Maladaptive inflammation is a major suspect in progressive neurodegeneration, but the underlying mechanisms are difficult to envisage in part because reactive glial cells at lesion sites secrete both proinflammatory and anti-inflammatory mediators. We now report that astrocytes modulate neuronal resilience to inflammatory insults through the phosphatase calcineurin. In quiescent astrocytes, inflammatory mediators such as tumor necrosis factor-α (TNF-α) recruits calcineurin to stimulate a canonical inflammatory pathway involving the transcription factors nuclear factor κB (NFκB) and nuclear factor of activated T-cells (NFAT). However, in reactive astrocytes, local anti-inflammatory mediators such as insulin-like growth factor I also recruit calcineurin but, in this case, to inhibit NFκB/NFAT. Proof of concept experiments in vitro showed that expression of constitutively active calcineurin in astrocytes abrogated the inflammatory response after TNF-α or endotoxins and markedly enhanced neuronal survival. Furthermore, regulated expression of constitutively active calcineurin in astrocytes markedly reduced inflammatory injury in transgenic mice, in a calcineurin-dependent manner. These results suggest that calcineurin forms part of a molecular pathway whereby reactive astrocytes determine the outcome of the neuroinflammatory process by directing it toward either its resolution or its progression.
- inflammatory cytokines
- insulin-like growth factor I
- neuronal death
Inflammation, a common denominator among the diverse list of neurodegenerative diseases, has been implicated recently as a key mechanism responsible for the progressive nature of neurodegeneration (Minghetti, 2005). As a result of a low turnover rate and little ability to recover from injury, neurons are extremely vulnerable to self-destructive inflammatory processes. This might hinder the recovery of neurological function at sites of inflammation or even exacerbate neuronal injury through a pathological cascade. The hallmark of brain inflammation is the activation of glial cells (Ridet et al., 1997). These cells constitute essential partners of neurons, providing functional support through a myriad of intercellular links (Seifert et al., 2006). However, during inflammation, glial cells may play a double-edged role because, depending on the course of the inflammatory process, glial-derived factors may result in being either beneficial or detrimental to neurons (Wyss-Coray and Mucke, 2002). Therefore, a balance between proinflammatory and anti-inflammatory signaling arising from a diversity of cells within the affected brain area will eventually determine the outcome of the inflammatory process. If the proinflammatory cascade, mounted mostly by activated microglia and astrocytes, is harnessed, the neuroprotective signals that these two types of glial cells produce, together with those produced by endothelial cells and neurons, will tilt the balance toward resolution of the process. Although the proinflammatory function of astroglia is usually considered not to be as prominent as that of microglia (Streit et al., 1999), astroglia become activated in response to immunologic challenges or brain injuries (Seifert et al., 2006). In fact, reactive astrogliosis is a marker of brain damage in general (Ridet et al., 1997). At the same time, activated astrocytes provide support for damaged neural tissues through several mechanisms, including release of neurotrophic factors and downregulation of proinflammatory messengers (Wyss-Coray and Mucke, 2002). However, the cellular and molecular pathways underlying interactions between proinflammatory and anti-inflammatory signaling within brain tissue are not well delineated.
While analyzing the anti-inflammatory actions of insulin-like growth factor I (IGF-I), a potent neuroprotective signal upregulated in all types of brain lesions (Torres-Aleman, 2000), we observed that inhibition of IGF-I expression at inflammatory sites exacerbated astroglial reactivity (Fernandez et al., 1997). This effect was probably related to the antagonistic actions of IGF-I on tumor necrosis factor-α (TNF-α) signaling onto astrocytes, an effect in which the phosphatase calcineurin appears to be involved (Pons and Torres-Aleman, 2000). In turn, TNF-α participates in glial-mediated inflammation through activation of the canonical nuclear factor κB (NFκB) pathway (Ginis et al., 2002). In the present work, we explored proinflammatory and anti-inflammatory pathways in astrocytes and their significance in neuronal demise and found that calcineurin plays a dual role in inflammatory responses.
Materials and Methods
Human IGF-I was from GroPep (Adelaide, Australia), bacterial lypopolysaccharide (LPS) and rat TNF-α were both from Sigma (St. Louis, MO), the proteasome inhibitor MG-132 (carbobenzoxy-l-leucyl-l-leucyl-l-leucinal), an IGF-I receptor antagonist [picropodophyllin (PPP)], and the calcineurin inhibitor cyclosporine A (CsA) were purchased from Calbiochem (Darmstadt, Germany). Antibodies against activated caspase 3 (a marker of apoptotic cells; 1:1000; Cell Signaling Technology, Danvers, MA), β3-tubulin (a neuronal marker; 1:5000; Promega, Madison, WI), GFAP (an astrocyte marker; 1:1000; Sigma), OX-42 (a microglia marker; 1:400; Serotec, Oxford, UK), A2B5 (an oligodendroglia marker; 1:500; Serotec), major histocompatibility complex II (MHCII) (1:300; Serotec), CD11b (1:500; Serotec), calcineurin (1:1000: Chemicon, Temecula, CA), β-actin (1:2000; Sigma), superoxide dismutase (SOD) Cu/Zn (1:2000; Stressgen, San Diego, CA), inducible nitric oxide synthase 2 (iNOS2) (BD Transduction Laboratories, San Diego, CA), cyclooxygenase 2 (Cox2) (Cayman Chemical, Ann Arbor, MI), IGF-I (GroPep), NFκBp65, nuclear factor of activated T-cells (NFAT4), GATA3, peroxisome proliferator-activated receptor γ (PPARγ), and ubiquitin (all from Santa Cruz Biotechnology, Santa Cruz, CA), and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (Affinity BioReagents, Golden, CO) were all used at a 1:1000 dilution except when indicated otherwise. Secondary antibodies were goat anti-rabbit or mouse HRP-coupled (1:20,000; Bio-Rad, Hercules, CA) or Alexa-coupled (1:1000; Invitrogen, Carlsbad, CA).
