Poly(ADP-ribose)polymerase-1 modulates microglial responses to amyloid β

Cell culture reagents were obtained from Cellgro/Mediatech (Herndon, VA), unless otherwise stated. Culture plates (24-well plates) and 75 cm² polystyrene culture flasks were from Falcon/Becton Dickinson (Franklin Lakes, NJ). N-(6-oxo-5,6-dihydrophenanthridin-2-yl)-N,N-dimethylacetamide (PJ34) was obtained from Sigma. (E)-3-(4-methylphenylsulfonyl)-2-propenenenitrile (BAY 11-7082) was obtained from Alexis Biochemicals. Amyloid beta _1-42 (Aβ), reverse amyloid beta _42-1 (rAβ), and carboxyfluorescein (FAM)-labeled amyloid beta _1-42 (FAM-Aβ), were obtained from Biopeptide Co. Inc. (San Diego, CA). Primary antibodies used were: rabbit polyclonal anti-poly(ADP-ribose) (PAR; Trevigen, Gaithersburg, MD), rabbit polyclonal anti-mouse ionized calcium binding adapter molecule 1(Iba-1; Waco), rabbit polyclonal anti-glial fibrillic acid protein (GFAP; Chemicon, Temecula, CA), rabbit polyclonal anti-microtubule-associated protein 2 (MAP2; Chemicon, Temecula, CA), mouse monoclonal anti-amyloid β 3D6 (Elan Pharmaceuticals) and rabbit polyclonal anti-Calbindin D-28k (Swant, Bellinzona, Switzerland). Secondary antibodies used were: anti-rabbit IgG conjugated with Alexa Fluor 488 or 594 (Molecular Probes Inc., Eugene, OR).

All animal studies were approved by the San Francisco Veterans Affairs Medical Center animal studies committee and follow NIH guidelines. PARP-1 ^-/- mice were derived from the 29S-Adprt1^tm1Zqw strain, originally developed by Z. Q. Wang [27], and obtained from Jackson Laboratory (Bar Harbor, ME). PARP-1 ^-/- mice used for cell culture studies were backcrossed for over 10 generations with wt CD-1 mice, and wt CD-1 mice were used as their controls. PARP-1 ^-/- mice used for in vivo studies and for generating the hAPP_J20/PARP-1 ^-/- mice were backcrossed to the C57BL/6 strain for over 10 generations. The hAPP_J20 mice on the C57BL/6 background were obtained from Dr. Lennart Mucke (Gladstone Institute). These mice express a hAPP minigene with the familial AD-linked Swedish (K670N, M671L) and Indiana (V717F) mutations, under control of the platelet-derived growth factor (PDGF) β-chain promoter [28]. The hAPP_J20 mice were crossed with the PARP-1 ^-/- mice to obtain the breeder genotypes: PARP ^+/- and hAPP_J20 /PARP-1 ^+/- . These were in turn crossed to generate subsequent generation breeder genotype mice along with the four genotypes of interest: wt, PARP-1 ^-/- , hAPP_J20 and hAPP_J20/PARP-1 ^-/- . Male mice 5 - 6 months of age were used for in vivo studies. Genotype was re-confirmed on each mouse using tissue obtained at euthanasia.

Neuron cultures were prepared as described previously [29]. In brief, cortices were removed from embryonic day 16 wt mice, dissociated into Eagle's minimal essential medium (MEM) containing 10 mM glucose and supplemented with 10% fetal bovine serum (Hyclone, Ogden UT) and 2 mM glutamine, and plated on poly-D-lysine-coated 24-well plates at a density of 7 × 10⁵ cells per well. After 2 days in vitro, 22 μM cytosine β-D-arabinofuranoside (Sigma, St. Louis, MO) was added to inhibit the growth of non-neuron cells. After 24 hours, the medium was removed and replaced with a 1:1 mixture of glial conditioned medium (GCM) and MEM. This medium was 50% exchanged with fresh medium after 5 days. The cultures contained about 97% neurons and 3% astrocytes as assessed by immunostaining for the neuron marker MAP2 and the astrocyte marker GFAP.

