Abstract
The transcription factor nuclear factor κB (NF-κB) is well known for its antiapoptotic action. However, in some disorders, such as cerebral ischemia, a proapoptotic function of NF-κB has been demonstrated. To analyze which subunit of NF-κB is functional in cerebral ischemia, we induced focal cerebral ischemia in mice with a germline deletion of the p52 or c-Rel gene or with a conditional deletion of RelA in the brain. Only RelA deficiency reduced infarct size. Interestingly, expression of the proapoptotic BH3 (Bcl-2 homology domain 3)-only genes Bim and Noxa in cerebral ischemia depended on RelA and the upstream kinase IKK (IκB kinase). RelA stimulated Bim and Noxa gene transcription in primary cortical neurons and bound to the promoter of both genes. Thus, the deleterious function in cerebral ischemia is specific for the NF-κB subunit RelA and may be mediated through Bim and Noxa.
Introduction
Members of the Bcl-2 family are sentinels of neuronal cell survival. Several studies have documented the role of Bcl-2 proteins in neuronal apoptosis after stroke (Graham et al., 2000; Cao et al., 2002). Bcl-2 proteins share four Bcl-2 homology domains, BH1 to BH4 (Strasser, 2005). Only a single BH3 domain is characteristic of a large subgroup of proapoptotic Bcl-2 family members, which includes Bim and Noxa. At the mitochondrial membrane, some BH3-only proteins directly activate the proapoptotic proteins Bax and Bak to release cytochrome c and other death signals that trigger the caspase cascade. The antiapoptotic Bcl-2 and Bcl-XL block Bax and Bak function. However, BH3-only proteins reverse this block and release Bak and Bax to trigger apoptosis. Thus, the ratio of proapoptotic and antiapoptotic Bcl-2 family members determines the fate of the cell.
Gene transcription regulates the ratio of proapoptotic and antiapoptotic Bcl-2 family members. DNA damage induces apoptosis by stimulating expression of the BH3-only genes Noxa and Puma through p53 (Oda et al., 2000; Yu and Zhang, 2003). Bim is upregulated by the proapoptotic transcription factors FOXO and E2F1 and by c-Jun N-terminal protein kinase (JNK)/c-Jun signaling (Gilley et al., 2003; Kuan et al., 2003; Sunters et al., 2003; Hershko and Ginsberg, 2004). However, antiapoptotic family members are also regulated at the transcriptional level. The induction of Bcl-2 and Bcl-XL by the antiapoptotic transcription factor nuclear factor κB (NF-κB) (Tamatani et al., 1999) is a well known phenomenon.
NF-κB consists of the five subunits p50, p52, RelA, c-Rel, and RelB. Surprisingly, in some paradigms, NF-κB promotes apoptosis (Kaltschmidt et al., 2000, 2002; Pizzi et al., 2002). In a mouse model of stroke, NF-κB was associated with neurodegeneration because germline deletion of the gene for the NF-κB subunit p50 reduced the infarct size (Schneider et al., 1999; Nurmi et al., 2004). Furthermore, expression of an NF-κB superrepressor was neuroprotective (Xu et al., 2002; Zhang et al., 2005), and inhibition or deletion of the upstream kinase IκB kinase (IKK) profoundly reduced the infarct size (Herrmann et al., 2005). The mechanisms that determine whether NF-κB is proapoptotic or antiapoptotic are not known. Possibly, the five subunits exert distinct effects on cell survival. Indeed, functional diversity is suggested by the unique phenotypes of knock-out mouse lines (Li and Verma, 2002). Recent evidence has shown that RelA exerts a proapoptotic and c-Rel an antiapoptotic effect in neurons (Pizzi et al., 2002, 2005).
To investigate the function of NF-κB subunits other than p50 in vivo, we induced cerebral ischemia in mice with a germline deficiency of p52 and c-Rel or with a conditional deletion of RelA in the brain. Deficiency of RelA but not of p52 or c-Rel reduced the infarct size. Moreover, RelA controlled the induction of the BH3-only genes Bim and Noxa in cerebral ischemia, providing a possible explanation for its deleterious effect. Our data suggest that, in addition to its known effect on antiapoptotic members, NF-κB may fine-tune cell survival by regulating the transcription of proapoptotic members of the Bcl-2 family.
Materials and Methods
Transgenic mice.
RelAflox/flox mice will be described elsewhere (R. M. Schmid, unpublished observations). LoxP sites were introduced between exons 6 and 7 and between exons 10 and 11 of the RelA gene. For controls, we used littermate RelAflox/flox mice that were negative for the nestin-Cre balancer gene (Betz et al., 1996). After focal cerebral ischemia, nestin-Cre balancer mice and wild-type littermates had the same infarct size (data not shown). For Southern blot analysis, DNA isolated from the tails and various tissues of mice was digested with BglII and separated in 1% agarose gels. For hybridization of DNA, a 431 bp genomic probe was used, and bands were detected by PhosphorImager (GE Healthcare, Little Chalfont, UK). c-Rel−/− mice (Liou et al., 1999) were backcrossed for seven generations on a C57BL/6 background and p52−/− mice (Paxian et al., 2002) for >10 generations. Therefore, we used C57BL/6 mice as controls.
