Contribution of Astrocyte-Derived IL-15 to CD8 T Cell Effector Functions in Multiple Sclerosis

Multiple sclerosis (MS) is an inflammatory disease of the CNS, pathologically characterized by activated glial cells, demyelination, and axonal loss (1). Immune cells, including macrophages and T lymphocytes, are detected within MS lesions (2). CD4 T lymphocytes are traditionally considered the main effectors in the pathogenesis of MS, but CD8 T lymphocytes have been suggested to play an important role in the tissue damage (3, 4). These cells are present in MS lesions in similar or greater amounts than their CD4 counterparts (5–9). CD8 T lymphocytes with polarization of their cytolytic granules are observed in close proximity to oligodendrocytes and demyelinated axons (6). These lymphocytes bear an activated effector memory phenotype, are oligoclonally expanded, and persist over time in the CNS of patients with MS (6, 9–12). Finally, animal models have illustrated the capacity of CD8 T lymphocytes to mediate MS-like diseases (13–17).

IL-15 is essential for the development, activation, and survival of CD8 T lymphocytes and NK cells (18, 19). Monocytes/macrophages and dendritic cells have been reported as the main source of this cytokine (20, 21). Intracellular IL-15 binds to the high-affinity IL-15Rα; this complex is subsequently transported to the cell surface for an efficient cross-presentation to other cells. The surface IL-15–IL-15Rα complex stimulates cells bearing the IL-2/IL15Rβ and common γ chain (22–24). Increased local expression of IL-15 has been suggested to contribute to the immunopathology of several human inflammatory diseases including rheumatoid arthritis, celiac disease, and inflammatory bowel disease (25–28).

The precise contribution of IL-15 to MS immunopathogenesis has not been elucidated, although elevated IL-15 in serum and in PBMCs at both the mRNA and protein levels in patients with MS has been reported (29–32). IL-15 mRNA has been detected in murine and human CNS tissue and in primary cultures of astrocytes and microglia (33–36). However, IL-15 protein levels do not correlate with mRNA detection, and cell membrane rather than secreted IL-15 is crucial in mediating in vivo effects (22, 37). Thus, it is essential to document functional surface IL-15 protein expression in human CNS cells. Although the IL-15 protein has been detected on astrocytes in inflammatory mouse models (36, 38), whether these nonmyeloid cells represent an important and functional IL-15 source in human CNS diseases specifically for CD8 T cells has not been addressed.

Astrocytes are implicated as contributors to both innate and adaptive immune responses taking place in the CNS (39). Our study shows that conditions mimicking CNS inflammation as observed in MS result in a robust increase in surface-bound IL-15 on human astrocytes in vitro. Moreover, this surface expression boosted the cytotoxicity of human CD8 T lymphocytes. We also demonstrate that IL-15 expression is upregulated on astrocytes within MS lesions. Taken together, our results demonstrate that astrocytes, via the production of IL-15, participate in enhancing effector responses of CNS-infiltrating CD8 T lymphocytes and thus are potential important contributors to tissue injury in MS.

Human fetal CNS tissue was obtained from Albert Einstein College of Medicine (Bronx, NY). Their ethical review boards and McGill University’s approved the studies. Astrocytes and microglia were isolated as previously described (40) and grown in DMEM containing 10% (v/v) FBS. Adult microglia were isolated from surgical resections performed for the treatment of non–tumor-related intractable epilepsy, in accordance with guidelines of the ethical board of McGill University and as previously described (41). Cells were treated for 24 h with IL-1β (10 ng/ml; Medicorp, Montreal, Quebec, Canada), IFN-γ (200 U/ml = 3.6 ng/ml; Biosource International, Camarillo, CA), TNF (2000 U/ml = 20 ng/ml; Biosource International), combinations of those cytokines, or IFN-γ followed by LPS (100 ng/ml; Sigma-Aldrich, Oakville, Ontario, Canada).

Informed consent was obtained from healthy donors for these studies that were approved by McGill University and Centre Hospitalier de l’Université de Montréal ethical boards (Montreal, Quebec, Canada). PBMCs isolated by Ficoll density gradient were plated in RPMI 1640 containing 10% (v/v) FBS, and floating cells were removed by washing after 2 h, leaving adherent monocytes.

