Honokiol Nanoscale Drug Delivery System Ameliorates the Cognitive Decits in Tgcrnd8 Mice of Alzheimer’s Disease via Inhibiting Neuropathology and Modulating Gut Microbiota

Honokiol (HO) exerts neuroprotective effects in several animal models of Alzheimer’s disease (AD), but the poor dissolution hampers its bioavailability and therapeutic ecacy. A novel honokiol nanoscale drug delivery system (Nano-HO) with smaller size and excellent stability was developed in this study to improve the solubility and bioavailability of HO. Aβ plaque-associated neuroinammation, JNK/CDK5/GSK-3β signaling pathway as well as gut microbiota in TgCRND8 transgenic mice. p < 0.001) and Oscillospira (F(4, 25) = 12.30, p < 0.001) were higher in TgCRND8 mice, as compared with the WT group. Nano-HO had equal ecacy as HO on reducing the relative abundances of Akkermansia, Allobaculum, Lactobacillus, Oscillospira, Mucispirillum and Parabacteroides in TgCRND8 mice, as compare with the Tg + vehicle group. IDE:insulin MCT:Medium-chain MWMT:Morris NEP:neprilysin; NFTs:neurobrillary PBS:phosphate-buffered saline; PDI:poly-dispersity index; TEM:transmission microscope; TNF-α:tumor necrosis factor; WT:wild type; ZP:zeta potential.

Nano-HO was prepared using HS-15, PEG-400, and MCT at the ratio of 4:2:1 (w/w/w). HO was dissolved in MCT (oil) and then mixed with HS-15 (surfactant) and PEG-400 (co-surfactant) in a gentle magnetic stirring at 300 rpm for 30 min at 25 °C. After pre-equilibrium at room temperature, the solution was diluted 100-fold with double-distilled water and stirred till clear and slightly bluish.

Characterization of Nano-HO
The droplet size, ZP and PDI were measured at 25 °C by a Zetasizer Nano ZS (Malvern Instruments, Britain) based on dynamic light scattering. The morphology of Nano-HO was determined by Hitachi-HT7700 transmission electron microscope (Hitachi-Technologies Corp., Tokyo, Japan). Samples with a 500-fold dilution were placed on a copper grid (400 mesh). After the samples were dried, they were stained with phosphotungstic acid (2%) for 30 s at room temperature to form a thin lm and then observe under transmission electron microscope (TEM).

Animals
Male Sprague Dawley (SD, weighing 230-250 g) rats were obtained from the Laboratory Animal Services Centre, The Chinese University of Hong Kong. Male TgCRND8 mice were crossed with female nontransgenic mice on the hybrid C3H/He-C57BL/6 background to breed a colony of experimental animals.
Non-transgenic littermates that did not express human APP transgene were identi ed as wild-type mice and used as negative controls for experiments. Both rats and mice were maintained on a 12 h light/dark cycle under controlled humidity (50 ± 10%) and temperature (24 ± 2 °C), with access to food and water ad libitum. The experimental procedures were approved by the Animal Experimentation Ethics Committee of The Chinese University of Hong Kong (Ref. No. 18/108/GRF).

In vitro release of Nano-HO and HO
The in vitro release of Nano-HO and HO was determined by a modi ed method described previously [32]. brie y, 5 mL of Nano-HO (containing 5 mg HO) and HO (5 mg HO suspended in 0.5% CMC-Na as control) were placed into a dialysis bag (molecular weight cut-off of 8000-14000 Da) surrounded by 100 mL of phosphate-buffered saline (PBS, pH 7.4) and incubated at 37 °C in an incubator shaker (100 rpm/min). Two hundred microliter of dialysates was collected at 0, 30, 60, 120, 240, 360, 480, 720 and 1440 min while same volume of fresh PBS (37 °C) was subsequently added into the dialysis solution. After centrifugation at 10000 rpm for 10 min, the dialysates were collected and passed through a 0.22 µm lter. For HPLC analysis, the samples were sonicated in 0.2 mL of methanol and detected three times by normalizing the results against the standard curve of HO. The HO released from Nano-DDS and free HO by percentages were plotted against time.

Pharmacokinetics study
Male SD rats (weighing 230-250 g) were randomly assigned into Nano-HO group and HO group (n = 5) containing the same content of HO (80 mg/kg). The dosage of HO was selected based on a previous report [32]. Under anesthetization with diethyl ether, the rat blood samples (0.30 mL each) were collected at 5, 15, 30, 45, 60, 90, 120, 240, 360, 480, 720, and 1440 min from the rat eye socket veins via heparinized capillary tubes after drugs treatment. After centrifugation at 3500 rpm for 10 min at 4 °C, plasma samples were collected and stored at -20 °C for further analysis. The method of plasma sample preparation was determined as previously described [32]. Brie y, 200 µL plasma was mixed with 50 µL docetaxel (800 µg/mL, internal standard) and 350 µL methanol in a vortex mixer for 30 s. The mixture was centrifuged at 12000 rpm for 15 min at 4 °C. Then, all supernatants were transferred to the autosampler vials for introduction into the HPLC system. The analysis was performed with a Shimadzu SIL-20 AHPLC system. Separation was achieved on a unisol C 18 column (5 µm, 100 A°, 4.6 × 250 mm, Agela Technologies, Tianjin, China) and eluted on an isocratic mobile phase composed of methanol and distilled water (76:24, v/v) at a constant ow rate of 1.0 mL/min. Analysis software DAS (Version 3.0; Data Analysis System, Shanghai, China) was used to assess the pharmacokinetic parameters according to the non-compartmental model. With the concentration time curve ranging from 0 to 12 hours (AUC 0 − 12 ), the maximum plasma concentration (C max ), and peak time (T max ) were obtained directly from the plasma concentration vs time curve. The mean residence time (MRT 0 − 12 ), and the biological half-life time (t 1/2 ) were estimated from the terminal linear portion of the plasma concentration-time pro le. The comparative t-test was applied using SPSS software to assess the statistical signi cance.
Polymerase chain reaction (PCR) for genotyping All mice were subjected to genotyping for the APP transgene before experiments as described in our previous study [33]. PCR analysis was performed on genomic DNA isolated from ear using the following primers: Forward-TGTCCAAGATGCAGCAGAACGGCTAC, Reverse -AAACGCCAAGCGCCGTGACT. Those mice with APP transgene were identi ed as transgenic mice, while those without APP transgene as wild type (WT) ones.