Generation of transgenic mice.
Regulated expression of a truncated, constitutively active form of calcineurin (ΔCnA) in astrocytes was achieved with the “Tet-off” system by generating two transgenic mouse lines. A first mouse line expressed tet-transactivator (tTA) under the astrocyte-specific GFAP promoter (pGFAP). A 1.8 kb of the murine pGFAP was cloned from genomic tail DNA using primers 5′ CTGTGAACTCGAGACCTGCTT 3′ and 5′ TCTCCGCTGAATTCTGCCCTGC 3′ that also generated the sites 5′ XhoI and 3′ EcoRI, respectively. The pGFAP was placed upstream of a XhoI–EcoRI fragment coding for tTA (pTet-off vector; Clontech, Mountain View, CA). The pGFAP–tTA fragment was cut from the vector by digestion with XhoI and HindIII. A second transgenic line expressed ΔCnA under the control of the tTA-sensitive TetO promoter. A cDNA encoding a truncated form of the murine calcineurin catalytic subunit Aα, ΔCnA (O'Keefe et al., 1992), was placed downstream from the TetO promoter (from pTRE vector; Clontech). The TetO promoter–ΔCnA fragment was cut from the vector with XhoI and HindIII. The pGFAP–tTA and TetO–ΔCnA DNA fragments were purified and independently injected into fertilized eggs of C57BL/6/SJL mice. Founder mice were analyzed by Southern blotting and PCR as described previously (Fernandez et al., 2001), using specific primers and backcrossed to C57BL/6 mice to generate the lines pGFAP–tTA (GFAP) and TetO–ΔCnA (ΔCnA). To generate the double-mutant mice [herein called astrocyte inducible calcineurin (AIC) mice], line GFAP7 (F3 generation) was crossed with line ΔCnA5 (F3 generation). Mice had access to food and plain water ad libitum or water complemented with doxycycline (Dox) (2 mg/ml; Sigma). Animals were kept under light/dark conditions following European Union guidelines (Directive 86/609/EEC) and handled according to institutionally approved procedures.
In vivo experiments.
AIC mice underwent brain injury following procedures described previously (Lopez-Lopez et al., 2004). Briefly, cortical stab wounds were performed in deeply anesthetized (2,2,2 tribromoethanol; Sigma) mice receiving (+Dox) or not doxycycline (−Dox) in their drinking water during 1 week before insult to control astrocyte expression of ΔCnA. A 26 gauge needle was inserted 3 mm from the surface of the brain (1.3–1.4 mm caudal to bregma, 1.5 mm lateral). Animals were allowed to survive for 5, 10, 20, 30, and 60 d. In a second group of experiments, AIC mice were treated with Dox for 1 week and submitted to brain injury. Two days later, Dox treatment was discontinued in a subset of animals, and the effect of delayed ΔCnA expression was determined 5 d after injury. Another group of animals treated for 1 week with ±Dox received an LPS injection (Escherichia coli serotype 055:B5; 0.5 μg/2 μl sterile pyrogen-free saline; Sigma) in the parietal cortex over 2 min with a Hamilton syringe at the following coordinates: 1.3 mm anteroposterior, 1.5 mm lateral, and 1 mm dorsoventral. Animals were allowed to survive for 1, 3, and 10 d. A second group of AIC mice also treated for 1 week before inflammatory challenge with or without Dox received a single intraperitoneal injection of LPS (1 mg/kg, in 100 μl vehicle) or vehicle solution. Animals were allowed to survive for 1 and 5 d. To determine perilesion changes, inflammatory markers and neuronal death were determined by Western blot (WB) and immunofluorescence. For biochemical studies, a block of brain parenchyma surrounding the stab wound or injection site (2 mm wide, 4 mm high) was dissected from each animal. Ten animals per group were used.
In vitro experiments.
Pure astroglial cultures [>95% GFAP-positive (GFAP+), OX-42-negative (OX-42−), A2B5− cells] were prepared as described previously (Pons and Torres-Aleman, 2000). Briefly, postnatal (day 3–4) brains were dissected and immersed in ice-cold DMEM/F-12; Invitrogen, Carlsbad, CA) supplemented with 10% fetal bovine serum (FBS). Cortex and hippocampus were removed and cut into 1 mm pieces. Tissue fragments were dissociated mechanically, and the resulting cell suspension was centrifuged (1000 rpm/10 min) and plated in DMEM/F-12 with 10% FBS, 5 × 105 cells per well. Cerebellar granule cells from 7-d-old rats were grown in plates covered with poly-l-lysine and Neurobasal plus B27 (Invitrogen), glutamine, and 25 mm KCl. Under these conditions, ∼95% of the cells are neurons (β3-tubulin+). Neurons were plated at 1 × 106 cells per well. Astrocyte–neuron cocultures were prepared as follows: astrocytes were plated, and, when they were 70% confluent, the medium was removed and neurons were plated onto them with freshly added medium. Cultures were allowed to grow for 4 d. In time course experiments with transfected neuronal cultures or neuron–astrocyte cocultures, cells were treated with cytotoxic doses of LPS (E. coli serotype 0127:B8; 1 μg/ml; Sigma) or TNF-α (20 ng/ml) for up to 72 and 24 h, respectively. The schedule of the experiments in transfected cocultures is detailed in Figure 1B. The calcineurin inhibitor CsA (500 nm) was added to transfected astrocytes 24 h before plating the neurons to avoid inhibition of neuronal calcineurin (see Fig. 1B). The proteasome inhibitor MG-132 (1 μm) was added 1 h before LPS or TNF-α stimulation of cocultures (see Fig. 1B). Cultured cells were transfected using Fugene reagent (Roche, Basel, Switzerland) when 70% confluent. Plasmids transfected were as follows: pGFAP–tTA (GFAP) and TetO–ΔCnA (ΔCnA); pNFκB–secreted alkaline phosphatase (SEAP) or pNFAT–SEAP and pCMV–ΔCnA. Mock transfection was done with pCMV vector.