Cortices were dissected from 1-day old mice and dissociated by mincing followed by incubation in papain (40 units) and DNase (2 mg) for 10 minutes at 37°C. After centrifugation for 5 minutes at 500 g, the cells were re-suspended and triturated with a fire-polished Pasteur pipette into Eagle's minimal essential medium (MEM) containing 5 mM glucose and supplemented with 10% fetal bovine serum (Hyclone, Ogden UT) and 2 mM glutamine. Cells were plated on 24-well plates or glass coverslips at a density of 2 × 10⁵ cells per well, or in 75 cm² flasks at a density of 5 × 10⁶ cells per flask, and maintained in a 37°C in a 5% CO₂ incubator. The medium was changed at 3 days in vitro and once per week thereafter. These cultures contained both astrocytes and microglia. Microglia were isolated from these cultures at age 2 to 3 weeks in vitro by shaking, and collecting the floating cells [24]. The cells were re-plated at a density of 5 × 10⁵ cells per well in 24-well plates for microglial monocultures, or at the density of 5 × 10⁴ cells per well on top of 6-day in vitro neuron cultures in 24-well plates for microglia-neuron co-cultures. The purity of the re-plated microglial monocultures was > 99%, and the microglia-neuron co-cultures contained about 7% microglia, 90% neurons, and 3% astrocytes as assessed by immunostaining for the microglial marker Iba-1, the neuron marker MAP2 and the astrocyte marker GFAP.

For in vitro use, 1 mM stock solutions of Aβ peptides (Aβ and rAβ) were diluted to 250 μM with MEM and incubated for 1 hour at 37°C to produce a mixture of Aβ monomers and oligomers [30]. For in vivo use Aβ peptides were diluted to 1 mg/ml (220 μM) with normal saline. The solution was prepared within one hour of use and kept at room temperature in order to maintain the peptides in oligomeric form (fibrils would block the syringe) [30, 31].

Neuron monocultures and microglia-neuron co-cultures were used at neuron day 7 in vitro. Microglial cultures were used at day 2-3 after re-plating. Cultures were incubated with 5 μM of Aβ or 5 μM of rAβ alone, or with inhibitors of PARP activation (PJ34, 400 nM) or NF-κB activation (BAY 11-7082, 5 μM) for the designated intervals. In some experiments, 5 μM of carboxyfluorescein-labeled amyloid β_1-42 (FAM-Aβ) was used to detect microglial phagocytosis of Aβ fibrils. All compounds were dissolved in MEM (microglia) or GCM/MEM mixture (neurons), and these solutions were used alone for control conditions.

All evaluations in this study were performed by observers blinded to the experimental conditions. Neuronal survival was determined by cell counting in 5 randomly selected phase contrast microscopic fields per culture well. Values were normalized to counts in control wells from the same 24-well plate. Microglia morphology was assessed by phase contrast microscopy of unfixed cells. Microglia with two or more thin processes were considered as ramified, resting microglia, and microglia with less than two processes, or with amoeboid cell soma, were classified as activated [24]. The numbers of resting and activated microglia were counted in 5 randomly selected fields per culture well. Immunostaining was performed with cultures fixed with 1:1 methanol:acetone at 4°C. Cultures were characterized with antibodies to GFAP and Iba-1 as previously described [24]. Antibody binding was visualized with suitable Alexa Fluor - conjugated anti-IgG. Negative controls were prepared by omitting the primary antibodies. For detection of poly(ADP-ribose), cultures were incubated with rabbit antibody to PAR. Microglial phagocytosis of Aβ was imaged using three-dimensional confocal imaging of cultures with microglia-astrocyte co-cultures exposed to 5 μM of FAM-Aβ. Microglial phagocytic activity in microglial monocultures was quantified as described [32] with minor modifications by measuring FAM fluorescence remaining in the cells after two washes with MEM. Nonspecific Aβ adherence to the culture plate surface was evaluated by measuring FAM fluorescence in cell-free culture wells that had been incubated with FAM-Aβ for 24 hours.

Microglial cultures were placed in 250 μl of MEM and incubated with Aβ or rAβ for 24 hours. Nitric oxide production was measured by using Griess reagent as previously described [25]. Cytokines and tropic factors were analyzed in 50 μl aliquots of cell culture medium using a Milliplex mouse multiplex immunoassay bead system according to the manufacturer's instructions (Millipore). Each sample was assayed in duplicate, and the fluorescent signal corresponding to each cytokine was measured with a BioPlex 200 system (Bio-Rad, Hercules, CA) in parallel with known standards. Nonspecific interactions between beads and test compounds were screened by running the immunoassay with test compounds dissolved in medium without cell culture exposure. The reverse sequence Aβ_42-1 (but not Aβ_1-42) was found to interfere with the assay in a non-specific manner, and thus rAβ-treated cultures could not be analyzed. Values for cytokine and trophic factor assays were normalized to the protein content of each well as determined by the bicinchoninic assay [33].