Experimental design and drugs treatment in TgCRND8 mice
Three-month-old male mice were divided into 5 groups with 9 mice in each group: (1) WT group; (2) TgCRND8 (Tg) + vehicle group; (3) Tg + HO (20 mg/kg) group; (4) Tg + Nano-HO (20 mg/kg) group; (5) Tg + Donepezil (5 mg/kg) group. The dosage of HO was selected based on the previous studies [16,34,35]. Donepezil was chosen as a positive control and dissolved in normal saline. HO was suspended in 0.5% sodium carboxymethylcellulose (CMC-Na). Mice were administered with HO, Nano-HO and donepezil by gavage once daily for 17 consecutive weeks, whereas mice in the WT group and Tg + vehicle group received the same volume of vehicle (0.5% CMC-Na) for the same duration. After drug treatment, the spatial learning and memory functions were assessed by Morris Water Maze test (MWMT). Figure 3A showed the experimental design and schedule.

Morris Water Maze test (MWMT)
MWMT was performed to assess spatial learning and memory functions [36]. The modular MWMT with a video tracking software of SuperMaze V2.0 was purchased from Xinruan Information Technology Co. Ltd (Shanghai, China). A tank was acted as a maze, and the diameter and the height of the maze were 180 and 70 cm respectively. The maze was lled with water at 25 °C and divided into four equal quadrants. A circular escape platform with 10 cm of diameter was xed in the midpoint of one quadrant that 2 cm beyond the water surface. The tank was located in a test room that contained various prominent visual color pictures (e.g., Triangle, circle, quadrate, etc.). The mice were trained for consecutive 4 days to nd the platform. There were 3 trials for each mouse per day, and the inter-trial interval of each trial was 60 s.
To minimize the performance differences caused by circadian rhythmicity, the MWMT was performed between 9:00 and 18:00. In each trial, we placed the mice gently in one quadrant randomly with its nose pointing toward the wall and allowed them to nd the escape platform. Each mouse was given 60 s to nd the platform and allowed to stay on it for 30 s. If a mouse did not nd the platform within 60 s, the mouse was placed on the circular platform for 30 s before the next trial, and the escape latency ( nding the submerged platform) was recorded as 60 s. To determine the ability of spatial learning, the time of the mouse spent to reach the platform was recorded. On day 5, a probe test of spatial memory was conducted by removing the platform, then the time spent in the target quadrant and the number of crossing the platform quadrant were recorded.

Brain sample processing
Twenty-four hours after MWMT, 6 mice in each group were euthanized with ketamine and xylene, then the brain tissues were removed rapidly and separated into two hemispheres equally for western blotting analysis and ELISA assay. All samples were stored at − 80 °C before further analysis.
For immuno uorescence analysis, 3 mice in each group were deeply anesthetized using xylene and ketamine and transcardially perfused with 0.9% saline followed by buffered 4% paraformaldehyde. Afterwards, the brain tissues were post-xed in 4% paraformaldehyde overnight at 4 °C, then stored in 30% at 4 °C sucrose until sectioned.

Cytokines determination
The brain tissues of mice were homogenized vigorously in 0.8 mL of lysis buffer (contained in kits). After incubation on ice for 20 min, the homogenates were centrifuged at 12000 rpm for 20 min at 4 °C. Protein concentrations were determined by Pierce™ BCA protein Assay kit (Catalog No.: 23227, Thermo Fisher Scienti c). The levels of TNF-α (Catalog No.: ab100747), IL-6 (Catalog No.: ab100712) and IL-1β (Catalog No.: ab100704) in the supernatants were determined using commercially available ELISA kits (Abcam, Cambridge, UK) according to the manufacturer's instructions. The levels of TNF-α, IL-6 and IL-1β were expressed as pg/mg protein.

