TGF-β1 Represses Activation and Resultant Death of Microglia via Inhibition of Phosphatidylinositol 3-Kinase Activity¹

Microglia are functionally equivalent to peripheral macrophages in the CNS (1). In various neuropathologies, microglia are found to be activated by cytokines and injured or dead neuronal cells (1, 2, 3, 4, 5). Activated microglia have been closely associated with various neurodegenerative diseases such as stroke, trauma, Alzheimer’s disease, multiple sclerosis, and HIV-associated dementia (6, 7, 8, 9, 10). Immunostimulated microglia react to those diseases through the secretion of bioactive molecules including reactive oxygen or nitrogen species and inflammatory cytotoxins. Sustained overproduction of those molecules would cause severe damage to the normal brain region as well as the pathological one. Therefore, the overactivation of microglia should be repressed by any means. Apoptosis of activated microglia is one way an organism can regulate immune and inflammatory responses. It has recently been reported that overactivation by LPS causes microglia apoptotic death (11). The balance between microglial activation and activation-induced cell death has to be tightly regulated by multiple cytokines and chemokines.

TGF-β1 is a pleiotropic cytokine that regulates cell growth and differentiation (12) and exert both pro- and antiapoptotic effects in cell type-dependent manners (13, 14, 15). Furthermore, TGF-β1 has been generally considered as a major anti-inflammatory cytokine, although it also has proinflammatory properties depending on the cell type, tissue of origin, or a variety of other factors (16). Increased TGF-β1 mRNA expression in rat brain following transient forebrain ischemia has been demonstrated (17). Besides, activation of microglia by proinflammatory factors leads to increased production of TGF-β1 that may counteract the proinflammatory reactions (18, 19). Nonetheless, the role of TGF-β1 in relation to anti-inflammatory and antiapoptotic functions in the CNS remains controversial. In the present study, we investigated the regulatory effect of TGF-β1 on the production of various kinds of inflammatory mediators and resultant suicidal death in activated microglial cells.

Out of many intracellular signaling molecules that have been identified, phosphatidylinositol 3-kinase (PI3K)³ is one of the key molecules in cell proliferation and it mediates survival signals in a wide range of cell types (20, 21, 22, 23). PI3K is also involved in regulation of inducible NO synthase (iNOS) expression. For example, treatment of macrophages with LPS increases PI3K activity, and PI3K inhibitors wortmannin and LY294002 (24, 25) up-regulate iNOS through sustained activation of NF-κB (26). Similarly, PI3K inhibition increases iNOS levels in the HT-29 epithelial cell line and LPS- or cytokine-stimulated C6 glial cells (27, 28). In LPS-stimulated RAW 264.7 cells, however, neither wortmannin nor LY294002 reduces iNOS protein production, suggesting that LPS-mediated NO production occurs via a PI3K-independent pathway (29). These findings indicate that the role of PI3K in NO production and inflammatory responses may depend on the cell type and are not yet clear in microglia cells.

Mitogen-activated protein kinases (MAPKs) (30, 31) and a transcription factor NF-κB (32, 33, 34) are important regulators of iNOS and TNF-α expression in LPS-activated inflammatory cells such as macrophages and rat primary microglia. Three MAPK families have been identified in mammalian cells (35): the p42/p44 MAPK 1 (MAPK kinase 1 (MEK1)), the c-Jun N-terminal kinases (JNKs), and the p38 MAPKs. LPS activates MEK1, JNK, and p38 MAPK and the involvement of p38 MAPK in iNOS induction has been discussed (31, 36). Activation of the NF-κB is also known to be a key component of production of NO and proinflammatory cytokines in activated macrophage and microglia (33, 37). Inhibition of NF-κB activity appears to be linked to anti-inflammatory mechanisms (38, 39).

In the present study, we examined the effects of TGF-β1 on LPS-stimulated inflammatory responses and cell death and the potentially critical role of PI3K in these processes in microglia. Understanding the processes of microglial activation and activation-induced cell death should delineate the therapeutic target molecule to reduce brain inflammation and resultant neuronal injury or death in neurodegenerative diseases.

