Critical Roles of NK and CD8⁺ T Cells in Central Nervous System Listeriosis¹

Listeria monocytogenes (LM)⁴ is a Gram-positive facultative intracellular bacterium, which causes gastroenteritis, mother-to-fetus infection, septicemia, and CNS infection in immunocompromised individuals, pregnant women, and newborns. Once symptoms develop, the mortality rate is 25–30% with CNS infection (CNS listeriosis) having the poorest prognosis of human listeriosis (1, 2, 3). Among various bacteria that infect the CNS, LM has the highest mortality rate because these infections exhibit not only meningitis but also encephalitis or brain abscesses (4).

Murine listeriosis has been widely used as a model system to investigate the mammalian immune response against intracellular pathogens, because LM is one of the best characterized and most easily manipulated pathogens (2, 5). During the early phase of infection, IFN-γ from NK and APCs plays a critical role in suppressing bacterial burdens by activating the bactericidal activity of phagocytes (6, 7, 8). This innate immune reaction is followed by acquired immune activity involving T lymphocytes.

Previous studies have illustrated a crucial role for αβ T cells in the subacute phase of systemic LM infection (9, 10). CD4⁺ T cells induce and maintain a Th1-type immune response by producing Th1 cytokines, most notably IFN-γ, which further activate macrophages to eradicate intracellular LM (11). LM-specific CD8⁺ T cells also appear to be essential for the complete clearance of LM and for the acquisition of subsequent memory responses (12). However, little is known about the role of each subset of T cells in CNS listeriosis.

A significant obstacle for the study of CNS listeriosis has been the lack of appropriate animal models. A useful CNS listeriosis model should induce CNS infection as a sequel to systemic listeriosis, thus mimicking the natural course of human listeriosis. However, it has been difficult to establish brain lesion models induced by i.v. or i.p. LM administration. One reason is that to establish CNS listeriosis following systemic infection, the bacterial load for systemic inoculation needs to be high, which is life-threatening for the hosts. The second reason is that the location in the brain and the timing of bacterial translocation to the brain after systemic infection vary widely among individual animals.

Another option for making a model of CNS listeriosis is to inoculate the bacteria directly into the brain. This method would establish CNS infection synchronously and achieve similar bacterial burdens among individual animals. CNS listeriosis triggered by such direct inoculation of LM into the CNS has been used to study host immunological reactions (13). In this model, however, animals succumb to severe ventriculitis within 1 wk unless immunized by a systemic inoculation of LM before CNS injection. Thus, animals die during the early phase of infection, preventing the study of acquired immunity against LM in the CNS. To overcome such problems, attenuated LM lacking ActA, InlA/B, or PlcB2 have been used for the CNS listeriosis model, allowing animals to live long enough for various analyses (13, 14, 15). However, it is more desirable to use wild-type LM to study the hosts’ natural immune reactions, since bacterial pathogenic components are crucial for the induction of host immune responses.

In this study, we have established a new model for CNS listeriosis in which LM infection is restricted to the brain parenchyma and meninges while avoiding the induction of ventriculitis. In this model, naive animals can survive the disease and eradicate LM without prior immunization. This new model enables us to analyze the acquired immune response against LM in the brain and show that the immune system in the brain is highly effective against intraparenchymal (i.pa.) LM infection. These results are contrary to the commonly accepted theory that the CNS is very vulnerable to LM infection, most likely because that conclusion was reached using the previous model, which results in fatal ventriculitis. We demonstrate here that IFN-γ is critically important in CNS listeriosis during the early phase of infection. In the later phase, CD8⁺ T cells are more important than CD4⁺ T cells in eradicating the bacteria from the CNS.

Six- to 8-wk-old female wild-type (WT) BALB/c and C57BL/6 mice and gld/gld mice on a C57BL/6 background were purchased from SLC. RAG2^−/− mice on BALB/c and C57BL/6 backgrounds were obtained from Taconic Farms. IFN-γ^−/− mice on a BALB/c background were purchased from The Jackson Laboratory. All mice were maintained in our specific pathogen-free animal facility, and experiments were performed between 6 and 12 wk of age in accordance with the guidelines of the Institutional Animal Care and Use Committee.

