J-Perk v6.10 serial key or number

The PERK-eIF2α branch of the Unfolded Protein Response (UPR) mediates the transient shutdown of translation in response to rising levels of misfolded proteins in the endoplasmic reticulum. PERK and eIF2α activation are increasingly recognised in postmortem analyses of patients with neurodegenerative disorders, including Alzheimer’s disease, the tauopathies and prion disorders. These are all characterised by the accumulation of misfolded disease-specific proteins in the brain in association with specific patterns of neuronal loss, but the role of UPR activation in their pathogenesis is unclear. In prion-diseased mice, overactivation of PERK-P/eIF2α-P signalling results in the sustained reduction in global protein synthesis, leading to synaptic failure, neuronal loss and clinical disease. Critically, restoring vital neuronal protein synthesis rates by inhibiting the PERK-eIF2α pathway, both genetically and pharmacologically, prevents prion neurodegeneration downstream of misfolded prion protein accumulation. Here we show that PERK-eIF2α-mediated translational failure is a key process leading to neuronal loss in a mouse model of frontotemporal dementia, where the misfolded protein is a form of mutant tau. rTg mice, which overexpress the PL tau mutation, show dysregulated PERK signalling and sustained repression of protein synthesis by 6 months of age, associated with onset of neurodegeneration. Treatment with the PERK inhibitor, GSK, from this time point in mutant tau-expressing mice restores protein synthesis rates, protecting against further neuronal loss, reducing brain atrophy and abrogating the appearance of clinical signs. Further, we show that PERK-eIF2α activation also contributes to the pathological phosphorylation of tau in rTg mice, and that levels of phospho-tau are lowered by PERK inhibitor treatment, providing a second mechanism of protection. The data support UPR-mediated translational failure as a generic pathogenic mechanism in protein-misfolding disorders, including tauopathies, that can be successfully targeted for prevention of neurodegeneration.

Endoplasmic reticulum (ER) stress and its associated Unfolded Protein Response (UPR) are emerging as major common themes in neurodegenerative disorders [13, 33]. The modulation of proteostasis is increasingly a potential therapeutic target in a wide range of human diseases associated with protein misfolding [12]. The UPR involves three major signalling cascades that enable the cell to deal with the build up of misfolded (or unfolded) proteins, which are detected by sensors within the ER [30]. One branch of the UPR, mediated by protein kinase RNA (PKR)-like/Pancreatic ER kinase (PERK), leads to the transient shutdown of protein synthesis. Activated (phosphorylated) PERK (PERK-P) phosphorylates the alpha subunit of eukaryotic initiation factor 2 (eIF2α); this inhibits the initiation of translation and induces expression of the transcription factor ATF4, and associated downstream signalling events, including induction of GADD34, the stress-induced eIF2α-P-specific phosphatase. (eIF2α can also be phosphorylated by other kinases as part of the integrated stress response (ISR), in response to cellular stresses such as viral infection and amino acid starvation). Increased PERK-P and eIF2α-P have been reported in postmortem analyses of brains from patients with Alzheimer’s disease (AD), Parkinson’s disease (PD) and the tauopathies Progressive Supranuclear Palsy (PSP) and Frontotemporal Dementia (FTD), as well as in prion disease [3, 15, 16, 37, 38]. In AD patients, UPR activation is thought to occur early in disease pathology in association with the accumulation of abnormal phosphorylated tau in neurons, but before neurofibrillary tangle formation [16, 26]. PERK-P staining has also been shown to be associated with phospho-tau-positive neurons in CA1 and dentate gyrus and subiculum regions of the hippocampus in FTD-tau cases [26].