LPS (Escherichia coli serotype 055:B5), LY294002, nitro-l-arginine methyl ester (L-NAME), and MTT were obtained from Sigma-Aldrich (St. Louis, MO). FBS and recombinant mouse TGF-β1 was purchased from Life Technologies (Grand Island, NY).

The murine BV2 cell line (generous gift from Tong Joh, Burke Institute, Cornell University, Ithaca, NY) that immortalized after infection with a v-raf/v-myc recombinant retrovirus exhibits phenotypic and functional properties of reactive microglial cells (40). BV2 cells were maintained at 37°C at 5% CO₂ in DMEM supplemented with 10% heat-inactivated endotoxin-free FBS, 2 mM glutamine, 100 μg/ml streptomycin, and 100 U/ml penicillin.

Cultures of primary microglial cells were established based on the differential adherence of cells harvested from fetus rat cortex. The methods were modified from Frei et al. (41). In brief, mixed cell cultures were prepared in T75 Falcon flasks from postnatal day −1 rat cerebral cortices in DMEM/F12 with 10% FBS. Cortices were dissociated by passing through a 130-μm nylon mesh and plated at 2 × 10⁵ cells/cm². Cultures were fed every 3–4 days with modified Eagle’s medium (MEM) supplemented with 10% fetal FBS. On days 12–13, the culture plate was shaken on a rotatory shaker at 200 rpm for 2 h (41). The suspended cells were plated on 24- or 48-well culture plates and incubated for 1 h at 37°C. The medium containing suspended cells was discarded and adherent cells were further incubated with 1% FBS-supplemented DMEM for future experiments. The enriched microglia were nearly pure (>95%) as judged by immunocytochemical staining for a microglia-specific marker, the CR3 complement receptor, detected by the Ab OX-42 (Roche Diagnostics, Indianapolis, IN).

NO production from activated microglia was determined by measuring the amount of nitrite, a relatively stable oxidation product of NO, as described previously (42). In brief, an aliquot of the conditioned medium was mixed with an equal volume of 1% sulfanilamide in water and 0.1% N-1-naphthylethylenediamine dihydrochloride in 5% phosphoric acid. The absorbance was determined at 570 nm. Sodium nitrite, diluted in culture medium at concentrations of 10–100 μM, was used to generate a standard curve.

Primary cultured rat microglia and BV2 cells were grown in 24-well plates at a concentration of 5 × 10⁴ cells/well followed by proper treatment. Morphological change was examined under the upright-type phase-contrast microscopy (Olympus, Melville, NY). To measure the cell viability, 50 μl of 5 mg/ml MTT in growth medium was added to each well. After incubation for 90 min at 37°C with MTT, cell medium was removed. The precipitated formazan, a product of the MTT tetrazolium ring by the action of mitochondrial dehydrogenases, was solubilized with DMSO and quantified spectrophotometrically at 550 nm. MTT assay reflects the metabolic activity of cells and serves as a helpful indicator of cell viability.

Total cellular RNA was extracted with TRIzol (Life Technologies) according to the manufacturer’s protocol. Total RNA (2 μg) was reverse transcribed for 1 h at 37°C in a reaction mixture containing 5 U RNase (Amersham, Piscataway, NJ), 0.5 mM dNTP (Boehringer Mannheim, Indianapolis, IN), 2 μM random hexamer (Stratagene, La Jolla, CA), 1× reverse transcriptase buffer, and 5 U reverse transcriptase (Qiagen, Valencia, CA). PCR was performed using primers for iNOS, IL-1β, TNF-α, IL-6, caspase 11, matrix metalloproteinase 9 (MMP9), IFN response factor (IRF-1), IFN-γ-inducible protein of 10 kDa (IP-10), Fas, and β-actin as below. Analysis of the resulting PCR products on 1% agarose gels showed single-band amplification products with expected sizes (Table I).

