Dual role of inflammatory stimuli in activation-induced cell death of mouse microglial cells. Initiation of two separate apoptotic pathways via induction of interferon regulatory factor-1 and caspase-11.

We have previously shown that mouse microglial cells undergo apoptosis upon inflammatory activation and that nitric oxide (NO) is the major autocrine mediator in this process (Lee, P., Lee, J., Kim, S., Yagita, H., Lee, M. S., Kim, S. Y., Kim, H., and Suk, K. (2001) Brain Res. 892, 380-385). Here, we present evidence that interferon regulatory factor-1 (IRF-1) and caspase-11 are the essential molecules in activation-induced cell death of microglial cells. The apoptogenic action of inflammatory stimuli such as lipopolysaccharide (LPS) and interferon-gamma (IFNgamma) was mediated through the induction of IRF-1 and caspase-11 expression in two separate events. Although IRF-1 was required for NO synthesis, caspase-11 induction was necessary for NO-independent apoptotic pathway. Microglial cells from IRF-1-deficient mice showed markedly decreased NO production, and they were partially resistant to apoptosis in response to LPS/IFNgamma but were sensitive to NO donor exposure. LPS/IFNgamma treatment resulted in the induction of caspase-11 followed by activation of caspase-11, -1, and -3. Inactivation of caspase-11 by the transfection of dominant-negative mutant or treatment with the caspase inhibitors rendered microglial cells partially resistant to LPS/IFNgamma-induced apoptosis. Inhibition of both NO synthesis and caspase-11 completely blocked LPS/IFNgamma-induced cytotoxicity. These results indicated that LPS/IFNgamma not only induced the production of cytotoxic NO through IRF-1 but also initiated the NO-independent apoptotic pathway through the induction of caspase-11 expression.

Microglial cells are ubiquitously distributed in the central nervous system and comprise up to 20% of the total glial cell population in brain (1). Although the ontogeny of microglial cells has long been debated, recent works using monoclonal antibodies specific for microglial cells indicated that these cells are closely related to monocytes and macrophages (2). As the primary immune effector cells in the central nervous system, microglial cells migrate to the site of tissue injury or inflammation, where they respond to invading pathogens or other inflammatory signals. Like monocytes/macrophages, they also secrete inflammatory cytokines and toxic mediators, which may amplify the inflammatory responses (3,4). In general, microglial cells function in a manner very similar to monocytes/ macrophages. Recently, activated macrophages have been shown to undergo apoptosis (5)(6)(7). It has been suggested that the apoptosis of activated macrophages is one mechanism whereby an organism may regulate immune and inflammatory responses involving macrophages (7). We and others have demonstrated that a similar regulatory mechanism exists for microglial cells (8,9) and astrocytes (10) as well. Microglial cells and astrocytes underwent apoptosis upon inflammatory activation, and nitric oxide (NO) 1 acted as an autocrine cytotoxic mediator in this process (9,10). Activation of glial cells may be intended to protect neurons at first. More frequently, however, activation of glial cells and inflammatory products derived from them have been implicated in neuronal destruction commonly observed in various neurodegenerative diseases (4). These deleterious effects of glial activation may be exacerbated by the failure of autoregulatory mechanisms. Thus, our understanding of the pathogenesis of neurodegenerative diseases may be enhanced by elucidation of the molecular mechanism underlying autoregulation of microglial activation.
IRF-1 is one of the major transcription factors that mediate IFN␥ responses (11). In particular, IRF-1 has been shown to mediate inducible nitric oxide synthase (iNOS) expression in response to IFN␥ (12). The expression of IRF-1 is induced by IFN␥ as well as other inflammatory mediators (13). Macrophages (12) and mixed glial cells (14) deficient in IRF-1 did not respond to LPS or IFN␥ for NO production, indicating that IRF-1 is an essential mediator in NO production of these types of cells. The involvement of IRF-1 in cellular apoptosis has also been reported. The role of IRF-1 in the induction of apoptosis by DNA damage or IFN␥ has been suggested (15)(16)(17), supporting the proapoptotic action of IRF-1. Previous work in our laboratory has also shown that IRF-1 plays a central role in IFN␥/ TNF␣-induced apoptosis of mouse pancreatic islet ␤-cells (18) and ME-180 human cervical cancer cells (19). However, the role of IRF-1 in microglial NO production and apoptosis has not been investigated.
Caspases play a central role in central nervous system cellular apoptosis (20). A variety of apoptotic stimuli lead to the activation of initiator caspases, which in turn triggers the caspase cascade, ultimately resulting in apoptotic cell death. Among more than a dozen caspases identified so far, caspase-11 has been first characterized as an activator of caspase-1 (21), and this caspase has been proposed to play an important regulatory role in both apoptosis and inflammatory responses (22). Activation of caspase-11 was crucial for the activation of caspase-1 (22); however, recent studies have demonstrated that caspase-11 activates caspase-3 as well under pathological conditions (23). Caspase-11 has been shown to carry out an essential function in apoptotic death of neuronal cells and oligodendrocytes (23)(24)(25). Caspase-11-deficient mice were partially resistant to the induction of experimental allergic encephalomyelitis (25) and showed reduced numbers of apoptotic cells after middle cerebral artery occlusion (23). The expression of caspase-11 was increased by inflammatory stimuli (21) and hypoxia (24); this increase in the expression is believed to auto-activate caspase-11 (23). Recombinant procaspase-11 has been shown to auto-process itself in vitro, suggesting that an elevated concentration of caspase-11 in stimulated cells may be sufficient for its auto-activation (23). These previous reports on the critical functions of caspase-11 in central nervous system cellular apoptosis led us to examine the role of this caspase in microglial apoptosis.
