Docosahexaenoic Acid-Derived Neuroprotectin D1 Induces Neuronal Survival via Secretase- and PPARγ-Mediated Mechanisms in Alzheimer's Disease Models

Neuroprotectin D1 (NPD1) is a stereoselective mediator derived from the omega-3 essential fatty acid docosahexaenoic acid (DHA) with potent inflammatory resolving and neuroprotective bioactivity. NPD1 reduces Aβ42 peptide release from aging human brain cells and is severely depleted in Alzheimer's disease (AD) brain. Here we further characterize the mechanism of NPD1's neurogenic actions using 3xTg-AD mouse models and human neuronal-glial (HNG) cells in primary culture, either challenged with Aβ42 oligomeric peptide, or transfected with beta amyloid precursor protein (βAPP)sw (Swedish double mutation APP695sw, K595N-M596L). We also show that NPD1 downregulates Aβ42-triggered expression of the pro-inflammatory enzyme cyclooxygenase-2 (COX-2) and of B-94 (a TNF-α-inducible pro-inflammatory element) and apoptosis in HNG cells. Moreover, NPD1 suppresses Aβ42 peptide shedding by down-regulating β-secretase-1 (BACE1) while activating the α-secretase ADAM10 and up-regulating sAPPα, thus shifting the cleavage of βAPP holoenzyme from an amyloidogenic into the non-amyloidogenic pathway. Use of the thiazolidinedione peroxisome proliferator-activated receptor gamma (PPARγ) agonist rosiglitazone, the irreversible PPARγ antagonist GW9662, and overexpressing PPARγ suggests that the NPD1-mediated down-regulation of BACE1 and Aβ42 peptide release is PPARγ-dependent. In conclusion, NPD1 bioactivity potently down regulates inflammatory signaling, amyloidogenic APP cleavage and apoptosis, underscoring the potential of this lipid mediator to rescue human brain cells in early stages of neurodegenerations.


Introduction
Alzheimer's disease (AD) is a neurodegenerative disease characterized by progressive cognitive impairment and, at the cellular level, by synaptic damage, intracellular neurofibrillary tangles and beta-amyloid precursor protein (bAPP) processing dysfunction that leads to overabundance of the 42 amino acid amyloid-beta (Ab42) peptide. Ab42 promotes neuroinflammation, synaptic toxicity, and apoptosis, and it transitions extracellularly from an oligomer to an aggregate that, in turn, become a major component of senile plaques [1][2][3][4][5][6]. Ab42 peptides are generated from bAPP via tandem cleavage by beta-and gamma-(b-and c-) secretases; alternatively an alpha-secretase distintegrin and metalloproteinase 10 (ADAM10) cleaves bAPP to yield a soluble form of bAPP, sAPPa, via the non-amyloidogenic or neurotrophic pathway. Docosahexaenoic acid (DHA; C22:6), an omega-3 essential fatty acid family member, is enriched in central nervous system (CNS), synaptic and other cellular membranes as an acyl chain of membrane phospholipids. DHA is involved in the building and function of the CNS, as well as synaptogenesis, cognition, neuroprotection, synaptic function and vision [7][8][9][10]. Current clinical trials favor a role for DHA in slowing cognitive decline in elderly individuals without dementia but not for the prevention or treatment of dementia, including AD [11,12]. Deficiencies in DHA biosynthesis by the liver correlate with cognitive impairment in AD patients [13], supporting the significance of the liver supply of DHA to the CNS in neurodegenerative diseases [13,14]. In AD transgenic mice dietary DHA restores cerebral blood volume, reduces Ab deposition, and ameliorates Ab pathology [15,16].
In the present study, we assessed DHA and NPD1 abundance in control and aged 3xTg-AD mouse hippocampus and used aging human neuronal-glial (HNG) primary cells to characterize NPD1 bioactivity on: neuroinflammatory events and apoptosis; to test the mechanism of NPD1-mediated regulation of Ab42 secretion; and to assess the significance of PPARc in the homeostatic bioactivity of NPD1. Here we provide evidence that, besides protecting against Ab42-induced neurotoxicity via anti-inflammatory and anti-apoptotic bioactivity, NPD1 down-regulates the amyloidogenic processing of bAPP, thus reducing Ab42 production. Moreover, NPD1 anti-amyloidogenic action through selective targeting of both the aand b-secretase-mediated processing of bAPP and anti-amyloidogenic action are exerted through PPARc receptor activation.

Materials and Methods
Studies and procedures were performed according to National Institutes of Health and Canadian Council on Animal guidelines, and animal protocols were approved by the Institutional Animal Care and Use Committee at the Louisiana State University Health Sciences Center, New Orleans (IACUC #2705, IBC# 08126 and 082303), and by the Laval University Animal Ethics Committee (approval ID = 07-113 and 07-061).

