YY-1224, a terpene trilactone-strengthened Ginkgo biloba, attenuates neurodegenerative changes induced by β-amyloid (1-42) or double transgenic overexpression of APP and PS1 via inhibition of cyclooxygenase-2

Ginkgo biloba has been reported to possess free radical-scavenging antioxidant activity and anti-inflammatory properties. In our pilot study, YY-1224, a terpene trilactone-strengthened extract of G. biloba, showed anti-inflammatory, neurotrophic, and antioxidant effects. We investigated the pharmacological potential of YY-1224 in β-amyloid (Aβ) (1-42)-induced memory impairment using cyclooxygenase-2 (COX-2) knockout (−/−) and APPswe/PS1dE9 transgenic (APP/PS1 Tg) mice. Repeated treatment with YY-1224 significantly attenuated Aβ (1-42)-induced memory impairment in COX-2 (+/+) mice, but not in COX-2 (−/−) mice. YY-1224 significantly attenuated Aβ (1-42)-induced upregulation of platelet-activating factor (PAF) receptor gene expression, reactive oxygen species, and pro-inflammatory factors. In addition, YY-1224 significantly inhibited Aβ (1-42)-induced downregulation of PAF-acetylhydrolase-1 (PAF-AH-1) and peroxisome proliferator-activated receptor γ (PPARγ) gene expression. These changes were more pronounced in COX-2 (+/+) mice than in COX-2 (−/−) mice. YY-1224 significantly attenuated learning impairment, Aβ deposition, and pro-inflammatory microglial activation in APP/PS1 Tg mice, whereas it significantly enhanced PAF-AH and PPARγ expression. A preferential COX-2 inhibitor, meloxicam, did not affect the pharmacological activity by YY-1224, suggesting that the COX-2 gene is a critical mediator of the neuroprotective effects of YY-1224. The protective activity of YY-1224 appeared to be more efficacious than a standard G. biloba extract (Gb) against Aβ insult. Our results suggest that the protective effects of YY-1224 against Aβ toxicity may be associated with its PAF antagonistic- and PPARγ agonistic-potential as well as inhibition of the Aβ-mediated pro-inflammatory switch of microglia phenotypes through suppression of COX-2 expression.


Background
Alzheimer's disease (AD) is the most common neurodegenerative disorder in the elderly and is associated with progressive memory loss and cognitive dysfunction [1][2][3][4]. Although the precise cause of the AD is unknown, βamyloid peptide (Aβ)-induced neurotoxicity, tau pathology, and neuroinflammatory responses by microglia are thought to contribute to the pathogenesis of AD [1,5]. Several studies have suggested that treatment with nonsteroidal anti-inflammatory drugs (NSAIDs) attenuates the loss of cognitive function and decreases the risk of developing AD [6,7]. NSAIDs act by inhibiting cyclooxygenase (COX), which is the rate-limiting enzyme for the conversion of arachidonic acid into inflammatory mediators, which include prostaglandin E2 (PGE2). COX-2 enzymatically mediates the inflammatory response, and its expression is significantly increased in the AD brain [8,9].
Platelet-activating factor (PAF) is a highly potent inflammatory mediator and a potential neurotoxin [10]. It plays an important role in excitotoxicity, production of free radicals and nitric oxide (NO), and regulation of pro-inflammatory cytokine genes [11][12][13][14]. The biological action of PAF is mostly mediated by binding to its G protein-coupled membrane-associated receptor (PAFR) [15]. In vitro studies have shown that activation of epidermal PAFR results in biosynthesis of COX-2 [16]. Both PAFR and COX-2 are involved in memory processing in vivo [17,18]. PAF is a short-lived molecule due to its rapid degradation by PAF acetylhydrolase (PAF-AH; EC 3.1.1.47) [19]. PAF-AH is an enzyme that hydrolyzes an acetyl ester at the sn-2 position of PAF, converting it into its inactive metabolite, 1-O-alkyl-sn-glycero-3-phosphocholine (lysoPAF) [20]. Three isoforms of PAF-AH were identified: plasma PAF-AH and PAF-AH II and I [21].
NSAIDs may regulate gene expression via their interaction with peroxisome proliferators-activated receptors (PPARs). There are three PPAR isoforms: PPAR α, β, and γ. PPARγ, in particular, has been implicated in inflammation and neurodegeneration [22]. The messenger RNA (mRNA) levels of PPARγ are increased in AD patients [23], suggesting that PPARγ may play an important role in modulating the pathophysiology of AD. The PPARγ isoform can be activated by NSAIDs, and its activation in the microglia suppressed the expressions of inflammatory cytokines, inducible NO synthase (iNOS), and COX-2 [24]. Aβ-induced activation of microglia was suppressed by a PPARγ agonist in vitro [24]. Furthermore, PPARγ agonists significantly decreased Aβ42 levels in vivo [25].
It has been demonstrated that a standardized G. biloba extract (Gb) contains flavonoids and terpene trilactones, and possesses free radical-scavenging and antioxidant activities [26,27]. Gb is commonly consumed as a dietary supplement for many disorders of the central nervous system (CNS), such as memory impairment, AD, and multi-infarct dementia [28][29][30][31]. Gb has been shown to protect against Aβ-induced neurotoxicity [32,33]. In addition, Gb showed anti-inflammatory activity through the antagonism of PAF [34,35]. Among the constituents of terpene trilactones, ginkgolides A, B, and C are highly selective and competitive PAFR antagonists [33].