In this work, we have demonstrated that both IRF-1 and caspase-11 are required for the activation-induced apoptosis of microglial cells and that these two molecules are strongly induced by inflammatory stimuli to participate in two separate events leading to microglial apoptosis. IRF-1 induction mediated the production of NO, which acted as an autocrine apoptotic mediator, whereas up-regulation of caspase-11 induced direct apoptotic pathways independently of NO production. NF-B was involved in NO production but not in apoptogenic action of NO or caspase-11.
Mice and Cells-Mice with a targeted mutation in the IRF-1 gene (homozygous mice and their heterozygous littermates) were kindly provided by Dr. Y. C. Sung at Postech (Pohang, Korea) (26) and bred in a virus-free facility at the Samsung Medical Center. The IRF-1 Ϫ/Ϫ, IRF-1 ϩ/Ϫ, and IRF-1 ϩ/ϩ colonies were maintained by mating IRF-1 ϩ/Ϫ to either IRF-1 Ϫ/Ϫ or IRF-1 ϩ/Ϫ mice. The genotype of all IRF-1 mice was determined by PCR of tail DNA using the following three primers, as described previously (26): IRF-1 forward primer, TTC CAG ATT CCA TGG AAG CAC GC; IRF-1 reverse primer, ATG GCA CAA CGG AAG TTT GCC; neo r reverse primer, ATT CGC CAA TGA CAA GAC GCT GG. PCR with these three primers amplify a 900-base pair sequence for IRF-1(ϩ) allele and a 700-base pair sequence for IRF-1(Ϫ) allele from genomic DNA. Mouse primary microglial cells were prepared as described previously with minor modifications (9,27). In brief, the forebrains of newborn wild-type or IRF-1-deficient mice were chopped and dissociated by trypsinization and mechanical disruption. The cells were seeded into poly-L-lysine-coated flasks. After in vitro culture for 10 days, microglial cells were detached by rapid and gentle shaking of the culture flasks and seeded into plastic surfaces. After an additional 1-h incubation, nonadherent cells were removed by replacing the culture medium. The purity of microglial cultures was greater than 90% as determined by OX-42 immunocytochemical staining (data not shown). The BV-2 mouse microglial cell line, originally developed by Dr. V. Bocchini at University of Perugia, Italy (28) was generously provided by Dr. E. Choi at Korea University (Seoul, Korea). RAW 264.7 cells were purchased from American Type Culture Collection (Manassas, VA). The cell lines were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 2 mM glutamine, and penicillin-streptomycin (Life Technologies, Inc.).
Assessment of Cytotoxicity by MTT Assay-Cells (3 ϫ 10 4 cells in 200 l/well for BV-2 cells, 2 ϫ 10 4 cells in 200 l/well for mouse primary microglial cells) were seeded in 96-well plates and treated with LPS and IFN␥ for the indicated time periods. The optimal concentrations for the cytotoxic action were 100 ng/ml for LPS and 100 units/ml for IFN␥ (9). In some experiments, cells were pretreated with caspase inhibitors, MG-132, or NAC for 1 h before LPS/IFN␥ treatment. After the treatment, the medium was removed, and 3-[4,5-dimethylthiazol-2-yl]-2,5diphenyltetrazolium bromide (MTT, 0.5 mg/ml) was added followed by incubation at 37°C for 2 h in CO 2 incubator. After a brief centrifugation, supernatants were carefully removed, and Me 2 SO was added to the cells. After insoluble crystals were completely dissolved, absorbance at 540 nm was measured using a Thermomax microplate reader (Molecular Devices).
NO Quantification-After cells (3 ϫ 10 4 cells in 200 l/well for BV-2 cells, 2 ϫ 10 4 cells in 200 l/well for mouse primary microglial cells) were treated with activating agents in 96-well plates, NO 2 Ϫ in culture supernatants was measured to assess NO production in microglial cells. Fifty l of sample aliquots were mixed with 50 l of Griess reagent (1% sulfanilamide, 0.1% naphthylethylene diamine dihydrochloride, 2% phosphoric acid) in 96-well plate and incubated at 25°C for 10 min. The absorbance at 550 nm was measured on a microplate reader. NaNO 2 was used as the standard to calculate NO 2 Ϫ concentrations. Morphological Analysis of Apoptotic Cells-Morphological changes in the nuclear chromatin of cells undergoing apoptosis were detected by staining with 2.5 g/ml of Hoechst 33342 fluorochrome (Molecular Probes, Eugene, OR) followed by examination on a fluorescence microscope (Olympus BX50).
