Perforin and Gamma Interferon-Mediated Control of Coronavirus Central Nervous System Infection by CD8 T Cells in the Absence of CD4 T Cells

Mice.

SCID and syngeneic BALB/c (H-2^d) wt mice were obtained from the National Cancer Institute (Frederick, Md.) when they were 6 weeks old. Homozygous GKO BALB/c mice at the 7th backcross generation were provided by Robert Coffman (DNAX Research Corporation, Palo Alto, Calif.). Homozygous PKO BALB/c mice were generated in our laboratory as previously described (5) and were used at the 11th backcross generation. Both GKO and PKO mice were maintained by intercrossing and were used at 6 to 7 weeks of age. SCID, GKO, and PKO mice were all maintained under sterile conditions.

Virus.

The neutralizing monoclonal antibody (MAb)-derived 2.2v-1 variant of JHMV (8) was used for infections. The virus was propagated in the presence of MAb J.2.2 (8) and was plaque assayed on monolayers of DBT cells as previously described (8). Mice were injected in the left hemisphere with a 30-μl volume containing 500 PFU of JHMV diluted in endotoxin-free Dulbecco's modified phosphate-buffered saline (PBS). For determination of CNS virus titers, one-half of the brains were homogenized on ice in Dulbecco's PBS by use of Tenbroeck tissue homogenizers. Following clarification by centrifugation at 500 × g for 7 min, homogenates were stored at −70°C and virus titers were determined by plaque assays on monolayers of DBT cells, as previously described (3-5). Clinical disease was graded as previously described (8), with grades as follows: 0, healthy; 1, hunched back; 2, partial hind limb paralysis or inability to maintain upright position; 3, complete hind limb paralysis; 4, moribund or dead.

T-cell purification and adoptive transfer.

BALB/c wt, GKO, and PKO donors were immunized by intraperitoneal injection with 1 × 10⁶ to 2 × 10⁶ PFU of JHMV. Donor cells were prepared from spleens at 4 to 16 weeks postimmunization. Spleen cells were depleted of B cells by adsorption onto 150-mm-diameter plates coated with goat anti-mouse immunoglobulin (ICN Pharmaceuticals, Costa Mesa, Calif.) or by adsorption to Immunolon columns (Biotecx Laboratories, Houston, Tex.). CD8⁺ T cells were purified by either positive or negative selection. Briefly, for negative selection, CD4⁺ T cells and residual B cells were depleted with magnetic beads coated with anti-CD4 and anti-CD19 (Miltenyi Biotec Inc., Auburn, Calif.) antibodies according to the manufacturer's instructions. Alternatively, following the removal of B cells, CD8⁺ T cells were purified by positive selection with anti-Lyt2-coated magnetic beads (Miltenyi Biotech). Purity was assessed by flow cytometry using phycoerythrin (PE)-labeled anti-CD8⁺ (clone 53-6.7), fluorescein isothiocyanate (FITC)-labeled anti-CD4 (clone GK1.5), and FITC-labeled anti-CD19 (clone 1D3) (BD PharMingen, San Diego, Calif.) antibodies. CD8⁺ T cells were enriched to >96% by both approaches, and no differences in antiviral effector function were detected. CD8⁺ T cells specific for the immunodominant N epitope ranged from 5 to 15% of the total CD8⁺ T cells, as determined by flow cytometry using FITC-labeled anti-CD8 antibody and a PE-labeled L^d MHC class I tetramer associated with the pN318-326 peptide (L^dN318; 0.1 to 0.2 μg/0.5 × 10⁶ to 1.0 × 10⁶ cells) (3). Recipients received 5 × 10⁶ purified donor CD8⁺ T cells per mouse by intravenous injection 24 h prior to viral challenge.

CMCs.

CNS-derived mononuclear cells (CMCs) were isolated from mice perfused with PBS by homogenization on ice with Tenbroeck tissue homogenizers as described above. Briefly, following centrifugation to obtain clarified homogenates for virus titer determination, cell pellets were suspended in RPMI medium containing 25 mM HEPES, pH 7.2, and adjusted to 30% Percoll (Pharmacia, Uppsala, Sweden). A 1.0-ml underlay of 70% Percoll was added prior to centrifugation at 800 × g for 20 min at 4°C. Cells were recovered from the 30%-70% interface and washed in RPMI medium prior to analysis.

Flow cytometry.