Cell culture.
Cortical neurons were prepared from embryonic day 16 (E16) mice. For transfection, cells were plated on 24-well plates precoated with poly-d-lysine (50 μg/ml) at a density of 200,000 cells per well. For RNA preparation, 2,000,000 cells per well were plated on six-well plates. Cells were incubated in Neurobasal medium (Invitrogen, Karlsruhe, Germany) supplemented with B27 (Invitrogen), l-glutamine (0.5 mm), penicillin (100 IU/ml), and streptomycin (100 μg/ml). In these cultures, >95% of cells were positive for the neuronal marker NeuN (neuronal-specific nuclear protein). Cells were used after 10 d in vitro.
Models of cerebral ischemia.
As an in vivo model of permanent focal cerebral ischemia, a distal middle cerebral artery occlusion (MCAO) was performed. All mice were male and were anesthetized at the age of 3–4 months by intraperitoneal injection of 150 μl of 2.5% Avertin (tribromoethanol) per 10 g of body weight. A skin incision was made between the ear and the orbit on the left side. The parotid gland and the temporal muscle were removed by electrical coagulation. The stem of the MCA was exposed through a burr hole and occluded by microbipolar coagulation (Erbe, Tübingen, Germany). Surgery was performed under a microscope (Hund, Wetzlar, Germany). A body temperature of 37°C was maintained in the mice by using a heating pad. After 48 h, mice were deeply reanesthetized with Avertin and perfused intracardially with Ringer's solution.
The procedure for infarct measurement on cryosections and correction for cerebral edema has been described previously (Herrmann et al., 2003). In a separate cohort of animals, the femoral artery was cannulated to measure arterial blood gases and mean arterial blood pressure. Arterial blood gases, glucose, and hemoglobin were measured 15 min before and 15 min into MCAO in a blood sample of 100 μl. For laser Doppler measurements, the probe (P415-205; Perimed, Piscataway, NJ) was placed 3 mm lateral and 6 mm posterior to the bregma. Relative perfusion units were determined (Periflux 4001; Perimed).
Drugs were administered 10 min before MCAO. Vehicle (0.9% saline) or BMS-345541 solution (both 2 μl) was injected into the lateral ventricle, using a 10 μl Hamilton syringe, 0.9 mm lateral, 0.1 mm posterior, and 3.1 mm deep relative to the bregma.
Oxygen glucose deprivation (OGD) was used as an in vitro model of ischemia. For OGD experiments, primary cortical neurons that had been in culture for 10 d were transferred into serum-free medium containing 5 mm 2-deoxy-d-glucose (Merck, Darmstadt, Germany) for 1 h. Then, the cells were placed in an anaerobic chamber that was flushed for 10 min with a mix of 95% N2 and 5% CO2. After incubation for the indicated times, cells were removed from the anaerobic chamber and incubated under normal conditions for another 24 h. The control group was cultured in parallel but did not receive 2-deoxy-d-glucose and was not flushed with N2/CO2. Then, RNA was extracted.
NF-κB DNA binding assay.
Mice were reanesthetized and perfused with Ringer's solution 4 h after MCAO. The ischemic core and periphery and contralateral cortices (each 18 mm2) were quickly dissected with a sample corer (Fine Science Tools, Foster City, CA). Nuclear fractions were isolated using commercially available reagents (Nuclear Extract Kit; Active Motif, Rixensart, Belgium). DNA binding activity of NF-κB subunits p50, p65, c-Rel, p52, and RelB was monitored using a commercially available ELISA-based assay (TransFactor; BD Biosciences, Mountain View, CA) (Pizzi et al., 2005). Briefly, nuclear samples (3 μg) were incubated in an ELISA plate coated with oligonucleotides containing an NF-κB consensus regulatory element sequence. The wells were washed and exposed to a primary antibody specific for each subunit of NF-κB p50, p65, and c-Rel included in the assay kit. To detect p52 and RelB, wells were exposed to anti-RelB antibody (1:200, sc-226X; Santa Cruz Biotechnology, Heidelberg, Germany) and to anti-p52 (1:200; sc-848X; Santa Cruz Biotechnology). Binding of the primary antibody to protein was detected through a chromogenic reaction involving the enzymatic breakdown of 3,3′,5,5′-tetramethylbenzidine via a horseradish peroxidase (HRP)-conjugated secondary antibody. Reactions were quantified spectrophotometrically at a wavelength of 655 nm. For each run, a series of positive and negative controls was performed to ensure the detection specificity of NF-κB DNA binding activity.