CD8 T cell lines specific for influenza matrix protein 1 aa 58–66 (influenza: GILGFVFTL) or myelin basic protein (MBP) aa 111–119 (MBP: SLSRFSWGA) were expanded as previously published (42). CD8 T cell lines were stimulated with autologous Ag-loaded B cells once a week, and IL-15 (10 ng/ml; PeproTech, Rocky Hill, NJ) was added both concurrently with fresh peptide loaded APCs and 3d later. CD8 T cell lines were considered specific when the amount of IFN-γ secreted after 24–36 h (ELISA from BD Biosciences, Mississauga, Ontario, Canada) in the presence of specific peptide-loaded autologous B cells was at least twice the amount secreted in the presence of irrelevant peptide-loaded B cells (Table I).

Astrocytes, microglia, and monocytes were detached and stained for surface and intracellular molecules as previously described (40) and subsequently acquired on an FACSCalibur or LSR II (BD Biosciences). Mouse mAbs directed at human protein either unconjugated or conjugated to FITC, Alexa Fluor 488, Pacific Blue, PE, PE-Cy5, or allophycocyanin were used. Surface staining targeted: IL-15, IL-15Rα (R&D Systems, Minneapolis, MN), MHC class I (MHC-I; HLA-A/B/C), CD14, and CD8 (BD Biosciences). Intracellular staining targeted: perforin, granzyme B (BD Biosciences), or glial fibrillary acidic protein (GFAP) (Molecular Probes, Invitrogen Life Technologies, Burlington, Ontario, Canada). In some experiments, dead cells were excluded using 7-aminoactinomycin D (BD Biosciences) according to the manufacturer’s instructions. Appropriate isotype controls were used in all steps. Flow cytometry analyses were performed using FlowJo (Tree Star, Ashland, OR), and the change in median fluorescence intensity (ΔMFI) was calculated by subtracting the fluorescence of the isotype from that of the stain.

Astrocytes were plated at 1 × 10⁵ cells in 500 μl/well in a 48-well plate and allowed to grow up to ∼70% confluence prior to being activated with cytokines (IL-1β + IFN-γ) for 24 h and finally treated with monensin (5 nM) for 1 h. After two washes, blocking anti–IL-15 Ab (20 μg/ml) (clone 34559, R&D Systems) or an isotype control was added for 1 h. The blocking anti–IL-15 mAb (clone 34559) used in our experiments is a well-characterized blocking reagent, as demonstrated by several groups (20, 43–45). Subsequently, 5 × 10⁴ CD8 T lymphocytes were added per well, and the coculture was maintained for 48 h in a 500-μl volume, after which cells were harvested, then stained and analyzed by flow cytometry as described above. For some experiments, we compared two commercially available IL-15–blocking Abs (clones 34559 and 34505, R&D Systems) and obtained similar results. To test the cytotoxic activity of CD8 T lymphocytes cocultured with astrocytes, autologous B cells previously loaded with specific or control Ag were added at a 1:1 ratio concurrently with anti-CD107a PE-conjugated Ab (BD Biosciences) and monensin A (5 nM) to prevent rapid internalization of CD107a by endocytosis, as used by others (46, 47). Five hours later, cells were harvested and stained for CD8 and analyzed by flow cytometry.

Total RNA was isolated using TRIzol (Invitrogen Life Technologies) and the Qiagen RNeasy mini kit (Qiagen, Mississauga, Ontario, Canada) according to manufacturer’s instructions and as previously described (40, 41). RNA samples were treated with DNase (Qiagen) and then transcribed into cDNA using random hexaprimers (Roche, Mississauga, Ontario, Canada) with the Moloney murine leukemia virus-RT enzyme (Invitrogen Life Technologies) at 42°C. IL-15 expression level was determined by quantitative real-time PCR (qPCR) using the ABI Prism 7700 Sequence Detection System (Applied Biosystems, Foster City, CA). Primers and TaqMan FAM-labeled MGB probes for IL-15 were obtained from Applied Biosystems (TaqMan Gene Expression Assays). β-Actin primers and cycling parameters have been previously published (41). qPCR cycling was performed according to the ABI Prism 7000 Sequence Detection System (Applied Biosystems) default temperature settings (2 min at 50°C and 10 min at 95°C, followed by 40 cycles of 15 s at 95°C and 1 min at 60°C) in a volume of 25 μl with 1× TaqMan Universal Master Mix (Applied Biosystems). For relative mRNA expression, IL-15 amplification was normalized to endogenous control and cDNA from cells found to have high IL-15 expression.

