Granulocyte-Macrophage Colony-Stimulating Factor Induces an Expression Program in Neonatal Microglia That Primes Them for Antigen Presentation¹ Free

Microglial cells originate from bone marrow elements that migrate into the brain in two successive waves. The first wave enters during the fetal period, whereas the second one colonizes the brain during the neonatal period, a few days after birth and before the blood-brain barrier is formed. Once settled in the brain parenchyma, microglial cells are forged under the influences of the microenvironment into a resident population whose characteristics are still largely unknown (1, 2, 3). Upon closure of the blood-brain barrier, any further transmigration into the parenchyma and exchange with circulating blood elements is restricted. An exception to this rule is a specific population of perivascular and submeningeal microglial cells, which continue to be replenished by infiltrating monocytes (4, 5). Once in the brain, the parenchymal microglial cells are shaped under the influence of growth factors and cytokines released by astrocytes, as well as neurochemicals released by neurons (1, 6, 7, 8).

Two hemopoietic CSFs, M-CSF and GM-CSF, are thought to play important roles in neonatal and postnatal microglial differentiation (9, 10, 11). M-CSF and GM-CSF both stimulate proliferation of neonatal microglia (12), but differently affect their morphology, phenotype, and Ag-presenting function. Morphologically, neonatal microglial cells cultured in M-CSF appear amoeboid, whereas those cultured in GM-CSF assume a dendriform shape (13). Phenotypically, neonatal microglial cells treated with M-CSF display some characteristics of macrophage (mφ)³ whereas those cultured in GM-CSF assume some of the markers associated with an immature dendritic cell (DC)-like phenotype (11, 13, 14). Importantly, the cytokine production capacity and competence as APCs are also differentially regulated by exposure to M-CSF and GM-CSF (13, 14, 15, 16, 17). Priming of microglial cells with GM-CSF enables them to become fully competent as APCs, upon contact with activated T cells or proinflammatory cytokines (18, 19). Although the competency of microglial cells for Ag presentation under noninflammatory conditions is still a matter of debate; it is apparent that GM-CSF endows both neonatal and adult microglial cells with an enhanced ability to process and present Ags (18, 19).

M-CSF is constitutively produced by astrocytes from embryo to adulthood (20). However, although GM-CSF is produced during the neonatal period, it is scarcely present in the adult CNS (20, 21). In the normal adult CNS, the only source of GM-CSF is from peripherally primed, infiltrating T cells, which patrol the brain parenchyma (22, 23). Because extravasation of activated T cells into the CNS is non-Ag dependent, the majority of T cells will exit without meeting their specific Ag. In contrast, if the T cells encounter a cognate self ligand, a potential autoimmune process could begin (23) with massive production of proinflammatory mediators. Under such conditions, microglial cells, already primed by T cell-released GM-CSF, will become fully competent APCs. Activated astrocytes and endothelial cells could also function as an additional source of GM-CSF (23, 24). Thus, it appears that M-CSF plays an important role in shaping the adult resident parenchymal microglia, whereas GM-CSF could have an important role in promoting their proinflammatory function. In this respect, the functional states of microglial cells elicited by M-CSF or GM-CSF have been proposed to correspond to the resting or “activated” form of microglia, respectively (11).

DNA microarray analysis can provide a very broad transcriptional profile of cells under different culture conditions or differentiation states. In this study, the transcript levels of ∼6000 genes have been analyzed, in primary murine neonatal microglial cells either untreated or cultured with GM-CSF or M-CSF. Both cytokines led to large-scale changes in the microglia expression profile. M-CSF treatment primarily induced transcription of genes involved in tissue organization and remodeling, maintenance, and regulation of brain homeostasis. By contrast, GM-CSF treatment induced transcription of genes important for T cell stimulation, chemotaxis, Ag processing, innate immunity, and immunosuppression, thus readying microglia for Ag presentation.

