Frequencies of Neuroantigen-Specific T Cells in the Central Nervous System Versus the Immune Periphery During the Course of Experimental Allergic Encephalomyelitis¹

The contribution of neuroantigen-specific T cells to the autoimmune disease process in experimental allergic encephalomyelitis (EAE)⁴ is particularly incomplete when it comes to an understanding of their functions in the target organ, the CNS itself. The detection of such cells has been largely confined to draining lymph nodes (dLN) and the spleen.

Most autoreactive T cells are thought to be deleted in the thymus. This is also the case for lymphocytes specific for the neuroantigen myelin basic protein (MBP) and proteolipid protein (PLP) (1, 2, 3). Low affinity cells, however, and T cells specific for determinants not expressed in the thymus escape this negative selection process. These lymphocytes persist as naive precursor cells, apparently ignorant of the endogenous autoantigen (4, 5, 6, 7), but upon priming they can give rise to autoimmune disease. Induction of EAE requires immunization with the neuroantigen itself or a cross-reactive foreign Ag. This supposedly mimics an infection that initiates the spontaneously developing autoimmunity in multiple sclerosis, for which EAE serves as a model. While the initial clonal expansion of the autoreactive precursors and their differentiation into effector cells occur in the dLN, little is known about their subsequent fate. Thus, it is not known whether the neuroantigen-specific T cells head directly for the target organ after becoming primed or whether they first randomly populate the immune periphery (8, 9). It is also not known what percentage of the autoimmune repertoire resides in the target organ as opposed to the immune periphery at various stages of the disease process. This information is of critical importance for various reasons. First, it is important to know whether there is a sizable peripheral reservoir of effector cells from which the autoimmune process in the target organ draws, as this would have to be included in immunotherapeutic considerations: T cells in the periphery should be more accessible to manipulation than those in the CNS. Second, it would be beneficial for immune-monitoring purposes to know how the numbers and functions of the neuroantigen-specific T cells in the periphery reflect those in the relevant target organ.

The population kinetics of these T cells are also unknown. While T cells that enter the CNS parenchyma are known to undergo apoptosis (10, 11), there is also evidence that during the inflammatory process in EAE, lymphoid neo-organogenesis is induced by lymphotoxin in the CNS (12). This may in turn give rise to ongoing stimulation of the neuroantigen-specific T cells and drive their clonal expansion. Moreover, these induced lymphoid tissues within the CNS may represent the site in which the amplification of the autoimmune process via determinant spreading occurs (13, 14, 15, 16). In this study, we address the organ distribution and fate of the first wave of peripherally primed effector cells, and the site of engagement of the second wave effector cells. Furthermore, we were interested in determining whether Th1/Th2 class-switching of the effector cells occurs in the course of EAE (17, 18, 19).

In this study, we set out to use a computer-assisted cytokine enzyme-linked immunospot analysis (ELISPOT) assay (20) to follow the PLP_139–151 peptide-specific CD4 effector cells in the CNS and the periphery over the natural course of PLP_139–151-induced EAE (21), measuring directly their frequency and cytokine signature at single cell resolution. Naive T cells do not secret cytokine, nor do memory cells in the absence of Ag (22). T cell differentiation into cytokine-producing effector cells depends on cycles of Ag-driven proliferation to open up the chromatin structure (23), requiring at least 3–4 days (20). In contrast, memory/effector T cells that underwent cytokine differentiation start producing the cytokine that they are precommitted to express within 24 h after Ag encounter. Therefore, the testing of freshly isolated T cells in ELISPOT assays of 24- to 48-h duration provides information on the frequency and the cytokine signature of the specific effector T cells in vivo. In previous work, we have shown that when freshly isolated LN or spleen cells were challenged with protein Ags or antigenic peptides, only in vivo primed CD4 memory cells produced cytokine in ELISPOT assays of such short duration (2, 20, 24, 25). Moreover, while we have established that accurate frequencies of the cytokine-producing T cells are measured in lymphoid tissues (20), we had to address whether this also holds true for T cells isolated from the CNS, where the Ag-presenting compartment fundamentally differs. While performing measurements of T cells isolated from the CNS, we had to account for conditions of optimal Ag presentation, including the contribution of the endogenously presented neuroantigen, and to address the concern of contamination with blood-borne T cells. After establishing the assay conditions for single cell CNS measurements, we studied the frequency and absolute numbers of PLP_139–151-specific T cells after PLP_139–151 immunization in the CNS and in the immune periphery. We provide evidence that the vast majority of effector cells reside in the periphery, not in the CNS, even at the peak of the disease. Furthermore, we show that the first wave of effector cells (the PLP_139–151-specific T cells engaged by peripheral immunization) eventually exhausts without undergoing a Th2 switch. Finally, our data suggest that second wave immunity to the endogenous PLP_178–191 peptide (15, 26) is engaged in the CNS itself.