Reverse Transcription-Polymerase Chain Reaction (RT-PCR)-Total RNA was extracted from BV-2 cells or primary microglial cells by a sequential addition of 4 M guanidinium thiocyanate, 2 M sodium acetate, and acid phenol/chloroform. Reverse transcription was carried out using Superscript (Life Technologies, Inc.) and oligo(dT) primer. PCR amplification using primer sets specific for IRF-1 was carried out at 60°C annealing temperature for 20 (BV-2 cells) or 30 cycles (primary microglia). PCR for all caspases and ␤-actin was carried out at 55°C annealing temperature for 30 -35 cycles. Nucleotide sequences of the primers were based on published cDNA sequences of mouse IRF- Western Blot Analysis-Cells were lysed in triple-detergent lysis buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.02% sodium azide, 0.1% SDS, 1% Nonidet P-40, 0.5% sodium deoxycholate, 1 mM phenylmethylsulfonyl fluoride). The protein concentration in cell lysates was determined using a Bio-Rad protein assay kit. An equal amount of protein for each sample was separated by 12% SDS-polyacrylamide gel electrophoresis and transferred to Hybond ECL nitrocellulose membranes (Amersham Pharmacia Biotech). The membranes were blocked with 5% skim milk and incubated sequentially with primary antibodies (polyclonal rabbit anti-mouse IRF-1, Santa Cruz Biotechnology; polyclonal rabbit anti-mouse caspase-11, R&D Systems) and horseradish peroxidase-conjugated secondary antibodies (anti-rabbit IgG, Amersham Pharmacia Biotech) followed by ECL detection. Caspase-11 polyclonal antibody (R&D Systems) was raised against synthetic peptide CGTMHSEKTPDVLQYD corresponding to amino acids 202 -217 of mouse caspase-11 (within the p20 region).
Assessment of Caspase Activity-Caspase-1 or -3-like activities were measured using a Caspase assay kit (PharMingen) according to the supplier's instruction. In brief, caspase-1 or -3 fluorogenic substrates (Ac-YVAD-AMC or Ac-DEVD-AMC) were incubated with LPS/IFN␥treated cell lysates for 1 h at 37°C, and then AMC liberated from Ac-YVAD-AMC or Ac-DEVD-AMC was measured using a fluorometric plate reader (Bio-Tek FLx800TB) with an excitation wavelength of 380 nm and an emission wavelength of 420 -460 nm. Caspase-8 or -11-like activities were similarly measured using caspase-8 or -11 fluorogenic substrates (Ac-IETD-AFC or Ac-LEHD-AFC) followed by detection of liberated AFC with an excitation wavelength of 400 nm and an emission wavelength of 480 -520 nm.
Generation of Dominant-Negative Mutant of Caspase-11-Dominantnegative mutant inhibitor of caspase-11 (caspase-11 C254G ) was constructed as described previously (21). To change the cysteine residue in the active site of the caspase into a glycine residue, two primers containing the mutations were synthesized: forward primer, GTC ATC ATT GTG CAG GCC GGC AGA GGT GGG AAC TCT GG; reverse primer, CCA GAG TTC CCA CCT CTG CCG GCC TGC ACA ATG ATG AC. Then, the mutant construct was generated by PCR with wild-type caspase-11 cDNA construct as a template (kindly provided by Dr. M. Miura, Brain Science Institute, RIKEN, Saitama, Japan) (21) and the mutant primer set using QuickChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). The occurrence of the desired mutation was confirmed by nucleotide sequencing.
Transient Transfection-BV-2 cells in 6-well plates were transfected with 1 g of dominant-negative mutant of caspase-11 fused with lacZ or co-transfected with 1 g of phosphorylation-defective dominant-negative mutant IBa (29) or NF-B p65 expression plasmid (kindly provided by Dr. D. W. Ballard, Vanderbilt University, Nashville, TN) (30) together with 0.2 g of lacZ gene (pCH110, Amersham Pharmacia Biotech) using LipofectAMINE reagent (Life Technologies, Inc.). Alternatively, in some experiments, BV-2 cells were co-transfected with wild-type caspase-11 fused with lacZ and dominant-negative mutant of caspase-11 fused with lacZ or NF-B p65. In this case, the total amount of plasmids transfected was fixed at 2 g. At 48 h after the transfection, the cells were treated with LPS and IFN␥. After another 24 h, the cells were fixed with 0.5% glutaraldehyde for 10 min at room temperature and stained with 5-bromo-4-chloro-3-indolyl ␤-D-galactopyranoside (Xgal; 1 mg/ml) in 4 mM potassium ferricyanide, 4 mM potassium ferrocyanide, 2 mM magnesium chloride at 37°C for detection of blue cells. At least 250 blue cells were counted for each experiment, and transfection efficiency was 15-30%.
NF-B Reporter Assays-NF-B reporter activity was measured using the Dual-luciferase reporter assay system (Promega, Madison, WI). In brief, BV-2 cells in 12-well plates were co-transfected with 0.5 g of NF-B-responsive reporter gene construct carrying two copies of B sequences linked to luciferase gene (IgG NF-B-luciferase, generously provided by Dr. G. D. Rosen, Stanford University, Stanford, CA) (31) together with 0.1 g of Renilla luciferase gene under herpes simplex virus thymidine kinase promoter (pRL-TK, Promega) using Lipo-fectAMINE reagent (Life Technologies, Inc.). At 24 h after the transfection, cells were treated with stimuli. After 5 h, activities of firefly luciferase and Renilla luciferase in transfected cells were measured sequentially from a single sample using Dual-luciferase reporter assay system (Promega). Results were presented as firefly luciferase activity normalized to Renilla luciferase activity. In some experiments, cells were co-transfected with NF-B p65 expression plasmid (0.5 g) along with NF-B-responsive reporter plasmid (0.5 g) and pRL-TK (0.1 g) before conducting the luciferase assays.
