CD8⁺ T Lymphocytes Mediate Borna Disease Virus-Induced Immunopathology Independently of Perforin

Abstract

Perforin-mediated lysis of target cells is the major antiviral effector mechanism of CD8⁺ T lymphocytes. We have analyzed the role of perforin in a mouse model for CD8⁺ T-cell-mediated central nervous system (CNS) immunopathology induced by Borna disease virus. When a defective perforin gene was introduced into the genetic background of the Borna disease-susceptible mouse strain MRL, the resulting perforin-deficient mice developed strong neurological disease in response to infection indistinguishable from that of their perforin-expressing littermates. The onset of disease was slightly delayed. Brains of diseased perforin-deficient mice showed similar amounts and a similar distribution of CD8⁺ T cells as wild-type animals. Perforin deficiency had no impact on the kinetics of viral spread through the CNS. Unlike brain lymphocytes from diseased wild-type mice, lymphocytes from perforin-deficient MRL mice showed no in vitro cytolytic activity towards target cells expressing the nucleoprotein of Borna disease virus. Taken together, these results demonstrate that CD8⁺ T cells mediate Borna disease independent of perforin. They further suggest that the pathogenic potential of CNS-infiltrating CD8⁺ T cells does not primarily reside in their lytic activity but rather in other functions.

The highly neurotropic Borna disease virus (BDV) is an enveloped virus with a single-stranded RNA genome of negative polarity that replicates and transcribes its genome in the nucleus of infected cells (5, 11). In horses and sheep it causes a nonpurulent meningoencephalitis and concomitant acute neurological disorder which has been termed Borna disease (BD) (25, 40, 47). A large number of warm-blooded animal species is susceptible to experimental infection with BDV (40). BDV is noncytolytic in vitro (18, 26) and in vivo (14, 41), and it has a strong tropism for neural tissue, where it can readily establish a persistent infection (9, 14, 15, 31). In naturally infected hosts and in experimentally infected rodents, neurological disease and behavioral abnormalities seem to result mainly from immunopathological processes (2, 15, 30). Strong perivascular and parenchymal infiltrations of CD4⁺ and CD8⁺ T cells were observed, and their appearance in the brain correlates with the onset of disease symptoms (30, 37, 49). Studies in the rat (3, 32, 46, 48) and mouse (15, 17) model systems and in naturally infected horses (2) indicated that immunopathology is mediated by CD8⁺ T cells which require help from the CD4⁺ T-cell subset. CD8⁺ T cells derived from the brains of diseased rats (36) and mice (17) have been shown to be cytolytically active against target cells infected with BDV or expressing BDV antigens. In adult infected rats, CD8⁺ T cells were shown to play a pivotal role in CNS tissue destruction resulting in cortical atrophy (3).

Perforin is considered to be the major cytotoxic effector mechanism of CD8⁺ T cells in antiviral immune responses (45, 51, 52). Together with granzymes A and B, it is the major component of the granule exocytosis pathway, which is one of the two pathways by which CD8⁺ T cells lyse target cells (44). In BDV infection of rats, increased expression of perforin mRNA was observed to correlate with an influx of CD8⁺ T cells into the brain (46), cytotoxic activity of brain-derived T cells, and elimination of virus in immunized rats (32). These kinetic data suggested a role for perforin in BDV-induced disease. However, no direct evidence has so far been available.

