APP/PS1 (APPswe, PSEN1dE9) double transgenic mice and their age matched wild-type (WT) littermates were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd. (Beijing, China). Mice were housed in specific pathogen free laboratory conditions with no more than four per cage. The mice were housed with controlled temperature (22–24 ${}^{\circ }$C) and humidity (55–65%) under a 12 h dark-light cycle, with food and water available ad libitum. All experiments were performed in accordance with The Guidelines for Animal Care and Use of China, and the experimental schemes were approved by the animal ethics committee of Shanghai Jiao Tong University, with every effect conducted to reduce animal suffering. After mice adapted to the experimenter and the environment, at the age of postnatal 8 months, we started the paeonol intervention study. Following the 3R principle (Replacement, Reduction, Refinement), we divided the mice into: vehicle-treated WT littermate control (WT), vehicle-treated APP/PS1 controls (APP/PS1) and paeonol-treated APP/PS1 mice (APP/PS1+ Pae). Each group contained 18 mice. Paeonol (Sigma-Aldrich, St. Louis, MO, USA, purity $\ge$99.5%) was dissolved in carboxymethylcellulose sodium and was administered orally at a dosage of 20 mg/kg by an oral gavage for three consecutive weeks. The vehicle group was treated with the same dosage of carboxymethylcellulose sodium. Behavioral analyses were performed after a 3-week administration of paeonol. The dosage was determined by previous study [16] and our preliminary experimental results had confirmed that this dose had little effect on behavioral changes on WT mice.

The MWM test is a hippocampus-dependent cognitive memory task assessed in an open field water maze as previously described [5]. The mice were trained to escape from the water by locating a hidden escape platform submerged approximately 1 cm below the water surface at a fixed location. The platform was located at the center of one of the four quadrants of the pool. Training consisted of 4 trials per day for 5 days. On day 6 of testing, the platform was removed and the mice were allowed to swim freely to explore the pool. The swimming parameters were recorded and calculated with a camera connected to a behavioral monitoring system (Noldus Information Technology, Wageningen, the Netherlands).

The object recognition test was performed in a square box under low light levels [17]. Mice were first habituated to the empty arena for 10 min per day for 3 days prior to the object recognition test. For the sampling phase, 2 identical objects were placed at 2 adjacent corners of the box, and the mice were allowed to explore freely in the box for 10 min. After 24 h, mice were reintroduced to the arena for 10 min for the test phase. During this test phase, the arena contained one object identical to those used in the sampling phase, and one novel object. The time spent exploring each object was recorded. A discrimination score was provided by calculating the time spent exploring the novel object minus the time spent exploring the familiar object, and dividing this by the total time spent exploring both novel and familiar objects during the test phase. The arena and objects were wiped down with 70% ethanol between trials to minimize olfactory cues. Ethovision XT (Noldus, Wageningen, Netherlands) was used to track the mice during testing.

After the brain sections were permeabilized with 0.3% Triton X-100 and blocked in 5% BSA, Thioflavin S (Thio S) staining was performed as previously described [18]. The hippocampal sections were immersed in Thio S solution (1%, 8 min) for staining. After washing, brain images were captured with a Nikon microscope. A$\beta$ plaque images were collected from six representative sections per mouse.

After mice were deeply anesthetized with isoflurane, the cortex and hippocampus brains tissues were quickly dissected and homogenized. After the homogenate was centrifuged, the supernatants were collected and the samples used to measure TNF-$\alpha$, IL-1$\beta$, A$\beta$40, and A$\beta$42 using commercial ELISA kits from Wako in accordance with the manufacturers’ instructions.

Stereotaxic surgery and adeno-associated virus (AAV) injection were performed as previously described [19]. Mice were placed on a stereotaxic apparatus (RWD science, Shenzhen, China) under isoflurane anesthesia, and a solution containing AAV-GFP was delivered to CA1 region: AP –1.8 mm; ML 1.25 mm; DV –1.25 mm in 100 nL doses. Infusions were performed at a rate of 20 nL/min. Spines of GFP-labeled neurons could be clearly visualized after a two-week recovery period.

Hippocampus tissue sections (400 $\mu$m thick) were prepared with a Leica VT 1200 s sectioning system. A platinum stimulated electrode with a bipolar enamel coating was placed in the Schaeffer collateral pathway afferents under a microscope. The glass recording electrode was placed in the stratum radiatum of area CA1. 50% of the maximum fEPSP was utilized to induce stable baseline synaptic transmission for 20–30 min prior to LTP induction. LTP was induced by three trains (5 min intertrain interval) of 100 Hz for 1 s. Data were collected using CLAMPEX 9.0 (Axon Instruments, Burlingame, CA, USA) and presented as the average slope of the fEPSP. Clampfit 9.0 (Axon Instruments, Burlingame, CA, USA) was used to analyze data. Two-way ANOVA was used for electrophysiological data analysis with p 0.05 as the significance threshold.