However, it is unclear what the pathological significance of these findings is for the human diseases; specifically, whether they reflect activation of a protective mechanism or the signature of a pathogenic process (see Scheper and Hoozemans, for review [33]). This is of key significance for both mechanistic understanding and for development of potential treatments. Important insights come from animal models. PERK branch UPR overactivation occurs in several mouse models of neurodegenerative disease, including AD [1, 9, 13], prion [24], tauopathy [1, 18] and ALS [3, 32]. In prion-diseased mice, accumulation of misfolded prion protein (PrP) causes persistently high levels of PERK-P and eIF2α-P, and hence sustained translational failure. This results in a catastrophic decline in levels of key synaptic proteins, leading to synaptic failure and ultimately to neuronal loss [24]. This is true for both PrP overexpressing mice and for wild-type mice, supporting that PERK-eIF2α overactivation is not an artefact of high levels of protein expression, but occurs at endogenous levels (consistent with observations in humans). Critically, both genetic manipulations that reduce eIF2α-P levels [24] and pharmacological inhibition of PERK using the inhibitor GSK [4], or with the small molecule inhibitor, ISRIB, which acts downstream of eIF2α-P, restored vital protein synthesis rates and prevented neurodegeneration and clinical disease in prion-infected mice [11, 23, 24]. In contrast, pharmacological treatment preventing the reduction of eIF2α-P levels accelerated disease and exacerbated toxicity in prion-diseased mice [24]. Given the range of neurodegenerative diseases in which PERK branch UPR dysregulation is seen, it is crucial to understand its role in mediating neuronal loss in protein-misfolding disorders more broadly, particularly in AD and other dementias. This will also help determine its potential as a therapeutic target in these disorders. Wider relevance of the pathway in AD is supported by the fact that both PERK haploinsufficiency [9] and genetic suppression of other eIF2α kinases [21] rescue memory deficits and neurodegeneration in mouse models of AD. Indeed, neuronal eIF2α-P levels—and hence rates of protein synthesis—in wild-type mice are key to learning and memory [6], and pharmacological inhibition of eIF2α-P signalling boosts cognition [35]. Further, with relevance for AD and other tauopathies, PERK branch UPR activation is also known to contribute to the phosphorylation of tau—which is central to pathology in these diseases—both in vitro and in vivo [14, 20, 39]. In vitro, this has been shown to be due to PERK-mediated induction of glycogen synthase kinase-3 (GSK3β), a serine/threonine kinase that phosphorylates tau at disease-relevant epitopes [25], an effect that can be reversed by PERK inhibition [39]. Thus, UPR activation may have a dual pathological role in tauopathies, impacting on two processes fundamental to neurodegeneration in these disorders.

We have now examined the role of PERK pathway activation in the rTg mouse model of FTD caused by the human tau PL mutation [29, 31]. Mice positive for the mutant tau transgene (tau⁺_PL) show high levels of PL tau expression throughout their lives, with age-related tau pathology, progressive memory deficits and, importantly, extensive forebrain neurodegeneration from  months of age [29, 31]. Tau⁺_PL rTg mice show abnormal conformations of tau from  months, with pre-tangle hyperphosphorylated tau detected by 4 months of age, with the appearance of mature neurofibrillary tangles and neurodegeneration in the hippocampus from  months. Neuronal loss becomes widespread throughout the forebrain after 7 months of age, leading to marked forebrain atrophy associated with clinical signs of poor grooming and motor impairment. Transgene-negative (tau⁻_PL) rTg mice have normal brain morphology, behaviour and appearance and do not show phosphorylated tau [29, 31]. Increased levels of PERK-P in the brain of these mice have been reported during later disease stages (from 9 months), although the role of PERK-P signalling in mediating neurodegeneration in these animals has not previously been examined [1].

We first confirmed activation of PERK-eIF2α signalling in rTg mice due to mutant tau expression. We then examined the effects of therapeutic inhibition of PERK signalling on both rates of neuronal protein synthesis and on tau phosphorylation and their role in neurodegeneration in these animals.

Mice

All animal work conformed to the UK Home Office regulations and institutional guidelines. rTg tau⁺_PL and tau⁻_PL mice were obtained from Eli Lilly and Company. Treatment with doxycycline was performed from 4 months of age, by oral dosing as described [31]. rTg mice were given two oral (bolus) doses of 10 mg/kg, then followed by Harlan Teklad T, diet with doxycycline  mg/kg in chow (n = 6 female mice). Treatment with GSK was by oral gavage twice daily with 50 mg/kg GSK suspended in vehicle ( % HPMC +  % Tween in H₂O at pH ) (n = 10 male), or with vehicle alone (n = 8 male) as described [23] from ~6 months of age.

Western blotting

Protein extraction from hippocampi and immunoblots of UPR and synaptic proteins were performed as described [23, 24]. Proteins were detected with the following antibodies: eIF2α-P (S51) (; , Cell Signaling) and eIF2α (L57A5) (; , Cell Signaling), ATF4 (CREB-2, ; sc, Santa Cruz), GADD34 (, AP, Proteintech), PERK-P (, , Santa Cruz), PERK (, , Cell Signaling), SNAP25 (,, ab, Abcam), PSD (, EPY, Millipore) total tau (Tau 5, , AHB, Invitrogen), GSK3β (, , Cell Signaling), pSer⁹-GSK3β (, , Cell Signaling) and pTyr^/-GSK3α/β (, , Millipore). Horseradish peroxidase (HRP)-conjugated secondary antibodies (,; DAKO) were applied, and protein was visualised using enhanced chemiluminescence (GE Healthcare) and quantitated using ImageJ. Antibodies against GAPDH (; sc, Santa Cruz) and Beta-III-tubulin (, MAB, Millipore) were used to determine loading.

Protein synthesis rates

Global translational levels (protein synthesis rates) were calculated by measuring ³⁵S-methionine incorporation into proteins in acute hippocampal slices, as described [23, 24]. Briefly, slices were incubated with  mBq of ³⁵S-methionine label in oxygenated artificial cerebrospinal fluid at 37 °C for 1 h. Incorporation of radiolabel was measured by scintillation counting (WinSpectral, Wallac Inc.). All biochemical analyses were performed in hippocampi from three to four mice.