Electrophoretic Mobility Shift Assays-Nuclear extracts were prepared from BV-2 cells or RAW 264.7 cells as described previously (32). Synthetic double-stranded oligonucleotides of consensus NF-B binding sequence, GAT CCC AAC GGC AGG GGA (Promega), were end-labeled with [␥-32 P]ATP using T4 polynucleotide kinase. Nuclear extract was incubated with the labeled probe in the presence of poly(dI-dC) in a binding buffer containing 20 mM HEPES at room temperature for 30 min. For competition assays, a 50-fold molar excess of unlabeled oligonucleotides was included in the reaction. DNA-protein complexes were resolved by electrophoresis in a 5% nondenaturing polyacrylamide gel, dried, and visualized by autoradiography.
Statistical Analysis-All data were presented as means Ϯ S.E. from three or more independent experiments. A statistical comparison between different treatments was done by either Student's t test or one-way analysis-of-variance with Dunnett's multiple comparison test using the Prism program (GraphPad Software Inc.). Differences with p value less than 0.05 were considered statistically significant.

Induction of IRF-1 by LPS/IFN␥ and Its Role in Microglial
NO Production and Subsequent Apoptosis-Previously, we observed that microglial cells activated by inflammatory stimuli underwent apoptosis as a possible autoregulatory mechanism for their own activation states and that NO produced by activated microglial cells was the primary cytotoxic mediator in this process (9). Because IRF-1 is thought to be one of the major transcription factors involved in iNOS induction and the resulting NO production (12), we evaluated the role of IRF-1 in apoptosis of activated microglial cells by first investigating the inducibility of this protein by inflammatory stimuli in microglial cells. Although IRF-1 messages were strongly induced by both LPS and IFN␥, IRF-1 protein was markedly induced only by IFN␥ (Fig. 1, A and B). A low level induction of IRF-1 protein by LPS was detected by a long exposure of the same blot (data not shown). A similar pattern of IRF-1 message induction was detected in BV-2 mouse microglial cells and mouse primary microglial cells (Fig. 1C). IRF-1-deficient mice were next employed to definitively address the significance of IRF-1 induction in microglial apoptosis. Compared with microglial cells from IRF-1 ϩ/ϩ mice, microglial cells from IRF-1 Ϫ/Ϫ mice were relatively resistant to LPS/IFN␥-induced apoptosis ( Fig.  2A). However, exogenous NO donors such as SNAP (0.5 mM) ( Fig. 2A) and SNP (1 mM) (data not shown) induced the death of microglial cells from IRF-1 ϩ/ϩ and Ϫ/Ϫ mice to a similar extent. Reduction in the cell viability of microglial cells exposed to LPS/IFN␥ was due to apoptotic death, as demonstrated by H33342 nuclear staining, where characteristic apoptotic morphology was induced by LPS/IFN␥ and attenuated in IRF-1 Ϫ/Ϫ mice (Fig. 2C). DNA ladder and subdiploid cells were also detected in microglial cells from IRF-1 ϩ/ϩ mice (9). As we have previously shown that LPS/IFN␥ induce apoptosis of microglial cells mostly through the production of NO (9) and that IRF-1 is known to mediate iNOS induction in mixed glial cells (14), we next sought to determine how microglial NO production is affected by IRF-1 deficiency. Microglial production of NO in response to LPS/IFN␥ was markedly reduced in IRF-1 Ϫ/Ϫ mice (Fig. 2B). These results indicate that LPS/IFN␥-induced IRF-1 may play an important role in the production of NO but not in cytotoxic action of NO.
Induction of Caspase-11 by Inflammatory Stimuli-Although LPS/IFN␥-activated microglial cells underwent apoptosis due to the autocrine NO production, the inhibition of NO production by NMMA, an iNOS inhibitor, did not completely block activation-induced cell death of microglial cells (9), indicating the involvement of other apoptotic mediators or pathways. Because caspase-11 has been shown to be induced by inflammatory stimuli and to play a critical role in the pathological activation of caspase-1 and -3, ultimately leading to cellular apoptosis (23), we next asked whether caspase-11 is involved in activation-induced apoptosis of microglial cells. Among the many caspases tested, the expression of caspase-7, -8, and -11 was induced by LPS/IFN␥ (Fig. 3, A and B). Although the induction of caspase-7 and -8 was modest, the expression of caspase-11 was strongly induced by LPS and IFN␥ in microglial cells, and this induction was not affected by IRF-1 deficiency (Fig. 3C). A similar level of caspase-11 induction was detected by RT-PCR in microglial cells from IRF-1 ϩ/ϩ and Ϫ/Ϫ mice. Induction of caspase-11 at the protein level was demonstrated by Western blot analysis (Fig. 4). LPS/IFN␥ treatment resulted in the enhancement of procaspase-11 protein expression as well as cleavage of the procaspase-11, suggesting LPS/IFN␥-induced activation of caspase-11. The caspase-11 locus encodes two polypeptides of 43 and 38 kDa (21). Approximately 30-and 20-kDa bands appear to be cleavage products derived from 43-kDa species (23).