SJL mice were purchased from The Jackson Laboratory (Bar Harbor, ME) and kept at the animal facility of Case Western Reserve University under specific pathogen-free conditions. Female mice were used at 6–10 wk of age. PLP peptide 139–151 and PLP peptide 178–191 were purchased from Princeton Biomolecules (Langhorn, PA); OVA was purchased from Sigma (St. Louis, MO); and IFA was purchased from Life Technologies (Grand Island, NY). CFA was made by mixing Mycobacterium tuberculosis H37RA (Difco, Detroit, MI) at 2.5 mg/ml into IFA. A total of 0.5 mM of each peptide or of OVA was injected in CFA, s.c., 100 μl emulsion per mouse. Pertussis toxin (List Biological Laboratories, Campbell, CA) was injected on days 0 and 1, 0.2 ng at each time point. Starting from day 3 after immunization, mice were assessed for the development of paralytic symptoms, and the severity was recorded according to the standard scale: grade 1, floppy tail; grade 2, hind-leg weakness; grade 3, full hind-leg paralysis; grade 4, quadriplegia; grade 5, death. To ensure the nourishment of paralyzed mice, elongated water tubes were used and food was placed in the bedding.

Cells from the CNS were prepared as follows. After sacrificing the animal, the spinal cord was removed from the entire vertebral column, placed into DMEM medium, and disrupted with the back of a syringe. The resulting cell suspension was filtered through a Falcon Cell Strainer 2350 (Becton Dickinson, San Jose, CA). The cells were washed twice with DMEM and subsequently counted. The cells were resuspended in HL-1 medium (BioWhittaker, Walkersville, MD) supplemented with 1% glutamine and plated typically at 5 × 10⁴ to 5 × 10⁵ cells/well, or at serial dilutions into ImmunoSpot M200 plates (Cellular Technology Limited, Cleveland, OH). Cells from the peritoneal cavity (PC) were obtained by lavage after injecting 7 ml of DMEM medium. Mouse blood was obtained by retroorbital bleeding, using heparin as an anticoagulant. The blood was diluted four times with sterile saline, and PBMCs were obtained by density-gradient centrifugation over a Ficoll gradient. Single cell suspensions from dLN and nondraining LN and the spleen were prepared as previously described (1). For flow cytometric studies of the cell populations obtained, we used directly labeled Abs specified, all from PharMingen (San Diego, CA). The stained cells were analyzed with a FACScan and Cell Quest software (Becton Dickinson). The cells isolated from spinal cords of mice with EAE contained 19–43% CD4 cells.

