Identification of natural products with neuronal and metabolic benefits through autophagy induction

Screen of novel autophagy inducers in a plant metabolite library

Using analytical chemistry approaches, we generated a pilot phytochemical library of more than 100 compounds by purifying glycoconjugate metabolites of different structures from several medicinal plants, including Panax ginseng, Helianthus annuus, Rosa rugosa and Veratrum nigrum. We preselected a subset of 19 phytochemicals for the autophagy inducer screen from our phytochemical library (Fig. 1A). The preselection was based on the documented hypoglycemic and/or neurotrophic effects of the medicinal plants against type 2 diabetes and/or neurodegeneration.Citation^14-26 The purity of all compounds was above 95% (Fig. S1), assessed by HPLC (high-performance liquid chromatography) with C18 columns, or HPLC with TSK columns plus the Molisch test, which meets the analytical standard (≥95 %) of commercial vendors.

Figure 1. Identification of a steroid glycoside compound Rg2 as a new autophagy inducer in vitro from a phytochemical library screen. (A) Screening paradigm for autophagy activators from plant metabolite products. (B) Chemical structure of Rg2. (C) Representative images of GFP-LC3 puncta in HeLa cells stably expressing GFP-LC3 cultured in normal or starvation medium, or treated with Rg2 in normal medium for 3 h. Scale bar: 20 μm. (D) Quantification of GFP-LC3 puncta in GFP-LC3-expressing HeLa cells cultured in normal or starvation medium, or treated with Rg2 in normal medium, for 30 min, 2 h and 3 h. Statistics compare each value to the one under the normal condition. (E) Western blot detection (upper) and quantification (lower) of SQSTM1 and LC3 in HeLa cells cultured in normal or starvation medium, or treated with Rg2 in normal medium, in the presence or absence of the lysosomal inhibitor bafilomycin A₁ (BafA1) for 3 h. Results represent mean ± s.e.m. ***, P < 0.001, t test.

Display full size

Using HeLa cells stably expressing GFP-tagged MAP1LC3B/LC3B (referred to as LC3 thereafter), an autophagosome marker, we screened the library for autophagy inducers. Upon autophagy induction, LC3 is converted from cytosolic nonlipidated form (LC3-I) to autophagosome-associated lipidated form (LC3-II), which can be visualized as punctate structures by fluorescence imaging. After a 3-h treatment with each chemical, we found that 3 groups of plant metabolites derived from Panax ginseng and Helianthus annuus, including Rg2, WGP (containing WGPN and WGPA), and AHP (containing AHPN, AHPA1 and AHPA2), significantly increased the number of autophagosomes (represented as GFP-LC3B puncta) (Fig. S2A and S2B), to a similar level as after 3 h of starvation. Autophagy induction is also confirmed by biochemical markers of autophagy, such as decreased levels of SQSTM1/p62, an autophagy substrate, and increased conversion of LC3-I to LC3-II by western blot (Fig. S2C). These compounds did not alter the levels of SQSTM1 and LC3 mRNAs transcriptionally (Fig. S2D), which suggests that the changes in SQSTM1 and LC3 occur posttranslationally due to autophagy regulation. To rule out compound toxicity in the secondary screen, we performed MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-2H-tetrazolium bromide] assays in 2 different cell lines, HeLa and PC12 cells, after 24 h of drug treatment (Fig. 1A, Fig. S3A). Compounds that dramatically reduce cell viability [such as GXVII (gypenoside XVII) and GLXXV (gypenoside LXXV)] were excluded from further testing. Thus, we identified a series of novel autophagy-inducing phytochemicals derived from medicinal plants (Fig. S2F).