Neonatal microglia was prepared from newborn SJL/J mice as previously described (13). Briefly, after careful removal of the meninges, brains were mechanically disrupted and passed through a 100-μm filter. Cells were seeded in MEM (Life Technologies, Rockville, MD) with the addition of 10% FCS, 5 μg/ml insulin (Life Technologies), and 2.0 mg/ml l-glucose (Sigma-Aldrich, St. Louis, MO) for 12–14 days. Mixed glial cultures were shaken overnight on an orbital shaker (first shake). Adherent glial cells were trypsinized, split, and reseeded for an additional 10–12 days of culture. The procedure was repeated twice (second and third shakes). The purity of each preparation was assessed by CD11b staining and was always >93%. GM-CSF (10 ng/ml) or M-CSF (5 ng/ml) were added at the beginning of the culture and again every 3 days. Microglial cells from the second shake were used for global gene expression analysis. In some experiments (Fig. 3, a and b), IFN-γ was added to the untreated control for 48 h. Surface staining was performed as previously described (13) using the following mAbs, hamster anti-mouse CD11c (clone HL-3), rat anti-mouse CD11b (clone M1/70), rat anti-mouse CD45 (clone 30-F11), rat anti-mouse CD24 (clone 30-F1), hamster anti-mouse B7-1 (clone 16-10A1), and rat anti-mouse B7-2 (clone GL1), all from BD PharMingen (San Diego, CA); rat anti-mouse DEC-205 (clone NLDC-145; American Type Culture Collection, Manassas, VA); goat anti-Tweak (Research Diagnostics, Flanders, NJ); and rat anti-mouse MARCO (clone ED31; a gift from P. Ricciardi-Castagnoli).

Total RNA was isolated from 10 million cells using TRIzol reagents (Life Technologies), and biotinylated cDNA was prepared and hybridized to Affimetrix microarrays using the manufacturer’s protocols. Transcripts exhibiting an average difference (between specific and control oligonucleotides) of 1500 were considered to be expressed: the percentages calculated using the Affimetrix “absolute call” were similar. Transcripts with average difference below this level were considered below the level of detection.

Total RNA was isolated using TRIzol reagents (Life Technologies). RNase protection assay was performed using 4–6 μg of total RNA using BD PharMingen Riboquant kit rCK-2 multiprobe template set following the manufacturer’s recommendations.

Spinal cord frozen sections (10 μm thick) were fixed in acetone at −20°C, washed in PBS, and blocked with BSA. The following primary Abs diluted in PBS/1% BSA were added to the slides and incubated overnight at 4°C: goat anti-Tweak, used at 5 μg/ml (Research Diagnostics), in combination with anti-mouse CD45-FITC conjugated (BD PharMingen). Sections were washed in PBS and incubated with rabbit anti-goat IgG conjugated to Alexa 594 (Molecular Probes, Eugene, OR) to detect Tweak. To amplify the signal from the green fluorescence, the amplification kit, Alexa Fluor 488 signal amplification system and Oregon Green Dye-conjugated probes (Molecular Probes) was used. Fluorescent microscopy was performed on an Olympus IX70 (Olympus, Melville, NY) with ×60 N.A. 1.4 infinity corrected optics and a PXL cold CCD camera (Roper Scientific, Tucson, AZ).

Microglial cells untreated or treated with GM-CSF (10 ng/ml), M-CSF (5 ng/ml), or IFN-γ (10 ng/ml) were pelleted and lysed in 1% Nonidet P-40, 150 mM NaCl, 50 mM Tris (pH 8), and supplemented with the protease inhibitor mixture (Complete Mini Roche), for 1 h at 4°C. Supernatants were collected by centrifugation and protein content was normalized. Proteins were run on a SDS-PAGE gel and transferred to polyvinylidene difluoride membrane. After transfer, the membrane was probed with CD74 (clone In-1; BD PharMingen) specific for the MHC-associated invariant chain or the anti-MHC class II mAb KL-304 (ATCC), followed by 1/5000 dilution of the secondary HRP-conjugated mAb and chemiluminescence detection.