ImmunoSpot M200 plates (Cellular Technology Limited) were coated overnight with the capture Abs in sterile PBS. R46A2 at 4 μg/ml (purified from hybridoma in our laboratory) was used for capturing IFN-γ; JES6-1A12 at 3 μg/ml (PharMingen) for IL-2; 11B11 at 2 μg/ml (purified from hybridoma) for IL-4; TRFK5 at 5 μg/ml (purified from hybridoma) was used for the capture of IL-5 and MP6-XT3 at 2.5 μg/ml for capture of TNF-α/β (PharMingen). The plates were blocked for 1 h with 1% BSA in PBS and washed three times with PBS. The freshly isolated cells from the various organs were plated in HL-1/1% glutamine-supplemented medium in the numbers specified, with or without Ag (7 μM, if not specified otherwise). In selected experiments, cells and Ags were titrated. The plates were cultured at 37°C, 7% CO₂ for 24 h (IFN-γ, IL-2, and TNF-α/β assays) or for 48 h (IL-4, IL-5 assays). The cells were then discarded, the plates washed with PBS first, then with PBS containing 0.025% Tween (PBST), and the detection Abs were added for overnight incubation. XMG1.2 biotin (purified from hybridoma in our laboratory) was used for IFN-γ; rat anti-mouse IL-4 biotin (BVD6-24G2; PharMingen) was used for IL-4; rat anti-mouse IL-2 biotin (JES6-5H4; PharMingen) was used for IL-2; biotinylated TRFK4 (PharMingen) was used for IL-5; and biotinylated MP6-XT3 at 3 μg/ml for the detection of TNF-α/β (PharMingen). The plates were then washed three times in PBST. Streptavidin-alkaline phosphatase (Dako, Carpenteria, CA) was added at a 1/1000 dilution in PBST as a third reagent for IL-2, IL-4, and IL-5, incubating for 2 h, followed by three washes in PBS. The plates were developed using nitroblue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate substrate (Kirkegaard & Perry Laboratories, Gaithersburg, MD) or HRP substrate AEC (3-amino-9-ethylcarbazole) (Pierce, Rockford, IL) for 30 min. The resulting spots were counted with an ImmunoSpot Series 1 Analyzer (Cellular Technology Limited) specifically designed for the ELISPOT assay. Digitized images were analyzed for the presence of areas in which color density exceeds background by a factor set on the basis of the difference between control wells (containing T cells and APC without Ag) and experimental wells (containing Ag in addition). After separating spots that touch or partially overlap, additional criteria of spot size and circularity were applied to gate out speckles and noise caused by spontaneous substrate precipitation, nonspecific Ab binding, and ELISA effects. Objects that did not meet these criteria were ignored, and areas that met them were recognized as spots, counted, and highlighted.

First we established the specificity of cytokine ELISPOT formation for the Ags used in this study: PLP_139–151, PLP_178–191, and OVA. SJL mice were injected with these Ags in CFA, and the dLN were tested 9 days later (Table I). In the presence of the immunizing Ag, IFN-γ- and IL-2-producing cells were detected in the frequency range 20/10⁶ to 150/10⁶; IL-4 and IL-5 were not detectable (<3/10⁶). Notably, the sensitivities of the IL-4 and IL-5 assays are equal to those of IFN-γ and IL-2 (20, 24). Vigorous peptide-induced TNF-α/β production (the Ab pair available does not distinguish between TNF-α and TNF-β) was also seen (data not shown), further substantiating the type 1 cytokine signature. The production of these cytokines was not cross-reactively triggered by any of the three test Ags, and was also not induced in naive mice (data not shown), suggesting specific detection of memory cells. These immunizations with CFA therefore induced highly polarized type 1 responses. Similar cytokine profiles were seen in the spleen and PC, with the exception of an additional IL-4 production in the spleen. This IL-4 was not produced by T cells, however, but by mast cells. IL-3 secreted by the Ag-activated T cells induced this bystander reaction (A. Y. K., manuscript in preparation).