Immunoblots.
Nuclear extracts (used for ELISA assay) or brain lysates from control and RelACNSKO mice were resolved by 10% or 15% SDS-PAGE. Prestained molecular size markers were used as molecular mass standards. The gels were electroblotted onto Hybond ECL nitrocellulose membrane (GE Healthcare). Immunodetection of the protein of interest was performed with the following antibodies: c-Rel, RelA, RelB, p52, extracellular signal-regulated protein kinase 2 (Erk2; Santa Cruz Biotechnology), p50 (Abcam, Paris, France), Bim (Calbiochem, Schwalbach, Germany), phospho-IKK1 (Ser180)/IKK2 (Ser181; Cell Signaling Technology, Beverly, MA), and actin (Sigma, Taufkirchen, Germany). The membranes were then probed with HRP-conjugated secondary antibody. For detection, we used enhanced chemiluminescence (ECL; GE Healthcare). For RelA and p50 detection in brain lysates of RelACNSKO or control mice, the ECL Advance Western Blotting Detection Kit (GE Healthcare) was used. Quantitative analysis was performed with the NIH ImageJ program.
Real-time reverse transcription-PCR.
Mice were reanesthetized and perfused with Ringer's solution at 6, 15, or 24 h after MCAO. The ischemic and contralateral cortices were quickly dissected and frozen on dry ice. Tissues were stored at −80°C. RNA from cortex or cultured cells was extracted with peqGOLD RNAPure (PEQLAB, Erlangen, Germany), according to the manufacturer's instructions. RNA (10 μg, cortex; 7.5 μg, cells) was transcribed with Moloney murine leukemia virus reverse transcriptase and random hexamers. Primers used for PCR amplification are described in Table 1.
Primers used for real-time RT-PCR
PCR was performed according to the following protocol: 10 min at 95°C, 15 s at 95°C, and 1 min at 60°C (40 cycles). Amplification was quantified with the Gene Amp 5700 sequence detector and the SYBR Green kit (PE Diagnostik, Weiterstadt, Germany). A linear concentration–amplification curve was established by diluting pooled samples. Quantified results for individual cDNAs were normalized to cyclophilin. With this procedure, we can quantify results relative to a control group. The purity of the amplified products was checked by the dissociation curve and gel electrophoresis of selected samples.
Immunohistochemistry and terminal deoxynucleotide transferase-mediated biotinylated UTP nick end labeling staining.
For immunohistochemical detection of active RelA, neurofilament-200, Bim, and Noxa, permanent MCAO was performed in wild-type mice. Two hours and 24 h after MCAO, mice were reanesthetized and perfused with 4% paraformaldehyde (PFA). Brains were incubated overnight in 4% PFA and then embedded in paraffin. Sections (1 μm) were incubated overnight at 4°C with the following antibodies: active RelA (Millipore, Billerica, MA), neurofilament-200 (Sigma), Bim (Stressgen Biotechnologies, Victoria, British Columbia, Canada), and Noxa (Imgenex, San Diego, CA). For detection, secondary fluorescein-linked anti-rabbit IgG antibody (Vector Laboratories, Burlingame, CA) and rhodamine-conjugated anti-mouse IgG antibody (Jackson ImmunoResearch Laboratories, West Grove, PA) were used. To exclude the possibility of false labeling, we omitted primary and secondary antibodies.
For terminal deoxynucleotide transferase-mediated biotinylated UTP nick end labeling (TUNEL) staining, sections prestained with anti-Bim or anti-Noxa antibody were incubated with 50 μl of TUNEL reaction mix (In Situ Cell Detection Kit, TMR red; Roche Diagnostics, Mannheim, Germany) for 1 h at 42°C. After washing three times, slides were mounted with medium containing 4′,6′-diamidino-2-phenylindole dihydrochloride (DAPI; Vectashield; Vector Laboratories).
Site-specific mutation and transfection.
The following plasmids have been described previously: Noxa–luciferase (−183/+146) (Oda et al., 2000) and Bim–luciferase (−3600/+96) (Bouillet et al., 2001). NF-κB binding sites were mutated in Noxa– and Bim–luciferase plasmids, using the QuikChange II Site-Directed Mutagenesis Kit (Stratagene, La Jolla, CA) as recommended. To generate Noxa–mutated binding sequence of the κB promoter (mκB)–luciferase and Bim–mκB–luciferase plasmids, the following primers were used: CGT CAC ATG ACG TCA CCG AAG AAG TCA CGA TAA AAT GCG AGA GCC (Noxa–mκB1), GCA GCC CGA GTC TTG GAA GAG CCC CAG AGC CCA GAT TGG (Noxa–mκB2), and GCC GTG GGG GGT AGG GCT TAT CTT CCG GCT TGC (Bim–mκB). Mutated plasmids were confirmed by sequencing.
