Autoimmune CD4+ T Cell Memory: Lifelong Persistence of Encephalitogenic T Cell Clones in Healthy Immune Repertoires 1
http://www.100md.com
免疫学杂志 2005年第13期
Abstract
We embedded green fluorescent CD4+ T cells specific for myelin basic protein (MBP) (TMBP-GFP cells) in the immune system of syngeneic neonatal rats. These cells persisted in the animals for the entire observation period spanning >2 years without affecting the health of the hosts. They maintained a memory phenotype with low levels of L-selectin and CD45RC, but high CD44. Although persisting in low numbers (0.01–0.1% of lymph node cells) they were sufficient to raise susceptibility toward clinical autoimmune disease. Immunization with MBP in IFA induced CNS inflammation and overt clinical disease in animals carrying neonatally transferred TMBP-GFP cells, but not in controls. The onset of the clinical disease coincided with mass infiltration of TMBP-GFP cells into the CNS. In the periphery, following the amplification phase a rapid contraction of the T cell population was observed. However, elevated numbers of fully reactive TMBP-GFP cells remained in the peripheral immune system after acute experimental autoimmune encephalomyelitis mediating reimmunization-induced disease relapses.
Introduction
Human autoimmune diseases, such as multiple sclerosis and rheumatoid arthritis, commonly take chronic, disabling courses. In some patients disease progresses steadily, while in others, relapses alter with remissions (1). According to present understanding, disease is caused and controlled by autoimmune T cells, which attack their target tissues, and thus determine activity and clinical character (2, 3). However, the nature of the autoimmune T cells over time remains largely unclear. In multiple sclerosis, for example, it is important to know whether different clones are recruited and activated from relapse to relapse. Newly recruited T cells could recognize distinct autoantigenic epitopes on the same autoantigen, or even be specific for different autoantigens, termed intra- and intermolecular determinant spreading, respectively (4, 5, 6, 7). As a less popular alternative, sequential disease episodes could be driven by members of the initial pathogenic clone(s) throughout the entire course of disease. Such a pathogenesis would invoke formation and persistence of autoimmune memory T cells (8).
Mechanisms of disease chronicity are difficult to study in model systems. This holds especially true for models induced by autoimmunization, because these either result in monophasic disease episodes with full and definite recovery, or, depending on large-scale irreversible tissue damage, take a chronic course. In this study, we established a model system which enabled us to study the fate of individual bona fide autoimmune memory cell clones in the Lewis rat. We transferred myelin basic protein (MBP) 3-reactive CD4+ T cells retrovirally engineered to express the gene of GFP (9, 10) into neonatal recipient animals. Using this technique we tracked and functionally characterized defined (autoreactive) memory T cell populations in vivo over time. The T cells integrated into the immune repertoire of immune-competent recipients leaving the endogenous immune system unaffected. They conferred T and B cellular reactivity to the recipient animals including specific T cell proliferation, proinflammatory cytokine release, and accelerated and increased autoantibody production. Importantly, the persisting autoreactive memory T cells raised the susceptibility toward autoimmune CNS inflammation and disease. After recovery from acute disease, they remained functionally fully intact in the peripheral immune system and, upon activation, evoked repeated disease episodes.
Thus, the data point toward long-term persistence of an autoimmunologic T cell memory as a critical factor in chronic autoimmune disease.
Materials and Methods
Animals, CD4+ TGFP cells, and neonatal transfer of TGFP cells
Lewis rats were obtained from the animal breeding facilities of the Max-Planck-Institute for Neurobiology (Martinsried, Germany) and were kept under standardized conditions. Ag-specific T cell clones used in the study were specific for guinea pig MBP and OVA. MBP was purified from guinea pig brains as described (11). Hen egg OVA was obtained from Sigma-Aldrich. All animal experiments were performed with the license of the Regierung von Oberbayern (No. 209.1/211-2531-56/99). CD4+ TGFP cells have been generated and tested for their phenotype, cytokine profile, and Ag specificity as described before (9). The lines consisted of oligoclonal CD4+ TCR+ Th1 cells and were highly specific for their respective Ags (see Table I, see Fig. 7D). Intraperitoneal transfer of TGFP cells (2 x 106 cells, in 0.5 ml/animal in Eagle’s HEPES medium) into newborns was performed under hypothermia within 48 h after birth. After T cell transfer, the newborn rats were kept under a 30°C humid atmosphere until fully recovered, then returned to their mother. Four different TMBP-GFP and three different TOVA-GFP cell lines were used for neonatal transfer. The activation state of the T cells did not influence the capacity of T cells to become memory cells. In most of the experiments, the T cells were transferred 4–5 days after restimulation with Ag.
Cell isolation, cytofluorometry, and FACS
Single-cell suspensions from organs were obtained as described previously (9, 10, 12). For cytofluorometric analysis, FACSCalibur operated by CellQuest software (BD Biosciences) was used (9, 10). The following mAbs were used for surface membrane analysis: W3/25 (anti-CD4; Serotec), R73 (TCR), OX-6 (rat MHC class II), OX-40 Ag (CD134), OX-39 (CD25, IL-2R chain), OX-22 (CD45RC), H-CAM (CD44), L-selectin (CD62L), V 8.2, 8.5, 10, 16 (all BD Biosciences). RPE-Cy-5-labeled anti-mouse Ab (DAKO) was used as secondary Ab.
Induction of adoptive transfer of experimental autoimmune encephalomyelitis (tEAE) and animal preparation
Adoptive transfer of the encephalitogenic T cell lines was performed by i.v. injection. The dose of T cells injected was adjusted to 5 x 106 T cell blasts/animal. Animals were monitored daily by measuring weight and examining disease scores (0, no disease; 1, flaccid tail; 2, gait disturbance; 3, complete hind limb paralysis; 4, tetraparesis; and 5, death).
Organ distribution of neonatally transferred TMBP-GFP and TOVA-GFP cells
Numbers of persisting TMBP-GFP following neonatal transfer were evaluated cytofluorometrically. The organs were carefully dissected and the relative numbers of GFP per organ cells were measured.