Animals and drug treatment
The present study was performed in accordance with the Institute for Laboratory Research (ILAR) Guidelines for the Care and Use of Laboratory Animals. COX-2 (−/−)-and COX-2 (+/+)-mice were described previously [43,44]. Breeding pairs of double transgenic mice expressing  PSEN1dE9) 85Dbo/J, The Jackson Laboratory, Bar Harbor, ME, USA] were bred and housed in an approved animal facility at Kangwon National University. Animals were maintained on a 12/12-h light/dark cycle and fed ad libitum and were adapted to these conditions for 2 weeks before the experiment.
Aβ  and Aβ (42-1) were purchased from American Peptide (Sunnyvale, CA, USA). Aβ  and Aβ  were dissolved in 0.1 M PBS at pH 7.4, and aliquots were stored at −20°C. Aβ peptides in each aliquot were aggregated by incubation in sterile distilled water at 37°C for 4 days. Six-month-old COX-2 (−/−)-and COX-2 (+/+)-mice were administered Aβ  and Aβ (42-1) (400 pmol, i.c.v. injection) according to the procedure established by Laursen and Belknap [45]. Each mouse was injected in the bregma using a 10-μl microsyringe (Hamilton, Reno, NV, USA) fitted with a 26-gauge needle inserted at a depth of 2.4 mm. The injection volume was 5 μl. The injection placement and needle track were visible and could be verified during dissection.
YY-1224 and a standard G. biloba extract (Gb) were obtained from the research center of Yuyu Pharma Inc. (Suwon, Republic of Korea). The content and representative HPLC chromatogram of each component in YY-1224 or Gb are shown in Table 1 and Additional file 1: Figure S1, respectively. YY-1224 (50 mg/kg) or Gb (50 mg/kg) was dissolved in 10% tween-80 and administered orally in a volume of 1 ml/kg. YY-1224 or Gb administration began 7 days before the Aβ i.c.v. injection, and the drug administration was continued once a day throughout the experimental period. The behavioral study commenced on day 3 after Aβ i.c.v. injection and was carried out sequentially. During the behavioral study, YY-1224 or Gb was administered 30 min after the behavioral test to avoid a direct effect on performance. The experimental design is shown in Fig. 1a and Additional file 1: Figure S2.
In addition, 6-month-old APPswe/PS1dE9 double Tg mice were treated with YY-1224 (50 mg/kg, p.o.) or Gb (50 mg/kg, p.o.) with or without the preferential COX-2 inhibitor, meloxicam (10 mg/kg, p.o.; Sigma-Aldrich, St. Louis, MO, USA), once a day for 3 months. Meloxicam was suspended in 0.5% sodium carboxymethyl cellulose (Na-CMC) immediately before use. The behavioral study was started when the mice were 9 months old, and additional treatment with YY-1224, Gb, or meloxicam was continued during the behavioral study. During the behavioral study, drugs were administered 30 min after the behavioral test to avoid a direct effect on performance. The experimental design is shown in Fig. 1b.

Y-maze test
The Y-maze test was performed as described previously [46]. Briefly, the Y-shaped maze was constructed of black acrylic with three identical arms separated by 120°. Each arm was 40 cm long, 12 cm tall, and 10 cm wide. The mouse was placed at the end of one arm and allowed to move freely through the maze during an 8-min session. The percent alternation was calculated as the ratio of actual to possible alternations (defined as the total number of arm entries minus two) multiplied by 100.

Novel object recognition test
The novel object recognition test was performed as described previously [47,48]. On the training trial, two different objects were fixed on the floor in a symmetric position from the center of the open field box (40 × 40 × 35 high cm), 20 cm from each other and 10 cm from the nearest wall, and each mouse was allowed to explore the objects for 10 min. Mice were subjected to the test trial 24 h after the training trial. One of the familiar objects used during training trial was replaced by a novel object, and each mouse was then allowed to explore the objects for 10 min. Time spent exploring each object was recorded and analyzed with a video tracking system (EthoVision, Noldus, The Netherlands). Novel object recognition was expressed as a "recognition index (%)," the percentage of time spent exploring the novel object over the total time spent exploring both objects.