To establish whether the PLP_139–151-induced type 1 cytokine measurements were obtained at single cell resolution, and to address whether these cytokines are produced in a cognate manner by the specific T cells (or also by involved bystander cells), we performed serial dilutions of primed cells on a fixed number of nonprimed APC. The number of primed dLN cells plated was linearly related to the number of IFN-γ and IL-2 spots detected (Fig. 1 A). Similar results were obtained when cells from spleens and the PC of PLP_139–151-primed mice were serially diluted (data not shown). Confirming published work (25), we found that the PLP_139–151-reactive cells resided in the CD4 cell fraction (data not shown). When freshly isolated bulk cell populations from primed mice were studied at high cell numbers (2.5 × 10⁵/well-10⁷/well),addition of APC did not affect the spot count (data not shown).Overall, the data suggest that Ag presentation in these compartments is saturating, and that accurate frequency measurements of PLP_139–151-reactive type 1 CD4 memory cells are made at single cell resolution in the absence of bystander cytokine production.

We next isolated mononuclear cells from the spinal cords of SJL mice with new onset, acute EAE induced by PLP_139–151 immunization. Vigorous production of IFN-γ and of IL-2 was induced by PLP_139–151 peptide (Fig. 1,B); PLP_178–191 did not induce cytokine over background at this early time point (see below). The signal-to-noise performance of the CNS assays approximated those of the LN (Fig. 1) with the exception of the first 2–3 days after the onset of clinical EAE, when the medium background in CNS isolates tended to be somewhat elevated compared with dLN. Image analysis showed that the size distribution of the IFN-γ and IL-2 spots generated by CNS isolates and dLN cells was comparable (data not shown). Also consistent with single cell resolution of the CNS measurements, serial dilutions of CNS cell isolates on a constant number of APC showed a linear relation for IFN-γ and IL-2 spot formation (Fig. 1 B). When the highest numbers of CNS-derived cells (5 × 10⁵ to 10⁷ per well) were tested with or without added APC, identical numbers of spots were seen (data not shown). These data suggest that the APC compartment in the inflamed CNS is fully competent in providing maximal stimulation/costimulation, revealing the accurate frequencies of PLP_139–151-specific IFN-γ- and IL-2-producing memory cells in the isolate in the absence of bystander reactions. The peptide-induced production of IFN-γ and of IL-2 was invariably accompanied by vigorous TNF-α/β spot formation in the CNS isolates, but this occurred over a high medium background, preventing accurate counting by image analysis. This finding is consistent with reports that activated microglial cells and infiltrating macrophages (27), rather than T cells, are the primary source of TNF in the CNS of mice with EAE (28). Also in accordance with this fact, and corroborating another report (29), we did not detect significant Ag-induced IL-4 or IL-5 production in such CNS recall assays at any stage of the disease. Therefore, cognate production of these type 2 cytokines by T cells did not seem to contribute to the disease.

We were concerned that the detected responses might derive from contamination by blood-borne CD4 cells as opposed to tissue-resident, CNS-infiltrating cells. Arguing against this possibility is the fact that PLP_139–151 peptide reactivity was not detected in CNS isolates before onset of EAE, although the chance for blood cell contamination at these time points should be comparable (Fig. 2, A vs B). Furthermore, the frequency of PLP_139–151-specific cells in the blood was 20–60 per million for IFN-γ- or IL-2-producing cells (H.H.H., manuscript in preparation), >10-fold lower than that seen in the CNS (Fig. 2 B). If blood contamination contributed to the results, the frequency of responses detected in the CNS isolates would be expected to be lower than that in the blood. Finally, in the case of blood contamination, the erythrocyte to CD4 cell ratio in the CNS isolate should be similar to that in the blood. We found that the ratio of CD4 cells to erythrocytes was ∼1:2 in the CNS material, while it was ∼1:3900 in the blood; hence, only 1 of every 2000 CD4 cells in the CNS isolate could be attributed to blood contamination.