CMC suspensions were blocked with anti-mouse CD16/CD32 (clone 2.4G2; BD PharMingen) antibody on ice for 15 min prior to staining. For three-color flow cytometric analysis, cells were stained with FITC-, PE-, and CyC-conjugated MAbs at 4°C for 30 min in PBS containing 0.1% bovine serum albumin. Where indicated, four-color analysis was conducted using FITC-, PE-, allophycocyanin (PerCP)-, and peridinin chlorophyll protein (APC)-conjugated MAbs. The expression of surface molecules was characterized by using the following MAbs (all obtained from BD PharMingen except where indicated): anti-CD8 (clone 53-6.7), anti-CD4 (clone GK1.5), anti-CD19 (clone 1D3), and anti-pan-NK (clone DX5). Neutrophils were identified with anti-Ly-6G/6C MAb (clone RB6-8C5). Virus-specific CD8⁺ T cells were detected by staining with FITC-labeled anti-CD8 antibody and PE-labeled L^dN318 tetramers. PE- or CyC-conjugated anti-CD45 (Ly-5) and anti-F4/80 (Serotec, Raleigh, N.C.) antibodies, or in some experiments, anti-CD11b (M1/70) antibody, distinguished microglia (CD45^lo F4/80⁺; CD45^lo CD11b⁺) from infiltrating or perivascular macrophages (CD45^hi F4/80⁺; CD45^hi CD11b⁺). MHC expression on microglia was determined by using an anti-H-2 D^d-specific MAb (clone 34-2-12), anti-I-A/I-E MAb (clone 2G9), or anti-I-A^d MAb (clone 39-10-8). Due to the consistent reactivity of the anti-F4/80 MAb with a subset of Ly-6G/6C⁺ cells, neutrophils were characterized as Ly-6G/6C⁺ class II⁻, whereas infiltrating macrophages were defined as F4/80⁺ class II⁺. Samples were analyzed on a FACStar or FACSCalibur flow cytometer (Becton Dickinson, Mountain View, Calif.). Forward and side scatter signals obtained in linear mode were used to establish a region (R1) containing live lymphocytes, macrophages, and neutrophils, while excluding dead cells and tissue debris. A minimum of 2.5 × 10⁵ viable cells were stained and 5 × 10⁴ to 1 × 10⁵ events were analyzed per sample.

Histopathological analysis.

Brains and spinal cords were removed and either fixed with Clark's solution (75% ethanol and 25% glacial acetic acid) and embedded in paraffin or snap frozen. Paraffin sections were stained with either hematoxylin and eosin or luxol fast blue. For identification of virus antigen-positive cells, mice were perfused with 4% paraformaldehyde in PBS (pH 7.4). Brains and spinal cords were removed, postfixed with 4% paraformaldehyde at 4°C for 24 h, and then equilibrated with 30% sucrose in PBS (pH 7.4) by overnight incubation at 4°C. Frozen sections were stained with MAb J.3.3, specific for the viral N protein, and a MAb specific for astrocytes (glial fibrillary acidic protein; Dako, Carpinteria, Calif.) or macrophages (F4/80; SeroTec), as previously described (3-5, 16, 17). JHMV antigen and glial fibrillary acidic protein double staining was visualized with peroxidase Vector VIP and Vector SG substrate kits (both from Vector Laboratories, Burlingame, Calif.), respectively. Viral antigen in F4/80-positive cells was detected with an alkaline phosphatase Vectastain ABC kit for MAb J.3.3 and a peroxidase Vectastain ABC kit with AEC substrate (both from Vector Laboratories) for MAb F4/80.

CD8⁺- and CD4⁺-T-cell infiltration was examined by immunoperoxidase staining of acetone-fixed frozen sections with rat anti-CD8a (Ly-2) and anti-CD4 (L3T4) MAbs (BD PharMingen) as previously described (14, 15). Primary MAbs were detected by immunoperoxidase staining with a biotinylated rabbit anti-rat antibody (Vector Laboratory) and a Vectastain ABC kit. Aminoethyl carbazol was used as the chromogen substrate (Vector Laboratories). Apoptotic cells were identified in acetone-fixed frozen sections by a terminal deoxynucleotidyltransferase-mediated dUTP-biotin nick end labeling assay using an Apop Tag kit (Oncor, Gaithersburg, Md.) with terminal deoxynucleotidyltransferase, as indicated by the manufacturer. Sections were scored in a blinded fashion for inflammation, viral antigen, and demyelination. Representative fields were identified based on average scores for all sections in each experimental group.