Histology

Paraffin-embedded brains were sectioned at 5 μm and stained with haematoxylin and eosin or NeuN antibody (; MAB, Millipore) for neuronal counts as described [23, 24]. CA1 pyramidal neuron counts were determined using three serial sections from three to four separate mice [24]. All neuronal counting was performed with the investigator being blind to the sample group being analysed. Immunohistochemistry for pSer/Thr tau was performed using AT8 (; MN, Thermo Scientific). Non-specific binding was blocked prior to primary antibodies using Histostain-Plus Bulk kit (Invitrogen). A biotinylated secondary antibody (Invitrogen) was used and stain visualised by diaminobenzidine reagent. All images were taken on using Axiovision software (Zeiss) and counted using Volocity imaging system (Version ) [24]. Immunofluorescence for pSer/Thr tau was performed using AT8 (; MN, Thermo Scientific) and PERK-P (, , Santa Cruz). Sections were blocked and permeabilized in blocking buffer (2 % donkey serum,  % Triton X in TBS) for 1 h and then incubated overnight at 4 °C with primary antibody in TBS+2 % serum. Alexa Fluor  nm and  nm (, Invitrogen) were then applied for 1 h. Slides were then mounted with VECTASHIELD Mounting media with DAPI (Vector laboratories). All images were taken using Zeiss LSM confocal microscope and counted using Volocity imaging system.

Phospho-tau biochemical analysis

To detect levels of phosphorylated tau in total lysates AlphaScreen assays (Amplified Luminescent Proximity Homogeneous Assay Screen) were performed as described in [2] and as per manufacturer’s instructions. Phospho-tau-specific AT8 (pSer/Thr tau) and PHF1 (pSer^/ tau) antibodies were used. Briefly, optimised antibody/acceptor bead mix was incubated overnight at 4 °C with sample lysate or standard diluted in AlphaScreen assay buffer. Next, streptavidin-coated donor beads were added and incubated in the dark at room temperature for 4 h. Plates were read at excitation  nm and emission – nm using an Envision plate reader.

XBP1-splicing assay

Total RNA was extracted from hippocampi with the mirVana isolation kit (Ambion). RNA samples were reversed-transcribed with oligo(dT) primers using ImProm-II Reverse Transcriptase (Promega). XBP1 mRNA was amplified with primers flanking the 26b intron (5′-GGAGTGGAGTAAGGCTGGTG and 5′-CCAGAATGCCCAAAAGGATA) using Phusin High-Fidelity taq polymerase (New England Biolabs) [23]. PCR products were resolved on 3 % agarose gels. MEF cells were treated with 5 μg/ml tunicamycin for 6 h and used as a positive control for XBP1 splicing.

Statistics

All statistical analyses were performed using Prism V6 software using Student’s t test for data sets with normal distribution and a single intervention. ANOVA testing was performed using one-way analysis with Tukey’s post hoc test for multiple comparisons.

High levels of total mutant tau expression lead to activation of the PERK branch of the UPR and sustained translational repression in rTg FTD mice

We assessed PERK branch UPR activation in the brains of rTg mice by measuring levels of eIF2α-P, ATF4, GADD34 and global protein synthesis rates over the course of disease. PL transgene-expressing mice (tau⁺_PL) express high levels of mutant tau from birth (repressible by doxycycline administration [31]). Characteristically, they show memory impairment and hippocampal neuronal loss by 6 months, with widespread forebrain atrophy and overt clinical signs by 8 months [29, 31] (Fig. 1a). We tested mice from 4 months of age. All tau⁺_PL mice expressed high levels of total tau compared to low levels in non-transgenic controls, which express only wild-type (murine) tau (Fig. 1b).

Levels of eIF2α-P and ATF4 were low in all animals until 6 months of age, by which time there was marked elevation of eIF2α-P and ATF4 levels in tau⁺_PL mice compared to non-transgenic littermates (Fig. 1b), consistent with activation of PERK signalling. Interestingly, GADD34 levels did not increase over time (Fig. 1b), which is unexpected given induction of ATF4, but consistent with the effects of PERK branch overactivation seen in prion-infected mice [24]. The failure to induce GADD34 would explain the persistent high levels of eIF2α-P seen in tau⁺_PL older mice (Fig. 1b), again similar to the sustained elevation of eIF2α-P levels seen in prion-diseased mice, both wild type and overexpressing [24]. We found no change in the other two signalling cascades of the UPR involving IREmediated XBP1 splicing (Fig. 1c) and cleavage of ATF6 (data not shown) throughout the time course of disease in rTg transgene-positive mice, consistent with findings in prion-infected mice [23].