Plasmid Constructs and Transient Transfection of HNG Cells
Plasmid containing APP695 cDNA bearing the Swedish mutation APP sw (Swedish double mutation APP695 sw , K595N, M596L) was a generous gift from Dr. T Golde of the Mayo Clinic (Jacksonville, FL). cDNA clones of full length hBACE1 genes were from Open Biosystem (Huntsville, AL). HNG cells were plated in 6well plates at 80% confluence and transiently transfected using Fugene HD transfection reagent (Roche Applied Science, Indianapolis, IN) with 2 mg per well of hAPP695 sw plasmid DNA alone or together with pEGFP (green fluorescent protein; BD Biosciences-Clontech), hPPARc, or hBACE1 at a DNA (mg):reagent (ml) ratio of 1:3. After 24 h, cells were typically incubated with 0, 50, 100 or 500 nM NPD1 or vehicle for 48 h before assay.
Small Interfering RNA-mediated Gene Silencing HNG cells were transfected with predesigned siRNA (Santa Cruz Biotechnology) to knock down human ADAM9 or ADAM10 mRNA. HNG cells over-expressing bAPP sw were transfected with a total of 60 pmol of ADAM9, ADAM10 or control siRNA using Lipofectamine 2000 transfection reagent (Invitrogen) and cultured for 24 h. The medium was replaced with a fresh one containing 500 nM of NPD1 and cells were cultured for another 48 h before assay.

Ab42 Oligomer Preparation
Ab42 peptides were initially solubilized in hexafluoroisopropanol (HFIP) (Sigma), aliquoted, and stored at 220uC as an HFIP film [42]. After evaporating HFIP, aliquoted peptide was resuspended with DMSO to 5 mM and diluted with phenol red free F12 media (Invitrogen) to a concentration of 100 mM. Peptide for the oligomer preparation was incubated at 4uC for 24 h prior to use [42]. The oligomeric status of Ab was verified by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE; Figure S1).

Immunocytochemistry and Imaging Analysis
HNG cells cultured on 8-well chamber slides (BD Biosciences, San Jose, CA) were fixed with 4% paraformaldehyde, then permeabilized and blocked with 0.125% Triton X-100 (Sigma) and 2% normal goat serum (GE Healthcare) in PBS at room temperature (RT) for 1 h. Cells were incubated overnight at 4uC with antibodies for b-tubulin III, GFAP, COX-2, B94 or APP-NT. Cells were washed 3 times with PBS and incubated for 3 h at RT with secondary antibodies conjugated with Cy3 or FITC fluorescein. After washing and drying, slides were applied with mounting medium (Vector Laboratories, Burlingame, CA) and observed under Zeiss Axioplan Inverted Deconvolution Fluorescent Microscope (Carl Zeiss, Oberkochen, Germany). Positivelystained cells were quantified using the manual counter function of the NIH ImageJ software.

Mediator lipidomic analysis
Lipids were extracted by homogenization of cells or tissues in chloroform/methanol and stored under nitrogen at 280uC [17,18,36]. For quantification, lipid extracts were supplemented with deuterated labeled internal standards, purified by solid-phase extraction, and loaded onto a Biobasic-AX column (Thermo-Hypersil-Keystone; 100 mm 62.1 mm; 5-mm particle sizes) run with a 45-min gradient protocol, starting with solvent solution A (40:60:0:01 methanol:water:acetic acid, pH 4.5; 300 ml/min); the gradient typically reached 100% solvent B (99.99:0.01 methanol:acetic acid) in 30 min, and was then run isocratically for 5 min. A TSQ Quantum (Thermo-Finnigan) triple quadrupole mass spectrometer and electrospray ionization was used with spray voltage of 3 kV and N 2 sheath gas (35 cm 3 /min, 350uC). Parent ions were detected on full-scan mode on the Q1 quadrupole. Quantitative analysis was performed by selective reaction monitoring. The Q2 collision gas was argon at 1.5 mTorr, and daughter ions were detected on Q3. Selected parent/daughter ion pairs for NPD1 and unesterified DHA were typically 359/153 m/z and 327/283 m/z, respectively. Calibration curves for NPD1 and DHA (Cayman Chemical) were acquired; NPD1 was generated via biogenic synthesis using soybean lipoxygenase and DHA, purified by HPLC, and characterized by LC-PDA-ESI-MS-MS according to reported biophysical criteria [9,17].
MTT cell viability assay, Hoechst staining, TUNEL assay and caspase-3 activity assay Cell viability was measured by 3-[4,5-dimethylthiazol-2-yl]-2,5diphenyl tetrazolium bromide (MTT) reduction assay (Sigma). HNG cells were incubated with 5 mM of Ab42 in the absence or presence of 50 nM of NPD1 for 48 h. MTT was added to a final concentration of 0.5 mg/ml and incubated for 2 h. Medium was then removed and equal volumes of isopropanol were added to dissolve the resulting formazan crystals. Absorbance was spectrophotometrically measured with a SpectraMax Microplate Reader (Molecular Devices, Sunnyvale, CA) at 570 nm. HNG cells were further incubated with 2 mM Hoechst 33258 (Invitrogen) for 45 min at 37uC before imaging. Cells were then viewed by using a Nikon DIAPHOT 200 microscope under UV fluorescence. Images were recorded by a Hamamatsu Color Chilled 3CCD camera and PHOTOSHOP 7.0 software. Positively stained cells were counted manually using ImageJ software. The apoptotic nuclei containing free 39-OH termini were detected using DeadEnd Fluorometric TUNEL Kit (Promega, Madison, WI). Samples were analyzed under a Zeiss Deconvolution Microscope. Caspase-3 activity from cell lysates was detected using Caspase 3 Colorimetric Assay Kit (Sigma). The absorbance was measured at 405 nm using a SpectraMax Microplate Reader.

