IFN-γ ameliorates autoimmune encephalomyelitis by limiting myelin lipid peroxidation

Significance

The mechanisms guiding tissue repair in autoimmune-mediated diseases of the CNS are not fully understood, but cytokines are believed to play an important role. The proinflammatory cytokine IFN-γ has been implicated in the pathogenesis of multiple sclerosis and its animal model, experimental autoimmune encephalomyelitis (EAE). Paradoxically, knockout studies in mice showed that EAE is exacerbated when the cytokine’s signaling is disrupted, but its mechanism of protection has not been resolved. Here, we show a critical role for IFN-γ in activating antigen-presenting cells toward a protective phenotype in the CNS by removing extracellular myelin debris and limiting the availability of substrate for oxidative stress-generated neurotoxic lipid peroxidation products.

Abstract

Evidence has suggested both a pathogenic and a protective role for the proinflammatory cytokine IFN-γ in experimental autoimmune encephalomyelitis (EAE). However, the mechanisms underlying the protective role of IFN-γ in EAE have not been fully resolved, particularly in the context of CNS antigen-presenting cells (APCs). In this study we examined the role of IFN-γ in myelin antigen uptake by CNS APCs during EAE. We found that myelin antigen colocalization with APCs was decreased substantially and that EAE was significantly more severe and showed a chronic-progressive course in IFN-γ knockout (IFN-γ^−/−) or IFN-γ receptor knockout (IFN-γR^−/−) mice as compared with WT animals. IFN-γ was a critical regulator of phagocytic/activating receptors on CNS APCs. Importantly, “free” myelin debris and lipid peroxidation activity at CNS lesions was increased in mice lacking IFN-γ signaling. Treatment with N-acetyl-l-cysteine, a potent antioxidant, abolished lipid peroxidation activity and ameliorated EAE in IFN-γ–signaling-deficient mice. Taken together the data suggest a protective role for IFN-γ in EAE by regulating the removal of myelin debris by CNS APCs and thereby limiting the substrate available for the generation of neurotoxic lipid peroxidation products.

IFN-γ has been implicated in the pathogenesis of multiple sclerosis (MS) and its animal model, experimental autoimmune encephalomyelitis (EAE) (1). The role of the Th1-derived cytokine IFN-γ in disease pathology was supported further by clinical evidence showing that MS patients treated with recombinant IFN-γ in clinical trials developed more severe inflammation (2). Paradoxically, EAE studies using IFN-γ knockout (IFN-γ^−/−) mice or neutralizing mAb (anti–IFN-γ mAb) showed that EAE was exacerbated under these conditions (3, 4). These results suggested that IFN-γ may be protective in EAE, for example by inducing T-cell apoptosis (5).

More recent studies suggest a central role for other cytokines and T-cell subsets, such as Th17 cells, in several autoimmune diseases, including EAE (6–8). Among the T-cell subsets with the highest pathogenic potential in both EAE and MS are Th17 cells coexpressing IFN-γ, known as “Th1-like” Th17 cells (9–12), or coexpressing GM-CSF (7, 8). The protective versus disease-promoting role of IFN-γ in EAE and MS, and in particular the potential protective role of this cytokine in the context of CNS antigen-presenting cells (APCs) in EAE, has not been fully resolved.

The effects of IFN-γ on APCs are pleiotropic and encompass up-regulation of MHC molecules, induction of reactive oxygen species (ROS) production, phagocytic activity, and increased production of other proinflammatory cytokines including TNF (13). CNS APCs, such as resident microglia and infiltrating peripheral macrophages and dendritic cells (DCs), have been shown to express MHC class II molecules as well as costimulatory molecules, including CD80 and CD86, that are important for the pathogenesis of EAE (14, 15). However, the roles of CNS APCs, including the mechanisms guiding the uptake and clearance of free myelin debris by CNS APCs during EAE, and the mechanisms that limit the extent of CNS damage during neuroinflammation are not well established.

It has been suggested that intervention in ROS generation potentially could represent a novel therapeutic strategy to reduce inflammation in the CNS during MS (16). Along these lines, it has been reported that cytokines such as IFN-γ can stimulate microglia and macrophages to break down oxidized lipids as a self-limiting mechanism to prevent host tissue from inadvertent oxidative damage (17).

Thus, in this study we investigated the mechanisms guiding myelin antigen (Ag) uptake and clearance by CNS APCs and found that the uptake of myelin Ag was related directly to the presence of IFN-γ. In C57BL/6 WT mice with EAE, most CNS APCs contained myelin Ag, whereas the number of myelin Ag⁺ APCs was dramatically decreased in IFN-γ^−/− and IFN-γ receptor (IFN-γR)^−/− mice, despite substantially increased disease severity. Importantly, free myelin debris at CNS lesions was increased and lipid peroxidation activity was enhanced in mice lacking IFN-γ signaling. Treatment with N-acetyl-l-cysteine (NAC), a potent antioxidant, abolished lipid peroxidation activity and ameliorated EAE in IFN-γ–signaling-deficient mice.

Taken together, the data suggest IFN-γ has a protective role in EAE by regulating myelin debris removal by CNS APCs and limiting the substrate available for the generation of neurotoxic lipid peroxidation products.

Results

IFN-γ Modulates Myelin Ag Uptake by CNS APCs.

Our previous work showed that ∼3% of microglia in the CNS of naive mice contain myelin Ag, including myelin basic protein (MBP), myelin oligodendrocyte glycoprotein (MOG), and proteolipid protein (PLP) (18). Upon induction of EAE, microglia were the first myeloid cells in the CNS to show increased myelin uptake. In contrast, DCs initially detected in the CNS after the induction of EAE did not contain myelin Ag, and myelin Ag-containing DCs were detected only several days later. The onset of clinical EAE coincided with a sharp increase in the numbers of CD4⁺ T cells as well as myelin Ag⁺ DCs. The timing and close association of increased myelin Ag uptake by CNS APCs and increased inflammatory infiltrates suggested that myelin Ag uptake was promoted by inflammatory mediators such as the proinflammatory cytokine IFN-γ.

To begin to test this hypothesis, WT, IFN-γ^−/−, or IFN-γR^−/− mice on the C57BL/6 background were immunized with MOG_35–55 peptide to induce EAE. Consistent with previous reports, IFN-γ^−/− and IFN-γR^−/− mice developed more severe clinical EAE than WT control animals (Fig. S1A). Furthermore, clinical signs of EAE persisted in IFN-γ^−/− and IFN-γR^−/− mice beyond day 25 postimmunization (p.i.), whereas EAE had remitted in WT mice by that time (Fig. S1A).

Fig. S1.

Increased EAE severity and CNS inflammation in IFN-γ^−/− and IFN-γR^−/− mice. IFN-γ^−/−, IFN-γR^−/−, and C57BL/6 WT mice were immunized with MOG_35–55 and scored daily for clinical EAE. (A) EAE clinical disease scores for WT, IFN-γ^−/−, and IFN-γR^−/− mice. Data are shown as the mean EAE score ± SD; n ≥ 15 mice per group over three independent experiments. The asterisk indicates a cumulative EAE score significantly increased in IFN-γ^−/− and IFN-γR^−/− mice versus WT mice for the time points designated by the horizontal line (P ≤ 0.05; Student’s t test). (B) CD4⁺ T-cell numbers in brain tissue in WT, IFN-γ^−/−, and IFN-γR^−/− mice. Data are shown as the mean number of cells per brain slice ± SD; n ≥ 12 mice per group over three independent experiments (*P ≤ 0.05; Student’s t test). (C) T-cell responses in spleen and CNS of IFN-γR^−/− and C57BL/6 WT mice at disease peak (day 21 p.i.) by cytokine ELISPOT assay for IFN-γ, IL-17, GM-CSF, and IL-5. Data are shown as the mean number of cytokine-producing cells (with medium control subtracted), ± SD; n ≥ 9 mice per group over three independent experiments. Asterisks indicate a significant increase in the number of spots over WT mice (P ≤ 0.05; Student’s t test).