Immunocytochemistry
Immunocytochemistry and staining quantification were performed as described previously [46,49,50]. Mice were perfused transcardially with 50 mL of ice-cold PBS (10 mL/10 g body weight) followed by 4% paraformaldehyde (20 mL/10 g body weight). Brains were removed and stored in 4% paraformaldehyde overnight. The brains were cut on a horizontal sliding microtome into 35-μm transverse free-floating sections. Every sixth section of the dorsal hippocampus (total of four sections) from each brain was collected for immunocytochemistry. Sections were blocked with PBS containing 0.3% hydrogen peroxide for 30 min and then incubated in PBS containing 0.4% Triton X-100 and 1% normal serum for 20 min. Antigen retrieval with formic acid (70%, 10 min) was performed to detect Aβ deposition in APPswe/PS1dE9 double Tg mice. After a 48-h incubation with primary antibody against PAFR (1:200; Cayman Chemical, Ann Arbor, MI, USA), PAF-AH (1:80; Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA), Aβ (1:50; Invitrogen, Thermo Fisher Scientific, Camarillo, CA, USA), or Iba-1 (1:500, Wako Pure Chemical Industries, Chuo-ku, Osaka, Japan), sections were incubated with the biotinylated secondary antibody (1:1000; Vector Laboratories, Burlingame, CA, USA) for 1 h. The sections were then immersed in a solution containing avidin-biotin peroxidase complex (Vector Laboratories) for 1 h, and 3,3′-diaminobenzidine was utilized as the chromogen. Digital images were acquired at ×4, ×10, or ×20 objective magnification using an Olympus microscope (BX51; Olympus) and a digital microscope camera (DP72; Olympus). ImageJ version 1.47 software (National Institutes of Health, Bethesda, MD, USA) was employed to quantify images. Briefly, images were converted to 8-bit grayscale images and were subjected to background subtraction to correct for uneven background. To measure PAFR-or PAF-AH-immunoreactivity, the pyramidal cell layer of CA1 and CA3 regions from each section were selected as the region of interest (ROI). Threshold values were set to select the immunoreactive area. The integrated density of each ROI was measured, and the value was expressed as density per micrometer square. For staining quantification of Aβ or Iba-1, the entire hippocampus and the cortical area containing primary somatosensory cortex were selected as the ROIs. Threshold values were set to identify immunoreactive areas, and particle analysis was employed to measure the area fraction of Aβ-or Iba-1immunoreactivity (Additional file 1: Figure S13).

Reverse transcription and real-time polymerase chain reaction (RT-rt-PCR)
Reverse transcription and real-time polymerase chain reaction (RT-rt-PCR) was performed as described previously [51]. Total RNA was isolated from the hippocampus using an RNeasy Mini Kit (Qiagen, Valencia, CA, USA) according to the manufacturer's instructions. Reverse transcription reactions were carried out using the RNA to cDNA EcoDry Premix (Clontech, Palo Alto, CA, USA) with a 1-h incubation at 42°C. For rt-PCR, 0.5 μL of complementary DNA (cDNA) was added to 50 μL of total PCR reaction mixture, containing 25 pmol of each primer and Quanti-Tect SYBR Green PCR Master Mix (Qiagen). rt-PCR was performed using a CFX96 Touch real-time PCR system (Bio-Rad Laboratories, Hercules, CA, USA). The reference gene (GAPDH) and target gene from each sample were run in parallel in the same plate with the same amount of cDNA. Primer sequences are listed in Table 2. After activation of HotStarTaq DNA polymerase at 95°C for 15 min, PCR was performed as 40 cycles of denaturation at 94°C for 1 min, annealing at 60°C for 1 min, and extension at 72°C for 1 min. The relative mRNA expression level was normalized to that of GAPDH (2 [Ct(GAPDH)-Ct(each gene)] ).

Determination of malondialdehyde
The amount of lipid peroxidation in the hippocampus was determined by measuring the level of thiobarbituric acid-reactive substance in homogenates and is expressed in terms of malondialdehyde (MDA) content. The MDA level was measured using HPLC-UV/VIS detection system (model LC-20AT and SPD-20A, Shimadzu, Kyoto, Japan) as described by Richard et al. [52] with slight modifications [49,53]. Briefly, each hippocampal tissue was homogenized in PBS, and 0.1 ml of this homogenate (or standard solutions prepared daily from 1,1,3,3-tetramethoxypropane) and 0.75 ml of the working solution (thiobarbituric acid 0.37% and perchloric acid 6.4%, 2:1, v:v) were mixed and heated to 95°C for 1 h. After cooling (10 min in ice water bath), the flocculent precipitate was removed by centrifugation at 3200 × g for 10 min. The supernatant was neutralized and filtered prior to injection. Isocratic separation was achieved on a 5-μm ODS column with mobile phase consisting of 50 mM PBS (pH 6.0): methanol (58:42, v/v) at a flow rate of 1.0 mL/min. The effluents were monitored at 532 nm.