Total RNA Extraction and RT-PCR
HNG cells were lysed and total RNA was extracted with TRIzol (Invitrogen). RNA quality and quantity were analyzed by using a 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA). 28S/ 18S ratio for each RNA sample was typically greater than 1.8. For reverse transcription, a Superscript III First-Strand SuperMix (Invitrogen) was used. 1 mg of total RNA was used as a template to react with 10 ml of 26RT Reaction Mix and 2 ml RT Enzyme Mix. Final total volume was 20 ml. Samples were incubated at 25uC for 10 min and then 50uC for 30 min. Reactions were stopped by heating to 85uC for 5 min, and RT product was amplified with Phusion High Fidelity  cycles, and final extension at 72uC for 7 min. PCR products were further analyzed on 1.5% agarose gels; relative band intensity was quantified using Quality One software (Invitrogen).

SDS-PAGE and Western Blotting
Conditioned media were collected from cultured HN cells after various treatments and protease inhibitor cocktail (Sigma) was added to 1% final concentration [9,27]. Cells were then washed twice with ice-cold DPBS and lysed and harvested in RIPA buffer (Sigma) supplemented with 2% protease inhibitor cocktail, then centrifuged at 10,0006 g for 15 min at 4uC. Supernatants were collected and quantified using Bio-Rad's (Hercules, CA) DC Protein Assay kit. 30 mg of cell lysate or 20 ml -of conditioned media were electrophoresed on 4-15% Tris-HCl gradient gels at 100 V for 80 min or 10-20% Tris-Tricine gels for the detection of CTFs at 50 mA for 120 min. Proteins were transferred to an Immobilon FL PVDF membrane (Millipore, Billerica, MA) at 100 V for 60 min. Membranes were incubated with primary antibody overnight at 4uC, followed by incubation with IRDye 800 or Alexa 680-conjugated secondary antibodies for 5 h at RT. After repeated washing with Tris-buffered saline, the membrane was then visualized by the Odyssey Infrared Imaging System (LI-COR, Lincoln, NE).

Sandwich ELISA Analysis of TNFa and Ab42
Secreted TNF-a, Ab42 and total Ab were detected using a Human TNF ELISA Kit (BD Biosciences, San Jose, CA), a human amyloid b 42 ELISA kit (Sigma) and a human amyloid b (1-x) assay kit (American Research Products), respectively. After reactions, the plates were immediately measured at 450 nm by a SpectraMax Microplate Reader.

Human Preadipocyte Differentiation Assay
Human preadipocytes maintenance and differentiation procedures were performed according to the manufacturer's instructions with modifications (Zen-Bio, Research Triangle Park, NC). Briefly, upon the initiation of the differentiation assay, preadipocytes were incubated in adipocyte medium supplemented with IBMX (0.5 mM) and NPD1, DHA or vehicle. A concentration range of 0.1-5 mM of each lipid was used. After 3-day incubation, the cell medium was replaced with the adipocyte medium without IBMX. Eight days after vehicle or lipid treatment, the media was removed and the cells were fixed with formalin (7% formaldehyde in PBS). Cells were then stained with Oil Red O (Sigma, Saint Louis, MS) and pictures were taken with a Nikon Eclipse TS100 inverted microscope (Nikon USA, Melville, NY). The Oil Red O-stained total lipid was then eluted with 100% isopropanol and quantified by measuring OD value at 500 nm with a SpectraMax Microplate Reader.

Cell-based PPARc Transactivation Assay
The two plasmids used for the transactivation assay (PPARc-GAL4 and MH100-tk-luc) were kindly provided by Dr. Ronald Evans of Salk Institute (La Jolla, CA) [32]. Luciferase assay was performed using Promega's Luciferase Assay System. Light units

Statistical Analysis
All experiments were repeated at least three times using independent culture preparations. Data are presented as mean 6 S.E. Quantitative data were statistically analyzed by one-way analysis of variance (ANOVA) followed by pair-wise comparisons using the Fisher's least significant difference test. A p,0.05 was considered significant.