We studied a total of 285 SJL mice with clinical disease at various time points after immunization with PLP_139–151 peptide. The basic findings reported in this work for this model were also reproduced in MBP_87–99-induced EAE of SJL mice and in myelin oligodendrocyte glycoprotein:35–55-induced EAE of C57BL/6 mice. First, we addressed whether, after priming, the neuroantigen-specific precursors directly migrate to the CNS, or whether they disseminate first in the immune periphery. Therefore, in addition to the dLN, we also tested the spleen, as well as the PC, representing a nonlymphoid compartment. The data for day 9 after immunization are shown in Fig. 2,A. High frequency IFN-γ-, IL-2 (and TNF-α/β)-producing cells were present in the dLN, the spleen, and the PC. At this time point, no cytokine-producing cells were detected in the CNS. The PLP_139–151-specific cells in the PC became detectable on day 3 after immunization, and their numbers peaked on day 9–12. From this point, their numbers gradually declined (Figs. 2, B–D, and 3). Therefore, in addition to the lymphoid tissues, a considerable reservoir of effector cells resided in extralymphoid tissues, as suggested by their high frequency in the PC; a considerable effector cell mass builds up in the periphery before the neuroantigen-specific T cells start to accumulate in detectable numbers in the CNS.

Clinical EAE developed at various time points after immunization, typically between days 11 and 20. With the onset of paralysis, PLP_139–151-specific T cells also became detectable in the CNS at high frequencies (Fig. 2,B). When the actual numbers of the neuropeptide-specific cells were calculated, however, it was found that the vast majority continued to be present in the periphery (Fig. 3). Their presence in the CNS vs the periphery did not seem to reflect compartmentalization of subpopulations according to pathogenicity (type A vs type B determinant specificity) (30, 31), because, reproducing data of others (7), we also found that the cells from the periphery could readily mediate EAE in adoptive transfers (data not shown). When the functional avidity of the PLP_139–151-specific cells was tested in the CNS vs lymphoid tissues by measuring the peptide concentration that induces 50% of the maximal spot numbers, no significant differences were seen at any stage of EAE (A. Y. K., manuscript in preparation). Therefore, neuroantigen-specific cells with higher avidity are not selectively retained in the CNS, while those with lower avidity continue to recirculate in the immune periphery. Moreover, because the absolute number of PLP_139–151-reactive T cells in the periphery continued to outnumber by far those in the CNS throughout the course of EAE (Fig. 3), it appears that the disease-mediating effector cell pool in the CNS draws from an extensive peripheral reservoir of effector cells.

The frequencies of PLP_139–151-specific cells in the periphery were highest before the onset of EAE and gradually declined thereafter (Fig. 2). This decline was also reflected in the absolute numbers of PLP_139–151-reactive cells in the periphery (Fig. 3). In the CNS, the frequencies and absolute numbers of PLP_139–151-specific cells were highest at the onset of the first paralysis and declined with the duration of the disease (Figs. 2 and 3). Invariably, for all mice tested individually later than 70 days postimmunization (n = 27), the reactivity to PLP_139–151 completely disappeared from their CNS, while low numbers of peptide-reactive cells continued to be present in the spleen (Figs. 2 D and 3). This overall decline in the numbers of the first wave effector cells, and their ultimate disappearance from the CNS was a function of time: it occurred irrespective of the number of relapses in this time frame, and was also seen in the two other EAE models studied (data not shown). As stated above, the cytokine signature of the PLP_139–151-specific T cells maintained its type 1 characteristic throughout the observation period, suggesting that the eventual decline of type 1 activity is not a consequence of type 2 switching, but rather reflects the drop in absolute numbers of the first wave T cells.

The above data suggest that the PLP_139–151-specific effector cells become exhausted in the course of EAE induced by immunization with this peptide, yet the mice continued to exhibit chronic paralysis. Thus, we were interested in testing whether determinant spreading had occurred, which could explain why the disease progressed. We tested mice during the first paralysis, and during the second episode of EAE for reactivity to PLP_178–191 (spreading from PLP_139–151 to 178–191 has been shown to be essential for establishment of chronic EAE (15, 26)). Reactivity to PLP_178–191 was systematically tested in the CNS, the spleen, the dLN (LN that drain the site of immunization), and the superficial cervical LN (that might be dLN for the brain). In four independent experiments, PLP_178–191-reactive cells were detected in the CNS during the secondary paralysis, while the peripheral lymphoid tissues did not show such responses (Table II). The data argue for determinant spreading to occur in the inflamed target organ itself. Confirming previous studies (15, 16), at later time points the response to PLP_178–191 emerged in the spleen (day 72, data not shown). While the natural history of the second wave response is beyond the scope of this study, these data support the notion that determinant spreading provides a mechanism for the persistence of the chronic disease after the first wave of effector cells is exhausted.