Among all the candidates, we decided to focus on a protopanaxatriol-type ginsenoside Rg2, a 785 Da steroid glycoside extracted from Panax ginseng (Fig. 1B). We selected Rg2 as the lead compound for functional analysis based on 3 reasons: 1st, Rg2 has a clear structure and a low molecular weight; 2nd, when applied at lower concentrations (such as 100 μg/ml), Rg2 induces higher levels of autophagy in vitro than WGPA and AHPA2 (Fig. S2A); 3rd, although WGPA and AHPA2 show robust in vitro autophagy-inducing capacity, Rg2 is the one that shows highest autophagy induction and lowest individual viability in vivo (Fig. S2E). We confirmed the autophagy-inducing capacity of Rg2 by GFP-LC3 assay (Fig. 1C and S4A), and found that Rg2 induced formation of GFP-LC3 puncta as early as 0.5 h after administration and continued to increase over time (Fig. 1D). In addition, the reported in vivo metabolite of Rg2, protopanaxatriol (PPT), also induced autophagy in GFP-LC3 HeLa cells after a 3-h treatment, as well as starvation and the parent compound Rg2 (Fig. S4B). Rg2 treatment also reduced SQSTM1 levels and induced LC3-II conversion (Fig. 1E). Cotreatment of the lysosomal inhibitor bafilomycin A₁ and Rg2 led to higher accumulation of LC3-II and SQSTM1 compared to normal conditions (Fig. 1E), suggesting that Rg2 treatment increases the autophagic flux. In addition, we also found that Rg2 induces autophagy in multiple other cell types, including the HepG2 human liver cell line derived from hepatoma and 2 cell lines of neuronal origin, Neuro2A cells from mouse neuroblastoma and PC12 cells from rat adrenal gland pheochromocytoma (Fig. S5). Although Rg2 potently induces bulk autophagy, it does not appear to induce selective autophagic clearance of mitochondria as strong as the mitochondrial uncoupler carbonyl cyanide m-chlorophenyl hydrazone (CCCP) (Fig. S4C). Finally, we ruled out cell toxicity of Rg2 by examining the viability of HeLa and PC12 cells treated with Rg2 for a time course of 3 h, 6 h, 12 h and 24 h (Fig. S3B). Therefore, altogether, we have identified Rg2, a plant-derived molecule, as a potent inducer of nonselective autophagy via cell-based assays.

Figure 2. Rg2 induces autophagy in multiple mouse tissues. (A) Representative images (left panel) and quantification (right panel) of GFP-LC3 puncta in the indicated organs of GFP-LC3 transgenic mice injected with vehicle (DMSO) or Rg2 once daily for 3 d. Results represent mean ± s.e.m. Statistics compare each value to the one with DMSO treatment. *, P < 0 .05; **, P < 0 .01; ***, P < 0 .001, t test. Scale bar: 20 μm. (B) Western blot detection (upper) and quantification (lower) of SQSTM1 and LC3 in the indicated organs of mice injected with vehicle (Vh) or Rg2. WAT, white adipose tissue; BAT, brown adipose tissue.

Display full size

Rg2, a ginseng-derived steroid glycoside, induces autophagy in mouse tissues

We next sought to investigate the autophagy-inducing and therapeutic potential of Rg2 in vivo. Panax ginseng has been used for thousands of years in East Asia as a traditional herbal medicine to treat diabetes and angiogenesis in cancer, and ginsenosides are proposed to be the major bioactive components.Citation^21,27 However, to date there has been a lack of clear experimental evidence characterizing the effectiveness and pharmacological mechanism of a variety of ginsenosides. Accordingly, we hypothesized that the ginsenoside Rg2 may carry out some of the beneficial effects of Panax ginseng by activating autophagy.

To test this hypothesis, we confirmed that Rg2 induces autophagy in vivo, by injecting Rg2 intraperitoneally (i.p.) into transgenic mice globally expressing GFP-LC3.Citation²⁸ We found that Rg2 injection at the dose of both 10 mg/kg and 20 mg/kg significantly increases the number of GFP-LC3 puncta (autophagosomes) in multiple tissues, including brain frontal cortex, heart, liver and muscle (Fig. 2A). In addition, Rg2 treatment resulted in increased LC3-I to LC3-II conversion and decreased levels of SQSTM1 in liver, muscle, brain, heart, white adipose tissue and brown adipose tissue (Fig. 2B), supporting that Rg2 induces autophagy in vivo. Importantly, Rg2 treatment does not alter food intake in these mice (Fig. S8B), demonstrating that autophagy induced by Rg2 is not due to fasting. Furthermore, we examined in vivo autophagy induction by Rg2 via GFP-LC3 puncta assays in both muscle and brain. We found that Rg2-induced autophagy peaks at 4 h and declines to the basal level after 8 to 16 h in muscle, whereas in the brain, autophagy activity peaks at 4 h and lasts for at least 24 h (Fig. S8A). Plasma protein binding assays demonstrated that when injected into mice, the percentage of Rg2 bound by plasma protein is approximately 42% at 10 μg/ml, 24% at 50 μg/ml, and 15% at 100 μg/ml (Fig. S6). Thus, the majority of Rg2 is not bound by plasma proteins and is freely available. Taken together, we identified Rg2 as a new in vivo autophagy inducer.