JHMV pathogenesis in immunodeficient SCID mice.

CD8⁺ T cells play a major role in the control of JHMV replication within the CNS via IFN-γ secretion and perforin-mediated cytolysis (5, 16, 21). Nevertheless, CD4⁺ T cells contribute to JHMV-specific CD8⁺ T cell expansion and survival within the CNS (34), in addition to being potent sources of IFN-γ secretion themselves (30). For evaluation of antiviral effector functions as well as the pathogenic potential of memory CD8⁺ T cells in the absence of CD4⁺ T cells, JHMV pathogenesis was analyzed in immunodeficient SCID mice. For initial confirmation of the importance of adaptive immunity in controlling JHMV infections (3, 4, 12, 15, 40), SCID mice were compared to wt mice. In infected wt mice, JHMV replication in the CNS peaked at 5 days postinfection (p.i.), declined beginning at day 7 p.i., and was below the level of detection by day 14 p.i. (Fig. 1). No clinical disease was detected in infected wt or SCID mice prior to day 10 p.i. Beginning at day 10 p.i., wt mice exhibited clinical evidence of encephalomyelitis which increased dramatically with time, consistent with previous data (8, 17, 21, 25). SCID mice exhibited barely detectable clinical disease throughout the infection, with a minimal average clinical score of ≤1.0. The majority of infected SCID mice succumbed between days 15 and 18 p.i., with vary rare individuals surviving until 22 days p.i. The unchecked JHMV replication from days 5 to 14 p.i. in infected SCID mice contrasts with the immune-mediated control of infectious virus from days 7 to 14 p.i. in wt mice (Fig. 1) and the partial control of virus in the CNS of mice deficient in the expression of IFN-γ, perforin, or both effectors (5, 16, 21). The absence of clinical disease in infected SCID mice is consistent with the suggestion that inflammatory responses contribute to clinically evident neurological symptoms (12, 13, 24).

A comparison of the pathological consequences of infection in the absence of adaptive immunity showed an absence of inflammation in the CNS of infected SCID mice, in contrast to a vigorous inflammatory response in wt mice (Fig. 2). The viral antigen level was dramatically higher in the CNS of SCID mice after day 3 p.i. and was abundant in all CNS cell types, including astrocytes, oligodendroglia, microglia/macrophages (Fig. 2A), and even occasional neurons (data not shown), at day 14 p.i. In contrast, only scattered cells positive for viral antigen, for which the level had dramatically decreased from the levels found during acute infection, were detected in the brains of infected wt mice at day 14 p.i. (Fig. 2B). Similarly, the viral antigen level was also dramatically increased in spinal cords of infected SCID mice relative to those of wt mice, and the antigen was present in all cell types in both gray and white matter, although more virus-infected cells were detected in the white matter than in the gray matter (Fig. 2C). In contrast to wt mice, occasional neurons were infected in the spinal cords of SCID mice, although the numbers did not increase in the rare survivors up to day 21 p.i. (data not shown). In spinal cords of wt mice, the residual viral antigen level decreased relative to the levels found during acute infections, and antigen was localized predominantly in white matter cells exhibiting morphological characteristics consistent with oligodendroglia (Fig. 2D). Coincident with viral clearance, demyelination in wt mice was apparent as early as day 10 p.i. (4, 8, 15, 25; data not shown). However, despite numerous infected oligodendroglia in the spinal cords of infected SCID mice, demyelination was absent from both the brains and spinal cords at day 14 p.i. (Fig. 2E and F). However, in contrast to the case for immunodeficient Rag1^−/− mice, which also show no myelin loss following JHMV infection (23, 24, 42), occasional small foci of demyelination were found in the spinal cords of the rare infected SCID mice which survived to later time points (data not shown). These small foci of myelin loss are consistent with the suggestion that oligodendroglial infection may contribute to focal demyelination (2, 12-15). Flow cytometric analysis of the cells derived from the CNS of infected SCID mice at day 10 p.i. showed ∼30% CD45^hi infiltrating cells, of which the vast majority (60 to 70%) were neutrophils, with 10% and 5 to 8% constituting NK cells and macrophages, respectively. By this time, infected wt mice had only ∼7% neutrophils and 2% NK cells in the CD45^hi population. Macrophages constituted ∼25%, while T cells comprised a predominant 65%, as previously described (5, 23; data not shown).