Histology
Histological analysis was performed as described (10). Briefly, after perfusion with 4% paraformaldehyde (PFA), lymph nodes (LNs) were prepared and postfixed in 4% PFA overnight. The frozen organs were cut into 20-μm sections. Following fixation for 30 min at room temperature with 4% PFA the sections were incubated with OX-33 (CD45RA, B cell marker, dilution 1/200; BD Biosciences) overnight at 4°C. Cy3-labeled anti-mouse antiserum (Dianova) was used as secondary Ab. The sections were counterstained with biotinylated anti-TCR Ab (R73, dilution 1/300), detected by Cy5-labeled streptavidin (Molecular Probes). Fluorescence analysis was performed with confocal laser-scanning microscopy (Leica). Immunohistochemical staining for GFP and analysis with a Zeiss EM10 transmission electron microscope was performed as described (13). For quantification of GFP+ cells, organs were treated as described (14). The inflammatory index in spinal cord lesions was determined after quantification of T cells (W3/13, dilution 1/50; Harlan Seralab) and monocytes/macrophages (ED1, dilution 1/1000; Serotec) on five randomly selected complete spinal cord cross-sections per animal. The section area was determined using a morphometrical grid. The values represent means (±SD) of two individual animals per group.
Ex vivo reactivity assays
LN cells were cultured in 96-well plates (in DMEM 1% rat serum) without Ag, with ConA (2.5 μg/ml), or in the presence of specific Ag (10 μg/ml MBP or OVA, respectively), or with the irrelevant Ag purified protein derivative (10 μg/ml), respectively. [3H]dT (2 Ci/mmol; Amersham-Buchler) was added to the cultures after 48 h. The radioactive label present in the different cultures was determined as described (10).
Amplification of ex vivo-isolated GFP+ memory T cells was measured by cytofluorometry. Their numbers were determined in relation to a known absolute number of added PE-labeled plastic beads (BD Biosciences). The amplification rate was calculated in relation to the GFP+ T cell numbers at day 0.
Intracellular IFN- staining, IFN- ELISPOT
Intracellular IFN- staining was performed with anti-mouse/rat IFN- Ab (clone DB-1; BD Biosciences) as described (15). Control IgG (mouse IgG MOPC31) was obtained from Sigma-Aldrich. IFN--ELISPOT analyses were performed as described (15) using polyclonal goat anti-rat-IFN- and biotinylated goat anti-rat-IFN- antiserum (R&D Systems), and an automated imaging system equipped with appropriate computer software (KS ELISPOT Automated Image Analysis System; Zeiss).
Quantitative PCR
TaqMan analysis was performed as reported (10) using the ABI Prim 7700 Sequence Detector TaqMan (Applied Biosystems). For quantification of cytokine mRNAs, the expression of the cytokine mRNA was set in relation to a housekeeping gene (-actin). All PCR data were obtained by two independent measurements. The cycle threshold (CT) value of the measurements did not differ >0.5 amplification cycles.
Ab measurements
ELISA. Serial blood samples were obtained from tails of the rats. After clotting at 4°C, the sera were collected by centrifugation and stored at –20°C. ELISA was performed according to the protocol of Stefferl et al. (16). The rat sera were diluted 1/1000. Mouse mAbs specific for rat Ig and isotypes IgM, IgG1, IgG2a, IgG2b, and IgE were obtained from Serotec. Secondary goat anti-mouse peroxidase conjugate was purchased from Dianova (dilution 1/8000). O-phenylenediamine dihydrochloride (Sigma-Aldrich) was used as substrate. The reaction was stopped with 3 M HCl, and OD was determined at 490 nm.
ELISPOT. Plasma cells were determined by reversing the present sandwich method (17). Briefly, ELISPOT plates (MAHA N45; Millipore) were coated with goat anti-rat IgG (H+L) Ab (10 μg/ml, 100 μl/well in carbonate buffer, pH 9.3; Jackson ImmunoResearch Laboratories). After blocking (5% BSA-PBS), cells were added in a serial dilution (100 μl/well) in triplicates and incubated for 24 h. After removal of cells, plates were incubated for 2 h with biotinylated Ag (MBP or OVA, respectively), or biotinylated goat anti-rat IgG (H+L) Ab (0.5 μg/ml, 100 μl/well in 0.5% BSA-PBS; Jackson ImmunoResearch Laboratories). Spots onto the ELISPOT plates were visualized using streptavidin-alkaline phosphatase with 5-bromo-4-chloro-3-indolyl phosphate/NBT as substrate (Sigma-Aldrich), and analyzed with an automated ELISPOT reader system (KS ELISPOT; Zeiss, Jena, Germany). Total Ig-spots were considered total numbers of plasma cells, MBP/OVA-spots represented MBP/OVA-specific plasma cells. No second biotinylated detector was used as negative control.
Results
Introduction of GFP-expressing CD4+ T cells into the immune repertoire by neonatal transfer
GFP-transduced CD4+ T cells (9) specific for MBP or OVA (OVA, TMBP-GFP, or TOVA-GFP, respectively), were injected i.p. into syngeneic Lewis rat pups (2 x 106 cells, within 48 h of life). Both TOVA-GFP and TMBP-GFP cells consisted of highly specific oligoclonal CD4+ T cell populations which displayed a Th1-like cytokine profile with high levels of IFN- and TNF-, but no IL-4 (Table I) (10). TMBP-GFP cells transferred clinical disease to adult animals, which in course and severity was indistinguishable from EAE transferred by their non-manipulated counterparts. However, neither of the MBP-specific T cells attacked newborn recipients (18, 19).
We traced the transferred, Ag-experienced T cells in the recipients over periods of >2 years. The cells settled throughout the immune system including LNs, spleen, thymus, and bone marrow (Fig. 1A), and they were also found in some non-immune organs including lung, liver, and gut. Importantly, however, they never spontaneously invaded the CNS (Fig. 1A). In LNs, the T cells populated the paracortical T cell areas and the medulla of the LNs. A few scattered green T cells were seen within lymph follicles (Fig. 1B). In the spleen, the persisting GFP+ T cells sat both in the periarteriolar sheaths of the white pulp and in the red pulp (data not shown). Within the thymus, most, if not all of the cells were located in the medulla (data not shown). The persisting T cells (Tmemory-MBP cells) did not demonstrably influence the host’s immune cell composition and immune competence (data not shown).