Consistent with PERK branch eIF2α-P activation, rates of global protein synthesis measured in brain slices of tau⁺_PL mice had declined to ~40 % of those seen in non-transgenic littermates by 6 months of age, remaining repressed until 11 months (Fig. 1d), after which time the animals were killed. Protein synthesis rates were normal up to 5 months of age, suggesting the onset of sustained translational repression is between 5 and 6 months (Fig. 1d). Levels of the pre- and post-synaptic proteins SNAP25 and PSD were equivalent in transgene-positive and -negative mice at 4 months, but reduced in tau⁺_PL mice at 6 months (Fig. 1e), reflecting the drop in protein synthesis due to high levels of eIF2α-P at this time. As in prion-diseased mice [24], the onset of eIF2α-P-mediated translational repression was closely associated with onset of neurodegeneration at 6 months in rTg mice, which is absent at 4 months (Fig. 1f).

To confirm that activation of PERK-eIF2α-P signalling is indeed the result of mutant tau expression in rTg tau⁺_PL mice, we tested the effect of repressing tau levels on eIF2α-P and ATF4 levels using doxycycline administration. Treatment of transgene-positive mice with doxycycline from 4 months resulted in marked reduction in total tau levels by 8 months (although not as low as levels in transgene-negative controls that express only murine tau) (Fig. 1g). This produced a corresponding marked reduction in eIF2α-P and ATF4 levels compared to levels seen in untreated tau⁺_PL mice, and equivalent to those seen in non-transgenic mice (Fig. 1g). Our data are, therefore, consistent with UPR activation being a direct result of elevated levels of mutant total tau in these mice. Consistent with this, reduced PERK-P immunostaining after doxycycline treatment in rTg mice has also been reported by others [1].

Thus, the onset of PERK/eIF2α-mediated translational repression is closely associated with synaptic protein loss and onset of neurodegeneration in mutant tau-expressing rTg mice, similar to the temporal relationship between these events in prion-diseased animals [24]. Importantly, although it is true that levels of mutant tau overexpression are very high in this model, the pR5 mouse model of tauopathy in which levels of mutant tau expression are much lower, at ~70 % endogenous tau, also shows UPR activation [17, 18]. This is similar to the detection of PERK/eIF2α activation in prion-diseased mice, which is seen in both PrP overexpressing and wild-type mice [24], and is, therefore, unlikely to reflect an artefact of protein overexpression.

PERK inhibitor treatment reduces eIF2α-P levels and restores global translation rates

Given the known neuroprotective effects of restoring protein synthesis rates in prion-diseased mice downstream of misfolded PrP accumulation [11, 23], we first asked if restoring translation through PERK inhibition in tau transgene-expressing mice was similarly neuroprotective. We treated tau⁺_PL mice with the specific PERK inhibitor GSK [4] or vehicle alone from ~6 months of age, when eIF2α-P levels are known to be elevated (Fig. 1b) and neuronal loss is beginning (Fig. 1f; [31]). (These were proof-of-principle studies, as GSK is known to produce pancreatic toxicity despite profound neuroprotection after prolonged treatment in prion-diseased mice [23]. The question was whether similar neuroprotective effects of reversing translational repression—achieved through PERK inhibition—are seen in this model; not whether GSK is a viable treatment).

rTg tau⁺_PL mice received 50 mg/kg GSK (n = 10) or vehicle (n = 8) by oral gavage twice daily, a dose optimised for good levels of brain penetration, as described in [23], for 2 months (Fig. 2a). At this time point, at 8 months of age, animals were killed for analysis. As predicted, and consistent with previous observations in prion-diseased mice [23], GSK treatment significantly reduced levels of PERK-P, eIF2α-P and ATF4 levels in tau⁺_PL mice, compared to vehicle-treated animals, which showed persistently elevated of levels of these proteins (Fig. 2b, c). Critically, PERK inhibitor treatment restored global protein synthesis rates to normal in 8-month-old tau⁺_PL mice, in contrast to vehicle-treated animals, which showed markedly reduced translation rates at this time point (Fig. 2d). (GSK does not affect eIF2α-P levels and global protein synthesis where the UPR is not activated [23]).

PERK inhibitor treatment is neuroprotective in FTD mice: reducing tau phosphorylation, and brain atrophy and abrogating clinical signs