Determination of protein carbonyls
The extent of protein oxidation in the hippocampus was assessed by measuring the content of protein carbonyl groups, which was determined spectrophotometrically with the 2,4-dinitrophenylhydrazine (DNPH)-labeling procedure, as described by Oliver et al. [54]. Results are expressed as nanomoles of DNPH incorporated/mg protein [49,53] based on the extinction coefficient for aliphatic hydrazones at 21 mM −1 ·cm −1 .

Determination of synaptosomal reactive oxygen species (ROS) formation
Synaptosomal fractions were obtained as described by Whittaker et al. [55], with minor modifications [49,53]. The protein concentration of the synaptosomal preparation was measured using the BCA protein assay reagent (Pierce, Rockford, IL, USA) and was found to be approximately 5 mg of protein per milliliter. ROS formation was assessed by measuring the conversion from 2′,7′-dichlorofluorescin diacetate (DCFH-DA) to dichlorofluorescin (DCF) as described by Lebel and Bondy [56]. Briefly, hippocampal synaptosomes were incubated with 5 μM DCFH-DA (Molecular Probes, Eugene, OR, USA) for 15 min at 37°C. Excess unbound probe was removed by centrifugation at 12,500 × g for 10 min. The fluorescence intensity due to ROS was measured at an excitation wavelength of 488 nm and an emission wavelength of 528 nm. DCF (Sigma-Aldrich) was used as a standard.

Determination of 4-hydroxy-2-nonenal (4-HNE) and protein carbonyl by slot blot analysis
Determination of 4-HNE was performed using slot blot analysis as described previously [49]. Following adsorption, the PVDF membranes were preincubated with 5% non-fat milk and incubated overnight at 4°C with anti-4-HNE antibody (1:2000, Calbiochem, La Jolla, CA, USA). After incubation with the primary antibody, membranes were incubated with a HRP-conjugated secondary antibody. Subsequent visualization was performed using an enhanced chemiluminescence system (ECL plus®, GE Healthcare). The amount of oxidized proteins was measured using the Oxyblot kit [Chemicon (EMD Millipore), Temecula, MA, USA] according to the instructions provided by the manufacturer. Briefly, the protein carbonyl content was labeled with protein hydrazone derivatives using 2,4-dinitrophenylhydrazide (DNP). Each blot was preincubated with 5% non-fat milk and then incubated with the primary antibody (1:100) specific to the DNP moiety, followed by incubation with a HRP-conjugated secondary antibody. Subsequent visualization was performed using an enhanced chemiluminescence system (ECL plus®, GE Healthcare) [49].

Results
Effect of YY-1224 or Gb on Aβ (1-42)-induced impairment of visual recognition learning in COX-2 (+/+)-and COX-2 The recognition test is based on the natural tendency of rodents to investigate a novel object instead of a familiar one. The choice to explore the novel object reflects the use of learning and (recognition) memory processes. There was no significant difference between Aβ-treated groups in exploratory preference for objects in the training trial. The exploratory preference for a novel object was significantly decreased [ The Y-maze test is based on the natural navigation behaviors of rodents and is used to evaluate spatial working memory, which is known to be dependent on hippocampal function. There was no significant change in the performance of the Y maze task between Aβ (42-1)-treated groups. In the COX-2 (+/+) mice, alternation behavior was significantly decreased [ Fig. 2b, ANOVA  (Fig. 2b, P < 0.01), Gb (Fig. 2b, P < 0.05), or genetic inhibition of COX-2 ( Fig. 2b, P < 0.05). YY-1224-mediated attenuation appeared to be more pronounced than Gb in COX-2 (+/+) mice; however, YY-1224 and Gb did not affect the Y-maze performance in Aβ (1-42)-treated COX-2 (−/−) mice. The results from the Morris water maze are comparable to those from the novel object recognition test and the Y-maze test (Additional file 1: Figure S7). and PAFR mRNA expression were significantly attenuated by Gb, YY-1224, or COX-2 gene knockout (Fig. 3a, b, ANOVA and post hoc pairwise comparisons showing the effect of Gb, YY-1224, or COX-2 gene knockout, P < 0.01).