We previously showed that in tEAE, most of the transferred effector T cells are lost both from the CNS and from the peripheral immune system (10). We now observed that Tmemory-MBP cells behave very differently. Although they were effectively cleared from the CNS following an acute EAE bout (Fig. 4C), they persisted in the periphery even in elevated numbers (Fig. 5A). Furthermore, these cells remained functionally intact. Reimmunization of memory rats with MBP/IFA 3 mo after a first EAE attack led again to clinical disease starting 7 days p.i., and this second EAE episode became more severe (weight loss, clinical score 2–3: hind limb pareses) than the preceding one (Fig. 7A). Control animals undergoing repeated immunization with MBP/IFA also developed EAE though with mild intensity (no weight loss, clinical score 0.5: partial tail paresis) indicating that the animals had raised and maintained autoreactive memory T cells after the first immunization (Fig. 7B).
We determined the numbers and the reactivity of Tmemory cells after repeated immunizations. Tmemory cells 2 mo after the first and 14 mo after the second immunization (>18 mo after neonatal transfer) persisted in frequencies of 0.17% (±0.02%) and 0.02% (±0.001%) of LN cells, respectively. Upon Ag exposure the Tmemory cells massively amplified (Fig. 7C, >1500-fold). Real-time analysis of the cells revealed that they maintained their Th-1 cytokine profile (Fig. 7, D and E). Further, they remained encephalitogenic: after expansion in vitro and transfer to healthy recipients, the Tmemory cells induced severe clinical EAE (maximal clinical score 3, data not shown).
Discussion
Formation of memory responses is a hallmark of the adaptive immune system. A first encounter of a particular Ag conditions the immune system to respond faster and more vigorously when exposed repeatedly to Ag. The memory response is due to the expansion and differentiation of Ag-specific memory cells which are more sensitive to the Ag and are able to produce effector molecules more efficiently than their naive progenitors (20).
Immunological memory has evolved to warrant rapid and efficient elimination of microbial agents that repeatedly enter the organism. As a rule, immunological memory builds up, following successful elimination or neutralization of the Ag from the organism. In contrast, persistence of Ag, like in chronic infectious diseases, often leads to the exhaustion of the immune response (21). In autoimmune responses, the target autoantigen is not eliminated, but persists throughout life. Thus, would encephalitogenic T cells build up memory against autoantigens, and if so, do memory cells play a role in the course of chronic or relapsing autoimmune disease? The long-lasting persistence of autoimmune T cell clones in human blood (22, 23, 24), and the memory phenotype displayed by at least some of these cells (25, 26, 27) may argue in favor of autoimmune memory. The opposite prediction would come from EAE studies, where priming of rats with MBP in CFA or IFA confers protection from subsequent EAE induction, but does not prime for enhanced inducibility (28).
A model to study autoimmune memory is required to allow identification and functional characterization of autoreactive memory T cell populations in an intact immune repertoire over extended periods of time. We describe here an experimental system which seems to satisfy these requirements. The model is based on the introduction of labeled autoreactive T cells into the neonatal, immature immune system, and their analysis in adulthood. The retrovirally transduced, GFP-expressing T cell lines lent themselves for our investigations, because they replenish their fluorescent label over years both in vivo and in vitro. Furthermore, as described before, retroviral manipulation does not interfere with the programmed function of the cells (9, 10). Although the GFP-labeled T cells produce a foreign protein which would cause their eventual rejection in adults, the cells are tolerated lifelong after transfer into neonatal hosts. Our model is based on previous work which had established that encephalitogenic T cells transferred into neonatal hosts neither induce EAE nor activate counterregulatory T cell loops (18, 19). Instead, the inoculated T cells persist in the recipients’ immune organs throughout life (29). The late formation of CNS myelin (30) may be one factor responsible for the resistance to transferred disease. A deficit in T cell responsiveness and susceptibility to tolerance induction of the young immune system may be another factor (31).
Neonatal tolerance of foreign cells is not a simple phenomenon of passive acceptance, but relies on several mechanisms. These may include deletion of the grafted T cells from the host’s repertoire, the activation of regulatory T cells, and the formation of anti-inflammatory milieus (32). Obviously, any of these mechanisms could influence the function of the myelin autoreactive T cells. However, this was not the case in our model. The GFP-labeled MBP-specific T cells maintained their functional properties throughout their persistence in the hosts’ immune tissues. Their responsiveness against Ag remained unimpaired and they maintained their full encephalitogenic potential.
Transfer of mature T cells into neonatal organisms may be a less artificial situation, as it may appear on first glance. It should be kept in mind that the introduction of foreign, i.e., maternal immune cells into a neonatal organism is a natural event. T cells enter the fetus via placenta, and the neonate via colostral milk (33). Maternal immune cells, which in humans preferentially display memory properties, are demonstrable for long periods of time (34). They may well influence reactivity of the maturing immune system for better or worse (35).
The phenotypic stability of in vivo persisting memory T cells was remarkable. Even after periods of >1.5 years, the T cells maintained their "memory phenotype", with high levels of CD44, but low levels of CD45RC and L-selectin (Fig. 2B). The cells never showed any reversion toward a "naive" phenotype, as has been described in murine models (36, 37). Furthermore, the memory T cells retained a stable cytokine response pattern. When isolated from adult rats, GFP-labeled MBP-specific T cells responded to specific Ag by prompt production of Th1-like cytokines, i.e., IFN-, TNF-, and IL2. Th1-like responses were also triggered in vivo. This phenotypic stability is in contrast with memory T cells transferred into adult RAG mutant mice, which turned either Th1 or Th2 depending on the nature of the antigenic stimulus applied (38). Did the autoimmune T cells persist as "central" or "effector" memory cells, categories that had been invoked in human (39) and mouse (40) studies? Although we did not study chemokine receptors, the low L-selectin levels and their functional properties would be compatible with the "effector memory" option.
Although our memory rats contained considerable numbers of highly autoreactive T cells, none of them ever developed spontaneous EAE bouts. In contrast to naive rats, however, memory rats were susceptible to EAE induction by MBP in IFA, and such treated rats responded with accelerated formation of anti-MBP Abs, especially of the IgG2a isotype. But, unexpectedly, the reaction of memory rats to MBP/CFA was barely enhanced. An MBP/CFA stimulus triggered a slightly accelerated EAE response (though without increased severity). The memory T cells, which respond to autoantigen so promptly in vitro, either are shielded from activating signals in vivo, or they are under the tight control of down-regulatory mechanisms.