Western blotting
For preparation of protein lysates, frozen brain tissues were homogenized in RIPA lysis buffer (Catalog No.: 89900, Thermo Fisher Scienti c) which contains 1% Protease/Phosphatase Inhibitor Cocktail (Catalog No.: 78442, Thermo Fisher Scienti c) for 30 min on ice. After centrifugation at 14,000 rpm at 4 °C for 15 min, the supernatants were collected. Protein concentrations were determined by Pierce™ BCA protein assay kit (Catalog No.: 23227, Thermo Fisher Scienti c). Equal amounts of proteins of different samples were loaded. The proteins were separated by SDS-PAGE and then transferred to PVDF membranes. After being blocked with 5% (w/v) non-fat milk in TBST at room temperature for 2 h, the PVDF membranes were incubated at 4 °C overnight with primary antibodies against CTFs ( Signaling Technology) and donkey anti-goat (Catalog No.: sc-2020, Santa Cruz) for 2 h at room temperature. After rinsing with TBST for 5 min × 3 times, the protein bands were visualized by the Pierce™ ECL western blotting substrate (Catalog No.: 32106, Thermo Fisher Scienti c). The intensity of each band was imaged by acer c300 (Azure systems, Mumbai, India) and analyzed using Image J software (NIH Image, MD, USA).
Immuno uorescence assay Coronal brain sections were sectioned at a thickness of 30 µm using cryostat (Leica CM1850, Leica Microsystems GmbH, Wetzlar, Germany) and stored at 4 °C in 0.1 M PB. Prior to staining, the sections were immersed in 0.25% trypsin and incubated at 37 °C for 30 min to achieve antigen retrieval. Then the sections were rinsed in PB three times for 15 min, followed by permeabilization in 0.1 M PB solution with 0.3% Triton, and subsequently incubated overnight at room temperature on a shaker with primary antibodies against anti-β-amyloid 17-24 antibody Catalog No.: A5213, Sigma, USA), anti-GFAP polyclonal antibody (Catalog No.: C106874, Sigma) and anti-IBA-1 antibody (Catalog No.: 019-19741, Wako) in the blocking solution. On the following day, the sections were rinsed with PB three times for 15 min. Next, the sections were incubated with donkey anti-mouse secondary antibody conjugated with Alexa Fluor 488, donkey anti-rabbit secondary antibody conjugated with Alexa Fluor 594 and donkey antimouse secondary antibody conjugated with Alexa Fluor 647 (1:500) (Life Technology/Thermo Fisher Scienti c, Waltham, MA) for 2 h at room temperature in dark, followed by rinsing with PB three times for 15 min. The sections were then mounted on microscope slides (Lab'IN Co, NT, Hong Kong) and coverslipped using uorescence mounting medium (Dako North America, Inc., CA, USA). Immuno uorescent images were captured using a Zeiss uorescent inverted microscope (Zeiss, Gottingen, Germany) equipped with an ORCA-Flash 4.0 v2 digital CMOS camera (Hamamatsu Photonics, Iwata City, Japan).
The quanti cation was analyzed by two investigators who were blinded to the animal grouping using Image J software (NIH, Bethesda, MD, USA). Molecular docking for HO on human BACE-1 SwissDock (URL: www.swissdock.ch) was used to perform the molecular docking analysis of HO on BACE-1. The 3D structure of HO was downloaded from Swissdock database. Crystal structure of BACE-1 in the complex with NLG919 analogue (PDB ID, 1SGZ) was downloaded from RCSB PDB Bank (http://www.pdb.org). The docking results were analyzed using UCSF Chimera 1.11.1 (RVBI, UCSF; San Francisco, CA, USA). Ligand binding results with negative △G values were regarded as having an a nity in the binding between HO and BACE-1. The number of possible hydrogen bonds and the bond lengths were determined by the Find H-Bond tool in UCSF Chimera. All docking procedures were performed using Windows 10.
Fecal DNA extraction and Illumina miseq sequencing Fecal samples of the mice were collected into 2 mL tubes and stored at -80 °C after frozen in liquid nitrogen. Fecal genomic DNA was extracted with OMGA-soil DNA kit as per the manufacturer's instruction. Hypervariant region V4 of bacterial 16S rRNA gene was ampli ed with the forward primer 515 F (5'-GTGCCAGCMGCCGCGGTAA-3') and reverse primer 806R (5'-GGACTACHVGGGTWTCTAAT-3') by PCR. Products were puri ed with Agencourt Ampure XP beads (AGENCOURT, Beckman coulter, US) to remove the unspeci c products. The quality of sequencing library was analyzed by Agilent 2100 bioanalyzer instrument (Agilent DNA 1000 Reagents, CA, USA) to determine the average molecular weights. Puri ed amplicons were sequenced pair end on the Illumina MiSeq PE300 System at Beijing Genomics Institute.
Raw fastq les were quality-ltered using QIIME61 (v1.17). Reads which could not be assembled were discarded. The taxonomy of each sequence was analyzed by RDP Classifer (v2.2) against Silva (v119) 16S rRNA database with 80% con dence threshold. Rarefaction analysis was performed by Mothur (v1.31.2) and α-diversity indexes were compared using rarefed data. Principal component analysis (PCA) plot was implemented by R programming language. Signi cant changes in relative abundance of microbial taxa were detected by linear discriminant analysis effect size (LEfSe).

Statistical analysis
All data were presented as the mean ± SEM. Group differences in the escape latency in the Morris water maze training task were analyzed using two-way analysis of variance (ANOVA) with repeated measures, with the factors being treatment and training day. The other data were analyzed using one-way ANOVA followed by Post-hoc Bonferroni's test to detect inter-group differences. Group differences between HO group and Nano-HO group were analyzed using unpaired t test. GraphPad Prism software (Version 8, GraphPad Software, Inc., CA, USA) was used to perform the statistical analysis. A difference was considered statistically signi cant when the p < 0.05.

Results
Droplet size, zeta potential, morphology and appearance of Nano-HO As shown in Fig. 1B-C, the mean droplet size of Nano-HO was 23.30 ± 0.46 nm with PDI of 0.087 ± 0.00 (n = 3), and the average zeta potential of Nano-HO was − 6.19 ± 1.70 mV (n = 3). As shown in Fig. 1D, the morphology of Nano-HO was observed by TEM, and it displayed that most microemulsion droplets were nearly spherical with a small size and dispersed homogeneously in aqueous medium. Additionally, Fig. 1E showed that Nano-HO was a transparent viscous liquid at room temperature (a) and formed a clear and transparent microemulsion after diluting with 100-fold distilled water (b). When the same content of HO was suspended in 0.5% CMC-Na solution, it was white turbid liquid (c). All these ndings indicated that Nano-HO could signi cantly increase the solubility of HO in water.
In vitro drug release The in vitro release of Nano-HO and HO were dialyzed against PBS (pH 7.4) at 37 °C ( Fig. 2A). The contents of HO in the dialysis buffer were quanti ed by HPLC with a C18 column while a standard curve was made for titration. The results demonstrated that HO and Nano-HO were gradually released into the dialysis buffer over a period of 24 h. The accumulative release rate of Nano-HO (86.3%) was farther than that of regular HO (27.0%) (p < 0.01).