In experiments using wild-type (wt) cultures, cells were exposed for 16 h to LPS or for 3 h to TNF-α before adding IGF-I (100 nm), and cell death was assessed 2 h later. When investigating the role of calcineurin in IGF-I neuroprotection in wt cells, CsA was added 1 h before IGF-I. In experiments using cells obtained from AIC mice, doxycycline (10 μg/ml) was added to the cultures at the same time that astrocytes were plated to inhibit ΔCnA expression from the onset of the experiment.
In vitro assays.
Production of reactive oxygen species (ROS) was determined by analyzing H2O2 levels in the cultures as described previously (Brera et al., 2000), with minor modifications. Briefly, the method uses the nonfluorescent cell-permeant compound 2′,7′-dichlorofluorescein diacetate (Invitrogen) which can be oxidized by peroxides producing the fluorescent compound 2′,7′-dichlorofluorescein. Generation of peroxides was measured in a FLUOstar plate reader (BMG Lab technologies, Offenburg, Germany) at an excitation wavelength of 485 nm and an emission of 520 nm.
To measure the activity of the transcription factors NFκB and NFAT, we used the SEAP reporter system that determines the levels of enzymatically active SEAP. Assays were performed according to the protocol of the manufacturer (Clontech). Transient transfections were performed with 250 ng of the plasmids pNFκB–SEAP and pNFAT–SEAP, and, after 24 h, luminescence was quantified using a Sirius Luminometer (Berthold Detection Systems, Pforzheim, Germany).
Calcineurin activity was determined using a calcineurin activity kit (Calbiochem). Culture cells and mouse brain tissue were homogenized in lysis buffer (50 mm Tris, 1 mm DTT, 100 μm EDTA, 100 μm EGTA, and 0.2% NP-40, pH 7.5) and protease inhibitor cocktail (Roche). To remove free phosphate, supernatant extracts were added to a desalting column resin. Phosphatase assays were performed in a total of 100 μl in assay buffer (200 mm NaCl, 100 mm Tris, 12 mm MgCl2, 1 mm CaCl2, 1 mm DTT, and 0.05% NP-40, pH 7.5). Recombinant human calcineurin at 50 U/μl, 25 μm calmodulin, and 10 μl RII phosphopeptide were used per assay. When indicated, EGTA was added (20 mm), and assays were incubated at 30°C for 30 min. Calcineurin phosphatase activity was measured by luminescent detection at 620 nm of free phosphate released from the calcineurin-specific RII phosphopeptide using the green reagent. Assays were performed in duplicate at each time point in three separate experiments.
Reverse transcription-PCR was performed using the QuantumRNA Universal 18S system from Ambion (Austin, TX). Total RNA was isolated from cultured cells or mouse tissues using Trizol (Invitrogen). Equivalent amounts of total RNA served as template for cDNA synthesis using M-MLV (Moloney murine leukemia virus) reverse transcriptase (Promega) at 42°C, followed by PCR using specific primers directed to the 3′ end of the GFAP promoter and the 5′ end of ΔCnA: forward, 5′ GATGTGCTTTACTAAGTCATC 3′; and reverse, 5′ CAGTTCATCGTCGGAGCAGA 3′. RNA 18S was quantitated as a loading control.
For RNA interference experiments [Silencer small interference RNA (siRNA) transfection II kit; Ambion], three different double-stranded CnA siRNAs (with sequences CnA1, 5′ GGAAGUGUUUGAUAACGAU 3′; CnA2, 5′ GGAAGGCAGGCUGGAAGAA 3′; CnA3, 5′ GGAUUCUCACCACAACAU 3′) were transfected (30 nm) with siPORT NeoFX (Ambion) into primary astrocytes (12-well plates containing 9 × 104 cells/ml). Seventy-two hours later, cultures were stimulated with LPS (1 μg/ml) overnight. Cells were lysed, and levels of CnA, iNOS2, Cox2, GAPDH (as a siRNA internal control), and β-actin were determined by WB. A nontargeting siRNA and a GAPDH siRNA were used as negative and positive controls, respectively. Assays were performed in triplicate.