LM EGD strain was provided by Dr. M. Mitsuyama (Kyoto University, Kyoto, Japan) and grown in brain-heart infusion broth (BD Biosciences). After washing with sterile pathogen-free PBS (pH 7.4), aliquots of log-phase cultures were stored at −80°C. For each experiment, LM was thawed from the stock solution and diluted appropriately in sterile pathogen-free PBS for either i.p., i.pa., or intraventricular (i.vt.) inoculation.

Three hundred to 3 × 10³ CFU of LM suspended in 1 μl of PBS were injected into the left forebrain. Anesthesia was provided by an i.p. injection of 40 mg/kg sodium pentobarbital (Dainippon Sumitomo Pharma) and 5 mg/kg xylazine (Sigma-Aldrich). A 1-cm longitudinal incision was made along the midline between ears and eyes to expose the bregma. For i.pa. and i.vt. LM inoculation, the injection point was 2 mm rostal and 1 mm lateral to the bregma and 0.5 mm caudal and 1 mm lateral to the bregma, respectively. A 27-gauge needle was used to deliver LM suspended in PBS into the brain. A total of 1 μl of suspension containing 3 × 10²–3 × 10³ CFU of LM was placed 1.5-mm deep from the external surface of the calvaria to prevent reflux during injection. Using this approach, bacteria were reproducibly deposited into the frontal lobe white matter or into the lateral ventricle. Control animals were implanted with sterile PBS. Incisions were closed using sterile surgical glue. The mortality rate associated with lesion generation was minimal, with >98% of animals surviving the procedure.

For the determination of the intracranial bacterial load, brains were isolated from euthanized animals at the indicated days postinfection (p.i.) and whole brain tissue was homogenized using tissue grinders. Ten-fold serial dilutions of the homogenate were seeded on Luria-Bertani broth (Sigma-Aldrich) agar plates. After a 24-h incubation at 37°C, bacterial colonies were counted.

Mice were immunized by i.p. injection of 0.2 ml of PBS containing 3 × 10⁴ LM or 0.2 ml of PBS alone. After 14 days, mice were challenged using i.pa. infection. In some experiments, mice were boosted i.p. with 1 × 10⁶ LM to obtain LM-specific CD8⁺ T cells for CTL assay.

T cells from the spleens of sex-matched mice were prepared by a negative or positive sorting method with magnetic beads. For whole T cell transfer, a negative sorting method was used. Splenocytes were stained with biotinylated anti-B220, anti-Gr1, anti-Mac1, anti-CD11c, anti-TER119 (erythrocytes) and anti-CD49b (NK cells) mAbs. The cells were then incubated with magnetic beads coupled with streptavidin and negatively sorted on an autoMACS (Miltenyi Biotec) to >95% purity. Four million cells were transferred into recipient mice via the orbital vein. For CD8⁺ T cell transfer, positive sorting was executed. Splenocytes were incubated with magnetic beads coupled with anti-CD8 mAb and positively sorted on an autoMACS to >95% purity. Two million cells were transferred into recipient mice. Recipient mice were then housed for at least 2 wk before infection.

On the indicated days after i.pa. infection, mice were deeply anesthetized using diethyl ether (Sigma-Aldrich) and perfused intracardially with ice-cold PBS to remove contaminating intravascular leukocytes from the brain. For flow cytometric analyses, brain tissue was dissected, cut into 1-mm cubes, and enzymatically digested for 20 min at 37°C with 1.0 ml of 0.75% type IV collagenase (Sigma-Aldrich) and DNase I (Sigma-Aldrich). After the digestion, the tissue was minced and passed through a 100-μm cell strainer and leukocytes were separated by Percoll (Amersham Biosciences) gradient centrifugation. Cells pooled at the interface between 1.072 and 1.088 g/ml were collected and washed with PBS.