MLR was performed combining different amount (from 25 to 200 × 10³ cells) of microglial cells either untreated or treated with GM-CSF or M-CSF, previously irradiated at 600 rad (Cs¹³⁷), with 200 × 10³ splenocytes from either B10BR, C57BL/6, or SJL/J (The Jackson Laboratory, Bar Harbor, ME). Proliferation was tested by [³H]thymidine incorporation. Bone marrow-derived DCs prepared as previously described (25) were used as positive control in each MLR.

Microglial cells were isolated from neonatal brain tissue of SJL mice, and cultured in the presence or absence of GM-CSF (10 ng/ml) or M-CSF (5 ng/ml) (13). Expression of RNA transcripts was analyzed for ∼6000 murine genes of known or predicted function from the National Center for Biotechnology Information Unigene database, using Affimetrix U74A microarrays. Marked differences were observed in the transcriptional profiles of the microglia after treatment with GM-CSF or M-CSF, as compared with each other or as compared with untreated cells. Over 600 transcripts were uniquely expressed under each of these conditions, with more than one-third of the transcripts exhibiting differential expression in largely nonoverlapping sets (Fig. 1,a). Large differences were observed in the transcript levels of expressed genes, with ∼16.5% of expressed genes exhibiting up- or down-modulation of at least 2.5-fold after cytokine treatment (Fig. 1,b). Separate experiments using independent primary microglia isolates showed a good correlation in expression levels for both GM-CSF and M-CSF treatments (R = 0.85, data not shown). In addition, many gene products were analyzed individually by flow cytometry or Western blotting to validate the transcript analysis (see Figs. 2–4 and 6). The high percentage of transcripts differentially expressed after treatment suggest that neonatal microglial cells undergo true lineage differentiation in response to M-CSF or GM-CSF, as opposed to simply a change in activation state.

Overall, the majority of genes substantially up-regulated in M-CSF-treated cells involved functions related to maintenance and regulation of brain homeostasis (Table I). These include genes involved in myelination growth and development, such as tenascin C (24-fold), ryudocan (21-fold), biglycan (9-fold), nerve growth factor (8-fold), and epidermal growth factor-like growth factor (36-fold) (26, 27, 28, 29); in processing of neuropeptides, such as peptidylglycine-α-amidating monooxygenase (11-fold) and carboxypeptidase H (10-fold) (30, 31); in tissue remodeling, such as collagens type I, IV, V, VI, and XI (from 3- to 11-fold); or in the heat shock response (hsp), for example, hsp47 (11-fold) and hsp105 (4-fold; Table I). In contrast, many of the gene products up-regulated in GM-CSF-treated cells appear to be involved in cellular processes related to immune cell activation, Ag presentation, and innate immunity. These are discussed in detail.

Although no proteins are known to be expressed uniquely on mφ or DCs, the pattern of expression of the integrins CD11b (Mac-1), CD11c, and the lectin-like receptor DEC-205 is considered characteristic, with expression of CD11c⁺, CD11b^low, and DEC-205⁺ associated with an immature DC phenotype, and a higher surface expression of CD11b^high observed for mφ. As previously reported (13), treatment of neonatal microglial cells with GM-CSF or M-CSF alters the expression of these markers toward a DC-like or mφ-like phenotype, respectively. These results matched well with the levels of the corresponding transcripts from the microarray analysis (data not shown).

Low-affinity receptors for GM-CSF are present on untreated microglial cells and are slightly up-modulated by GM-CSF and down-modulated by M-CSF (Table II). Interestingly, the high-affinity GM-CSFR (common β-chain among IL-3, IL-5, and GM-CSFR) is absent on untreated cells, but is up-regulated >26-fold upon GM-CSF treatment.