DHA and NPD1 deficits in 3xTg-AD mouse hippocampus
DHA and NPD1 levels were assayed in the hippocampus of 3xTg-AD mice, harboring the PS1 (M146V), APP (Swe) and tau (P301L) human transgenes that model several human AD features [38,39]. DHA and NPD1 levels were analyzed using LC-PDA-ESI-MS-MS-based lipidomic analysis as previously described ( Figure 1) [9,18]. Both DHA and NPD1 showed age-related changes in 4-month old versus 12-13 month old 3xTg-AD animals. DHA concentration in the hippocampus was found to be reduced 2-fold between 4-month old control versus 4-month 3xTg-AD animals, and 3-fold between 4 and 12-13 month old control animals ( Figure 1A). NPD1 in the hippocampus showed dramatic reductions both in aging control animals; a 12-fold reduction between 4 and 12-13 month controls and a 3-fold reduction between 4 and 12-13 month 3xTg-AD mice ( Figure 1B).

NPD1 protects HNG cells from Ab42-induced apoptosis
Phase contrast and immunofluorescence of differentiated HNG cells expressing the neuronal marker b-tubulin III and the astrocyte marker glial fibrillary acidic protein (GFAP) revealed neuronal-glial co-cultures containing about 50% neurons under these conditions (Figure 2). NPD1 was shown to counteract Ab42 oligomer-induced apoptosis in HNG cells using MTT, Hoechst 33258 staining, TUNEL and assay of caspase-3 activity ( Figure 3A-G). These assays showed that over 48 h Ab42 oligomer triggers about 50% cell death with concomitant nuclear compaction and striking apoptotic changes ( Figure 3). Ab42 peptides also enhanced caspase-3 activity at least 6-fold, an effect that was reduced in the presence of NPD1. Co-incubation of 50 nM NPD1 with Ab42 oligomer resulted in enhanced cell viability and attenuation of Ab42 peptide-mediated apoptosis and cytotoxicity ( Figure 3).

NPD1 down-regulates Ab42 oligomer-induced proinflammatory gene expression
Our previous DNA microarray-based analysis suggested antiinflammatory bioactivity of NPD1 in HNG cells, as shown by their attenuation of Ab42 peptide-induced elevation of the proinflammatory genes COX-2, TNF-a and B94 [9]. Here we extended these studies by exploring NPD1 actions at both the mRNA and protein levels using RT-PCR, Western assay, ELISA assay and immunocytochemistry (Figures 3 and 4). The relative basal abundance of TNF-a mRNA was low, B94 mRNA increased during incubation at 6 h, and constitutive expression of COX-2 mRNA occurred during incubation. Ab42 increased mRNA abundance of COX-2, TNF-a and B94 at 3, 6 and 12 h ( Figure 4A,B). COX-2 mRNA stood out because it displayed immediate early-inducible gene behavior upon Ab42 peptide exposure [9]. Protein expression of cellular COX-2, TNF-a secreted to the incubation media, and immunocytochemistry of COX-2 and B94 showed Ab42-stimulated enhancement and NPD1 (50 nM) markedly reduced Ab42 oligomer-stimulated mRNA increases as well as COX-2, TNF-a and B94 protein expression ( Figure 4C). NPD1 therefore elicits potent downregulation in the expression of a specific set of pro-inflammatory and pro-apoptotic genes known to be up-regulated in AD hippocampus and in stressed HNG cell models of AD [9,41,[43][44][45][46]. Messenger RNA, Western, ELISA and immunohistochemisty data are presented in Figures 4 and 5. NPD1 represses amyloidogenic processing of bAPP with concomitant stimulation of non-amyloidogenic processing Ab42-peptides are secreted from human brain cells as they age or in response to physiological stress [4,9,27,47,48]. The processing of bAPP holoenzyme and secretion of bAPP fragments is controlled in large part by alpha-, beta-and gamma-(a-,b-and c-) secretases [3,4]. To assess the effects of NPD1 on secretase-mediated Ab42 peptide generation, we used HNG cells transiently-transfected with bAPP sw and assayed for the abundance of the a-secretasegenerating enzymes precursor-ADAM10 (pro-ADAM10), mature-ADAM10 (m-ADAM10), b-amyloid cleavage enzyme (BACE1) and the gamma-secretase presenilin-1 (PS1) (Figure 6). Western blot analysis revealed that the steady-level of BACE1 was reduced by 500 nM of NPD1. Meanwhile, the active and mature form of ADAM10 (m-ADAM10), the putative a-secretase, was dosedependently increased in response to NPD1. We did not find changes in the pro-ADAM10, the inactive precursor or in the mRNA abundance of ADAM10 (data not shown; Figure 6). The undergoing changes in these two secretases are in agreement with alterations in Ab42 peptide abundance, and in other cleavage products of bAPP (Figure 7). Interestingly, both m-ADAM10 and BACE1 levels were elevated in bAPP-over-expressing cells ( Figure 6). Presenilin 1 (PS1), the main catalytic component for c-secretase, remains unchanged after different bAPP sw or NPD1 treatments ( Figure 6). This same pattern was also seen in their Cterminal counterparts, CTFb and CTFa; importantly, no change was observed in the steady-state level of the neural cell abundant bAPP (holo-bAPP; see Figure 7A). NPD1-mediated up-regulation of m-ADAM-10 and down-regulation of BACE1 was apparent with maximal effect at 500 nM, the highest concentration used in these experiments (Figures 6 and 7).