Pharmacokinetics study
The mean plasma concentration-time curve pro les after administration with Nano-HO and HO was presented in Fig. 2B, and the pharmacokinetics parameters acquired by the non-compartmental analysis were listed in Fig. 2C. The results demonstrated that the T max was similar in Nano-HO (0.78 ± 0.05) and HO (0.80 ± 0.07). The half-life (t 1/2 ) of Nano-HO (1.63 ± 0.31) was prolonged about 1.50-fold as compared to that of HO (1.09 ± 0.22). Moreover, the peak concentration (C max ) of Nano-HO (0.78 ± 0.09 µg/mL) was enhanced nearly 1.77-fold than that of HO (0.44 ± 0.02 µg/mL) (p < 0.01). The mean residence time (MRT 0 − 12 ) of Nano-HO (2.83 ± 0.15 h) was slightly longer than that of the HO (2.58 ± 0.10 h). The area under the concentration-time curves from 0 to 12 h (AUC 0 − 12 h) of Nano-HO and HO were 2.20 ± 0.06 µg·h/ mL and 1.18 ± 0.05 µg·h/ mL, respectively, yielding a relative bioavailability of 186.44% (p < 0.01) for Nano-HO. These ndings indicated that when given the same content, Nano-HO could effectively improve the oral bioavailability of HO and prolonged its circulation time in rats.

Nano-HO and HO improved cognitive de cits in TgCRND8 mice
The spatial learning and memory functions of mice was assessed using MWMT. In the training trials, all groups were trained to seek the hidden platform and gradually shortened their escape latency to reach the platform. As shown in Fig. 3B, a signi cant difference was found in the mean latency between training days (F(3, 160) = 40.80, p < 0.001) and between treatments (F(4, 160) = 7.319, p < 0.001), but no interaction was observed between training day and treatment (F(12, 160) = 0.3841, p > 0.05). However, mice in Tg + vehicle group exhibited prolonged escape latency compared with WT mice from day 3 (F(4, 40) = 2.234, p < 0.05) and day 4 (F(4, 40) = 2.262, p < 0.05). As shown in Fig. 3C-D, TgCRND8 mice stayed less time in the target quadrant (F(4, 40) = 7.139, p < 0.001) and crossed through the hidden platform with fewer frequency (F (4, 40) = 8.340, p < 0.001) than WT mice in the probe test. Mice in the HO and Nano-HO groups spent more time in the target quadrant (p < 0.05 and p < 0.01 respectively) and increased the frequency of crossing platform (p < 0.05 for both) when compared to the vehicle-treated TgCRND8 mice. After treatment with donepezil (5 mg/kg), the frequency of crossing platform was higher (p < 0.05) and the time spent in the target quadrant longer (p < 0.05) than those in the Tg + vehicle group.
Nano-HO and HO reduced Aβ deposition and inhibited Aβ plaque-associated neuroin ammation As shown in Fig. 5A, signi cant increase in the microglia density was observed in the hippocampus (F(4, 10) = 44.88, p < 0.001) and the cortex (F(4, 10) = 71.58, p < 0.001) of TgCRND8 mice, as compared with the WT group. Treatment with HO and Nano-HO markedly decreased the microglia density both in the hippocampus (p < 0.01 for both) and cortex (p < 0.01 for both) of TgCRND8 mice, as compared with the Tg + vehicle group. In addition, there were also marked increase in the astrocyte density in the hippocampus (F(4, 10) = 41.04, p < 0.001) and cortex (F(4, 10) = 80.69, p < 0.001) in TgCRND8 mice, when compared with the WT group (Fig. 5B). The HO and Nano-HO treatment signi cantly attenuated the astrocyte density both in the hippocampus (p < 0.01 for both) and cortex (p < 0.05 and p < 0.01, respectively), as compared with the Tg + vehicle group. Furthermore, as shown in Fig. 5C, Aβ plaque burdens were signi cantly elevated in the hippocampus (F(4, 10) = 109.6, p < 0.001) and the cortex (F(4, 10) = 84.75, p < 0.001) of TgCRND8 mice, as compared with the WT mice. The Aβ plaque burdens in the HO and Nano-HO groups signi cantly decreased in the hippocampus (p < 0.05 and p < 0.01 respectively) and the cortex (p < 0.01 for both), as compared with the Tg + vehicle group. Donepezil (5 mg/kg) signi cantly inhibited the microglia and astrocytes in ltration, and also attenuated the Aβ plaque burden in the hippocampus (p < 0.01) and the cortex (p < 0.01) of TgCRND8 mice. Interestingly, Nano-HO more markedly decreased the astrocyte density both in the hippocampus (p < 0.01) and the cortex (p < 0.01), and reduced the Aβ plaque burdens in the hippocampus (p < 0.05) of TgCRND8 mice than the HO group.
The above results demonstrated that Nano-HO could signi cantly inhibit the expression of BACE-1. Thus, a molecular docking of HO with BACE-1 was conducted to investigate whether HO was a BACE-1 inhibitor.
The molecular docking results showed that HO could form three hydrogen bonds at speci c residues (Lys107, Asp216 and VAL170) with BACE-1 protein with binding energy of -6.64, -6.75, -6.85 kcal/mol, respectively, and with bond lengths of 2.503 Å, 2.382 Å and 2.215 Å, respectively (Fig. 6E). The docking results were consistent with the above western blot data, indicating that HO is a BACE-1 inhibitor.