After 10 d in vitro, cortical neurons were transfected using Lipofectamine 2000 (Invitrogen), 0.2 μg per well of the Noxa or Bim plasmid, and 0.8 μg per well of the RcCMV-p65 or pBluescript plasmids as described previously (Potrovita et al., 2004). To normalize the transfection efficiency, 0.02 μg per well Renilla luciferase (phRLTK) control plasmid (Promega, Mannheim, Germany) was used. After 52 h, cells were harvested, and firefly and Renilla luciferase were measured using Dual Luciferase Reporter Assay (Promega).
Electrophoretic mobility shift assay.
Cos cells on 9 cm plates were transfected with RcCMV–p65 for 24 h. Cell extracts were prepared as described previously (Schreiber et al., 1989). Protein and a double-stranded oligonucleotide from the κB-enhancer (GAT CCA GAG GGG ACT TTC CGA GA) labeled with 32P by a Klenow fill-in reaction were incubated at 4°C in the following buffer: 1 mm EDTA, 0.5 μg/μl BSA, 10 mm Tris-HCl, pH 7.5, 50 mm NaCl, 1 mm DTT, and 5% glycerol (total volume, 20 μl). For competition, the following double-stranded oligonucleotides were used: mκB, GAT CCA GAC CAT GGT ATC CGA GA; Bim, GAT CGG GGG TGG GGC TTA CCT TCC GGC; mBim, GAT CGG GGG TAG GGC TTA TCT TCC GGC; NoxaκB2, GAT CGT CTT GGG GGA GCC CCA GAG CCC; and mNoxaκB2, GAT CGT CTT GGA AGA GCC CCA GAG CCC. Protein–DNA complexes were resolved on a 6% nondenaturing polyacrylamide gel at 280 V.
Results
To find out which subunits of NF-κB are activated in cerebral ischemia, we performed an ELISA-based analysis of DNA binding of the NF-κB subunits. All five subunits were activated in the ischemic hemisphere, although the relative activation between the core and the periphery of the ischemic area differed between the subunits (Fig. 1A). Similar results were obtained by immunoblotting of the same nuclear extracts (Fig. 1B, left). Immunohistochemistry with an antibody that specifically recognizes active RelA (Kaltschmidt et al., 1995) confirmed that RelA is activated in neurofilament-200-positive neurons (Fig. 1B, right) (Herrmann et al., 2005).
Activation and function of NF-κB subunits in cerebral ischemia. A, All five subunits of NF-κB were activated after 4 h of MCAO. DNA binding in the ischemic and the corresponding contralateral cortex was measured with an ELISA-based assay. Values are means ± SEM (n = 6) and expressed as a percentage of the respective contralateral control. Core and periphery were dissected according to a standardized protocol (see Materials and Methods). B, Increased nuclear concentrations of the five subunits of NF-κB after cerebral ischemia. Immunoblots of nuclear extracts of the ischemic core (IC), ischemic periphery (IP), and the contralateral cortex (CC) after 4 h of MCAO are shown (left). Nuclear translocation of active RelA (green) was detected in neurofilament-200-positive neurons (NF-200; red) 2 h after permanent MCAO by immunohistochemistry (right). C, Genetic strategy to generate RelACNSKO mice. The nestin-Cre balancer gene was introduced into mice that were homozygous for the floxed RelA allele. Exons of the RelA gene are depicted as gray boxes. D, RelACNSKO mice were mosaics for the deleted RelA gene in several organs. However, in cerebellum, forebrain, and eye, the mutated allele predominates, as shown by a Southern blot. ko, Knock-out; wt, wild type. E, RelA protein concentration was significantly reduced in the noninfarcted hemisphere of RelACNSKO mice compared with controls (p < 0.05; n = 13–15; two-tailed t test). Representative immunoblots are shown. F, Conditional deletion of RelA in the brain reduced infarct volume, whereas deletion of p52 or c-Rel had no effect on the infarct size. The infarct volume was measured 48 h after permanent MCAO. *p < 0.05 (two-tailed t test). ns, Not significant. Values are means ± SEM. The sample size is noted in the boxes.