One regulatory loop was established by the work of Cohen and colleagues (41) who showed that intact or attenuated encephalitogenic T cell lines protect adult Lewis rats from subsequent attempts to induce EAE. The vaccination is due to the activation of CD8+ regulatory T cells, which suppress encephalitogenic T cells in vivo and destroy them in vitro (42, 43). In the mouse, CD8+ regulatory T cells recognize clonotypic structures of the target CD4+ T cells in the context of atypical MHC class Ib proteins (44, 45). This mechanism is not effective in our system, because transfer of encephalitogenic T cells into neonatal recipients fails to recruit protective CD8+ T cells (19).
Another regulatory loop which has resurfaced recently is CD4+CD25+ regulatory T cells, which down-regulate the pathogenic potential of autoreactive cells in vivo and in vitro via membrane-dependent contacts (46, 47). Control by regulatory cells, however, is less likely in our system, as demonstrated by strong proliferation and cytokine production of the GFP+ memory T cells in vitro as well as in vivo.
Autoimmune T cells undergo radical changes on their way into the target organ. They must be maximally activated to become pathogenic (48). But when introduced into a naive recipient, the T cells first settle in peripheral immune organs, where they assume a "migratory phenotype", which involves a profound reorganization of the membrane protein pattern including down-regulation of activation markers IL-2R and OX-40 and induction of chemokine receptors. Upon arrival in the CNS, the autoimmune T cells become reactivated again (10). The memory T cells described here resemble resting T cells with regard to their membrane phenotype, but functionally, they are distinct. Even after repeated immunizations and persistence for >1.5 years, the proliferative response of memory T cells to Ag by far exceeded the one of resting cultured T line cells or ex vivo isolated migratory T cells (Figs. 3 and 7). Memory T cells amplified several 100-fold, in contrast to the other T cell stages with maximally 20–50 amplification rounds. One possible explanation could lie in an intrinsic property of the proliferative machinery of the memory T cells allowing higher cell division rates. Alternatively, lower levels of apoptosis, e.g., based on elevated levels of antiapoptotic molecules might enable the cells to proliferate even under suboptimal stimulation conditions (49, 50).
Which are the factors that keep MBP-specific memory T cells in vivo alive over years? The roles of Ag-dependent signals and cytokines of the local milieu for persistence of memory CD4 T cells are matters of debate. Cytokines, such as IL-7 and IL-15, are considered as potential candidates as well for CD8+ and CD4+ memory T cells (51, 52). Several transfer studies using recipients lacking either relevant Ag or MHC products argue against a major role of continuous antigenic signaling (53, 54, 55). However, other work indicates that contact with Ag helps to maintain Ag reactivity of memory cells over time (56). It is not established whether MBP-specific memory T cells would find their nominal Ag presented by APCs in the peripheral lymphoid tissues, where they persist. Definitely, MBP-related genes are expressed in cells of the lymphoid system (57, 58), but to date immunogenic presentation of these proteins has not been shown with certainty.
The present memory model allows investigations of the fate of autoimmune T cells beyond a disease episode, a clinically important question, which so far has eluded detailed investigation. In previous studies we analyzed the migratory behavior of GFP-labeled encephalitogenic T cells on their way to and within the CNS tissue (9, 10, 15). However, due to antigenicity of the GFP protein, it was impossible to reliably trace the cells for periods longer than 1 wk. Our results now show that memory T cells integrated in a naive host’s immune system can be activated in vivo to contribute to a transient EAE attack. Most important, while numerous effector cells are deleted in the lesion, another substantial proportion of the cells remains within the lymphoid system, where it can be reactivated at a later time point.
In conclusion we show here that autoreactive memory T cells can persist in a healthy organism essentially throughout life. Under "regular" circumstances, they do not spontaneously produce disease, but they clearly participate in induced autoimmune attacks. It is therefore reasonable to conclude that autoimmune memory T cells may participate in human chronic autoimmune disease, such as in multiple sclerosis, and as we hope, animal models of autoimmune memory may contribute to better understanding of these long-lasting disorders.
Acknowledgments
We thank Simone Bauer, Ingeborg Haarmann, Katrin Vogt, Dietmute Büringer, and Karin Brückner for excellent technical assistance. We thank Drs. Klaus Dornmair, Edgar Meinl, and David Johnson for critically reading the manuscript.
Disclosures
The authors have no financial conflict of interest.
Footnotes
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
1 The work was supported by the Deutsche Forschungsgemeinschaft (SFB 455) and the European Community (Mechanisms of Brain Inflammation: QLG3-CT-2002-00712).
2 Address correspondence and reprint requests to Dr. Alexander Flügel, Max-Planck-Institute of Neurobiology, Department of Neuroimmunology, Am Klopferspitz 18, 82152 Martinsried, Germany. E-mail address: Fluegel@neuro.mpg.de
3 Abbreviations used in this paper: MBP, myelin basic protein; p.i., postimmunization; EAE, experimental autoimmune encephalomyelitis; tEAE, adoptive transfer of EAE; PFA, paraformaldehyde; LN, lymph node.
Received for publication December 30, 2004. Accepted for publication April 6, 2005.
References
Compston, A., A. Coles. 2002. Multiple sclerosis. [Published erratum appears in 2002 Lancet 360: 648.]. Lancet 359: 1221-1231.
Wekerle, H.. 1998. Immunology of multiple sclerosis. A. Compston, and H. Lassmann, and W. I. McDonald, and B. Matthews, and H. Wekerle, eds. McAlpine’s Multiple Sclerosis 379-407 Churchill Livingston, London. .
Hohlfeld, R.. 2003. Hunting (auto)immune T cells in neuroimmunological diseases. Brain 126: 2-4.
Berger, T., S. Weerth, K. Kojima, C. Linington, H. Wekerle, H. Lassmann. 1997. Experimental autoimmune encephalomyelitis: the antigen specificity of T-lymphocytes determines the topography of lesions in the central and peripheral nervous system. Lab. Invest. 76: 355-364.
Tuohy, V. K., M. Yu, B. Weinstock-Guttman, R. P. Kinkel. 1997. Diversity and plasticity of self recognition during the development of multiple sclerosis. J. Clin. Invest. 99: 1682-1690
Lehmann, P. V., T. Forsthuber, A. Miller, E. E. Sercarz. 1992. Spreading of T-cell autoimmunity to cryptic determinants of an autoantigen. Nature 358: 155-157.
