Poly(ADP-ribose) polymerase-1 activity promotes NF-κB-driven transcription and microglial activation: implication for neurodegenerative disorders

Authors

  • Alberto Chiarugi,

    1. Department of Neuroscience, Stroke and Neurovascular Regulation Laboratory, Massachusetts General Hospital, Harvard Medical School, Charlestown, Massachussets, USA
    Search for more papers by this author
  • Michael A. Moskowitz

    1. Department of Neuroscience, Stroke and Neurovascular Regulation Laboratory, Massachusetts General Hospital, Harvard Medical School, Charlestown, Massachussets, USA
    Search for more papers by this author

Address correspondence and reprint requests to Alberto Chiarugi, Department of Cellular and Molecular Physiology, University of Massachusetts Medical School, 55 Lake Avenue North, Worcester, MA 01655, USA. E-mail: alberto_chiarugi@hotmail.com

Abstract

Excessive release of proinflammatory products by activated glia causes neurotoxicity and participates in the pathogenesis of neurodegenerative disorders. Recently, poly(ADP-ribose) polymerase-1 (PARP-1) has been shown to play a key role in nuclear factor kappa B (NF-κB)-driven expression of inflammatory mediators by glia during the neuroimmune response. Here we report the novel finding that the enzymatic activity of PARP-1 promotes, in an β-nicotinamide adenine dinucleotide-dependent fashion, the DNA binding of NF-κB in microglia exposed to lipopolysaccharides, interferon-γ or β-amyloid 1–40. Consistently, we found that targeting NF-κB-dependent glial activation with pharmacological inhibitors of PARP-1 enzymatic activity reduces expression of inflammatory mediators such as inducible nitric oxide synthase, interleukin 1β, tumor necrosis factor α and amyloid precursor protein, and reduces the neurotoxic potential of activated glia in vitro. Importantly, pharmacological inhibition of lipopolysaccharide-induced poly(ADP-ribose) formation in vivo suppresses neuroinflammation and related neural cell death. Our findings build on prior published reports in PARP-1 null mice and highlight the importance of PARP-1 enzymatic activity in transcriptional control during glial activation, identifying PARP-1 activity-dependent regulation of NF-κB as a novel pharmacological target for therapeutic intervention in the treatment of acute and chronic neurodegenerative disorders.

Abbreviations used

beta-amyloid

AD

Alzheimer's disease

APP

amyloid precursor protein

BA

benzoic acid

BZD

benzamide

DMEM

Dulbecco's modified Eagle medium

ERK

extracellular signal-regulated kinase

FBS

fetal bovine serum

FITC

fluorescein isothiocyanate

GFAP

glial fibrillary acidic protein

HMG-I(Y)

high mobility group proteins-I(Y)

HS

horse serum

iNOS

inducible nitric oxide synthase

IRF

IFNγ regulatory factor

LDH

lactate dehydrogenase

MAP2

microtuble-associated protein 2

NF-κB

nuclear factor-κB

NAD+

β-nicotinamide adenine dinucleotide

PARP-1

poly(ADP-ribose)-polymerase-1

PBS

phosphate-buffered saline

PBST

PBS + Triton X-100

PHE

6(5H)-phenanthridinone

STAT

signal transducer and activator of transcription

The role of glial cells in the pathophysiology of CNS diseases has been intensively investigated during the last several years. It is known that glial cells respond to changes in the brain's microenvironment by elaborating numerous mediators and growth factors to restore and maintain neuronal homeostasis (Nguyen et al. 2002). However, astrocytes and microglia may also participate in the pathogenesis of disorders such as multiple sclerosis, HIV encephalitis and stroke (Barone and Feuerstein 1999; Gonzalez-Scarano and Balutch 1999; Iadecola and Alexander 2001) and become activated following amyloid β-protein (Aβ) deposition in Alzheimer's disease (AD) and related neurodegenerative disorders (McGeer and McGeer 2001). Indeed, glial cell stimulation by Aβ or other activating factors leads to the synthesis of reactive radicals and cytokines such as nitric oxide (NO), superoxide anion, interleukin (IL)-1β, IL-6, tumor necrosis factor (TNF)-α, and not yet identified molecules that trigger apoptotic and excitotoxic death programs in neurons (Dawson and Dawson 1998; Liberatore et al. 1999; Allan and Rothwell 2001; Bezzi et al. 2001; Combs et al. 2001; Heneka and Feinstein 2001). Accordingly, a major effort has been directed towards clarifying the molecular and cellular mechanisms of glia–glia and glia–neuron communication in order to develop novel therapeutic strategies to reduce neurotoxicity (Barone and Feuerstein 1999; Combs et al. 2000; Bacskai et al. 2001; McGeer and McGeer 2001; Watterson et al. 2001; Heneka et al. 2002).