Consistent with their key role as sensors of CNS injury, neonatal microglia express mRNA encoding receptors for several proinflammatory cytokines including IL-6R, IFN-γR, and TNF-αR (Table II). Neither of the receptors appeared to be regulated by M-CSF or GM-CSF (Table II) with the exception of IL-6R, which is slightly down-regulated by both the hemopoietic factors.

Microglial cells also expressed receptors for anti-inflammatory cytokines, particularly the IL-10R and the secreted form of the IL-4R (Table II); both of which are slightly down-regulated by M-CSF but not GM-CSF.

The role of cytokine in the CNS goes beyond the regulation of microglia immune function. In this respect, cytokines like IL-7 and IL-11 have been shown to induce proliferation on cultured astroglia (32) as well as to induce glial differentiation (33).

Overall, the cytokine receptors profile of microglial cells do not appear to be significantly changed by treatment with both hemopoietic factors, with exception of IL-7R and IL-11R, absent in untreated microglial cells, but up-regulated by GM-CSF (Table II).

Under inflammatory degenerative and traumatic conditions in the brain, microglial cells are the resident primary source of both pro- and anti-inflammatory cytokines. Untreated microglial cells display transcripts for strongly immunosuppressive and/or proapoptotic cytokines like TGF-β1 and Tweak (Fig. 2, b and c), reflecting a strategy that appear to limit immune-mediated events within the CNS. Transcripts for TGF-β and Tweak were both decreased by M-CSF treatment. Tweak was also down-regulated by GM-CSF. Transcripts encoding for several chemokines were detected in untreated microglia, including mφ inflammatory protein (MIP)-1α, MIP-1β, MIP-1γ, MIP-2, monocyte chemoattractant protein (MCP)-1, MCP-3, and MCP-5. Treatment with GM-CSF markedly increased transcripts for chemokines C10, MIP-1α, MIP-1γ, and MIP-2, while at the same time down-regulating MCP-5. C10, MIP-1γ, and MIP-2 were also up-regulated by M-CSF treatment, albeit to a lower degree. MCP-1 was up-regulated only by M-CSF. Both global transcript analysis (Fig. 2,b) and RNase protection assay (Fig. 2,a) demonstrated induction of the proinflammatory IL-1α, IL-1β, and the IL-1R antagonist upon treatment with GM-CSF but not M-CSF. Both hemopoietic factors up-regulate mRNA encoding for IFN-β (Fig. 2 b).

Microglial cells are the primary APCs resident in the brain. In the resting condition, they express very low levels of class II MHC protein and the class-II-associated invariant chain, but these can be up-regulated by treatment with the inflammatory cytokines such as IFN-γ. Previously, we have reported that GM-CSF skewing of microglial cells made them more inclined to up-regulate class II MHC expression in response to proinflammatory cytokines (13). Due to a mismatch between the MHC haplotype of the microglia cells (H-2^s) and the microarray target mRNA (H-2^b), MHC expression could not be monitored by the global gene analysis. Therefore, class II MHC levels were monitored by Western blotting. Class II MHC protein, which is almost undetectable in untreated cells, was increased to a barely detectable level after treatment with GM-CSF, but not M-CSF, although class II MHC surface staining was not increased (data not shown). As expected, IFN-γ was a much better inducer of class II MHC proteins (Fig. 3,a). Similar results were observed for the class II-associated invariant chain chaperone Ii, for which both alternately spliced p31 and p41 variants as well as partially processed p10 and p12 forms were observed after GM-CSF but not M-CSF treatment (Fig. 3,b). This pattern for invariant chain was observed also by transcript analysis (Fig. 3 c). Transcript levels for the class II MHC-associated peptide exchange factor H2-M were low but detectable in each of the samples, and were not changed by either treatment. No significant change in class I MHC protein was observed after treatment with either M-CSF or GM-CSF (data not shown).