Rg2-mediated autophagy induction is AMPK-ULK1-dependent and MTOR-independent

To explore the potential mechanisms by which Rg2 activates autophagy, we examined the signaling pathways regulated by Rg2 treatment. Activation of AMPK (AMP-activated protein kinase) and suppression of MTOR are 2 major upstream signaling pathways that induce autophagy in mammalian cells.Citation^29-33 We found that compared either with normal or starvation conditions, Rg2, as well as 2 other autophagy-inducing plant compounds AHP and AHPA1, potently induced phosphorylation of AMPK in HeLa cells (Fig. 3A). However, compared with starvation, during which MTOR and its downstream substrates RPS6KB (p70 ribosomal S6 kinase) and EIF4EBP1 (eukaryotic translation initiation factor 4E binding protein 1) were dephosphorylated and suppressed, the MTOR pathway remained phosphorylated and active in the presence of Rg2 (Fig. 3A), suggesting that MTOR deactivation is not required for Rg2-induced autophagy. To exclude a cell line-dependent effect, we also detected similar roles of Rg2 in other cell types, including HepG2, Neuro2A and PC12 cells (Fig. 3B). Thus, we proposed that Rg2 induces autophagy via AMPK rather than MTOR. To further test this hypothesis, we analyzed the phosphorylation of ULK1, an essential kinase for autophagy. ULK1 can be phosphorylated by AMPK at S555 (pro-autophagy) or by MTOR at S757 (anti-autophagy) under different conditions, and autophagy activity is induced by phosphorylation of S555 or dephosphorylation of S757.Citation^34,35 In contrast to starvation, we found that Rg2 treatment significantly induced AMPK phosphorylation of ULK1 at S555, but did not suppress S757 phosphorylation by MTOR (Fig. 3C), which supports our hypothesis that Rg2 induces autophagy through the AMPK-ULK1 pathway but not MTOR signaling. This will be important for further study of the therapeutic potential, given that MTOR suppression elicits a number of unwanted side effects, such as immunosuppression caused by the well-studied MTOR inhibitor and autophagy inducer rapamycin.

Figure 3. Rg2 induces autophagy by activating the AMPK-ULK1 pathway. (A) AMPK and MTOR signaling pathways were analyzed by western blot (left) and quantified (right) in HeLa cells cultured in normal or starvation medium, or treated with the indicated chemicals in normal medium for 3 h. (B) AMPK and MTOR signaling pathways were analyzed by western blot in multiple cell lines, including HepG2, Neuro2A and PC12 cells, cultured in normal or starvation medium, or treated with Rg2 in normal medium for 3 h. (C) Phosphorylation of ULK1 was analyzed by western blot in HeLa cells cultured in normal, nutrient-rich or starvation medium, or treated with Rg2 in normal medium for 3 h. (D and E) Representative images (D) and quantification (E) of GFP-LC3 puncta in HeLa cells stably expressing GFP-LC3 cultured in normal or starvation medium, or treated with Rg2 and the indicated siRNA in normal medium for 3 h. Cells were transfected with control (NC) or PRKAA1 and PRKAA2 (the AMPK subunits) siRNAs 48 h prior to Rg2 treatment. Knockdown of the AMPK subunits was assessed by western blot analyses in (D). Scale bar: 20 μm. (F) Western blot detection (left) and quantification (right) of SQSTM1 and LC3 in AMPK WT or knockout (prkaa1^−/− prkaa2^−/−) MEFs cultured in normal or starvation medium, or in normal medium supplied with Rg2 or rapamycin for 3 h. Results represent mean ± s.e.m. ***, P < 0 .001, t test.

Display full size

To confirm that autophagy induction by Rg2 is AMPK-dependent, we analyzed Rg2-induced autophagy activity in AMPK knockdown or knockout cells. We found that siRNA knockdown of the AMPK subunits PRKAA1 and PRKAA2 significantly decreased the number of autophagosomes (represented by GFP-LC3 puncta) induced by Rg2 (Fig. 3D and E). In addition, in comparison with starvation or rapamycin treatment, Rg2-induced degradation of SQSTM1 and formation of LC3-II was decreased in prkaa1^−/− prkaa2^−/− double knockout mouse embryonic fibroblasts (MEFs) (Fig. 3F), suggesting that AMPK is indispensable for Rg2-mediated autophagy upregulation. Altogether, these data suggest that Rg2 induces autophagy through activating the AMPK-ULK1 pathway.

Rg2 increases insulin sensitivity in HFD-fed mice

Various species of Ginseng have been used as an alternative medicine to treat type 2 diabetes, and their glycemic lowering effects seem to depend on the profiles of ginsenoside composition.Citation^19-22 Accumulating evidence also indicates that upregulation of autophagy may be metabolically beneficial.Citation^36-38 Accordingly, to determine whether the ginsenoside Rg2 plays a role in metabolic regulation, we examined the insulin receptor signaling in Rg2-treated HeLa cells. We found that Rg2 treatment enhanced insulin-stimulated phosphorylation of INSR (insulin receptor) and the trend of phosphorylation of its downstream kinase AKT/protein kinase B (pan) (Fig. 4A), suggesting insulin sensitization by Rg2. Thus, these data indicate that Rg2 may sensitize the insulin pathway by upregulating autophagy.