Quantification of bAPP Fragments
As Ab42 peptide generation is regulated by differential bAPP processing, NPD1-mediated Ab42 peptide reduction is due to altered bAPP processing, and thereby altered bAPP cleavage products should confirm these catabolic outcomes. To test this idea, we used HNG cells over-expressing bAPP sw , and measured levels of N-terminal (sAPPa and sAPPb sw ) and C-terminal fragments (CTFa and CTFb) of bAPP as well as holo-bAPP protein upon exposure to increasing concentrations of NPD1. We show that NPD1 lowers sAPPb sw secretion and elevates sAPPa in a dose-dependent manner (Figure 7). This observation is paralleled by a decrease in CTFb and an increase in CTFa in the same cellular fractions and a significant 3.4-fold increase in mADAM10 (Figure 7). Silencing of ADAM9 and ADAM10 and overexpression of BACE1 Collectively, these data suggests the participation and modulation of BACE1 and ADAM10 activities in NPD1-mediated regulation of bAPP processing. Just like ADAM10, ADAM9 is also endowed with a-secretase activity [6,49], and changes in BACE1 abundance may also contribute to Ab42 peptide reduction. We therefore investigated whether ADAM 9, ADAM10 and BACE1 are essential to NPD1's regulation of bAPP processing by knocking down siRNA-targeted ADAM9 and ADAM10 genes. We also over-expressed BACE1 by transfecting HNG cells with a plasmid bearing the human BACE1 full length cDNA. We then measured total bAPP and other bAPP cleavage fragments in the presence of NPD1 with or without ADAM 9 siRNA or ADAM10 siRNA knockdown or BACE1 over-expression. As seen in Figure 8, when compared to controls (control siRNA) no changes occurred in ADAM9 siRNA group while ADAM10 siRNAmediated knockdown almost completely abrogated the induction of sAPPa and CTFa by NPD1. Similarly, moderate overexpression of BACE1 overturned NPD1-induced reduction in Ab42 peptides along with sAPPb sw and CTFb fragments. These results in combination with those shown in Figures 6 and 7 strongly suggests that NPD1's regulatory action targeting bAPP processing may be mediated in part through coordinated upregulation of the a-secretase ADAM10 and down-regulation of BACE1 enzymatic activity.

NPD1 is a PPARc activator
PPARc is a key anti-inflammatory and Ab-lowering mediator, and several polyunsaturated fatty acids and their derivatives are ligands for PPARc. Thus, we asked whether NPD1 influences PPARc actions, and if this could be related to its neuroprotective bioactivity. We first tested NPD1 as a potential PPARc activator using primary human adipocyte differentiation. PPARc is an adipogenesis modulator, and PPARc agonists induce adipocyte differentiation. Adipogenesis assay was used for screening potential PPARc-active compounds. Primary human pre-adipocytes were treated with 0.1, 1 and 5 mM of NPD1 or DHA during differentiation induction (see Figure 9). Ligand-induced differentiation was assessed by Oil Red O staining. NPD1 led to enhancement of differentiation in the primary human preadipocytes, while equivalent doses of its precursor DHA displayed little adipogenic activity, suggesting that NPD1 does display PPARc activity (Figure 9 and Figure S2). To further evaluate the activity of NPD1, we used a cell-based PPARc transactivation reporter assay. HNG cells co-transfected with hPPARc-GAL4 and MH100-tk-luc were incubated with increasing concentrations (0.1, 1.0, 5.0 and 10 mM) of NPD1 or DHA for 24 h. NPD1, but not its precursor DHA, increased reporter activity in a dose-dependent manner indicating that NPD1 acts as an activator of an inducible PPARc response ( Figure 10A).