Beginning to investigate immunopathology of the CNS in these animals, we observed that CD4⁺ T cells persisted in high numbers in IFN-γ^−/− and IFN-γR^−/− mice beyond day 25 p.i., whereas their numbers had decreased substantially in WT mice at this time point (Fig. S1B). Because it has been suggested that IFN-γ is proapoptotic for T cells, possibly via the induction of reactive oxygen/nitrogen species (5, 19), we tested whether T-cell apoptosis was altered in the CNS of IFN-γR–deficient mice. However, we observed only a slight, statistically insignificant, change in the percentage of apoptotic CD4⁺ T cells by caspase-3 and TUNEL staining in IFN-γR^−/− mice compared with WT mice at day 23 p.i. (Fig. S2), a point at which CNS T cells were reduced significantly in WT animals (Fig. S1B). Furthermore, we observed that the frequencies of MOG_35–55–reactive T cells producing IFN-γ, IL-17, or GM-CSF by cytokine ELISPOT assay were increased only slightly in the spleen or CNS of IFN-γR^−/− mice compared with WT animals (Fig. S1C); this increase did not appear substantial enough to account fully for the strongly enhanced disease severity in the absence of IFN-γ signaling.

Fig. S2.

No significant difference in T-cell apoptosis was seen in IFN-γ–signaling-deficient mice compared with WT mice. EAE was induced in C57BL/6 WT and IFN-γR^−/− mice with MOG_35–55, and mice were killed between the peak and chronic stage of disease (day 23 p.i.). (A) CD4⁺ T cells (green) were costained with either caspase-3 (red) (Upper Row) or TUNEL assay (red) (Lower Row) to identify double-positive apoptotic CD4⁺ T cells (small arrows) in brain tissue slices from EAE mice at day 23 in WT and IFN-γR^−/− mice by immunofluorescence staining. (Scale bars, 20 µm.) (B) Quantification of total CD4⁺ T cells (black bars) or apoptotic CD4⁺ T cells (gray bars) as determined by caspase-3 (Upper) or TUNEL assay (Lower). Data are shown as the mean number of cells per brain tissue slice ± SD; n ≥ 9 over three independent experiments. ns, no significant change in the percentage of double-positive (apoptotic) cells in IFN-γR^−/− mice versus the WT control group (P ≤ 0.05; Student’s t test).

Next, we investigated whether myelin Ag uptake by CNS APCs was altered in the absence of IFN-γ signaling, because this alteration conceivably could contribute to disease severity. To begin to address this issue, EAE was induced by immunization of WT, IFN-γ, IFN-γR^−/−, and WT mice treated with neutralizing anti–IFN-γ mAb with MOG_35–55 peptide. Representative animals were killed at key time points after the onset of clinical EAE (days 13, 17, 21, and 25 p.i.). Staggered brain tissue sections were stained by immunofluorescence using mAbs to identify myeloid lineages (CD11c⁺ DCs, CD11b⁺ microglia/macrophages) and were analyzed by confocal laser-scanning microscopy for colocalization with myelin Ag (MBP, MOG, and PLP) as previously described (18).

Consistent with our previous study, large numbers of CD11b⁺ microglia/infiltrating macrophages and CD11c⁺ DCs were present in the CNS of WT and mutant mice after the onset of EAE (around day 13 p.i.), and the number of APCs peaked at approximately day 20–21 p.i. (Fig. 1A, black bars) (18). The majority (>60%) of APCs in the CNS of WT mice colocalized with one or more myelin Ag (Fig. 1 shows representative data for MBP). Of note, the percentage of APCs containing myelin Ag in WT mice remained relatively constant over the course of disease, including at disease remission when the overall number of APCs declined sharply. In contrast, IFN-γ^−/− and IFN-γR^−/− mice showed a significantly smaller percentage of MBP⁺ APCs at most time points (Fig. 1A) and overall (average of 24% and 21% across all time points for IFN-γ^−/− and IFN-γR^−/− mice, respectively) (Fig. 1A), and large extracellular accumulations of myelin Ag frequently were present in CNS lesions (Fig. 1B, small arrows). Of note, the results in WT mice treated with anti–IFN-γ mAb were similar to those in mutant mice, suggesting that early and late effects of impaired IFN-γ signaling were comparable.

Fig. 1.

Altered colocalization of myelin Ag with CNS APCs in mice with impaired IFN-γ signaling. EAE was induced in C57BL/6 WT, IFN-γ^−/−, and IFN-γR^−/− mice and in WT mice treated with anti–IFN-γ neutralizing antibody with MOG_35–55 peptide immunization as described in Materials and Methods. (A) Quantification of myelin Ag-positive (MBP⁺) CD11b⁺ microglia/macrophages (Upper Row) and CD11c⁺ DCs (Lower Row) from reconstructed confocal laser-scanning microscopy z-stacks. Numbers shown represent the mean number of cells per brain tissue slice ± SD; n ≥ 9 mice per group per time point over three independent experiments. Asterisks indicate a significant difference in the percentage of MBP⁺ versus WT APCs at each time point (P ≤ 0.05; Student’s t test). (B) Internal slices from original z-stack images showing myelin Ag (red) and CNS APCs (green) in lesions during EAE. Colocalization of myelin Ag with APCs (yellow, large arrows) is indicated in representative cells. Clusters of extracellular myelin debris are indicated by small arrows. (Scale bars, 20 µm.)

These findings were not limited to the C57BL/6 background and the MOG peptide-induced EAE model, because the numbers of myelin Ag⁺ CNS APCs from recipient SJL mice treated with anti–IFN-γ mAb were reduced similarly, despite a significant increase in clinical EAE severity compared with PBS-treated mice (Fig. S3) following adoptive transfer of PLP_139–151–specific CD4⁺ T cells in SJL mice as previously described (18).

Fig. S3.

The effect of IFN-γ deficiency is not restricted to the C57BL/6 background or the active immunization model. EAE was induced in SJL recipient mice by adoptive transfer of PLP_139–151 T cells, and mice were treated with PBS or anti–IFN-γ mAb throughout the EAE disease course. (A) Mice were observed and EAE disease was scored for 20 d post adoptive transfer (p.t.). Data are shown as the mean EAE score ± SD; n ≥ 10 mice per group over two independent experiments. The asterisk indicates the cumulative EAE score increased significantly in SJL mice treated with anti–IFN-γ versus PBS-treated control mice for the time points designated by the horizontal line (P ≤ 0.05; Student’s t test). (B) Brain tissue was recovered from representative animals in A at onset (day 7 p.t.), peak (day 14 p.t.), and recovery (day 20 p.t.). Shown are internal slices from original z-stack images showing myelin Ag (red) and CD11c⁺ APCs (green) in lesions during EAE. Colocalization of myelin Ag with APCs (yellow, large arrows) is seen in WT mice. Accumulation of extracellular myelin debris not colocalized to APCs is found in SJL mice treated with anti–IFN-γ mAb (small arrows). (Scale bars, 50 µm.) (C) Quantification of the number of myelin Ag⁺ (MBP⁺) CD11b⁺ microglia/macrophages or CD11c⁺ DCs in SJL mice treated with anti–IFN-γ mAb or control mice treated with PBS. Data are shown as the mean number of cells per brain tissue slice ± SD; n ≥ 8 mice per group per time point over two independent experiments. Asterisks indicate a significant decrease in the percentage of MBP⁺ APCs in mice treated with anti–IFN-γ versus PBS-treated control animals (P ≤ 0.05; Student’s t test). (D) Quantification of extracellular versus intracellular myelin Ag in SJL mice treated with anti–IFN-γ mAb and in PBS-treated control mice. Data are shown as the mean of the total free MBP ± SD; n ≥ 10 mice per group over two independent experiments. The asterisk indicates a significant change in the volume of free MBP over WT (P ≤ 0.05; Student’s t test).