Nano-HO and HO suppressed apoptosis and tau protein hyperphosphorylation
The protein level of caspase-3 (F(4, 10) = 44.72, p < 0.001) in the brain tissues of TgCRND8 was signi cantly elevated, while Bcl-2 expression (F(4, 10) = 7.979, p < 0.01) was decreased, when compared with the WT group ( Fig. 7A-B, respectively). After treatment with HO, Nano-HO and donepezil, the expressions of caspase-3 were effectively mitigated (p < 0.01 for all). In addition, Nano-HO and donepezil treatment also signi cantly increased the Bcl-2 expression (p < 0.01 for both), when compared with the Tg + vehicle group.
Nano-HO and HO regulated the JNK/CDK5/GSK-3β signaling pathway As clearly shown in Fig. 8, as compared to the WT group, the ratio of p-JNK/JNK was notably upregulated in the Tg + vehicle group (F(4, 10) = 46.49, p < 0.001). HO and Nano-HO treatments was able to down-regulate the ratio of p-JNK/JNK (p < 0.01 for both), as compared with the Tg + vehicle group.
Additionally, signi cant increase in the ratio of p-35/CDK5 (F(4, 10) = 17.71, p < 0.001) was observed in the Tg + vehicle group as compared to WT group, which was remarkably attenuated by HO and Nano-HO treatment (p < 0.05 and p < 0.01, respectively). On the other hand, the ratio of p-GSK-3β (Ser9)/GSK-3β were markedly decreased in the brain tissues of TgCRND8 mice (F(4, 10) = 36.44, p < 0.001), as compared with the WT group. Treatment with HO and Nano-HO obviously increased the ratio of p-GSK-3β (Ser9)/GSK-3β (p < 0.01 for both). Treatment with donepezil (5 mg/kg) signi cantly decreased the ratio of p-35/CDK5 (p < 0.01) and recovered the ratio of p-GSK-3β (Ser9)/GSK-3β (p < 0.01). On the other hand, Nano-HO showed more potency in inhibiting the activation of GSK-3β via elevating the ratio of p-GSK-3β (Ser9)/GSK-3β (p < 0.05) than regular HO. Differences in gut microbiota pro le among WT, Tg and HO treatment groups The system clustering tree (Fig. 9A) revealed signi cant differences among ve groups. Samples in HO and Nano-HO groups were clustered separately from Tg + vehicle group, re ecting that HO and Nano-HO prevented the changes of gut microbiota in TgCRND8 mice.
For the α-diversity analysis, the Shannon index was signi cantly decreased (F(4, 25) = 3.96, p < 0.05) and Simpson index was remarkably increased (F(4, 25) = 8.887, p < 0.001) in TgCRND8 mice (Fig. 9B-C). Nano-HO and HO treatments improved the Shannon index although the improvement failed to reach a signi cant difference (p > 0.05 for both), while the treatment signi cantly decreased the Simpson index (p < 0.05, p < 0.01 respectively), indicating that Nano-HO and HO could improve the diversity and species evenness in the fecal samples of TgCRND8 mice.
In addition, principal coordinate analysis (PCoA) and partial least squares discrimination analysis (PLS-DA) both yielded well separated positions among the groups (Fig. 9D-E). Notably, the bacterial communities in the Nano-HO group were more closely clustered with the WT mice than HO group, which differed from TgCRND8 mice, suggesting that the bacterial communities in TgCRND8 mice were changed and the gut microbiota composition differed among the ve groups.
Gut microbiota composition at different levels among ve experimental groups Figure 10 and Fig. 11 illustrated the gut microbiota community composition and dominant bacterial distribution at different levels in fecal samples.
At the order level, a total of 19 genera were identi ed in all samples (Fig. 10C). As shown in Fig. 11C, Clostridiales was of predominance in all samples among ve groups and showed a high abundance in the Tg + vehicle group (F(4, 25) = 9.757, p < 0.001). The relative abundances of Campylobacterales (F(4, 25) = 25.07, p < 0.001) and Desulfovibrionales (F(4, 25) = 25.64, p < 0.01) were signi cantly higher in TgCRND8 mice, whereas the relative abundances of YS32 (F(4, 25) = 20.63, p < 0.001) and Bi dobacteriales (F(4, 25) = 6.76, p < 0.001) remarkably decreased in TgCRND8 mice, as compared with the WT group. The proportion of Turcibacterales was decreased in TgCRND8 mice but failed to show a difference (F(4, 25) = 1.538, p > 0.05), as compared with the WT group. Nano-HO exerted similar inhibitory effect to HO on the relative abundance of Clostridiales, Campylobacterales and Desulfovibrionales, but had better enhancement effect than HO on the relative abundance of YS32 and Bi dobacteriales in TgCRND8 mice, as compare with the Tg + vehicle group.

