-B-Th" interactions. Due to the assymmetry idiotypic activation stops at the first Th level. A Th clone cannot activate a subsequent B cell clone: if the B cells recognise the Th cells, they see idiotype but get no help; if the Th cells see the B cells, the B cells are helped but see no idiotype. The percolation along "B-B-B" pathways in these two models is next analysed. Two B cells clones, each helped by one Th clone, are connected by a symmetric idiotypic interaction. It turns out that in both models the second (i.e. anti-idiotypic) B cells (B2) never proliferate.(ABSTRACT TRUNCATED AT 400 WORDS)     

Staphylococcal enterotoxin-mediated human T-T cell interactions.

Staphylococcal enterotoxins (SE) are known to stimulate a large proportion of T cells. SE bind to MHC-class II molecules on APC and a particular segment of certain TCR V beta and V gamma gene products. Resting human T cells do not express HLA class II Ag and therefore cannot present SE to T cells. Activated human T cells, however, do express HLA-DR, -DP, and -DQ Ag and could consequently serve as APC for SE. As such, local immune responses to SE might be regulated and/or abrogated by SE-mediated T-T cell interactions leading to T cell destruction. We have investigated if such SE-mediated T-T cell interactions can occur in vitro using human cytolytic TCR-alpha beta+ and TCR-gamma delta+ T cell clones. We demonstrate that the TCR-alpha beta+ T cell clones can efficiently present staphylococcal enterotoxin A (SEA) to each other: T cell clones coated with SEA are lysed by SEA-reactive T cell clones but not by a SEA-nonreactive T cell clone. Furthermore, the SEA-reactive TCR-alpha beta+ clones (but not the SEA-nonreactive clone) destruct themselves in the presence of SEA at low concentrations of SEA (less than 0.01 microgram/ml). Also, SEA-coated T cell clones can induce proliferative responses although such responses are much weaker than those induced when B cells are used as stimulator cells. In contrast, the SEA-reactive TCR-gamma delta+ T cell clones are resistant to autokilling in the presence of SEA and they do not lyse SEA-coated TCR-gamma delta+ targets. However, such targets can be lysed by TCR-alpha beta+ effector cells. These results indicate that TCR-gamma delta+ cells are relatively resistant to lysis and that during local nonspecific immune responses triggered by SE, which induces HLA-class II expression by the responding T cells, SE-mediated T-T cell interactions may play a role in the regulation and/or abrogation of these immune responses.     

The race between initial T-helper expansion and virus growth upon HIV infection influences polyclonality of the response and viral set-point

Infection with HIV is characterized by very diverse disease-progression patterns across patients, associated with a wide variation in viral set-points. Progression is a multifactorial process, but an important role has been attributed to the HIV-specific T-cell response. To explore the conditions under which different set-points may be explained by differences in initial CD4 and CD8 T-cell responses and virus inoculum, we have formulated a model assuming that HIV-specific CD4 cells are both targets for infection and mediators of a monoclonal or polyclonal immune response. Clones differ in functional avidity for HIV epitopes. Importantly, in contrast to previous models, in this model we obtained coexistence of multiple clones at steady-state viral set-point, as seen in HIV infection. We found that, for certain parameter conditions, multiple steady states are possible: with few initial CD4 helper cells and high virus inoculum, no immune response is established and target-cell-limited infection follows, with associated high viral load; when CD4 clones are initially large and virus inoculum is low, infection can be controlled by several clones. The conditions for the dependence of viral set-point on initial inoculum and CD4 T-helper clone availability are investigated in terms of the effector mechanism of the clones involved.     