Immunocytochemistry was performed as described previously (Trejo et al., 2001). Animals (n = 10 per group) were deeply anesthetized with pentobarbital (50 mg/kg) and perfused transcardially with 4% paraformaldehyde in 0.1 m phosphate buffer, pH 7.4 (PB). Coronal 40-μm-thick sections were cut in a vibratome and collected in PBS. For in vitro experiments, cells were plated on 20 mm coverslips and fixed. Coverslips or free-floating brain sections were blocked with 5% normal horse serum and incubated overnight at 4°C with the respective primary antibody in PB containing 0.1% bovine albumin, 3% horse serum, and 0.2% Triton X-100. After several washes in PB, sections were incubated with an Alexa-coupled secondary antibody (1:1000). Omission of primary antibody was used as control. Neuronal death was determined by a single investigator in uncoded brain sections and cultures, as described (Fernandez et al., 1998). β3-Tubulin+/activated caspase 3+ double-stained cells were scored and expressed as percentage of total β3-tubulin+ cells. Activated caspase 3+ neurons were scored in six consecutive sections in a 1 mm2 area or in six different microscope fields for each coverslip when counting cell cultures. MHCII+ and CD11b+ cells in double-stained brain sections from lesioned areas were scored and expressed as number of cells per square millimeter. Confocal analysis was performed in a Leica (Bensheim, Germany) microscope.
Animals were perfused transcardially with saline before collection of brain samples for biochemical analysis. Cells or cerebral cortex pieces (1 mm around lesion or injection) were homogenized in ice-cold buffer consisting of 20 mm Tris HCl, 150 mm NaCl, 5 mm EDTA, 10% glycerol, 1% NP-40, 400 μm sodium orthovanadate, and protease inhibitors (200 mm PMSF, 1 μg/ml aprotinin, and 1 μg/ml leupeptin). WB and immunoprecipitation were performed as described previously (Pons and Torres-Aleman, 2000). WB membranes were reblotted with either the same antibody used for immunoprecipitation or with β-actin as internal standards and to normalize for protein load. Levels of the protein under study were expressed relative to protein load in each lane. Densitometric analysis was performed using Quantity One software (Bio-Rad). IGF-I was quantified in cell lysates and media by sandwich ELISA using mouse IGF-I as a standard (IBT, Reutlingen, Germany). Briefly, 96-well plates were covered with a monoclonal anti-mouse IGF-I antibody (1:1000; R & D Systems, Minneapolis, MN) at 4°C overnight. After several washes with PBS, pH 7.4, wells were incubated with a blocking solution containing PBS, pH 7.4, 5% Tween 20, 5% sucrose, and 0.05% sodium azide, 2 h at room temperature. After three washes with PBS, pH 7.4, plates were placed at 4°C overnight. Samples were added and maintained at 4°C overnight. After washes, plates were incubated with a biotinylated anti-mouse IGF-I antibody (1:1000; R & D Systems) at 4°C overnight. After washes, plates were incubated with a streptavidin horseradish peroxidase conjugate (1:20,000; GE Healthcare, Little Chalfont, UK). After washing, o-phenyllenediamine dihydrochloride (Sigma) was added during 30 min at room temperature. Absorbance was measured in a spectrophotometer at 450 nm.
A protein array (1.1; Raybiotech, Norcross, GA) to quantitate relative expression of 40 different inflammation-related factors was used following the instructions of the manufacturer. Membranes were blocked for 30 min and incubated with 50 μg of AIC astrocyte culture lysate or with 500 μg of AIC tissue homogenate for 2 h at room temperature and then washed for 30 min and incubated with a biotin-conjugated antibody (1:250) for 2 h. After consecutive washes, streptavidin-conjugated peroxidase (1:1000) was added for 1 h at room temperature. Membranes were washed thoroughly and exposed to ECL substrate (GE Healthcare), followed by apposition of the membranes with autoradiographic film (GE Healthcare) for 1 min. The film was scanned, and spots were digitized into pixel densities using the Quantity One imaging software (Bio-Rad). The ratio of relative expression was established after subtraction of the background intensity and comparison with the positive spots available in the membrane.
Statistical analyses were performed with a t test when comparing two groups and a two-way ANOVA for comparing multiple groups. Statistical significance was established when p < 0.05. Results shown are mean ± SEM.
Calcineurin participates in astrocyte-mediated neuroprotection against inflammatory damage
As mentioned above, we previously found that the phosphatase calcineurin mediates the anti-inflammatory effects of IGF-I on astrocytes (Pons and Torres-Aleman, 2000). To gain additional insight into the biological significance of this anti-inflammatory activity, we transfected wt rat astrocytes with a constitutively active form of the catalytic subunit of calcineurin (ΔCnA). Transfected astrocytes showed high levels of ΔCnA together with marked increases in calcineurin activity (Fig. 1A). Next, we tested the neuroprotective activity of ΔCnA-transfected astrocytes against inflammatory damage. When rat cerebellar neurons were cocultured with mock-transfected astrocytes [cytomegalovirus (CMV)-transfected astrocytes] and submitted to TNF-α or LPS challenge, they died within 24–72 h, respectively (Fig. 1C, left histograms). However, when neurons were cocultured with astrocytes expressing ΔCnA, neuronal death after either LPS (1 μg/μml) or TNF-α (20 ng/ml) was markedly abrogated (Fig. 1C, right histograms) (p < 0.01 vs CMV-transfected astrocytes). Neuroprotection was mediated at least in part by humoral factors because conditioned medium from ΔCnA-expressing astrocytes protected neurons against inflammatory damage after 16 h of exposure to LPS (36 ± 4% of neurons alive in conditioned medium from ΔCnA astrocytes vs 1 ± 1.2% neurons alive in conditioned medium from CMV astrocytes; p < 0.05) or 3 h after TNF-α (data not shown). This protective action was not observed when neurons themselves expressed ΔCnA (supplemental Fig. 1A, available at www.jneurosci.org as supplemental material). At the doses used, LPS or TNF-α did not affect astrocyte survival (data not shown). In addition, when calcineurin activity was inhibited in ΔCnA-expressing astrocytes by adding the calcineurin inhibitor CsA 24 h before coculturing them with neurons, neuroprotection against LPS or TNF-α (data not shown) was significantly diminished (72 ± 5% of neurons alive compared with 35 ± 5% alive when CsA was added; p < 0.05). Furthermore, neuronal production of ROS in response to LPS or TNF-α challenge was completely abrogated when neurons were cocultured with ΔCnA-expressing astrocytes but not when neurons were cultured alone or with mock-transfected astrocytes (p < 0.001 vs CMV transfected) (Fig. 1D).