Poly(ADP-ribose) polymerase-1 (PARP-1) is a nuclear enzyme activated by DNA strand breaks. PARP-1 synthesizes polymers of poly(ADP-ribose) (PAR) from β-nicotinamide adenine dinucleotide (NAD+) to promote DNA repair (for reviews see de Murcia and Menissier-de Murcia 1994; D'Amours et al. 1999; Herceg and Wang 2001). In conditions of genotoxic stress, however, intense PARP-1 activation depletes NAD+ and adenosine triphosphate (ATP) stores, eventually leading to necrotic cell death by energy failure (Szabo and Dawson 1998; Ha and Snyder 2000; Herceg and Wang 2001). However, recent findings also demonstrate that PARP-1 activation triggers mitochondrial release of apoptosis-inducing factor (AIF), a powerful proapoptotic protein (Yu et al. 2002). Accordingly, PARP-1 inhibition decreases tissue injury after experimental treatment of disorders characterized by genotoxic stress such as ischemia/reperfusion, diabetes, shock, neurotoxicity, and inflammation (Szabo 1998; Szabo and Dawson 1998; Ha and Snyder 2000; Virag and Szabo 2002). PARP-1 also participates in the regulation of gene transcription (for reviews see D'Amours et al. 1999; Ziegler and Oei 2001; Chiarugi 2002). In particular, a growing body of evidence highlights the key role of PARP-1 in the regulation of nuclear factor-κB (NF-κB), a transcription factor essential for immune and stress responses within the brain (Mattson and Camandola 2001). Indeed, NF-κB-driven transcription of proinflammatory cytokines is impaired in PARP-1 knockout animals (Oliver et al. 1999; Ha et al. 2002) and NF-κB and PARP-1 co-immunoprecipitate (Hassa and Hottinger 1999). Despite the relevance of PARP-1 protein to transcriptional regulation, the importance of PARP-1 enzymatic activity to NF-κB trans-activation remains to be established. For example, Hassa et al. (2001) concluded that the enzymatic activity of PARP-1 and its binding to DNA is not required for κB-dependent transcription, whereas Chang and Alvarez-Gonzalez (2001) reported that the PARP-1 inhibitor 3-amino benzamide reduces the DNA binding of NF-κB. Therefore, to gain insight into the role of PAR formation in transcriptional regulation, we investigated whether the impaired NF-κB-dependent immune response already shown in PARP-1–/– mice (Oliver et al. 1999; Ha et al. 2002) could be achieved by using different chemical inhibitors of PARP-1 enzymatic activity. Given that NF-κB processing is crucial for glial cell activation and related neurotoxicity (Akama and Van Eldik 2000; Stern et al. 2000; Combs et al. 2001; Nguyen et al. 2002), we asked whether pharmacological targeting of PARP-1 might represent an alternative way to regulate NF-κB within glia and the detrimental consequences of the neuroimmune response.

We report here that pharmacological inhibition of PARP-1 reduces both NF-κB activity and the synthesis of inflammatory mediators during glial activation. Importantly, these effects correlate with reduced neurotoxicity of activated glial cells in vitro and in vivo. Our results underscore the importance of PARP-1 enzymatic activity and poly(ADP-ribosyl)ation during glial cell transcriptional activation and identify PARP-1 activity-dependent regulation of NF-κB as a possible treatment target for neuroinflammatory and neurodegenerative disorders.

Methods

Rat glial and neuronal cultures

Primary mixed cultures of astrocytes and microglia (referred as ‘glial cells’, unless otherwise indicated) were prepared as previously described (Bal-Price and Brown 2001) and grown in Dulbecco's modified Eagle medium (DMEM) + 10% fetal bovine serum (FBS). Pure microglial cultures (95% pure, as assessed by morphology and immunohistochemistry) were obtained by shaking mixed glial cell culture to dislodge microglia. Glial cells were identified evaluating their morphology and immunoreactivity to glial fibrillary acidic protein (GFAP, astrocytes) and OX-42 and CD45 (microglia). Cells were subcultured in 24-well plates for 48 h before stimulation with 0.3 µg/mL lipopolysaccharide (LPS) plus 100 U/mL rat interferon gamma (IFNγ; Sigma, St Louis, MO, USA) or 30 µm Aβ1–40 (Bachem, King of Prussia, PA, USA), unless otherwise stated. Under these experimental conditions, there was no evidence of cytotoxicity [3-(4,5-dimethylthiazal-2-yl)-2,5-diphenyltetrazolium bromide (MTT) and lactate dehydrogenase (LDH) release assays]. PARP-1 inhibitors (Sigma) dissolved in dimethylformamide were added to the culture medium 1 h before activation. Conditioned media were collected for nitrites, IL-1β and TNFα measurement (Akama and Van Eldik 2000), whereas cells were lysed for western blotting. For immunohistochemistry, glial cells were subcultured on glass slides. Neurotoxicity was determined in three separate experiments conducted in quadriplicate according to Bal-Price and Brown (2001) with minor modifications. Briefly, mixed glia at confluence activated in 6-well plates with LPS + IFNγ/12 h were washed twice with phosphate-buffered saline (PBS) and transferred to primary cultures of cortical neurons prepared as previously described (Moroni et al. 2001). Aliquots of the co-culture medium were collected after 48 h to measure LDH activity (Koh and Choi 1987). MTT reduction was evaluated according to Lobner (2000). Organotypic hippocampal slices (Moroni et al. 2001) were stimulated at 7 days in vitro with LPS (0.1 µg/mL) plus IFNγ (50 U/mL)/6 h. Each experiment was repeated at least three times unless otherwise stated.

Western blotting

Western blotting was performed according to McDonald et al. (1998). Ten or 50 µg of glial cell proteins were loaded per lane to resolve proteins from LPS + IFNγ- or Aβ1–40-stimulated cultures, respectively. Striatal sections collected in Eppendorf tubes were sonicated in a lysis buffer [50 mm Tris pH 7.4, 1 mm EDTA, 1 mm phenylmethylsulphonyl fluoride (PMSF), 4 µg/mL aprotinin and leupeptin, 1% sodium dodecyl sulfate (SDS)] and centrifuged (11 000 g/10 min/4°C). Twenty micrograms of protein/lane were loaded. After 4–20% SDS polyacrylamide gel electrophoresis (SDS–PAGE) and blotting, membranes were probed with primary antibodies [polyclonal anti-iNOS (inducible nitric oxide synthase), anti-IL-1β, and anti-HMG-I(Y) (Santa Cruz, Palo Alto, CA, USA); monoclonal antiβ-actin (Sigma); polyclonal antiphospho-ERK1/2, anti-ERK1/2, anti-p65, and anti-PARP-1 (BD Transduction Laboratories, Lexington, KY, USA), monoclonal anti-PAR (Alexis, San Diego, CA, USA), and monoclonal anti full-length amyloid precursor protein (APP; Abcam, Cambridge, UK)]. For immunoprecipitation, cells were lysed with PBS/1% Nonidet P-40 (PBS/N) and 150 µg soluble proteins were pre-cleared (1 h) in 200 µL PBS/N containing 5 µL rabbit serum and 30 µL Protein-G-sepharose. After centrifugation (3500 g/3 min), proteins were incubated with 2 µg anti-PARP-1 antibody and, 4 h later, 30 µL Protein-G-sepharose were added and immunocomplexes precipitated after 1 h (5000 g/5 min). Proteins were resolved by western blotting. Densitometric analysis was performed by using the NIH Image software.