Role of Caspase Activation in Microglial Apoptosis-To evaluate the significance of caspase activation in microglial apoptosis, the effects of caspase inhibitors on microglial cell death was assessed. A broad spectrum caspase inhibitor (z-VADfmk), caspase-1-specific inhibitor (z-YVAD-fmk), and caspase-3-specific inhibitor (z-DEVD-fmk), but not caspase-8-specific inhibitor (z-IETD-fmk), significantly inhibited LPS/IFN␥-induced apoptosis of BV-2 cells, suggesting that activation-induced apoptosis of microglial cells was caspase-dependent and was mediated through caspase-1 and -3 pathways (Fig. 5A). Because the activation of caspase-11 is required for caspase-1 activation (22), these results also indicated the involvement of caspase-11 in the apoptosis of activated microglial cells. An inhibitor specific for caspase-3 also inhibited SNAP-induced apoptosis, whereas inhibitors specific for caspase-1 and -8 were without effects, suggesting that NO-induced apoptosis was mediated through caspase-3 but not caspase-1 or -8 (Fig. 5B). None of these caspase inhibitors, however, influenced LPS/ IFN␥-induced NO production, indicating that activation of specific sets of caspases was involved in LPS/IFN␥-or NO-induced apoptosis but not in NO production (Fig. 5C). Taken together, the results obtained using peptide caspase inhibitors suggested that LPS/IFN␥ induced microglial apoptosis through caspase-11, -1, and -3, whereas NO-induced apoptosis was directly mediated through caspase-3 activation. To further test for the possible involvement of caspase-11 in LPS/IFN␥-induced microglial apoptosis, we next asked whether caspase-11, which is induced by inflammatory stimuli, is indeed activated during LPS/IFN␥-induced apoptosis of microglial cells. Induction of

FIG. 2. Role of IRF-1 in activationinduced cell death of microglial cells.
Microglial cells from IRF-1 ϩ/ϩ or Ϫ/Ϫ mice were treated with LPS (100 ng/ml) plus IFN␥ (100 units/ml) or SNAP (0. 5 mM) for 24 h, and then cell viability was assessed by MTT assays (A), NO production was measured by Griess reaction (B), or nuclear morphology was determined by Hoechst 33342 staining (C). The viability of cells treated with culture media was set to 100%. Arrows in C indicate apoptotic cells with condensed or fragmented nuclei. Microglial cells from IRF-1 Ϫ/Ϫ mice were partially resistant to LPS/IFN␥-induced cell death but not to SNAP-induced death. LPS/IFN␥-induced production of NO was markedly decreased by IRF-1 deficiency. Results are the mean Ϯ S.E. of three independent experiments. The asterisks indicate statistically significant differences (p Ͻ 0.05).
caspase-11-like activity in LPS/IFN␥-treated microglial cells was detected by in vitro cleavage of the specific fluorogenic substrates (Fig. 5D). The fluorogenic peptide substrate for caspase-11 was based on published results (23). Nevertheless, because Ac-LEHD-AFC could also be a substrate for caspase-9, a firm conclusion on the induction of caspase-11 activity cannot be drawn by in vitro caspase assay alone. LPS/IFN␥-induced activation of caspase-11 was supported by the detection of autoprocessed caspase-11 by Western blot analysis (Fig. 4). Activation of caspase-1 and -3, but not caspase-8, was similarly detected in LPS/IFN␥-treated BV-2 cells by in vitro caspase assays using fluorogenic substrates (Fig. 5D). In contrast, SNAP activated caspase-3 but not caspase-1, -8, or -11 in in vitro caspase assays (data not shown).
Essential Role of Caspase-11 in NO-independent Microglial Apoptosis-Our results in Fig. 5 suggested that activation of caspase-11 is involved in LPS/IFN␥-induced microglial apoptosis but not in NO-induced apoptosis. To further analyze the role of caspase-11 in activation-induced apoptosis of microglial cells, we conducted a transfection of BV-2 cells with wild-type as well as the dominant-negative mutant inhibitor of caspase-11. A dominant-negative mutant of caspase-11 inhibitor significantly inhibited LPS/IFN␥-induced apoptosis but not SNAP-induced apoptosis, further supporting the unique role of caspase-11 in LPS/IFN␥-induced apoptosis (Fig. 6A). Inactivation of caspase-11 by the dominant-negative mutant was verified by co-transfection of the wild-type and mutant forms. Transfection of BV-2 cells with the wild-type caspase-11 alone induced apoptosis (Fig. 6B). Co-transfection of BV-2 cells with the wild-type and the mutant caspase-11 abolished apoptosis-inducing effects of the wild-type caspase-11 (Fig. 6B). Because the treatment of BV-2 cells with LPS/IFN␥ could also induce NO production, the effects of caspase-11 inactivation on the microglial apoptosis was also assessed in the presence of NMMA in order to exclude the effect of NO produced endogenously (Fig. 6C). Inhibition of both NO production and caspase-11 almost completely blocked activationinduced cell death of microglial cells, indicating that the apoptosis of activated microglial cells is mediated through both NO-dependent and -independent pathways and that caspase-11 plays an essential role in NO-independent microglial apoptosis. Collectively, our results suggest that inflammatory stimuli initiate two separate apoptotic pathways in microglial cells, one being dependent on IRF-1 induction and subsequent NO produc-tion, and the other being dependent on caspase-11 induction.