Tissue sections of postmortem brains from donors without CNS disease and patients diagnosed clinically and confirmed by neuropathological examination as having MS were obtained from the NeuroResource Tissue Bank, University College London Institute of Neurology, London, U.K. CNS tissues were donated to the tissue bank with informed consent following ethical review by the London Research Ethics Committee, London, U.K. This study was approved by the Centre Hospitalier de l’Université de Montreal Ethical Committee. Snap-frozen coded sections (∼1-cm² and 10-μm thick) were cut from blocks of normal control and MS brain tissues. Sections cut before and immediately after the ones used for the immunofluorescence studies were stained with Oil Red O (ORO) and hematoxylin and scored as previously described (48) (Table II). Serial sections were air-dried and fixed in cold acetone for 10 min and then blocked for nonspecific binding for 1 h in HBSS containing 2% (v/v) horse serum, 2% (v/v) FBS, 1 mM HEPES buffer, and 0.1% (w/v) sodium azide, and then blocked for endogenous biotin or biotin-binding proteins using the Blocking Kit from Vector Laboratories (Burlington, Ontario, Canada) following the manufacturer’s instructions. All Abs were diluted in HBSS containing 2% (v/v) horse serum, 2% (v/v) FBS, 1 mM HEPES buffer, and 0.1% (w/v) sodium azide, and washes were done in PBS. Staining for IL-15 was performed overnight at 4°C using monoclonal mouse anti-human IL-15 Ab (25 μg/ml; R&D Systems clone 34559). Sections were subsequently incubated for 2 h with biotinylated polyclonal goat anti-mouse Abs (DakoCytomation, Mississauga, Ontario, Canada), then incubated for 1.5 h with Cy3-conjugated streptavidin (Jackson ImmunoResearch Laboratories, West Grove, PA) and Alexa Fluor 488-conjugated anti-GFAP mouse mAb (10 μg/ml; Molecular Probes), FITC-conjugated anti-CD68 mouse mAb (2 μg/ml; Biolegend, San Diego, CA), rabbit anti–Iba-I Ab (5 μg/ml; Wako Chemicals USA, Richmond, VA), or allophycocyanin-conjugated anti-CD8 (BD Biosciences) mouse mAb for at least 2 h at room temperature. Iba-I staining was followed by Alexa Fluor 488-conjugated goat anti-rabbit Abs (Invitrogen Life Technologies), whereas CD8 staining was followed by rabbit anti-allophycocyanin Abs and FITC-conjugated donkey anti-rabbit Abs for 2 h each. Sections were incubated for 30 min with RNase (100 μg/ml; Sigma-Aldrich) and then TO-PRO-3 iodide (3.3 μM; Invitrogen Life Technologies) for 10 min. To quench tissue autofluorescence, sections were treated with 1% (w/v) Sudan black (Sigma-Aldrich) in 70% ethanol for 3 min and washed thoroughly. Control stainings were concurrently carried out on adjacent sections using appropriate primary isotype controls (for IL-15, mouse IgG₁ [Sigma-Aldrich]; for GFAP, Alexa Fluor 488-conjugated mouse IgG₁ [Caltag Laboratories, Burlingame, CA]; for CD68, FITC-conjugated mouse IgG_2a [BD Biosciences]; for CD8, allophycocyanin-conjugated mouse IgG₁ [BD Biosciences]; for Iba-1, rabbit Igs [Jackson ImmunoResearch Laboratories]) at the same concentrations. Slides were observed using a Leica confocal microscope SP5 (Leica Microsystems, Richmond Hill, Ontario, Canada). Confocal images were acquired simultaneously in different channels throughout a 4–8 μm z-stack every 0.2–0.5 μm. We validated staining specificity by lack of signal when the corresponding laser was turned off but with the others still on. Several fields (more than five) containing cells positive for the relevant specific cell marker were taken for each section and used to assess the proportion of cells expressing IL-15. Moreover, we confirmed the absence of bleed-through by re-examination of selected sections via sequential scanning.

Data handling and analysis (Student t test) were performed using Prism 3.0 (GraphPad, La Jolla, CA).