Whole-cell protein lysates of BV2 cells were prepared in lysis buffer (10 mM Tris, 140 mM NaCl, 1% Triton X-100, 0.5% SDS, and protease inhibitors, pH 8) and cleared from cellular debris by centrifugation. The supernatants were aliquoted and stored at −70°C for further use. Samples were assayed for protein concentration using the Bradford (Bio-Rad, Richmond, CA) assay. Protein samples (20–40 μg for each) were separated by 10% SDS-PAGE and transferred to polyvinylidene difluoride membranes. The membrane was blocked with 5% BSA in TBST solution. The blots were incubated with polyclonal caspase 3 Ab (Santa Cruz Biotechnology, Santa Cruz, CA), anti-phospho-MEK1 (New England Biolabs, Beverly, MA), anti-phospho-p38 MAPK (New England Biolabs), anti-phospho-JNK (New England Biolabs), anti-MEK1 (Santa Cruz Biotechnology), anti-p38 MAPK (Santa Cruz Biotechnology), anti-Akt, anti-phospho-Akt (Cell Signaling Technology, Beverly, MA), and anti-JNK (Santa Cruz Biotechnology) Abs in blocking solution overnight at 4°C according to the manufacturer’s direction for dilution. After washing with TBST, HRP-conjugated secondary Abs (1/3000 dilution in TBST; New England Biolabs) were applied and the blots were developed by the ECL detection system (Amersham). Protein content was determined with the bicinchoninic acid reagents from Pierce (Rockford, IL) using BSA as standard.

BV2 cells (2 × 10⁷) were treated with LPS and/or TGF-β1 or LY294002 for 30 min. Nuclear proteins were prepared according to the method of Dignam et al. (43). Protein concentration was estimated using the Bradford assay (Bio-Rad) with BSA as the standard. The prepared nuclear extracts were stored in small aliquots at −80°C until use. The dsDNA oligonucleotide probe contained the consensus NF-κB binding site (Promega, Madison, WI) labeled by polynucleotide kinase (New England Biolabs). Radiolabeled double-stranded oligonucleotides were purified through Sephadex G-25 spin columns. Probes were stored at −20°C until use. Aliquots of nuclear protein (20 μg) were incubated with labeled oligonucleotide in binding buffer (50 mM KCl, 12.5 mM HEPES (pH 7.6), 6.25 mM MgCl₂, 0.05 mM EDTA, 0.5% Nonidet P-40, 0.5 mM DTT, 5% glycerol, and 2 μg of poly(dI-dC)) for 30 min on ice. The reaction samples were separated on 5% polyacrylamide gel (acryl:bisacrylamide, 30:1) with 0.5× Tris-borate-EDTA buffer containing 2% glycerol, and the gel was electrophoresed at 200 V at room temperature for ∼1.5 h. The gel was then dried and exposed to x-ray film (Kodak, Rochester, NY) with an intensifying screen at −80°C.

Transfection of the NF-κB gene into BV2 cells was performed using Lipofectamine plus transfection reagent per the manufacturer’s instruction (Invitrogen, Carlsbad, CA). The NF-κB reporter plasmid contained three copies of the κB-binding sequence fused to firefly luciferase gene (Clontech Laboratories, Palo Alto, CA).

BV2 cells were lysed in ice-cold lysis buffer (10 mM Tris-HCl (pH 7.4), 0.5% Nonidet P-40, 150 mM NaCl, 1 mM CaCl₂, 1 mM MgCl₂, 10 μg/ml aprotinin, 50 μM leupeptin, 2 mM sodium vanadate, 1 mM PMSF, 1 mM DTT, and 5% glycerol) and centrifuged at 13,000 × g. The cleared supernatant was assayed for protein quantification and 500 μg of total proteins was immunoprecipitated with anti-p85 subunit of PI3K Ab (Upstate Biotechnology, Lake Placid, NY). The precipitated proteins bound on protein A-Sepharose beads were resuspended in kinase buffer containing phosphatidylinositol 4,5-bisphosphate (0.2 mg/ml), 10 μCi of [γ-³²P]ATP, and 20 mM MgCl₂ for 10 min (44). The reaction was terminated by adding chloroform:methanol:12 M HCl (1:1) and freeze dried. The products were resuspended in 15 ml of chloroform, resolved by TLC in chloroform:methanol:ammonium hydroxide:water (86:76:10:14), and visualized by autoradiography.