Enabled by a new generation of ELISPOT assays, we followed the natural history of the autoimmune T cell response to PLP_139–151 in EAE induced by immunization with this peptide, addressing the clonal sizes and cytokine signatures of these first wave effector cells in the immune periphery vs the CNS itself.

There is renewed interest as to what extent neuroantigens expressed in the thymus and the immune periphery contribute to the shaping of the preimmune T cell repertoire (1, 2, 32, 33). In particular, the PLP isoform that is expressed in the thymus of SJL mice does not include the region that encodes the 139–151 peptide, and subsequently precursor T cells specific for this peptide do not undergo negative selection, but occur in relatively high frequencies in the immune periphery (32, 33). Moreover, it was shown that in unimmunized mice, these T cells are driven by the endogenous Ag into a preactivated/memory state (33). The cytokine signature of these precursor cells remained unresolved, however. The activation of cytokine genes as required for T cell differentiation into an effector cell requires extensive proliferation to open the chromatin structure (23). We did not detect PLP_139–151-induced cytokine (IL-2, IL-4, IL-5, or IFN-γ)-producing cells in unimmunized SJL mice. Therefore, the PLP_139–151-specific precursor cells do not seem to have undergone cytokine differentiation in unimmunized hosts, and hence do not qualify as effector cells.

The extent of engagement of cytokine-producing specific effector cells after immunization was striking. By day 9, the frequency of PLP_139–151-specific IFN-γ-producing CD4 cells in dLN and spleen alone increased from <1/10⁶ (Table I) to ∼200/10⁶ in the dLN, and ∼600/10⁶ in the spleen (Fig. 2,A). When corrected for absolute numbers, this translates into an increase from <100 memory cells from the preimmune state to a total of ∼75,000 cells in dLN and the spleen. These numbers are still likely to underrepresent the magnitude of clonal expansion, because the memory cells are thought to disseminate early into extralymphoid tissues such as the lung, the liver, the gut, the skin, and the bone marrow, these being preferential sites for memory cell homing (8). Indeed, studying the PC, we found high frequencies of peptide-specific cells on day 9 after the immunization (Fig. 2 A). Provided the PC is representative of other extralymphoid sites from which isolation of lymphocytes still awaits new techniques to be developed, this finding suggests that a considerable effector cell mass resides in extralymphoid tissues. While this clonal expansion of CD4 cells engaged by immunization is impressive, it does not reach the dimensions of CD8 cell responses in viral infections, in which as many as 10–20% of all T cells can be stained with the specific tetramer (34, 35, 36, 37).

While the total number of PLP_139–151-specific T cells peaked in lymphoid tissues before the onset of clinical EAE (Fig. 3), and while high frequencies of PLP_139–151-specific cells were present in the extralymphoid PC at these early time points, we could not detect these cells in the CNS itself before the actual onset of the disease (Figs. 2 A and 3). After becoming primed in the dLN, the neuroantigen-specific T cells seem to disseminate throughoutthe entire organism, as well as seeding into the PC. Being T cell blasts, these lymphocytes should also be able to cross the blood-brain barrier and to randomly seed in the CNS. Most of these PLP_139–151-specific T cell blasts end up in the CNS parenchyma, where they, like T cell blasts of all specificities that enter there, are prone to undergo apoptosis (11). While the PLP_139–151-reactive T cells, therefore, are initially likely to randomly migrate to the brain and to other sites of the body, apoptosis of these cells in the immune-privileged CNS may explain why, at early time points, functional T cells can be detected only in the immune periphery, and not the CNS.