Dissected hippocampi were harvested and lysed using RIPA buffer containing protease and phosphatase inhibitors (Beyotime, China). The concentration of protein was measured with a BCA protein assay kit (Pierce, Rockford, USA). Later, equal amounts of protein were separated by SDS-PAGE and transferred to PVDF membranes (BioRad). After blocking with 5% skimmed milk for 2 h at room temperature, they were incubated overnight at 4 ${}^{\circ }$C with the primary antibody (anti-PSF-95 1:1000, anti-SYN 1:2000, anti-Map-2 1:800 or anti-$\beta$-actin 1:1000). On the following day, the membranes were incubated with HRP-conjugated secondary antibody (1:5000). Each sample was analyzed by immunoblotting at least twice. The positive bands were visualized by enhanced chemiluminescence (Pierce Biotechnology; Sigma-Aldrich Chemical). ImageJ software (National Institutes of Health, Bethesda, MD, USA) was used for the quantification of protein levels.

After being perfused transcardially with ice-cold 0.1 M phosphate-buffered saline (PBS) and post fixed with 4% paraformaldehyde (PFA) [20], 4-$\mu$m thick coronal sections were used for Nissl staining [21], and 15-$\mu$m coronal sections were prepared for immunohistochemistry. Subsequently, sections were blocked in 10% normal goat serum (30 min) and incubated with primary antibodies: anti-Iba-1 (Abcam, Ab178847, 1:400) at 4 ${}^{\circ }$C overnight. The secondary antibody was allowed to react with the sections. After washing with PBS three times with 0.1 M PBS, the sections were further reacted with HRP and visualized with DAB.

Data in this study were analyzed using GraphPad Prism 7.0 software (GraphPad Inc., La Jolla, CA, USA) and results are shown as mean $\pm$ SEM. One-way ANOVA or Two-way RM ANOVA was used for the multigroup comparison according to experimental requirements; p 0.05 was considered statistically significant.

Impaired cognitive performance is well observed in the APP/PS1 AD model. After paeonol or vehicle treatment, the behavior of the three groups were tested using the Morris Water Maze (MWM) and novel object recognition (NOR). In the MWM, the results showed that the paeonol treated APP/PS1 mice took less time to reach the hidden platform than the vehicle-treated APP/PS1 mice during the training phase (Fig. 1A,B). The average swimming speed did not differ among the three groups. In the following probe trials, the APP/PS1 mice explored a shorter time, as well as performing fewer crossings into the target quadrant than the WT group. However, both parameters were significantly improved by paeonol treatment (Fig. 1C,D). In the NOR test, as expected, vehicle-APP/PS1 mice could not distinguish the novel from the familiar object, while paeonol-treat APP/PS1 mice displayed a significantly higher discrimination index, with better cognitive performance (Fig. 1E,F).

Fig. 1.

Paeonol improves neurobehavioral outcomes in APP/PS1 transgenic mice. (A) The time to climb on the hidden platform during the training phase of Morris water maze. (B) Trace map of the mice during the Morris water maze test. (C) The time of exploring hidden platform in the Morris water maze. (D) The number of crossing hidden platform in the Morris water maze. (E) Time of exploring same object in the novel object test. (F) Recognition index performance in the novel object test. Mean $\pm$ SEM was shown (n = 8). **p 0.01, ***p 0.001.

After behavioral testing, the mice were sacrificed, and A$\beta$ pathology was visualized using immunofluorescence (Fig. 2A). Cortical and hippocampal sections of APP/PS1 mice showed apparent A$\beta$ plaque deposits (Fig. 2B,C). However, sections of paeonol-treated APP/PS1 mice exhibited reduced positive area of A$\beta$ plaque (Fig. 2B,C). Soluble A$\beta$40 and A$\beta$42 concentrations in the cortex and hippocampus were assessed using sandwich ELISAs. As shown in Fig. 2D,E, the level of soluble A$\beta$40 and A$\beta$42 were significantly lower within the cortex and hippocampus of the paeonol-treated group than in the vehicle control group.

Fig. 2.

