Decline of cognitive function is the hallmark of Alzheimer’s disease (AD), regardless of the pathological mechanism. Traditional Chinese medicine has been used to combat cognitive impairments and has been shown to improve learning and memory. Radix Polygalae (RAPO) is a typical and widely used herbal medicine. In this study, we aimed to follow the β-amyloid (Aβ) reduction activity to identify active constituent(s) of RAPO. We found that Onjisaponin B of RAPO functioned as RAPO to suppress Aβ production without direct inhibition of β-site amyloid precursor protein cleaving enzyme 1 (BACE1) and γ-secretase activities. Our mechanistic study showed that Onjisaponin B promoted the degradation of amyloid precursor protein (APP). Further, oral administration of Onjisaponin B ameliorated Aβ pathology and behavioral defects in APP/PS1 mice. Taken together, our results indicate that Onjisaponin B is effective against AD, providing a new therapeutic agent for further drug discovery.

Traditional Chinese medicine has long been used to treat dementia [ 20 – 23 ]. Among those historically used herbal drugs, Radix Polygalae (RAPO) has been demonstrated to exhibit nootropic activity [ 20 , 24 , 25 ]. Moreover, our previous data regarding the “Smart soup” containing Rhizoma Acori Tatarinowii, Poria cum Radix Pini and Radix Polygalae showed systematic beneficial effects against AD and RAPO function to decrease Aβ production [ 24 , 26 ]. Herein, we explored the major component(s) of RAPO and the related underlying mechanism.

The amyloid cascade hypothesis is supported by accumulating studies based on cell culture and animal experiments [ 3 ]. In amyloid hypothesis, the maturation, processing and degradation of APP and the consequential production and clearance of Aβ initiate AD pathogenesis [ 4 ]. APP can be sequentially cleaved by BACE1 and γ-secretase [ 5 , 6 ] and finally yield Aβ species. This makes BACE1 and γ-secretase key-players in AD pathogenesis. In brains of Alzheimer’s disease patients, large amounts of Aβ are produced and aggregated, mainly in the hippocampus and prefrontal cortex, causing neuronal death and impairment of cognitive function [ 7 , 8 ]. Thus, for the past two decades, to identify a solution for this devastating disease, researchers have focused on modulating BACE1 or γ-secretase activities. The cellular Aβ level can be reduced, and cognitive function impairments have been shown to be ameliorated in transgenic AD model mice treated with secretase inhibitors or modulators [ 9 ]. However, treatments with these compounds have been discontinued because of severe adverse effects in recent clinical trials [ 10 , 11 ]. Aside from functions in Aβ generation, BACE1 and γ-secretase are also involved in multiple physiological processes including cell adhesion and Notch signaling. These accumulating evidences suggest that the strategy of directly inhibiting the enzymatic activity of BACE1 and γ-secretase may need to be optimized. APP and its proteolytic products undergo degradation via protein degradation pathways [ 12 – 14 ]. The proteasome is the major organelle for protein degradation in cells [ 15 ], and cleavage through this pathway reduces Aβ production [ 16 – 18 ]. Furthermore, our previous work has shown that interfering with the interaction between BACE1 and γ-secretase, thereby blocking the sequential process and reducing Aβ production, may have some advantages in modifying the disease [ 19 ].

Alzheimer’s disease is a complex and currently incurable age-related neurodegenerative disease and is highly prevalent in aged cohorts worldwide [ 1 ]. It is the most common late-age mental failure in humans and currently exerts great economic and political pressure on modern society. Aβ deposition, tau tangles and cognitive degeneration are the hallmarks of the disease [ 2 ], therefore reducing Aβ production and improving cognitive function has be considered as an effective disease- and symptom-modifying therapeutic strategy.

All experiments were repeated at least three times. All data are presented as the mean ± s.e.m. and analyzed using GraphPad Prism 6.01 (San Diego, CA, USA). The unpaired Student’s t-test was used for comparisons of two groups. Group differences were analyzed with one-way analysis of variance (ANOVA). The results of Morris water maze hidden platform training were analyzed using two-way ANOVA. Differences were considered significant when p < 0.05.