Discussion
HO has been reported to improve cognitive de cits in several animal models of AD via clearing Aβ deposition, inhibiting AChE activity and suppressing neuroin ammation [14,16,17,37,38], but the poor water solubility badly limited its bioavailability and potential medicinal application. Nano-DDS is bene cial for prolonging exposure time, increasing drug e cacy and overcoming poor bioavailability of drugs, which makes it appealing as a universal vehicle for lipophilic drugs. It is well-known that droplet size of nanoparticles (10-50 nm) is a critical factor since it is closely associated with the rate, extent and absorption of drug release. Our results showed that Nano-HO could form nano-sized microemulsion droplets (23.30 ± 0.46 nm) when diluted with distilled water (Fig. 1B). Meanwhile, low PDI re ects the uniformity of particle size. The closer the PDI value is to zero, the more homogeneous the droplets are [39]. Additionally, stability of nanoparticles partially depends on the surface zeta potential, a parameter that gives the magnitude of the electrostatic repulsive interactions between particles [40]. A higher value of zeta potential usually hinders the probability of coalescence, thereby maintaining homogeneity of droplet size [41]. Our results indicated that Nano-HO formulation exhibited a relatively high negative average zeta potential and a low PDI value, suggesting that it met the required zeta potential prerequisite for a stable microemulsion. In addition, the accumulative release rate of HO from Nano-HO (86.3%) in PBS (pH 7.4) was signi cantly higher than that from regular HO (27.0%) over a period of 24 h (Fig. 2A).
Possible reasons may include that small droplet size of Nano-HO provided a large surface area for drug release into the aqueous phase. On the other hand, the pharmacokinetics study was investigated in rats to compare the bioavailability of Nano-HO with that of regular HO. The results demonstrated that the t 1/2 and MRT 0 − 12 were both prolonged in Nano-HO group than in the regular HO group, suggesting that the oral bioavailability of Nano-HO was greatly improved as compared with regular suspension. Moreover, the AUC 0 − 12 h of Nano-HO (2.20 ± 0.06 µg·h/ mL) was signi cantly increased when compared with HO (1.18 ± 0.05 µg·h/ mL), resulting in a relative bioavailability of 186.44% to HO. These ndings indicated that the improved bioavailability of Nano-HO was predominantly owing to the increased solubility. Moreover, Nano-HO exerted better improving effects on cognitive de cits in TgCRND8 mice than HO, and these ndings were believed to be related to the improved oral bioavailability of Nano-HO.
Neuroin ammation is widely considered as one of the major pathological factors of AD. Microglias are the primary in ammatory cells in the brain. Astrocytes, the most abundant glial subtype in the central nervous system, also play a critical role in the pathogenesis of AD. Growing lines of evidence have demonstrated that Aβ accumulation in AD causes microglia activation and astrocyte recruitment, thereby inducing the release of pro-in ammatory cytokines including TNF-α, IL-6 and IL-1β [42][43][44]. Meanwhile, in ammation could induce the expression of BACE-1, promote Aβ deposition, and exacerbate tau protein hyperphosphorylation and neurons loss. Therefore, in ammation is the core driver of AD pathogenesis. In this study, we found that both Nano-HO and regular HO could prevent the microgliosis, astrogliosis and Aβ deposits in the hippocampus and cortex of TgCRND8 mice, as well as suppress the release of TNF-α, IL-1β and IL-6 in the brain tissues of TgCRND8 mice ( Fig. 4 and Fig. 5). Interestingly, Nano-HO inhibited astrogliosis in both hippocampus and cortex of TgCRND8 mice in a more potent manner than HO (Fig. 5B). These ndings indicated that the amelioration of Nano-HO on hippocampal-dependent memory function was attributable to its anti-in ammatory property.
It is well-known that Aβ deposition is a key pathogenic hallmark in AD pathogenesis. Increased production of Aβ peptides and formation of Aβ plaques through sequential cleavage of APP by the β-and γ-secretases contribute to the pathological basis of AD [45]. Speci cally, p-APP (Thr 668), as observed near the plaques, may increase the Aβ levels by facilitating the exposure and cleavage by β-secretase BACE-1 [46]. PS-1 and APH-1 are vital catalytic submits of γ-secretase responsible for APP cleavage to Aβ [47,48]. Increasing evidence revealed that proteolytic degradation is a particularly important determinant of cerebral Aβ levels, and Aβ-degrading enzymes including IDE and NEP play critical roles in Aβ degradation [49]. Therefore, inhibition of β-or γ-secretase or enhancement of Aβ-degrading enzymes could help to reduce the Aβ production. Our results demonstrated that Nano-HO showed similar effect in inhibiting the protein expressions of APH-1 and PS-1 as HO (Fig. 6). Interestingly, Nano-HO showed better effect on inhibiting the protein expressions of p-APP (Thr 688) and BACE-1, and enhancing the protein expressions of IDE and NEP than that of HO. These results suggested that Nano-HO may modulate APP processing and phosphorylation through suppressing the activities of β-and γ-secretases and enhancing the activities of Aβ-degrading enzymes to clear the Aβ deposition in the brains of TgCRND8 mice. Furthermore, our molecular docking results demonstrated that HO was well docked with BACE-1 at three active sites including Lys 107, Asp 216 and VAL 170 (Fig. 6E), suggesting that HO may be a BACE-1 inhibitor.
Abnormally high level of hyperphosphorylated tau protein is another typical pathological hallmark of AD, which also leads to oxidative stress via increasing the reactive oxygen species (ROS) production.
Increased ROS could promote in ammatory response, then induce neuronal apoptosis or loss, ultimately resulting in learning and memory impairments [50,51]. It has been reported that the phosphorylation of tau protein is abnormally accentuated at different sites of Thr 205 (7.61 times increase), Ser 396 (4.95 times increase) and Ser 404 (2.97 times increase) in the postmortem brain tissues of AD patients [52]. In addition, up-regulation of caspase-3 is directly responsible for cellular apoptosis in AD [53]. Thus, inhibition of tau protein hyperphosphorylation and neuronal apoptosis may be potential therapeutic targets for AD. Our results revealed that Nano-HO could inhibit tau protein hyperphosphorylation at Thr 205 and Ser 404 sites, as well as the protein expression of caspase-3, but enhance the protein expression of Bcl-2 in the brain tissues of TgCRND8 mice (Fig. 7), indicating that the inhibitory effect of Nano-HO on speci c hyperphosphorylation of tau protein and apoptosis may be the underlying molecular mechanisms of its cognitive function improving effects.
Activation of JNK pathway has been consistently found in the surrounding area of the Aβ plaques in AD patients and transgenic mice via facilitating p-APP (Thr 668) in culture cell lines [54][55][56] and exaggerating p-tau (Thr 205) [57]. In our present study, both Nano-HO and HO signi cantly downregulated the ratio of protein expressions of p-JNK/JNK in brain tissues of TgCRND8 mice (Fig. 8A). In addition, JNK pathway is also closely involved in the activation of GSK-3β, which is considered to be a key kinase responsible for APP phosphorylation in neuronal cells and intimately associated with AD progression [58,59]. Hyperactivation of GSK-3β has been found in the brains of AD patients [60]. Suppressing GSK-3β activity has been demonstrated to decrease the generation and accumulation of Aβ in APP transgenic mice of AD [61]. Moreover, GSK-3β is also a major kinase associated with the aberrant phosphorylation of tau [62], which could be inactivated by phosphorylation at Ser 9 site, suggesting that agents with ability to up-regulate p-GSK-3β (Ser 9) maybe potential candidates for the treatment or prevention of AD [63,64]. Our results indicated that Nano-HO had better e cacy than HO on enhancing the ratio of p-GSK-3β (Ser9)/GSK-3β in the brain tissues of TgCRND8 mice (Fig. 8C).