Unreasonable implications of reasonable idiotypic network assumptions

We first analyse a simple symmetric model of the idiotypic network. In the model idiotypic interactions regulate B cell proliferation. Three non-idiotypic processes are incorporated: (1) influx of newborn cells; (2) turnover of cells: (3) antigen. Antigen also regulates proliferation. A model of 2 B cell populations has 3 stable equilibria: one virgin, two immune. The twodimensional system thus remembers antigens, i.e. accounts for immunity. By contrast, if an idiotypic clone proliferates (in response to antigen), its anti-idiotypic partner is unable to control this. Symmetric idiotypic networks thus fail to account for proliferation regulation. In high-D networks we run into two problems. Firstly, if the network accounts for memory, idiotypic activation always propagates very deeply into the network. This is very unrealistic, but is an implication of the “realistic” assumption that it should be easier to activate all cells of a small virgin clone than to maintain the activation of all cells of a large (immune) clone. Secondly, graph theory teaches us that if the (random) network connectance exceeds a threshold level of one interaction per clone, most clones are interconnected. We show that this theory is also applicable to immune networks based on complementary matching idiotypes. The combination of the first “percolation” result with the “interconnectancr” result means that the first stimulation of the network with antigen should eventually affect most of the clones. We think this is unreasonable. Another threshold property of the network connectivity is the existence of a virgin state. A gradual increase in network connectance eliminates the virgin state and thus causes an abrupt change in network behaviour. In contrast to weakly connected systems, highly connected networks display autonomous activity and are unresponsive to external antigens. Similar differences between neonatal and adult networks have been described by experimentalists. The robustness of these results is tested with a network in which idiotypic inactivation of a clone occurs more generally than activation. Such “long-range inhibition” is known to promote pattern formation. However, in our model it fails to reduce the percolation, and additionally, generates semi-chaotic behaviour. In our network, the inhibition of a clone that is inhibiting can alter this clone into a clone that is activating. Hence “long-range inhibition” implies “long-range activation”, and idiotypic activation fails to remain localized. We next complicate this model by incorporating antibody production. Although this “antibody” model statically accounts for the same set of equilibrium points, it dynamically fails to account for state switching (i.e. memory). The switching behaviour is disturbed by the autonomous slow decay of the (long-lived) antibodies. After antigenic triggering the system now performs complex cyclic behaviour. Finally, it is suggested that (idiotypic) formation of antibody complexes can play only a secondary role in the network. In conclusion, our results cast doubt on the functional role of a profound idiotypic network. The network fails to account for proliferation regulation, and if it accounts for memory phenomena, it “explodes” upon the first encounter with antigen due to extensive percolation.     

Hierarchy of cytotoxic T cell clones. I. Reversal of resistance to lysis and killing between anti-hapten cytotoxic T cells

Cells from clones of anti-hapten cytotoxic T lymphocytes (CTL) can act as both effector cells and, when treated with the specific hapten, as target cells. Individual clones can kill haptenated cells only from other clones that are less efficient killers. Clones specific for both fluorescein and trinitrophenol could be ordered in a single hierarchy in which resistance to lysis correlated with lytic efficiency. When the killing efficiency was reduced with phorbol myristate acetate (PMA) or the colchicine analogue, Colcemid, the degree of resistance to lysis was also reduced. The use of PMA-treated fluoresceinated targets greatly enhanced intraclonal killing and similarly lead to a repositioning of clones within the hierarchy of normal cells. By the haptenation of appropriate clones, efficient CTL could kill cells from other clones in a direction apparently opposite to recognition. The results demonstrate that effects other than antigen recognition of the target cell may result in variations in the nature of T cell immune responses.     

Immune networks modeled by replicator equations

In order to evaluate the role of idiotypic networks in the operation of the immune system a number of mathematical models have been formulated. Here we examine a class of B-cell models in which cell proliferation is governed by a non-negative, unimodal, symmetric response function f(h), where the field h summarizes the effect of the network on a single clone. We show that by transforming into relative concentrations, the B-cell network equations can be brought into a form that closely resembles the replicator equation. We then show that when the total number of clones in a network is conserved, the dynamics of the network can be represented by the dynamics of a replicator equation. The number of equilibria and their stability are then characterized using methods developed for the study of second-order replicator equations. Analogies with standard Lotka-Volterra equations are also indicated. A particularly interesting result of our analysis is the fact that even though the immune network equations are not second-order, the number and stability of their equilibria can be obtained by a superposition of second-order replicator systems. As a consequence, the problem of finding all of the equilibrium points of the nonlinear network equations can be reduced to solving linear equations.     