Effects of YY-1224 or Gb on changes in PPARγ expression induced by Aβ
As mentioned above, it has been suggested that PPARγ is related to inflammatory and neurodegenerative processes [22]. To explore whether PPARs modulatory actions are involved in the anti-inflammatory effect of YY-1224, the mRNA or protein expression of PPARα or PPARγ was examined after Aβ    (Fig. 8b, c, ANOVA and post hoc pairwise comparisons showing the inhibitory effect of YY-1224, Gb, and COX-2 gene knockout, P < 0.01). YY-1224 mediated significant attenuation than Gb in COX-2 (+/+) mice (Fig. 8b, c, ANOVA and post hoc pairwise comparisons showing the difference between Gb and YY-1224, P < 0.05). YY-1224 and Gb did not show additional effects compared to COX-2 gene depletion (Fig. 8). The results from RT-rt-PCR of the PPARα or PPARγ mRNA expression were comparable to those from RT-PCR (Additional file 1: Figure S10). Effects of meloxicam, a preferential COX-2 inhibitor, on the pharmacological effect of YY-1224 on memory dysfunction and Aβ plaque deposition in APP/PS1 Tg mice To understand the pharmacological effects of YY-1224, we examined the effect of Gb or YY-1224 on memory dysfunction and Aβ plaque deposition in APP/PS1 Tg mice. ANOVA and post hoc pairwise comparisons showed that treatment with YY-1224 or Gb significantly increased the alternation ratio in the Y-maze test (Fig. 9a, YY-1224 or Gb, P < 0.01 or P < 0.05) and the recognition index in the novel object recognition test (Fig. 9b, YY-1224 or Gb, P < 0.01 or P < 0.05). A preferential COX-2 inhibitor, meloxicam, also significantly enhanced cognitive function in the Y-maze (Fig. 9a, P < 0.01) and novel object recognition tests (Fig. 9b, P < 0.01). Consistent with results obtained from Aβ-treated mice, YY-1224 was more effective at enhancing novel object recognition in APP/ PS1 Tg mice (Fig. 9b, ANOVA and post hoc pairwise comparisons showing the difference between Gb and YY-1224, P < 0.05), and meloxicam treatment did not affect the performance of YY-1224-or Gb-treated APP/ PS1 Tg mice (Fig. 9a, b).
YY-1224, Gb, or meloxicam treatment significantly attenuated Aβ plaque deposition in the hippocampus and cortex of APP/PS1 Tg mice (Fig. 9c, d, ANOVA and post hoc pairwise comparisons showing the inhibitory effect of YY-1224, Gb, and meloxicam, P < 0.01 for hippocampus and cortex). Attenuation mediated by YY-1224 was more pronounced than attenuation mediated by Gb (Fig. 9c, d, ANOVA and post hoc pairwise comparisons showing the difference between Gb and YY-1224, P < 0.05 for hippocampus and cortex). Meloxicam did not affect responses to pharmacological activity in YY-1224-or Gb-treated APP/PS1 Tg mice (Fig. 9c, d).
Effects of meloxicam on pharmacological activity mediated by YY-1224 through PAF-AH I and PPARγ expression in the hippocampi of APP/PS1 Tg mice Because YY-1224 exerted its pharmacological effect through the enhancement of PAF-AH I and PPARγ expression in Aβ (1-42)-treated mice, we examined the effect of YY-1224 on the mRNA and protein expression of PAF-AH I and PPARγ in the hippocampus of APP/PS1 Tg mice. In line with the results from Aβ-treated mice, ANOVA and post hoc pairwise comparisons indicated that PAF-AH I-immunoreactivity was significantly increased by YY-1224 (Fig. 10a, b, P < 0.05 for CA1 and CA3), Gb (Fig. 10a, P < 0.05 for CA1), and pharmacological inhibition (i.e., meloxicam) of COX-2 (Fig. 10a, b, P < 0.05 for CA1 and CA3) in APP/PS1 Tg mice. Since the α2 subunit of PAF-AH I was positively regulated by YY-1224 in Aβ-treated mice, we examined the mRNA expression of this subunit in APP/PS1 Tg mice. ANOVA and post hoc pairwise comparisons revealed that the mRNA expression of the PAF-AH I α2 subunit was significantly enhanced by YY-1224 (Fig. 10c, P < 0.01), Gb (Fig. 10c, P < 0.01), and COX-2 inhibition (i.e., meloxicam; Fig. 10c, P < 0.01). YY-1224 enhanced PAF-AH I-immunoreactivity and PAF-AH I α2 mRNA expression more than Gb (Fig. 10a-c, ANOVA and post hoc pairwise comparisons showing the difference between Gb and YY-1224, P < 0.05). The results from RT-rt-PCR of PAF-AH I α2 subunit mRNA expression were comparable to those from RT-PCR (Additional file 1: Figure S11a).