Importantly, PERK inhibitor treatment was neuroprotective in rTg mice, as in prion-diseased mice [23], partially restoring neuronal numbers, maintaining total brain weight and preventing clinical signs in 8-month-old animals, and preventing any progression of neuronal loss from 6 months of age, when treatment was begun (Fig. 3a–c). Thus, all vehicle-treated tau⁺_PL mice (8/8) showed poor grooming, hunched posture, hind-leg clasping and/or poor mobility by 8 months of age (Fig. 3a, panel ii). In contrast, all PERK inhibitor-treated tau⁺_PL mice (10/10) showed normal grooming, posture and movement at this stage (Fig. 3a, panel iii). Indeed, PERK inhibitor-treated tau⁺_PL mice were indistinguishable clinically from their transgene-negative tau⁻_PL littermates (Fig. 3a, panel i). Histological examination confirmed marked neuroprotection in tau⁺_PL mice treated with GSK, compared to vehicle-treated mice, which showed profound hippocampal neuronal loss (Fig. 3a, compare panels v and vi), characteristic of the extensive forebrain neurodegeneration described in these mice at this stage [31]. GSK treatment resulted in preservation of ~60 % of CA1 neurons at 8 months, compared to ~25 % in vehicle-treated mice at this time (Fig. 3b). The protective effect is notable especially given that treatment begun at 6 months, when neuronal loss is already beginning in the hippocampus (Fig. 1f; [29, 31]). On a macroscopic scale, PERK inhibitor treatment significantly reduced brain atrophy, with greater total brain weights compared to untreated transgene-expressing animals at this stage (Fig. 3c). Both the number of CA1 neurons and brain weights of PERK inhibitor-treated animals at 8 months were very similar to neuronal numbers reported in untreated tau⁺_PL rTg mice at  months by other workers [29, 31

Ginseng is one of the most commonly used herbs that is believed to have a variety of biological activities, including reducing blood sugar and cholesterol levels, anti-cancer, and anti-diabetes activities. However, little is known about the molecular mechanisms involved. In this study, we showed that protopanaxadiol (PPD), a metabolite of the protopanaxadiol group ginsenosides that are the major pharmacological constituents of ginsengs, significantly altered the expression of genes involved in metabolism, elevated Sestrin2 (Sesn2) expression, activated AMPK, and induced autophagy. Using CRISPR/CAS9-mediated gene editing and shRNA-mediated gene silencing, we demonstrated that Sesn2 is required for PPD-induced AMPK activation and autophagy. Interestingly, we showed that PPD-induced Sesn2 expression is mediated redundantly by the GCN2/ATF4 amino acid-sensing pathway and the PERK/ATF4 endoplasmic reticulum (ER) stress pathway. Our results suggest that ginseng metabolite PPD modulates the metabolism of amino acids and lipids, leading to the activation of the stress-sensing kinases GCN2 and PERK to induce Sesn2 expression, which promotes AMPK activation, autophagy, and metabolic health.

Autophagy is an evolutionarily conserved self-digestive process via which cells adapt to nutrient starvation and other stress conditions^1,2. Defects in autophagy is associated with a number of pathological conditions, including cancer, infectious diseases, myopathies, and neurodegenerative disorders^3,4. Under normal growth conditions, autophagy is kept at a basal level mainly for housekeeping purposes such as degradation of long-lived proteins and turnover of damaged cellular organelles. Under stress conditions such as nutrient starvation, ER stress, oxidative stress, and hypoxia, autophagy is induced to provide cells with additional internal nutrient supplies and promote cellular survival^5,6,7,8. The molecular pathway that links nutrient depletion and stress to autophagy involves the energy sensor AMP-activated protein kinase (AMPK) and the mechanistic target of rapamycin complex 1 (mTORC1). AMPK was shown to be associated with ULK1 (unclike kinase 1) and phosphorylation of ULK1 by AMPK was required for autophagy initiation⁹. In addition, AMPK can also promote autophagy by inhibition of autophagy-negative regulator mTORC1¹⁰.

Sestrin2 (Sesn2) is a member of the highly conserved antioxidant proteins, which is induced by a variety of cellular stress and functions to activate AMPK, inhibit mTORC1, and suppress reactive oxygen species accumulation^11,12,13,14. This allows cells to adjust their metabolism to adapt to different cellular stresses. Consistent with this, in vivo studies revealed that Sesn2 is important for metabolic homeostasis. Sesn2-deficient mutants showed diverse age- and obesity-associated metabolic pathologies such as accumulation of lipid droplets and protein aggregates, mitochondrial dysfunction, and insulin resistance. These pathologies were related to defective autophagy and the misregulation of AMPK and mTORC1 signaling^15,16,17. Importantly, restoration of AMPK activation suppressed the observed defects in Sestrin-deficient fly and mouse mutants^16,17,18, suggesting activation of AMPK is a key in vivo function of Sesn2. In addition, Sesn2 can serve as a Leu sensor and regulate mTORC1 activity directly through modulation of GATOR complexes¹⁹.

In response to different environmental stresses, the α-subunit of eukaryotic initiation factor 2 (eIF2α) is phosphorylated by distinct kinases, which represses global translation initiation but selectively enhances the translation of ATF4, a master regulator controlling the transcription of downstream target genes essential for adaptive responses, including genes involved in metabolism, cell survival, apoptosis, and autophagy^20,21,22. In mammals, four protein kinases, GCN2, PERK, HRI, and PKR, are known to phosphorylate eIF2α in response to distinct upstream stress signals to modulate translation initiation. The GCN2 kinase, which is activated by binding to the uncharged tRNA, phosphorylates eIF2α in response to amino acid deprivation, whereas PERK, which is a transmembrane protein of the ER, phosphorylates eIF2α in response to ER stress^21,23.