Sensitization of Microglial Cells by Inflammatory Stimuli to NO-induced Apoptosis-Because LPS and IFN␥ have been
shown to up-regulate many different genes in addition to IRF-1 and caspase-11 in microglial cells, we hypothesized that some of those gene products may also affect the apoptosis of activated microglial cells. To test this hypothesis, we sought to determine whether pretreatment of microglial cells with LPS/IFN␥ modulates their sensitivity to cytotoxic action of NO. To rule out the possibility that LPS/IFN␥ pretreatment induces NO production, this experiment was done in the presence of NMMA. LPS/IFN␥ pretreatment under these conditions sensitized microglial cells to nontoxic dose of NO donors, indicating that inflammatory stimuli not only induce the production of cytotoxic mediator (NO) but also render the target cells more sensitive to cytotoxic action of NO through yet unidentified pathways in addition to IRF-1 and caspase-11 induction (Fig.  7). Involvement of other pathways in the increased sensitivity of microglial cells to NO is supported by our results that caspase-11 acted independently of NO (Figs. 5 and 6) and that microglial cells from IRF ϩ/ϩ and Ϫ/Ϫ mice responded similarly to NO ( Fig. 2A).
Role of NF-B in Microglial NO Production and Apoptosis-NF-B has been shown to mediate iNOS induction by LPS in macrophages (33) and to be involved in anti-apoptotic signal transductions in many different cell types (34). Particularly, in ME-180 human cervical cancer cells, NF-B played a cytoprotective role, which was inhibited by inflammatory stimuli such as IFN␥ (19) and IFN␣ (35). Thus, we asked what role NF-B activation plays in microglial apoptosis following inflammatory activation, whether it is proapoptotic or antiapoptotic? LPS induced NF-B activation, which was enhanced by IFN␥ and suppressed by NAC treatment (Fig. 8). LPS/IFN␥-induced NF-B was necessary for NO induction and the subsequent microglial apoptosis, because NF-B-inhibiting agents such as MG-132 and NAC inhibited microglial NO production as well as cell death (Fig. 9, A and B). Transfection of microglial cells with phosphorylation-defective dominant-negative mutant IB (also called the "super-repressor" of NF-B) also inhibited LPS/ IFN␥-induced cell death (Fig. 9C). To determine whether NF-B activation confers cytoprotection against NO, BV-2 cells transfected with the p65 subunit of NF-B were exposed to NO donors. Transfection of p65 did not influence microglial cell death induced by NO (Fig. 9D) or caspase-11 transfection (Fig.  9E). NF-B activation by LPS/IFN␥ treatment or p65 subunit transfection was confirmed by NF-B reporter assays (Fig. 9, F  and G). These results indicate that NF-B activation in microglial cells seems to merely mediate inflammatory NO production, possibly through iNOS induction, but it does not play a cytoprotective role against NO or caspase-11 induction. Induction of iNOS expression by LPS/IFN␥ in BV-2 cells has been demonstrated previously (9). Thus, it is likely that NF-B is involved in NO production but not in apoptosis-inducing actions of NO or caspase-11.

DISCUSSION
Here, we present evidence that IRF-1, NF-B, and caspase-11 play a central role in the apoptosis of mouse microglial cells following inflammatory activation. LPS/IFN␥ treatment induced apoptosis of activated microglial cells through two independent events: they induced production of the autocrine toxic mediator (NO) in microglial cells in an IRF-1/NF-B-dependent manner, and concurrently they initiated NO-independent apoptotic pathways through caspase-11 induction (Fig. 10).
Activation-induced cell death (AICD) is an active process. T and B lymphocytes undergo activation-induced cell death as an autoregulatory mechanism for the body to remove unwanted activated cells after making appropriate use of them (36,37). Compared with lymphocytes, glial cells in the central nervous system have not been not well studied in this respect. Recently, however, the results in our laboratory indicated that both microglial cells (9) and astrocytes (10) might be under the control of a similar regulatory mechanism. In particular, microglial cells that are closely related to macrophages underwent NOdependent apoptosis upon inflammatory activation. In contrast to the AICD of T lymphocytes, where Fas-Fas ligand interaction plays a central role, neither Fas-Fas ligand interaction nor TNF␣ was important in the AICD of microglial cells. Instead, NO produced by activated microglial cells themselves was the  6. Essential role of caspase-11 in microglial apoptosis. A, inactivation of caspase-11 by transfection of dominant-negative caspase-11 inhibitor rendered BV-2 microglial cells resistant to LPS/IFN␥-induced apoptosis as demonstrated by counting blue cells co-expressing lacZ at 24 h after LPS/IFN␥ treatment (LPS, 100 ng/ml; IFN␥, 100 units/ml). The number of blue cells upon transfection with a control vector expressing lacZ only (pCH110) without LPS/IFN␥ treatment was set to 100%. B, transfection of BV-2 cells with wild-type caspase-11 alone induced apoptosis in a dose-dependent manner, and co-transfection of the cells with wild-type and dominant-negative caspase-11 (1.5 g) abolished cytotoxicity of caspase-11, confirming that the mutant caspase-11 C254G indeed acted in a dominant-negative fashion. The pcDNA3 was added to fix the total amounts of plasmids transfected at 2 g. The pCH110 was used as a control vector. C, transfection of