The anti-amyloidogenic effect of NPD1 is PPARcdependent
We further examined whether PPARc is involved in the regulation by NPD1 of bAPP processing. First, we studied the effect of PPARc on Ab42 peptide production in HNG cells overexpressing bAPP sw by either transiently-transfecting PPARc cDNA or using the PPARc agonist, rosiglitazone. The efficiency of transfection and subcellular localization of both proteins were monitored by immunofluorescence and Western blotting. The majority of PPARc expression was found to be associated with the HNG cell nuclei (Figure 10B,C). In both PPARc-transfected and PPARc agonist-treated HNG cells, we observed a decrease in the amount of secreted total Ab and Ab42. The decrease was comparable to that conferred by NPD1 treatment. To determine whether PPARc is required in this action, HNG cells were also incubated with the PPARc antagonist, GW9662. GW9662 reversed the Ab peptide reduction in NPD1-treated cells and in PPARc over-expression or PPARc agonist-treated cells as well ( Figure 11A). These results suggest that PPARc is required for NPD1's regulation of Ab release. To further define the action of PPARc on bAPP processing and its implication in the antiamyloidogenic effect of NPD1, we analyzed the levels of bAPP fragments using the above treatments. Similar to NPD1-induced reduction in sAPPb and CTFb, in cells over-expressing PPARc or treated with PPARc agonist, these two b-secretase cleavage products were substantially down-regulated ( Figure 11B,D,E). Just as in the case of Ab peptide release, this down-regulatory effect was reversed by the addition of PPARc antagonist in all relevant treatment groups ( Figure 11B,D,E). Note that GW9662 alone caused no changes in either sAPPb sw or CTFb ( Figure 11C). In contrast, unlike NPD1, PPARc overexpression or PPARc agonist did not modify the levels of sAPPa or CTFa. Nor did the PPARc antagonist abolish the NPD1-induced increase in these fragments ( Figure 11B,D,E). Meanwhile, no changes in holo-bAPP by PPARc were observed ( Figure 11B). These data suggest that PPARc is involved in NPD1's regulation via the bsecretase pathway but not via the a-secretase pathway. We next examined the levels of ADAM10 and BACE1, the putative aand b-secretase that are actively involved in NPD1's modulation of APP processing. In agreement with the alterations in levels of bAPP fragments, PPARc activation reduced the steady-state level of BACE1 expression but did not affect ADAM10. PPARc antagonism abolished the NPD1-induced decrease in BACE1 but was not able to reverse the increase in mature ADAM10 level ( Figure 11B,F).