During EAE, T cells accumulate in the perivascular space, which is of mesenchymal origin. As opposed to the parenchyma, this tissue does not predispose T cells to apoptotic cell death (11). The neuroantigen-specific T cells that end up in this compartment will be selectively retained there as the consequence of autoantigen recognition (38). The production of cytokines and chemokines by these (initially few) T cells will lead to the initiation of a local inflammatory reaction, promoting the further recruitment of blood-borne lymphocytes and macrophages. Thus, it has been shown that the development of the perivascular infiltrate in EAE depends on the activation of microglia by IFN-γ-producing T cells, resulting in TNF-α and chemokine production by the microglia (28, 39), and the induction of adhesion molecules (VCAM and ICAM) that facilitate traffic of mononuclear cells across the endothelium. The time required for the active development of this inflammatory infiltrate in the CNS may explain why these neuroantigen-specific T cells accumulate with a considerable delay in the target organ itself relative to their presence in extralymphoid sites.

It has been a matter of controversy how many of the CNS-infiltrating T cells are neuroantigen-specific vs bystander cells with different specificity. For early parenchymal lesions, up to 50–80% specific cells have been reported (40, 41). Other studies found the specific cells to be a minor fraction of the CNS infiltrate (42, 43, 44). These differences might be related to the stage of EAE studied. As opposed to the requirement for T cells to be in the blast stage for them to enter the noninflamed CNS, in the chronically inflamed CNS, under the local influence of lymphotoxin, lymphoid structures develop that facilitate the general recruitment of T cells (12). Our data suggest that at any stage of active EAE, functional neuroantigen-specific T cells are a minority (<1 in 1000 cells, Fig. 2 B) within the infiltrate. Functional measurements such as the one we performed, even if made at single cell resolution, will miss cells that are anergized or undergo apoptosis. There might be a higher number of PLP_139–151-specific cells present in the early infiltrate, but in this case, these cells do not seem to be functional effector cells.

Calculating the total numbers of PLP_139–151-specific T cells in the CNS vs those in the peripheral lymphoid tissues, we found that, throughout EAE, less than 20% of these cells resided in the CNS (Fig. 3). We propose that the presence of PLP_139–151-specific cells in these two compartments reflects general migrational properties of memory T cells rather than the selective compartmentalization of subpopulations. We base this presumption on three observations. First, the cells isolated from the immune periphery of mice with EAE can adoptively transfer the disease. This provides evidence that they have the migrational properties to enter the CNS, they have the specificity to recognize the endogenous (type A) conformation of the peptide (30, 31), and they secrete the pathogenic set of cytokines (which could extend beyond those that we measured). Indeed (secondly), when the cytokine signatures of the PLP_139–151-specific cells were directly compared, they were identical for the CNS, the dLN, the peritoneal lavage cells, and the spleen (the apparently peptide-specific IL-4 production in the spleen was found to be a cytokine-driven bystander reaction). Third, the peptide dose that induced 50% maximal spot numbers in the CNS isolates vs splenocytes was identical, suggesting comparable functional avidity spectra for PLP_139–151-specific cells in both compartments. This strongly argues against the retention of high affinity clones in the CNS or their preferential expansion/exhaustion by the autoantigen in the CNS in the course of the disease. The vast majority of effector cells, therefore, seems to constitute a reservoir in the periphery, from which the autoimmune process in the target organ draws. This reservoir is largest before the onset of the disease, waning over 2–3 mo (Fig. 3). One explanation for this observation is that the effector cells enter the CNS at the inflamed perivascular cuffs, from where they migrate into the parenchyma. There they recognize Ag, possibly become anergic, and undergo apoptosis (45, 46). This continuous depletion of the effector cells in the CNS results in the drainage of the peripheral effector cell pool until it eventually exhausts (Fig. 3). Because the PLP_139–151-specific T cells maintained the IFN-γ⁺, IL-2⁺, TNF-α/β⁺ cytokine signature over the entire observation period without converting to IL-4 and IL-5 production, we conclude that the loss in numbers of cells producing the pathogenic type 1 cytokines does not reflect type 2 switching.