Semi-quantitative RT–PCR

One microgram of total RNA was reverse transcribed and the DNA mixture subjected to PCR using the following oligonucleotide primers: IL-1β, 5′-GAAAGACGGCACACCCCACC-3′ (sense) and 5′-AAACCGCTTTTCCATCTTCTTCT-3′ (antisense); iNOS, 5′-AGCAGGCACACGCAATGATG-3′ (sense) and 5′-CGCCAAG-AACGTGTTCACCA-3′ (antisense); TNFα, 5′-ATGAGCACAGA-AAGCATGATC-3′ (sense) and 5′-CAGAGCAATGACTCCAAAGTA-3′ (antisense); GAPDH, 5′-CCCTCAAGATTGTCAGCAATG-3′ (sense) and 5′-GTCCTCAGTGTAGCCCAGGAT-3′ (antisense). Numbers of cycles (94°C 30 s, 58°C 30 s, 72°C 1 min, 5 min last extension), selected after determining the linear working range for the reaction, were 25 for GAPDH and 28 for iNOS, IL-1β and TNFα. PCR products were separated on 1.8% agarose gels.

Intrastriatal injection of LPS and edema determination

One microliter of saline containing 0.5 µg LPS or saline alone were injected in the striatum of pentobarbital-anesthetized Sprague–Dawley rats (approximatley 200 g, n = 4 per group) according to Stern et al. (2000) and local animal care regulations. Three additional groups of rats (n = 5) were injected intraperitoneally (i.p.) after LPS injection and every 8 h with 6(5H)-phenanthridonone (PHE; 60 mg/kg), benzamide (BZD; 300 mg/kg) or dimethyl sulfoxide (DMSO; vehicle group) and killed 24 h later. Rats were anesthetized, perfused, and brains removed and sectioned using a cryostat microtome and rostrocaudal landmarks. Sections (10 µm) were mounted on slides and air dried. Striatal volume was measured by an image analysis system (M4, St Catharines, Ontario, Canada) on hematoxylin and eosin-stained sections, and calculated by summing the striatal area of each section multiplied by 0.5 (interval between two sections in mm). Edema was expressed as the volume increase in the injected versus contralateral striatum.

Immunocytochemistry and immunohistochemistry

Glial cells activated with LPS + IFNγ/12 h were washed in PBS and ethanol fixed. Endogenous peroxidase was blocked (hydrogen peroxide 0.3%/30 min) and cells were incubated 1 h with 10% horse serum (HS) diluted in PBS/Triton X-100 0.3% (PBST) and then 3 h with PBST/2% HS containing an anti-iNOS polyclonal antibody (1 : 100, BD Transduction Laboratories). Binding was revealed with a Cy3-conjugated corresponding antibody (1 : 200). Then, cells were incubated with a fluorescein isothiocyanate (FITC)-conjugated anti-OX-42 antibody (1 : 100; Serotec, Oxford, UK). The ABC kit (Vector Laboratories, Burlingame, CA, USA) was used to reveal binding of the anti-CD45 (Pharmingen, San Diego, CA, USA) and antiphospho-ERK1/2 antibodies (1 : 100) in mixed glial cultures, and binding of the anti-PAR antibody (1 : 100) in cultures of pure microglia exposed to LPS + IFNγ/3 h or hydrogen peroxide (1 mm/1 h). Expression of iNOS and OX-42 was quantified by optical density analysis of immunostainings using NIH Image software and expressed as the mean of four fields per slides of three different experiments.

Brain sections mounted on slides were ethanol-fixed, pre-incubated 1 h with PBST/10% HS and then overnight with the primary antibody (1 : 100 in PBST/2% HS). After visualizing the binding with the corresponding secondary antibody, sections were subjected to a further 6 h incubation for double-labeling experiments. Antibodies were as follows: monoclonal and polyclonal anti-PAR (Alexis), monoclonal anti-MHC II (Pharmingen), monoclonal Cy3-conjugated anti-GFAP (Sigma), monoclonal anti-MAP-2 (Chemicon, Tamecula, CA, USA). Cy3- and Cy2-conjugated secondary antibodies were from Jackson ImmunoResearch Laboratories (West Grove, PA, USA). Terminal deoxynucleotidyl-transferase-mediated dUDP-biotin nick end-labeling (TUNEL, Promega, Madison, WI, USA) was performed according to the manufacturer's instructions. Seven sections per striatum (one section every 0.5 mm) were analyzed and cell count expressed as the sum of TUNEL + cells per striatum. Immunolabeling was analyzed by confocal microscopy (Leica DMRB/Bio-Rad MCR 1024, Bio-Rad Laboratories, Hercules, CA, USA). Specificity of immune detection was established by the omission of the primary antibody.