As shown in Fig. 11, ANOVA and post hoc pairwise comparisons showed that the mRNA and protein expressions of PPARγ were significantly increased by treatment with YY-1224 (Fig. 11b, c, P < 0.01), Gb (Fig. 11b, c, P < 0.01 for mRNA, P < 0.05 for protein), or meloxicam (Fig. 11b, c, P < 0.01) in APP/PS1 Tg mice, and these results were consistent with the results from Aβ-treated mice. The pharmacological action of YY-1224 was more effective than that of Gb in PPARγ protein expression (Fig. 11c, ANOVA and post hoc pairwise comparisons showing the difference between Gb and YY-1224, P < 0.05). In addition, meloxicam did not affect the expression of PAF-AH or PPARγ in YY-1224-or Gb-treated APP/PS1 Tg mice (Figs. 10 and 11). The results from RT-rt-PCR of PPARα or PPARγ mRNA expression were comparable to those from RT-PCR (Additional file 1: Figure S11b and c).
Effects of meloxicam on the pharmacological activity of YY-1224 in response to reactive microgliosis in APP/PS1 Tg mice To elucidate the role of microglia in the pharmacological effects of YY-1224, we examined the effect of Gb and YY-1224 on activated patterns of microglia (Fig. 12). Treatment with Gb and YY-1224 significantly attenuated Iba-1-labeled microglia in the hippocampus and cerebral cortex of 0.5% Na-CMC-treated APP/PS1 Tg mice (Fig. 12a, b, ANOVA and post hoc pairwise comparisons showing the inhibitory effect of YY-1224, Gb, and meloxicam, P < 0.01 for hippocampus and cortex). Meloxicam did not alter significantly the Iba-1 immunoreactivity mediated by YY-1224 or Gb in APP/PS1 Tg mice (Fig. 12).
Different roles of microglia in tissue repair or damage may be due to distinct microglial subsets, i.e., "classically activated" pro-inflammatory (M1) or "alternatively activated" anti-inflammatory (M2) cells. Treatment with Gb or YY-1224 significantly decreased mRNA levels of M1 markers (CD16, CD32, and CD86) in the hippocampus of 0.5% Na-CMC-treated APP/PS1 Tg mice (Fig. 13a-c, ANOVA and post hoc pairwise comparisons showing the inhibitory effect of YY-1224 and Gb, P < 0.01). In contrast, Gb and YY-1224 significantly enhanced the mRNA levels of YM1 and CD206 of the M2 phenotype in the hippocampus of 0.5% Na-CMC-treated APP/PS1 Tg mice (Fig. 13d, e, ANOVA and post hoc pairwise comparisons showing the inhibitory effect of YY-1224 and Gb, P < 0.01). In meloxicam-treated groups, mRNA levels of CD16, CD32, and CD86 were significantly decreased in the hippocampus, while YM1 and CD206 Fig. 9 Activity of YY-1224 (YY) or Gb on memory impairment and Aβ deposition in APP/PS1 Tg mice. a Effects of meloxicam on the pharmacological activity of YY or Gb in response to Y-maze performance and b novel object recognition. c, d Effect of meloxicam on the pharmacological activity of YY or Gb in response to Aβ-immunoreactivity in the hippocampus (c) and cortex (d). Veh vehicle for YY or Gb (10% tween-80 in sterile saline). Each value is the mean ± S.E.M of 10 (a, b) or 5 (c, d) animals. # P < 0.05, ## P < 0.01 vs. vehicle + 0.5% Na-CMC; & P < 0.05 vs. Gb + 0.5% Na-CMC (two-way ANOVA was followed by Fisher's LSD pairwise comparisons). Scale bar = 200 μm mRNA levels were significantly increased (Fig. 13, ANOVA and post hoc pairwise comparisons showing the inhibitory effect of meloxicam, P < 0.01). In addition, meloxicam did not significantly alter the mRNA expression of CD16, CD32, CD86, YM1, and CD206 mediated by YY-1224 or Gb in APP/PS1 Tg mice (Fig. 13). The results from RT-rt-PCR of the mRNA expression of microglial phenotypic markers were comparable to those from RT-PCR (Additional file 1: Figure S12).