Isolated cerebral or splenic leukocytes were analyzed by double or triple immunofluorescence staining followed by flow cytometry. All of the following Abs were purchased from BD Biosciences: anti-CD3ε-FITC (clone 145-2C11), anti-CD11b-FITC (clone M1/70), anti-CD11c-FITC (clone HL3), anti-CD4-PE (clone GK1.5), anti-CD8α-PE (clone 53-6.7), anti-CD19-PE (clone 1D3), anti-Ly-6G and Ly-6C (Gr-1)-PE (clone RB6-8C5), anti-FasL-PE (clone MFL3), anti-CD8α-biotin (clone 53-6.7), anti-CD45-biotin (clone 30-F11), anti-Ly-6G and Ly-6C (Gr-1)-biotin (clone RB6-8C5), anti-TCRδ-biotin (clone GL3), and anti-CD49b-biotin (clone DX5). To block nonspecific binding of Abs to Fc receptors, isolated cells were first incubated with anti-CD16/32 Abs (clone 2.4G2) at 4°C for 10 min. Subsequently, the cells were stained with a mixture of fluorochrome-labeled Abs and biotin-conjugated Abs at 4°C for 20 min. Then the cells were washed and stained by streptavidin-allophycocyanin (BD Biosciences) at 4°C for 15 min. Flow cytometry was performed on a FACSCalibur (BD Biosciences) and the data were analyzed with FlowJo software (Tree Star). Murine microglia were defined as CD45^intCD11b⁺, macrophages as CD45^highCD11b⁺Gr-1⁻, and neutrophils as CD45^highCD11b⁺Gr-1⁺. H-2K^d listeriolysin O (LLO) tetramer was purchased from Medical and Biological Laboratories. Staining for the tetramer was conducted according to the manufacturer’s recommended protocols.

For H&E staining and immunohistochemistry, mice were perfused intracardially with ice-cold PBS under deep diethyl ether anesthesia. Brains of three animals per group were dissected and blocks were mounted in plastic cases with OCT compound (Sakura Finetek), snap frozen on a metal bar precooled in liquid nitrogen, and stored at −80°C. Sections (6 μm) were stained with H&E (Sakura Finetek). Immunohistochemistry was performed on frozen sections (6 μm). Sections were fixed in acetone (Sigma-Aldrich) for 5 min. Nonspecific binding was blocked for 30 min with Blockase (a casein digestion product; Dainippon Pharmaceutical). Endogenous peroxidase activity was blocked with 3% H₂O₂ (Santoku Chemical Industries) in PBS for 5 min. Sections were stained by an indirect immunoperoxidase protocol using biotin-conjugated rabbit anti-LM polyclonal Ab (Virostat) as a primary Ab. Immunoperoxidase staining was conducted using the Vectastain Elite ABC kit (Vector Laboratories) and diaminobenzidine substrate (Vector Laboratories). Sections were lightly counterstained with hematoxylin.

Mice were depleted of CD4⁺ T cells, CD8⁺ T cells, or NK cells by i.v. injection with 0.3 mg of anti-CD8 mAb (clone 2.43, rat IgG2b), 0.5 mg of anti-CD4 mAb (clone GK1.5, rat IgG2a), or 0.5 mg of rabbit anti-asialo GM1 polyclonal Ab (Wako Pure Chemical Industries), respectively, for 3 successive days. Mice were further treated every 3 days with an additional 0.3 mg of 2.43, 0.5 mg of GK1.5, or 0.5 mg of anti-asialo GM1 for CD4⁺ T cells, CD8⁺ T cells, or NK cell depletion, respectively, resulting in >98% depletion of CD8⁺, CD4⁺, or DX5⁺ cells as confirmed by flow cytometric analysis (data not shown). An irrelevant mAb, anti-CD45.1 (clone A20-1.7, rat IgG2b) was used as a control Ab.

Whole blood was obtained from either lateral tail veins or orbital veins of the indicated animals. After incubation at 4°C overnight, blood samples were centrifuged, the sera were collected, and samples were analyzed for IFN-γ using an ELISA kit purchased from R&D Systems. All assays were conducted according to the manufacturers’ recommended protocols.

Primary cultures of murine microglia were prepared as described previously (16). In brief, a mixed glial culture was prepared from the brains of neonatal BALB/c mice and maintained for 10–20 days in DMEM (Sigma-Aldrich) containing 10% FCS, 10 U/ml penicillin (Invitrogen) and 10 mg/ml streptomycin (Invitrogen). Microglia were obtained as floating cells over the mixed glial culture. The floating cells were collected by a gentle shake and 5 × 10⁴ cells/well were transferred to 96-well dishes, and the attached cells were used for CTL assays.

Statistical significance was evaluated by the Mann-Whitney U test for the comparison of bacterial burdens, serum IFN-γ levels, and survival rates between two groups. A value of 28 days was assigned to survivors living >28 days after LM infection. The Wilcoxon t test was applied for the comparison of mean fluorescence intensity between the sample groups stained with either FasL or isotype-matched control IgG.