Cells harvested by scraping were suspended in cell lysis buffer containing 50 mM Tris (pH 7.4), 0.03% Nonidet P-40, and 1 mM DTT and the lysates were cleared from cellular debris by centrifugation. Protein concentration was determined and 50 μg of protein per assay was examined for caspase 3 activity in microtiter plates. Caspase 3 colorimetric substrate Ac-DEVD-pNA (p-nitroanilide; Sigma-Aldrich) (45) was added to a final concentration of 200 mM and absorption was read on a microplate reader at 405 nm at the indicated time points. The specificity of the caspase substrate was tested by pre- and coincubation with the tetrapeptide inhibitor Ac-DEVD-CHO. The values were expressed as relative units.

The data are expressed as the mean ± SEM and analyzed for statistical significance using ANOVA, followed by Scheffe’s test for multiple comparison. A p < 0.05 was considered to be significant.

It has recently been reported that overactivation by LPS causes microglia apoptotic death (11). We examined the effect of TGF-β1 on LPS-mediated cell death in BV2 cells. BV2 cells were pretreated for 1 h with TGF-β1 and further treated with 0.2 μg/ml LPS for up to 48 h. Morphological studies using phase-contrast microscopy showed that a majority of cells appeared dead after a 48-h treatment with LPS. Inclusion of TGF-β1 completely prevented LPS cytotoxicity (Fig. 1,A). MTT assay showed a significant death of BV2 cells 48 h after LPS treatment, which was also completely blocked by TGF-β1 (Fig. 1,B). NO was previously reported as an autocrine mediator of inflammation-induced death of microglia (46). LPS markedly induced cell death and NO production in BV2 cells (Fig. 1, A and B). We further found that TGF-β1 down-regulated LPS-induced iNOS mRNA expression (Fig. 1,C) and NO production (Fig. 1,D). However, LPS-induced cell death was not significantly reduced by the nonselective NO synthase inhibitor nitro-l-arginine methyl ester (l-NAME; 1 mM), implying that LPS triggers another major cell death pathway(s) that can be regulated by TGF-β1 (Fig. 1, A and B).

In a wide range of cell types, PI3K has been reported to play a crucial role in cell survival signaling (20, 21, 22, 23). Microscopic observation and MTT assay showed that 1-h pretreatment with a PI3K inhibitor, LY294002 (20 μM), almost completely blocked LPS (0.2 μg/ml, 48 h)-induced BV2 cell death (Fig. 2, A and B). Like TGF-β1, LY294002 largely down-regulated LPS-induced iNOS mRNA expression (Fig. 2,C) and NO production (Fig. 2 D). Similar results in cell death and NO generation were also obtained with another PI3K inhibitor, wortmannin (100 nM; data not shown).

We examined whether TGF-β1 and LY294002 could inhibit LPS-stimulated cytotoxicity and NO production in primary cultured rat microglial cells. Pretreatment with TGF-β1 (5 ng/ml) or LY294002 (20 μM) for 1 h reduced the cell death (Fig. 3,A) and NO production caused by LPS (1 μg/ml; Fig. 3 B).

Activated microglial cells are known to produce various kinds of cytokines and cell death signaling molecules (47, 48). Therefore, we examined the effect of TGF-β1 and LY294002 on production of such molecules by LPS. BV2 cells were pretreated for 1 h with TGF-β1 (5 ng/ml) or LY294002 (20 μM) and stimulated with 0.2 μg/ml LPS for 24 h. mRNA levels were measured by RT-PCR using total RNA. We found that LPS stimulated IL-1β, TNF-α, IL-6, MMP9, caspase 11, IRF-1, IP-10, and Fas mRNA expression in BV2 cells (Fig. 4,A). Both TGF-β1 and LY294002 inhibited expression of most of these mRNA species, except IL-1β. Since TNF-α is a major molecule induced by various inflammatory stimuli and it plays a central role in various inflammatory diseases (49), we measured the amount of TNF-α released into the culture medium after 24 h of incubation. Consistent with the RT-PCR result, pretreatment of BV2 cells with TGF-β1 or LY294002 significantly inhibited LPS-induced TNF-α production (Fig. 4 B).