Ginseng is one of the most commonly used herbs with a variety of biological activities. Evaluation of randomized controlled trials revealed that Panax ginseng showed promising results for improving glucose metabolism, reducing inflammation, and preventing cancer recurrence²⁴. However, little is known of how ginseng may exert its various effects. It is generally accepted that the major pharmacological constituents of ginsengs are ginsenosides with the majority being the protopanaxadiol (PPD) group ginsenosides²⁵. After oral ingestion, the bioavailability of ginseng parent compounds is low due to the low level of absorption. Instead, the parent compounds are transformed by the intestinal microbiome to the more easily adsorbed metabolites such as compound K and PPD^26,27,28. We showed previously that PPD significantly affected expression of genes involved in lipid and steroid biosynthesis, and induced ER stress, pro-death mechanisms such as p53, and pro-survival mechanisms such as autophagy²⁹. Recent studies suggest that defective autophagy, which is regulated by AMPK and mTOR, is linked with metabolic diseases such as obesity, diabetes, and its complications³⁰. In this study, we characterized the mechanisms by which PPD induces autophagy.

PPD source and quality

PPD was synthesized and purified as described²⁶. The high-performance liquid chromatography-determined purity was %.

Cell culture, chemicals, and reagents

Human colorectal cancer cells HCT were obtained from the American Type Culture Collection. Cells were maintained in Dulbecco’s modified Eagle’s medium supplemented with 5% fetal bovine serum (Hyclone Laboratories), 50 IU of penicillin/streptomycin (Gemini Bio-Products), and 2 mmol/l of l-glutamine (Invitrogen) in a humidified atmosphere with 5% CO₂ at 37 °C. The p53^−/− HCT cells were described previously^31,32. PERK^+/+ and PERK^−/− mouse embryonic fibroblasts (MEFs) were described previously^33,34. NAC (N-acetylcysteine) was obtained from Sigma.

Western blot analysis

After being treated for the desired period of time, HCT cells were collected and washed twice with phosphate-buffered saline and lysed in RIPA buffer (20 mM Tris-HCl,  mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% NP, 1% sodium deoxycholate, 1 mM phenylmethlsulfonyl fluoride,  mM sodium pyrophosphate, 1 mM β-glycerophosphate, 1 mM sodium vanadate, 1 µg/ml leupeptin). Equal amounts of protein were loaded and the immunoblotting was detected by Li-Cor Odyssey image reader as described³². Anti-β-actin antibody was obtained from Santa Cruz Biotechnology. Antibodies against LC3-I/II, phospho-AMPKα (Thr), ACC, phospho-ACC (Ser79), and PERK were obtained from Cell Signaling Technology. Anti-Sesn2 antibody was obtained from Proteintech.

TFEB-GFP nuclear localization analysis

Wild-type (WT) or Sesn2 knockout cells were transfected with TFEB-GFP plasmid (a gift from Shawn Ferguson, Addgene plasmid #). Forty-eight hours after transfection, cells were treated with vehicle control or 30 µM PPD for 20 h and the localization of TFEB-GFP were analyzed by fluorescence microscopy. To determine the percentage of cells with nuclear TFEB-GFP, the localization of randomly selected 20 cells for each treatment groups were determined and repeated 3 times.

RNA isolation and RT-PCR

Total RNA was extracted with the RNeasy Mini Kit according to the manufacturer’s instructions (Qiagen). cDNA was synthesized using M-MLV reverse transcriptase from Promega. Quantitative real-time PCR was carried out as described³⁵. Primer pairs used for reverse-transcription PCR (RT-PCR) are: human Sesn2, 5′-CTCCTG GCGCCACTACATTG-3′ (sense) and 5′-ACTCAGGGTCACCACCAGT-3′ (antisense); human CHOP, 5′-TGGAAGCCTGGTATGAGGAC-3′ (sense) and 5′-TGTGACCTCTGCTGGTTCTG-3′ (antisense); human ATF3, 5′-GCAGAGCTAAGCAGT CGTGG -3′ (sense) and 5′-CGCCT TGATGGTTCTCTGCT-3′ (antisense); human GAPDH, 5′-CTCTGACTTCAACAGCGACAC-3′ (sense) and 5′-CATACCAGGAAATGAGCTTGACAA-3′ (antisense); mouse Sesn2, 5′-GCACCTTCGCCTCCCAGTGA-3′ (sense) and 5′-GCAGGCTCTCCTGAAGCTGC-3′ (antisense); mouse GAPDH, 5′-GCACAGTCAAGGCCGAGAAT-3′ (sense) and 5′-GCCTTCTCCATGGTGGTGAA-3′ (antisense).