Previous work has shown that the subunit p50 contributes to ischemic brain damage (Schneider et al., 1999; Nurmi et al., 2004). Because germline deletion of RelA causes embryonic death (Beg et al., 1995), we used a conditional knock-out approach to delete RelA in neurons and glial cells. We deleted RelA in neural cells (RelACNSKO) by crossing mice carrying floxed RelA alleles with transgenic mice expressing the Cre recombinase under control of the nestin promoter (Betz et al., 1996) (Fig. 1C). RelACNSKO mice were healthy and viable. The deletion in the RelACNSKO mice did not affect the expression of the adjacent gene on the 3′ side of RelA, sipa, as evaluated by reverse transcription (RT)-PCR (sipa1 mRNA, 13.8 ± 3.0 relative units in control mice vs 14.1 ± 8.1 relative units in RelACNSKO mice; n = 3–4). The nestin-Cre transgenic mouse line has been demonstrated to induce a mosaic phenotype in several organs (Betz et al., 1996). Indeed, Southern blot analysis showed that the RelA gene was partially deleted in several tissues of RelACNSKO mice (data not shown). However, this partial deletion had no effect on various physiological parameters, which may affect brain damage in MCAO (Table 2). In RelACNSKO mice, a marked deletion of the RelA gene was observed in forebrain, cerebellum, and the eye (Fig. 1D). Because the nestin-Cre transgene targets mainly neurons and astrocytes but not endothelial or microglial cells, a complete loss of RelA in brain extracts is not expected. Immunoblots of RelA in extracts of the nonischemic right hemisphere demonstrated a significant reduction of RelA protein levels by 35% (p < 0.05; n = 13–15) (Fig. 1E). In the same extracts, there was a nonsignificant trend toward lower protein concentrations of the NF-κB subunit p50 (Fig. 1E), which may be attributable to the regulation of p50 expression by NF-κB (Ten et al., 1992). The infarct size after 48 h of MCAO was significantly reduced in RelACNSKO mice, showing that RelA promotes ischemic damage (Fig. 1F, right). To investigate the functional consequences of p52 activation, we used mice with a germline deficiency of the subunit p52 (Paxian et al., 2002). These mice showed no statistically significant difference in infarct volumes compared with wild-type controls (Fig. 1F, left). The role of c-Rel was studied in mice harboring a germline deletion of the c-Rel gene (Liou et al., 1999). Germline deletion of c-Rel had no effect on the infarct volume 48 h after onset of MCAO either (Fig. 1F, middle). Because of poor health, RelB-deficient mice were not included in this study (Weih et al., 1995).
Physiological parameters 15 min before and 15 min after MCAO in RelACNSKO mice and littermate controls
To search for genes that may mediate the neurodegenerative function of RelA, we measured the expression of various BH3-only genes by real-time RT-PCR 24 h after permanent MCAO in mice. Of nine BH3-only genes, Bid, Noxa, Bik, and Bim were significantly upregulated after MCAO (Fig. 2A). There was no difference between the induction of Bim and Noxa mRNA in the core and periphery of the ischemia (Fig. 2B). Alternative splicing generates three isoforms of Bim (BimEL, 23 kDa; BimL, 20 kDa; and BimS, 18 kDa). Our real-time RT-PCR is specific for BimS, which has the highest apoptotic potency of the isoforms. Immunoblots supported the induction of BimS and demonstrated that the other isoforms were also enhanced by MCAO (Fig. 2C). Moreover, immunohistochemistry confirmed the upregulation of Bim and Noxa 24 h after permanent MCAO. Bim and Noxa staining was partially colocalized with TUNEL, suggesting that Bim and Noxa are implicated in apoptotic cell death after cerebral ischemia (Fig. 2D).
Expression of BH3-only genes after MCAO. A, mRNA accumulation in the ischemic cortex after 24 h of MCAO was quantified by real-time RT-PCR. Bottom, Values are means ± SEM (n = 12) and expressed relative to the contralateral nonischemic side. *p < 0.02; **p < 0.003 (two-tailed t test). Gel electrophoresis verified the correct size of amplified PCR products (top). B, mRNA accumulation of Bim and Noxa was not different between core and periphery of the infarcted area after 24 h of MCAO. ns, Not significant. C, Immunoblotting showed upregulation of Bim 24 h after MCAO. The three bands correspond to BimEL (23 kDa), BimL (20 kDa), and BimS (18 kDa). D, Immunohistochemistry (IHC) demonstrated upregulation of Bim and Noxa after cerebral ischemia. Low levels of Noxa and Bim were detected in the cortex of sham-operated mice (top). Bim or Noxa (green) was expressed by TUNEL-positive cells (red) 24 h after MCAO (bottom). *, Double-positive cells.
In cerebral ischemia, apoptotic cell death mainly affects neurons. To approach the question of whether BH3-only genes are induced in neurons, we exposed primary cortical mouse neurons to camptothecin, a topoisomerase I blocker that models DNA damage in cerebral ischemia. Camptothecin stimulated the expression of Noxa and Bim at 16 and 24 h of exposure, whereas the expression of Bid and Bik was only enhanced after 24 h of stimulation (Fig. 3A). A similar picture emerged after exposure of cortical neurons to OGD, an in vitro model of cerebral ischemia. Bim, Noxa, and Bik were significantly induced at the mRNA level by 6 h of OGD (Fig. 3B). However, Bid expression was not stimulated by OGD in cortical neurons. Thus, Bik, Noxa, and Bim are induced by in vivo and in vitro models of cerebral ischemia.