BV-2 cells with dominant-negative mutant caspase-11 was also carried out in the presence of NMMA (0.5 mM). Inactivation of caspase-11 by dominant-negative mutant transfection and a simultaneous inhibition of NO production by NMMA almost completely blocked LPS/IFN␥-induced death of the microglial cells. Results are the mean Ϯ S.E. of three independent experiments. The asterisks indicate statistically significant differences (p Ͻ 0.05). major cytotoxic mediator. However, the presence of an NOindependent cytotoxic mechanism was suggested by the incomplete protection by iNOS inhibitor. Here, we present evidence that caspase-11 induction is such an NO-independent apoptotic pathway. Inflammatory stimuli such as LPS and IFN␥ were more than just cellular activators to microglial cells. They not only activated the cells to produce inflammatory mediator such as TNF␣ and NO but also induced autoregulatory apoptosis. The involvement of similar (overlapping) signaling pathways in cellular activation and apoptosis has been suggested by previous studies. Bcl-x L , one of the well known apoptotic inhibitors, has been shown to inhibit both cellular activation and apoptosis in macrophages (38,39). Furthermore, inhibition of Akt, which is known to mediate survival signaling in a number of different cell types, enhanced LPS-induced cellular activation in astrocytes (40). Thus, it appears that inflammatory stimuli induce or activate a specific group of genes and signaling pathways, some of which are commonly involved in both activation and apoptosis. Our results suggested that caspase-11 may be an example of such genes. The induction of caspase-11 may be involved in the production of proinflammatory cytokines such as interleukin-1␤ and interleukin-18 by activating the processing enzyme required (caspase-1). We and others have previously demonstrated the production of interleukin-1␤ (41) and interleukin-18 (42) by mouse microglial cells. In addition to its role in microglial activation, caspase-11 induction also causes auto-activation, which in turn may initiate caspase cascade leading to microglial apoptosis. Therefore, inflammatory stimuli seem to activate microglial cells to produce various inflammatory mediators and, concomitantly, to activate a built-in autoregulatory mechanism by both direct and indirect means (NO-independent and -dependent pathways). This, however, may not be the whole story. Our results that pretreatment of microglia with inflammatory stimuli enhanced their sensitivity to exogenous NO (Fig. 7) suggest the presence of yet other mechanisms whereby inflammatory stimuli affect the microglial AICD process (see below).
In this work, we have focused on the effects of LPS and IFN␥ as representative inflammatory stimuli. Treatment of microglial cells with LPS or IFN␥ alone also induced cell death, NO production (9), and induction of caspase-11 and IRF-1, albeit with a modest difference in the degree of response ( Figs. 1 and 3). This suggests that the AICD-inducing mechanism was not different among the single treatments (LPS or IFN␥ alone) and the cotreatment (LPS plus IFN␥). Astrocytes responded to LPS and IFN␥ in the same way as microglial cells; LPS/IFN␥ induced IRF-1 and caspase-11 as well as NO production in astrocytes, and moreover, TNF␣ also induced caspase-11 expression in astrocytes. 2 Thus, the mechanism of LPS/IFN␥-induced AICD of microglial cells proposed in the current studies may also be applicable to astrocytes and other inflammatory stimuli. According to our model of activation-induced apoptosis of microglial cells, inflammatory signals that activate microglia may also initiate internal death program (Fig. 10). One interesting question that can be raised then is how microglial cells could survive inflammatory activation. It should be kept in mind that microglial cells in vivo are heterogeneous and interact with other glial cells as well as neurons. There is also growing evidence that activated microglial cells proliferate in vivo as one way of replenishment (1). Thus, not all microglial cells may respond to the inflammatory signals in the same fashion. Upon inflammatory activation, individual microglial cells in heterogeneous population may either undergo AICD or return to the resting state via other regulatory mechanisms depending on the specific microenvironment under which they react to the signals. Although many activated microglial cells may be eliminated, some would survive to be deactivated. Whatever the mechanism of down-regulation is, this may be an excellent autoregulatory system for the microglial activation. One can easily imagine pathological situations where this type of autoregulatory mechanism goes wrong. Failure of the autoregulation of "over-activated" microglial cells may result in pathological destruction of bystander cells (neurons and other glial cells) exposed to toxic mediators produced by activated microglia. Recently, up-regulated Bcl-x L expression has been detected in the reactive microglia of patients with neurodegenerative diseases (43). Authors have proposed that a high level of Bcl-x L protein might render microglia more resistant to cytotoxic environment such as areas of neurodegeneration (43). Expression of anti-apoptotic Bcl-2 protein has been also associated with the aged brain and with neurodegenerative diseases (44). The importance of physiological regulation of microglial activation by AICD is supported by these previous reports.