The whole-organism heavy chain B cell repertoire from Zebrafish self-organizes into distinct network features

Background  

The adaptive immune system is based on selected populations of molecularly distinct individual B and T cell clones. However, it has not been possible to characterize these clones in a comprehensive and informatics manner to date; attempts have been limited by the number of cells in the adaptive immune system and an inability to quantify them. Recently, using the Zebrafish (ZF) Danio rerio as a model organism and parallel sequencing as the quantifying technology, Weinstein et al. overcame this major hurdle and quantified the entire heavy chain B-cell repertoire in ZF. Here, we present a novel network analysis of the data from the Weinstein group, providing new insights into the network structure of the B-cell repertoire.     

Analysis of the state of differentiation of two rat colon carcinoma clones with distinct tumorigenic properties.

The presence of some characteristics of normal rat intestinal epithelial cells was studied on two clones originating from a single rat colon carcinoma. These clones differed by their tumorigenic properties in the syngeneic host. However, they grow similarly in vitro and in immuno-deprived animals. The PROb clone which had the ability to form progressive tumors in the syngeneic host appeared to possess more features of differentiated cells than the REGb clone which was immunologically rejected by syngeneic hosts. Indeed, the morphology of the cells was different, the REBb cells having a more fibroblastic appearance while the PROb cells had the capacity to form domes characterizing the functional polarization of the cell layer. The two clones could also be distinguished by their expression of proteins of intermediate filaments. Both expressed cytokeratins showing their epithelial origin, but only REGb cells displayed vimentin which is characteristic of mesenchymal or poorly differentiated epithelial cells. Furthermore, analysis of the expression of a series of glycoconjugate tissue antigens and an unknown protein (p120) showed that the PROb cells resembled more the normal adult digestive epithelium than the REGb cells did. In conclusion, it appears that in this model, the most aggressive cells, those resisting to the constraints imposed by the immune system, are also the more differentiated ones.     

The network of hemopoietic regulatory proteins in myeloid cell differentiation.

There are clones of myeloid leukemic cells that can be induced to undergo terminal cell differentiation to macrophages by normal hemopoietic regulatory proteins. Induction of differentiation in two different clones of myeloid leukemic cells with interleukin 6 (IL-6) or granulocyte-macrophage colony-stimulating factor (GM-CSF) resulted in induction of mRNA for the hemopoietic regulatory proteins IL-6, GM-CSF, interleukin 1 alpha and interleukin 1 beta, tumor necrosis factor, and transforming growth factor beta 1. In one of these clones, induction of differentiation with GM-CSF was also associated with induction of mRNA for macrophage colony-stimulating factor (M-CSF) but not for the receptor for M-CSF (c-fms), whereas in the other clone, induction of differentiation with IL-6 was associated with induction of mRNA for both c-fms and M-CSF. The clones also differed in their responsiveness to these regulators. There was no induction of mRNA for granulocyte colony-stimulating factor or interleukin 3 during differentiation of either clone. The results indicate that the genes for a nearly normal network of positive and negative hemopoietic regulatory proteins are induced during differentiation of these myeloid leukemic cells and that there are leukemic clones with specific defects in this network.     

Genetic mechanisms for dominant VH gene expression. The VHB512 gene.

A total of 37 mAb with reactivity for dextran B512 have been studied; 30 of them were products of independent rearrangements and 21 made use of the same VH gene, the VHB512 gene. These results unambiguously established that the immune response to dextran in the high responder mouse strain C57BL/6 was restricted. Idiotypic determinants are located all over the Ig V region. Many but not all Id described so far can be ascribed to protein structures encoded by VH or VL gene segments. The expression of the major Id, 17-9 Id, in C57BL/6 was not absolutely correlated with the expression of the dominant VHB512 gene in the same mouse strain. Inspection of amino acid sequences of the CDR3 of idiotypic positive and negative clones suggested that idiotypic structures may be associated with the expression of Tyr at position 95 and Phe or Leu at position 96 in the H and L chains, respectively. Therefore the indiscriminate use of idiotypic markers to characterize VH genes and the relevance of idiotypic regulation in VH gene expression are questioned. Id-positive and Id-negative clones displayed similar affinity values for dextran, indicating that idiotypic and binding structures were probably separated. The exchange of Asp65 for Gly65 in one of the clones reduced affinity for dextran, suggesting the involvement of CDR2 in dextran binding. The dominant expression of VH genes can be explained by somatic and/or genetic mechanisms. Because somatic mechanisms such as idiotypic regulation or selection based on affinity for dextran did not seem to influence the expression of the VHB512 gene we favor a genetic alternative. We discuss a model based on the distance between VH genes and D and JH elements. This model is compatible with somatic and genetic regulation in other systems and provides a new theoretical approach to the understanding of immune VH dominance and low responsiveness.     