Lehmann, P. V., E. E. Sercarz, T. Forsthuber, C. M. Dayan, G. Gammon. 1993. Determinant spreading and the dynamics of the autoimmune T cell repertoire. Immunol. Today 14: 203-208.(Naoto Kawakami*, Francesc)
We embedded green fluorescent CD4+ T cells specific for myelin basic protein (MBP) (TMBP-GFP cells) in the immune system of syngeneic neonatal rats. These cells persisted in the animals for the entire observation period spanning >2 years without affecting the health of the hosts. They maintained a memory phenotype with low levels of L-selectin and CD45RC, but high CD44. Although persisting in low numbers (0.01–0.1% of lymph node cells) they were sufficient to raise susceptibility toward clinical autoimmune disease. Immunization with MBP in IFA induced CNS inflammation and overt clinical disease in animals carrying neonatally transferred TMBP-GFP cells, but not in controls. The onset of the clinical disease coincided with mass infiltration of TMBP-GFP cells into the CNS. In the periphery, following the amplification phase a rapid contraction of the T cell population was observed. However, elevated numbers of fully reactive TMBP-GFP cells remained in the peripheral immune system after acute experimental autoimmune encephalomyelitis mediating reimmunization-induced disease relapses.
Introduction
Human autoimmune diseases, such as multiple sclerosis and rheumatoid arthritis, commonly take chronic, disabling courses. In some patients disease progresses steadily, while in others, relapses alter with remissions (1). According to present understanding, disease is caused and controlled by autoimmune T cells, which attack their target tissues, and thus determine activity and clinical character (2, 3). However, the nature of the autoimmune T cells over time remains largely unclear. In multiple sclerosis, for example, it is important to know whether different clones are recruited and activated from relapse to relapse. Newly recruited T cells could recognize distinct autoantigenic epitopes on the same autoantigen, or even be specific for different autoantigens, termed intra- and intermolecular determinant spreading, respectively (4, 5, 6, 7). As a less popular alternative, sequential disease episodes could be driven by members of the initial pathogenic clone(s) throughout the entire course of disease. Such a pathogenesis would invoke formation and persistence of autoimmune memory T cells (8).
Mechanisms of disease chronicity are difficult to study in model systems. This holds especially true for models induced by autoimmunization, because these either result in monophasic disease episodes with full and definite recovery, or, depending on large-scale irreversible tissue damage, take a chronic course. In this study, we established a model system which enabled us to study the fate of individual bona fide autoimmune memory cell clones in the Lewis rat. We transferred myelin basic protein (MBP) 3-reactive CD4+ T cells retrovirally engineered to express the gene of GFP (9, 10) into neonatal recipient animals. Using this technique we tracked and functionally characterized defined (autoreactive) memory T cell populations in vivo over time. The T cells integrated into the immune repertoire of immune-competent recipients leaving the endogenous immune system unaffected. They conferred T and B cellular reactivity to the recipient animals including specific T cell proliferation, proinflammatory cytokine release, and accelerated and increased autoantibody production. Importantly, the persisting autoreactive memory T cells raised the susceptibility toward autoimmune CNS inflammation and disease. After recovery from acute disease, they remained functionally fully intact in the peripheral immune system and, upon activation, evoked repeated disease episodes.
Thus, the data point toward long-term persistence of an autoimmunologic T cell memory as a critical factor in chronic autoimmune disease.
Materials and Methods
Animals, CD4+ TGFP cells, and neonatal transfer of TGFP cells
Lewis rats were obtained from the animal breeding facilities of the Max-Planck-Institute for Neurobiology (Martinsried, Germany) and were kept under standardized conditions. Ag-specific T cell clones used in the study were specific for guinea pig MBP and OVA. MBP was purified from guinea pig brains as described (11). Hen egg OVA was obtained from Sigma-Aldrich. All animal experiments were performed with the license of the Regierung von Oberbayern (No. 209.1/211-2531-56/99). CD4+ TGFP cells have been generated and tested for their phenotype, cytokine profile, and Ag specificity as described before (9). The lines consisted of oligoclonal CD4+ TCR+ Th1 cells and were highly specific for their respective Ags (see Table I, see Fig. 7D). Intraperitoneal transfer of TGFP cells (2 x 106 cells, in 0.5 ml/animal in Eagle’s HEPES medium) into newborns was performed under hypothermia within 48 h after birth. After T cell transfer, the newborn rats were kept under a 30°C humid atmosphere until fully recovered, then returned to their mother. Four different TMBP-GFP and three different TOVA-GFP cell lines were used for neonatal transfer. The activation state of the T cells did not influence the capacity of T cells to become memory cells. In most of the experiments, the T cells were transferred 4–5 days after restimulation with Ag.
Cell isolation, cytofluorometry, and FACS
Single-cell suspensions from organs were obtained as described previously (9, 10, 12). For cytofluorometric analysis, FACSCalibur operated by CellQuest software (BD Biosciences) was used (9, 10). The following mAbs were used for surface membrane analysis: W3/25 (anti-CD4; Serotec), R73 (TCR), OX-6 (rat MHC class II), OX-40 Ag (CD134), OX-39 (CD25, IL-2R chain), OX-22 (CD45RC), H-CAM (CD44), L-selectin (CD62L), V 8.2, 8.5, 10, 16 (all BD Biosciences). RPE-Cy-5-labeled anti-mouse Ab (DAKO) was used as secondary Ab.
Induction of adoptive transfer of experimental autoimmune encephalomyelitis (tEAE) and animal preparation
Adoptive transfer of the encephalitogenic T cell lines was performed by i.v. injection. The dose of T cells injected was adjusted to 5 x 106 T cell blasts/animal. Animals were monitored daily by measuring weight and examining disease scores (0, no disease; 1, flaccid tail; 2, gait disturbance; 3, complete hind limb paralysis; 4, tetraparesis; and 5, death).
Organ distribution of neonatally transferred TMBP-GFP and TOVA-GFP cells
Numbers of persisting TMBP-GFP following neonatal transfer were evaluated cytofluorometrically. The organs were carefully dissected and the relative numbers of GFP per organ cells were measured.