Significance was defined as p < 0.05. Data are expressed as the mean ± SD, unless otherwise stated.

To examine whether the location of LM infection in the brain is an important factor for the progress of CNS listeriosis, WT BALB/c mice were infected with LM (3 × 10² CFU/head) either in the forebrain parenchyma or in the lateral ventricle (see Fig. 1). Whereas all mice infected in the ventricle succumbed by day 4 p.i., 100% of mice infected in the forebrain parenchyma survived the infection (Fig. 1,A). In addition, bacterial burdens in the brain were significantly lower in mice infected in the forebrain parenchyma compared with mice infected in the ventricle (Fig. 1 B). Mice infected in the forebrain parenchyma cleared LM from the brain by day 14 p.i.

Immunohistochemical analyses of i.pa. and i.vt. infections (Fig. 2) showed that i.vt. infection induced severe ventriculitis with massive leukocyte infiltration and widespread LM distribution. In contrast, i.pa. infection led to focal encephalitis at the site of bacterial injection with mild meningitis but no ventriculitis. On day 3 p.i., the i.pa. infection site was apparent in the forebrain of the injected side of the hemisphere (Fig. 2,A, arrow), and no leukocyte infiltration into the ventricle was observed (Fig. 2,E). In contrast, a large number of leukocytes infiltrated into the ventricle after i.vt. infection (Fig. 2, B and F). In addition, severe meningitis was induced in i.vt.- infected mice but not i.pa.-infected mice (Fig. 2, C and D). Many LM were found in the ventricular space in i.vt.-infected brain, whereas few LM were detected in the ventricle in i.pa.-infected mice (Fig. 2, G and H). It should be noted that LM were almost completely confined to the brain parenchyma in the i.pa. infection model (data not shown).

To quantitate host immune reactions in this new CNS listeriosis model, the kinetics of leukocyte infiltration in the i.pa.-infected brain of BALB/c mice were determined by flow cytometric analyses and compared with those in the spleen (Fig. 3). Leukocyte infiltration in the brain was transient, peaking at day 7 p.i. Significantly different populations were observed in the brain compared with the spleen. For example, although B cells were the major population in the spleen, there were only small numbers of B cells in the brain. On the other hand, many γδ T cells infiltrated into the brain, whereas γδ T cells were a minor population in the spleen. As expected, on day 3 p.i., innate immune cells such as macrophages and NK cells occupied a relatively large proportion compared with the later phase of infection. On day 7 p.i., both innate and acquired immune cells were present in the brain. On day 10 p.i., the predominant cell types in the brain were T cells, including γδ T cells. Infiltrating cell numbers decreased by day 14 p.i., paralleling the clearance of bacteria.

To test the importance of acquired immunity in CNS listeriosis, WT BALB/c mice were immunized by i.p. inoculation of 3 × 10⁴ LM 14 days before CNS infection. Immunized and naive WT mice were then infected i.pa. with LM. When 3 × 10² CFU of LM were used for infection, both naive and immunized mice survived (Fig. 1,A and data not shown). It should be noted that whereas it took ∼14 days for naive mice to eliminate LM from the brain, immunized mice cleared the bacteria by day 3 p.i. Peak bacterial burdens in naive mice were ∼20-fold higher than those of immunized mice (data not shown). When 3 × 10³ CFU of LM were administered, most naive mice died within 2 wk while all immunized mice survived the infection (Fig. 4,A). Intracerebral bacterial burdens were significantly lower in immunized mice compared with naive mice and immunized mice eradicated the bacteria by day 3 p.i. (Fig. 4 B). These data clearly demonstrate that systemic immunization activates acquired immunity, which then functions in the infected brain.

Since LM-specific CD8⁺ T cells play an important role in systemic listeriosis (12), we examined LM-specific CD8⁺ T cells in the i.pa.-infected mice using the H-2K^d tetramer complexed with a peptide derived from LLO (Fig. 4, C and D). Mononuclear cells isolated from the brain and the spleen on day 7 p.i. were stained with anti-CD8 mAb and H-2K^d LLO tetramer (Fig. 4,C). Approximately 25% of CD8⁺ T cells infiltrating the brain were tetramer⁺ LM-specific CD8⁺ T cells. Such LM-specific T cells were also detected in the spleen (Fig. 4,C). The peak of tetramer⁺ T cells in the brain was observed on day 7 p.i., consistent with the peak of leukocyte infiltration (Fig. 4 D). These data collectively indicate that i.pa. infection of LM effectively induces LM-specific adaptive immune responses in the brain.