Free myelin in the extracellular environment has been implicated in CNS neurotoxicity, and its removal by CNS microglia has been suggested to enhance regeneration in lesions (20, 21). We observed fewer CNS APCs containing myelin Ag in lesions in IFN-γ^−/− and IFN-γR^−/− mice with EAE. Conceivably, this observation could be caused by accelerated myelin Ag breakdown in APCs in which IFN-γ signaling is disrupted; under such conditions one would predict less extracellular myelin Ag at lesions. Alternatively, IFN-γ–signaling-deficient APCs could be impaired in phagocytosing myelin Ag; this impairment would be predicted to result in increased extracellular myelin Ag in CNS lesions. To differentiate between these possibilities, we used 3D remodeling of APC cell surface and quantification of MBP volume (as a surrogate marker for myelin Ag) both within cells and in the surrounding lesions in confocal images of the CNS during EAE (18). The results show that most MBP at lesions was contained within or at the surface of APCs in C57BL/6 WT mice (Fig. S4). In contrast, most myelin Ag within CNS lesions in IFN-γ^−/− and IFN-γR^−/− mice was located extracellularly, and only a small percentage of the total MBP was contained within APCs (Fig. S4). Similar results were obtained in the PLP_139–151–induced adoptive transfer EAE model in SJL mice (Fig. S3).

Fig. S4.

Increased extracellular myelin is present in IFN-γ^−/− and IFN-γR^−/− mice during EAE. EAE was induced in C57BL/6 WT, IFN-γ^−/−, and IFN-γR^−/− mice with MOG_35–55 peptide immunization as described in Materials and Methods. The relative percentages of the volume of free myelin Ag found either within APCs (intracellular) or outside APCs (extracellular) in the CNS lesions of WT, IFN-γ^−/−, and IFN-γR^−/− EAE mice at the peak of disease were quantified by confocal microscopy imaging analysis. Data are shown as the mean MBP volume [total number of voxels positive for red signal found within (intracellular) or outside (extracellular) APCs, expressed in cubic millimeters], ± SD; n ≥ 12 mice per group over three independent experiments. Asterisks indicate a significant change in MBP volume versus WT (P ≤ 0.05; Student’s t test).

Taken together, the data showed that in the absence of IFN-γ or IFN-γR the number and percentage of CNS APCs containing myelin Ag was decreased despite substantially enhanced clinical EAE, suggesting that myelin Ag uptake by APCs was inversely correlated with clinical disease severity in IFN-γ^−/− and IFN-γR^−/− mice. Moreover, the data suggested impaired myelin Ag uptake by CNS APCs in the absence of IFN-γ signaling.

EAE Severity Is Decreased Under Conditions of Limited Availability of Myelin Ag.

Our studies revealed a significant increase in free myelin Ag in the CNS of mice with disrupted IFN-γ signaling and a corresponding significantly enhanced EAE severity. These results raised the question of whether increased extracellular myelin Ag contributed mechanistically to increased disease severity. To begin to address this question, we took advantage of MBP-deficient shiverer mice backcrossed to the HLA-DR2b (DRB1*15:01) background. These mice were backcrossed to the C57BL/6 background and allow the induction of EAE with MOG_35–55 peptide (18). In early characterization of the shiverer mutation, it was shown that MBP^+/shi mice express ∼50% of the MBP found in WT MBP^+/+ mice (22). We reasoned that if free myelin Ag increased the severity of EAE, the disease should be less severe in animals in which IFN-γ signaling is disrupted, because less MBP, and therefore less extracellular myelin debris, should be available during conditions of neuroinflammation and demyelination.

HLA-DR2b transgenic (tg) MBP^+/+ and MBP^+/− mice were immunized with MOG_35–55 to induce EAE. Respective groups of animals were treated with anti–IFN-γ mAb as described in Materials and Methods, and the animals were observed for clinical disease for up to 30 d. Treatment with mAb was initiated on day 8 p.i. to allow the development and expansion of MOG_35–55–reactive T cells and therefore to test the effect of IFN-γ specifically on the progression of disease rather than on disease initiation. EAE was significantly less severe in HLA-DR2b tg MBP^+/− mice than in MBP^+/+ littermates (Fig. 2A). Furthermore, EAE remitted substantially earlier in MBP^+/− mice than in MBP^+/+ mice. Importantly, treatment with anti–IFN-γ mAb showed only a mild disease-enhancing effect in MBP^+/− mice as compared with MBP^+/+ mice (Fig. 2A, open triangles versus open circles).

Fig. 2.

Reduced availability of myelin Ag correlates with decreased EAE severity. (A) EAE clinical disease scores of HLA-DR2b MBP^+/− or MBP^+/+ mice treated with PBS or anti–IFN-γ mAb. The horizontal line indicates days of treatment with anti–IFN-γ mAb. The asterisk indicates significantly (*P ≤ 0.05; Student’s t test) increased cumulative EAE score in MBP^+/+ mice treated with anti–IFN-γ mAb. (B) Numbers of CD4⁺ T-cells in MBP^+/+ and MBP^+/− mice following indicated treatment. The asterisk indicates a significant (*P ≤ 0.05) decrease in MBP^+/− compared with MBP^+/+ mice. (C) Frequencies of cytokine-producing T cells (SFC) by ELISPOT assay in spleen and CNS of MBP^+/+ or MBP^+/− HLA-DR2b mice treated with PBS or anti–IFN-γ mAb. Asterisks indicate a significant (*P ≤ 0.05; Student’s t test) increase. (D, Left) Representative z-stack images of brain tissue slices showing myelin Ag colocalization (yellow, large arrows) of MBP (red) and APCs (green) in CNS lesions of HLA-DR2b MBP^+/− or MBP^+/+ mice treated with PBS or anti–IFN-γ mAb. Extracellular myelin Ag not colocalized to APCs is seen in mice treated with anti–IFN-γ mAb (small arrows). (Scale bars, 20 µm,) (Right) Quantification of total free MBP in MBP^+/+ and MBP^+/− mice after treatment with anti–IFN-γ mAb or PBS. Asterisks indicate a significant (*P ≤ 0.05; Student’s t test) decrease. (E) Quantification of MBP⁺CD11b⁺ microglia/macrophages (Upper Row) and CD11c⁺ DCs (Lower Row) from reconstructed confocal laser-scanning microscopy z-stacks. Asterisks indicate a significant (*P ≤ 0.05; Student’s t test) decrease in the percentages of MBP⁺CD11b⁺ APCs in MBP^+/− mice compared with MBP^+/+ mice. (F) Quantification of the total volume of intracellular MBP found within APCs or extracellular MBP in CNS lesions of PBS or anti–IFN-γ mAb MBP^+/+ or MBP^+/− mice treated with PBS or anti–IFN-γ. Asterisks indicate a significant change (*P ≤ 0.05; Student’s t test) in total MBP volume. Data are presented as mean ± SD for n ≥ 9 mice per group over three independent experiments; ns, not significant.

On examination of CNS tissues from MBP^+/+ and MBP^+/− mice, we found that a significantly lower number of CD4⁺ T cells had infiltrated into the parenchyma of MBP^+/− mice in both treated and untreated groups (Fig. 2B, Right), as compared with MBP^+/+ mice (Fig. 2B, Left), correlating with reduced disease severity overall (Fig. 2A, triangles versus circles). To investigate whether autoreactive T-cell responses were affected in MBP^+/− mice, we performed cytokine ELISPOT assays on both spleen and CNS tissue and found comparable frequencies of IFN-γ– and IL-17–producing MOG_35–55–reactive CD4⁺ T cells in PBS-treated MBP^+/+ and MBP^+/− mice (Fig. 2C, black bars). Similar to our earlier results (Fig. S1C), the cytokine profiles and frequencies of MOG_35–55–reactive T cells were not strikingly affected in MBP^+/− mice as compared with MBP^+/+ mice (Fig. 2C) and therefore were not likely to account for the observed decrease in EAE severity.