The surface-bound form of IL-15 is mainly responsible for cellular signaling to neighboring cells (22). Thus, to determine whether human CNS cells are a relevant source of this cytokine, we assessed the presence of surface IL-15 on human astrocytes. Highly enriched primary cultures of human astrocytes (>95% pure) (Fig. 1A) were either left alone or incubated for 24 h with IL-1β, IFN-γ, TNF, IFN-γ + TNF, or IFN-γ + IL-1β prior to being stained for IL-15 and IL-15Rα. A small proportion of astrocytes (1–15%) expressed surface IL-15 when left untreated or treated with individual cytokines IL-1β, IFN-γ, or TNF (Fig. 1B). However, treatment with the combinations of IFN-γ + TNF or IFN-γ + IL-1β both resulted in surface IL-15 expression in >55% of astrocytes (Fig. 1B). These observations were confirmed with astrocytes obtained from multiple donors (Fig. 1C; untreated [10.0 ± 2.9%] versus IFN-γ + TNF [55.6 ± 10.7%] or IFN-γ + IL-1β [56.4 ± 9.0%]; *p < 0.05), although the absolute percentage observed varied from donor to donor. All treatments that induced IL-15 were associated with significantly increased MHC-I expression on the astrocytes (Fig. 1D, untreated versus cytokine treatment; *p < 0.02). We also confirmed using qPCR that cytokine combinations IFN-γ + TNF and IFN-γ + IL-1β increased mRNA level coding for IL-15 in human astrocytes (Fig. 1E). A large proportion of astrocytes basally expressed IL-15Rα, and this percentage increased for some donors following cytokine treatments compared with untreated cells (Fig. 1B, 1F). We also performed ELISA to detect IL-15 in astrocyte-conditioned media, but secreted levels were below the minimum detection threshold of the assay (10 pg/ml). Moreover, when we added soluble rIL-15 to astrocytes, we could not detect surface binding of the cytokine after 5 h of incubation at 37°C. This suggests that astrocytes synthesized IL-15, which was then translocated and expressed on the cell surface bound to its IL-15Rα, similar to other cell types (22).

Previous studies (33–36) reported that microglia express IL-15 mRNA without evaluating whether this cytokine was present on the cell membrane. Thus, we evaluated whether human microglia also provide surface IL-15 and compared adult microglia to human peripheral monocytes, which derive from the same lineage. Adult microglia and monocytes were either untreated or stimulated for 24 h and then stained for surface IL-15, MHC-I, CD11c (for microglia), and CD14 (for monocytes) (Fig. 2A). We confirmed on numerous samples that microglia concurrently express both CD68 and CD11c (data not shown). A proportion of human adult microglia did express surface IL-15 (∼40%) (Fig. 2A) under basal culture conditions, but this expression did not increase posttreatment [IFN-γ + LPS, IFN-γ + IL-1β, IFN-γ + TNF (Fig. 2A and data not shown)] despite increased MHC-I expression, as we have previously shown (49). Surface IL-15 expression was nearly undetectable in untreated monocytes; however, following their activation with IFN-γ + LPS, a well-known combination that activates these myeloid cells, most cells expressed surface IL-15 as previously shown (20).

We also compared IL-15 expression by fetal astrocytes and microglia from the same donor following IFN-γ + TNF or IFN-γ + IL-1β exposure (Fig. 2B). A small percentage (9.7%) of human fetal microglia isolated from the same material as human fetal astrocytes expressed IL-15 at basal levels (Fig. 2B). The proportions of fetal microglia expressing surface IL-15 increased following IFN-γ + TNF or IFN-γ + IL-1β (33.2 or 19.2%) treatment (Fig. 2B, one donor representative of three) as did astrocytes from the same donor (10.3% untreated versus 31.4 and 36.3%). Thus, both human adult and fetal microglia can provide surface IL-15.