Electrophoresis mobility gel shift assay

NF-κB DNA binding activity was evaluated according to Hehner et al. (1998) using the Gel Shift Kit (Promega) and 10 µg of protein extract. The binding mixture was incubated at room temperature for 20 min 4% non-denaturing polyacrylamide gels and a [32P]-labeled double-stranded oligonucleotide (5′-AGTTGAGGGGACTTTCCC-AGGC-3′) were used. In some experiments, NAD+ and PARP-1 inhibitors were added to the binding mixture 15 min prior to the addition of the labeled probe.

Statistical analysis

Statistical analysis of group differences in striatal edema or numbers of TUNEL-positive cells after treatment was performed using anova + Tukey's w-test. Differences between densitometric values were determined by Kruskall–Wallis + Dunn's post-test. P-values less than 0.05 were considered statistically significant.

Results

Inflammatory activation of glia requires poly(ADP-ribosyl)ation

LPS + IFNγ

Expression of IL-1β and iNOS appeared hours after adding LPS + IFNγ to mixed cultures of microglia and astrocytes, and the response was time-dependent (Fig. 1). Twelve hours after LPS + IFN-γ, only OX-42+ cells (with the typical microglial morphology) were iNOS+ in the mixed culture of astrocytes and microglia (Figs 2C and D). PARP-1 inhibitors, PHE and BZD, 3-amino BZD and nicotinamide (Banasik et al. 1992) suppressed the expression of IL-1β and iNOS. iNOS and IL-1β expression were reduced by 62 ± 13 and 59 ± 9%, respectively, after 30 µm PHE (p < 0.05 vs. LPS + IFNγ), or 48 ± 11 and 56 ± 8% after 1 mm BZD (p < 0.05 vs. LPS + IFNγ), 12 h after LPS + IFNγ exposure (Figs 1a and b). 3-amino BZD (1 mm) reduced the expression of iNOS and IL-1β by 43 ± 8 and 67 ± 7% (p < 0.05 vs. LPS + IFNγ), respectively, and 10 mm nicotinamide suppressed iNOS expression by 71 ± 13% (not shown, p < 0.05 vs. LPS + IFNγ). Both PHE and BZD also reduced iNOS immunoreactivity in cultured microglial cells (Figs 2Be–h and C).

Figure 1.

Inhibitors of PARP-1 activity reduced iNOS and IL-1β expression in glial cultures and organotypic hippocampal slices exposed to LPS + IFNγ. Mixed astrocytes and microglia were activated by LPS + IFNγ and iNOS and IL-1β expression evaluated over time. The addition of a PARP-1 inhibitor PHE (a) or BZD (b) reduced iNOS and IL-1β expression. β-Actin is show as a loading control. Glial cultures exposed for 24 h to PHE (30 µm) or BZD (1 mm) did not exhibit evidence of toxicity as reflected by LDH release or impairment of methylthiazolyl tetrazolium (MTT) reduction (not shown). (c) PHE and BZD dose-dependently inhibited the expression of iNOS and IL-1β in organotypic hippocampal slices exposed to LPS + IFNγ. Each blot is representative of at least three experiments.

Figure 2.

PHE and BZD reduced iNOS expression and PAR formation during microglial activation. (A) Immunohistochemical characterization of mixed glial cultures. GFAP-positive astrocytes (red), flat and irregularly shaped, tended to join in clusters, whereas OX-42-positive microglia (green) were scattered among astrocytes. (B) In a mixed glial culture, OX-42+-positive microglia (a, c, e, g) did not express iNOS under control conditions (b), but expressed iNOS after adding LPS + IFNγ (arrowheads) (d). PHE (f) or BZD (h) reduced iNOS expression without affecting OX42 immunoreactivity (e and g, respectively) in LPS + IFNγ-activated cultures. (C) Optical density readings are shown for the above experiment and each column represents the mean ± SEM of 12 determinations. *p < 0.05, **p < 0.01 vs. LPS + IFNγ. (D) PAR formation in microglia was abrogated by PARP-1 inhibitors after LPS + IFNγ stimulation. (a) PAR in resting microglia. (b) Robust PAR nuclear immunoreactivity after exposure to hydrogen peroxide. (c) LPS + IFNγ increased PAR immunoreactivity. PHE (d) or BZD (e) decreased PAR formation after LPS + IFNγ. An experiment representative of 2 is shown. Bar = 40 µm (A and D), 100 µm (B).

The ability of PARP-1 inhibitors to reduce iNOS and IL-1β expression suggested that PARP-1 enzymatic activity promoted microglial activation. In agreement with this, we detected PAR nuclear staining after incubating (3 h) cultures of pure microglia with LPS + IFNγ, and suppression of this immunoreactivity when PHE (30 µm) or BZD (1 mm) were added to the culture medium (Fig. 2D). Importantly, iNOS and IL-1β expression induced by LPS + IFNγ in organotypic hippocampal slices were reduced by 30 µm PHE (46 ± 12 and 34 ± 13%, respectively, p < 0.05 vs. LPS + IFNγ) and 1 mm BZD (22 ± 9 and 8 ± 7%, respectively, p < 0.01 vs. LPS + IFNγ; Fig. 1c). Hence, these novel findings establish the importance of PARP-1 activity to iNOS and IL-1β expression in both glial cultures and hippocampal tissue.

The effects of PARP-1 inhibitors were not shared by 10 mm benzoic acid (BA, an inactive structural analog of BZD) or 300 µm novobiocin [an inhibitor of mono(ADP-ribosyl)ation; Banasik et al. 1992; data not shown]. Hence, the effects we observed were probably related to PARP-1 inhibition. Moreover, the drugs were effective at doses consistent with their PARP-1 inhibitory potency (Banasik et al. 1992) and, notably, their effects are in line with the impairment of cytokine expression found in PARP-1–/– mice (Oliver et al. 1999; Ha et al. 2002). At these concentrations, we found that PHE and BZD did not affect cytoplasmic signaling events involved in cytokine expression such as IκBα degradation, signal transducer and activator of transcription (STAT)-1 phosphorylation or IFNγ regulatory factor (IRF)-1 expression in microglial cells exposed to LPS + IFNγ for 1–6 h (not shown), suggesting specific nuclear effects for the PARP-1 inhibitors. Hence, although we cannot rule out possible effects of PHE and BZD on other ADP-ribosylating enzymes, the findings reported herein suggested that pharmacological inhibition of PARP-1 blocked glial cell immune activation following LPS + IFNγ.

