Immune Reactivity in the Nervous System: Modulation of T-lymphocyte Activation By Glial Cells

The brain, spinal cord and peripheral nervous system are the sites of functions that are the basis of human mind and behaviour. An organ system with such an elaborate function, would consequently be expected to have commensurately demanding housekeeping requirements. This seems indeed to be the case for the supply of nutritives, clearance of metabolites, structural support and compartmentalization, and ion equilibrium, to list just a few functions contributed by glial cells: the non-neural components of the nervous system.

Paradoxically, however, the tissues of the nervous system have been thought to be neglected by the surveillance mechanisms of the immune system. Lack of lymphatic drainage, the tight endothelial blood–brain barrier (BBB), lack of major histocompatibility complex (MHC) antigens and decreased graft rejection responses have been quoted in support of this concept (Wekerle, Linington, Lassman & Meyermann, 1986). In this article, a series of experiments will be discussed, the results of which suggest that the nervous system is not inaccessible to the immune system. An important conclusion will be, that glial cells, particularly astrocytes, have the capacity efficiently to interact with migrating immune cells. Astrocytes, are able not only to activate circulating T-lymphocytes but, in addition, efficiently to down-regulate these cells to form key elements at a regulator interface between the immune and nervous systems.

One way to study directly the immune cell circulation within the CNS parenchyma, is to use permanent lines of T-lymphocytes that recognize monospecifically defined molecules on CNS structures.

T-lymphocytes, selected for reactivity against myelin basic protein (MBP) (a major structural protein of CNS myelin) are especially suitable immunological probes. First, it has been known for many years that immunization of inbred Lewis rats with MBP will lead to autoimmune inflammation of the CNS, experimental autoimmune encephalomyelitis (EAE) (Roboz-Einstein, Robertson, DiCaprio & Moore, 1962), and that this disease can be transferred to naive syngeneic recipient rats by immune lymphocytes, but not by humoral factors (Paterson, 1960). Second, myelin basic protein is an autoantigen which has been thoroughly characterized at molecular and cellular levels. Its sequence is known in several species (Eylar & Jackson, 1974) and its gene has recently been cloned (Roach et al. 1983). Third, the immune response against MBP is under the strict control of immune response (Ir) genes in the rat (Williams & Moore, 1973; Gasser, Palm & Gonatas, 1975; Günther, Odenthal & Wechsler, 1978) and in the mouse (Fritz et al. 1985).

MBP-specific T-lymphocyte lines have been established both from pre-immunize Lewis rats (Ben-Nun, Wekerle & Cohen, 1981) and from unprimed, normal Lewis lymph node cells (Schluesener & Wekerle, 1985). All Lewis rat T-lines selected for reactivity against intact MBP share important properties. First, they all express a membrane phenotype which is typical for the T-helper subset of T-lymphocytes: they bind the monoclonal marker antibody W3/25, but exclude antibody OX8 (Mason et al. 1983). Then, all these T-lines are exclusively specific for ‘their’ selecting antigen, MBP. Furthermore, they recognize MBP in the molecular context of major histocompatibility complex class II antigens, la determinants (Fierz et al. 1985. Finally, T-lines, selected for reactivity against the intact MBP molecule, are not specific for rat epitopes that are randomly positioned on the MBP molecule. In the Lewis rat, all MBP-specific T-lines are reactive against one circumscript molecular epitope represented by amino acid sequence 66–68 (Sun & Wekerle, 1985. This epitope has been defined as the main encephalitogenic region for the Lewis rat (Kibler et al. 1977). In the SJL mouse, the encephalitogenic MBP region of the SJL strain is located within amino acid sequence 89–169, whereas PL mice preferentially respond to sequence 1–37 (Fritz et al. 1985). Thus, individual immune systems which are controlled by distinct Ir gene sets will recognize individual epitopes on a given myelin autoantigen.

Lewis rat-derived MBP-specific T-lines are not only autoreactive, but auto-aggressive. In vitro, activated MBP-specific T-line cells are able to mediate violent, acute CNS-specific immune inflammation. Typical clinical and histological autoimmune encephalomyelitis (EAE) can be transferred to normal syngeneic recipient rats with cell doses as low as 1×10⁴cells/recipient (Schluesener & Wekerle, 1985). In these cases, EAE is relatively mild and the recipient will recover spontaneously. At higher doses, starting at 1–2× 10⁶cells/recipient, the disease will take an unfailingly lethal course.