Figure 4. Rg2 treatment protects mice from high-fat diet-induced insulin resistance. (A) Rg2 enhances insulin signaling. Phosphorylation of AKT and INSR (insulin receptor) with or without 10 min of 50 nM insulin incubation in HeLa cells was detected by western blot analyses (left) and quantified (right). Prior to insulin stimulation, cells were sensitized overnight in serum-free medium, and then cultured in normal (N) or starvation (S) medium, or in normal medium supplied with Rg2 for 3 h. Results represent mean ± s.e.m. of 3 independent experiments. (B and C) Glucose tolerance test (GTT, B) and insulin tolerance test (ITT, C) of WT (left panel) and BCL2^AAA (right panel) mice pretreated with high-fat diet (HFD) for 3 wk, followed by concurrent treatment with HFD and Rg2 or vehicle for 4 wk. (D and E) Body weight (D) and body fat mass (E) of WT and BCL2^AAA mice after HFD feeding for 3 wk followed by cotreatment of HFD and Rg2 for 4 wk. WAT, white adipose tissue. Results represent mean ± s.e.m. Statistics compare each value in HFD and Rg2-treated groups to the one in HFD and vehicle-treated groups. N=5-6. *, P < 0 .05; **, P < 0 .01; ***, P < 0 .001; NS, not significant, t test.

Display full size

To further test this hypothesis in vivo, we used the HFD-induced type 2 diabetes model. We found that 4 wk of Rg2 cotreatment with HFD in WT mice significantly decreased HFD-induced glucose intolerance (by glucose tolerance tests) and insulin resistance (by insulin tolerance tests), whereas Rg2 treatment failed to do so in the autophagy-deficient BCL2^AAA mice (Fig. 4B and C). BCL2^AAA mice contain 3 knock-in alanine mutations (Thr69Ala, Ser70Ala and Ser84Ala) at the phosphorylation residues of BCL2 (a binding partner and inhibitor of BECN1/Beclin 1), which block BCL2 phosphorylation, BECN1 release from inhibitory binding of BCL2, and autophagy induction in response to stress.Citation³⁶ Thus, these data suggest that activation of autophagy is essential for Rg2-mediated prevention of insulin resistance induced by HFD. In addition, Rg2 treatment strongly reduced the size of liver lipid droplets, body weight gain and fat mass under HFD, to a level similar to regular diet feeding (Fig. S9, Fig. 4D and 4E). Nonetheless, different from insulin sensitivity and liver lipid droplets, the regulation of body weight and fat mass by Rg2 seems to be independent of autophagy activity, as Rg2 exerted similar effects on both WT and the autophagy-defective BCL2^AAA mice. Thus, these data suggest that Rg2 improves several metabolic parameters upon HFD feeding, including insulin sensitivity, lipid droplet size, fat mass and obesity. Specifically, Rg2 ameliorates HFD-induced lipid droplet deposition and insulin resistance through activating autophagy.

Rg2 promotes clearance of aggregated-prone proteins in cells and in brain

The accumulation of protein aggregates has been implicated in the pathogenesis of several neurodegenerative diseases. One approach by which these aggregates can be eliminated is through the activation of autophagy.Citation^39-43 A number of studies have also reported an associative role of Panax ginseng on cognition improvement in rodents and humans.Citation^16-18 In addition, we found that Rg2 is able to enter the brain by the pharmacokinetics study (Fig. S7). Its level peaks 15 min after injection, gradually decreases over 8 h postinjection, and is still detectable after 24 h postinjection. Thus, we propose that the ginsenoside Rg2 may mediate the protective effects against neurodegeneration by enhancing the autophagic clearance of aggregate-prone proteins.

Accordingly, we next determined whether Rg2 affects the clearance of proteinaceous inclusions in vitro, using HeLa cell lines expressing tetracycline-repressible expanded polyglutamine (polyQ)-repeat protein HTT (huntingtin), HTT65Q and HTT103Q.Citation⁴² Unlike the HTT protein with the normal number of glutamine repeats (HTT25Q), HTT65Q and HTT103Q formed insoluble polyQ aggregates larger than 0.2-μm diameter, which were detected by filter trap assays (Fig. 5A). We found that Rg2 treatment decreased accumulation of both HTT65Q and HTT103Q aggregates. The effect of Rg2 was abolished upon knockdown of the essential autophagy gene ATG7 (Fig. 5A), suggesting that Rg2-induced aggregate reduction is ATG7-dependent, and thus autophagy-dependent. This was also confirmed by immunofluorescence imaging analysis, which showed that the number of cells positive for HTT aggregates was reduced upon Rg2 administration (Fig. 5B). The Rg2-mediated reduction in HTT-positive cells was not due to reduced cell viability, as cells treated for 24 h (the same treatment time used to measure HTT aggregate clearance) grew as well as the control cells without mutant HTT expression (+Tet) (Fig. S10A). Thus, these data suggest that autophagy induction by Rg2 enhances clearance of HTT aggregates in vitro.