Antiviral CD8⁺-T-cell effector function within the CNS.

JHMV replication is inhibited via both IFN-γ secretion and perforin-mediated cytotoxicity. Although CD8⁺ T cells exert these antiviral activities, their functioning is partially dependent upon CD4⁺ T cells (34). For evaluation of the sole antiviral functions of CD8⁺ T cells and dissection of the relative contributions of IFN-γ and perforin-mediated cytolysis in the absence of CD4⁺-T-cell-mediated help, CD8⁺ T cells derived from immunized wt, PKO, and GKO donors were adoptively transferred into SCID recipients. Although CD8⁺ T cells activated in vitro or obtained from actively infected donors traffic into the JHMV-infected CNS (32, 34, 36), memory populations were chosen to diminish the loss of activated T cells in the liver and potential activation-induced cell death within the infected CNS (28). Antiviral functioning was assessed by comparing viral replication within the CNS of infected SCID mice and those of infected SCID CD8⁺-T-cell recipients. CD8⁺ T cells derived from wt donors reduced virus replication in the CNS of SCID recipients at day 10 p.i. compared to untreated controls and continued to be effective at controlling infectious virus to near the limits of detection to day 14 p.i. (Fig. 3). These data demonstrate that CD8⁺ T cells are recruited into the JHMV-infected CNS in the absence of CD4⁺ T cells and that they express effective antiviral functions. No evidence for antiviral activity was detected in the CNS at day 10 p.i. following the adoptive transfer of purified CD8⁺ T cells from naïve donors into SCID recipients, and only a small effect (<1 log₁₀ reduction) was detected at day 14 p.i. (data not shown). The results confirmed that antiviral effects reside in the virus-specific memory CD8⁺ T cells rather than in primed naïve CD8⁺ T cells contained within the transferred population.

The relative contribution of a perforin-dependent mechanism for controlling JHMV replication in the absence of IFN-γ was examined by the transfer of CD8⁺ T cells derived from immunized GKO donors. GKO CD8⁺ T cells were less efficient at controlling viral replication than were wt CD8⁺ T cells at day 10 p.i. (Fig. 3), and their impaired antiviral function was even more apparent by day 14 p.i. (Fig. 3). The inability of GKO CD8⁺ T cells to sustain control of viral replication was confirmed after the transfer of in vitro-activated CD8⁺ T cells (data not shown). The reduced ability of GKO CD8⁺ T cells to control virus replication in SCID recipients is consistent with results for transfer into PKO/GKO recipients (5) and the inability of infected GKO mice to clear infectious virus from the CNS (21).

CD8⁺ T cells from PKO donors that are competent in IFN-γ secretion but unable to mediate cytolysis were approximately as efficient as CD8⁺ T cells derived from both wt and GKO donors at controlling viral replication within the CNS at day 10 p.i. (Fig. 3). However, in contrast to GKO CD8⁺ T cells, PKO CD8⁺ T cells suppressed viral replication to similar levels as the wt donor cells did at 14 days p.i. (Fig. 3). The ability of PKO CD8⁺ T cells to control JHMV replication within the CNS is consistent with the delayed, but complete, clearance of JHMV from the CNS of infected PKO mice (16) but contrasts with the less effective virus control following transfer into PKO/GKO hosts containing CD4⁺ T cells (5). This apparent dichotomy may be due to homeostatic parameters resulting in enhanced expansion and CNS recruitment of donor T cells in SCID mice compared to spatial constraints in PKO/GKO hosts. Cells derived from the CNS of infected SCID recipients of both wt and GKO donors exhibited ex vivo JHMV-specific cytotoxic activity at day 7 p.i. (data not shown). In addition, cells derived from the CNS of infected SCID recipients of both wt and PKO CD8⁺ T cells secreted IFN-γ following in vitro stimulation (data not shown). The ability to express cytolytic activity, secrete IFN-γ, and importantly, control CNS virus replication demonstrates that memory CD8⁺ T cells exert functional activity within the CNS of infected SCID recipients even in the absence of CD4⁺-T-cell support.

Altered CNS inflammation in the presence of donor CD8⁺ T cells.