The prevalence of type 1 cytokine activity in the CNS during the entire course of EAE has also been observed by others (28, 29). The data are consistent with studies on IL-4 knockout mice, suggesting no critical role for this cytokine in the course of EAE (47, 48). Elevated IL-4 levels were measured in the CNS by PCR-based methods, however (17, 49). PCR-based methods do not provide information on frequencies of cytokine-producing cells, and do not excel in providing quantitative results. Measuring actual frequencies, we found that the cells that produce IFN-γ, IL-2, and TNF-α/β outnumbered at least 200-fold those producing IL-4 or IL-5. Type 2 T cell activity by the first wave of effector cells, therefore, does not prevail at any stage of EAE.

While our monitoring of the first wave effector cells showed that they are exhausted within 2–3 mo, the mice continued to exhibit clinical signs of EAE. It seemed conceivable, therefore, that a secondary wave of effector T cells, engaged by determinant spreading, mediated the later stages of the disease, as proposed by the “dynamic autoimmune repertoire hypothesis” (14, 16). In the model we studied, spreading to PLP_178–191 has been shown to occur, and is in fact required for disease progression (15, 26). Confirming these findings, we also observed spreading to the PLP_178–191 peptide in PLP_139–151-induced EAE. It has been a matter for debate, however, where determinant spreading occurs. Our data seem to suggest that second wave autoimmunity becomes engaged in the inflamed target organ itself. We detected reactivity to PLP_178–191 in the CNS before this response became detectable in the periphery (Table II). The development of the aforementioned lymphoid structures in the inflamed perivascular compartment of the CNS might be required for the spreading reaction to occur and might also explain the delay with which it occurs. Cervical LN are also candidates for the site of second wave priming. We tested superficial cervical LN and we did not detect responses there. We did not succeed in isolating deep cervical LN. From these (Fig. 2) and our previous studies (20, 25) of the kinetics and topography of the T cells primed in LN, we learned that responses in the spleen follow those in the dLN with only 1–2 days of delay. Moreover, when priming occurred in the periphery, the responses became detectable in the periphery (including the spleen) before the CNS (Fig. 2). Therefore, the fact that the second wave response to PLP_178–191 was detected in the CNS before it was seen in the periphery (Table II) does not prove, but argues for engagement of determinant spreading in the target organ itself. This notion may explain why, frequently, cryptic determinants are targeted (13). Involving different types of APC that are, in addition, activated by the chronic local T cell-mediated inflammation, the determinant hierarchy in the inflamed CNS is likely to fundamentally differ from that displayed following Ag presentation by the resting APC lineages present in lymphoid tissues or the blood (14).

In summary, while semiquantitative T cell proliferation assays (which assess primarily IL-2-driven bystander cell proliferation (50)) suggested early on that autoimmune responses might be dynamic, involving exhaustion and spreading (13, 14, 16, 51), in this work we report the first direct monitoring of the autoreactive T cell pool at single cell resolution. These results provide direct evidence for a dynamic autoimmune response. Monitoring the frequency and cytokine signature of the first wave of effector cells in the periphery and the target organ itself during the course of EAE, we provide evidence for a large peripheral reservoir of memory/effector cells that outnumber by far those in the target organ. This reservoir seems to fuel the inflammatory process in the CNS, leading eventually to the exhaustion of the primary effector cell pool while engaging second wave autoimmunity in the CNS itself.

We thank R. Trezza and T. Ansari for excellent technical help, Earl R.Sigmund for editorial assistance, and Drs. P. Heeger and M. Tary-Lehmann for valuable discussions.