Figure 5. Rg2 decreases polyglutamine aggregation in an autophagy-dependent manner. (A) Filter trap assay (upper) and quantification (lower) of stable HeLa cells conditionally expressing HTT25Q-CFP, HTT65Q-CFP or HTT103Q-CFP in a Tet-off system, in the presence or absence of Rg2 or the indicated siRNA. Cells were transfected with nontargeting control (NC) or ATG7 siRNA 24 h prior to Rg2 treatment for another 24 h. HTT aggregates were analyzed by lysate filtration through 0.2 μm nitrocellulose membrane. Cells treated with tetracycline served as negative control. (B) Representative images (upper panel) and quantification (lower panel) of inclusions formed by CFP-tagged polyglutamine HTT in cells as in (A). Blue, DAPI. Results represent mean ± s.e.m. Scale bar: 20 μm. Statistics compare each value to the one under the “-” condition. **, P < 0 .01; ***, P < 0 .001; NS, not significant, t test.

Display full size

To analyze whether Rg2 is protective in neurodegenerative disorders caused by aggregate-prone proteins, we used a mouse model of Alzheimer disease, the 5XFAD mice, which overexpress a combination of 5 familial Alzheimer disease (FAD) mutations in human APP (amyloid precursor protein) and human PS1 (presenilin 1) proteins.Citation⁴⁴ These mice demonstrate early and aggressive phenotypes of intraneuronal Aβ42 aggregates, β-amyloid plaques and neurodegeneration, and represent a good model for our study. Using dot blot assays of the brain lysates, we found that compared with vehicle (DMSO) treatment, Rg2 injection for 16 wk in 2-mo-old male 5XFAD mice effectively decreased the level of brain Aβ42 (Fig. 6A), the aggregate-prone peptide cleaved from APP, whereas expression of the precursor APP remained unaffected (Fig. S10B). Microscopy data further showed that in Rg2-treated 5XFAD mice, there was a significantly fewer number of amyloid plaques, stained by either thioflavin or Aβ42 antibody (Fig. 6B). Collectively, these data demonstrated that Rg2 reduces intracellular Aβ42 levels and extracellular plaque formation in mouse brain.

Figure 6. Rg2 ameliorates cerebral Aβ accumulation and learning and memory deficits in a 5XFAD Alzheimer mouse model. (A) Representative images (left) and quantification (right) of dot-blot assays on total Aβ42 levels in brain samples of 5XFAD mice treated with vehicle (DMSO) or Rg2 for 4 mo, immunostained with either anti-Aβ42 antibody or IgG as control. Total protein loading was labeled by Ponceau S. Triplicate experiments from 4 mice in each group were shown. (B) Representative images (left) and quantification (right) of amyloid deposits stained by thioflavin or anti-Aβ42 antibody in brain of 5XFAD mice treated with vehicle or Rg2 for 4 mo. (C) Morris water maze test of 5XFAD mice treated with vehicle or Rg2 for 4 mo. Escape latency and total distance traveled in visible platform test and hidden platform test are shown. Results represent mean ± s.e.m. Statistics were analyzed by the IBM SPSS Statistics Tools comparing differences between 2 curves. (D) Contextual fear conditioning test of 5XFAD mice treated with vehicle or Rg2 for 4 mo. Fear conditioning training exposed mice to 2-s electrical foot shocks separated by 1-min interval through a grid floor at the bottom of the chamber. On the following day, the mice were returned to the same chamber and their movements were recorded with a video camera to test for contextual conditioning. Freezing (very low levels of movement) in the training environment indicate context-associated fear. WT mice without APP transgene were used as negative control in the above studies. N = 8. *, P < 0 .05; **, P < 0 .01, t test.