Histology
Histological analysis was performed as described (10). Briefly, after perfusion with 4% paraformaldehyde (PFA), lymph nodes (LNs) were prepared and postfixed in 4% PFA overnight. The frozen organs were cut into 20-μm sections. Following fixation for 30 min at room temperature with 4% PFA the sections were incubated with OX-33 (CD45RA, B cell marker, dilution 1/200; BD Biosciences) overnight at 4°C. Cy3-labeled anti-mouse antiserum (Dianova) was used as secondary Ab. The sections were counterstained with biotinylated anti-TCR Ab (R73, dilution 1/300), detected by Cy5-labeled streptavidin (Molecular Probes). Fluorescence analysis was performed with confocal laser-scanning microscopy (Leica). Immunohistochemical staining for GFP and analysis with a Zeiss EM10 transmission electron microscope was performed as described (13). For quantification of GFP+ cells, organs were treated as described (14). The inflammatory index in spinal cord lesions was determined after quantification of T cells (W3/13, dilution 1/50; Harlan Seralab) and monocytes/macrophages (ED1, dilution 1/1000; Serotec) on five randomly selected complete spinal cord cross-sections per animal. The section area was determined using a morphometrical grid. The values represent means (±SD) of two individual animals per group.
Ex vivo reactivity assays
LN cells were cultured in 96-well plates (in DMEM 1% rat serum) without Ag, with ConA (2.5 μg/ml), or in the presence of specific Ag (10 μg/ml MBP or OVA, respectively), or with the irrelevant Ag purified protein derivative (10 μg/ml), respectively. [3H]dT (2 Ci/mmol; Amersham-Buchler) was added to the cultures after 48 h. The radioactive label present in the different cultures was determined as described (10).
Amplification of ex vivo-isolated GFP+ memory T cells was measured by cytofluorometry. Their numbers were determined in relation to a known absolute number of added PE-labeled plastic beads (BD Biosciences). The amplification rate was calculated in relation to the GFP+ T cell numbers at day 0.
Intracellular IFN- staining, IFN- ELISPOT
Intracellular IFN- staining was performed with anti-mouse/rat IFN- Ab (clone DB-1; BD Biosciences) as described (15). Control IgG (mouse IgG MOPC31) was obtained from Sigma-Aldrich. IFN--ELISPOT analyses were performed as described (15) using polyclonal goat anti-rat-IFN- and biotinylated goat anti-rat-IFN- antiserum (R&D Systems), and an automated imaging system equipped with appropriate computer software (KS ELISPOT Automated Image Analysis System; Zeiss).
Quantitative PCR
TaqMan analysis was performed as reported (10) using the ABI Prim 7700 Sequence Detector TaqMan (Applied Biosystems). For quantification of cytokine mRNAs, the expression of the cytokine mRNA was set in relation to a housekeeping gene (-actin). All PCR data were obtained by two independent measurements. The cycle threshold (CT) value of the measurements did not differ >0.5 amplification cycles.
Ab measurements
ELISA. Serial blood samples were obtained from tails of the rats. After clotting at 4°C, the sera were collected by centrifugation and stored at –20°C. ELISA was performed according to the protocol of Stefferl et al. (16). The rat sera were diluted 1/1000. Mouse mAbs specific for rat Ig and isotypes IgM, IgG1, IgG2a, IgG2b, and IgE were obtained from Serotec. Secondary goat anti-mouse peroxidase conjugate was purchased from Dianova (dilution 1/8000). O-phenylenediamine dihydrochloride (Sigma-Aldrich) was used as substrate. The reaction was stopped with 3 M HCl, and OD was determined at 490 nm.
ELISPOT. Plasma cells were determined by reversing the present sandwich method (17). Briefly, ELISPOT plates (MAHA N45; Millipore) were coated with goat anti-rat IgG (H+L) Ab (10 μg/ml, 100 μl/well in carbonate buffer, pH 9.3; Jackson ImmunoResearch Laboratories). After blocking (5% BSA-PBS), cells were added in a serial dilution (100 μl/well) in triplicates and incubated for 24 h. After removal of cells, plates were incubated for 2 h with biotinylated Ag (MBP or OVA, respectively), or biotinylated goat anti-rat IgG (H+L) Ab (0.5 μg/ml, 100 μl/well in 0.5% BSA-PBS; Jackson ImmunoResearch Laboratories). Spots onto the ELISPOT plates were visualized using streptavidin-alkaline phosphatase with 5-bromo-4-chloro-3-indolyl phosphate/NBT as substrate (Sigma-Aldrich), and analyzed with an automated ELISPOT reader system (KS ELISPOT; Zeiss, Jena, Germany). Total Ig-spots were considered total numbers of plasma cells, MBP/OVA-spots represented MBP/OVA-specific plasma cells. No second biotinylated detector was used as negative control.
Results
Introduction of GFP-expressing CD4+ T cells into the immune repertoire by neonatal transfer
GFP-transduced CD4+ T cells (9) specific for MBP or OVA (OVA, TMBP-GFP, or TOVA-GFP, respectively), were injected i.p. into syngeneic Lewis rat pups (2 x 106 cells, within 48 h of life). Both TOVA-GFP and TMBP-GFP cells consisted of highly specific oligoclonal CD4+ T cell populations which displayed a Th1-like cytokine profile with high levels of IFN- and TNF-, but no IL-4 (Table I) (10). TMBP-GFP cells transferred clinical disease to adult animals, which in course and severity was indistinguishable from EAE transferred by their non-manipulated counterparts. However, neither of the MBP-specific T cells attacked newborn recipients (18, 19).
We traced the transferred, Ag-experienced T cells in the recipients over periods of >2 years. The cells settled throughout the immune system including LNs, spleen, thymus, and bone marrow (Fig. 1A), and they were also found in some non-immune organs including lung, liver, and gut. Importantly, however, they never spontaneously invaded the CNS (Fig. 1A). In LNs, the T cells populated the paracortical T cell areas and the medulla of the LNs. A few scattered green T cells were seen within lymph follicles (Fig. 1B). In the spleen, the persisting GFP+ T cells sat both in the periarteriolar sheaths of the white pulp and in the red pulp (data not shown). Within the thymus, most, if not all of the cells were located in the medulla (data not shown). The persisting T cells (Tmemory-MBP cells) did not demonstrably influence the host’s immune cell composition and immune competence (data not shown).