The course of transferred EAE is peculiar. Only after a symptom-free period of 3 days does the treated host develop limb paralysis, which is initially restricted to the tail and then ascends cranially via hind- and forelimbs. Usually, within 5–6 days, the encephalitic rat will fall into stupor and die. The extremely acute course of target tissue destruction has been followed semiquantitatively by electrophysiological methods in T-line-mediated experimental autoimmune neuritis (EAN), a system sharing its fundamentals with T-line-mediated EAE. In EAN, T-lines are selected for recognizing the P2 protein of peripheral nerve myelin (Linington et al. 1984). Like MBP, the P2 molecule is a basic protein located at the cytoplasmic face of the myelin membrane (Omlin, Webster, Palkovits & Cohen, 1982). Lewis rats respond against a defined molecular sequence, determined by Uyemura et al. (1982). In a manner similar to MBP-specific T-lines, injection of activated P2-specific T-cells does not lead to clinical symptoms within the first 3 days post-injection. Continuous monitoring of the transmitting function of sciatic nerves revealed that from day 3, nerve function was abruptly reduced; within 12h all functions were lost. Complete recovery of these nerves took at least 10 weeks (Heininger et al. 1986). The electrophysiological development was faithfully paralleled by clinical paralysis (Linington et al. 1984) and the development of typical mononuclear infiltration (Izumo, Linington, Wekerle & Meyermann, 1985).

The fact that lethally autoaggressive T-lymphocyte lines can be isolated from completely normal immune cell populations has profound immunological implications. Assuming phenotypic stability of the recombined gene of the T-cell receptor variable region (Ikuda, Ogura, Shimizu & Honjo, 1985; Leiden & Strominger, 1986), one can conclude that the antigen receptor specificity of the T-line cells is identical with that of their precursors in the initial unselected lymphocyte population. This strongly supports the concept that the normal T-cell immune repertoire contains potentially autoaggressive T-cell clones as normal components. Consequently, suppressive regulation rather than physical elimination of autoaggressive T-lymphocyte clones should be the basis of immunological self-tolerance.

The histological lesion arising during T-line-mediated EAE cannot be distinguished from that seen in acute EAE induced by active immunization. In both models, the key morphological features are mononuclear infiltrations predominantly concentrated around post-capillary CNS vessels. In addition, some diffuse infiltration is seen in the adjacent areas (Vass, Lassman, Wekerle & Wisniewski, 1986). The second main feature of the EAE lesion is the activation of astrocytes. Demyelination (a hallmark of multiple sclerosis and chronic EAE) is, however, inconspicuous and restricted to areas close to acute infiltration (Lassmann, 1983). Immunocytochemistry has demonstrated that most, if not all, of the infiltrating lymphocytes are T-cells of the same subset as the injected T-lines. Nevertheless, transfers of isotope-prelabelled encephalitogenic T-line cells, and autoradiographic localization of the injected cells within the inflammatory lesions, revealed that only a small minority of the infiltrating T-cells were indeed the progeny of the injected pathogenic T-cells. The overwhelming majority, however, must have been contributed by the host immune system (Wekerle et al. 1986).

EAE has been thought to result from mechanisms involving delayed type hypersensitivity (DTH). These are based on complex interactions between immune cells of different subsets, on the one hand, and macrophages and other accessory cells on the other. Is the recruitment of host-derived inflammation thus essential for EAE development? We used lethally irradiated (750 R) Lewis rats as immunosuppressed T-line recipients to answer this question. These animals are severely depleted both of their own immune cells and of their haemopoietic elements, and unless restored with exogenous haemopoietic stem cells, will die within 2–3 weeks of anaemia or infections. In these experiments, lethally irradiated Lewis rats were injected with sublethal doses of MBP-specific T-line cells and were scored daily for clinical symptoms (Fig. 1). They were compared to unirradiated, control animals that had been treated with the same dose of activated T-line cells. Remarkably, both animal groups developed EAE and in both the course of the disease was practically identical. Like unirradiated controls, irradiated animals showed a lag phase of 3 days preceding onset of clinical symptoms. Thereafter, neurological signs developed abruptly in both groups. Most unexpectedly, however, the T-cell-mediated disease was self-limiting, even in irradiated animals. All animals recovered from neurological symptoms by day 11.