We sought to clarify the contribution of perforin to immunopathogenesis by introducing a mutated perforin gene into the genetic background of MRL mice. The MRL mouse strain is highly susceptible to BDV-induced neurological disease (15). Its high susceptibility is determined by the H-2k haplotype and by additional unidentified genetic traits. After five backcrosses of a mutated perforin gene to MRL mice, the effect of a heterozygous (+/0) or homozygous (0/0) perforin gene defect was tested by infecting these animals as newborns with a mouse-adapted variant of BDV as previously described (15). Mice were monitored for disease development for up to 8 weeks, and thereafter, the brains of the majority of these mice were histologically analyzed for lymphocytic infiltration and viral load. Mice lacking functional perforin developed neurological disease to the same extent as animals with an intact perforin allele (Fig. 1A). The number of animals succumbing to severe disease (score of 3) was relatively low in both groups when infection was done in newborns with the mouse-adapted BDV variant (Fig. 1A). Next, we infected juvenile MRL mice with a BDV variant that underwent an additional passage in newborn rat brain to increase the titer of the stock (stock no. 82). When these infection conditions were used, the number of severely diseased animals increased about threefold (Fig. 1B). Again, perforin 0/0, perforin +/+, and perforin +/0 mice showed similar disease susceptibilities (Fig. 1B and data not shown). However, a statistically significant difference was observed concerning the incubation time of BDV-induced neurological disease. When incubation times of mice whose disease scores are shown in Fig. 1A were compared, disease onset occurred 3 days earlier on average in perforin +/0 mice than in perforin 0/0 mice (34.3 versus 37.5 days on average; P = 0.03). In the second experiment, wild-type mice that were infected with virus stock no. 82 (disease scores of these mice are shown in Fig. 1B) also developed disease 4 days earlier than their no. 82-infected perforin-deficient counterparts (25.4 versus 29.7 days; P = 0.02). Thus, perforin might contribute to the kinetics of disease development.

FIG. 1.

BDV-induced neurological disease occurs in the absence of perforin. MRL mice harboring one (+/0) or both (+/+) intact alleles of the perforin gene or MRL mice lacking a functional perforin gene (0/0) were intracerebrally inoculated with a BDV variant passaged four times in the brains of 4-week-old mice (A) or with a BDV stock that underwent an additional passage in the brains of newborn rats (B). Animals were infected either as newborns (A) or at 11 to 17 days of age (B). They were monitored daily for disease symptoms. The severity of disease was scored on an arbitrary scale ranging from 0 to 3. 0, no symptoms; 1, low degree of ataxia, increased anxiety; 2, clear ataxia, torticollis, unphysiological and uncontrolled movements of extremities when the animal was held up by the tail, rough fur or hunched posture, characteristic position of hind limbs when animal was lifted by the tail (15); 3, pronounced weight loss, severe ataxia and torticollis, paraparesis, apathy, moribund. (C) Brain hemispheres of all animals shown in Fig. 1A were analyzed histologically for mononuclear inflammatory infiltrates. The severity of inflammation was scored on an arbitrary scale ranging from 0 to 3. 0, no infiltrates; 1, up to three perivascular infiltrates per brain section with one or two layers of cells, some mononuclear cells in meninges; 2, up to five perivascular infiltrates per brain section with multilayer appearance and spread into parenchyma, intermediate meningitis; 3, more than six perivascular infiltrates per brain section with multiple layers of cells and strong infiltration of parenchyma at multiple sites, strong meningitis.

The brains of BDV-infected perforin 0/0 mice were analyzed histologically for mononuclear infiltrates. They all showed clear CNS inflammation like perforin +/0 mice (Fig. 1C) or perforin +/+ mice (data not shown and Fig. 2A to D). The great majority of perforin 0/0 mice (80%) showed severe meningoencephalitis with a score of 3 (Fig. 1C), indicating that BDV-induced neuroinflammation does not depend on perforin-mediated cytotoxic activity. The percentage of animals with the strongest degree of CNS inflammation was always higher than the percentage of animals with severe neurological disorder in all groups (Fig. 1A and C). It therefore appears that the overall degree of inflammation does not strictly correlate with the clinical outcome. This suggests that inflammation has to occur at certain sites of the CNS to result in clear neurological disease, whereas inflammation at less critical sites might be clinically inapparent. On the other hand, clearly diseased animals uniformly had intermediate to strong meningoencephalitis, confirming previous results (15).

FIG. 2.

Distribution of T-cell subsets is similar in brains of perforin-expressing and perforin-deficient MRL mice. Perforin +/+ (A to D) and perforin-deficient (E to H) MRL mice were intracerebrally infected with BDV as newborns and sacrificed at the peak of BDV-induced neurological disease, approximately 4 weeks p.i. Sagittal cryosections of brain hemispheres were stained with monoclonal antibody RM4.5 against CD4 (A, E, C, and G) or monoclonal antibody 53-6-7 against CD8 (B, F, D, and H) (all monoclonal antibodies were from BD Pharmingen, San Diego, Calif.) diluted 1:500 in PBS–5% normal goat serum. Bound antibodies were visualized using a biotinylated secondary anti-rat immunoglobulin G antibody and the streptavidin-peroxidase conjugate detection system (Vector Laboratories, Burlingame, Calif.). Sections of the cerebellum (A, B, E, and F) and the neocortex (C, D, G, and H) are shown. Magnification, ×240.