To further investigate the contribution of acquired immunity in CNS listeriosis, RAG2^−/− mice on a BALB/c background (BALB/c-RAG2^−/− mice) were i.pa. infected with 3 × 10² CFU of LM. RAG2^−/− mice survived at least 7 days after the infection, but succumbed to the disease by day 14 p.i. (Fig. 5,A). Nearly identical results were obtained with RAG2^−/− mice on a C57BL/6 background (data not shown). There was a significant difference in intracerebral bacterial burdens between WT and RAG2^−/− mice on day 5 p.i. and thereafter. Although the number of bacteria in the brain was reduced in WT mice by day 8 p.i., RAG2^−/− mice were unable to control bacterial growth and harbored increased numbers of bacteria in the brain (Fig. 5 B).

To examine the lymphocyte subset critical for the process of LM elimination, naive T cells from WT BALB/c mice were adoptively transferred to BALB/c-RAG2^−/− mice before LM infection. As shown by the closed circles in Fig. 5,A, RAG2^−/− mice transferred with T cells survived the infection and excluded LM from the CNS. Next, to determine the importance of CD4⁺ or CD8⁺ T cells for the clearance of LM from the brain, each cell population was depleted from WT BALB/c mice by the administration of specific Abs throughout the course of i.pa. infection. As shown in Fig. 6,A, all mice infected with 3 × 10² CFU of LM survived CNS listeriosis without CD4⁺ T cells. In contrast, the depletion of CD8⁺ T cells resulted in the death of all mice by day 16 p.i. (Figs. 6,A and 8,A). When mice were infected with 1 × 10³ CFU of LM, the depletion of CD4⁺ T cells resulted in a lower survival rate than control mice (Fig. 6,B). Under the same condition, the depletion of CD8⁺ T cells showed more severe effects (Fig. 6,B). Although no significant difference in bacterial burdens was observed on day 5 p.i., mice without CD8⁺ T cells harbored a significantly higher bacterial burden than CD4⁺ T cell-depleted mice on day 10 p.i. (Fig. 6,C). In addition, the depletion of CD4⁺ T cells resulted in the increase of bacterial burden compared with control mice (Fig. 6,C). To confirm these results, purified CD4⁺ or CD8⁺ T cells from WT BALB/c mice were adoptively transferred into BALB/c-RAG2^−/− mice before i.pa. infection with 3 × 10² CFU of LM. Consistent with the results shown in Fig. 6 A, RAG2^−/− mice survived LM infection after CD8⁺ T cell transfer, while CD4⁺ T cell transfer did not rescue RAG2^−/− mice (data not shown). These results clearly indicate the importance of CD8⁺ T cells compared with CD4⁺ T cells in the control of CNS listeriosis.

Flow cytometric analyses using anti-CD8 Ab and H-2K^d LLO tetramer showed the presence of LM-specific T cells in the brain even in the absence of CD4⁺ T cells (Fig. 6,D). The proportion of LLO-specific CD8⁺ T cells in the infected brain on day 7 p.i. was higher than that in the spleen, indicating that LLO-specific CD8⁺ T cells expanded and accumulated in the brain without CD4⁺ T cell help. However, the proportion of LM-specific T cells was significantly lower in mice depleted of CD4⁺ T cells than control mice (Figs. 6,D and 4 C). These results suggest that CD4⁺ T cells play some roles but are not essential for the generation and migration of LM-specific CD8⁺ T cells and the clearance of LM from the brain. This is consistent with the result shown for systemic listeriosis (18).

We next examined the mechanisms by which CD8⁺ T cells function to eliminate bacteria from the brain in CNS listeriosis. Since microglia are the major brain cell population infected by LM, we performed CTL assays to examine whether CD8⁺ T cells exhibit cytotoxic activity against microglia infected with LM. Primary cultures of microglia were loaded with heat-killed LM in vitro and used as target cells for CD8⁺ T cells isolated from either naive or LM-immunized mice. At the E:T ratio of 10 and 50, LM-immunized CD8⁺ T cells showed definitely higher cytotoxic activity than naive CD8⁺ T cells (Fig. 7 A). These results suggest that Ag-specific CD8⁺ T cells exclude LM from the brain at least in part through cytotoxic activity on LM-infected cells.