Astrocyte calcineurin protects against brain inflammatory injury
We next designed in vivo“proof-of-concept” experiments to establish an anti-inflammatory action of calcineurin in astrocytes. Double transgenic mice with Dox-regulated expression of ΔCnA in astrocytes were generated by crossbreeding mice expressing a murine GFAP promoter-driven tTA transgene with mice expressing a TetO-driven ΔCnA transgene. We first corroborated that in vitro expression of these transgenes allowed Dox-regulated calcineurin activity in wt astrocytes cotransfected with them (Fig. 2A). In vivo, double transgenic AIC mice expressed ΔCnA in the brain (Fig. 2B) and markedly enhanced calcineurin activity in a Dox-regulated manner (Fig. 2C). Detailed characterization of AIC mice confirmed the expression of the two transgenes (GFAP–tTA and TetO–ΔCnA) (supplemental Fig. 1B, available at www.jneurosci.org as supplemental material), expression of ΔCnA mRNA in the brain (supplemental Fig. 1C, available at www.jneurosci.org as supplemental material), and Dox-regulated calcineurin immunoreactivity in astrocytes (supplemental Fig. 1D, available at www.jneurosci.org as supplemental material). Expression of ΔCnA mRNA in AIC brain appeared early after birth and was maintained thereafter (Fig. 2D).
Importantly, when AIC mice received Dox in the drinking water, ΔCnA mRNA was inhibited within 24–48 h (Fig. 2E). Because expression of ΔCnA in cultured AIC astrocytes was also inhibited by adding Dox (10 μg/ml) at the onset of the culture (Fig. 2F), we concluded that calcineurin activity in AIC astrocytes was regulated by Dox in vivo and in vitro. That calcineurin activity is regulated by Dox was indirectly confirmed by the observation that Dox regulated the release by AIC astrocytes of humoral neuroprotective signals such as Cu/Zn SOD or IGF-I that are known to be functionally associated with calcineurin (Tokheim and Martin, 2006; Alfieri et al., 2007). Thus, in response to proinflammatory challenge with LPS, AIC astrocytes secreted larger amounts of SOD and IGF-I when cultured without Dox (Fig. 2G,H). Similar results were obtained in wild-type rat astrocytes transduced with ΔCnA and challenged with LPS (supplemental Fig. 1E, available at www.jneurosci.org as supplemental material) or TNF-α (data not shown), respectively. When the IGF-I receptor antagonist PPP (120 nm) was added to LPS-stimulated cocultures containing ΔCnA-expressing astrocytes, the neuroprotective action of astrocytes was moderately reduced: ∼70% of the neurons were alive in ΔCnA-expressing astrocyte cocultures, whereas after adding PPP, ∼60% of neurons were alive. However, in cocultures with mock-transfected astrocytes, the presence of PPP reduced to 50% the neuronal survival induced by astrocytes after LPS insult (from ∼30 to ∼15% neurons alive).