To analyze the distribution of CD4⁺ and CD8⁺ T cells in the brains of perforin 0/0 mice, brain hemispheres from diseased animals were cryopreserved, and sections were stained with monoclonal antibodies to CD4 or CD8 antigens. In inflamed brains of perforin +/+ animals, CD4⁺ T cells were the main constituents of perivascular cuffs (Fig. 2A and C) and meningeal infiltrates (data not shown). In contrast, CD8⁺ T cells were present in significantly lower numbers than CD4⁺ T cells in perivascular cuffs and the majority of CD8⁺ T cells appeared to have entered the CNS parenchyma (Fig. 2B), a pattern which had also been observed in brains of horses naturally infected with BDV (2, 8) and in experimentally infected rats (3). The same pattern was found in perforin 0/0 mice (Fig. 2E and F), indicating that the lack of perforin did not result in a redistribution of T-cell subsets in BDV-induced meningoencephalitis. This pattern was most prominent in the cerebellum (Fig. 2A, B, E, and F) and was similar in other CNS regions, like the neocortex (Fig. 2C, D, G, and H), thalamus, and hippocampus. However, outside the cerebellum, more CD4⁺ T cells were found in the CNS parenchyma (Fig. 2C and G). Again, this picture was similar for perforin +/+ mice (Fig. 2C) and perforin 0/0 mice (Fig. 2G), demonstrating that the lack of the major lytic effector mechanism did not alter the distribution pattern of T-cell subsets.

Gross neuronal damage was not observed in either wild-type mice or perforin 0/0 mice with strong CNS inflammation, providing further support to the concept that lytic effector functions of CD8⁺ T cells do not play a prominent role in the development of clinical signs of disease. Experimental BD in the mouse model in this respect is similar to the natural disease in horses, where signs of neurodegeneration are also rare despite severe CNS inflammation (2, 14).

The kinetics of viral spread in the CNS of perforin +/+ and perforin 0/0 mice was compared in a time course using two mice each per time point starting from day 8 after infection until day 31 postinfection (p.i.). Animals were killed at the indicated times, and total RNA was isolated from one brain hemisphere and analyzed by Northern blotting using a BDV-specific probe (15) that detects the transcripts encoding the two most abundant BDV proteins in infected cells, the nucleoprotein p40 and the phosphoprotein p24. The first clear signals appeared on day 16 p.i. and then quickly increased to maximum levels between days 16 and 27 in perforin +/+ as well as perforin 0/0 mice (Fig. 3). Some animals of both genotypes showed reduced levels of viral transcripts after day 16 p.i. compared to the levels found in the majority of animals analyzed after this time point. Although some of these animals with reduced viral RNA levels showed severe disease at the time they were euthanatized, there is no strict correlation between acute disease and a drop in the abundance of viral transcripts (Fig. 3A). Taken together, these experiments showed that the kinetics of viral spread in the brains of perforin +/+ and perforin 0/0 mice were similar. At later times (up to 90 days p.i.), the amounts of viral transcripts in the brains of perforin +/+ and perforin 0/0 mice also remained comparable (data not shown). Thus, the lack of perforin did not allow faster spread of virus in the CNS, which might indicate that perforin plays no significant role in the putative ability of CD8⁺ T cells to restrict viral replication during the course of CNS infection.

FIG. 3.

BDV replicates with similar kinetics in brains of wild-type and perforin-deficient mice. Perforin +/+ mice (A) and perforin 0/0 MRL mice (B) were intracerebrally infected with mouse-adapted BDV at an age of 13 to 17 days and sacrificed at the indicated time points p.i. RNA from one brain hemisphere of each animal was prepared and analyzed by Northern blotting as previously described (41). Membranes were first hybridized to a radiolabeled rat GAPDH (glyceraldehyde-3-phosphate dehydrogenase) probe as loading control (bottom) and reprobed with a BDV-specific radiolabeled cDNA comprising nt 1 to 1873 of the BDV He/80 genome (top) to detect the major viral transcripts. Positions of BDV-specific transcripts of 0.8, 1.2, and 1.9 kb are indicated.