In an effort to model the CD8 T lymphocytes found in MS lesions as closely as possible, we evaluated the effects of IL-15 on Ag-specific CD8 T lymphocytes (3). Human Ag (MBP or influenza)-specific CD8 T cell lines (see typical examples in Table I) were characterized by flow cytometry according to their effector memory profile (loss of CCR7), expression of lytic enzymes (perforin and granzyme B), and expression of NKG2D, a coactivating receptor we have previously shown to be involved in oligodendrocyte killing (40). Ag-specific CD8 T lymphocytes were cocultured with either untreated astrocytes or astrocytes pretreated with IFN-γ + IL-1β (inducing a high proportion of surface IL-15–expressing cells, as per Fig. 1) for 48 h in the presence of either an Ab against IL-15 or an isotype control. Flow cytometric data from 1 representative CD8 T cell line are illustrated in Fig. 3A–C, and data from 10 CD8 T cell lines are summarized in Fig. 3D–F. The proportion of CD8 T lymphocytes expressing granzyme B and perforin was low in untreated CD8 T lymphocytes (Fig. 3A, 6%; 3D, average of 10 CD8 T cell lines: 3.6 ± 0.5%), but increased at least 3-fold in the presence of rIL-15 (Fig. 3A, 24%, top panels; 3D, average of 10 CD8 T cell lines: 22.8 ± 1.4%). CD8 T lymphocytes cultured on untreated astrocytes showed a modest increase in the proportion of cells expressing both perforin and granzyme B (Fig. 3A, alone 6% versus untreated astrocytes: 14%; 3D, average of 10 CD8 T cell lines: 3.6 ± 0.5 versus 12.7 ± 1.0%). This effect was most likely due to low basal expression of IL-15 by untreated astrocytes because it was efficiently blocked by anti–IL-15 Ab (Fig. 3A, 7%; 3D, average of 10 CD8 T cell lines: 5.4 ± 0.6%). The ΔMFI of granzyme B was also elevated following coculture with astrocytes compared with untreated CD8 T lymphocytes (Fig. 3E, data from 10 CD8 T cell lines: alone, 135 ± 25 versus untreated astrocytes, 496 ± 60). CD8 T lymphocytes cultured on treated astrocytes in the presence of the isotype control displayed a pronounced 5-fold increase in the proportion of cells expressing perforin and granzyme B (Fig. 3A, 30%; 3D, data for 10 CD8 T cell lines: alone, 3.6 ± 0.5% versus treated astrocytes plus isotype, 28.0 ± 1.7%) and in the level of granzyme B per cell (Fig. 3E, ΔMFI of granzyme B: alone, 135 ± 25 versus treated astrocytes plus isotype, 1171 ± 208; ***p < 0.001). Culture of CD8 T lymphocytes on treated astrocytes in the presence of anti–IL-15 Ab significantly reduced the percentage of CD8 T lymphocytes expressing perforin and granzyme B (Fig. 3A, 8%; Fig. 3D, average of 10 CD8 T cell lines 8.5 ± 0.8%; ***p < 0.001 compared with isotype). Thus, surface IL-15 provided by astrocytes significantly elevated the expression of lytic enzymes by human CD8 T lymphocytes.

CD8 T cells responding to IL-15 provided as a recombinant protein or by treated astrocytes were determined to be effector memory cells because cells that acquired granzyme B expression did not express CCR7 (Fig. 3B). We also observed that CD8 T cells cultured in the same conditions expressed elevated levels of NKG2D compared with CD8 T cells cultured alone (Fig. 3C, 3F, average 10 CD8 T cell lines ΔMFI of NKG2D: alone, 10.3 ± 1.1; treated astrocytes plus isotype control, 31.6 ± 1.9; treated astrocytes with anti–IL-15, 17.6 ± 1.0; isotype versus anti–IL-15 ***p < 0.001). These results indicate that IL-15 provided by astrocytes enhanced NKG2D levels on human CD8 T lymphocytes. Supernatants of activated astrocytes were not able to increase perforin, granzyme B, or NKG2D expression levels in CD8 T lymphocytes, suggesting that the observed effects by astrocytes were mediated by membrane-bound IL-15 (data not shown). Moreover, cocultures of astrocytes and CD8 T lymphocytes separated by a transwell did not reproduce these results, supporting the cell–cell contact dependence of our observations (data not shown). We obtained similar results with MBP- and influenza-specific CD8 T cell lines, suggesting that regardless of their Ag specificity, CD8 T lymphocytes infiltrating the CNS respond to IL-15 provided by astrocytes (Fig. 3). To confirm that astrocyte-provided IL-15 had an impact on ex vivo-purified CD8 T cells, IFN-γ + IL-1β–activated astrocytes were cocultured for 6 d with CFSE-labeled ex vivo-isolated heterologous CD8 T lymphocytes. The addition of Ab against IL-15 did significantly decrease the proportion of CD8 T cells expressing granzyme B compared with the isotype control (anti–IL-15: 12% versus isotype 21%; p < 0.05), without significantly affecting the proliferation of CD8 T cells (24 versus 21%). To rule out that only IL-15 pre-exposed T lymphocytes were sensitive to IL-15–mediated effects, cell lines cultured with IL-2 from the day of isolation were also incubated with rIL-15. The response of these cells was the same as cells cultured with IL-15 (data not shown).