Once we observed that calcineurin protects neurons against inflammatory death in vitro, we explored whether calcineurin in AIC astrocytes protected neurons against brain inflammation in vivo. AIC mice submitted to a penetrating brain injury that produces an inflammatory response in the lesioned area were analyzed 5, 10, 20, 30, and 60 d later. At 60 d, weak glial reactivity was still present in AIC mice not expressing ΔCnA, but the rest of the changes analyzed were resolved within 1 month. Because all pathological changes examined peaked within 5–10 d of the lesion, we analyzed them in detail 5 d after the lesion. In AIC mice with suppressed ΔCnA expression by treatment with Dox (+Dox AIC mice), neuronal death (identified as double-stained β3-tubulin+-activated caspase 3+ cells) 5 d after brain injury was prominent (Fig. 3A). Dox-treated AIC mice also showed reactive astroglial cells (GFAP+ cells) in the lesion site. These reactive astrocytes abundantly expressed proinflammatory markers such as Cox2 (Fig. 3B,C) and iNOS2 (Fig. 3C). In addition, Dox-treated AIC mice show detectable levels of the reactive microglia marker MHCII, as well as the presence of MHCII+ and CD11b+ cells in the lesioned area (Fig. 3D). Notably, MHCII levels, as well as MHCII+ and double-stained MHCII+/CD11b+cells, but not CD11b+ cells that remained elevated, were significantly attenuated in AIC mice expressing ΔCnA before injury (p < 0.001 vs +Dox AIC mice; n = 10) (Fig. 3D, −Dox) (supplemental Fig. 2A, available at www.jneurosci.org as supplemental material). At 5 d after lesion, the number of apoptotic neurons in the lesioned area was significantly decreased when ΔCnA was expressed: 20.6 ± 1.0 β3-tubulin+-activated caspase 3+cells/mm2 in −Dox vs 82.6 ± 1.7 cells/mm2 in +Dox AIC mice (p < 0.001; n = 10). Screening of proinflammatory protein levels with a protein array at day 5 after lesion (Raybiotech) (supplemental Results, available at www.jneurosci.org as supplemental material) confirmed a general downregulation by ΔCnA expression (Table 1). Microglia reactivity and macrophage infiltration at the damaged site, determined by levels of MHCII and number of MHCII+ and CD11b+ cells, was also attenuated (Fig. 3D). Furthermore, as seen in vitro, the number and the intensity of staining of IGF-I-immunoreactive astrocytes (IGF-I+/GFAP+ cells) in the vicinity of the lesion was greater in AIC mice expressing ΔCnA (from 83 ± 16 astrocytes/mm2 in +Dox to 102 ± 27 astrocytes/mm2 in −Dox; p < 0.05). Notably, when expression of ΔCnA was allowed in astrocytes after brain trauma, once the inflammatory cascade is active, neuronal death was still markedly attenuated. Thus, in AIC mice in which Dox treatment was discontinued 2 d after injury (±Dox mice) to allow ΔCnA expression (that starts within 24 h of Dox withdrawal), the number of apoptotic neurons 5 d after injury was markedly reduced, close to the reduction seen when ΔCnA expression was allowed before injury (−Dox mice). There were 32 ± 0.7 apoptotic neurons/mm2 in ±Dox, 17 ± 2.2 neurons in −Dox, but 75 ± 5.3 neurons in +Dox AIC mice (p < 0.001 vs ±Dox and −Dox groups; n = 5 per group). In ±Dox AIC mice, microglia reactivity was also significantly reduced (Fig. 3D, bottom histograms) (p < 0.001 vs +Dox group).
Injection of bacterial endotoxins (LPS) into the brain is a well established model of inflammatory processes associated with human neurotropic infections (Fassbender et al., 2004; Felts et al., 2005). After brain LPS injection to AIC mice, pathological alterations peaked within 1–3 d and were resolved 10 d later. We examined the response to ΔCnA expression 3 d after LPS challenge and found an attenuated inflammatory response when ΔCnA was expressed. Significantly reduced neuronal death after LPS injection was seen in AIC mice without Dox before endotoxic challenge: 12.0 ± 1.8 β3-tubulin+/activated caspase 3+ cells/mm2 in −Dox vs 63.3 ± 1.0 β3-tubulin+/activated caspase 3+ cells/mm2 in +Dox AIC mice (p < 0.001; n = 10). As seen after brain trauma, markers of glial reactivity were also reduced in ΔCnA-expressing AIC mice. These included reduced activation of astrocytes (GFAP+/iNOS2+) (Fig. 4A), decreased levels of proinflammatory mediators in the lesioned area (p < 0.001 vs AIC +Dox; n = 10) (Fig. 4B, Table 1) (supplemental Fig. 2B, available at www.jneurosci.org as supplemental material), and reduced microglia activation (data not shown). Furthermore, systemic administration of LPS to mimic central inflammatory responses to infection, a common complication in elderly demented patients (Perry et al., 2003), resulted also in an attenuated brain inflammatory response when AIC mice expressed ΔCnA. Five days after intraperitoneal injection of LPS, the levels of Cox2 and iNOS2 in cortical tissue of AIC (−Dox) mice were reduced compared with AIC mice receiving Dox in the drinking water (Fig. 4C) (supplemental Fig. 2C, available at www.jneurosci.org as supplemental material) (p < 0.001 vs AIC +Dox; n = 10).
Calcineurin inhibits a canonical proinflammatory pathway in astrocytes
We then explored mechanisms underlying reduced inflammatory responses when calcineurin is activated in astrocytes before inflammatory challenge. In these studies, we used an in vitro approach to determine mechanisms involved at the molecular level. First, we validated the system by determining whether expression of ΔCnA in cultured rat astrocytes also ameliorated the subsequent response to inflammatory challenge. Indeed, as seen in vivo in AIC mice, astrocytes expressing ΔCnA before inflammatory challenge show reduced increases of the inflammatory effectors Cox2 and iNOS2 16 h after LPS (Fig. 5A) or 3 h after TNF-α (data not shown). Other proinflammatory signals were also inhibited (Table 1). With this system, we next observed that, in ΔCnA-transfected astrocytes, but not in mock-transfected ones, LPS- or TNF-α-induced activation of NFκB and NFAT was inhibited even below basal levels (p < 0.001 vs CMV astrocytes) (Fig. 5B). This inhibitory effect of ΔCnA was not seen in neurons (supplemental Fig. 3A, available at www.jneurosci.org as supplemental material). Confirming these results in rat astrocytes, we found that NFκB and NFAT activity after inflammatory challenge was again inhibited below basal levels in astrocytes from AIC mice expressing ΔCnA, i.e., cultured in the absence of Dox (Fig. 5C).