Discussion
DHA partially counteracts cognitive decline in the elderly [11]. Moreover, omega-3 essential fatty acid-rich diets are associated with a trend in reduced risk for MCI and with MCI conversion to AD, whereas DHA has been shown to be beneficial in transgenic AD models [8,10,11,16,50]. The 15-lipoxygenase-1-(15-LOX-1) DHA-derived NPD1 displays neuroprotective bioactivity in brain and retinal cells against various insults, including oxidative injury, ischemia-reperfusion and inflammation [9,17,18,[51][52][53]. Both AD brain [9] and the 3xTg-AD mouse exhibit reductions in DHA and NPD1 (Figure 1). In this study we further characterized the antiinflammatory and anti-apoptotic activity of NPD1 in co-cultures of HNG cells stressed with Ab42 oligomer, and studied the NPD1mediated modulation of aand b-secretase activity that resulted in reduced shedding of Ab42.
AD is marked by synaptic damage, neuronal atrophy and cell death in the hippocampus and entorhinal cortex [4,[54][55][56]. Neurotoxicity induced by Ab42 aggregates appears to drive microglial-mediated neuroinflammatory responses and apoptosis [3,4,50,57]. Oxidative stress, calcium overload, mitochondrial dysfunction and membrane impairments, along with activation of caspases and cell death are associated with Ab42 up-regulation [55]. We found that NPD1 induces HNG cell survival after Ab42oligomer-mediated stress and reduced Ab42-triggered apoptosis. NPD1 attenuated caspase-3 activation and decreased compacted nuclei and fragmented DNA [18,19] (Figure 3). These observations are in agreement with the NPD1-mediated up-regulation of anti-apoptotic Bcl-2, Bcl-xl and Bfl-1 expression and the decrease in the pro-apoptotic expression of Bax, Bad and Bik [9,18].
Neuroinflammatory neurodegeneration associated with Ab42 is an important contributory event to AD neuropathology [54,56]. In these experiments primary HNG cells were used, as human primary neurons do not survive well in the absence of glial cells [9,29] (Figure 2). While we cannot exclude the possibility that glial cells are providing some neuroprotective 'shielding', both neuronal and glial cells release cytokines when exposed to Ab42 that, in turn, activate more microglia and astrocytes that reinforce pathogenic signaling. NPD1 is anti-inflammatory and promotes inflammatory resolution [17,18,37,53]. In HNG cell models of Ab42 toxicity, microarray analysis and Western blot analysis revealed down-regulation of pro-inflammatory genes (COX-2, TNF-a and B94), suggesting NPD1's anti-inflammatory bioactivity targets, in part, this gene family [9]. These effects are persistent, as shown by time-course Western blot analysis in which protein expression was examined up to 12 h after treatment by Ab42 and NPD1.
Although counteracting Ab42-induced neurotoxicity is a promising strategy for AD treatment, curbing excessive Ab42 release during neurodegeneration is also desirable. DHA could lower Ab42 load in the CNS by stimulating non-amyloidogenic bAPP processing, reducing PS1 expression, or by increasing the expression of the sortilin receptor, SorLA/LR11 [8,21,41,58]. In contrast to a previous report by Green et al. [16] that suggested that Ab peptide reductions in whole brain homogenates of 3xTg AD after dietary supplementation of DHA were the result of decreases in the steady state levels of PS1, our experiments in primary HNG cells showed no effects of NPD1 on PS1 levels, but a significant increase in ADAM10 coupled to a decrease in BACE1 ( Figure 5). These later observations were further confirmed by both activity assays (Figures 6 and 7) and siRNA knockdown ( Figure 8). NPD1 reduces Ab42 levels released from HNG cells over-expressing APP sw in a dose-dependent manner. Our examination of other bAPP fragments revealed after NPD1 addition, a reduction in the b-secretase products sAPPb sw and CTFb occurred, along with an increase in a-secretase products sAPPa and CTFa, while levels of bAPP expression remained unchanged in response to NPD1. Hence these abundance-and activity-based assays indicate a shift by NPD1 in bAPP processing from the amyloidogenic to non-amyloidogenic pathway. Previously sAPPa has been found to promote NPD1 biosynthesis from DHA [9], while in the present study NPD1 works to stimulate sAPPa secretion, creating positive feedback and neurotrophic reinforcement. Secreted sAPPa's beneficial effects include enhanced learning, memory and neurotrophic properties [6]. NPD1 further down-regulated the b-secretase BACE1 and activated ADAM10, a putative a-secretase. Our ADAM10 siRNA knockdown and BACE1 over-expression-activity experiments confirmed that ADAM10 and BACE1 are required in NPD1's regulation of bAPP. NPD1 therefore appears to function favorably in both of these competing bAPP processing events.
PPARc activation leads to anti-inflammatory, anti-amyloidogenic actions and anti-apoptotic bioactivity, as does NPD1. Some fatty acids are natural ligands for PPARc, which have a predilection for binding polyunsaturated fatty acids [59][60][61]. Our hypothesis that NPD1 is a PPARc activator was confirmed by results from both human adipogenesis and cell-basedtransactivation assay (Figures 9 and 10). NPD1 may activate PPARc via direct binding or other interactive mechanisms [33,62]. Analysis of bAPP-derived fragments revealed that PPARc does play a role in the NPD1-mediated suppression of Ab production. Over-expressing PPARc or incubation with a PPARc agonist led to reductions in Ab, sAPPb and CTFb similar to that with NPD1 treatment, while a PPARc antagonist abrogated these reductions. Activation of PPARc signaling is further confirmed by the observation that PPARc activity decreased BACE1 levels, and a PPARc antagonist overturned this decrease. Thus, the antiamyloidogenic bioactivity of NPD1 is associated with activation of the PPARc and the subsequent BACE1 down-regulation. The difference between the bioactivity of NPD1 concentrations for anti-apoptotic and anti-amyloidogenic activities (50 nM vs. 500 nM) may be due to the different cell models used (i.e., Abpeptide stressed vs. bAPP sw -over-expressing HNG cells) and/or related mechanisms.