To determine whether the increased content of lytic enzymes in CD8 T lymphocytes correlated with augmented cytotoxic activity, we used the well-characterized CD107a-mobilization flow cytometry-based assay (Fig. 3G). Lytic enzymes are stored in intracellular granules bearing CD107a. In the presence of appropriate target cells, these enzymes are released, resulting in the transient expression of CD107a on the surface of effector cells. The CD107a mobilization assay has been shown to correlate with degranulation and killing activity (as measured by [⁵¹Cr] release) of human CD8 T lymphocytes (46, 47, 50). To evaluate the consequences of the CD8 T cell–astrocyte interaction on the subsequent encounter of CD8 T cells with their specific target cells, we added autologous B cells loaded with either specific or control peptide in the astrocyte–CD8 T lymphocyte coculture. CD8 T lymphocytes cultured with rIL-15 showed a 3-fold greater proportion of CD107a surface-expressing cells in presence of target cells pulsed with the appropriate Ag in comparison with CD8 T lymphocytes left alone (12 versus 4%) (Fig. 3G). IL-15 provided by treated astrocytes also increased the proportion of CD8 T lymphocytes expressing CD107a reaching 10% with an isotype control; this proportion was significantly diminished when anti–IL-15–blocking Ab was included in the coculture (6%) (Fig. 3G). Importantly, degranulation of CD8 T lymphocytes was Ag-specific because the addition of control Ag-loaded target cells did not trigger any significant response even in the presence of an IL-15 source (1–4%) (Fig. 3G).

To assess whether astrocytes could provide IL-15 in vivo, especially within MS lesions, we performed immunohistochemistry on postmortem brain tissues obtained from patients with MS and normal controls (description of samples provided in Table II). Sections were stained with GFAP and IL-15–specific Abs or appropriate isotype controls. At least five distinct fields (at ×400, each field comprises 0.09 mm²) per section from three nonneurologic disease controls and six patients with MS containing GFAP-positive cells were thoroughly analyzed, and representative fields are illustrated (Fig. 4). All MS sections examined (n = 6 donors) displayed strong immunoreactivity for GFAP, especially within and at the edges of demyelinated lesions. A high proportion of astrocytes (>80–90% as assessed by quantification of 21 distinct fields from 6 MS lesions for >220 GFAP⁺ cells) expressed high levels of IL-15 (Fig. 4H, 4J, 4M, 4O, white arrows). IL-15 was detected in astrocytes in acute (Fig. 4, block 5) and subacute/chronic (Fig. 4, block 10) MS lesions. Colocalization of IL-15 and GFAP was confirmed by scanning the tissue at different z planes. We also detected strong IL-15 immunoreactivity in astrocytes near blood vessels with or without perivascular cuffs (data not shown). In all sections from controls, we also detected GFAP⁺ astrocytes. However, the majority of these cells (>80%) did not express detectable IL-15, with only a few showing weak expression (Fig. 4B, 4C, white arrow).

We also observed IL-15 expression in GFAP-negative cells in MS sections (Fig. 4H, 4M, orange arrowheads). To identify these cells, adjacent sections were stained for IL-15 and Iba-1 or CD68, two markers characteristic of macrophages/microglia. The number of Iba-1⁺ cells (Fig. 5A, 5E, 5I) was proportional to the ORO scoring, a well-characterized immunohistopathological staining that detects lipids in phagocytic macrophages (Table II); in MS lesions with an ORO score >3 (blocks 4–6 and 9), >250 macrophages/microglia per mm² were detected, whereas <100/mm² were observed in the other sections (blocks 8 and 10). Seventy-five percent of the 450 macrophages/microglia counted from a total of 24 fields of seven MS lesions (60–90% for individual sections) expressed IL-15 (Fig. 5G, 5K), although some myeloid cells had undetectable IL-15 (Fig. 5K, orange arrow). Similar results were obtained when CD68 was used to identify macrophages/microglia (data not shown). In these sections, we also observed numerous IL-15⁺ cells with an astrocyte-like morphology that were not macrophages/microglia (Fig. 5G, white arrowheads), consistent with our results presented above showing colocalization of IL-15 with GFAP-expressing cells. Fewer Iba-1⁺ cells were detected in control sections (Fig. 5) compared with MS tissues, and a proportion of them expressed IL-15 (34 of 53 microglia counted in a total of 15 fields from the 3 control sections (64%) (Fig. 5A–C). Thus, overall, the number of Iba-1⁺ cells expressing detectable IL-15 levels per surface area was strongly reduced in controls compared with MS tissues.