Inflammatory responses were compared in the CNS of infected SCID mice and infected SCID CD8⁺-T-cell recipients by immunohistochemistry at day 14 p.i. to assess the degree of inflammation as well as the distribution of CD8⁺ T cells. No inflammatory cells or CD8⁺ T cells were detected in the CNS of infected SCID mice (Fig. 4A and B). In contrast, wt CD8⁺ T cells induced vigorous inflammatory responses in both brains and spinal cords (Fig. 4C and D). CD8⁺ T cells from wt donors were present in the brain parenchyma and the white matter of the spinal cord (Fig. 4D), consistent with the reduced infectious virus level (Fig. 3). In contrast to unmanipulated infected SCID mice, no infected neurons were noted in recipients of wt CD8⁺ T cells. CD8⁺ T cells deficient in either IFN-γ or perforin also induced vigorous CNS inflammation (Fig. 4E to H) and were distributed within the brain parenchyma and the spinal cord white matter, similar to those derived from wt donors. Less efficient control of virus replication (Fig. 3) could thus not be attributed to impaired tissue access. An increased frequency of apoptotic cells with the morphology of lymphocytes was detected within the parenchyma of both wt and PKO CD8⁺-T-cell recipients compared to untreated infected SCID mice. However, reduced numbers of apoptotic cells were detected in the CNS of GKO T-cell recipients (data not shown), consistent with the notion that IFN-γ diminishes CD8⁺-T-cell survival (1). Overall, these data support the concept that CD8⁺ T cells entering the infected CNS in the absence of CD4⁺ T cells exhibit an increased frequency of apoptosis (32). CD8⁺ T cells derived from wt donors induced demyelination in the CNS of infected SCID recipients (Fig. 4C), similar to infected wt mice (Fig. 2) and PKO/GKO recipients of wt CD8⁺ T cells (5). However, inflammation by GKO or PKO CD8⁺ T cells did not induce myelin loss in SCID recipients (Fig. 4E and G). The predominant localization of all three populations of CD8⁺ T cells within the spinal cord white matter (Fig. 4F and H) supports the notion that demyelination is not associated with the presence of CD8⁺ T cells capable of either IFN-γ or perforin secretion alone.

Differential susceptibility of glial cells to CD8⁺-T-cell-mediated effector mechanisms.

An analysis of JHMV infection in PKO/GKO recipients of virus-immune CD8⁺ T cells (5) supported the concept that perforin-mediated cytotoxicity controls virus replication in astrocytes and macrophages/microglia while IFN-γ controls virus replication in oligodendroglia (16, 21). The CNS of infected SCID recipients were examined to determine if donor CD8⁺ T cells alone, in the absence of other potential adaptive antiviral immune effectors, specifically CD4⁺ T cells, regulated virus replication in specific CNS cell types via distinct effector mechanisms. In infected SCID mice, the viral antigen was prominent in both astrocytes and oligodendroglia, with approximately equal numbers of infected cells (Fig. 2 and 5). Concomitant with a reduction in recoverable infectious virus in recipients of wt CD8⁺ T cells (Fig. 3 and 4), an ∼70% reduction in virus-infected astrocytes and oligodendroglia was observed compared to untreated SCID mice (Fig. 5B). Similar to wt donor CD8⁺ T cells, GKO CD8⁺ T cells reduced the number of infected astrocytes in spinal cords (Fig. 5D). However, no reduction was found in infected oligodendroglia compared to the case for untreated SCID mice. In contrast, in SCID recipients of PKO CD8⁺ T cells the number of infected astrocytes was comparable to that in untreated SCID mice; however, the frequency of infected oligodendroglia was dramatically reduced (Fig. 5C). Despite suppressed virus replication (Fig. 3) and reduced infected oligodendroglia, rare infected neurons were detected in PKO CD8⁺-T-cell recipients, in contrast to SCID recipients of either wt or GKO CD8⁺ T cells. Although few infected F4/80⁺ cells were detected in the CNS of any recipient group compared to wt mice, the overall distribution of F4/80⁺ cells in the CNS was similar. The low frequency of this population, as detected by immunohistochemistry (data not shown), precluded determination of the effects of IFN-γ and perforin-mediated cytolysis on virus replication in this cell type. These data are nevertheless consistent with the inability of perforin-mediated cytolysis to control JHMV replication in oligodendroglia and the inability of infected PKO mice to inhibit virus replication in astrocytes, as previously observed with CD4-competent mice (5, 16).