Our current work indicates that inflammatory stimuli play a dual role in AICD of microglial cells. They not only induce the indirect apoptotic pathway via production of NO, but they also induce the direct apoptotic pathway in a manner independent of NO production. Although caspase-11 was required for NOindependent apoptotic pathway, IRF-1 and NF-B were involved in the NO-dependent apoptosis of microglial cells mainly by mediating NO synthesis. We and others have previously reported that STAT1/IRF-1 pathway plays a central role in cellular apoptosis by inflammatory cytokines (15-19, 35), and caspase induction has been suggested as a possible down- Inflammatory signals such as LPS and IFN␥ not only activate microglial cells but also induce their apoptosis. LPS/IFN␥ initiates two apoptotic signaling pathways, one being the production of NO via IRF-1 induction and NF-B activation (left), the other being caspase-11 induction (right). Both NO and caspase-11 activation induce activation of effector caspases resulting in apoptotic death of microglial cells. The third action of inflammatory stimuli may be modulation of apoptotic mediators such as p53 and downstream Bcl-2 family proteins, through which NO action may be mediated (middle). The broken lines (arrows and box) indicate speculations that are not based on the experimental data presented in this study. See the text ("Discussion") for details. stream event following IRF-1 induction in IFN␥-induced apoptosis (17). In microglial cells, however, the absence of IRF-1 did not affect the inducibility of caspase-11 that mediated apoptosis of activated cells. Although we have not tested the inducibility and activation of other caspases in IRF-1-deficient microglial cells, a role of IRF-1 in microglial apoptosis seems to be related to NO production rather than caspase induction or activation (Figs. 2B and 3C). Our results also demonstrated that NF-B plays a role similar to IRF-1 in microglial apoptosis. NF-B activation appeared to be involved in inflammatory NO production; however, it did not influence the apoptotic process initiated by LPS/IFN␥ or NO donors (Fig. 9). It has been reported previously that IRF-1 and NF-B interact in vitro as well as in vivo for the cooperative induction of inflammatory genes (33,45,46). In particular, the two transcription factors interacted to cooperatively enhance iNOS gene transcription in macrophages stimulated by LPS and IFN␥ (33). Cooperative induction of iNOS expression by interaction between the two transcription factors may also occur in LPS/ IFN␥-treated microglial cells, and this interaction between IRF-1 and NF-B may well be responsible for cytotoxic NO production in response to LPS/IFN␥.
Our results here demonstrate that caspase-11 plays a critical role in NO-independent actions of inflammatory stimuli during AICD of microglial cells. Caspase-11 was strongly induced by LPS/IFN␥ and inactivation of the caspase blocked NO-independent apoptosis. Furthermore, over-expression of caspase-11 alone induced microglial apoptosis to a certain extent. These results are in agreement with previous studies in which overexpression of caspase-11 in Rat-1 fibroblasts induced apoptosis (21). The results also suggest that up-regulation of caspase-11 expression alone may be sufficient for the initiation of apoptotic pathways in microglial cells, possibly through auto-activation of caspase-11 followed by cascade activation of downstream executioner caspases. This possibility was also suggested in a previous report (23), where caspase-1 and -3 were proposed as downstream caspases cleaved by auto-activated caspase-11. A similar involvement of caspase-1, -3, and -11 was demonstrated in microglial apoptosis. Although NO donor-induced apoptosis appeared to involve only the caspase-3 pathway, apoptotic pathway initiated by LPS/IFN␥ seemed to involve caspase-1, -3, and -11 (Fig. 5). Thus, inflammatory stimuli initiate two divergent apoptotic pathways in microglial cells, one being NO production followed by its apoptogenic action and the other being caspase-11 induction. However, these two pathways ultimately converge on the activation of caspase-3 for the execution of apoptosis. Future studies using microglial cells from caspase-11-deficient mice will give the definitive answer on the role of the caspase-11 in AICD of microglial cells.
Although IRF-1 and caspase-11 play a major role in activation-induced apoptosis of microglial cells, the involvement of other genes up-regulated by inflammatory stimuli cannot be excluded. In fact, pretreatment of microglial cells with LPS/ IFN␥ enhanced sensitivity of the cells to NO-induced toxicity, indicating the presence of other apoptotic mediators modulated by inflammatory stimuli. IRF-1 and caspase-11 are not likely to play an important role under these conditions for the following reasons. First, because the pretreatment with LPS/IFN␥ was done in the presence of NMMA, a role of endogenous NO and IRF-1, which predominantly mediated its production, is minimal. Second, there was no difference in the sensitivity of microglial cells to NO donors between IRF-1 ϩ/ϩ and Ϫ/Ϫ mice. Third, an exogenous NO induced the apoptosis in microglial cells that constitutively expressed very low levels of caspase-11, indicating that caspase-11 induction was not necessary for NO-induced apoptosis. Finally, NO-induced apoptosis was not blocked by caspase-1-specific inhibitor, indicating that NOinduced apoptosis was not mediated through caspase-11/ caspase-1 pathways. Taken collectively, our results strongly suggest that LPS/IFN␥-inducible gene products other than IRF-1 and caspase-11 may also participate in the AICD of microglial cells. Candidate genes include p53 and Bax, which are known to be involved in NO-induced apoptosis (47) (Fig.  10). Further studies that address this issue are in progress in our laboratory.
In conclusion, microglial cells undergo apoptosis upon inflammatory activation. In the activation-induced apoptosis of microglial cells, inflammatory stimuli play a dual role by initiating two separate apoptotic pathways. They induce indirect apoptotic pathway through the production of NO. They also initiate a direct apoptotic pathway via caspase-11 induction in a manner independent of NO. Eventually, however, the two pathways appear to share a common executioner caspase (caspase-3) for the ultimate demise of the cells.