We investigated the presence of CD8 T lymphocytes in the vicinity of IL-15–expressing cells (Fig. 6) and noted whether these T cells were located in the CNS parenchyma or within perivascular cuffs. Although the number of CD8 T lymphocytes per surface area varied between donors, CD8 T cells were present in the CNS parenchyma of all the MS lesions examined (Fig. 6D) and in perivascular cuffs (Fig. 6G). Moreover, CD8 T cells were observed outside of blood vessels, only where cells displayed evident IL-15 expression (Fig. 6F, 6I). In control sections, no CD8 T lymphocytes could be observed in the CNS parenchyma, and only rare CD8 T lymphocytes were associated with blood vessels (Fig. 6A).

Our study demonstrates that human astrocytes in a proinflammatory milieu express sufficient surface IL-15 so as to enhance effector functions of Ag-specific CD8 T lymphocytes. Although IL-15 mRNA and protein expression in murine and human CNS cells has been reported (33–36), our results are the first, to our knowledge, to demonstrate that surface IL-15 expressed by human glial cells is functional and acts on human CD8 T lymphocytes. Our in situ data specifically document the prominent expression of IL-15 in MS lesions. In addition, CD8 T lymphocytes are localized to areas with abundant IL-15–expressing cells in MS lesions.

In our in vitro studies, treatment with single cytokines (IFN-γ, IL-1β, or TNF) did not alter IL-15 levels on astrocytes, although these treatments did increase MHC-I expression (Fig. 1D). We conclude that such treatments did not trigger either the appropriate signaling cascade or amount of intracellular signaling to induce surface IL-15 expression (51). However, IFN-γ + TNF and IFN-γ + IL-1β combinations significantly increased the proportion of astrocytes expressing surface IL-15, and these cytokines have been shown to induce other immune mediators in human primary astrocytes (52, 53). Activated T lymphocytes and macrophages/microglia, abundantly present in MS lesions (1), are classic in vivo sources of these cytokines (1, 54–57).

Our in vitro results demonstrate that activated human astrocytes augment the expression of multiple effector molecules (granzyme B, perforin, NKG2D) and MHC-I–restricted Ag-specific cytotoxicity by CD8 T lymphocytes in a contact-dependent manner (Fig. 3). Thus, local IL-15 may enhance the propensity of parenchymal CD8 T cells to lyse MHC-I–expressing target cells such as oligodendrocytes, the cellular target of MS demyelination. Our previous work demonstrates that oligodendrocytes in MS lesions express NKG2D ligands (40). Moreover, we have shown that the same cytokine combinations (e.g., IFN-γ + TNF) that increase surface IL-15 on human astrocytes, leading to elevated NKG2D and lytic enzyme expression by CD8 T lymphocytes, augment NKG2D ligands on human oligodendrocytes (40). Thus, the concomitant presence within MS lesions of IL-15–expressing astrocytes and NKG2D ligand-expressing oligodendrocytes may enhance CD8 T lymphocyte effector functions and consequently exacerbate the CD8 T lymphocyte-mediated killing of oligodendrocytes. Supernatants from activated astrocytes, or insertion of a transwell separating both cell types, did not reproduce the coculture results. Cells from both immune (monocytes and dendritic cells) (21, 58–60) and nonimmune origin (synovial fibroblasts and endothelial cells) (26, 61, 62) have previously been shown to activate T lymphocytes via contact-dependent IL-15–mediated mechanisms. Our functional studies were conducted using fetal astrocytes, as we are unable to isolate adult astrocytes in sufficient number or purity. However, our immunohistochemistry data demonstrate that human adult astrocytes express significant amounts of IL-15, especially in MS lesions (Fig. 4).