To study whether PI3K is a critical signaling molecule in the cytoprotective effect of TGF-β1, BV2 cells were pretreated with TGF-β1 (5 ng/ml) or LY294002 (20 μM) for 1 h and then further treated with LPS (0.2 μg/ml) for 1 h. PI3K was immunoprecipitated from total cell lysates and lipid kinase activity was measured in vitro. We found that LPS increased PI3K activity and LY294002 inhibited this response, as expected (Fig. 5,A). Interestingly, TGF-β1 also inhibited LPS-induced PI3K activity, suggesting the PI3K involvement in the TGF-β1 signaling pathway. Thus, we further tested the involvement of the PI3K pathway on TGF-β1-mediated anti-inflammatory reaction by examining the phosphorylation of Akt, a well-known target of PI3K. LPS-stimulated Akt phosphorylation was significantly suppressed by both LY294002 and TGF-β1 (Fig. 5 B).

Activation of p38 MAPK plays a critical role in LPS-mediated NO generation (31). LPS increased the activities of MEK1, p38, and JNK, as observed previously by us (31). In the present study, we found that TGF-β1 and LY294002 down-regulated the LPS-induced p38 and JNK phosphorylation (Fig. 6). In addition, phosphorylation of MEK1 was inhibited by LY294002.

The transcription factor NF-κB has been associated with iNOS induction in macrophages (50) and microglia (51) and apoptotic signaling in cancer cells (52). In addition, many anti-inflammatory mediators act by modifying the activity of NF-κB. Thus, we further tested the regulation of NF-κB DNA binding activity by both TGF-β1 and LY294002 in BV2 cells. We found that LPS stimulated NF-κB DNA binding activity and that this activity was enhanced by TGF-β1 and LY294002 (Fig. 7A). We assayed NF-κB transcriptional activity by transfecting BV2 cells with a plasmid containing a triple NF-κB binding site and a luciferase reporter gene. We found that both TGF-β1 and LY294002 enhanced LPS-stimulated luciferase activity 2- to 3-fold (Fig. 7 B). Interestingly, however, the chemical NF-κB inhibitor pyrrolidine dithiocarbamate repressed NO production and cell death (our unpublished result). Thus, NF-κB appears to be involved in LPS-induced NO production and cell death in a complex manner by interacting with other molecules.

Activation of caspase 3 requires cleavage of the 32-kDa inactive precursor to generate the active 17-kDa form. It was previously reported that a caspase 3-specific inhibitor reduced activation-induced apoptosis in BV2 cells, whereas inhibitors specific for caspases 1 and 8 had no effect (48). In the present study, we observed that the active caspase 3 subunit appeared in total cell lysate of LPS-activated cells, and this appearance was inhibited by either TGF-β1 or LY294002 (Fig. 8,A). Inhibition of caspase 3 enzymatic activity by TGF-β1 and LY294002 was also observed in an in vitro assay determined in cell lysates using a specific fluorescence substrate (Ac-DEVD-pNA) (Fig. 8 B).

TGF-β1 is produced by microglial and astroglial cells under various pathological conditions such as ischemia and Alzheimer’s disease (53, 54, 55). TGF-β1 has been shown to protect neurons from a variety of insults, but precisely how TGF-β1 protects neurons is not fully understood. Although TGF-β1 has been reported to have both pro- and anti-inflammatory functions depending on cell types and experimental conditions, its prominent role in suppression of inflammation is well established. In particular, the inhibitory effect of TGF-β1 on NO biosynthesis in activated macrophages and microglia has been well documented (56, 57, 58). Nonetheless, understanding how TGF-β1 regulates iNOS or other inflammatory mediators remains to be answered. In the present study, we showed that TGF-β1 strongly suppressed LPS-induced cell death as well as production of NO and other proinflammatory mediators in primary rat microglial cells and BV2 murine microglial cells.

Our investigation on the mechanisms by which TGF-β1 suppresses LPS-mediated microglia activation focused on our findings that TGF-β1 and LY294002 had common effects on LPS-mediated inflammation and cell death in microglia. We also found that another PI3K inhibitor, wortmannin, had a similar effect on LPS-induced NO generation and apoptosis (data not shown). Since PI3K is in general involved in early intracellular signaling, it is likely that inhibition of PI3K by TGF-β1 blocked early signaling involved in LPS-stimulated induction of NO, TNF-α, IL-6, IRF-1, IP-10, MMP9, Fas, and caspase 11. On the contrary, however, some previous studies showed that inhibition of PI3K by LY294002 or wortmannin increased NO synthesis in LPS-stimulated peritoneal macrophages and RAW 264.7 cells (26, 59). In those studies, activation of macrophages with LPS triggered PI3K, but PI3K inhibitors up-regulated iNOS expression mainly through a mechanism involving sustained NF-κB activity. The discrepancy between our present and the previous results of others could be due to the different cell types and/or interaction with other signaling molecules.