Activation of T cells requires costimulatory interactions in addition to generation and recognition of MHC-peptide complexes. Several costimulatory receptors are present on microglial cells. As previously reported, neonatal microglia express CD80 (B7-1) and CD86 (B7-2; Fig. 4, a and b), which provide a CD28-dependent costimulatory signal (18). The expression of B7-1 was slightly up-regulated by GM-CSF, whereas B7-2 profile was not affected. The heat-stable Ag, CD24, has recently been added to the list of costimulatory molecules involved in T cell activation (34). CD24 play an important role in CNS immunity since its targeted mutation completely abrogated development of experimental allergic encephalitis (EAE), although its mutation does not prevent the induction of autoreactive T cells (35). In this study, we report that CD24 was dramatically increased (122-fold) in GM-CSF-treated cells (Fig. 4, a and b). CD24 up-regulation on microglial cells could represent a novel, CD28-independent, costimulatory signal for expansion of autoreactive T cells.

Altogether, GM-CSF-skewed cells were endowed with enhanced T cell stimulatory function, as assessed in an MLR assay (Fig. 3 d).

Cathepsins are a large family of cysteine and aspartic proteases implicated in endosomal protein degradation and generation of peptides for Ag processing. Using global gene expression analysis, differential expression of many of the cathepsins was found. Cathepsin F, L, and S, which are involved in invariant chain degradation and peptide loading, are differently distributed among APCs. Cathepsin S is present in B cells, DCs, and mφ, whereas cathepsin F is more selectively expressed in tissue mφ. Neonatal microglial cells have been reported to express cathepsin S and L, but not F (13), while resting microglial cells have been reported to express cathepsin E (36). Global gene expression analysis confirms these observations. Interestingly, both cathepsins L and F were up-regulated by GM-CSF (Fig. 5,a), while cathepsin E was down-regulated (Fig. 5,a). Neonatal microglial cells expressed cathepsin B, C, D, and S, with levels not significantly regulated by either GM-CSF or M-CSF (Fig. 5 a).

Matrix metalloproteinases (MMP) are a family of proteases that have been implicated in processing of extracellular matrix during a variety of pathological conditions. In DC and Langerhans cells, MMPs enhance cell migration (37, 38). Treatment of microglia with GM-CSF strongly up-regulated expression of MMP-12 (mφ metalloelastase), MMP-9, and MMP-11 (stromelysin-3; Fig. 5 b), whereas M-CSF selectively up-regulates MMP-9. These proteases are potent collagenase, and may play an important role in remodeling the brain microenvironment during disease processes as well as affecting cell trafficking into the brain (39, 40). It is also tempting to speculate that these proteases may mediate extracellular Ag processing. Notably, antigenic determinants derived from myelin basic protein have been reported to be presented in the CNS through a pathway that does not involve internalization and endosomal processing (41).

Microglial cells respond to foreign pathogens by phagocytosis after receptor-mediated recognition of microbial Ags as well as producing several antimicrobial factors. Untreated microglial cells express mRNA encoding for several endocytic receptors, including Fcγ and complement receptors, the mφ scavenger receptor, and CD14, a receptor for LPS (Fig. 6,a). Among these, both FcγRI and RIII appeared to be down-modulated by the treatment with GM-CSF (Fig. 6,a). MARCO is a receptor expressed on both mφ and DC (F. Granucci and P. Ricciardi-Castagnoli, manuscript in preparation), which is important in DC for clearance of bacteria. MARCO transcripts are expressed in untreated microglial cells and up-regulated >15-fold in GM-CSF, but not in M-CSF-treated cells (Fig. 6,a). Surface staining with an anti-MARCO Ab confirmed the up-regulation observed by global transcript analysis (Fig. 6 b).

Microglia can produce antimicrobial peptides, which act as endogenous antibiotics, and are important contributors to regional innate immunity. Untreated microglial cells express mRNA encoding for natural resistance-associated mφ protein 1 and 2, as well as Cryptidin and cathelin-like peptides. The majority of these antimicrobial peptides are down-regulated by GM-CSF treatment, but unaffected by M-CSF.