Although Ab-lowering effects of PPARc have been reported, the molecular mechanism of this action remains unclear. Induction of bAPP ubiquitination, which leads to enhanced bAPP degradation and reduced Ab peptide secretion, has been suggested [60]. Alternatively, Ab clearance might be involved, or regulation by PPARc may be due to enhancement of insulin sensitivity and increases in brain insulin degrading enzyme [59]. Our results suggest that decreases in BACE1 may be the cause for Ab reduction [27,63]. A reason for these conflicting reports may be that cell models and culture conditions used varied; in our study, we used HNG cells transiently over-expressing bAPP sw while previous reports employed cell lines using stable bAPP expression. Similar to the model of Sastre et al. [63], our cells underwent increases in ab overproduction. Excessive Ab causes inflammatory responses in both neuronal and glial cells [27]. Since inflammatory signaling plays a role in AD pathogenesis, we believe HNG cell cultures are a valuable model for Ab42 -mediated cellular actions. The fact that comparable results of our study were obtained at a much lower drug concentration (0.5 mM of rosiglitazone vs. 10-30 mM in previous reports) ( Figure 10) underscores the highly sensitive nature of HNG cells after bAPP transfection. It is still possible that PPARc may repress BACE1 by antagonizing activities of other transcription factors that promote BACE1 expression, such as STAT1, NF-kB and AP1 [64]. It is noteworthy that BACE1 expression in HNG cells was increased after bAPP over-expression. The fact that PPARc did not affect the levels of sAPPa and CTFa besides PPARc antagonist being unable to reverse NPD1-elicited increase in these fragments, clearly show that PPARc is not essential for NPD1's regulation on the nonamyloidogenic pathway. Further analysis of ADAM10 showed no change occurring in ADAM10 following PPARc activation, nor did PPARc antagonists affect NPD1-enhanced expression of mature ADAM10. Therefore, modulation by NPD1 of a-secretase and bAPP processing are independent of PPARc. ADAM10 is synthesized as an inactive zymogene and is processed to its mature form by cleavage of the pro-domain by pro-protein convertases (PPCs), such as furin and PC7 [65]. Other evidence also demonstrated that protein kinase C (PKC) and mitogen-activated protein (MAP) kinase, particularly extracellular signal-regulated kinases (ERK1/2), are involved in regulation of a-secretase activity [62,66,67]. No cross-talk between the PCs and PKC or MAP kinases has been reported. Since in our study only the mature ADAM10 was increased, it is likely that the PPCs are implicated in NPD1 actions.
PPARc antagonist GW9662 also failed to reverse the antiapoptotic effect of NPD1, indicating that PPARc is not implicated in NPD1 anti-apoptotic bioactivity ( Figure 10). NPD1 attained this neuroprotection at a concentration of 50 nM, at which its PPARc activity is far from physiologically relevant in the in vitro system. Other mechanisms have been proposed to explain DHA's anti-apoptotic and anti-inflammatory effects, including maintenance of plasma membrane integrity, activation of Akt signaling [68], and conversion into other derivatives [23,50]. These findings also provide clues for NPD1's potential targets. NPD1 inhibits NF-kB activation and COX-2 expression in brain ischemia-reperfusion [17], while Ab peptide-induced apoptosis is associated with ERK and p38 MAPK-NF-kB mediated COX-2 up-regulation [44]. Neuroprotection mediated by NPD1 may further involve components of signaling pathways upstream of NF-kB activation and DNA-binding [9].
Our results provide compelling evidence that NPD1 is endowed with strong anti-inflammatory, anti-amyloidogenic, and antiapoptotic bioactivities in HNG cells upon exposure to Ab42 oligomers, or in HNG cells over-expressing bAPP sw . These results suggest that NPD1's anti-amyloidogenic effects are mediated in part through activation of the PPARc receptor, while NPD1's stimulation of non-amyloidogenic pathways is PPARc-independent. Suggested sites of NPD1 actions are schematically presented in Figure 11. NPD1 stimulation of ADAM10 coupled to suppression of BACE1-mediated Ab42 secretion clearly warrants further study, as these dual secretase-mediated pathways may provide effective combinatorial or multi-target approaches in the clinical management of the AD process. Figure S1 Characterization of Ab peptide preparations using LMW-Western analysis. (A) Lanes 1 and 2 represent duplicate Ab peptide preparations prepared and analyzed by one of the authors (WJL), and (B) Ab peptide preparations prepared and analyzed completely independently by another one of the authors (YZ); both Ab peptide preparations were prepared and Figure 11. NPD1 promotes non-amyloidogenic, neurotrophic bioactivity via pleiotrophic mechanisms. Membrane esterified DHA is excised by phospholipase A2 (PLA2) to yield free DHA; in turn free DHA is 15-lipoxygenated to generate NPD1 which then enters a neuroprotective cycle. These events are mediated, in part, by inhibiting apoptosis, by blocking inflammatory signaling, by promoting cell survival and by shifting bAPP processing from an amyloidogenic into a neurotrophic, non-amyloidogenic pathway. BACE1 activity is suppressed and a-secretase (ADAM10) activity is stimulated, thus down-regulating Ab42 peptide release from membranes. Augmentation of BACE1 and ADAM10 by NPD1 may be mediated via other neuromolecular factors. We note that the ADAM10 cleavage product sAPPa further induces the conversion of free DHA into NPD1, thus constituting a positive, neurotrophic feedback loop. doi:10.1371/journal.pone.0015816.g011 analyzed according to the HFIP (hexafluoroisopropanol; 1,1,1,3,3,3-hexafluoro-2-propanol) preparative and gel analytical methods described by Stine et al., (J Biol Chem. 278:11612-22,2003). No higher order Ab fibrils are evident in either (A) or (B). Panel (C) shows relative toxicity of monomer, fibril and oligomer preps shown in (B) as analyzed using MTT [3-(4,5dimethylthiazol-2-yl)-2,5-diphenyltetrazolium; Invitrogen] cell viability assay.