Paeonol adminstration decreased A$\beta$ plaques burden in APP/PS1 mice. (A–C) Immunofluorescent labelling and quantification results of A$\beta$ plaque deposition in the cortex and hippocampus. Scale bar = 200 $\mu$m. (D,E) ELISA results showed a significant reduction in soluble A$\beta$40/42 in both the cortex and hippocampus. Mean $\pm$ SEM was shown (n = 7 mice). **p 0.01, ***p 0.001.

We carried out Iba1 immunofluorescence staining to assess whether paeonol could inhibit microglial activation in APP/PS1 mice. Results showed that the activation of microglia in the hippocampus of paeonol-treated mice were inhibited compared with vehicle-treated APP/PS1 mice (Fig. 3A) and this was further confirmed by immunoblot analysis (Fig. 3B). We also studied the effects of paeonol on the expression of inflammation-related molecules in the cortex and hippocampus of APP/PS1 mice, with levels of TNF-$\alpha$ and IL-1$\beta$ were detected by ELISA. Consistent with the above results, it was clearly evident that the cytokine inflammatory response in the paeonol-treated APP/PS1 mice was less prominent (Fig. 3C,D).

Fig. 3.

Paeonol treatment caused microglia inactivation and decreased neuroinflammation in APP/PS1 mice. (A) Immunohistochemistry for microglia markers ionized calcium-binding adapter molecule 1 (Iba1) in the hippocampus. Scale bars, 20 $\mu$m. (B) Western blot band results of Iba1 expression and (E–G) the results of statistical analyses. (C,D) TNF-$\alpha$ and IL-1$\beta$ levels in the cortex and hippocampus were detected by ELISA. Mean $\pm$ SEM was shown (n = 5 mice). **p 0.01, ***p 0.001.

Previous results prompted us to further evaluate hippocampus neural damage. According to the results of Nissl staining, the decreased number of Nissl-positive cells within CA1 and CA3 hippocampal subregions revealed a significant protective effect of paeonol (Fig. 4A). It was proved that APP/PS1 mice administered paeonol exhibited a significantly decreased proportion of Nissl-positive cells (Fig. 4B). This provides direct evidence that paeonol confers a neuroprotective effect. APP/PS1 mice have been shown to undergo spine morphology changes and decreases in spine density. To examine whether paeonol treatment achieved a therapeutic effect on synaptic connectivity, hippocampal neuronal spines from each group were GFP-labelled for morphological analysis (Fig. 4C). Spine loss was represented by diminished GFP-labelling, and this was observed in the hippocampus pyramidal neurons of APP/PS1 mice. Paeonol administration reversed this spine density reduction, implying a neuroprotective effect of paeonol on synaptic impairment (Fig. 4D).

Fig. 4.

Paeonol treatment inhibited neuronal damage in the hippocampus of APP/PS1 mice. (A) Nissl staining results of the hippocampal CA1 and CA3 regions. (B) Quantification of Nissl-positive dead cells in the hippocampus of CA1 and CA3 regions. (C) Representative photomicrographs of EGFP-labeling in CA1 neurons under treatment with Paeonol. Scale bar: 10 $\mu$m. (D) Quantification of the density of dendritic spines from randomly selected dendritic segments of neurons. Mean $\pm$ SEM was shown (n = 5 mice). **p 0.01, ***p 0.001.

Long-term potentiation (LTP) was induced within the hippocampus to test the hypothesis that paeonol improves cognitive recovery by rescuing functional synaptic connections in this brain region in APP/PS1 mice. The slope of field excitatory postsynaptic potential (fEPSP) induced by stimulation was calculated in each study group. Paeonol could effectively rescue excitatory synaptic current deficit in APP/PS1 mice after the high-frequency stimulation protocol, as demonstrated in the improved LTP found in paeonol-treated mice (Fig. 5A,B). The expression of molecular markers of synaptic damage was further examined. As shown in Fig. 5C–F, we observed that compared with controls APP/PS1 mice, the levels of post-synaptic protein-95 (PSD-95), synaptophysin (SYN), and dendritic-related molecular microtubule-associated protein-2 (MAP-2) were significantly increased in mice receiving paeonol treatment. These molecular expression findings further validate the synaptic protection of paeonol witnessed in functional performance tests.

Fig. 5.

Paeonol administration promote synaptic plasticity in APP/PS1 mice. (A) Traces of hippocampal LTP induced with high frequency stimulation. (B) Normalized fEPSP slopes of HFS-induced LTP. (C–F) Representative immunoblots and the relative levels of PSD-95, SYP, and MAP-2. Mean $\pm$ SEM was shown (n = 5 mice). **p 0.01, ***p 0.001.