One of the molecular mechanisms of CD8⁺ T cell-mediated cytotoxicity is Fas-FasL interaction. In fact, the expression of FasL is readily detected on CD8⁺ T cells in the brain on day 7 p.i. (Fig. 7,B). To examine the in vivo importance of FasL on CD8⁺ T cells, CD8⁺ T cells from either WT or gld/gld mice, which lack a functional FasL gene product, were adoptively transferred to RAG2^−/− mice on a C57BL/6 background and those mice were infected with 3 × 10² CFU of LM i.pa. Although RAG2^−/− mice transferred from WT mice survived the CNS listeriosis, 60% of RAG2^−/− mice transferred from gld/gld mice succumbed to the infection (Fig. 7 C). These data indicate that FasL expressed on CD8⁺ T cells contributes to the clearance of LM from the brain in vivo.

IFN-γ produced by host cells plays an important role in protection against LM, particularly during the early phase of systemic infection (6, 7, 8). However, little is known about the importance of IFN-γ in CNS listeriosis. To determine whether IFN-γ plays a protective role against CNS listeriosis, we infected BALB/c- and B6-IFN-γ^−/− mice i.pa. with LM. As shown in Fig. 8,A, BALB/c-IFN-γ^−/− mice (○) died within 4 days p.i., indicating that IFN-γ is also critical in the control of CNS listeriosis. The same results were obtained for C57BL/6-IFN-γ^−/− mice (data not shown). When the serum IFN-γ levels in WT BALB/c mice i.pa infected with LM were examined by ELISA, the peak of serum IFN-γ was observed on days 3–4 p.i. (Fig. 8,B). These results suggest that IFN-γ secreted in the acute phase seems to be important in the control of LM during the early phase of CNS listeriosis. Furthermore, since the peak of IFN-γ production was observed on days 3–4 p.i., it is likely that IFN-γ is produced by innate immune cells such as NK cells and/or dendritic cells (7, 8). In addition, CD8⁺ T cells are also known to produce IFN-γ in the innate phase of listeriosis (26). To estimate the relative contribution of NK cells and CD8⁺ T cells to host survival and IFN-γ production in our CNS listeriosis model, we depleted NK cells and/or CD8⁺ T cells from WT BALB/c mice using anti-asialo GM1 Ab and/or anti-CD8 mAb (clone 2.43) and infected i.pa. with 3 × 10² CFU of LM. Since the administration of anti-asialo GM1 Ab is known to deplete not only NK cells but also asialo GM1⁺CD8⁺ T cells (19), the contribution of NK cells to the infection was estimated by comparing mice depleted of only CD8⁺ T cells with mice depleted of both CD8⁺ T cells and asialo GM1⁺ cells. As observed in Fig. 6, mice devoid of CD8⁺ T cells survived the acute phase of infection and lived significantly longer than IFN-γ^−/− mice and mice treated with anti-asialo GM1 Ab (Fig. 8,A). There was little difference in survival between mice depleted of asialo GM1⁺ cells and mice depleted of both CD8⁺ T cells and asialo GM1⁺ cells. Serum IFN-γ was significantly lower in anti-asialo GM1 Ab-treated mice than in control mice during the first 3 days (Fig. 8,C). Although serum IFN-γ was significantly lower in CD8⁺ T cell-depleted mice than in control mice during the first 3 days, they still produced a significantly higher level of IFN-γ than asialo GM1⁺ cell-depleted mice (Fig. 8 C). These results strongly indicate that IFN-γ produced by NK cells is critically important in host defense in the acute phase of CNS listeriosis and that the contribution of CD8⁺ T cell-derived IFN-γ is relatively lower than NK cell-derived IFN-γ.

When T cells were adoptively transferred from BALB/c-IFN-γ^−/− mice into BALB/c-RAG2^−/− mice before challenge with LM, the recipient mice survived the infection and cleared the bacteria from the brain (Fig. 8 A, •). These results also indicate that T cell-derived IFN-γ is dispensable for the control of CNS listeriosis in the presence of IFN-γ derived from innate immune cells.