Meanwhile, CDK5 plays a crucial role in the development of central nerve system and AD progression [65]. Under pathological conditions, CDK5 was activated via direct binding to its neuronal speci c activators p-35, and then aggravate tau hyperphosphorylation by enhancing GSK-3β, exacerbate neuronal loss and subsequently lead to neurodegeneration [66][67][68][69][70][71]. Therefore, agents that suppress the CDK5 activity may ameliorate plaque pathology, neuro brillary and neuronal loss in AD. Our results indicated that Nano-HO suppressed the ratio of p-35/CDK5 in the brain tissues of TgCRND8 mice, suggesting that the cognitive de cits improving effects of Nano-HO were associated with its ability to inhibit the CDK5 activity, and the nding was also consistent with the improvement on kinase activity GSK-3β of Nano-HO.
The bacteria community in the gut can directly re ect the health status of the host by maintaining a certain proportion to protect the bacterial ora balance. The changes in bacterial diversity and richness can lead to the dysfunctions of the bacterial community, and trigger brain-gut axis dysbiosis, contributing to the occurrence of neurodegenerative disorders like AD [72]. In our study, the decreased Shannon index and increased Simpson index suggested that TgCRND8 mice were associated with the diversity and evenness deduction of the bacterial community, as compared with the WT mice ( Fig. 9B-C), and the observation was consistent with the similar decline of bacterial diversity in AD patients [73,74]. The structural variability or similarity among different treatment groups was assessed by system clustering tree, PCA and PLS-DA in our study ( Fig. 9B and 9E-F). These results showed that the mice in Nano-HO group clumped visibly far away from the TgCRND8 mice, emphasizing that the bacteria community tended to recover to normal. This observation was consistent with the nding of the changed intestinal bacteria in AD patients as reported before [75].
Several studies have demonstrated an essential role of gastrointestinal microbes in the development of cerebral Aβ amyloidosis along with a peripheral in ammatory state [76,77]. Bacteria living in the intestinal tract adhere to the intestinal mucosal surface of epithelial cells, forming bacterial ora, thereby affecting the intestinal integrity and permeability [78]. When the harmful bacteria destroyed the integrity of intestinal epithelial cells, the in ammatory reaction was triggered or aggravated accompanied with an increase in in ammatory cytokine (e.g., IL-6 and TNF-α) levels [79]. Our study showed that the bacterial community altered, which coincided with the productions of TNF-α, IL-1β and IL-6, along with increase of Aβ plaques in brains. These results implied that TgCRND8 mice might cause the damage of the brain via changing the bacteria condition in the gut.
Reduction of given bene cial bacteria increased the in ammation, which can be harmful to the intestinal structure. Such reduction can be characterized in Firmicutes spp. and Bi dobacteria spp. [80][81][82].
Metabolites secreted by Firmicutes spp. decreased the production of pro-in ammatory factors such as TNF-α, thus suppressed the occurrence of in ammation [83]. Probiotics such as Lactobacillales spp. and Bi dobacteriales spp. improved the conditions of in ammation and intestinal epithelial barrier function impairment [80,82]. In AD mouse model, acetate (a metabolite of Bi dobacterium breve strain A1) has been reported to ameliorate cognitive disturbances [84]. It is worth noting that when compared to the WT mice (53.6%, in phylum level, 1.18% in order level, and 2.61% in genus level,), there was a decline of Firmicutes, Bi dobacteria and Lactobacillus by 42.4%, 87.3%, and 69.7%, respectively, in TgCRND8 mice (Fig. 11A, C and E), revealing that the reduction of bene cial bacteria was a potential cause of intestinal in ammation in TgCRND8 mice.
Additionally, fewer Actinobacteria, but more Bacteroidetes and Proteobacteria were found in the intestinal microbiota of AD patients or APP/PS1 transgenic mice when compared to healthy controls [74,77], suggesting that bacterial dysbiosis was positively associated with the progression of AD. In our study, we noticed that the relative abundance of Firmicutes, Proteobacteria and Bacteroidetes were major community at the phylum level, which account to almost 90%, followed by Actinobacteria and Cyanobacteria. The relative abundance of Actinobacteria had an 88% decrease in TgCRND8 mice, as compared with the WT mice, while the proportion of the Proteobacteria, Bacteroidetes and Cyanobacteria visibly increased by 235.3%, 99.7% and 125% respectively in TgCRND8 mice, as compared to the WT mice. Those alterations were in accordance with the previous reports [85,86]. Nano-HO inhibited the relative abundance of the Firmicutes, Proteobacteria, Bacteroidetes and Cyanobacteria in TgCRND8 mice as similar to that of regular HO. Interestingly, Nano-HO enhanced the relative abundance of Actinobacteria in TgCRND8 mice in a more potent manner than HO (Fig. 11A).
Recently, the effect of chronic Helicobacter pylori infection on AD has been demonstrated by the release of massive in ammatory mediators [87]. Helicobacter pylori ltrate could cause tau protein hyperphosphorylation in mouse neuroblastoma N2a cells and brains of rats via activation of GSK-3β [88].
Our results demonstrated that the relative abundance of Helicobacteraceae (at family level, Fig. 11D) in TgCRND8 group was augmented by 532.3% as compared to the WT group. Interestingly, Nano-HO reversed this change in TgCRND8 mice in a more potent manner than HO.
Mucin-degrading bacteria are identi ed as microbial drivers. Among them, Prevotella degrades mucin and Desulfovibrio enhances the rate-limiting sulfatase step by hydrolyzing glycosyl sulfate esters. Ruminococcus is also able to degrade mucins [89]. As probiotics strains, Akkermansia can secrete immunoglobulin A (IgA) and antibacterial peptides by immunological rejection to resist pathogen damage to the intestine, thereby possessing anti-in ammatory and barrier-improving properties [90,91].
Our results showed that the relative abundance of Desulfovibrionales (at order level, Fig. 11C), Prevotellaceae (at family level, Fig. 11D) and Ruminococcaceae (at family level, Fig. 11D) drastically increased to 457.6%, 325.6% and 139.3%, respectively, in TgCRND8 mice, as compared with the WT group, and the changes may be of relevance to the increased transmembrane permeability. The relative abundance of Akkermansia (at genus level, Fig. 11E) signi cantly decreased in TgCRND8 group, as compared with the WT group. Nano-HO and HO reversed these changes in TgCRND8 mice. Figure 12 schematically summarized the molecular mechanisms underlying the cognitive de cits ameliorating actions of Nano-HO and HO in TgCRND8 mice.