Aβ stimulates cytokine production in glial cells via an NF-κB-dependent pathway (Combs et al. 2001; Akama and Van Eldik 2000). To examine the importance of PARP-1 to Aβ-induced cytokine expression in glia, we added PARP-1 inhibitors to glial cultures treated with Aβ1–40, an immunogenic fragment of Aβ, and evaluated the chemical's effects on cytokine and APP expression. A 24-h exposure to Aβ1–40 increased iNOS and IL-1β expression, albeit less robust than after LPS + IFNγ(Fig. 3A). The expression of iNOS and IL-1β was reduced in the presence of PHE (30 µm) or BZD (1 mm; Fig. 3A). The two compounds also reduced the production of nitrites, IL-1β and TNFα in the culture medium dose-dependently, whereas BA did not (Fig. 3B). As shown in Fig. 3(C), PHE (30 µm) and BZD (1 mm) suppressed increased APP expression after 12 h incubation with Aβ1–40.

Figure 3.

PHE and BZD suppressed microglial activation by Aβ1–40. (A) iNOS and IL-1β expression was increased in cultures of glia exposed to Aβ1–40. The expression of iNOS and IL-1β was decreased by PHE (− 86 ± 11 and − 90 ± 15%, respectively) or BZD (− 67 ± 9 and − 81 ± 16%, respectively). The expression following LPS + IFNγ stimulation is shown for comparison. (B) PHE and BZD but not BA dose-dependently reduced nitrite (a) as well as TNFα (b) and IL-1β (c) levels in the conditioned medium exposed to Aβ1–40. Each column represents the mean ± SEM of three experiments conducted in triplicate. *p < 0.05, **p < 0.01 vs. Aβ1–40 (anova followed by Tukey–Kramer w-test). (Ca) Full-length APP expression in glial cultures exposed to Aβ1–40/12 h; (Cb) PHE or BZD reduced expression of APP in cultures activated by Aβ1–40. (Da) Microglial expression of CD45 (arrow), a marker of immune cells within the brain. Immunoreactivity of phospho-ERK1/2 was barely detected in microglia under control conditions (Db, arrow) and highly increased by Aβ1–40 (30 µm/15 min; Dc, arrow). Astrocytes (arrowhead) were not immunoreactive. Note the similar morphology of CD45 and phospho-ERK1/2-positive cells shown in (Da) and (Dc), respectively. Bar = 40 µm (E) PHE (30 µm) and BZD (1 mm) did not affect Aβ1–40-induced ERK1/2 phosphorylation in glial cultures exposed for 15 min to the peptide. The expression levels of ERK are shown.

Extracellular signal-regulated kinase (ERK)1/2 phosphorylation in microglia is a crucial signal transduction event after Aβ exposure, downstream of tyrosine kinase activation but upstream to NF-κB activation (McDonald et al. 1998). Accordingly, we found that Aβ1–40-induced ERK1/2 phosphorylation in microglia (Fig. 3D) was not reduced by PARP-1 inhibitors (Fig. 3E). These results provide evidence that both BZD and PHE specifically targeted signaling event(s) downstream of cytoplasmic kinase signaling to reduce expression of inflammatory mediators.

PARP-1 regulates microglia activation at the transcriptional level

We next investigated the effects of PHE and BZD on transcriptional activation in microglia. Both PARP-1 inhibitors dose-dependently reduced iNOS, IL-1β and TNFα transcript levels in microglial cells challenged with LPS + IFNγ/3 h (Fig. 4A). The mRNAs were also reduced by 3-amino BZD (1 mm) and nicotinamide (10 mm; not shown).

Figure 4.

PARP-1 inhibitors regulate transcription in activated microglia. (A) PHE and BZD dose-dependently reduced transcript levels of iNOS, TNFα, and IL-1β in cultures activated with LPS + IFNγ. GAPDH is show as a loading control. (B) PARP inhibitors modified NF-κB binding activity in extracts of microglial cells activated with LPS or Aβ1–40. (a) The LPS-induced DNA binding activity of NF-κB appears as a retarded band (p65) reduced by an antibody raised against the NF-κB subunit p65 (RelA) or by the addition of 50-fold molar excess of cold probe. (b) Aβ1–40 dose-dependently increased the DNA binding activity of NF-κB. PHE and BZD dose-dependently reduced NF-κB binding by LPS (c) or Aβ1–40 (d). (e) BA did not decrease NF-κB binding activity. (f) DNA binding activity of NF-κB is increased dose-dependently by NAD+ and reduced by PHE and BZD. The addition of NAD+ also caused the appearance of supershifted bands that were suppressed by PHE or BZD. The non-specific DNA binding activity was decreased by NAD+ and partially restored by PHE and BZD. (C) p65 and HMG-I(Y) co-immunoprecipitated with PARP-1 in protein extracts from cultured microglial cells. Note that LPS activation increases PARP-1-HMG-I(Y) interaction, an effect prevented by PHE or BZD. An experiments representative of 2 is shown.