In this study, we have established a new model for CNS listeriosis, which has several advantages over previously used models. First, our new protocol induces focal encephalitis with mild meningitis in WT mice. The mice survive the infection and eliminate LM from the brain in contrast to the previous model where mice succumb to infection. Such distinctions are likely due to the differences in the infection protocols between those two models. In the previous model, the volume of bacterial suspension was ∼30 μl (20), which seemed too excessive to confine bacteria within the parenchyma (see Fig. 1,E). In our new model, LM were suspended in a very small volume of PBS (1 μl) and injected into the forebrain parenchyma at a position distant from the ventricle (see Fig. 1,C), which enabled us to keep LM localized to the parenchyma without leaking into the ventricular system. When we intentionally injected a bacterial suspension into the lateral ventricle (see Fig. 1 D), a completely different pathology was induced. The survival rate, time course of bacterial burdens, and histological characteristics in the i.vt.-infected mice were very similar to those of the previous CNS listeriosis model induced by direct injection of a relatively large volume of bacterial suspension into the brain parenchyma. It is thus likely that when LM were applied to the brain parenchyma in a large volume of suspension, bacteria leaked into the ventricular system and caused ventriculitis, leading to the early death of infected animals (14, 20, 21, 22). The reasons for the distinctly severe symptoms and high mortality caused by i.vt. LM are not clear at the moment. One possibility is that under such conditions, bacteria can grow freely in the extracellular cerebrospinal fluid space.

A second advantage is that our new model seems to reflect the natural course of CNS listeriosis better than the previous model. In human and domestic animals, it is well known that encephalitis and meningitis but not ventriculitis are the major listeriosis-related brain pathologies (4, 23). In our new i.pa. infection model, mice suffer from focal encephalitis and mild meningitis without ventriculitis and recover from the disease. Therefore, this model more closely mimics naturally occurring listerial brain infection than the previous model.

Third, because mice recover from the infection in our model, acquired immunity can be easily studied during the entire course of infection. In the previous model that results in fatal ventriculitis, naive mice died within 5 days (13), a period too short to study the role of acquired immunity. Our study shows that the kinetics of bacterial burdens in the brain is different from that of systemic infection. The peak of the bacterial load occurred at ∼5–7 days p.i. and 10–14 days were required for the total exclusion of bacteria from the infected brains (Fig. 1 B). In contrast, in the systemic LM infection model induced by i.v. or i.p. bacterial administration, the peak bacterial burden is on about day 3 p.i., and 7 days are usually sufficient to exclude bacteria from the spleen and liver (2, 24). Patterns of leukocyte infiltration in the brain seem to correlate with the kinetics of bacterial clearance and our data illustrate the overall process by which CNS listeriosis is controlled by various cell populations. During the early phase of infection such as on day 3 p.i., macrophages and NK cells likely play roles in the brain, reflecting the activation of innate immunity. On day 7 p.i., acquired immune cells increase in the brain, suggesting that day 7 p.i. is the transitional time point from the innate to the acquired immune responses. Bacterial clearance is evident once T cells are recruited to the brain, suggesting an important role for T cells in the clearance of bacteria.

We noted that the numbers of B cells are very low compared with those of other cell types, suggesting that this population does not play important roles in clearing LM from the brain (Fig. 3). Another notable characteristic is the enhanced infiltration of γδ T cells, which have been reported to play a protective function against LM in the systemic infection model (25). The role of γδ T cells in CNS listeriosis and the mechanisms by which γδ T cells are recruited to the brain remain to be elucidated.

The production of IFN-γ by innate immune cells is important in limiting bacterial growth during the early phase of infection (5). Our new CNS listeriosis model provides us with an opportunity to examine the relative contributions of innate vs adaptive immunity as well as the role of IFN-γ against CNS listeriosis. As shown in Fig. 5,A, RAG2^−/− mice survived for at least 7 days after LM inoculation in the brain, longer than IFN-γ^−/− mice, which succumbed to the disease within 4 days despite the presence of adaptive immunity (Fig. 8,A). The facts that the peak level of IFN-γ in the sera of WT as well as RAG2^−/− mice upon i.pa. LM infection was observed on days 3–4 p.i. (Fig. 8,B and data not shown) and that IFN-γ^−/− mice succumb to the infection within 4 days strongly suggest that the production of IFN-γ in the early phase of infection is critical for the control of the initial growth of bacteria in CNS listeriosis. Interestingly, RAG2^−/− mice adoptively transferred with T cells from IFN-γ^−/− mice survive CNS listeriosis (Fig. 8 A), suggesting that IFN-γ produced by T cells is dispensable.