For a more quantitative determination of subsets of inflammatory cells, CNS-derived cells from infected SCID mice and all three recipient groups were analyzed by flow cytometry (Fig. 6). Compared to those for infected control SCID mice, CNS yields were consistently higher for recipient groups, but varied in absolute numbers between experiments. In the CNS of infected untreated SCID mice, CD45^hi bone-marrow-derived inflammatory cells comprised ∼30% of the cells at day 10 p.i. (Fig. 6A). The majority of CD45^hi cells were Ly-6G/6C⁺ and MHC class II⁻, consistent with the case for neutrophils (Fig. 6B). NK cells and MHC class II⁺ F4/80⁺ macrophages each comprised <10% of the infiltrating cells (Fig. 6B). No CD4⁺ or CD8⁺ T cells were detected in the CNS of infected control SCID mice. CD45^hi cells derived from the CNS cells of all CD8⁺-T-cell recipient groups ranged from 25 to 40% (Fig. 6A). However, these populations contained significantly reduced percentages of neutrophils and NK cells. These data suggest that CD8⁺ T cells, irrespective of the ability to secrete IFN-γ or perforin, reduce the accumulation and/or retention of neutrophils and NK cells. In contrast, relative to the levels detected in untreated SCID mice, the frequency of infiltrating CD45^hi MHC class II⁺ F4/80⁺ macrophages was elevated in recipients of IFN-γ-secreting CD8⁺ T cells (Fig. 6B). No increase in infiltrating macrophages was found in recipients of GKO CD8⁺ T cells, supporting the concept that CNS inflammation by CD8⁺ T cells able to secrete IFN-γ enhances the recruitment and/or retention of bone-marrow-derived macrophages (22, 39). Note the equivalent frequencies of macrophages recruited into the brains of recipients of both wt- and PKO-derived CD8⁺ memory T cells. Their frequency was even higher in spinal cords of wt and PKO CD8⁺-T-cell recipients, constituting up to 50% of infiltrating cells (data not shown). Nevertheless, demyelination was only present in the CNS of wt recipients, while no demyelination was detected in the CNS of recipients of PKO CD8⁺ T cells (Fig. 4 and 5).

Consistent with the control of virus replication and reduced viral antigen, 35 to 55% of the inflammatory cells recovered from the CNS of all recipients comprised CD8⁺ T cells (Fig. 6C). Virus-specific cells comprised >50% of the total CD8⁺ T cells within the CNS of both wt and GKO recipients. The frequencies of both total and virus-specific CD8⁺ T cells in CNS-derived CD45^hi cells from PKO recipients were consistently increased compared to those in recipients of wt- and GKO-derived CD8⁺ T cells, possibly reflecting the inhibitory role of perforin on CD8⁺-T-cell expansion (1, 31). The increased frequency of virus-specific CD8⁺ T cells derived from PKO donors may also reflect ongoing stimulation by antigen-positive astrocytes, suggesting that continued expression of viral antigens may contribute to the magnitude and duration of CD8⁺-T-cell responses.

One possibility for ineffective CD8⁺-T-cell-mediated clearance is insufficient MHC upregulation on resident CNS cells, although NK cells, a potential source of IFN-γ, are present in the CNS of infected SCID mice. MHC expression was therefore compared on resident CD45^lo microglia from control and recipient mice at day 10 p.i. Microglia from naïve mice express few, if any, MHC class I molecules (5, 29). However, the majority of microglia in infected SCID mice expressed class I molecules, and transferred CD8⁺ T cells did not dramatically alter their expression (Fig. 7). IFN-γ-competent wt and PKO CD8⁺ T cells had slightly enhanced expression, while GKO donor cells had a reduced frequency as well as intensity of class I expression on microglia. Thus, the specific inability of cells from GKO donors to control JHMV infections could not be attributed to an absence of MHC class I expression on resident CNS cells. In contrast to MHC class I, MHC class II expression was dramatically affected by IFN-γ secretion by donor CD8⁺ T cells. Class II molecule expression was undetectable on microglia in both infected control and GKO CD8⁺-T-cell recipients. In contrast, wt and PKO CD8⁺ T cells induced class II expression on 70 and 96% of microglia, respectively, thus providing direct in vivo evidence for IFN-γ secretion. In contrast to microglia, class II expression by infiltrating macrophages was higher in all groups of mice, with the highest expression detected in recipients of IFN-γ-competent donor T cells. The effects of IFN-γ were thus manifested by the enhanced overall recruitment of CD45^hi class II⁺ F4/80⁺ macrophages and class II induction on microglia, in addition to the reduction in virus load observed in PKO recipients.