Next, we investigated myelin Ag localization to CNS APCs in these mice during EAE using immunofluorescence and confocal microscopy analyses (Fig. 2 D–F). As expected, confocal microscopy analysis showed significantly less free myelin Ag available in lesions of MBP^+/− mice as compared with MBP^+/+ mice (Fig. 2D, Right). Also, significantly fewer microglia and macrophages were detected in PBS-treated MBP^+/− mice than in PBS-treated MBP^+/+ animals, corresponding to the less severe EAE observed (Fig. 2E, Upper Row). In contrast, the numbers of DCs were comparable in the two groups (Fig. 2E, Lower Row). Analysis of myelin Ag uptake by confocal microscopy showed a significant decrease in myelin Ag⁺ APCs in MBP^+/+ mice after treatment with neutralizing anti–IFN-γ mAb, similar to the levels observed in IFN-γ^−/− and IFN-γR^−/− mice (compare Fig. 2 D and E and Fig. 1). Similar to the observations in MBP^+/+ animals (Fig. 2E, compare third and far left panels), treatment of MBP^+/− mice with anti–IFN-γ mAb resulted in a decrease in the percentage of APCs containing myelin Ag⁺ as compared with PBS-treated mice at most time points (Fig. 2E, compare panels at far right and second from the right).

Finally, we observed a proportionally similar increase in extracellular free myelin Ag in lesions in MBP^+/− mice as compared with MBP^+/+ mice upon treatment with anti–IFN-γ mAb (Fig. 2F), suggesting a comparable efficiency in Ag uptake. However, the total amount of myelin Ag (combined intracellular and extracellular free MBP) was significantly decreased in lesions of MBP^+/− mice.

Thus, the results showed that disruption of IFN-γ signaling similarly impaired the ability of CNS APCs to scavenge free myelin Ag in MBP^+/− versus MBP^+/+ mice. However, when the availability of myelin Ag was limited, the overall amount of free myelin Ag was significantly decreased in MBP^+/− mice, corresponding to the substantially ameliorated disease observed in these animals. Collectively, the results support the view that increased levels of free myelin Ag contribute to the severity of EAE.

IFN-γR^−/− APCs Show Decreased Phagocytosis and Increased Activation of CD4⁺ T Cells.

Next, we tested the role of IFN-γ signaling for phagocytosis and/or Ag presentation in vivo during EAE and in vitro by determining the extent of the expression of phagocytic markers and costimulatory markers and Ag-induced T-cell activation. IFN-γR^−/− and WT mice were immunized with MOG_35–55 peptide to induce EAE. The brain and spinal cord were removed at the peak of disease (∼day 21 p.i.) and were processed into single-cell suspensions for flow cytometry analysis. CD45 expression was used to discriminate between CD11b⁺ CD45^lo CNS-resident microglia and CD11b⁺ CD45^hi infiltrating macrophages as previously described (18).

Of note, IFN-γR^−/− CNS-resident microglia (CD11b⁺CD45^lo) showed a significant reduction in the expression of the phagocytic markers Trem2 (Triggering receptor expressed on myeloid cells 2) and CD68 and significant up-regulation of CD172a, a negative regulator of phagocytosis (Fig. 3A). CNS-infiltrating macrophages (CD11b⁺CD45^hi) also showed reduced expression of CD68, but there was no significant difference in the expression of the other markers of phagocytosis. In contrast, DCs (CD11c⁺) were not significantly affected in their expression of phagocytosis markers TREM2 and CD68 in the absence of IFN-γ signaling (Fig. 3A). However, expression of CD172a was dramatically decreased.

Fig. 3.

Cellular phenotypes of CNS-resident and CNS-infiltrating immune cells are altered during EAE when IFN-γ signaling is impaired. (A and B) CNS-resident CD11b⁺CD45^lo microglia, CNS-infiltrating CD11b⁺CD45^hi macrophages, and CD11c⁺ DCs show altered phenotypes of phagocytosis (A) or Ag presentation (B) based on the presence or absence of IFN-γ signaling. Data are shown as the mean of the absolute number of infiltrating cells positive for the indicated marker ± SD; n ≥ 10 mice per group over five independent experiments. Asterisks indicate a significant change in marker expression in IFN-γR^−/− versus WT cells (*P ≤ 0.05; **P ≤ 0.01; Student’s t test; ns, not significant). (C) Altered expression of indicated activation markers on MOG_35–55–specific T-cell receptor (TCR) transgenic CD4⁺ T cells from 2D2 mice activated by WT or IFN-γR^−/− spleen APCs in the presence of MOG_35–55 Ag. Data are shown as the mean of the absolute number of cells positive for the respective surface marker in IFN-γR^−/− vs. WT cells, ± SD. Asterisks indicate a significant change in marker expression in T cells activated by WT versus IFN-γR^−/− APCs over three independent experiments (*P ≤ 0.05; Student’s t test). (D) T-cell responses by cytokine ELISPOT assay for 2D2 CD4⁺ T cells activated by WT or IFN-γR^−/− APCs in the presence of MOG_35–55 Ag. Data are shown as the mean of the number of cytokine spots with medium control subtracted, ± SD. Asterisks indicate a significant increase in number of spots over WT (*P ≤ 0.05; Student’s t test) over three independent experiments.

Investigating molecules involved in T-cell activation and Ag presentation, we observed that the expression of MHC class II molecules was significantly decreased in resident microglia and infiltrating macrophages in the absence of IFN-γ signaling (Fig. 3B). In contrast, DC expression of MHC class II was not affected significantly. Additionally, whereas CD40 expression was decreased on all populations, the expression of the costimulatory molecule CD86 was decreased in both microglia and macrophages but was increased strongly in DCs (Fig. 3B). Thus, DCs appeared to be sufficiently capable of both phagocytosis and Ag presentation in the absence of IFN-γ signaling. Consistent with this notion, 2D2 CD4⁺ T cells cultured with IFN-γR^−/− spleen APCs showed a marked increase in the activation markers CD44 and CD69 by CD4⁺ T cells (Fig. 3C). Additionally, cytokine ELISPOT assay results confirmed that presentation of MOG_35–55 peptide by IFN-γR^−/− spleen APCs induced strong proinflammatory cytokine production by 2D2 T cells (Fig. 3D).

Taken together, these results indicated differential effects of IFN-γ signaling on resident microglia and infiltrating macrophages in the CNS during EAE, compared with DCs, in their ability to uptake myelin Ag and present it effectively to CD4⁺ T cells, resulting in the activation of pathogenic T cells.

Increased Lipid Peroxidation in the Absence of IFN-γ Signaling.

Our results thus far strongly indicated a role for IFN-γ in limiting disease in EAE via modulating myelin Ag uptake and the removal of free myelin Ag by CNS APCs. It has been suggested previously that IFN-γ potentially could limit tissue damage during inflammation by limiting oxidative processes such as peroxidation of lipids (17). Myelin is lipid-rich; thus it was conceivable that peroxidation of lipids and/or proteins could be increased in the CNS of animals with disrupted IFN-γ signaling during EAE, in particular during the chronic phase of disease. To address this possibility, IFN-γR^−/− and WT mice were immunized with MOG_35–55 peptide to induce EAE, and the WT mice were treated with anti–IFN-γ or isotype control mAb. Lipid peroxidation of CNS tissue during EAE was determined by thiobarbituric acid reactive substances (TBARS) assay and expression of 4-hydroxynonenal (4-HNE). The antioxidant role of IFN-γ in EAE was tested by determining whether disease and lipid peroxidation could be inhibited by administering the antioxidant NAC (150 mg/kg, i.p.) daily, beginning on day 8 p.i., to half of the mice from each group.

As shown in Fig. 4, and consistent with previous results, when treated with anti–IFN-γ mAb both WT and IFN-γR^−/− animals exhibited a significant increase in EAE severity and proceeded to a chronic inflammatory phase rather than the monophasic disease course observed in WT animals treated with isotype control mAb (Fig. 4A). In contrast, and in agreement with previous reports, treatment with NAC ameliorated EAE in isotype mAb-treated control WT mice (Fig. 4A, Left, open symbols). Importantly, treatment with NAC strongly curtailed disease severity in both mice treated with anti–IFN-γ mAb and IFN-γR^−/− mice (Fig. 4A, Center and Right, open symbols).