Display full size

Rg2 improves cognitive function in a mouse model of Alzheimer's disease

To determine the effects of Rg2 treatment in vivo, we performed several behavioral assays in the 5XFAD mouse model. For cognitive function, we performed Morris water maze tests. In water maze tests with a visible platform there was no significant difference in either escape latency or distance among WT, vehicle (DMSO)-treated mice or Rg2-treated mice (Fig. 6C, Video S1 to S3), suggesting there was no visual abnormality among all groups. In contrast, in tests with a hidden platform, vehicle-treated 5XFAD mice showed apparent deficiency in learning and memory of the platform location over the time of 5 d, whereas Rg2-treated mice had significantly improved performance in both escape latency and distance similar to WT mice (Fig. 6C, Video S4 to S6). These data suggest that the learning and memory impairment caused by amyloid accumulation is ameliorated by Rg2 treatment. It should be noted that we observed no rescue in the locomotor activity of Rg2-treated 5XFAD mice compared to vehicle-treated ones, as assayed quantitatively by the distance traveled in the open-field tests (Fig. S10C), suggesting that the shortened escape latency in the hidden platform test is not due to overall improvement in physical ability. In addition, in 5XFAD mice with or without Rg2 injection, we tested contextual fear memory, which is mediated by hippocampal function and impaired by amyloid deposition.Citation^45,46 We found that treatment of Rg2 rescued the freezing responses of 5XFAD mice when they were placed in the same context of the foot-shock conditioning chamber (Fig. 6D), suggesting an improvement of the hippocampal-dependent contextual fear memory in these mice. Overall, we concluded that Rg2 treatment improves mouse learning and memory function that is impaired by Aβ aggregation.

Mice

All mice were housed on a 12-h light/dark cycle. 5XFAD mice and the littermate wild-type controls were produced with a C57BL/6JxSJL background, and all other mice with a C57BL/6J background. The global BCL2^AAA knockin and 5XFAD mice have been described previously.Citation^36,44 The BCL2^AAA mice contain 3 knockin mutations in BCL2 that abolish its phosphorylation (Thr69Ala, Ser70Ala and Ser84Ala). The 5XFAD transgenic mice coexpress mutant human APP with 3 FAD mutations, the Swedish (K670N, M671L), Florida (I716V) and London (V717I) mutations, and mutant human PS1 containing 2 FAD mutations, M146L and L286V. Both transgenes are driven by the mouse Thy1 promoter for brain overexpression. Rodent diet containing 60% fat (Research Diets, D12492) was used for high-fat diet treatment. All animal protocols were approved by the Northwestern University Institutional Animal Care and Use Committee (IACUC).

Cell lines

HeLa cell lines were obtained from ATCC, prkaa1 prkaa2 null MEFs were from B. Levine (University of Texas Southwestern Medical Center), and HeLa cells conditionally expressing CFP-tagged HTT with polyQ repeats were from A. Yamamoto (Columbia University). Cells were cultured in DMEM medium (Gibco, 11995073) supplemented with 10% fetal bovine serum (FBS). Tetracycline-free FBS was used for HeLa cells stably expressing HTT (Takara Bio USA, 631107), and regular FBS was used for all other cells (HyClone, SH30070.03HI).

Isolation of plant-derived metabolites

The plant chemicals were prepared in the following manner: (1) The ginseng polysaccharide (WGP, water-soluble ginseng polysaccharide) was obtained from the roots of Panax ginseng C. A. Meyer by hot-water extraction followed by precipitation with 80% ethanol and deproteination by the Sevag reagent. WGP was resolved into 2 fractions, WGPN (water-soluble ginseng polysaccharide neutral fraction) and WGPA (water-soluble ginseng polysaccharide acidic fraction), by anion-exchange chromatography (GE Healthcare Life Sciences, 17070905, Sweden).Citation⁷⁷ (2) AHP (acidic-soluble Helianthus polysaccharide) was prepared from the disk of sunflower (Helianthus annuus L.) by extraction with 0.2% oxalic acid followed by precipitation with 60% ethanol. AHP was also fractionated into 2 fractions, AHPN (acidic-soluble Helianthus polysaccharide neutral fraction) and AHPA (acidic-soluble Helianthus polysaccharide acidic fraction), by anion-exchange chromatography. AHPA was further purified with gel filtration chromatography using Sepharose CL-6B column into 2 fractions, AHPA-1 (acidic-soluble Helianthus polysaccharide acidic fraction 1) and AHPA-2 (acidic-soluble Helianthus polysaccharide acidic fraction 2). (3) The polysaccharide of Veratrum nigrum L, VP (Veratrum nigrum polysaccharide), was extracted by hot water and 80% ethanol precipitation from the roots of Veratrum nigrum L. VP was further graded into 2 fractions, VPN (Veratrum nigrum polysaccharide neutral fraction) and VPA (Veratrum nigrum polysaccharide acidic fraction), through DEAE-cellulose column chromatography. (4) WRPP (water-soluble polysaccharide from bee pollen of Rosa rugosa) was obtained from bee pollen of Rosa rugosa by extraction with hot water and precipitation with 80% ethanol. WRPP was further purified through DEAE-cellulose column chromatography into 3 fractions, WRPP-N (water-soluble polysaccharide from bee pollen of Rosa rugosa neutral fraction), WRPP-1 (water-soluble polysaccharide from bee pollen of Rosa rugosa acidic fraction 1) and WRPP-2 (water-soluble polysaccharide from bee pollen of Rosa rugosa acidic fraction 2).Citation⁷⁸ (5) The ginsenoside Rg2 was purified from the leaf-stems of Panax ginseng C. A. Meyer. Total ginsenosides were extracted with hot water and then concentrated. After purification by a macroporous resin column, Rg2 was separated through an HPLC system, using analytical Shim-pack PREP-ODS (H) column (4.6 mm × 250 mm, 5 μm) connected to an HPLC system (LC-10AT vp pump, SPD-10A vp UV–VIS detector, Shimadzu, Kyoto, Japan), eluted at the flow rate of 1.0 ml/min with the following gradient program, 0 to 10 min, 32% acetonitrile (in distilled water, v/v), 10 to 40 min, 32% to 60% acetonitrile, and 40 to 50 min, 60% acetonitrile, monitored by the absorbance at 203 nm. (6) The gypenoside XVII (GXVII) was transformed from ginsenoside Rb1, which was purified from the roots of Panax ginseng C. A. Meyer. The transformation product GXVII was separated through an HPLC system, using an analytical Shim-pack PREP-ODS (H) column (4.6 mm × 250 mm, 5 μm) connected to an HPLC system (LC-10AT vp pump, SPD-10A vp UV–VIS detector, Shimadzu, Kyoto, Japan), eluted at the flow rate of 1.0 ml/min with the following gradient program, 0 to 10 min, 32% acetonitrile (in distilled water, v/v), 10 to 40 min, 32% to 60% acetonitrile, and 40 to 50 min, 60% acetonitrile, monitored by the absorbance at 203 nm.