Induction of BH3-only genes in primary cortical neurons in in vitro models of cerebral ischemia. A, Camptothecin (CPT; 10 μm) exposure enhanced mRNA accumulation of Bid, Bik, Noxa, and Bim in a time-dependent manner. Values are means ± SEM (n = 7–8) and expressed relative to the unstimulated control group. *p < 0.006 [ANOVA; least significant difference (LSD) post hoc]. B, OGD stimulates mRNA accumulation of Bik, Noxa, and Bim in a time-dependent manner. Values are means ± SEM (n = 5–6) and expressed relative to the group not exposed to OGD. *p < 0.01 (ANOVA; LSD post hoc).
To investigate the effect of RelA on the induction of BH3-only genes, we measured mRNA levels of the four BH3-only genes that were found to be upregulated in wild-type mice. The induction of Bim and Noxa in wild-type mice was abolished in RelACNSKO mice, whereas the upregulation of Bik persisted in RelACNSKO mice (Fig. 4A). Previous work has shown that NF-κB is activated in cerebral ischemia by the IKK complex. BMS-345541, a specific IKK inhibitor (Burke et al., 2003), protected against ischemic brain damage (Herrmann et al., 2005). IKK activity is reflected by the phosphorylation state of IKK. Intracerebroventricular injection of BMS-345541 reduced IKK phosphorylation 4.5 h after MCAO in the periphery of the ischemic area, confirming that BMS-345541 is an effective inhibitor of IKK in vivo (Fig. 4B). To explore the function of IKK in the induction of Bim and Noxa, we injected mice intracerebroventricularly with BMS-345541 or vehicle and measured Bim and Noxa mRNA at three time points after MCAO. Bim and Noxa were first upregulated after 6 h of MCAO; at this time, BMS-345541 inhibited the induction. However, BMS-345541 lost its effect after 24 h of MCAO (Fig. 4C). These data suggest that Bim and Noxa are target genes of IKK and RelA.
Induction of Bim and Noxa by MCAO depends on the IKK/NF-κB signaling pathway. A, In RelACNSKO mice, Bim and Noxa mRNA accumulation was not enhanced 24 h after MCAO. *p < 0.004 compared with sham group. +p < 0.02 compared with wild-type mice with MCAO. n = 3–4 [ANOVA; least significant difference (LSD) post hoc]. B, Phospho-IKK1/2 in the periphery (P) of the ischemic cortex was decreased in mice injected with 12 μg of BMS-345541 intracerebroventricularly as shown by immunoblot. In the core (C) of the ischemia, phospho-IKK1/2 was not affected by MCAO. Mice were analyzed 4.5 h after permanent MCAO. Erk2 was used as control for protein loading. ns, Nonspecific band. C, The IKK inhibitor BMS-345541 transiently inhibited Bim and Noxa mRNA accumulation after MCAO. BMS-345541 was injected intracerebroventricularly immediately before MCAO. *p < 0.05 compared with vehicle-treated mice with MCAO (ANOVA; LSD post hoc; n = 4–6). ns, Not significant. Values are means ± SEM.
To further evaluate the possibility that NF-κB stimulates gene transcription of Noxa and Bim in neurons after cerebral ischemia, we transfected primary cortical neurons with reporter fusion genes, in which the 5′-flanking sequences of the mouse Bim (−3600/+96) or the mouse Noxa gene (−183/+146) direct luciferase expression. Indeed, cotransfection of an expression plasmid for RelA stimulated Bim transcription approximately fourfold (Fig. 5A) and Noxa transcription approximately twofold (Fig. 5B). A search for putative NF-κB binding sites in the 5′-flanking sequence of the Bim and Noxa genes using the TRANSFAC database found one potential binding site in the promoter of the mouse Bim gene (−62/−53) and two close to the transcription start site of the Noxa gene (−49/−40; +87/+96), with the second site being almost palindromic. Mutation of the NF-κB binding site in the Bim promoter significantly reduced luciferase expression (Fig. 5A). The residual effect of RelA after mutation of the NF-κB site suggests that the mutation may not destroy RelA binding completely or there may be other sites in the 5′-flanking sequence. Mutation of the first NF-κB binding site (κB1) in the Noxa promoter had no effect, but mutation of the second binding site (κB2) blocked the stimulation by RelA (Fig. 5B). To further test whether RelA binds to the identified response elements in the Bim and Noxa promoter region, we performed electrophoretic shift assays. The binding of RelA expressed in Cos cells to a consensus NF-κB binding site was inhibited by wild-type Bim and Noxa sites. The mutations that interfered with RelA function also diminished competition of RelA DNA binding (Fig. 5C). These data support the notion that RelA binds to the promoter of Bim and Noxa and stimulates their transcription.