Fig. 4.

Lipid peroxidation is increased in mice lacking IFN-γ signaling and is reduced by NAC treatment. (A) EAE clinical disease scores of C57BL/6 WT mice treated with isotype or anti–IFN-γ mAb (Left and Center) and of IFN-γR^−/− mice (Right). Mice were treated with PBS (closed symbols) or NAC (open symbols) from day 8 p.i. until the end of study (day 30 p.i.) and were scored daily for disease. Data are shown as the mean EAE score ± SD. Asterisks indicate cumulative EAE scores that are significantly decreased in NAC-treated versus PBS-treated mice. (B) CD4⁺ T-cell numbers in NAC-treated mice at the chronic stage of disease (day 25 p.i.). Data are shown as the mean number of cells per brain tissue slice ± SD. (C) Frequencies of cytokine-producing T cells by ELISPOT in spleen and CNS of WT (Left), anti–IFN-γ mAb (Center), or IFN-γR^−/− (Right) mice treated with indicated isotypes. Data are shown as the mean number of cytokine spots with medium control subtracted ± SD. (D) TBARS assay for spinal cords from chronic-phase EAE mice (day 25 p.i.). Data are shown as the mean relative fluorescence units ± SD. Asterisks indicate a significant decrease. Data in A–D are shown as mean ± SD for n ≥10 mice per group over two or three independent experiments (*P ≤ 0.05; **P ≤ 0.01; Student’s t test; ns, not significant). (E) 4-HNE staining of brain tissue sections from chronic-phase EAE (day 25 p.i.) WT mice treated with isotype mAb (Left) or anti–IFN-γ mAb (Center) and IFN-γR^−/− mice (Right). (Upper Row) PBS-treated mice. (Lower Row) NAC-treated animals. (Scale bars, 50 μm.)

When we examined the effect of NAC treatment on the accumulation of inflammatory cells, we found significantly fewer CD4⁺ T cells in the CNS of NAC-treated animals with impaired IFN-γ signaling (Fig. 4B, Middle and Right), whereas the numbers of T cells remained basically unchanged in isotype control mAb-treated WT mice. To test whether NAC treatment affected the frequencies or cytokine differentiation of MOG_35–55–specific CD4⁺ T-cell responses, we performed cytokine ELISPOT assays for IFN-γ, IL-17, GM-CSF, and IL-5. However, no significant differences in CD4⁺ T-cell frequencies or cytokine profiles were detected in the spleen or CNS of mice treated with NAC or PBS control, whether or not IFN-γ signaling was disrupted (Fig. 4C).

We then examined the relative amount of lipid peroxidation occurring in EAE among all groups and treatments using two complementary approaches (Fig. 4 D and E). First, we evaluated lipid peroxidation of myelin in the spinal cords of mice with EAE by the commonly used TBARS assay, which measures the presence of malondialdehyde in tissues, indicative of lipid peroxidation of that tissue (Fig. 4D). Second, we used indirect immunofluorescence staining of the brain tissue of EAE mice with an antibody specific for 4-HNE (Fig. 4E), a product generated specifically by the lipid peroxidation process (23).

The results showed that lipid peroxidation was present in WT mice with EAE at late stages of the disease (day 25 p.i.); however, the amount of thiobarbituric acid reactivity detected was relatively low (Fig. 4D, dark bar in the left pair of bars). Furthermore, 4-HNE staining was detected in these mice (Fig. 4E, Upper Left). NAC treatment did not significantly affect the amount of lipid peroxidation observed in the CNS of WT animals by TBARS assay (16% reduction; Fig. 4D, left pair of bars) or 4-HNE staining (Fig. 4E, Left, compare Upper and Lower Rows). In strong contrast, a significant increase in lipid peroxidation activity was observed in WT mice treated with anti–IFN-γ mAb and in IFN-γR^−/− mice (Fig. 4D, center and right dark bars).

Importantly, NAC treatment resulted in a highly significant reduction of lipid peroxidation products in both WT mice treated with anti–IFN-γ mAb (80% reduction by TBARS; Fig. 4D, center pair of bars) and IFN-γR^−/− mice (82% reduction by TBARS; Fig. 4D, right pair of bars) combined with a strong decrease in 4-HNE staining observed near lesions (Fig. 4E, Center and Right, compare Upper and Lower Rows).

Last, we sought to examine the effect of reduced lipid peroxidation on the colocalization of myelin Ag with CNS APCs and the levels of free myelin debris (Fig. 5). The results showed that NAC treatment dramatically reduced the amount of free myelin Ag in animals lacking IFN-γ signaling (Fig. 5A). Furthermore, NAC-treated mice showed an increased in myelin Ag uptake by CNS APCs compared with PBS-treated animals (>26% increase in CD11b⁺ cells and >18% in CD11c⁺ cells; Fig. 5 B and C). Additionally, the amount of extracellular myelin in NAC-treated animals was reduced to levels similar to those found in WT mice (Fig. 5D).

Fig. 5.

Decreased myelin debris and increased number of myelin Ag⁺ CNS myeloid cells in NAC-treated animals with EAE. (A) Quantification of total free MBP in C57BL/6 WT, anti–IFN-γ, and IFN-γR^−/− mice treated with PBS or antioxidant NAC. Data are shown as the mean of the total free MBP volume [mean total volume of red channel (in cubic micrometers) containing voxels for red signal], ± SD. Asterisks indicate a significant change (*P ≤ 0.05; **P ≤ 0.01; Student’s t test). (B) Internal slices from original z-stack images showing myelin Ag (red) and CNS APCs (green) in lesions during EAE. Colocalization of myelin Ag with APCs (yellow, large arrows) is seen in WT mice treated with PBS, whereas a large amount of extracellular myelin debris (red, small arrows) not colocalized to APCs is found in mice with disrupted IFN-γ signaling. (Scale bars, 50 µm.) (C) Quantification of the number of MBP⁺CD11b⁺ microglia/macrophages (Upper Row) and CD11c⁺ DCs (Lower Row) in PBS- and NAC-treated animals from reconstructed confocal laser-scanning microscopy z-stacks. Data are shown as the mean number of cells per brain tissue slice, ± SD. Asterisks indicate a significant change in the percentage of MBP⁺ APCs in NAC-treated compared with PBS-treated mice (**P ≤ 0.01; Student’s t test). (D) Quantification of the total volume of intracellular MBP within APCs or extracellular MBP in CNS lesions of PBS- and NAC-treated EAE mice. Data are shown as the mean of the total free MBP ± SD. Asterisks indicate a significant decrease in total free MBP in NAC-treated mice compared with the PBS-treated control (*P ≤ 0.05; Student’s t test). In A–D, n ≥ 9 mice per group over three independent experiments.

Taken together, the results show that lipid peroxidation was strongly increased in the CNS during EAE when IFN-γ signaling was disrupted. Antioxidant treatment reversed lipid peroxidation and strongly ameliorated EAE, as is consistent with an important role for IFN-γ in limiting CNS demyelinating disease via this mechanism.

Discussion

In this study, we show that IFN-γ is a critical mediator of the uptake of myelin Ag by CNS APCs during EAE. Using mice deficient for IFN-γ or IFN-γR or treating mice with neutralizing IFN-γ mAb, we found that the number of APCs containing myelin Ag in the CNS of mice with EAE was decreased dramatically in the absence of IFN-γ signaling, despite significantly more severe clinical disease, suggesting a previously unrecognized role for IFN-γ in controlling the disease process via modulating myelin Ag scavenging in the CNS. The observation that clinical EAE was more severe in the absence of IFN-γ signaling is consistent with previous studies by Dalton and colleagues (4, 5) and has been ascribed to a proapoptotic effect of this cytokine for T cells. However, the finding that myelin Ag uptake by CNS APCs was decreased dramatically in the absence of IFN-γ signaling points to a divergence between IFN-γ effects on different myeloid lineages in the CNS in terms of myelin Ag phagocytosis, Ag presentation to encephalitogenic T cells, and regulation of disease severity. Our results suggest that disease severity is determined not only by Ag presentation in the CNS and T-cell activation but also by additional factors, such as the accumulation of neurotoxic myelin debris in the CNS microenvironment.