Filter trap assay

HeLa cells stably expressing CFP-HTT25Q, CFP-HTT65Q and CFP-HTT103Q were treated with using 100 ng/ml tetracycline (IBI Scientific, IB02200), or 0.1 mM Rg2 with control or ATG7 siRNA (GE Dharmacon ON-TARGETplus control or ATG7 SMARTpool siRNA) for 48 h. Cells were then collected and lysed in lysis buffer (50 mM Tris-HCl, pH 8.8, 100 mM NaCl, 5 mM MgCl₂) containing 0.5% NP-40 (Sigma, 74385) for 30 min at 4°C. After centrifugation, the pellet was digested with 0.5 mg/ml DNaseI (Sigma, D5025; in 20 mM Tris-HCl, pH 8.0) for 1 h at 37°C, and dissolved into lysates by the addition of 2% SDS (Sigma, L3771), 50 mM DTT, 20 mM EDTA. The lysates containing insoluble aggregates were then added onto 0.2-μm nitrocellulose membrane that was preequilibrated with 2% SDS-TBS (20 mM Tris-HCl, pH 7.5, 500 mM NaCl) for 30 min, and were filtered through the membrane by gentle vacuum using the Bio-Dot SF microfiltration apparatus (Bio-Rad, Hercules, California, USA). The signal was detected by immunostaining with the HRP-conjugated GFP antibody (Santa Cruz Biotechnology, sc9996; 1:1000).

Glucose and insulin tolerance tests

Mice were fasted for 4 h in glucose tolerance tests (GTT) and insulin tolerance tests (ITT) with free access to water. Glucose tolerance tests were performed by intraperitoneal (i.p.) D-glucose injection at the dose of 1.5 g/kg body weight, and insulin tolerance tests were performed by i.p. insulin (Sigma-Aldrich, I0516) injection at the dose of 0.75 U/kg body weight. Blood was drawn from tail veins at baseline and 15, 30, 60 and 120 min after injection, and the serum glucose level was measured by the OneTouch Ultra Glucometer (Abbott, Chicago, IL, USA). The data were plotted as blood glucose concentrations over time.

Morris water maze testing

Before testing, 8-wk-old 5XFAD male mice were treated with i.p. injection of 20 mg/kg Rg2 (or DMSO vehicle) once/d, 5 d/wk, for 16 wk. Age-matched wild-type (WT) male mice were used as a negative control. The Morris water maze test consists of 2 sections: the visible platform testing and hidden platform testing. During the tests, mice were placed in the water tank filled with opaque water (maintained at 25°C), with their heads facing toward the tank wall. In the visible platform section, a black platform extending 2 cm above the water level was used for these trials. For each trial, the platform was randomly positioned, and the mouse was placed in the tank at different start positions. The trial was stopped after the mouse found and climbed onto the platform, and the escape latency was recorded. The trial was stopped if the mouse did not climb onto the platform in 60 s, and the experimenter guided it to the platform. Mice were tested for 4 d with 8 trials per day. In the hidden platform section, a transparent platform underneath the water level was used instead of the black one during all trials, mice were tested with a fixed platform location over 5 d period with 6 trials per day, and they were allowed to search the platform in 60 s. In the tests, 2 parameters were evaluated: the trail duration (s) and distance to the platform (m).