RelA binds to the Bim and Noxa promoter to stimulate transcription. A, Transient cotransfection of primary cortical neurons with an expression vector for RelA stimulated luciferase expression controlled by 3600 bp of the Bim promoter. Mutation of the NF-κB site in the promoter significantly reduced luciferase expression. Values are means ± SEM. +p < 0.0001 [n = 8; ANOVA; least significant difference (LSD) post hoc]. B, RelA overexpression stimulated luciferase expression controlled by the Noxa promoter. Mutation of the NF-κB binding site 2 (κB2) but not of the binding site 1 (κB1) blocked luciferase expression. *p < 0.01 (n = 8; ANOVA; LSD post hoc). The location of NF-κB binding sites relative to other response elements is depicted at the top. wt, Wild type. C, Electrophoretic mobility shift assays showed that RelA binding to a consensus NF-κB binding site was blocked by an excess of the consensus site (κB), the Noxa κB2, and the Bim NF-κB binding sites, but not by the mutated sequences. RelA was expressed in Cos cells.
Discussion
Several investigators have reported on the activation of NF-κB in cerebral ischemia (Mattson and Camandola, 2001). Here, we specify this finding, showing that all five subunits are activated. In view of this broad activation of NF-κB subunits, it was interesting to compare their effects on ischemic brain damage. Our data support the concept that NF-κB subunits differ in their functional significance. In our paradigm of focal cerebral ischemia, conditional deletion of RelA in the brain reduced the infarct size, whereas a germline knock-out of the genes for the subunits c-Rel and p52 had no effect. Notably, our data do not exclude an effect of c-Rel and p52 on parameters other than infarct size or a potential compensation by redundant mechanisms. Together with the previous finding that p50−/− mice have smaller infarcts (Schneider et al., 1999; Nurmi et al., 2004), the reduced infarct size in RelACNSKO mice suggests that p50/RelA heterodimers or p50/p50 and RelA/RelA homodimers promote ischemic brain damage. A slight reduction of p50 levels in brains of RelACNSKO mice (Fig. 1E) might have contributed to protection in cerebral ischemia. The predominant role of p50 and RelA is in accordance with their abundance in the brain and in other tissues (Hoffmann et al., 2003; Kaltschmidt et al., 2005). In the canonical NF-κB pathway, p50 and RelA activation is triggered by the upstream IKK kinase complex with the subunits IKK1, IKK2, and NEMO (NF-κB essential modulator). Indeed, IKK2 deletion or inhibition was shown to ameliorate ischemic brain damage (Herrmann et al., 2005). Thus, interference with the NF-κB pathway at various levels reveals a deleterious role of NF-κB in cerebral ischemia, which is in apparent contrast to what has been found in other paradigms. There is ample evidence that NF-κB inhibits apoptosis in tumor cells, hepatocytes, and many other cell types, including neurons (Pizzi et al., 2002; Fridmacher et al., 2003; Tarabin and Schwaninger, 2004). In stroke models, however, protective effects of NF-κB, which may be mediated by c-Rel (Pizzi et al., 2002), are overrun by the deleterious action of p50 and RelA. In accordance with this concept, deletion of c-Rel had no effect on infarct size (Fig. 1F). In contrast, the protective role of NF-κB seems to prevail in short cerebral ischemia, which leads to ischemic preconditioning (Blondeau et al., 2001).
NF-κB promotes cell survival through induction of Bcl-2 and Bcl-XL (Lee et al., 1999; Tamatani et al., 1999; Mattson and Camandola, 2001). Previous studies reported either an increase or a decrease in Bcl-2 and Bcl-XL expression in cerebral ischemia (Krajewski et al., 1995; Gillardon et al., 1996; Chen et al., 1997). In our stroke model, there was no change in Bcl-2 or Bcl-XL expression (data not shown). However, we found a reproducible upregulation of the proapoptotic BH3-only genes Bik, Bim, and Noxa by cerebral ischemia. The induction of Bim by cerebral ischemia has been reported previously (Shibata et al., 2002; Gao et al., 2005).
Bim is localized in neurons (Shibata et al., 2002). Indeed, in primary cortical neurons, in vitro OGD induced Bim and Noxa expression. Interestingly, we found that conditional deletion of RelA in neural cells or pharmacological IKK inhibition blocked Bim and Noxa expression but had no effect on Bik induction by cerebral ischemia. RelA seems to directly regulate transcription of the Bim and Noxa gene by binding to the promoter because mutation of the NF-κB motif near the transcription start site of the Bim and Noxa gene both interfered with the induction by RelA and reduced the binding of RelA to the DNA. The sequence of the NF-κB site in the 5′-untranslated region of the Noxa gene (GGGGAGCCCC) matches the consensus binding site GGGRNNYYCC of NF-κB exactly, where R represents purine, Y represents pyrimidine, and N represents any base (Ghosh et al., 1998). The sequence in the Bim promoter (GGGGCTTACC) deviates by a purine in position 8 from the consensus sequence. However, the same variation occurs in the NF-κB motif of the GM-CSF (granulocyte/macrophage colony-stimulating factor) gene, which is a high-affinity binding site (Schreck and Baeuerle, 1990).