The neonatal microglial cell phenotype closely resembles that of circulating monocytes (2, 11, 12, 13), and can be skewed to a more mφ-like or DC-like phenotype by M-CSF or GM-CSF, respectively (13). Using global gene expression analysis, we now show that upon treatment of microglial cells with either GM-CSF or M-CSF, a large proportion of expressed genes is up-regulated or down-regulated >2.5-fold (>16%). Such a large-scale shift in gene expression indicates that neonatal microglial cells undergo a marked transformation by either GM-CSF or M-CSF, and is further evidence that neonatal cells exist in a relatively undifferentiated state. It has been previously shown that in T cells activated by a specific super Ag, the number of genes differentially expressed is ∼4.4% at best (42). Also, in the DC response to different pathogens, the number of regulated genes is always <10% (43). Thus, it is likely that microglial cell skewing with either GM-CSF or M-CSF induces a true lineage differentiation as opposed to a different state of activation.

Although skewing of microglia toward a mφ-like phenotype occurs following exposure to M-CSF within the CNS, both growth factors are needed for full maturation of other tissue mφ. M-CSF^−/− mice are largely devoid of alveolar mφ and bone osteoclasts (although interestingly, they do retain near normal CNS microglial populations; Refs. 44 and 45). In contrast, alveolar mφs from GM-CSF^−/− animals are normal in number but exhibit deficits in function. GM-CSF^−/− alveolar mφ are unable to effectively degrade surfactant and are defective in their ability to clear pathogens (46).

The biological activity of GM-CSF is initiated by its binding to a specific receptor formed by an α and β subunit. The α subunit binds GM-CSF with low affinity (47). The β subunit does not bind GM-CSF alone, but in combination with the α subunit will form a high-affinity GM-CSFR (48). The interplay between the two GM-CSFRs currently is incompletely understood, although it is thought that in humans both receptors mediate myeloid lineage differentiation, but only the high-affinity one mediates cell growth (49). The low-affinity GM-CSFR was present on untreated microglial cells and was down-regulated by M-CSF, consistent with a skewing toward a mφ phenotype. Interestingly, the high-affinity receptor appeared only after GM-CSF treatment, again consistent with a differentiation away from the resting phenotype. Overall, the majority of transcripts regulated by M-CSF were involved in lipid, carbohydrate, or protein metabolism (Table I). Of particular interest was the up-regulation of a series of proteoglycans involved in tissue remodeling, as well as molecules important in glial-neuron interaction. In general, it appeared that M-CSF regulates functions already evident in untreated cells, and related to the role of microglial cells in brain homeostasis under physiological conditions. This is in accord with the fact that M-CSF is constitutively present in the brain parenchyma.

GM-CSF up-regulated a large number of transcripts strongly implicated in CNS immunity. Interestingly, both neonatal and adult microglial cells require a multistep activation process to reach full competency as APCs, wherein a preactivation with GM-CSF is followed by a secondary activation with a proinflammatory cytokine (19) or CD40 ligand (18). Under physiological conditions in the adult brain, the only source of GM-CSF is activated T cells, which survey the brain parenchyma (22, 23). The amount of GM-CSF released by these cells could be sufficient to induce a preinflammatory state or precompetency of local microglial cells. If this is followed by a sufficient number of infiltrating T cells, which either specifically or by cross-reaction recognize a cognate ligand, an autoimmune process could ensue. In contrast, if no productive interaction is generated, the activated T cells will leave the brain without inciting inflammation (23). The interaction between infiltrating T cells and microglia has been studied in a model of host vs graft disease, without interference from infiltrating mφ/monocytes. In this model, when a sufficient number of T cells crossed the blood brain barrier, the cytokine milieu produced was sufficient to activate resident microglial cells (50).