Our studies indicate that the accumulation of myelin debris may provide a substrate for ROS released during CNS inflammation, resulting in lipid peroxidation, which IFN-γ signaling may naturally serve to limit by promoting phagocytosis of free myelin. Lipid peroxidation is a self-propagating chain-reaction (24, 25), and the initial oxidation of only a few lipid molecules can result in significant tissue damage. It is well known that ROS can oxidize the lipid tails of the phospholipids present in the membrane bilayer of myelinating oligodendrocytes, introducing a break and producing an aldehyde or a straight-chain alkane. The MBP-induced fusion-formed myelin sheath then becomes unstable in the presence of even small amounts of these products of lipid degradation (25). A number of studies have shown that increased lipid peroxidation occurs in patients with MS (26, 27), although this view has been challenged (28, 29). More recently, studies in EAE have shown evidence of lipid peroxidation processes in mice that can be reduced with antioxidant treatment (30, 31). However, the mechanisms that regulate and curtail lipid peroxidation have largely remained unresolved.

IFN-γ is known to prime mononuclear phagocytes for enhanced production of ROS (32) and induces a number of enzymes, including the kynurenine pathway. Some metabolites, such as quinolinic acid, participate in the lipid peroxidation process and produce free radicals (33). However, specifically in macrophages and microglia, IFN-γ additionally induces the production of the metabolite 3-hydroxyanthranilic acid (3-HAA), a powerful antioxidant (34). Because 3-HAA is a robust free-radical scavenger (17), it was previously proposed that induction of this pathway by IFN-γ may represent a localized, extracellular antioxidant defense to protect host tissue from inadvertent oxidative damage that could occur during inflammatory conditions. This concept is in agreement with our studies, in which we found that the absence of IFN-γ signaling during EAE resulted in significantly enhanced lipid peroxidation products, as compared with WT mice, and corresponded to a more severe disease course that proceeded to a chronic inflammatory state. Importantly, exogenous antioxidant treatment with NAC dramatically reduced the amount of lipid peroxidation products and free myelin debris found at lesions and reduced disease severity in mice with disrupted IFN-γ signaling to below WT levels. The results support the view that IFN-γ has a key protective role in EAE, particularly during the progressive phase of disease, via self-resolution of the disease by self-limiting ROS production, with the net effect being downregulating the lipid peroxidation of myelin. Additionally, our studies advance this concept by suggesting a central role for IFN-γ–driven myelin Ag phagocytosis as a key protective mechanism in limiting lipid peroxidation. This concept is in agreement with MS studies showing that myeloid cells become antiinflammatory once they have ingested myelin (35).

Oxidized lipids and DNA are highly enriched in MS plaques (36) and have been associated with oxidative injury of oligodendrocytes and active demyelination as well as axonal or neuronal injury. It is noteworthy that antioxidant therapy in patients with MS has shown benefits in some studies but not in others, with one study showing no effect in the secondary progressive MS subtype but another showing a positive effect on patients with milder disease (37, 38). Similarly, the role of antioxidant therapy in EAE has not been fully resolved, with some studies, including ours, showing benefit (30, 39) but others failing to show benefit (40). Interestingly, a metabolite of the kynurenine pathway seems to influence EAE beneficially (41), in agreement with IFN-γ’s role in the generation of this metabolite. The conflicting results on the efficacy of antioxidant therapy in MS/EAE may indicate that antioxidants must be administered early enough to limit destructive lipid peroxidation. Further studies are needed to determine if timing affects the efficacy of antioxidant administration in EAE mice, with respect to IFN-γ signaling.

Our data suggest that once a certain threshold number of CNS APCs containing (and presenting) myelin Ag is reached, sufficient activation of encephalitogenic T cells is achieved to promote the disease. We cannot tell conclusively from our studies whether a small number of APCs containing large amounts of myelin Ag are required to generate a threshold number of myelin Ag:MHC II complexes per cell for T-cell activation, or whether it is sufficient for myelin Ag:MHC II complexes to reach a cumulative threshold when spread out over a larger number of APCs. In either case, IFN-γ–mediated signaling, uptake of myelin Ag by CNS APCs, free extracellular myelin Ag, and EAE severity were strongly correlated. Clearly, APCs, and in particular the presentation of myelin peptide by MHC class II molecules expressed by APCs, are critically required to induce EAE. In mice unable to express MHC class II molecules (42), EAE is abolished. Furthermore, DCs seem to be more critical than microglia for the induction of EAE (15, 43). Several other cell types with potential functions in Ag presentation, such as astrocytes, neutrophils, or mast cells, have been implicated in EAE pathogenesis. However, based on our data and supported by evidence from the literature, it seems unlikely that Ag presentation is the key contribution of these cells to the disease process. We have not tested B cells in our studies, because they play a marginal role in the induction and propagation of MOG peptide-induced EAE (44). Instead, our results are consistent with an alternative role for CNS APCs during the disease process. Enhanced EAE severity, despite decreased Ag colocalization to CNS APCs, suggests these cells have an important role in the clearance of myelin Ag released from damaged and/or dying oligodendrocytes. The failure to clear free myelin debris could inhibit the differentiation of oligodendrocyte precursors and could be toxic to neurons (20, 21). Furthermore, myelin debris may contain inhibitory molecules that antagonize axonal regeneration (45), and the removal of axonal debris by microglia increases axonal regeneration (46, 47). Additionally, microglial phagocytosis of myelin debris can promote resolution of the CNS inflammatory response (16, 48), and myelin-phagocytosing microglia express genes involved in the activation, migration, proliferation, and differentiation of oligodendrocyte precursor cells (49). Recently, it has been shown that if CNS-resident microglia cannot clear degenerated myelin from affected axons, remyelination is impaired, whereas infiltrating macrophages did not contribute to this function (50).

We observed low cellular expression of TREM2 during EAE in all APC subsets, with only microglia being significantly affected by the disruption of IFN-γ signaling. Loss of TREM2 function impairs microglial phagocytosis (51, 52), and enhanced TREM2-dependent phagocytosis of degenerated myelin can ameliorate EAE (51). TREM2 is known to interact with the signaling adapter protein DAP12 (DNAX-activation protein of 12 kDa) (53); however its ligand in apoptotic cells and other cellular debris is unknown. Similarly we noted that CD68 expression was decreased in both microglia and macrophages but was not significantly affected in DCs. CD68 is a member of the scavenger receptor family and is thought to promote phagocytosis and clear cellular debris, e.g., by the uptake of oxidized LDL (54). Therefore, a lower expression of this molecule specifically in these APC populations would be consistent with a reduced ability to ingest myelin-derived lipids. Mononuclear phagocytes from CD68^−/− mice have been shown to exhibit a trend toward enhanced antigen presentation to CD4⁺ T cells (55), similar to the trend we observed in APCs in IFN-γ–signaling-deficient mice during EAE. Additionally, GM-CSF has been shown to cause monocytes to increase ROS production and decrease phagocytosis (56). We noted a mild increase in the frequencies of GM-CSF–secreting T cells in the absence of IFN-γ signaling, along with increased lipid peroxidation and decreased phagocytosis, which may contribute to the phenotypic and functional changes observed in APCs in our model. Last, in mice with disrupted IFN-γ signaling the phagocytosis inhibitory receptor CD172a/Sirpα was increased in microglia, less so in macrophages, and was decreased in DCs. Ligation of macrophage CD172a by CD47 expressed on interacting cells prevents macrophage phagocytosis, whereas impaired CD172a engagement by CD47 promotes macrophage engulfment (57–59). Along these lines, it has been shown recently that CD47 is expressed by intact myelin and myelin-forming oligodendrocytes and may protect them from phagocytosis by microglia and macrophages via interaction with CD172a. This mechanism might be detrimental when clearance of myelin debris is desired (45). Thus, the increased expression of CD172a on microglia, and to a lesser extent on macrophages, may contribute to the lack of myelin Ag uptake observed in IFN-γ–deficient EAE mice in our studies. Importantly, CD172a expression by DCs was decreased significantly in IFN-γR^−/− mice, suggesting that DCs have a continued ability to ingest myelin debris despite the lack of IFN-γ signaling. Additionally, the continued high expression of MHC class II and costimulatory CD86 molecules on DCs, but not on microglia and macrophages, suggests that these DCs may remain capable of presenting myelin Ag, even if devoid of IFN-γ signaling. Interestingly, it has been shown that oxidized lipids induce a mature phenotypic signature in DCs with up-regulation of CD86, whereas at the same time phagocytic activity is inhibited (60). Similarly, oxidized lipids have been shown to promote the differentiation of a macrophage cell line into a DC phenotype in vitro (61), as is consistent with our data showing an increase in the number of DCs in IFN-γR^−/− mice and WT mice treated with anti–IFN-γ mAb. Thus, increased MHC class II and costimulatory CD86 expression and the increased number of DCs may act in concert to promote proinflammatory T-cell responses in the CNS in the absence of IFN-γ signaling. This effect may be amplified by the impaired ability of microglia to phagocytose free myelin. Along these lines, it is well known that microglia must be preactivated or primed by inflammatory challenge to become efficient phagocytes (62, 63), and our studies suggest an important signal may be IFN-γ in the context of CNS inflammation such as EAE. Furthermore, myelin-phagocytosing macrophages have been shown to inhibit T-cell proliferation in culture (64).