Irradiated rats receiving lethal doses of MBP-specific T-cells showed morphological lesions with interesting histological patterns. In contrast to the CNS lesions of unirradiated control animals, EAE lesions in the spinal cords of irradiated rats lacked completely the dense perivascular infiltrations. Instead, scattered mononuclear cells, presumably activated T-lymphocytes, infiltrated the adjacent parenchyma as when EAE was transferred by primed T-cells into irradiated rats (Sedgwick, Brostoff & Mason, 1987) (Fig. 2). Also, in contrast to the normally simplified pattern of cell infiltration, there was a severe defect of the BBB as reflected by multiple focal bleedings through CNS blood vessel walls. These and severe oedema are the main morphological characteristics of severe clinical changes.

Development of T-line-mediated EAE thus does not depend on the dense perivascular and parenchymal infiltrations by host-derived cells. In contrast, it appears that the encephalitogenic T-line cells are autonomous in causing the CNS defects of EAE. We assume that these defects are caused by direct cytotoxic interactions with local glial cells (Sun & Wekerle, 1986). Moreover, these experiments indicate that migration of host-derived ‘suppressor cells’ into EAE lesions, which occurs predominantly in the late phases of the disease (Hickey & Gonatas, 1984), is not critical in terminating EAE. As will be shown below–as an interesting alternative - local glial cells may be involved in down-regulation of the local T-cell response.

One of the most impressive aspects of T-line-mediated autoimmune disease is its tissue specificity. MBP-specific T-lines transfer an autoimmune disease which is exclusively restricted to the CNS, whereas inflammation mediated by P2-specific T-lines affects the peripheral nervous system (PNS) but not adjacent CNS tissues (Izumo et al. 1985; Wekerle et al. 1986). Since in transferred EAE the pathogenic T-lymphocytes are injected intravenously and subsequently circulate through the blood vascular system, the BBB endothelium, which separates the circulation from the CNS parenchyma, can be expected to play a critical role in target retrieval by the encephalitogenic T-cells. CNS endothelia could thus present MBP molecules (in the context of la determinants) on their luminal surface, and these autoimmunogenic molecular complexes could act as a target signal for circulating T-line cells. Endothelial la has been reported by several groups (Traugott, McFarlin & Raine, 1986; Sobel, Natale & Schneeberger, 1987), although others (Matsumoto, Hara, Tanaka & Fujiwara, 1986; Vass et al. 1986) have been unable to confirm this (Fig. 3). In vitro assays have shown that cells from CNS microvessel preparations have the capacity to present antigen to primed T-cell populations (McCarron, Kempski, Spatz & McFarlin, 1985), whereas pure, clonal populations of rat CNS endothelia were completely inert (H. Wekerle & W. Risau, unpublished results).

There seems to be an alternative, antigen-independent pathway, leading T-lymphocytes through the BBB. Activated, but not resting T-cells, have the capacity to break through CNS endothelia and to invade the parenchyma irrespective of their antigen specificity.

It has been known for years that T-line cells mediate EAE with stunning efficiency, provided they are preactivated in vitro (Schluesener & Wekerle, 1985). In contrast, resting MBP-specific T-line cells will not transfer disease, even if injected in extremely high doses (Naparstek et al. 1983). Activation-dependent, but antigenindependent, BBB transmigration by T-lymphocytes was formally documented using [¹⁴C]thymidine-labelled T-line cells of different antigenic specificities and by their subsequent autoradiographic localization in CNS tissues. We found that a small but constant number of MBP-specific activated T-lymphoblasts appeared in the brain parenchyma within 24 h post-injection. During the following 48 h their proportion remained roughly constant. At the beginning of neurological symptoms (approx. 3 days post-injection) a second, massive wave of cellular invasion occurred. This led to the immigration of millions of unlabelled host-recruited immune cells. Surprisingly, the first (but not the second) peak of invasion was also seen after injection of isotope-labelled, activated T-line cells which were specific for ovalbumin (a model antigen for foreign determinants which do not occur naturally within the host nervous system). Again, labelled ovalbumin-specific T-cells were demonstrable by 24 h post-injection (Fig. 4). These cells were diluted out during the following few days without any subsequent second invasion peak.