Discussion
Increasing evidence has shown that Aβ-triggered microglial inflammatory activation damages neurons in AD pathogenesis [5,57]. In this study, YY-1224 improved cognitive function and inhibited neuroinflammatory activation in COX-2 (+/+) mice after Aβ (1-42) treatment. YY-1224 inhibited the production of pro-inflammatory factors and oxidative stress through PAFR antagonism and activation of the PPARγ-COX-2 signaling pathway. YY-1224 appeared to be more effective than Gb against cognitive dysfunction and Aβ pathology in APP/PS1 Tg mice as well as against Aβ-induced memory impairment via microglia polarization. Previous studies have suggested that the antioxidant activity of Gb contributes to a memory-enhancing effect in AD patients or aged rats [58,59]. Like NSAIDs, the constituents of Gb attenuated the production of inflammatory mediators such as COX-2, iNOS, and TNF-α in vitro [60]. We hypothesize that YY-1224 facilitates memory-enhancing and anti-inflammatory effects by increasing the ratio of terpenoids such as bilobalide and ginkgolides [36]. Bilobalide has been reported to exert protective effects on neurons [37][38][39][40], to facilitate synaptic transmission and plasticity in hippocampal subfields [61,62], and to enhance spatial learning and memory [63]. Ginkgolide B exerts neuroprotection by alleviating neurotoxicity [42] and neuronal apoptosis [41], in addition to acting as a PAF antagonist [64]. c Effects of meloxicam on the pharmacological activity of YY or Gb in response to PAF-AH I α2 mRNA expression in the hippocampus. Veh vehicle for YY or Gb (10% tween-80 in sterile saline). Each value is the mean ± S.E.M of five animals. # P < 0.05, ## P < 0.01 vs. vehicle + 0.5% Na-CMC; &P < 0.05 vs. Gb + 0.5% Na-CMC (two-way ANOVA was followed by Fisher's LSD pairwise comparisons). Scale bar = 200 μm PAF is a highly active phospholipid mediator of inflammation [65]. The initial step of PAF formation is activation of phospholipase A2 (PLA2) in a calciumdependent manner, yielding lyso-PAF. During this step, arachidonic acid is also released and can be converted to its respective cyclooxygenase and lipoxygenase products. The lyso-PAF is then acetylated in position 2 of the glycerol backbone by a coenzyme A (CoA)-dependent acetyltransferase. The majority of PAF's effects are attributed to interaction with PAFR, which is expressed by multiple peripheral cell types; however, the effects of PAF are regionally restricted to microglia subpopulations in the CNS [66]. The inhibition of PAF activity is mediated by a deacetylation reaction catalyzed by PAF-AH [19]. PAF-AH I, originally identified in the brain, consists of three subunits (α1, α2, and β), in which the α subunits provide the catalytic activity. The complex has received attention in part because the subunit that modulates enzymatic activity (the β subunit) is the product of the LIS1 gene. LIS1 expression may be related to the regulation of PAF AH activity in the brain [67]. PAF-AH I has been implicated in neuronal development, neuronal functions, Alzheimer's disease, bipolar disorder, and tolerance to hypoxia [68]. Although the most structurally well-characterized component of PAF AH I is the α1 subunit [69], a highly specific role of the α2 subunit for PAF hydrolysis has been recognized [70]. In this study, the α2 subunit was the most sensitive to Aβ (1-42) exposure or double transgenic overexpression of APP and PS1. Therefore, the α2 subunit may be a therapeutic target for YY-1224 or Gb in response to Aβ (1-42) exposure or double transgenic overexpression of APP and PS1via positive modulation of COX-2.
Among two major groups of constituents of YY-1224, flavonoid (24%) and terpenoid (12%) ginkgolides are known as potent antagonists of PAF [33]. YY-1224 treatment significantly attenuated decreases in the mRNA expression of PAF-AH I α2 and PAFR in response to Aβ  in COX-2 (+/+) mice. This suggests that the induction of PAF-AH I α2 and the PAF antagonistic activity of YY-1224 may contribute to its protective effects against Aβ (1-42) toxicity and possibly the double transgenic overexpression of APP and PS1.
Our current findings are consistent with previous reports that activation of PPARγ regulates the expression of PAF-AH [71]. We also observed that repeated treatment with YY-1224 attenuates the Aβ (1-42)-induced decrease in PPARγ and PAF-AH expression in COX-2 (+/+) but not COX-2 (−/−) mice, suggesting that the PPARγ-PAF-AH pathway may mediate the protective effects of YY-1224 via positive modulation of COX-2. This notion is supported by previous findings that PAF induces the expression of COX-2 [72]. The connection between PAF and PGE2 production by COX-2 has also been demonstrated in astrocytes. Stimulation of these Veh vehicle for YY or Gb (10% tween-80 in sterile saline). Each value is the mean ± S.E.M of five animals. # P < 0.05, ## P < 0.01 vs. vehicle + 0.5% Na-CMC; &P < 0.05 vs. Gb + 0.5% Na-CMC (two-way ANOVA was followed by Fisher's LSD pairwise comparisons) cells with the non-hydrolyzable analog methylcarbamyl-PAF increased the secretion of PGE2, which was reduced by inhibitors of COX-2 but not COX-1 [73]. In this manner, PAF, which is a short-lived molecule due to its rapid degradation by PAF-AH, is able to have a long lasting effect on neurons through induction of COX-2. We found that the deleterious effect of Aβ (1-42) on brain oxidative stress and inflammatory markers are found in COX-2 expressing-but not COX-2-null mice. In addition, the protective effect of YY-1224 is present only in COX-2 expressing-but not COX-2 null mice or APP/PS1 Tg mice treated with the selective COX-2 inhibitor, meloxicam. This suggests an important role of the PPAR-PAF-COX-2 pathway in the deleterious effect of Aβ (1-42) as well as beneficial effects of ginkgolides and YY-1224. Inhibition of this pathway could reduce the formation of pro-inflammatory products of COX-2 such as prostaglandins resulting in decreased levels of pro-inflammatory mediators. We also found that COX-2 inhibition might interfere with a possible feedforward loop to amplify the inflammatory process by reducing PAF signaling through induction of its degrading enzyme, PAF-AH I, and reduction in expression of the PAF receptor, PAFR. This potential feedforward loop could exist in parallel with a previously described feedback loop, where pro-inflammatory mediators formed by COX-2 induce the expression of PAF-AH, resulting in reduced PAF and resolution of inflammation [74]. In this context, it is interesting to note that unlike PAF-AH I, PAF-AH II is induced by Aβ , and such induction is abolished in COX-2 null mice, although this enzyme is highly expressed in the liver and kidney [75]. More studies are needed to determine how COX-2 metabolites such as prostaglandins differentially regulate PAF-AH isoforms, resulting in propagation or modulation of inflammation.
Besides its effects on neurons, Aβ can activate microglial cells. This could lead to neuroinflammation and progression of neurodegeneration [76]. Microglia/macrophages are capable of undergoing phenotypic polarization to the M1 phenotype to produce pro-inflammatory cytokines, or to the M2 phenotype to produce anti-inflammatory cytokines [77,78]. A prevalence of M1 over M2 microglia/ macrophages has been reported in neurodegenerative pathologies, such as AD, and in the elderly brain [79]. In the present study, Aβ (1-42) induced an increase in Fig. 12 Activity of YY-1224 (YY) or Gb on Iba-1-immunoreactivity in the APP/PS1 Tg mice. a, b Effects of meloxicam on the pharmacological activity of YY or Gb in response to Iba-1-immunoreactivity in the hippocampus (a) and cortex (b). Veh vehicle for YY or Gb (10% tween-80 in sterile saline). Each value is the mean ± S.E.M of five animals. # P < 0.01 vs. vehicle + 0.5% Na-CMC; &P < 0.05 vs. Gb + 0.5% Na-CMC (two-way ANOVA was followed by Fisher's LSD pairwise comparisons). Scale bar = 200 μm inflammatory cytokines including TNF-α and IL-1β in microglia from the COX-2 (+/+) mice. PPARγ agonists are able to prevent the detrimental polarization of microglia, or switch the polarization of reactive microglia from the pro-inflammatory to the anti-inflammatory phenotype [87,80]. Therefore, our results suggest that inhibiting the M1 pro-inflammatory phenotype or activating microglial polarization toward an M2 anti-inflammatory phenotype might contribute to the memory-enhancing effect of YY-1224.