Mechanisms whereby calcineurin may block the activation of NFκB and NFAT by proinflammatory stimuli were then analyzed. As seen previously after IGF-I-induced activation of calcineurin (Pons and Torres-Aleman, 2000), expression of ΔCnA in astrocytes blocked phosphorylation of IκBα by TNF-α (Fig. 5D). Furthermore, expression of ΔCnA blocked the increase in the levels of NFκBp65 and NFAT4 (an isoform found in astrocytes) (Jones et al., 2003) elicited by 16 h of exposure to LPS or 3 h after TNF-α (data not shown) in mock-transfected astrocytes (Fig. 5E) (supplemental Fig. 3B) (p < 0.001 vs CMV-transfected astrocytes; n = 6). We next explored mechanism whereby expression of ΔCnA results in reduced levels of NFκBp65 and NFAT4 after inflammatory challenge. First, we inhibited the proteasome pathway with MG-132 and found that the effect of ΔCnA was blocked (Fig. 5F) (supplemental Fig. 3C, available at www.jneurosci.org as supplemental material) (p < 0.001 vs ΔCnA astrocytes; n = 6). This suggested that the proteasome pathway is involved in the reduction of the levels of NFκBp65 and NFAT4 produced by ΔCnA. In agreement with this possibility, we observed that ΔCnA-expressing astrocytes had increased NFκBp65 and NFAT4 complexed with ubiquitin, a protein tag to the proteasome pathway (p < 0.001 vs CMV-transfected astrocytes; n = 6) (Fig. 5G) (supplemental Fig. 3D, available at www.jneurosci.org as supplemental material). At the same time, the inhibitory action of ΔCnA was accompanied by increased association of NFAT4 with GATA3 (an isoform of brain tissue) (Pandolfi et al., 1995) and of NFκBp65 with PPARγ (Fig. 5H) (supplemental Fig. 3E, available at www.jneurosci.org as supplemental material) (p < 0.001 vs CMV astrocytes; n = 6). These two transcription factors modulate the activity of NFκB and NFAT, respectively (Avni et al., 2002; Bailey and Ghosh, 2005). Similar changes were seen in astrocytes obtained from AIC mice and cultured in the presence or absence of Dox (data not shown).
Astrocyte calcineurin is recruited by both proinflammatory and anti-inflammatory signals
Although our results indicated that active calcineurin counteracts a subsequent inflammatory stimulus, this phosphatase is known to participate in inflammatory responses (Staruch et al., 1998). Indeed, as seen in other cell types, in quiescent wt astrocytes, calcineurin is activated by LPS and TNF-α (Fig. 6A). Moreover, when endogenous calcineurin is depleted in wt astrocytes by transfection of an siRNA against calcineurin, the production of inflammatory mediators such as iNOS2 or Cox2 in response to LPS is abrogated (Fig. 6B). LPS and TNF-α also stimulated the activity of the proinflammatory transcription factors NFκB and NFAT and in a calcineurin-dependent manner, as evidenced by its blockade with CsA (Fig. 6C).
However, an anti-inflammatory factor such as IGF-I, which inhibits TNF-α-stimulated NFκB in astrocytes (Pons and Torres-Aleman, 2000), also enhanced calcineurin activity in wt astrocytes (Fig. 6A). At the same time, IGF-I was able to inhibit the increase in NFκB and NFAT activity after inflammatory stimuli (Fig. 6D). That activation of calcineurin in astrocytes is essential for the neuroprotective actions of IGF-I against inflammation was evidenced by the fact that IGF-I rescued neurons of TNF-α-induced death (Fig. 6E) or block neuronal production of ROS after TNF-α (Fig. 6F) only when astrocytes were present. A similar protective effect of IGF-I was seen after LPS (data not shown). Furthermore, calcineurin was essential for IGF-I neuroprotection because cyclosporine A inhibited its actions (Fig. 6E). Because IGF-I is found in reactive astrocytes in brain lesions in which inflammation occurs (supplemental Fig. 3F, available at www.jneurosci.org as supplemental material), it may activate astrocyte calcineurin during brain injury.
Glial cells are a rich source of trophic signals for neurons, and its potential use as therapeutic targets in neurodegenerative diseases has already been outlined (Ranaivo et al., 2006). However, a better knowledge of the pathways involved in glia-to-neuron protective communication is needed before we can properly address this goal. Our present observations indicate that the phosphatase calcineurin in astrocytes modulates neuronal damage associated with inflammation, a common condition in the neurodegenerative process (Minghetti, 2005). Intriguingly, in brain diseases in which inflammation is involved such as Alzheimer's disease (Mrak and Griffin, 2005) or amyotrophic lateral sclerosis (Weydt et al., 2002), calcineurin activity has been reported to be abnormally low (Ladner et al., 1996; Ferri et al., 2001), whereas in situ generation of constitutively active calcineurin is neuroprotective in brain ischemia (Shioda et al., 2006).
Although the role of calcineurin in inflammation has long been known and its inhibitors are used in the clinic (Shanley, 2002), its concomitant involvement in anti-inflammatory signaling was not documented (Fig. 7). This ambivalent mechanism of astrocyte calcineurin, reminiscent of that reported for CD14 in immune cell responses (Devitt et al., 1998), may help explain the long-held notion that an interplay between proinflammatory and anti-inflammatory signals dictates the outcome of the inflammatory process and, ultimately, its associated neuronal loss (Wyss-Coray and Mucke, 2002). Indeed, signals that activate calcineurin, including neuroprotective factors such as IGF-I, together with cytokines such as TNF-α are present in inflammatory lesions in a distinct time-dependent manner. Whereas TNF-α increases within hours after damage (Yin et al., 2003), IGF-I levels peak after days (Beilharz et al., 1998). This pattern of activation would theoretically allow a calcineurin-dependent termination of the acute inflammatory phase and resultant generation of ROS by an anti-inflammatory action of IGF-I and other protective signals. Other signals are clearly involved because inhibition of the action of calcineurin-induced IGF-I is not sufficient to abrogate the neuroprotective actions of calcineurin-expressing astrocytes. Among these, SOD appears as a likely candidate because its production was enhanced, and this can be related to ROS inhibition and neuroprotection because SOD is a potent free radicals inhibitor and neuroprotective signal.