Conclusions
Our study demonstrated for the rst time that Nano-HO could improve cognitive de cits in a more potent manner than HO in TgCRND8 mice via inhibiting the in ltration of astrogliosis and β-secretase, upregulating Aβ-degrading enzymes, suppressing tau protein hyperphosphorylation at site Thr 205, inhibiting JNK pathway and activating GSK-3β pathway. The multi-target effects of Nano-HO against cognitive de cits in TgCRND8 mice were mediated, at least in part, via inhibiting neuroin ammation and tau hyperphosphorylation, modulating APP processing and phosphorylation through suppressing the activation of JNK/CDK5/GSK-3β signaling pathway. Furthermore, Nano-HO could regulate the compositions and structures of gut microbiota to protect the gut micro ora and its stability. Taken together, Nano-HO is a promising Nano-based formulation with natural compound worthy of further development into AD treatment.

Consent for publication
All authors have consented for publication.

Availability of data and materials
All the primary data supporting the conclusions of this study are availablefrom the corresponding author on a reasonable request.

Competing interests
The authors declare no con icts of interest with respect to this article.    Data were expressed as mean ± SEM (n = 9). #p< 0.05 and ##p< 0.01 when compared with the WT group; * p< 0.05 and ** p< 0.01 when compared with the Tg + vehicle group.

Figure 4
Effects of Nano-HO and HO on the levels of cytokines includingTNF-α (A), IL-1β (B) and IL-6 (C) in the brain tissues of TgCRND8 mice. Data were expressed as mean ± SEM (n = 6). ##p< 0.01 compared with WT group; ** p< 0.01 compared with Tg + vehicle group.  Effects of Nano-HO and HO on the APP processing and APP phosphorylation in the brain tissues of

Figure 12
A schematic drawing depicting the molecular mechanisms underlying the cognitive de cits ameliorating actions of Nano-HO and HO in TgCRND8 mice. Firstly, the transmembrane APP was processed in the amyloidogenic pathway, in which APP was sequentially cleaved by β-secretase, γ-secretase and Aβ degrading enzymes, leading to the production of Aβ peptide and formation of Aβ plaques. Nano-HO and HO reduced Aβ deposition by inhibiting β-secretase andγ-secretase and enhancing theactivity of Aβ degrading enzymes, thereby reducing the Aβ-associated activation of microgliosis and astrogliosis, as well as decreasing secretion of pro-in ammatory cytokines. Also, Nano-HO and HO inhibit tau hyperphosphorylationvia preventing the activation of JNK/CDK5/GSK-3β signaling pathway and preventing apoptosis. Moreover, Nano-HO and HO regulated the gut dysbiosis to reach a balance and protect the microbiota ora stability in TgCRND8 mice. These molecular actions of Nano-HO and HO nally contributed to the improvements in spatial learning and memory therapeutic effects in TgCRND8 mice.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.