Our immunohistochemistry studies show that IL-15 expression is significantly increased in MS lesions compared with controls and that astrocytes are major sources of these augmented cytokine levels (Fig. 4). Moreover, enhanced IL-15 levels were observed in both acute (blocks 4–6 and 9) and subacute/chronic (blocks 7, 8, and 10) MS lesions, suggesting that this cytokine is upregulated for extended periods of time within the inflamed CNS during MS. Astrocytes are the most abundant glial cell type within the CNS. Moreover, these cells are strategically localized at the blood–brain barrier and thus can interact with incoming immune cells, including CD8 T lymphocytes as soon as they enter the CNS. Several functions have been attributed to astrocytes (53). Our data also indicate that the numerous macrophages and microglia in MS lesions provide IL-15 to infiltrating immune cells, such as CD8 T lymphocytes, in the CNS parenchyma. Moreover, several studies have documented the capacity of human monocytes/macrophages for providing functional IL-15 to T cells (20, 23), supporting the notion that CNS macrophages and most likely microglia do the same. In line with our findings, astrocytes have been shown to be the main source of IL-15 in LPS-injected mice, with reactive microglia being an additional source (38). Dendritic cells, found mainly in the perivascular spaces in MS lesions, could serve as an additional source of IL-15 (21). Finally, CD8 T cells detected in the parenchyma and in perivascular cuffs of MS lesions were in close proximity to IL-15–expressing cells (Fig. 6).

Blockade of systemic IL-15 (or IL-15 signaling receptor) in animal models featuring an inflammatory autoimmune response (e.g., rheumatoid arthritis, diabetes, psoriasis) has been shown to decrease disease development and severity (63–66). Recently, Gomez-Nicola and colleagues (67) observed aggravated experimental autoimmune encephalomyelitis in IL-15 knockout mice compared with wild-type littermates, although the maximum clinical score (1.6) they obtained for wild-type animals was low compared with other reports. However, because IL-15 null mice displayed marked reductions in the numbers of NK and memory CD8 T cells (68), indicating an abnormal immune system, interpretations of the role IL-15 using these mice should be made with extreme caution. Other groups have demonstrated that nonmicroglial CNS cells, especially astrocytes via NF-κB signaling, respond to CNS inflammation and play a deleterious role in autoimmune CNS inflammation during experimental autoimmune encephaylomyelitis (69, 70). We have previously shown that human astrocytes upon TLR3 ligation produce chemokines and cytokines (i.e., IL-6) (41) that can activate T cells. Our data assign a role to these cells as stimulators of infiltrating CD8 T lymphocytes in the inflamed CNS during MS. Moreover, we are currently investigating whether other factors derived from astrocytes contribute in maintaining or dampening T cell activation. Although not addressed in this study, IL-15 could participate in other aspects of neuroinflammation given its chemoattractive and prosurvival properties (71, 72). For example, IL-15 has been previously shown to be a potent T cell chemoattractant during rheumatoid arthritis, resulting in enhanced T cell migration to inflamed joints (73, 74). Its prosurvival properties could contribute to aspects of neuroinflammation.

Studies in humans support a role for IL-15 in the target organs of rheumatoid arthritis (26, 62, 74) and celiac disease (75–78). In the latter, prolonged exposure of CD8 T lymphocytes to IL-15 in the inflamed gut epithelium greatly enhances their cytotoxicity toward target cells. This was suggested to be a mechanism at the basis of the tissue destruction observed in patients with celiac disease. Our data support the proposal that a similar mechanism could potentially be at work in the inflamed CNS of patients with MS. In conclusion, we demonstrate that in the target organ of MS pathogenesis, astrocytes via the production of IL-15 contribute to perpetuating the activation of damaging CD8 T lymphocytes.

We thank Ellie McCrea, Diane Beauseigle, Manon Blain, Caroline Lambert, and Janet Laganière for technical assistance. We also thank Drs. André Oliver and Jeffrey Hall for providing adult human CNS tissues for the in vitro studies, Dr. Alexandre Prat for access to the Leica confocal microscope and the LSRII flow cytometer, and Drs. Christine Vande Velde and Réjean Lapointe for scientific discussion.

Disclosures The authors have no financial conflicts of interest.