Costimulatory molecules play a fundamental role in the interaction between T cells and APCs. CD24, expressed on APC, provides a CD28-independent costimulation by interacting with either p-selectin, or in a homotypic interaction, with another CD24 molecule expressed on T cells (51). The functional significance of this newly described costimulatory pathway is as yet not understood, although it appears that CD24, different from members of the B7 family, is not required for the induction phase of an immune response (52). In the animal model of EAE, it was shown that targeted mutation of CD24 does not affect the priming and proliferation of self-reactive T cells; however, it does effect EAE development (35). Also, passive transfer of myelin oligodendrocyte glycoprotein 35-55-specific CD24⁺ T cells in CD24^−/− mice is not sufficient for EAE induction, because a still not well understood homotypic CD24 interaction must occur for extravasation of self-reactive T cells into the CNS (35). Because CD24 expression on microglial cells is dramatically up-regulated by GM-CSF, it seems likely that upon entry of activated GM-CSF-secreting T cells into the brain, CD24 would be readily up-regulated on microglial cells, which in turn would acquire higher costimulatory capacity to sustain T cell expansion.

Chemokines are small chemoattractant molecules, which regulate cell trafficking and homing in every organ. Microglial cells produce a basal level of several chemokines essential for regulating T cell infiltration and patrolling of the brain parenchyma. GM-CSF, much more than M-CSF, up-regulated the transcripts for several of these chemokines, further establishing its role as priming factor for brain immunity.

Microglial cells play a pivotal role in protein degradation and tissue remodeling, which occurs during degenerative, traumatic, and inflammatory CNS conditions. A series of aspartic proteases, cysteine proteases, and metalloproteinases are involved in this process. MMP-9, MMP-11, and MMP-12 are strongly up-regulated with GM-CSF and may participate in tissue remodeling and microglial cell migration. Cathepsin S, L, and F are involved in Ag processing and invariant chain degradation (53, 54, 55), both important for the Ag-presenting role of microglial cells. Both cathepsin F and L were up-regulated by GM-CSF. However, neither GM-CSF nor M-CSF altered the expression of cathepsin S, which was previously shown to degrade myelin basic protein at both neutral and acidic pH, and to be up-regulated by neurotrophic factors in microglial cells (56). Other proteins important for Ag processing and presentation also were up-regulated by GM-CSF, including both class II MHC protein and its associated chaperone invariant chain. Although up-regulation of class II MHC proteins by GM-CSF does not result in a large increase in surface expression, the small degree of up-regulation that was observed was sufficient to promote much stronger T cell proliferation in an MLR assay.

Several substances that are produced in the normal brain account for the maintenance of an immunosuppressive environment. Normal neuronal activity and secreted neurotrophin inhibit the microglia expression of MHC class II and costimulatory molecules, and this tends to counteract the activation of microglia and astrocytes by IFN-γ and LPS (7, 8). As observed in this study, microglial cells also produce TGF-β and Tweak, immunosuppressive cytokines that have been shown to be strongly apoptotic toward activated T cells. Consistent with this observation, activated T cells entering the brain parenchyma are known to be more likely to undergo apoptosis than to proliferate. Consistent with its role as preactivator of microglial cells, GM-CSF decreases expression of Tweak, and up-regulates proinflammatory cytokines like IL-1α and β along with several chemokines that can promote leukocyte migration into the brain.

In summary, the global gene expression analysis reported in this study establishes the molecular basis for the previously recognized role of GM-CSF in promoting microglial participation in adaptive immune responses. GM-CSF treatment leads to up-regulation of molecules involved in T cell costimulation (CD24), chemotaxis (C10, MIP-1α, MIP-1γ), Ag processing class II MHC, invariant chain, cathepsin F and L), innate immunity (MARCO), and to down-regulation of apoptosis-inducing factors (Tweak, TGF-β), which together constitute a differentiation process that prime microglia cells for Ag presentation.

We thank Hans-George Fischer and Gaby Reichmann for helpful suggestions in the preparation of the manuscript.