Although NO is known to be a major cytotoxic mediator involved in apoptosis in activated microglia, we found that inhibition of NO production did not block LPS-induced cell death, implying that inflammatory molecules in addition to NO play a role in activation-induced microglial death. Inflammatory stimuli induce or activate a specific group of genes and signaling pathways, some of which are commonly involved in both activation and cell death. It is more likely that activation-induced cell death is caused by the combinatorial effect of these inducible molecules. Therefore, the cytoprotective effect of TGF-β1 and LY294002 results from direct inhibition of apoptotic molecule expression or signaling, or indirect reduction of inflammatory signals. Although identification of direct mediators and interplay of various cytokines or molecules that contribute to activation-induced cell death requires further studies, previous reports suggest that combined treatment of LPS and IFN-γ induced the production of cytotoxic NO through IRF-1 expression and also initiated the NO-independent apoptotic pathway through the induction of caspase 11 expression (39, 48). Therefore, the NO production and death of microglial cells by LPS could be independently regulated at least in part through inhibition of IRF-1 and caspase 11 expression by TGF-β1 and LY294002 (39, 48).

NF-κB has been associated with iNOS induction and antiapoptotic signal transduction (32, 34, 50). Similarly, in our preliminary study, we found that suppression of NF-κB activity by the chemical inhibitor pyrrolidine dithiocarbamate repressed LPS-induced NO production and inflammation-induced cell death in microglial cells (our unpublished result). Interestingly, however, in the present study, we found that LPS-induced NF-κB activation was further enhanced by TGF-β1 and LY294002 both. Although the possible role of NF-κB in LPS-induced inflammation and cell death in microglia requires further investigation, the cell type- and experimental condition-dependent effect could be due to the interplay of NF-κB with other factors such as MAPK.

Previously, we reported that p38 MAPK is closely associated with LPS-induced iNOS expression in BV2 cells (31). Our present results further showed that TGF-β1 and LY294002 exerted their anti-inflammatory and antiapoptotic effects in part by down-regulating p38 activation in LPS-activated microglia. Although the role of JNK in microglia activation is not defined, its critical role in LPS-induced iNOS expression has been demonstrated in macrophages by several workers (30, 60). The iNOS gene promoter region contains several binding sites for NF-κB, AP-1, and other transcription factors, and AP-1 can be activated by JNK signaling. Our unpublished result showed that the JNK-specific inhibitor SP600125 partially inhibited LPS-induced microglial cell death but not NO production, suggesting potential involvement of this pathway in the antiapoptotic effect of TGF-β1 and LY294002.

The present study is the first to show the positive regulatory role of PI3K in inflammatory reactions and activation-induced cell death in microglia. More importantly, by showing the inhibitory effect of TGF-β1 on LPS-induced PI3K activation, these studies reveal a part of the mechanism of suppressive action of TGF-β1 in brain inflammation. Because the PI3K pathway regulates very early inflammatory and apoptotic events, inhibition of PI3K will result in suppression of downstream events triggered by LPS. The present data correlating PI3K activity with cell death are of particular interest, because PI3K activity has previously been considered as a survival signal (20, 21, 22, 23). The cytoprotective effect of TGF-β1 produced in the CNS following the application of various stimuli may be consistent with the low numbers of apoptotic microglia found in in vivo models of CNS inflammatory diseases (61). Abnormal overactivation and resultant death of microglial cells may cause more neurotoxicity in various neurodegenerative diseases (62, 63). TGF-β1 may render activated microglia to return to inactivated resting forms. Therefore, our data indicate that the cross-talk between TGF-β1 and PI3K signaling pathways may be important therapeutic targets for treatment of neuroinflammatory diseases.