Trans-activation of genes encoding iNOS, IL-1β and TNFα is typically driven by NF-κB, a transcription factor positively modulated by PARP-1 protein (Hassa and Hottinger 1999; Oliver et al. 1999; Chang and Alvarez-Gonzalez 2001; Hassa et al. 2001; Ha et al. 2002). We therefore investigated whether this positive modulation was dependent upon PARP-1 enzymatic activity. In extracts of microglia exposed to LPS (2 h) or Aβ (3 h), the DNA binding activity of NF-kB appeared as a retarded band that was reduced by the addition to the binding mixture of an antibody raised against the NF-κB subunit p65. The band was also reduced by 50-fold molar excess of cold probe, thereby indicating specificity of binding (Figs 4Ba and b). We found that the LPS- or Aβ-induced DNA binding activity of NF-κB was reduced by PHE or BZD but not BA (Figs 4Bc–e). Taken together, these novel findings establish that PARP-1 enzymatic activity was a necessary prerequisite for the DNA binding activity of NF-κB during microglial activation.

Because PARP-1 binds to DNA–protein complexes (Nirodi et al. 2001; Butler and Ordahl 1999; Akiyama et al. 2001; Chang and Alvarez-Gonzalez 2001; Hassa et al. 2001; Soldatenkov et al. 2002), we explored whether the enzyme was present and actively regulating NF-κB binding activity within microglial extracts. We therefore added the PARP-1 substrate NAD+ to the microglial binding mixture in vitro (see Methods) and found that the pyridine nucleotide enhanced NF-κB binding in a dose-dependent manner. NAD+ also promoted the formation of DNA–protein complexes (supershifted bands). Both effects were prevented by adding PARP-1 inhibitors to the binding mixture (Fig. 4Bf), suggesting that PAR assisted NF-κB–DNA interaction and the formation of supramolecular complexes recognizing the κB element. Poly(ADP-ribose) glycohydrolase, the PAR-degradating enzyme (potentially active in our binding mixture), may have also modulated the NAD-induced PAR-dependent NF-κB DNA binding activity. NAD+ also reduced non-specific binding of nuclear proteins to the NF-κB recognition sequence, an effect partially restored by inhibitors of PARP-1 (Fig. 4Bf). The identity of the protein(s) involved in constitutive non-specific binding to the κB oligoprobe is unknown. Nevertheless, the finding is consistent with the observation that PAR formation differentially affects protein-DNA binding, and that poly(ADP-ribosyl)ation both promotes and reduces transcription depending upon the gene under investigation and the transcription factor involved (Ziegler and Oei 2001; Chiarugi 2002).

PARP-1 activity promoted the association of the transcription-assisting high mobility group protein (HMG)-I(Y) within a protein complex containing PARP-1 and NF-κB. As shown in Fig. 4(C), the NF-κB subunit p65 and HMG-I(Y) co-immunoprecipitated with PARP-1 in extracts of untreated microglia. Following activation, even greater amounts of HMG-I(Y) co-immunoprecipitated with PARP-1, an effect suppressed by the presence of PHE (30 µm) or BZD (1 mm) during LPS exposure (2 h; Fig. 4C). Hence, these novel findings demonstrate that PARP-1 enzymatic activity is essential for NAD+-dependent PARP-1 facilitation of NF-kB DNA binding and assembly in transcription-regulating complexes.

PARP-1 promotes the neuroinflammatory response and neurotoxicity

To investigate whether PARP-1 activity contributes to glia-induced neurotoxicity, astrocytes and microglia activated with LPS + IFNγ were added to primary cultures of mixed cortical cells (neurons/glia, see Methods) and the MK-801-sensitive release of LDH in conditioned medium was taken as an index of neurotoxicity (Koh and Choi 1987). Activated glial cells significantly increased LDH activity [LDH: 312 ± 67%, p < 0.01 vs. 109 ± 24% (resting glia), n = 12, Student's t-test]. When the NMDA receptor antagonist MK-801 (10 µm) was added to the co-culture, LDH levels decreased to 171 ± 68% (n = 12), thereby indicating that neurons were a significant source of LDH and died by excitotoxicity because glia do not express NMDA receptors. Activated glial cells dying during the neuroinflammatory response may also have contributed to LDH release. These findings are consistent with the notion that activated glia release glutamate and neurotoxins leading to NMDA receptor overactivation (Bal-Price and Brown 2001; Bezzi et al. 2001). When neurons were exposed to glia previously activated with LPS + IFNγ plus PHE (30 µm) or BZD (1 mm), LDH activity in the co-culture medium was reduced [143 ± 51% (PHE), 166 ± 40% (BZD), n = 12, p < 0.05 vs. LPS + IFNγ-activated glia]. These novel findings indicate that PARP-1 activity is essential for neurotoxicity of activated glial cells in culture.

We next evaluated the importance of poly(ADP-ribosyl)ation and cytokines to mechanisms of neural cell killing in vivo. To this end, we microinjected LPS into the rat striatum, a classical in vivo model of immune-induced neurodegeneration. We first determined whether PARP-1 was activated during neuroinflammation. Six hours following LPS microinjection into the striatum, numerous PAR+ nuclei were found, whereas striatal PAR was undetectable after saline injection (not shown); immunoreactivity sustained for at least 24 h (Fig. 5A). PAR was formed in iNOS positive cells (Figs 5Ba–c), GFAP+ astrocytes (Figs 5Bd–f) or microtubule-associated protein-2 (MAP-2)+ neurons (Figs 5Bg–i). Parallel experiments demonstrated that iNOS+ cells completely co-localized with those expressing major histocompatibility complex class II antigen (not shown), a marker of activated microglia/macrophages within brain. In addition, PAR co-localized in cells labeled by TUNEL (Figs 5Bl–n), a marker of DNA damage and apoptotic cell death after intracerebral injection of LPS (Matsuoka et al. 1999).

Figure 5.