The analysis of clonal expansions in normal and autoimmune B cell repertoires

Clones are the fundamental building blocks of immune repertoires. The number of different clones relates to the diversity of the repertoire, whereas their size and sequence diversity are linked to selective pressures. Selective pressures act both between clones and within different sequence variants of a clone. Understanding how clonal selection shapes the immune repertoire is one of the most basic questions in all of immunology. But how are individual clones defined? Here we discuss different approaches for defining clones, starting with how antibodies are diversified during different stages of B cell development. Next, we discuss how clones are defined using different experimental methods. We focus on high-throughput sequencing datasets, and the computational challenges and opportunities that these data have for mining the antibody repertoire landscape. We discuss methods that visualize sequence variants within the same clone and allow us to consider collections of shared mutations to determine which sequences share a common ancestry. Finally, we comment on features of frequently encountered expanded B cell clones that may be of particular interest in the setting of autoimmunity and other chronic conditions.     

Role of iron in T cell activation: TH1 clones differ from TH2 clones in their sensitivity to inhibition of DNA synthesis caused by IgG Mabs against the transferrin receptor and the iron chelator deferoxamine.

TH1 and TH2 helper T cell clones have been studied with respect to their sensitivity to inhibition of DNA synthesis by an IgG anti-transferrin receptor antibody (ATRA), the iron chelator deferoxamine, and the combination of the two reagents. TH1 clones are very sensitive to ATRA-mediated inhibition of DNA synthesis while TH2 clones are very resistant, but both TH1 and TH2 clones show significant down-modulation of surface transferrin receptors after ATRA exposure. TH2 clones exhibit larger chelatable iron storage pools than TH1 clones, however, and even partial chelation of TH2 cell storage iron does not fully convert a TH2 clone to the ATRA sensitivity pattern of a TH1 clone. It is therefore proposed that the greater resistance of TH2 clones to ATRA mediated inhibition derives from the combined effects of larger and less labile iron storage pools. These studies provide novel evidence indicating that nonuniform iron metabolism can exist within the T cell compartment and thus raise questions as to why such differences exist and how they can be integrated into models of the T cell activation process. These studies also suggest that the cell-mediated immune response in vivo, which is known to be sensitive to iron deficiency, may be evoked by effector cells which resemble TH1 clones insofar as iron metabolism is concerned.     

Breaking T cell tolerance with foreign and self co-immunogens. A study of autoimmune B and T cell epitopes of cytochrome c.

The initiation of autoimmune B cell and T cell responses by self Ag or by foreign pathogens (molecular mimics) is not well understood. In the present study, cytochrome c (cyt c) was used as a model autoantigen to investigate how self-proteins are involved in the priming of autoimmune T cell responses. Immunization with foreign cyt c has been extensively analyzed in previous studies as a model for both humoral and cellular immune responses. Mice do not, however, make antibody or T cell responses to immunization with self (mouse) cyt c. In addition, T cell tolerance can be broken by autoreactive B cells that are readily elicited by immunization with cross-reactive foreign cyt c. These immune B cells presumably bind self cyt c and process and present the self Ag to stimulate an autoreactive T cell response. Autoreactive T cell clones derived by this mechanism are all specific for determinants within amino acids 1-80 of the cyt c protein presented by I-Ek. No T cell responses were observed to the carboxyl terminal 81-104 fragment that dominates the response to foreign cyt c. All clones derived in this study are stimulated by a polypeptide encompassing amino acids 54-68 and utilized the V beta 8.2 TCR gene. In contrast, T cells stimulated by foreign cyt c did indeed respond to fragment 81-104 and appear to utilize alternate TCR genes. Our data demonstrate that B cells specific for linear determinants distributed along the entire length of the foreign cyt c molecule can provide the stimulus required for breaking T cell tolerance to self cyt c. The applications of this work to understanding the mechanisms of autoimmune disease are discussed.