Activated, but not resting, T-lymphocytes, thus seem to have access to the CNS parenchyma, irrespective of their antigen specificity. There are interesting parallels between migration of normal T-lymphoblasts and the behaviour of brain-metastasizing tumour cells (Nicolson, 1984). In all these examples, migration through the BBB seems to involve a similar, stepwise mechanism. First, the cells must attach to the luminal side of the BBB endothelium. Then, due to a shared pattern of induced and secreted proteolytic and glycolytic enzymes, the elaborate system of intercellular tight junctions opens (Nagy, Peters & Hiittner, 1984) and, finally, the cells pass through the surrounding basal membranes (Naparstek, Cohen, Fuks & Vlodavsky, 1984). As will be discussed below, this migration pathway may be the basis for the specialized version of immune surveillance in the CNS.

If T-cells on their way to the CNS parenchyma do not identify their target organ as the endothelial level, the signal of target specificity must be at some stage beyond the BBB. Assuming that this signal depends on immunogenic presentation and recognition of the relevant (myelin) autoantigen, the normal CNS should contain, contrary to more prevailing beliefs, antigen-presenting cells (APC).

Astrocytes are likely candidates for more than one reason. First, one subset of astrocytes is tightly arranged around CNS blood vessels (Janzer & Raff, 1987), an ideal site for a strategically positioned APC, and second, astrocytes have been shown to interact intimately with immune cells. It has been claimed, for example, that they respond to several immune mediators, including interleukin 1 (Giulian & Lachman, 1985), and also that they can be induced to produce such factors (Fontana et al. 1982).

Astrocytes were tested for their potential antigen-presenting capacity by using purified astroglial cultures derived from perinatal Lewis rat brain. When such astrocyte monolavers were cultured together with svngeneic antigen-specific T-line cells, no conspicuous interactions were noted. However, immediately after addition of the relevant soluble antigen (MBP or ovalbumin) the specific T-cells aggregated around the astrocvtes, and within 24 h they were transformed to enlarged, strongly proliferating lymphoblasts (Fontana, Fierz & Wekerle, 1984). Detailed analyses have known that T-line cell activation was the consequence of immunospecific presentation and recognition of antigen. The reaction depended on the dose of antigen added, and antigen recognition was restricted by la determinants of the MHC.

Ia-restricted interactions were difficult to reconcile with the established observation that CNS parenchymal cells express little, if any, class I (Schachner & Sidman, 1973) and no class II MHC antigen (Hart & Fabre, 1981). Indeed, our cultured astrocytes were also negative for la antigens before presentation. However, when stained after interaction with lymphocytes, they had acquired la in considerable amounts (Fierz et al. 1985). Astrocytes, and a number of other cell types outside the nervous system, can be induced to synthesize and express la determinants either by mere contact with activated T-lymphocytes, as in our experiments (Fontana et al. 1984), or by T-lymphoblast-derived mediators such as interferon-y (Hirsch, Wietzerbin, Pierres & Goridis, 1983; Wong et al. 1985). These interactions not only induce la expression on the surface, but unlike many other cells, enable the astrocytes to take up, process and immunogenically present soluble antigen. Astrocytes thus qualify as inducible, facultative APC.