Thus, our results expand the role of IFN-γ from a T-cell–centric proinflammatory role to include an additional important protective role via regulating the function of myelin Ag uptake by CNS-resident microglia as well as CNS-infiltrating macrophages and DCs. A protective role for IFN-γ in EAE is fully compatible with literature reports suggesting that this cytokine has a protective role in EAE via inhibition of T-cell proliferation and induction of apoptosis (3, 4). However, our studies also suggest a previously unidentified, CNS APC-based, mechanism via clearance of free myelin debris and reduction of lipid peroxidation for the protection conferred by IFN-γ in EAE, and this mechanism may be important for the chronic progression of the disease observed in human MS. Although previously it was demonstrated that the administration of exogenous IFN-γ was unable to provide CNS protection in MS (2), it is possible that peripherally administrated IFN-γ has limited CNS bioavailability (65) and therefore was unable to exert protection under these conditions.

Along these lines, we propose that in CNS inflammation IFN-γ fulfills several roles with functionally opposing outcomes. Initially, IFN-γ contributes to the activation of APCs and up-regulation of MHC II expression and costimulatory molecules on these cells. This function is important for protection of the CNS from microbial pathogens; however, this mechanism promotes pathology and is detrimental in CNS autoimmune diseases. Secondarily, and probably later during the disease course, IFN-γ may provide protection by limiting oxidative stress-mediated damage to CNS tissues and may promote repair of the damaged tissues by removing free myelin Ag released by damage to the myelin sheath and oligodendrocytes. Free myelin Ag is detrimental because of its neurotoxic effects and its inhibition of oligodendrocyte precursor differentiation, and it is likely that it may act as a substrate for lipid and protein peroxidation. Our results in MBP^+/− shiverer mice showing decreased EAE under conditions of limited free myelin debris are consistent with this view and are in line with relative EAE resistance reported for mice deficient in other myelin Ags (66). Moreover, when levels of MOG protein were not decreased in mice heterozygous for the mog gene, these animals remained fully susceptible to EAE (67). Myelin lipid peroxidation may be targetable by antioxidant treatment and in turn may limit myelin debris accumulation, but the timing (early treatment) would be critical; the requirement for early treatment could explain the mixed results with antioxidant treatment in patients with MS.

In conclusion, in this study we provide evidence that IFN-γ provides an APC-centered mechanism of protection by promoting the removal of free myelin debris. Furthermore, our data suggest a previously unrecognized function for IFN-γ in promoting CNS protection by limiting the damage caused by lipid peroxidation.

Materials and Methods

Mouse Models.

Six- to eight-week-old C57BL/6, IFN-γ^−/−, SJL, and 2D2 tg mice were obtained from The Jackson Laboratory. IFN-γR^−/− mice were provided as a generous gift from Bernard Arulanandam, The University of Texas at San Antonio, San Antonio, TX. The HLA-DR2b (DRB1*1501) tg WT (MBP^+/+) and heterozygous shiverer (MBP^+/−) mice were bred in-house. All animals were maintained in pathogen-free conditions in the American Association for Laboratory Animal Science facility at the University of Texas at San Antonio. All experiments were approved by the University of Texas, San Antonio Institutional Animal Care and Use Committee and were performed in accordance with the relevant guidelines and regulations. Mice were fed and watered ad libitum. The induction of EAE (active and passive) and the injection of mice with anti–IFN-γ mAb or NAC are described in SI Materials and Methods.

Isolation of Cellular Subpopulations, Analysis of Cell Death, and Analysis of Myelin Ag-Containing CNS APCs.

Procedures for magnetic or flow cytometry isolation of specific cellular subsets, detection of apoptotic cells, and quantification analysis of myelin Ag-containing cells are described in SI Materials and Methods.

SI Materials and Methods

Mouse in Vivo Procedures.

To induce EAE, mice were immunized s.c. with 200 μg MOG_35–55 peptide (United Biochemical Research) emulsified 1:1 in complete Freund’s adjuvant (CFA) containing 5 mg/mL heat-killed Mycobacterium tuberculosis H37Ra (DIFCO). Pertussis toxin (400 ng/mL) was given i.p. on day 0 and day 2 p.i. Clinical scores were monitored daily and assigned a score based on the following symptoms: 0, no clinical disease; 1, flaccid tail; 2, partial hind limb paralysis; 3, total hind limb paralysis; 4, front and hind limb paralysis; 5, moribund or dead. For adoptive transfer, donor SJL mice (Jackson Labs) were immunized s.c. with 100 μg PLP_131–151 peptide emulsified in CFA. On day 10, inguinal and popliteal lymph nodes and spleens were obtained, and single-cell suspensions were made. Cells were adjusted to a concentration of 1 × 10⁷ cells/mL in DMEM containing 10% (vol/vol) FCS and 1% l-glutamine, were placed into a 24-well culture plate with 10 μg/mL PLP_131–151 antigen and 20 ng/mL recombinant IL-23 (rIL-23, eBioscience), and were transferred into a 37 °C incubator with 5% CO₂. On day 4, cells were removed from culture, washed two times, and resuspended in serum-free medium. Approximately 2 × 10⁷ cells were injected i.p. into each recipient mouse.

Anti–IFN-γ treatment.

Mice were injected i.p. with 500 μg of rat anti-mouse IFN-γ mAb in 200 μL of PBS (anti–IFN-γ mAb; R4-6A2; Bio X Cell) or 200 μL of PBS as control starting on day 0 or day 8 p.i. and then twice a week for the rest of the study. For adoptive transfer experiments only recipient mice were treated with anti–IFN-γ. For the lipid peroxidation study, 500 μg rat IgG2a isotype control mAb (Bio X Cell) in 200 μL of PBS was used as control. PBS was used as a control for NAC administration as described below.

Antioxidant treatment of EAE mice.

NAC (Sigma) was injected i.p. at a concentration of 150 mg/kg in 100 μL of PBS per mouse daily beginning on day 8 p.i. and continuing throughout the EAE disease course, as indicated. Control animals received 100 μL of PBS i.p. daily.

Antibodies Used in This Work.

Antibodies to MBP, MOG, and PLP were purchased from Neuromics (MO20009, GT15141, and CH23008, respectively). Fluorochrome-conjugated antibodies to CD4, CD11b, CD11c, and MHC class II were purchased from eBioscience, Inc. (53-0041, 53-0112, 53-2051, and 11-5322, respectively).

Isolation of Cellular Subpopulations.

Flow cytometry and cell sorting.