To directly demonstrate the lack of cytolytic activity of CD8⁺ T cells from the CNS of severely diseased perforin 0/0 animals, we tested them as effector cells in an in vitro cytotoxicity assay. Brain lymphocyte preparations from the brains of heavily diseased perforin +/+ and 0/0 animals yielded similar numbers of lymphocytes per brain (2 × 10⁶ to 4 × 10⁶), and the average percentage of CD8⁺ T cells varied between 20 and 40% in preparations from both types of mice as determined by flow cytometric analysis (data not shown). In a chromium-51 release assay with murine L929 target cells expressing p40, the major cytotoxic T-lymphocyte target protein of BDV, by a recombinant vaccinia virus, no lytic activity could be detected with lymphocytes from perforin 0/0 mice (Fig. 4A). In contrast, lymphocytes from +/+ mice efficiently lysed p40-expressing L929 target cells (Fig. 4A). Negative control L929 cells infected with a recombinant vaccinia virus expressing influenza A virus neuraminidase were not specifically lysed by either perforin +/+ or perforin 0/0 brain lymphocyte preparations (Fig. 4B). A lack of BDV-specific lytic activity by effector cells from perforin 0/0 mice was also observed on target cells pulsed with the recently identified H-2k-restricted immunodominant BDV-p40 epitope TELEISSI (42), whereas these cells were efficiently lysed by lymphocytes from perforin +/+ animals (Fig. 4C). L929 cells pulsed with the control H-2k-restricted peptide SEVLNQII derived from the reverse transcriptase of human immunodeficiency virus type 1 (HIV-1) were not lysed by either type of effector lymphocytes (Fig. 4D). Interestingly, the background level of lysis that we reproducibly observed when vaccinia virus-infected target cells were incubated with BDV-specific brain lymphocytes from perforin +/+ mice (Fig. 4B; 17) was absent when brain lymphocytes from perforin 0/0 mice were used (Fig. 4B). This indicates that the background chromium-51 release represented perforin-dependent cell lysis. Since this lysis occurred without presentation of BDV-specific antigens, natural killer (NK) cells that also lyse via perforin exocytosis might be responsible for the observed background lytic activity in brain lymphocyte preparations. Vaccinia virus infection has been shown to sensitize L929 cells for NK cell-mediated lysis (6), which could explain the fact that only vaccinia virus-infected target cells, and not peptide-pulsed target cells, showed background lysis (Fig. 4). NK cells have been found in the brains of BDV-infected, acutely diseased rats (16) and lysis of the NK target cell line YAC-1 by brain lymphocyte preparations of diseased MRL mice has been observed (data not shown).

FIG. 4.

CD8⁺ T cells from brains of perforin 0/0 mice exhibit no lytic activity in vitro. Lymphocyte preparations from three strongly diseased BDV-infected perforin +/+ mice (filled symbols) and three perforin 0/0 mice (open symbols) were tested individually ex vivo for cytolytic activity in standard 6-h ⁵¹Cr release assays using 4 × 10³ target cells per well. L929 (H-2k) target cells were infected with recombinant vaccinia virus expressing p40 (VV-p40) (A) or with a vaccinia virus recombinant expressing the N1 neuraminidase of influenza virus A/FPV/1/34 as an irrelevant control protein (VV-NA) (B). Alternatively, L929 target cells were pulsed with the immunodominant BDV-p40 peptide epitope TELEISSI (C) or with the irrelevant H-2K^k-binding peptide SEVLNQII derived from the reverse transcriptase of HIV-1 (D).

The classical model to study CD8⁺ T cell-mediated immunopathology in the CNS is the intracerebral infection of mice with the noncytopathic lymphocytic choriomeningitis virus (LCMV) (7, 12). LCMV mainly infects ependymal, choroid plexus, and meningeal cells after intracerebral infection and causes a strong lymphocytic infiltration in the CNS of immunocompetent mice. When perforin-deficient mice are infected with LCMV, a marked cerebral infiltration of CD4⁺ and CD8⁺ T cells is also observed, but the mice do not develop acute neurological disease (20, 21). Thus, perforin-mediated lysis seems to be the major pathogenetic mechanism in lymphocytic choriomeningitis but not in BDV-induced neurological disease.