We previously showed that in tEAE, most of the transferred effector T cells are lost both from the CNS and from the peripheral immune system (10). We now observed that Tmemory-MBP cells behave very differently. Although they were effectively cleared from the CNS following an acute EAE bout (Fig. 4C), they persisted in the periphery even in elevated numbers (Fig. 5A). Furthermore, these cells remained functionally intact. Reimmunization of memory rats with MBP/IFA 3 mo after a first EAE attack led again to clinical disease starting 7 days p.i., and this second EAE episode became more severe (weight loss, clinical score 2–3: hind limb pareses) than the preceding one (Fig. 7A). Control animals undergoing repeated immunization with MBP/IFA also developed EAE though with mild intensity (no weight loss, clinical score 0.5: partial tail paresis) indicating that the animals had raised and maintained autoreactive memory T cells after the first immunization (Fig. 7B).
We determined the numbers and the reactivity of Tmemory cells after repeated immunizations. Tmemory cells 2 mo after the first and 14 mo after the second immunization (>18 mo after neonatal transfer) persisted in frequencies of 0.17% (±0.02%) and 0.02% (±0.001%) of LN cells, respectively. Upon Ag exposure the Tmemory cells massively amplified (Fig. 7C, >1500-fold). Real-time analysis of the cells revealed that they maintained their Th-1 cytokine profile (Fig. 7, D and E). Further, they remained encephalitogenic: after expansion in vitro and transfer to healthy recipients, the Tmemory cells induced severe clinical EAE (maximal clinical score 3, data not shown).
Discussion
Formation of memory responses is a hallmark of the adaptive immune system. A first encounter of a particular Ag conditions the immune system to respond faster and more vigorously when exposed repeatedly to Ag. The memory response is due to the expansion and differentiation of Ag-specific memory cells which are more sensitive to the Ag and are able to produce effector molecules more efficiently than their naive progenitors (20).
Immunological memory has evolved to warrant rapid and efficient elimination of microbial agents that repeatedly enter the organism. As a rule, immunological memory builds up, following successful elimination or neutralization of the Ag from the organism. In contrast, persistence of Ag, like in chronic infectious diseases, often leads to the exhaustion of the immune response (21). In autoimmune responses, the target autoantigen is not eliminated, but persists throughout life. Thus, would encephalitogenic T cells build up memory against autoantigens, and if so, do memory cells play a role in the course of chronic or relapsing autoimmune disease? The long-lasting persistence of autoimmune T cell clones in human blood (22, 23, 24), and the memory phenotype displayed by at least some of these cells (25, 26, 27) may argue in favor of autoimmune memory. The opposite prediction would come from EAE studies, where priming of rats with MBP in CFA or IFA confers protection from subsequent EAE induction, but does not prime for enhanced inducibility (28).
A model to study autoimmune memory is required to allow identification and functional characterization of autoreactive memory T cell populations in an intact immune repertoire over extended periods of time. We describe here an experimental system which seems to satisfy these requirements. The model is based on the introduction of labeled autoreactive T cells into the neonatal, immature immune system, and their analysis in adulthood. The retrovirally transduced, GFP-expressing T cell lines lent themselves for our investigations, because they replenish their fluorescent label over years both in vivo and in vitro. Furthermore, as described before, retroviral manipulation does not interfere with the programmed function of the cells (9, 10). Although the GFP-labeled T cells produce a foreign protein which would cause their eventual rejection in adults, the cells are tolerated lifelong after transfer into neonatal hosts. Our model is based on previous work which had established that encephalitogenic T cells transferred into neonatal hosts neither induce EAE nor activate counterregulatory T cell loops (18, 19). Instead, the inoculated T cells persist in the recipients’ immune organs throughout life (29). The late formation of CNS myelin (30) may be one factor responsible for the resistance to transferred disease. A deficit in T cell responsiveness and susceptibility to tolerance induction of the young immune system may be another factor (31).
Neonatal tolerance of foreign cells is not a simple phenomenon of passive acceptance, but relies on several mechanisms. These may include deletion of the grafted T cells from the host’s repertoire, the activation of regulatory T cells, and the formation of anti-inflammatory milieus (32). Obviously, any of these mechanisms could influence the function of the myelin autoreactive T cells. However, this was not the case in our model. The GFP-labeled MBP-specific T cells maintained their functional properties throughout their persistence in the hosts’ immune tissues. Their responsiveness against Ag remained unimpaired and they maintained their full encephalitogenic potential.
Transfer of mature T cells into neonatal organisms may be a less artificial situation, as it may appear on first glance. It should be kept in mind that the introduction of foreign, i.e., maternal immune cells into a neonatal organism is a natural event. T cells enter the fetus via placenta, and the neonate via colostral milk (33). Maternal immune cells, which in humans preferentially display memory properties, are demonstrable for long periods of time (34). They may well influence reactivity of the maturing immune system for better or worse (35).
The phenotypic stability of in vivo persisting memory T cells was remarkable. Even after periods of >1.5 years, the T cells maintained their "memory phenotype", with high levels of CD44, but low levels of CD45RC and L-selectin (Fig. 2B). The cells never showed any reversion toward a "naive" phenotype, as has been described in murine models (36, 37). Furthermore, the memory T cells retained a stable cytokine response pattern. When isolated from adult rats, GFP-labeled MBP-specific T cells responded to specific Ag by prompt production of Th1-like cytokines, i.e., IFN-, TNF-, and IL2. Th1-like responses were also triggered in vivo. This phenotypic stability is in contrast with memory T cells transferred into adult RAG mutant mice, which turned either Th1 or Th2 depending on the nature of the antigenic stimulus applied (38). Did the autoimmune T cells persist as "central" or "effector" memory cells, categories that had been invoked in human (39) and mouse (40) studies? Although we did not study chemokine receptors, the low L-selectin levels and their functional properties would be compatible with the "effector memory" option.
Although our memory rats contained considerable numbers of highly autoreactive T cells, none of them ever developed spontaneous EAE bouts. In contrast to naive rats, however, memory rats were susceptible to EAE induction by MBP in IFA, and such treated rats responded with accelerated formation of anti-MBP Abs, especially of the IgG2a isotype. But, unexpectedly, the reaction of memory rats to MBP/CFA was barely enhanced. An MBP/CFA stimulus triggered a slightly accelerated EAE response (though without increased severity). The memory T cells, which respond to autoantigen so promptly in vitro, either are shielded from activating signals in vivo, or they are under the tight control of down-regulatory mechanisms.