Spleen, lymph nodes, brain, and/or spinal cord tissues were removed from naive or EAE mice following cardiac perfusion as previously described (18). Single-cell suspensions were obtained from tissues by mechanical isolation. Extracellular myelin was removed from all brain and spinal cord suspensions using myelin removal beads (Miltenyi Biotec), according to the manufacturer’s instructions. For surface staining, cell suspensions were blocked with 2% mouse serum for 20 min on ice and then were stained with directly labeled antibodies for 45 min. Following staining, cells were washed with a 10× volume of PBS and fixed using fixative agent (eBioscience). For intracellular staining, fixed cells then were permeabilized in 1× permeabilization buffer (eBioscience) for 15 min and were incubated with directly labeled antibodies for 45 min. Cells then were washed with 10× volume of 1× permeabilization buffer and analyzed or sorted on a BD FACSAria II (BD Bioscience). Data were analyzed with FACSDiva software (BD Bioscience).

In vitro T-cell and APC coculture.

Spleen was procured from 6- to 8-wk-old naive WT or IFN-γR^−/− mice, and inguinal and popliteal lymph nodes were procured from 6- to 8-wk-old 2D2 TCR tg mice (Jackson Labs). Single-cell suspensions were obtained from procured tissues by mechanical disruption. CD4⁺ T cells were sorted as described above from single-cell suspensions prepared from inguinal and popliteal lymph nodes of 2D2 mice and cocultured overnight with WT or IFN-γR^−/− splenocytes as APCs and 10 µg/mL MOG_35–55 peptide and 100 ng/mL LPS. Cells then were stained and analyzed by flow cytometry as described above. In parallel, 2D2 cells were tested by cytokine ELISPOT assay for the frequencies of cytokine-producing cells as described.

Detection of CD4⁺ T-Cell Apoptosis.

The number of CD4⁺ T cells was quantified using immunofluorescence staining with antibodies against caspase-3 (Abcam) and CD4 as described above to detect caspase-3⁺CD4⁺ cells. Additionally, we used the ApopTag Red in Situ Apoptosis Detection Kit (Millipore), which is based on the TUNEL assay and detects DNA fragmentation, on brain tissue cryosections obtained from WT and IFN-γR^−/− EAE mice. Briefly, following fixation and permeabilization of tissues as described above, we added the equilibration buffer for 5 min and then incubated tissues in the terminal deoxynucleotidyl transferase enzyme for 1 h at 37 °C. Stop/Wash buffer was added for 10 min; then slides were washed in PBS. Then tissue sections were incubated in anti-digoxigenin conjugated with rhodamine for 30 min, washed in PBS, and counterstained with directly labeled anti-CD4 as described above.

Detection of Lipid Peroxidation Activity.

Lipid peroxidation was measured fluorometrically by the estimation of malondialdehyde (MDA) concentration based on the reaction with thiobarbituric acid (TBA), as previously described (68), compared with water control using an OxiSelect TBARS Assay Kit (Cell Biolabs). Briefly, spinal cords were removed from WT and IFN-γR^−/− EAE mice at the chronic stage of disease (day 30 p.i.) following cardiac perfusion. Spinal cord tissue was resuspended in 0.05% butylated hydroxytoluene to prevent further oxidation from occurring and was homogenized by sonication on ice. Supernatants from tissue homogenates were reacted with TBA at 95 °C for 45 min, and the fluorometric shift was read on a Synergy HT Multimode Microplate Reader (BioTek) at 540-nm excitation and 590-nm emission. Relative MDA content then was determined by comparison with the predetermined MDA standard curve. Additionally, the lipid peroxidation product 4-HNE was detected using indirect immunofluorescence following a procedure similar to that described above. Ab against 4-HNE (HNEJ-2) was purchased from Abcam.

Immunofluorescence Staining.

DAPI for immunofluorescence staining was purchased from Sigma Aldrich. For immunofluorescent staining, murine brains and spinal cords were obtained at the designated time points, placed in OCT freezing compound, and frozen at −80 °C. Tissue was cut into 4- to 20-μm-thick sections and was staggered onto glass slides. Alternatively, tissue was cut into 20- to 60-μm-thick sections, placed into fixative agent (eBioscience) in 24-well untreated tissue culture plates (Fisher), and then placed on slides following the staining procedure. Sections were fixed with an eBioscience fixative agent on ice and then were permeabilized with an eBioscience permeabilization buffer and blocked with appropriate serum before staining. Purified primary antibodies were allowed to incubate on sections overnight at 4 °C in a closed, humid chamber. Secondary antibodies and directly labeled primary antibodies then were added for 1 h at RT in a closed, humid chamber. Slides were rinsed between incubations with PBS, PBS + 0.05% Tween-20 (PBST), or PBS + 0.01% Triton-X, depending on the Ag of interest. Coverslips were mounted with Fluoromount-G (Southern Biotec), and slides were allowed to dry and cure overnight at RT in the dark.

Quantification of Myelin Ag-Containing CNS APCs.

Image deconvolution was performed using AutoQuant X version 2.2.2 software, and subsequent image analysis was performed using Imaris 3D/4D software version 7.2 from Bitplane as previously described (18). The 3D structure of APCs in the CNS was reconstructed using Imaris software from confocal laser-scanning z-stack images taken 0.2 μm apart from a depth of 1.2–59.6 μm, depending on the thickness of the CNS tissue being imaged, as previously described (18). Quantification of the absolute number of APCs was performed as previously described (18). Briefly, the number of nuclei present in each 3D-stacked image, which could be expanded radially to colocalize with APC cell-surface markers within 5–10 μm in >75% of all possible directions, were counted as APCs using Imaris 7.2 software by Bitplane. Myelin Ag associated with APCs was distinguished from Ag in the extracellular microenvironment by digitally enclosing APCs using the reconstructed 3D surface and then determining the volume (in cubic millimeters) of Ag contained within (intracellular) or outside (extracellular) the modeled APC.

Cytokine ELISPOT Assay.

Cytokine ELISPOT assays were performed as described previously (18). Briefly, ELISPOT plates (Multiscreen IP; Millipore) were coated with IFN-γ–specific capture Ab (AN-18; eBioscience), IL-17–specific capture Ab (17F3; Bio X Cell), IL-5–specific capture Ab (TRFK5; eBioscience), or GM-CSF–specific capture Ab (MP1-22E9; BD Bioscience) diluted in PBS. The plates were blocked with 1% BSA in PBS for 1 h at RT and then were washed four times with PBS. Cells were added with or without Ag and incubated for 24 h at 37 °C. The plates were washed three times with PBS and four times with PBST, and IFN-γ–specific biotinylated detection Ab (R4-6A2; eBioscience), IL-17–specific biotinylated detection Ab (TC11-8H4; BioLegend), IL-5–specific biotinylated detection Ab (TRFK4; eBioscience), or GM-CSF–specific biotinylated detection Ab (MP1-31G6; BD Bioscience) was added and allowed to incubate overnight. The plate was washed four times with PBST and incubated with streptavidin-alkaline phosphatase (Invitrogen). Cytokine spots were visualized by 5-bromo-4-chloro-3-indolyl phosphate/nitroblue tetrazolium phosphatase substrate (Kirkegaard & Perry Laboratories). Image analysis of ELISPOT assays was performed on a Series 2 Immunospot analyzer and software (Cellular Technology) as described previously (18). In brief, digitized images of individual wells of the ELISPOT plates were analyzed for cytokine spots, based on the comparison of experimental (containing T cells and APC with Ag) and control wells (T cells and APC, no Ag). After separation of spots that were touching or partially overdeveloped, nonspecific noise was gated out by applying spot size and circularity analysis as additional criteria. Spots that fell within the accepted criteria were highlighted and counted.

Statistical Analysis.

Statistical evaluations were performed using JMP SAS 9 software. Phenotypic marker expression, T-cell responses and activation, and comparison across groups of infiltrating and MBP-associated cells were evaluated using a Student’s t test. Cumulative EAE scores were calculated, as previously described (69), and statistical significance was calculated by Student’s t test. P values ≤ 0.05 (*) or ≤ 0.01 (**) were considered significant.