HEK293/APPswe cells or HEK293T cells transiently transfected with myc-NotchΔE or PSGL1-HA were treated with chemicals for 4 hours, and membrane fractions or total lysates were analyzed for APP-CTF, NICD and PSGL1-CTF levels. Culture medium from HEK293/APPswe cells was subjected to Western blotting analysis for sAPPα and sAPPβ and to ELISA analysis for total Aβ. The in vitro C99 assay was carried out as described previously [ 32 ]. The in vitro APP processing assay was performed using the membrane fraction of HEK293/APPswe cells incubated with the indicated chemicals in vitro at 37°C for 2 hours and then subjected to Western blot for APP-CTF analysis. Membrane fractions or total lysates of mouse brain were extracted and subjected to Western blotting analysis for ADAM10, BACE1, PS1, APLP1 and its CTF, and E-Cadherin CTFs. Flotillin or Actin was blotted as a loading control.

Total membrane fractions were extracted from 293T cells or APP/PS1 mouse brain and used in ELISA-based secretase assays or fluorogenic substrate assays to measure BACE1 or γ-secretase activity. Fluorogenic substrate assays and the ELISA-based γ-secretase assay were carried out as previously reported [ 30 , 31 ]. For the ELISA-based BACE1 assay, membrane fractions of HEK293T cells were collected after lysis in buffer A. Supernatants containing 20 mg protein were centrifuged at 25000 g for 1 hour. The resulting membrane pellets were then resuspended in BACE1-assay buffer (50 mM sodium acetate, pH 4.5 and 0.5 mM biotinylated APP-TM peptide). After being incubated at 37°C for 30 minutes, the reaction mixtures or biotinylated standard peptides DK-16 were neutralized by Tris-Na 2 HPO 4 buffer and added into streptavidin-coated 96-well plates (Pierce) and incubated at room temperature for one hour. Anti-Aβ 82E1 antibodies (IBL) were added to the plate and then incubated for another hour. After three washes, horseradish peroxidase-labeled anti-mouse antibodies were added, and the plates were incubated for one hour. Ultra-TMB (Pierce) was used as the substrate for horseradish peroxidase and the absorbance at 450 nm was recorded. Concentrations of samples were calculated according to standard curves. APP-TM peptide (amino acid sequence: SGLTNIKTEEISEVNLDAEFRHDSGYEVHHQK-biotin) and DK-16 (amino acid sequence: DAEFRHDSGYEVHHQK-biotin) were synthesized by GL Biochem.

HEK293/APPswe cells were treated with chemicals at the indicated concentrations and durations. The conditioned medium was then collected and subjected to a sandwich ELISA for the total Aβ level. Human Aβ40 and Aβ42 in APP/PS1 mouse brains were extracted as previously reported [ 29 ] and measured with human Aβ ELISA kits according to the manufacturer’s guidelines. ELISA kits for total human Aβ, human Aβ40 and human Aβ42 were obtained from ExCell Bio.

HEK293T, HEK293 and A431 cells were previously purchased from ATCC, and HEK293MSR cells were a kind gift from Sanofi-Aventis Research and Development. All cell lines were maintained under the same condition as described previously [ 19 ]. In detail, HEK293T, HEK293MSR, HEK293/APPswe and A431 were cultured in Dulbecco’s modified Eagle’s medium (DMEM) with 10% (w/v) heat-inactivated fetal bovine serum in a humidified incubator with 5% CO2/95% air (v/v) at 37°C. HEK293 cells were cultured in MEM under the same condition. The Swedish mutant form of APP was transfected into HEK293 using Fugene HD (Roche) following the manufacturer’s instructions. A cell line stably expressing Swedish mutant APP was established in the presence of 1 mg/ml G418. Penicillin-Streptomycin solution (Life Technologies) was added according to the manual. All constructs were the same as reported previously and were verified by sequencing [ 19 , 28 ].