Synthesis of PAR during the neuroinflammatory response to LPS in vivo. (A) LPS caused a time-dependent (6–24 h) increase of PAR nuclear immunoreactivity surrounding the needle track after microinjection into the rat striatum (arrowheads and 6 h, inset). At 72 h, only non-specific immunoreactivity was visible (not shown). (B) iNOS+ (a, green) and PAR+ (b, red) cells partly co-localized (c, yellow) 24 h after the intrastriatal injection as did (f, yellow) GFAP+ astrocytes (d, red) and PAR+ nuclei (e, green). MAP-2+ neurons (g, red) and PAR+ nuclei (h, green) co-localized (i, yellow) in the same area. An almost complete overlap (n, yellow) occurred between TUNEL+ (l, green) and PAR+ (m, red) nuclei. Arrowheads indicate cells/nuclei immunopositive to PAR. Bar = 200 µm (A), 50 µm (B).

To understand the pathophysiological significance of PAR formation during the neuroimmune reaction, rats were treated with PHE or BZD after LPS microinjection into the striatum (see Methods). As shown in Fig. 6(a), striatal PAR levels increased 24 h after LPS injection (226 ± 61% of saline), along with iNOS and IL-1β expression. Treatment with PHE or BZD decreased PAR formation in this tissue (121 ± 16 and 82 ± 25% of saline, respectively; Fig. 6a), demonstrating that drug administration inhibited poly(ADP-ribosyl)ation in vivo. Interestingly, PHE and BZD also reduced the expression of iNOS by 60 ± 11 and 55 ± 13%, respectively, and IL-1β by 38 ± 9 and 51 ± 17%, respectively (Fig. 6a), consistent with the in vitro data. Furthermore, striatal edema was significantly reduced by treatment with PARP-1 inhibitors (Fig. 6b). Finally, PHE and BZD reduced the number of TUNEL+ cells in the LPS-challenged striatum (Fig. 6c), consistent with a model of cell death induced by proinflammatory mediators (Gonzalez-Scarano and Balutch 1999; Allan and Rothwell 2001).

Figure 6.

PHE and BZD reduce PAR formation, iNOS and IL1β expression, edema formation as well as number of TUNEL+ cells following LPS microinjection into the rat striatum. (a) PHE or BZD reduced PAR formation and expression of iNOS and IL1β induced in the striatum 24 h after LPS microinjection. Numbers (kDa) on the left refer to the migration level of molecular weight standards. β-Actin is show as a loading control. Two representative animals per group are shown. PHE or BZD treatment also reduced edema (b) and TUNEL+ cells (c) in the LPS-injected striatum. In (b) and (c) each column represents the mean ± SEM of five animals. *p < 0.05 vs. DMSO.

Discussion

We report the novel finding that PARP-1 enzymatic activity promotes NF-κB DNA binding during glial activation by LPS + IFNγ or Aβ1–40. Consistent with these results, drugs inhibiting PARP-1 activity reduce NF-κB-dependent transcription of IL1β, iNOS, and TNFα in glia and protect neurons and brain tissue in models of immune-induced neurodegeneration. The present results are in line with the emerging importance of glia in the pathogenesis of neurodegenerative disorders (Allan and Rothwell 2001; Nguyen et al. 2002) as well as with the impairment of NF-κB activity in PARP-1–/– mice (Oliver et al. 1999; Hassa et al. 2001; Ha et al. 2002) and their resistance to brain injury and peripheral inflammation (Szabo and Dawson 1998; Ha and Snyder 2000; Herceg and Wang 2001). By showing the importance of PARP-1 enzymatic activity to NF-κB-driven trans-activation and related neurotoxicity, our findings build on previous work demonstrating that suppressing PARP-1 protein expression impairs glial activation (Ha et al. 2002) and abrogates neurotoxicity of microglia in NMDA-treated hippocampal slices (Ullrich et al. 2001a).

We confirm prior work showing that poly(ADPribosyl)ation promotes the interaction among PARP-1, p65 (the trans-activating subunit of NF-κB) and transcriptional co-activators HMG-I(Y) in activated microglia (Ullrich et al. 2001a). We also report the novel finding that PARP-1 substrate NAD+ promotes NF-κB DNA binding activity in extracts of activated microglia in a poly(ADP-ribosyl)ation-dependent manner, in line with results obtained with purified PARP-1 and NF-κB (Chang and Alvarez-Gonzalez 2001). These findings, together with evidence that HMG-I(Y) promotes NF-κB-driven transcription of proinflammatory mediators (Perella et al. 1999; Ullrich et al. 2001a), indicate that PARP-1 enzymatic activity is essential to co-ordinate the assembly of NF-κB into DNA-binding, transcriptionally active supramolecular complexes during glial activation. Interestingly, we also show that NAD+ promoted the formation of supershifted bands in gel shift assays (Fig. 4Bf). Although the molecular characterization of these additional bands is currently under investigation, it is worth noting that PAR-binding domains have been detected in NF-κB subunits (Pleschke et al. 2000) and that HMG-I(Y) are well known nuclear targets of PARP-1 (Giancotti et al. 1996; D'Amours et al. 1999). These findings, along with the evidence of PARP-1/HMG-I(Y)/NF-κB-p65 complex formation and PARP-1 binding to specific DNA elements (Butler and Ordahl 1999; Akiyama et al. 2001; Nirodi et al. 2001), suggest that the three proteins might be components of the supershifted bands. Although not investigated here, NF-κB subunits such as p50, Rel-B, and C-Rel may interact with PARP-1 in addition to p65. Thus, consistent with a previous study on transcription enhancer factor-1 and β-myosin expression (Butler and Ordahl 1999), we suggest that the NAD+-dependent increase of NF-κB binding as well as formation of supershifted bands may be the result of direct PAR targeting to transcriptional protein complexes (i.e. specific and general transcription factors and components of the basal transcriptional machinery). The ability of PARP-1 inhibitors to prevent both NF-κB DNA binding and supershifted band formation further emphasizes that the synthesis of PAR is critical for modulating the assembly of transcription-regulating multiprotein complexes. Accordingly, PARP-1 has been identified as the transcription co-activator TFIIC (Slattery et al. 1983) and its activity regulates expression of iNOS (Le Page et al. 1998), chemokines (Nirodi et al. 2001; Hasko et al. 2002), integrins (Ullrich et al. 2001a) as well as muscle proteins (Butler and Ordahl 1999). In addition, recent evidence suggests physiological roles for PARP-1 under homeostatic conditions. Indeed, PARP-1 binding to specific promoter elements and poly(ADP-ribosyl)ation mediate transcriptional regulation (Butler and Ordahl 1999; Akiyama et al. 2001; Nirodi et al. 2001; Zhang et al. 2002), and PAR formation and targeting also occur in the absence of DNA damage (Kun et al. 2002). The possibility that certain promoters lack PARP-1 binding element(s) might explain the important finding by Snyder and his group that transcriptional regulation by PARP-1 is cell type- and gene-specific (Ha et al. 2002). Although complementary mechanisms based on non-enzymatic actions have been proposed to explain transcriptional control by PARP-1 (Hassa et al. 2001), our findings are consistent with a NAD+/ADP-ribosylation-dependent model of transcriptional regulation by PARP-1.