In initial stages of the inflammatory process, calcineurin will be activated by TNF-α (and other mediators of inflammation). Once IGF-I (and other neuroprotective mediators such as SOD) produced mostly by activated microglia (and other local, i.e., astrocytes, and peripheral sources) start to accumulate at the lesion site, calcineurin will be further stimulated (Fig. 6A). This IGF-I activation of already active calcineurin would result in cancellation of the inflammatory cascade and potentiation of the neuroprotective pathway (Fig. 7). The fact that transgenic expression of constitutively active calcineurin both in vivo and in vitro results also in neuroprotection suggests that a certain level of calcineurin activity (compare the levels of active calcineurin in Figs. 1A and 6A) is required for neuroprotection. Although the initial calcineurin-dependent release of proinflammatory mediators after insult by astrocytes is considered to be beneficial, its subsequent harnessing by calcineurin-dependent anti-inflammatory pathways may be critical to avoid maladaptive responses.
Previous reports indicated both beneficial and detrimental actions of calcineurin in the response of astrocytes to insults, including inflammation (Pons and Torres-Aleman, 2000; Norris et al., 2005). Gene expression analysis of mixed astrocyte–neuronal cultures transduced with ΔCnA showed a reactive/inflammatory astrocyte phenotype. However, expression of ΔCnA in mixed cultures also induced neuroprotective signals, and neuronal proapoptotic genes were downregulated (Norris et al., 2005). Collectively, it seems that activation of calcineurin in resting astrocytes leads to an inflammatory phenotype, whereas its activation in stimulated astrocytes (after stimuli such as inflammatory mediators) results in abrogation of inflammatory cascades. This cell-context-dependent pattern of calcineurin signaling, a common feature of intracellular signaling pathways (Natarajan et al., 2006), may help understand the apparently contradictory observation that calcineurin can regulate NFAT/NFκB activity in opposite ways (Rao et al., 1997; Biswas et al., 2003; Kim et al., 2004; Martinez-Martinez et al., 2006). Calcineurin is also known to play opposing roles in apoptotic responses, depending on cell context and type of stimulus. Indeed, modulation of calcineurin in astrocytes affects their response to apoptotic stimuli (Kaminska et al., 2004), which could theoretically explain its neuroprotective effect because astrocytes are essential protective partners for neurons (Takuma et al., 2004). However, the fact that the proinflammatory stimuli used in the present study did not produce astrocyte death makes it unlikely that astrocyte calcineurin protected neurons by modulating astrocyte apoptosis.
Although the routes involved in stimulation of NFAT and NFκB are well described, the pathways involved in their inactivation are less known. The present results suggest that calcineurin inhibits NFκB by dephosphorylation of nuclear factor κB inhibitor (IκB) and interactions with PPARγ, and NFAT through interactions with GATA. Both transcription factors have been shown previously to interact with NFAT and NFκB, respectively (Chinetti et al., 1998; Wada et al., 2002). Intriguingly, recent evidence indicates that not only calcineurin (see above) but also PPARγ activity is diminished in Alzheimer's disease patients (Sastre et al., 2006). Calcineurin may also direct NFAT and NFκB to the proteasome pathway because lower levels of both transcription factors together with increased association to ubiquitin was observed in the presence of ΔCnA. Our results also confirm the stimulatory role of calcineurin on neuroprotective signals such as SOD and IGF-I (McCall et al., 2003; Norris et al., 2005), providing a conceptual framework to understand the contribution of astroglia to neuronal death in diseases such as amyotrophic lateral sclerosis (Clement et al., 2003), in which these neuroprotective signals are involved (Wilczak et al., 2003; Turner et al., 2005).
Altogether, it seems that calcineurin plays a dual role in activating proinflammatory cascades in response to proinflammatory signals such as TNF-α or canceling them and at the same time promoting neuroprotection in response to trophic signals such as IGF-I. At any rate, the pathways underlying this dual action of calcineurin, which probably involves interactions with scaffolding and modulatory proteins (Klauck et al., 1996; Fuentes et al., 2000; Abbasi et al., 2006), and release of various neuromodulators warrant additional studies.
In summary, astrocyte calcineurin may constitute a key element in the process whereby an acute inflammatory response progresses toward a neuroprotective effect and is resolved or develops into a sustained insult.
This work was supported by Ministerio de Educacion y Ciencia Grants SAF2001-1722 and 2004-0446 and by Neuropharma. We thank Dr. A. Villarroel (Consejo Superior de Investigaciones Científicas, Bilbao, Spain) for his kind gift of the ΔCnA plasmid. We appreciate the invaluable help of C. Hernandez with the confocal analysis and C. Lopez-Lopez for technical advice.
- Correspondence should be addressed to I. Torres-Aleman, Cajal Institute, Consejo Superior de Investigaciones Científicas, Avenida Dr. Arce 37, 28002 Madrid, Spain.