Onjisaponin B (purity > 98%) was purchased from Biopurity, and RP granules were purchased from Jiangyin Tianjiang Pharmaceutical Co., Ltd. L-685,458 (purity > 96%) and (S)-(+)-ibuprofen (purity > 99%) were purchased from Sigma, DAPT (purity > 99%) was purchased from Selleck and BACE1 inhibitor IV (purity > 98%) was purchased from Calbiochem. E2012 was synthesized by Ginkgo Pharma. MG132 (purity > 97%) was purchased from Selleck and lactacystin (purity > 98%) from Santa Cruz. CellTiter-Glo was purchased from Promega. Fugene HD and Effectene Transfection Reagent were purchased from Roche and QIAGEN, respectively. Immunoblotting was performed with the following antibodies: anti-ADAM10 (a Disintegrin and metalloproteinase domain-containing protein 10) C-term (Sigma); anti-PS1 N (1–65) (EMD); anti-BACE1 N-term (Abgent); anti-APP-CTF (Sigma); anti-Flag (Sigma); anti-sAPPα (secreted Amyloid Precursor Protein-α) (IBL); anti-sAPPβ (secreted Amyloid Precursor Protein-β) (IBL); anti-HA (Sigma); anti-c-Myc (Santa Cruz); anti-NICD (Notch intracellular domain) (Cell Signaling); anti-E-Cadherin-CTF (BD Transduction Laboratories); anti-APLP1 (APP-like protein 1) C-Terminal (643–653) (Calbiochem). Secretase activity assays were performed with anti-Aβ40 (EMD Millipore) and anti-Aβ (82E1, IBL). Immunohistochemistry was performed with anti-Aβ, 1–16, 6E10 (Covance) and anti-GFAP (Glial fibrillary acidic protein) (Dako).

The Morris water maze analysis was performed as previously reported [ 27 ], and the animals were randomly numbered among genotypes and grouped for the test. The apparatus was a 120-cm-diameter circular pool filled with water containing small white plastic particles, with cues of four different shapes posted on four directions of the inner pool wall. The water temperature was maintained at 23.0 ± 0.5°C and the room temperature at 25.0 ± 0.5°C during the whole procedure. A transparent platform 11 cm in diameter was placed 1 cm below the water surface at a fixed position in the target quadrant. The training consisted of 4 trials per day for 7 consecutive days. On day 4, probe trials were conducted after the fourth training trial. On day 8, a single round of probe trial was performed. An automated tracking system (Ethovision XT software) was used to monitor the mouse swimming paths and other parameters.

All animal experiments were performed according to the National Institutes of Health Guide for the Care and Use of Laboratory Animals. The animal protocols were approved by the Biological Research Ethics Committee, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences. Effort was made to minimize animal pain and discomfort. The IACUC approved this research under the approval number SIBCB-NAF-14-002-S309-015.

Results

Systematic fractionation identifies RAPO-1-3 as the active fraction of RAPO that reduces Aβ production In our previous study, we showed that Radix Polygalae reduced Aβ generation in vivo and in vitro [26]. To identify the active components, we first separated RAPO into 3 fractions using an ethanol-water gradient (Fig 1A). The concentrations of the fractions were calculated according to the yield and are presented as weight/volume (w/v) of original raw material. The concentrations of the different RAPO fractions applied to HEK293/APPswe cells are 0.1, 0.3 and 1 mg/ml (relative RAPO concentration). This relative concentration is converted based on the yield of each fractionation step. For example, according to the fractionation scheme, RAPO-1 is composed of 10.1588% dry weight of RAPO (43.6% × 23.3% = 10.1588%), and 1 mg/ml (relative to RAPO) of RAPO-1 is equivalent to 0.1 mg/ml (actual concentration). We then treated HEK293/APPswe cells with various concentrations of each fraction. HEK293/APPswe cells stably overexpress human APP protein carrying a Swedish mutant (K595N/M596L) and show elevated Aβ secretion [33]. Medium from HEK293/APPswe cells treated with RAPO fractions was collected to detect the total Aβ level using a sandwich ELISA. Meanwhile, cells were lysed and subjected to the CellTiter-Glo assay to evaluate cell viability. The 8-hour treatment with RAPO-1 or RAPO-2 did not affect cell viability, whereas RAPO-3 reduced cell viability to 48% (S1A Fig). Compared to vehicle treatment, treatment with either RAPO-1 or RAPO-3 markedly reduced the Aβ level in the conditioned medium, by 48% and 90%, respectively (S1A Fig). To test whether the Aβ-reducing activity and cytotoxicity might be separated, we further fractioned RAPO-1 and RAPO-3 into three sub-fractions (Fig 1A). The cell viability was severely compromised upon RAPO-3-2 treatment, decreasing by 68% (S1B Fig), whereas the fraction RAPO-1-3 reduced Aβ generation without obvious cytotoxicity (S1B Fig). These results indicate that RAPO-1-3 is the representative fraction of RAPO that reduces Aβ production without affecting cell viability. Therefore, we focused on RAPO-1-3 in our further studies. PPT PowerPoint slide