One regulatory loop was established by the work of Cohen and colleagues (41) who showed that intact or attenuated encephalitogenic T cell lines protect adult Lewis rats from subsequent attempts to induce EAE. The vaccination is due to the activation of CD8+ regulatory T cells, which suppress encephalitogenic T cells in vivo and destroy them in vitro (42, 43). In the mouse, CD8+ regulatory T cells recognize clonotypic structures of the target CD4+ T cells in the context of atypical MHC class Ib proteins (44, 45). This mechanism is not effective in our system, because transfer of encephalitogenic T cells into neonatal recipients fails to recruit protective CD8+ T cells (19).
Another regulatory loop which has resurfaced recently is CD4+CD25+ regulatory T cells, which down-regulate the pathogenic potential of autoreactive cells in vivo and in vitro via membrane-dependent contacts (46, 47). Control by regulatory cells, however, is less likely in our system, as demonstrated by strong proliferation and cytokine production of the GFP+ memory T cells in vitro as well as in vivo.
Autoimmune T cells undergo radical changes on their way into the target organ. They must be maximally activated to become pathogenic (48). But when introduced into a naive recipient, the T cells first settle in peripheral immune organs, where they assume a "migratory phenotype", which involves a profound reorganization of the membrane protein pattern including down-regulation of activation markers IL-2R and OX-40 and induction of chemokine receptors. Upon arrival in the CNS, the autoimmune T cells become reactivated again (10). The memory T cells described here resemble resting T cells with regard to their membrane phenotype, but functionally, they are distinct. Even after repeated immunizations and persistence for >1.5 years, the proliferative response of memory T cells to Ag by far exceeded the one of resting cultured T line cells or ex vivo isolated migratory T cells (Figs. 3 and 7). Memory T cells amplified several 100-fold, in contrast to the other T cell stages with maximally 20–50 amplification rounds. One possible explanation could lie in an intrinsic property of the proliferative machinery of the memory T cells allowing higher cell division rates. Alternatively, lower levels of apoptosis, e.g., based on elevated levels of antiapoptotic molecules might enable the cells to proliferate even under suboptimal stimulation conditions (49, 50).
Which are the factors that keep MBP-specific memory T cells in vivo alive over years? The roles of Ag-dependent signals and cytokines of the local milieu for persistence of memory CD4 T cells are matters of debate. Cytokines, such as IL-7 and IL-15, are considered as potential candidates as well for CD8+ and CD4+ memory T cells (51, 52). Several transfer studies using recipients lacking either relevant Ag or MHC products argue against a major role of continuous antigenic signaling (53, 54, 55). However, other work indicates that contact with Ag helps to maintain Ag reactivity of memory cells over time (56). It is not established whether MBP-specific memory T cells would find their nominal Ag presented by APCs in the peripheral lymphoid tissues, where they persist. Definitely, MBP-related genes are expressed in cells of the lymphoid system (57, 58), but to date immunogenic presentation of these proteins has not been shown with certainty.
The present memory model allows investigations of the fate of autoimmune T cells beyond a disease episode, a clinically important question, which so far has eluded detailed investigation. In previous studies we analyzed the migratory behavior of GFP-labeled encephalitogenic T cells on their way to and within the CNS tissue (9, 10, 15). However, due to antigenicity of the GFP protein, it was impossible to reliably trace the cells for periods longer than 1 wk. Our results now show that memory T cells integrated in a naive host’s immune system can be activated in vivo to contribute to a transient EAE attack. Most important, while numerous effector cells are deleted in the lesion, another substantial proportion of the cells remains within the lymphoid system, where it can be reactivated at a later time point.
In conclusion we show here that autoreactive memory T cells can persist in a healthy organism essentially throughout life. Under "regular" circumstances, they do not spontaneously produce disease, but they clearly participate in induced autoimmune attacks. It is therefore reasonable to conclude that autoimmune memory T cells may participate in human chronic autoimmune disease, such as in multiple sclerosis, and as we hope, animal models of autoimmune memory may contribute to better understanding of these long-lasting disorders.
Acknowledgments
We thank Simone Bauer, Ingeborg Haarmann, Katrin Vogt, Dietmute Büringer, and Karin Brückner for excellent technical assistance. We thank Drs. Klaus Dornmair, Edgar Meinl, and David Johnson for critically reading the manuscript.
Disclosures
The authors have no financial conflict of interest.
Footnotes
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
1 The work was supported by the Deutsche Forschungsgemeinschaft (SFB 455) and the European Community (Mechanisms of Brain Inflammation: QLG3-CT-2002-00712).
2 Address correspondence and reprint requests to Dr. Alexander Flügel, Max-Planck-Institute of Neurobiology, Department of Neuroimmunology, Am Klopferspitz 18, 82152 Martinsried, Germany. E-mail address: Fluegel@neuro.mpg.de
3 Abbreviations used in this paper: MBP, myelin basic protein; p.i., postimmunization; EAE, experimental autoimmune encephalomyelitis; tEAE, adoptive transfer of EAE; PFA, paraformaldehyde; LN, lymph node.
Received for publication December 30, 2004. Accepted for publication April 6, 2005.
References
Compston, A., A. Coles. 2002. Multiple sclerosis. [Published erratum appears in 2002 Lancet 360: 648.]. Lancet 359: 1221-1231.
Wekerle, H.. 1998. Immunology of multiple sclerosis. A. Compston, and H. Lassmann, and W. I. McDonald, and B. Matthews, and H. Wekerle, eds. McAlpine’s Multiple Sclerosis 379-407 Churchill Livingston, London. .
Hohlfeld, R.. 2003. Hunting (auto)immune T cells in neuroimmunological diseases. Brain 126: 2-4.
Berger, T., S. Weerth, K. Kojima, C. Linington, H. Wekerle, H. Lassmann. 1997. Experimental autoimmune encephalomyelitis: the antigen specificity of T-lymphocytes determines the topography of lesions in the central and peripheral nervous system. Lab. Invest. 76: 355-364.
Tuohy, V. K., M. Yu, B. Weinstock-Guttman, R. P. Kinkel. 1997. Diversity and plasticity of self recognition during the development of multiple sclerosis. J. Clin. Invest. 99: 1682-1690
Lehmann, P. V., T. Forsthuber, A. Miller, E. E. Sercarz. 1992. Spreading of T-cell autoimmunity to cryptic determinants of an autoantigen. Nature 358: 155-157.
Lehmann, P. V., E. E. Sercarz, T. Forsthuber, C. M. Dayan, G. Gammon. 1993. Determinant spreading and the dynamics of the autoimmune T cell repertoire. Immunol. Today 14: 203-208.(Naoto Kawakami*, Francesc)