Fear conditioning testing

5XFAD and WT mice were treated with Rg2 or DMSO (vehicle) the same as for Morris water maze analyses. On the training day (d 1), the mouse was placed in the fear conditioning chamber and then given 5 trace fear conditioning trials. On each trial, the tone (30-s duration, 75 dB, Hiss-pulsed at 5 Hz, 50% duty cycle, 5-ms rise/fall time) was followed by a 20-s empty trace interval and then a shock (1-s duration, 0.7-mA intensity) through the grid floor. After the final trial, the mouse was immediately returned to its cage. The test chamber was wiped with cleaner between the testing of each mouse. On the testing day (d 2), the mouse was placed in the same conditioning chamber for a 360-s period, during which no stimuli were presented and its movements were recorded. The result was presented as percentage of freezing duration in the total duration.

Dot blot assay

Mice were euthanized and perfused with phosphate-buffered saline (PBS; Sigma-Aldrich, D8537). Brain tissues were dissected and homogenized. The supernatant was extracted in 5 M GuHCl, 50 mM Tris, pH 7.4 overnight at room temperature, added on 0.2-μm nitrocellulose membrane, and dried for 1 h at 37°C. The membrane was stained with Ponceau S, and the dot blot signal on the membrane was detected by immunostaining with Aβ42 (Invitrogen, 700254; 1:1000) antibody and HRP-conjugated secondary antibody (Santa Cruz Biotechnology, sc2004; 1:2000).

Immunofluorescence microscopy

Frozen brain sections (5 μm) from WT and 5XFAD mice treated with Rg2 or DMSO (vehicle) were immunostained with Aβ antibody (Invitrogen, 700254; 1:500) and Alexa Fluor 594 goat anti-rabbit antibody (ThermoFisher Scientific, A11012). Additional sections were immnunostained with 1% thioflavin S (Sigma-Aldrich, 230456), and analyzed by fluorescence microscopy (ECLIPSE Ti-E Inverted Microscope, Nikon, Melville, New York, USA) under 10 x objective.

Autophagy analyses

HeLa cells stably expressing GFP-LC3 were cultured in normal medium, starvation (EBSS, Earle's balanced salt solution; Sigma-Aldrich, E7510) medium, or normal medium with the indicated drugs for 3 h or 16 h. The cells were fixed with 2% paraformaldehyde (PFA) for 20 min, and GFP-LC3 puncta were quantified by fluorescence microscopy. For assessment of autophagy in vivo, 8-wk-old male GFP-LC3 mice were starved for 48 h, or treated with the indicated drugs once daily for 3 d by i.p. injection, and then anaesthetized by isoflurane and perfused with 4% PFA. Brain samples were fixed in 4% PFA overnight, 15% sucrose for 4 h and 30% sucrose overnight before frozen sections were prepared. The number of GFP-LC3 puncta per unit area of tissue was quantified by fluorescence microscopy. Autophagy in vivo was also analyzed by western blot analysis of brain tissue extracts with antibodies against LC3 and SQSTM1 (see below for details).

Electron microscopy (EM)

Mice were euthanized and liver tissues were rapidly fixed in 2.5% glutaraldehyde in 0.1 M cacodylate buffer. EM was performed on the Tecnai Spirit G2 microscope (FEI, Hillsboro, Oregon, USA) as previously described.Citation⁷⁹

Western blot analyses

Cell or mouse brain extracts were prepared in lysis buffer containing 50 mM Tris, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100 (Bio-Rad, 161-0407), proteinase inhibitor cocktail (Roche Applied Sciences, 11873580001) and halt phosphatase inhibitor cocktail (ThermoFisher Scientific, 78420), and subjected to western blot analysis with anti-LC3 (Novus Biologicals, NB100-2220), anti-SQSTM1 (Abnova, H00008878-M01), anti-p-INSR (insulin receptor; Cell Signaling Technology, 3024), anti-p-AKT (Ser473) (Cell Signaling Technology, 4060), anti-INSR (Cell Signaling Technology, 3025), anti-AKT (pan) (Cell Signaling Technology, 4691), or anti-ACTB/β-actin-HRP (Santa Cruz Biotechnology, sc47778 HRP) antibodies.