An indirect mechanism through which NF-κB is able to regulate Noxa gene transcription was suggested by Aleyasin et al. (2004). These authors found that, when DNA had been damaged, NF-κB stimulated Noxa and Puma gene transcription through induction of p53 expression. Focal cerebral ischemia only stimulated Noxa but not Puma mRNA accumulation, suggesting an underlying mechanism other than p53. In addition to the direct induction of Noxa and Bim by the IKK/NF-κB pathway that we propose here, indirect interactions with other signaling cascades known to be involved in Bim expression are possible. IKK has been shown to inhibit the transcription factor FOXO3a (Hu et al., 2004), which stimulates Bim gene transcription (Gilley et al., 2003; Sunters et al., 2003). Likewise, NF-κB inhibits the effect of JNK on gene transcription (De Smaele et al., 2001; Tang et al., 2001; Park et al., 2004). JNK has been shown to control Bim transcription in cerebral ischemia (Kuan et al., 2003; Okuno et al., 2004). Thus, IKK/NF-κB cross talk with FOXO3a or JNK may counteract the induction of Bim transcription caused by the direct stimulation. However, the role of such interactions must be further investigated.
Of note, NF-κB directs gene transcription of Bcl-2 family members with opposite effects on mitochondrial permeability that are either proapoptotic (Bax, Bim, and Noxa) or antiapoptotic (Bcl-2 and Bcl-XL). The relative control of these genes may be the switch that determines the enigmatic role of NF-κB for cell survival. It is possible that the composition of the NF-κB dimer or the activation pattern of other transcription factors determines whether the induction of proapoptotic or antiapoptotic Bcl-2 family members prevails.
Footnotes
-
We thank Philippe Bouillet (The Walter and Eliza Hall Institute of Medical Research, Melbourne, Australia) for the Bim–luciferase construct and Tadatsugu Taniguchi (University of Tokyo, Tokyo, Japan) for the Noxa–luciferase construct. BMS-345541 was kindly provided by James Burke (Bristol-Myers Squibb, Princeton, NJ). We received a kind contribution of nestin-cre mice from Soizic Bourteele and Oliver Planz (Friedrich-Loeffler-Institute, Tübingen, Germany).
- Correspondence should be addressed to Markus Schwaninger, Department of Neurology, University of Heidelberg, Im Neuenheimer Feld 400, 69120 Heidelberg, Germany. markus.schwaninger{at}med.uni-heidelberg.de
References
- Aleyasin et al., 2004.↵
- Beg et al., 1995.↵
- Betz et al., 1996.↵
- Blondeau et al., 2001.↵
- Bouillet et al., 2001.↵
- Burke et al., 2003.↵
- Cao et al., 2002.↵
- Chen et al., 1997.↵
- De Smaele et al., 2001.↵
- Fridmacher et al., 2003.↵
- Gao et al., 2005.↵
- Ghosh et al., 1998.↵
- Gillardon et al., 1996.↵
- Gilley et al., 2003.↵
- Graham et al., 2000.↵
- Herrmann et al., 2003.↵
- Herrmann et al., 2005.↵
- Hershko and Ginsberg, 2004.↵
- Hoffmann et al., 2003.↵
- Hu et al., 2004.↵
- Kaltschmidt et al., 2000.↵
- Kaltschmidt et al., 2002.↵
- Kaltschmidt et al., 2005.↵
- Kaltschmidt et al., 1995.↵
- Krajewski et al., 1995.↵
- Kuan et al., 2003.↵
- Lee et al., 1999.↵
- Li and Verma, 2002.↵
- Liou et al., 1999.↵
- Mattson and Camandola, 2001.↵
- Nurmi et al., 2004.↵
- Oda et al., 2000.↵
- Okuno et al., 2004.↵
- Park et al., 2004.↵
- Paxian et al., 2002.↵
- Pizzi et al., 2002.↵
- Pizzi et al., 2005.↵
- Potrovita et al., 2004.↵
- Schneider et al., 1999.↵
- Schreck and Baeuerle, 1990.↵
- Schreiber et al., 1989.↵
- Shibata et al., 2002.↵
- Strasser, 2005.↵
- Sunters et al., 2003.↵
- Tamatani et al., 1999.↵
- Tang et al., 2001.↵
- Tarabin and Schwaninger, 2004.↵
- Ten et al., 1992.↵
- Weih et al., 1995.↵
- Xu et al., 2002.↵
- Yu and Zhang, 2003.↵
- Zhang et al., 2005.↵