Conclusions
In conclusion, we have demonstrated that terpene trilactone-activating G. biloba YY-1224 ameliorates AD pathogenesis by inhibiting oxidative stress and neuroinflammation via positive modulation of microglial activation. These protective effects of YY-1224 were mediated by its PAF antagonistic potential and PPARγ activity through the COX-2 regulatory signaling pathway. Thus, YY-1224 may be a potential candidate for development of novel drugs for AD-like neurodegenerative disorders. , and CD86 (c) mRNA expressions of M1 phenotype microglia/macrophages. d, e Effect of meloxicam on the pharmacological activity of YY or Gb in response to YM1 (d) and CD206 (e) mRNA expressions of M2 phenotype microglia/macrophages. Veh vehicle for YY or Gb (10% tween-80 in sterile saline). Each value is the mean ± S.E.M of five animals. # P < 0.01 vs. vehicle + 0.5% Na-CMC; &P < 0.05, &&P < 0.01 vs. Gb + 0.5% Na-CMC (two-way ANOVA was followed by Fisher's LSD pairwise comparisons)

Additional files
Additional file 1: Supplemental figures. Figure S1. Representative HPLC chromatograms of ginkgo flavone glycosides and terpene trilactones. Figure S2. Experimental design for evaluating the effects of YY-1224 on Aβ (1-42)-induced learning impairments in COX-2 (+/+) and COX-2 (−/−) mice. Figure S3. Effects of YY-1224 or Gb on changes in the protein expression of BDNF, GDNF, NGF, or IGF-1 after k252a or JB-1 treatment in the hippocampus of the COX-2 (+/+) mice. Figure S4. Effects of YY-1224 or Gb on changes in SOD-1 or GPx-1 protein expression after DDC or MS treatment in the hippocampi of the COX-2 (+/+) mice. Figure S5. Effect of YY-1224 or Gb on changes in COX-2 mRNA expression in the hippocampi of the COX-2 (+/+)-mice and on changes in COX-2 protein expression in PC12 cells or mixed cortical cells after treatment with Aβ (1-42). Figure S6. Effect of YY-1224 or Gb on Aβ (1-42)-induced cell death in PC12 cells or mixed cortical cells. Figure S7. Effects of YY-1224 or Gb on Aβ (1-42)-induced memory impairment in COX-2 (+/+) and COX-2 (−/−) mice. Figure S8. Effects of YY-1224 or Gb on Aβ (1-42)-induced changes in PAFR and PAF-AH mRNA levels in the hippocampus. Figure S9. Effects of YY-1224 or Gb on Aβ (1-42)-induced pro-inflammatory genes in the hippocampi of mice. Figure S10. Effects of YY-1224 or Gb on Aβ (1-42)-induced PPAR mRNA expressions in the hippocampi of mice. Figure S11. Effects of YY-1224 or Gb on the mRNA level of PAF-AH I α2 and PPAR in the hippocampus of APP/PS1 Tg mice. Figure  S12. Effects of YY-1224 or Gb on mRNA expressions of microglial phenotype markers in the hippocampus. Figure S13