The present study emphasizes that transcriptional repression of inflammatory proteins by PARP-1 inhibitors modulates inflammation and cell death in vitro and in vivo. Indeed, we found that suppression of NF-κB DNA binding activity and iNOS, IL-1β and TNFα transcription in glia by PARP-1 inhibitors correlated with both reduced LDH release from co-cultured neurons and number of striatal TUNEL+ cells after LPS microinjection. These results are in line with the ability of glia to kill neurons by releasing neurotoxic factors such as NO, IL-1β, and TNFα (Barone and Feuerstein 1999; Allan and Rothwell 2001; Bal-Price and Brown 2001; Bezzi et al. 2001; Combs et al. 2001), with the role of PARP-1 in apoptotic cell death (Simbulan-Rosenthal et al. 1998; Chiarugi 2002; Yu et al. 2002), and indicate for the first time that targeting formation of PAR in glia is a novel, indirect mechanism that provides neuroprotection. However, in light of the antiapoptotic effects of NF-κB on neurons (Mattson and Camandola 2001), excessive repression of NF-κB trans-activation by sustained inhibition of poly(ADP-ribosyl)ation may result in neurotoxicity. Increased production of APP([an NF-κB-driven gene; Grilli et al. 1996) and inflammatory mediators in response to Aβ1–40 is thought to be a key determinant of brain pathology (Carlson et al. 2000; McGeer and McGeer 2001). The finding that PAR immunoreactivity is increased in AD brains (Love et al. 1999), plus the novel finding that PARP-1 inhibitors decreased cytokines and full-length APP expression in glia exposed to Aβ1–40 (Fig. 3), suggest that PARP-1 activity is of pathophysiological significance in amyloid-related diseases. Furthermore, PARP-1 gene deletion protects mice from methyl-phenyl-tetrahydropyridine (MPTP)-induced parkinsonism (Mandir et al. 1999), strengthening the link to neurodegenerative diseases.

The ability of PARP-1 inhibitors to reduce expression of proinflammatory mediators and edema in the LPS-injected striatum (Figs 6a and b), indicates that poly(ADP-ribosyl)ation is a key event in the development of neuroinflammation in vivo as well. Furthermore, the fact that PAR+ nuclei (also found in MAP-2+ neurons, Figs 5Bg–i) completely co-localized with TUNEL+ cells (Figs 5Bl–n) and that these latter were reduced by the treatment with PARP-1 inhibitors (Fig. 6c) is an indirect evidence that PAR formation actively participates in neurotoxicity during in vivo neuroinflammation. The novel finding that PAR was increased 6 h after intrastriatal LPS injection (Fig. 5A) confirms our result that poly(ADP-ribosyl)ation is an early step in the neuroinflammatory response in vitro (Fig. 2C). The synthesis of PAR during in vitro and in vivo glial activation may be due to several possible stimuli such as DNA damage (Berton et al. 1991), metabolic stress (Ullrich et al. 2001b), increases in intracellular calcium (Homburg et al. 2000) or activity by other newly identified PARPs (Chiarugi 2002). In addition, the appearance of PAR in sub rather than entire populations of cells (e.g. neurons, glia; Fig. 5) may be due to rapid PAR hydrolysis by poly(ADP-ribose) glycohydrolase and very short half-life of polymer chains (D'Amours et al. 1999).

Taken together, our results indicate that the first wave(s) of PARP-1 activity and PAR formation (Figs 2C and 5A) promote(s) synthesis of mRNAs (Vispèet al. 2000) and proinflammatory mediators (Figs 1 and 6a). Accordingly, PARP-1 has been recently identified as a RNA polymerase II-binding protein (Carty and Greenleaf 2002) and PAR formation itself may regulate PARP-1 binding to DNA elements important to transcriptional events (Soldatenkov et al. 2002). The inhibition of poly(ADP-ribosyl)ation early on, may therefore abort neuroinflammation and subsequent DNA damage-induced PARP-1 overactivation, ATP depletion, and cell death. In conclusion, we demonstrate that the mechanisms by which PARP-1 operates in transcriptional regulation within glial cells depends upon poly(ADP-ribosyl)ation and can be pharmacologically targeted to reduce inflammation and obtain neuroprotection.

Acknowledgements

This study was supported by NIH grants 5 P50 NS10828 and 5 R01 HL62602.

Ancillary