NK cells and APCs such as dendritic cells and macrophages are innate immune cells that are capable of producing IFN-γ in systemic listeriosis (7, 8, 26). In the CNS listeriosis model established in this study, NK cells were identified as cells playing a major role for acute phase IFN-γ production and survival (Fig. 8, A and C). The effect of NK cell depletion was more prominent on CNS listeriosis than on systemic listeriosis (Fig. 8 and Refs. 8 and 26). Using CD11c-DTR-transgenic mice in which the expression of diphtheria toxin receptor is driven by the CD11c promoter (27), we generated CD11c⁺ cell (dendritic cells)-depleted mice by administering diphtheria toxin. These mice produced levels of serum IFN-γ similar to WT mice upon i.pa. infection of LM, whereas systemic LM infection resulted in a lower serum IFN-γ level in CD11c⁺ cell-depleted mice than control mice (data not shown). These results indicate that NK cells play a dominant role in the production of IFN-γ during the early phase of CNS listeriosis. However, as shown in Fig. 8 C, depletion of CD8⁺ T cells resulted in decreased IFN-γ production, suggesting that CD8⁺ T cells contribute to the early immune response to some extent. The precise characteristics of T cells contributing to this early immune response are currently obscure. γδ T cells or memory type T cells may contribute to this response.

Our results clearly show the importance of CD8⁺ T cells in adaptive immune responses during the later phase of infection. A more important role for CD8⁺ T cells over CD4⁺ T cells in the control of systemic listeriosis was reported in previous studies (9, 10). In the classical CNS listeriosis model in which LM infection results in fatal ventriculitis, both CD4⁺ and CD8⁺ T cells were shown to be equally indispensable (20). In contrast, our results using a new i.pa. infection model show that CD4⁺ T cells play some roles but are dispensable and that CD8⁺ T cells play a critical role in the clearance of LM in CNS listeriosis as shown in Figs. 6 and 7. Furthermore, we demonstrate the induction of Ag-specific CD8⁺ T cells in the brain using H-2K^d LLO tetramer staining as observed in systemic listeriosis where Listeria-specific CD8⁺ T cells are required for the exclusion of bacteria (28). Indeed, we show the induction of Ag-specific CTLs capable of exhibiting cytotoxicity against microglia loaded with LM.

The up-regulation of FasL was significant only on brain-derived CD8⁺ T cells, but not on spleen-derived CD8⁺ T cells in CNS listeriosis (Fig. 7,B). In addition, our results indicate that FasL plays a crucial role in the clearance of LM from the brain in vivo (Fig. 7 C). Fas-FasL-mediated apoptosis of microglia is known to be augmented by IFN-γ in vitro (29). These results collectively suggest that the Fas-FasL system is an important molecular pathway, leading to microglial apoptosis. In addition to their cytotoxic effect in CNS listeriosis, another function of CD8⁺ T cells may be IFN-γ production as cytokine (30)- or TLR (31)-dependent IFN-γ production by CD8⁺ T cells has been reported. However, as discussed above, the contribution of CD8⁺ T cells in the production of IFN-γ seems secondary.

From the present results, we envision the following scenario: when LM invades brain parenchyma, inherent brain cells block LM from freely proliferating and allow the prompt influx of macrophages and NK cells from circulating blood. Macrophages and microglia activated by NK cell-derived IFN-γ limit bacterial growth in the early phase of infection. After Ag- specific CD8⁺ T cells are activated and recruited to the infection site on around day 7 p.i., the bacterial burden decreases and LM are eventually eliminated from the infected brain by day 14 p.i. There are several questions that remain to be answered in future studies. How is phase-specific infiltration of cell type controlled after the infection? What is the most significant APC in CNS listeriosis? What is the role of γδ T cells? Our new model provides a powerful tool with which to answer these questions.

We thank Dr. S. Kohsaka and Dr. K. Ohsawa of the National Institute of Neuroscience for valuable discussions on the preparation of microglia, Dr. K. F. Tanaka of the National Institute of Physiology for helpful advice in the management of microglial experiments, and K. Takei for animal care. Thanks are also due to Dr. L. K. Clayton for critically reading this manuscript.

S.K. is a consultant for MBL. The authors otherwise have no financial conflicts of interest.