PowerPoint slide PNG larger image

larger image TIFF original image Download: Fig 1. Systemic fractionation identifies RAPO-1-3 as the active fraction of RAPO that reduces Aβ production. (A) The extraction and fractionation scheme of RAPO. (B) RAPO-1-3 significantly reverses the spatial memory deficit of AD mice. (C) Representative tracks of each group of mice in probe trial test at day 8. (D) The latency to platform in probe trial for each group of mice at day 8. (E) No differences in the swimming distance and velocity among the groups. (F) The time spent by mice in the target quadrant. (G) SDS-soluble and FA-soluble Aβ40 and Aβ42 levels in the mouse hippocampi and cortices were measured by sandwich ELISA and normalized to control. Data are presented as the mean ± s.e.m. * p < 0.05, ** p < 0.01 and *** p < 0.001. Two-way ANOVA with Bonferroni's multiple comparison test (B, F), one-way ANOVA with Bonferroni's multiple comparison test (D, E) and two-tailed t-test (G). https://doi.org/10.1371/journal.pone.0151147.g001 We then tested the in vivo efficacy of RAPO-1-3. It has been reported that APP/PS1 double transgenic mice begin to display accumulated Aβ plaque deposition in the hippocampus and cortex, as well as age-dependent deficits in cognitive function at 6 months of age [34–36]. We chronically administered RAPO-1-3 to APP/PS1 mice. Gender- and age-matched APP/PS1 transgenic mice and their transgene-negative littermates were grouped and treated with RAPO-1-3 (0.15 g/kg/day) or Vehicle (50% PEG400 in distilled water) by oral gavage. During the drug administration, animals’ body weights were recorded. The body weights in the RAPO-1-3 group showed no significant change compared to that of the vehicle group (data not shown), suggesting that there was no obvious toxic effect of the treatment. Three months later, the Morris Water Maze analysis was performed to evaluate the spatial learning and reference memory of the animals [27, 37, 38]. There was no obvious difference among animals in swimming distance or velocity (Fig 1E), implying that RAPO-1-3 treatment did not change mouse locomotor activity. As shown in Fig 1B, consistent with previous reports, APP/PS1 mice exhibited significantly impaired learning and memory ability compared to their wild-type littermates in the hidden platform phase (Veh APP/PS1 -/- vs. Veh APP/PS1 +/+, p < 0.0001) [36]. Interestingly, these spatial learning and memory deficits of APP/PS1 mice were ameliorated by chronic administration of RAPO-1-3 (RAPO-1-3 APP/PS1 +/+ vs. Veh APP/PS1 +/+, p < 0.0001). During the probe trial at day 8, the mice treated with RAPO-1-3 took less time to reach the position of the platform (RAPO-1-3 APP/PS1 +/+ vs. Veh APP/PS1 +/+, p = 0.2479) (Fig 1D), spent more time in the target quadrant (RAPO-1-3 APP/PS1 +/+ vs. Veh APP/PS1 +/+, p = 0.2116) (Fig 1F), and crossed more frequently within the platform area (Fig 1C). These data indicate that RAPO-1-3 effectively ameliorates the spatial learning and reference memory deficiency of APP/PS1 transgenic mice. Soluble Aβ oligomers are deleterious and correlate to cognitive deficits in Alzheimer’s disease [39, 40]. Hence, SDS-soluble and FA-soluble Aβ40 and Aβ42 in mouse cortices and hippocampi were quantified by sandwich ELISA. SDS-soluble Aβ40 and Aβ42 were moderately reduced in the cortex and hippocampus, while FA-soluble Aβ40 and Aβ42 were significantly reduced (Fig 1G). These data indicate that RAPO-1-3 functions as the active fraction of RAPO that reduces Aβ production in vivo and in vitro, as well as attenuating the learning and memory deficits in APP/PS1 mice.