Microglia were infected with lentivirus encoding destabilized, enhanced green fluorescence protein driven by the NF-κB promoter (Lenti-κB-dEGFP) [34] at 8-9 days in vitro, while still in co-culture with astrocytes. Infection was performed in culture medium with viral titer of 6.4 × 10^-8 pg of p24 antigen/ml. The microglia were isolated and re-plated 5-6 days later, and used for experiments 2 days after re-plating. Photographs were prepared at the designated intervals after Aβ exposure, and the percent of cells expressing green fluorescent protein (GFP) were counted in five random fields within each well.

Wt and PARP-1^-/- mice were given stereotaxic injections of Aβ, rAβ, or saline vehicle into hippocampus (anteroposterior 2.0 mm, mediolateral 1.5 mm, and dorsoventral 2.0 mm from bregma and cortical surface) with a Hamilton syringe. Mice received 1 μg of Aβ (or rAβ) in a 1 μl injection volume. Injections were made over a 5 minute period and the needle was withdrawn after an additional 5 minutes. Some animals received i.p. injection of PARP inhibitor (PJ34, 15 mg/kg) 15 minutes prior the Aβ injections. In a subset of experiments FAM-Aβ was used to confirm uniform injection volumes and identify the area Aβ diffusion. Mice were euthanatized 6 hours after Aβ injections, and brains were removed after transcardial perfusion with a 0.9% saline and 4% formaldehyde. Brains were post-fixed in 4% formaldehyde overnight, cryoprotected by immersion in 20% sucrose for 24 hours, and stored at -80°C.

One hemisphere (forebrain) was removed after saline perfusion, frozen, and stored at -80°C for biochemical studies. The remaining hemisphere was post-fixed in 4% formaldehyde, cryoprotected in sucrose, and cryostat sectioned into 30 μm coronal sections for immunostaining. Immunostaining was performed with 30 μm coronal sections as described previously [25, 35]. Microglia were stained using Iba-1 antibody, Aβ plaques were stained with 3D6 antibody and calbindin expression was detected with Calbindin D-28k antibody. Primary antibody staining was visualized with suitable goat anti-IgG antibody conjugated with either Alexa Fluor 594 or 488. Brain sections were mounted on cover slips with DAPI-labeled mounting media (Vectashield) to facilitate recognition of brain structures. Negative controls were prepared by omitting the primary antibodies. Microscope imaging settings were kept uniform for all samples. Microglial morphology was analyzed in hippocampal CA1 and DG areas and in perirhinal cortex, with the exception of Aβ-injected brains, where microglial morphology was evaluated in 250 × 200 μm area starting 100 μm lateral to the needle track. Microglial activation was scored according to morphology and cell number (Table 1), as modified from [25]. Calbindin expression was determined by measuring the mean optical density in the designated, uniform-sized regions of interest with the ImageJ program (NIH). Values were measured on three comparable sections from each mouse, background values were subtracted, the resulting values averaged to give one value per mouse. For cytokine assays (Milliplex multiplex assays, Millipore) the forebrain hemispheres were homogenized 1:3 weigh to volume in M/PIER Mammalian Protein reagent (Thermo Scientific) with complete protease inhibitor (Sigma), following by centrifugation. Cytokine levels determined using standards in each assay plate, and values were normalized to protein content of the supernatants.

The lysates used for cytokine assay were further processed with guanidine buffer. ELISAs were performed as described [36] and normalized to total protein content. We used antibodies that recognize species referred to as Aβ_1-42 and Aβ_1-X (Elan Pharmaceuticals). The Aβ_1-42 ELISA detects only Aβ_1-42, and the Aβ_1-X ELISA detects Aβ_1-40, Aβ_1-42, and Aβ_1-43, as well as C-terminally truncated forms of Aβ containing amino acids 1-28.

Novel object recognition was tested in a white square plastic chamber 35 cm in diameter under a red light, as previously described [37]. Mice were transferred to the test room and acclimated for at least 1 hour. On the first day, mice were first habituated to the testing arena for 15 minutes and then each mouse was presented with two identical objects in the same chamber and allowed to explore freely for 10 min a training. On the second day, mice were placed back into the same arena for the 10 min test session, during which they were presented with an exact replica of one of the objects used during training (familiar object) and with a novel, unfamiliar object of different shape and texture. Object locations were kept constant during training and test sessions for any given mouse. Arenas and objects were cleaned with 70% ethanol between each mouse. Frequency of object interactions and time spent exploring each object was recorded with an EthoVision video tracking system (Noldus Information Technology, Leesbug, VA). Frequency of object interactions was used for analyses.