CNS infection with the neurotropic JHM strain of mouse hepatitis virus represents another model for virus-induced encephalomyelitis (29). Direct cytotoxic effects of viral replication do not seem to play a major role in JHM-induced pathogenesis (19). Rather, various immune cell subsets contribute to neurological disease after intracerebral JHM virus infection (19, 24). With this system, it was found that the overall mortality of mice was similar in the presence or absence of perforin although mortality was moderately delayed in perforin-deficient animals, suggesting some contribution of perforin to fatal encephalitis (24). Moreover, encephalomyelitis and demyelination following subacute infection was not prevented by a lack of perforin (24), indicating that other immune effector mechanisms were operative in JHM virus pathogenesis. The role of perforin in intracerebral infection of mice with Theiler's virus, a murine picornavirus, is more complex in that it exhibits disease-promoting and disease-restricting properties depending on the phase of infection. Perforin-mediated cytotoxicity is beneficial during acute infection. It promotes H-2D^b-restricted viral clearance from the CNS (39). Mice lacking perforin die within 18 days p.i. with high viral loads and strong inflammation of the CNS (39). Direct viral cytotoxicity does not play a role in this outcome since major histocompatibility complex class I-deficient mice develop a persistent infection without neurological deficits (27, 38). Demyelination-associated disease, which develops in the late phase of Theiler's virus infection under certain conditions, is dependent on perforin. In mice lacking perforin, neurologic injury is much less pronounced than in their wild-type counterparts (28). It thus appears that the mechanism of BDV-induced neurological disease is clearly different from the T-cell-mediated immunopathology models based on LCMV or Theiler's virus. There is more similarity to the pathogenesis following cerebral JHM virus infection; however, there is no dominating role of CD8⁺ T cells in JHM virus-induced immunopathology (23).

The aforementioned findings raise the question about other possible CD8⁺ T-cell effector mechanisms in BDV-induced CNS disease. The second major effector mechanism of CD8⁺ T cells is upregulation of Fas ligand surface expression to induce Fas receptor (CD95)-mediated cell death in Fas-expressing target cells (45). Apart from the fact that Fas expression on neurons might be limited (10, 33), the major role of the Fas system seems to reside in T-cell development and in T-cell homeostasis during the course of an immune reaction rather than viral clearance (50, 51). It thus seems likely that the Fas/FasL system does not significantly contribute to BDV-induced neurologic disease. This view is supported by the results of a recent pilot study which showed that MRL-lpr/lpr mutant mice lacking functional Fas are highly susceptible to BDV-induced neurological disorder (unpublished data). Moreover, it has been shown that the Fas/FasL system plays no role in the partly CD8⁺ T cell-mediated CNS disease induced by JHM virus (35). A third important effector mechanism of CD8⁺ T cells is secretion of cytokines, predominantly gamma interferon (IFN-γ) and tumor necrosis factor alpha. It is unclear whether BD is indeed caused by the deleterious action of cytokines released by activated T cells. It has been shown that IFN-γ contributes to viral clearance in the CNS in the case of Sindbis virus, recombinant vaccinia virus, and JHM virus (4, 22, 34). In a murine model of rotavirus infection, CD8⁺ T cells were shown to clear rotavirus from mice lacking perforin, Fas, and IFN-γ (13), indicating that apart from the known mechanisms there may exist further CD8⁺ T-cell effector pathways that could mediate either protection or immune damage. Possible candidates are ligand/receptor systems like the IFN-γ-modulated TRAIL apoptosis system (1, 43).

We have demonstrated in the present study that BDV-induced CD8⁺ T cell-mediated neurological disease occurs in the absence of perforin, which is considered the major lytic mechanism of antivirally active CD8⁺ T cells. Extensive application of the knockout approach will further help to elucidate the CD8⁺ T-cell effector functions mediating BDV-induced CNS pathology and possibly also BDV clearance.