RAPO-1-3 and Onjisaponin B promote APP degradation Studies show that interference with APP metabolism alters Aβ production [16, 18, 48–51]. Thus, we monitored the APP cleavage pattern on western blots. Treatment with BSI IV completely abolished the production of sAPPβ and C99 and led to the accumulation of mature APP and C83. GSI L685,458 treatment caused the accumulation of C99 and C83, while the GSM E2012 exerted little effect on any of the APP metabolic products (Fig 3A). Conversely, treatment with RAPO, RAPO-1 or RAPO-1-3 led to a reduction in the level of full-length mature APP, and accordingly the protein levels of the cleavage products (including sAPPα, C83, sAPPβ and C99), while the immature APP level remained unchanged (Fig 3A). Onjisaponin B treatment resulted in a similar APP processing pattern (Fig 3B) as that of RAPO-1-3. Nevertheless, neither BACE1 expression nor its maturation changed significantly upon Onjisaponin B treatment (S3A Fig). These data suggest that Onjisaponin B may not reduce the cellular Aβ level by blocking APP maturation. To test whether Onjisaponin B inhibited APP or secretase expression, we performed quantitative RT-PCR analysis. As shown in S3C Fig, we found no obvious differences among the mRNA levels of those proteins. Then, we tested whether it was the alteration of APP protein levels that led to the reduction of Aβ production. The in vitro APP processing in the presence of active BACE1 was monitored. HEK293/APPswe cell membrane fractions were extracted and incubated with the indicated compounds for 2 hours in BACE1-assay buffer. Treatment with BSI IV completely blocked sAPPβ generation and caused the accumulation of full-length mature APP and C99, while GSI L685,458 did not change the sAPPβ level but caused minor accumulation of APP and C99 (Fig 3C). However, both treatment with either Onjisaponin B or Tenuifolin reduced the level of sAPPβ while leaving the level of full-length APP unchanged (Fig 3C). Further, in the presence of Onjisaponin B together with BSI IV, the full-length mature APP accumulated to a level comparable to BSI IV alone, whereas the level of sAPPβ was decreased (Fig 3D). Co-treatment with L685,458 and Onjisaponin B showed no synergetic effect compared to L685,458 treatment alone (Fig 3D). Moreover, an in vitro γ-secretase activity analysis showed that the AICD production was not affected by Onjisaponin B treatment (S3B Fig). These data suggest that Onjisaponin B reduced Aβ production neither by the direct inhibition of secretase activities nor by modulating protein expression. The proteasome proteolytic pathway has also been shown to be involved in the degradation of full-length APP [13, 52, 53]. Thus, we treated HEK293/APPswe cells with BSI IV or L685,458 in the presence of the proteasome inhibitor MG132 and monitored the APP processing pattern. Consistent with previous reports [54], MG132 showed no obvious effect on the APP processing pattern in the presence of BACE1 or γ-secretase inhibitors (Fig 3E). Interestingly, the decreases in full-length mature APP and subsequently in C99 and C83 by Onjisaponin B were partially rescued in the presence of MG132 (Fig 3E). The Aβ levels of the cells treated with MG132 and Onjisaponin B were comparable to those of the cells treated with MG132 alone (Fig 3F). To rule out the possibility that the rescue by MG132 was due to its GSI-like activity, we tested another, more specific proteasome inhibitor, lactacystin, which has been reported to show no GSI-like activity even at 100 μM [55]. As shown in Fig 3F, the Aβ levels of the cells treated with Onjisaponin B in the presence of lactacystin were significantly higher than those of the cells treated with Onjisaponin B alone. Together, these data indicate that Onjisaponin B and the active fractions of RAPO reduce Aβ production by promoting APP degradation through the proteasome pathway. Additionally, as we recently reported that blocking the PS1/BACE1 interaction could also contribute to reduced production of Aβ [19], we tested whether Onjisaponin B possessed a similar inhibitory activity. Co-immunoprecipitation analysis of PS1 and BACE1 was performed. As shown in S4A Fig, Onjisaponin B interfered with the interaction between PS1 and BACE1. Further, FRET [19] (S4B Fig) and Split-TEV assay [19, 56] (S4D Fig) also showed that Onjisaponin B reduced the interaction between PS1 and BACE1. These data indicate that Onjisaponin B may possess multiple functions to reduce Aβ production.