Spatial learning and memory were tested by the Morris Water Maze test, using a circular pool (122 cm in diameter, filled with opaque water at 24°C as describe previously [25, 35]. The mice were trained first to locate a platform with a visible cue (days 1 - 2), and then to locate a hidden platform (days 3 - 5) using large spatial cues in the room. The platform was moved to a new quadrant in each session during the visible platform cue training. The platform remained in the same quadrant throughout all the sessions during hidden platform training. The mice received two training sessions per day for five consecutive days. Each session consisted of three one-minute trials with a 10-minute inter-trial interval. The interval between the two daily sessions was 3 hours. Once the mice located the platform they were allowed to remain on it for 10 seconds. Mice that failed to find the platform within one minute were manually placed on the platform for 15 seconds. Time to reach the platform (latency), distance traveled (path length), and swim speed (velocity) were recorded with a video tracking system (Noldus).

For in vivo studies, the "n" denotes the number of mice in each group, and for cell culture studies the "n" denotes the number of independent experiments, each performed in triplicate or quadruplicate. All data are expressed as the mean ± SEM. Microglial morphological changes were evaluated with the Kruskal-Wallis test followed by the Dunn's test for multiple group comparisons. Data form Morris Water Maze test was analyzed by repeated measures one-way ANOVA. All other data were compared with ANOVA followed by the Bonferroni's test for multiple group comparisons.

The hAPP_J20 mouse expresses human amyloid precursor protein with AD-linked mutations [28]. The hAPP_J20 mice were crossed with PARP-1 ^-/- mice to evaluate the effects of PARP-1 gene deletion in this mouse model of AD. Spatial memory decline in hAPP_J20 mice correlates with loss of calbindin in the hippocampus [38]. A loss of calbindin in the hAPP mouse hippocampus was likewise observed in the present study (Figure 1A). This loss was attenuated in the hippocampal CA1 pyramidal layer of the hAPP_J20/PARP-1 ^-/- mice, but not in the dentate gyrus (Figure 1). Cognitive testing confirmed deficits in the hAPP_J20 mice as assessed by both the novel object recognition test and the Morris water maze test of spatial memory (Figure 2). The hAPP_J20/PARP-1 ^-/- mice performed better than the hAPP_J20 mice in the novel object recognition test, but not in the Morris water maze test (Figure 2).

The hAPP_J20 mice exhibit Aβ accumulation and scattered amyloid plaque formation by age 6 months [28]. These mice also show accumulation of amoeboid microglia at the amyloid plaques, and increased number of activated microglia throughout cortex and hippocampus (Figure 3). Despite comparable levels of Aβ accumulation in hAPP_J20 and hAPP_J20/PARP-1 ^-/- mice (Figure 3E), microglial activation was reduced in the hAPP_J20/PARP-1 ^-/- mice, in both amyloid plaques and in non-plaque areas (Figure 3). The total number of microglia was not statistically different between genotypes, in either amyloid plaque areas (hAPP_J20 vs. hAPP_J20/PARP-1 ^-/- ; 7.06 ± 0.94 vs. 6.22 ± 1.36 cells per mm²) or in non-plaque areas (Figure 3).

Cytokine levels in the hAPP_J20 mouse brains were not significantly different than in wt brains, but some cytokines were altered in the PARP-1 ^-/- and the hAPP_J20/PARP-1 ^-/- brains (Table 2).

We considered the possibility that ageing hAPP_J20 mice might express other factors, in addition to Aβ, that promote microglial activation. To directly determine the effects of PARP-1 deficiency on Aβ-induced microglial activation, we injected oligomeric Aβ directly into the hippocampus of wt and PARP-1 ^-/- mice. The Aβ injections induced soma enlargement and process retraction characteristic of activated microglia, and also increased microglial number in the area of injection (Figure 4). These changes were evident within 6 hours of the Aβ injections and were restricted to the area where Aβ was detected, i.e. ~500 μm from the injection needle track. In contrast, mice injected with vehicle (saline) or with a control, reverse-sequence Aβ (rAβ) showed microglial activation only in the immediate vicinity of the needle track lesion. Aβ injected into either PARP-1 ^-/- mice or wt mice treated with the PARP-1 inhibitor PJ34 produced substantially less microglial activation than Aβ injected into untreated wt mice (Figure 4).