Microarray data analysis

The data from six microarrays of control or PPD-treated samples were obtained and normalized as described previously³⁶. The normalized data were analyzed using ArrayTools Version , to identify genes significantly affected by 20–25 µM of PPD treatment. Replicates were averaged by geometric mean. Parametric p-values were computed using t-tests on the logarithms of gene expression intensities. Genes with significantly altered expression levels, grouped by Kegg biochemical pathway annotations and ordered by the smallest p-value within the pathway group, are shown in Supplementary Table S1. Significance Analysis of Microarrays was performed with a delta of and target false discovery rate of ³⁷. Genes were excluded if <20% of expression data exhibited a fold change from the gene’s median or >50% of data were missing. A minimum intensity threshold was set at Genes identified by the analysis are listed in Supplementary Table S2.

Lentiviral preparation and transduction

The pLKO.1 lentiviral sh-RNA expression system was used to generate short hairpin RNA (shRNA) constructs as described previously³⁸. The sequences of shRNA and SgRNA used in this study included the following: sh-GFP (5′-ACGTCTATATCATGGCCGACA-3′), sh-Sesn2 (5′-GGTCCACGTGAACTTGCT GC-3′). Lentiviral packaging was done according to the previously described protocol³⁹. Briefly, expression plasmids pCMV-dR and pCMV-VSV-G were co-transfected into T cells using the calcium phosphate method at  μg (for a 10 cm dish). The transfection medium containing calcium phosphate and plasmid mixture was replaced with fresh complete medium after incubation for 5 h. Media containing virus was collected 48 h after transfection and then concentrated using 20% sucrose buffer at 20, × g for 4 h. The virus pellet was re-dissolved in the proper amount of complete growth medium and stocked at −80 °C. Cells were infected with the viruses at the titer of % infection in the presence of Polybrene (10 μg/ml) for 48 h and were treated as desired.

The Lenti-Crispr V2 system was used to generate knockout clones⁴⁰: Sg-GCN2 (5′-GGAGAGCTACCCGCAACGAC-3′), Sg-PERK (5′-TGGAGCGCGCCATCAGCCCG-3′), Sg-ATF4 (5′-AGTGAAGTGGATATCACTGA-3′), Sg-Sesn2 (5′-GGACTACCTGCGGTTCGCCC-3′), Sg-AMPKα1 (5′-ATTCGGAGCCTTGATGTGGT-3′), and Sg-AMPKα2 (5′-ATTCGCAGTTTAGATGTTGT-3′). To generate HCT cells with single-gene knockout, cells were infected with SgRNA viruses at the titer of % infection in the presence of Polybrene (10 μg/ml) for 48 h and then single-cell clones was selected by serial dilution in well plate. To generate GCN2 and PERK double mutant clones (DKO GCN2&PERK), sgRNA against PERK was introduced into a clone of GCN2 single knockout cells and then single-cell clones was selected by serial dilution in well plate. The single-cell clones were confirmed by genomic DNA sequencing or western blotting.

PPD induces Sesn2 expression, which is required for PPD-induced autophagy

We examined PPD-induced gene expression alterations to gain insights into the mechanisms by which PPD may regulate autophagy. Analysis of microarray expression profiling data for significantly altered expression of individual genes nominated differences in the metabolism of various amino acids, purine and pyrimidine, propanoate and butanoate, Glycerophospholipid, in the biosynthesis of steroids and aminoacyl-tRNA, and in mitogen-activated protein kinase and Insulin signaling (Supplementary Table S1). Significance analysis of microarrays for PPD-altered genes identified Sesn2, an important regulator of metabolic homeostasis, in addition to several genes involved in amino acid transport and synthesis such as SLC7A11, SLC6A9, and Asparagine Synthetase (Supplementary Table S2). Sesn2 has been shown to regulate AMPK and mTOR, which are key regulators of autophagy. Indeed, Sesn2 expression was highly induced by PPD (Fig. 1a). As Sesn2 is known to regulate AMPK and autophagy, and we previously showed that PPD induced autophagy, we determined whether Sesn2 is required for PPD-induced autophagy by examining the effects of Sesn2 inactivation. PPD treatment significantly increased the level of LC3-II (Fig. 1c, d, sh-GFP and WT). TFEB is transcription factor that regulates expression of autophagy-regulated genes and its nuclear localization can be used to evaluate autophagic flux⁴¹. We used TFEB-GFP cytoplasmic and nuclear localization to further analyze autophagy induction by PPD⁴². TFEB-GFP was excluded from the nucleus in control-treated WT cells; PPD treatment induced TFEB-GFP to be present in the nucleus (Fig. 1f, g, WT), similar to treatment with serum-free medium that is known to induce autophagy (Supplementary Fig. S1). These results suggest that PPD induces autophagy in WT HCT cells. Interestingly, inactivation of Sesn2 either by shRNA knockdown (Fig. 1b) or by CRISPR/CAS9-mediated knockout (Fig. 1e) significantly inhibited PPD-induced LC3-II accumulation (Fig. 1c, d) and blocked PPD-induced presence of TFEB-GFP in the nucleus (Fig. 1f, g). These results suggested that PPD-induced Sesn2 is required for its induction of autophagy.