DOWN-REGULATION OF ACTIVATED T-LYMPHOCYTES BY ASTROCYTES

The finding that EAE (transferred by encephalitogenic T-lines to irradiated recipients) produced a monophasic, self-limiting course of the disease, suggested the existence of counter-regulatory mechanisms acting on the pathogenic T-lymphocytes which were independent of the host immune system. Local glial cells could also be involved in these interactions. Indeed, screening a panel of long-term, cultured and cloned astrocyte lines revealed immunosuppressive activity. These clones were derived from a standard astrocyte culture isolated from neonatal Lewis rat brain. After 8 months of bulk culture, the GFAP⁺ astrocytes were dissociated and cloned by limiting dilution techniques. All of these astrocyte sublines expressed GFAP in varying intensity and were inducible to express la. When screened for their capacity immunogenically to present auto- or foreign antigens to relevant T-line cells, the clones differed strongly in their APC capacity. This was documented most clearly in dose–response experiments using graded numbers of antigen-presenting astrocytes, with constant numbers of responding T-lymphocytes. Some clones were highly efficient at all cell densities, others presented antigen only at low cell densities, whereas a third group was completely ineffective. The incompetence of the latter to present antigen was not due to lack of la antigens on their surface, or to a defect in their capacity to present antigen, but rather reflected active suppression of T-cell activation. Non-reactive astrocyte clones, in contrast to antigen-presenting ones, were found strongly to suppress conventional T-cell activation by ‘normal’ APC (macrophages or dendritic cells from immune organs) plus antigen (Fig. 5). Preliminary evidence suggests that the suppressive activity is released into the culture supernatants. The molecular nature of the putative mediator is, however, unknown. Since these suppressive clones are fully active in the presence of indomethacin, prostaglandins do not seem to be involved in suppression. Possible candidates for suppressing activity are, however, either apolipoprotein E, which has been shown to be released by astrocytes upon interferon-y pretreatment (Fig. 6: Oropeza, Wekerle & Werb, 1987) or another, unrelated, astrocyte-derived suppressor factor (Schwyzer & Fontana, 1985).

Our findings that activated T-lymphocytes can cross the BBB and that astrocytes can be induced to up- and down-regulate the activity of intra-CNS T-lymphocytes have important implications. We have proposed that these observations reflect a specialized mode of immune surveillance in the CNS, which may have developed as an adaptation to the specific requirements of the vulnerable CNS tissues. Potential ‘bystander damage’ inherent to immune surveillance reactions is minimized by reducing the number of patrolling lymphocytes to the few activated T-cells in the blood circulation, and by focusing their response to perivascular areas. Moreover, the fact that only inducible APC can present antigen in the normal CNS may serve strictly to regulate and limit cellular immune responses (Wekerle et al. 1986).

The validity of this concept can be tested by examining its implicit predictions. Brain grafts, for example, should be accessible to activated, but not resting T-lymphocytes. This is indeed the case, for tissue transplants are promptly rejected after peripheral immunization with grafts of the same specificity (Head & Griffin, 1985). Also in graft-vs-host disease, T-cells have also been shown to invade CNS tissues (Hickey & Kimura, 1987). In addition, small, but significant, numbers of lymphocytes should be found in normal CNS tissue, and these T-cells should be in a (post-)activated state: predictions which have been confirmed immunocytochemistry and the use of appropriate lymphocyte markers (Booss, Esiri, Tourtellotte & Mason, 1983; Hafler et al. 1985). Furthermore, astrocytes, having interacted with (auto-)immune T-cells, should show at least some la antigens on their surface, as has been demonstrated in several cases of myelin-specific autoimmune diseases (Traugott et al. 1986; Hickey, Osborn & Kirby, 1985). The fact that macrophages and/or microglial cells are normally more intensively stained in la immunocytochemistry (Vass et al. 1986) supports the existence of marked counter-regulatory properties of astrocytes, as demonstrated in our present experiments.

Several fascinating, but so far unanswered, questions arise from the present observations. Apart from their role in immunosurveillance of the CNS, astrocytes that interact with CNS-patrolling lymphocytes could have physiological roles in modulating immune reactivity. After contacts with astrocytes, T-cells may ‘export’ CNS-derived stimuli to the peripheral immune system. Conversely, astrocyte may be modulated in their activity by contacts with local T-cells. It has been shown repeatedly that astrocytes respond to immune mediators (Fontana et al. 1982; Giulian & Lachman, 1985; Hirsch et al. 1983; Wong et al. 1985). Since our knowledge of the astrocyte function within the CNS is limited, the present interpretations of lympho-glial interactions remain speculative. Could human behaviour be conditioned by the immune status via the T-cell/astrocyte junction? Could general responses like alertness/sleep, temperature and muscular tonus be influenced by such mechanisms? More detailed information on lymphocyte/glia interactions will help to elucidate the pathogenesis of diseases, which so far have been enigmatic. Narcolepsy–a sleep disorder with a strikingly close association with immune response genes - (Seignalet, 1986) may be one of them.

There is no conclusive answer to these questions at present, but we are confident that an experimental approach to these problems will emerge.

We are grateful to all our colleagues who contributed to the research in our group and to Mrs B. Goebel, who typed this manuscript. This unit is supported by funds from the Hermann and Lilly Schilling-Stiftung.