Results of the studies presented above suggest that the protective effects of PARP-1 deficiency are attributable to attenuated activation of PARP-1 ^-/- microglia. However, since PARP-1 ^-/- mice also lack PARP-1 in neurons, astrocytes, and other cell types, it is alternatively possible that the attenuated microglia response in these mice is secondary to effects of PARP-1 gene deletion in other cells. We therefore used cell cultures to assess the direct effects of PARP inhibition on microglia. Aβ stimulation of wt microglia induced transformation to either the fully activated amoeboid appearance or to a partially activated morphology, with enlarged soma and fewer, thickened processes. By contrast, PARP-1 ^-/- microglia retained the resting, ramified morphology, as did microglia of either genotype treated with vehicle or with the control peptide, rAβ (Figure 5). Microglial proliferation and viability were not affected by Aβ incubation (not shown). A rapid accumulation (within 1 hour) of poly(ADP-ribose) (PAR) was detected in Aβ-stimulated wt microglia, indicating enzymatic PARP-1 activity. The accumulation of PAR was blocked by co-incubation with the PARP inhibitor, PJ34 (Figure 5). PJ34 also blocked morphological transformation in microglia treated with Aβ exposure, supporting a requisite role for microglial PARP-1 activity in this process (Figure 5).

Microglial activation by Aβ and other stimuli can promote neuronal death [34, 39‐41]. We evaluated the role of PARP-1 in microglial neurotoxicity using neuron-microglia co-cultures. Twenty-four hours incubation with 5 μM Aβ caused no significant cell death in neuron monocultures, but killed more than 50% of neurons cultured with wt microglia. The microglia-mediated Aβ toxicity was abolished in cultures treated with the PARP-1 inhibitor, PJ34, and in wt neurons co-cultured with PARP-1 ^-/- microglia (Figure 6).

The transcription factor NF-κB is involved in many aspects of microglial inflammatory responses [42], and PARP-1 regulates the transcriptional activity of NF-κB [15, 19]. Microglia cultures were transfected with an NF-κB-driven eGFP reporter gene [34] to evaluate the effects of Aβ and PARP-1 on NF-κB transcriptional activation in microglia. Aβ produced a large increase in the number of microglia expressing eGFP when assessed at either 90 minutes or 24 hours, and this increase was prevented by PARP inhibition (Figure 7A,B). Nitric oxide release and TNFα release are both regulated by NF-κB in myeloid cells [40, 43]. Accordingly, microglial release of NO and TNFα were found to be stimulated by Aβ, and blocked by the NF-κB inhibitor, BAY 11-7082 [44]. The release was also blocked by the PARP-1 inhibitor PJ34 and in PARP-1 ^-/- cells (Figure 7C,D). PJ34 and BAY 11-7082 also reduced microglial release of NO and TNFα in the absence of Aβ stimulation although basal release was not reduced in PARP-1 ^-/- microglia (data not shown). Aβ stimulation also increased release of other NF-κB regulated cytokines (KC, RANTES, MCP-1and MIP-1β; Table 3). The magnitude of increase was reduced by PARP-1 abrogation, but the statistical significance was not reached or was lost after correction for the multiple group comparisons (Table 3).

Activated microglia can also release, in addition to neurotoxic agents, several cytokines and trophic factors that can promote neuronal survival [8, 45‐47]. In particular, vascular endothelial growth factor (VEGF) and transforming growth factor β (TGFβ) are released by microglia [48‐50] and have beneficial effects in experimental AD ([51, 52], but see also [53]). Here, Aβ was found to reduce microglial release of both VEGF and TGFβ. This reduction was reversed by inhibitors of PARP-1 and NF-κB (Figure 8). These treatments also increased basal VEGF and TGFβ release (not shown).

We examined the possibility that the reduced microglial activation produced by PARP-1 inhibition might also result in reduced clearance of Aβ peptides, using FAM-labeled Aβ. Cultured microglia rapidly engulfed and accumulated the FAM-Aβ peptides, and this was unaffected by PARP-1 inhibition or PARP-1 ^-/- genotype (Figure 9). Of note, PARP-1 ^-/- microglia with engulfed Aβ peptide maintained the resting, ramified morphology, unlike the wt microglia (Figure 9C).