Sesn2 is required for PPD-induced AMPK activation

Sesn2 is known to regulate AMPK activity⁴³. The induction of Sesn2 by PPD suggest that PPD may induce AMPK activation, which can be monitored by phosphorylation on AMPKα T⁴⁴. Indeed, PPD treatment significantly increased p-AMPKα levels (Fig. 1h) and significantly increased phosphorylation of ACC on Ser79 (Fig. 1i), a key target of AMPK in fatty acid biosynthesis⁴⁵. These results show that PPD induces AMPK activation. Interestingly, PPD also slightly increased levels of AMPKα (Fig. 1h), which is consistent with the report that Sesn2 also modulates AMPK subunit expression⁴⁶. Importantly, knockdown of Sesn2 significantly inhibited PPD-induced phosphorylation of AMPKα on T (Fig. 1j). These results indicate that Sesn2 is also required for PPD-induced AMPK activation.

TORC1 and ULK1 kinase complexes are key regulators of autophagy. As AMPK has been shown to regulate both TORC1 and ULK1 kinase complexes, we tested whether AMPK activation is involved in PPD-induced autophagy. We used CRISPR/CAS9 approach to knock out both AMPKα1 and AMPKα2. Double knockout (DKO) AMPKα1/α2, which removed both AMPKα1 and AMPKα2 proteins (Fig. 1k), significantly decreased PPD-induced LC3-II accumulation (Fig. 1l). Taken together, these results suggest that Sesn2 is required for PPD-induced AMPK activation, which contributes to the autophagy induction.

PPD-induced Sesn2 expression requires ATF4

Sesn2 has been shown to be induced by different stress insults, including p53, oxidative stress, and ER stress. To determine the mechanism by which PPD induces Sesn2, we examined the effects of inactivating p53, addition of antioxidant NAC, or inactivation of ATF4, a key transcription factor in the PERK branch of ER stress response. Knockout of p53 or addition of NAC did not affect Sesn2 induction by PPD (Fig. 2a, b). These observations suggest that PPD-induced Sesn2 is not mediated by p53 or oxidative stress. In contrast, knockdown of ATF4 significantly decreased PPD-induced Sesn2 expression (Fig. 2d, e). In addition, PPD also induced expression of ATF3 and CHOP (Fig. 2c), two ATF4-regulated transcription factors regulating the expression of genes involved in the stress response program²¹. As expected, induction of ATF3 and CHOP by PPD was also significantly inhibited by ATF4 knockdown (Fig. 2d). In addition, we used CRISPR/CAS9 approach to knock out ATF4 and generated an ATF4 mutant line with a 2 bp deletion in the open reading frame (Fig. 2f). Knockout of ATF4 blocked PPD-induced expression of Sesn2, ATF3, and CHOP (Fig. 2g, h). Furthermore, inactivation of ATF4 but not addition of NAC inhibited PPD-induced LC3-II accumulation (Fig. 2i–k). These results suggest that ATF4 is required for PPD-induced Sesn2 expression, which in turn is required for PPD-induced autophagy.

Inactivation of PERK alone does not inhibit PPD-induced Sesn2 expression

As ATF4 is activated by PERK during ER stress and as PPD can induce ER stress and expression of PERK branch of ER stress targets CHOP and ATF3, we further examine the effect of inactivating PERK. Two independent PERK small deletion lines that caused frame-shift mutations were generated using CRISPR/CAS9 approach (Fig. 3a). Both lines eliminated PERK protein (Fig. 3b). However, knockout PERK did not decrease PPD-induced expression of Sesn2, CHOP, or ATF3 (Fig. 3c–e). Tunicamycin (Tu), a well-characterized ER stress-inducing agent, induces ER stress because of the inhibition of N-linked glycoslylation and the accumulation of unfolded glycoproteins in the ER. To make sure that the PERK-knockout cell lines did have the expected effects on ER stress, we characterize their effects on Tu-induced ER stress. Tu treatment in WT cells induced expression of Sesn2, CHOP, and ATF3. Knockout of PERK significantly inhibited Tu-induced expression of these genes (Fig. 3f, g). Therefore, the PERK knockout cells failed to inhibit PPD-induced Sesn2 expression, despite having the expected function of blocking Tu-induced expression of PERK branch of ER-stress target genes. Furthermore, PPD-induced similar levels of Sesn2 expression in matched PERK^+/+ and PERK^−/− MEFs, even though Tu-induced Sesn2 was reduced in the PERK^−/− cells (Fig. 3h, i). Taken together, these results suggest that PPD may induce ATF4 activation and Sesn2 expression through a pathway in parallel to PERK.

Both amino acid deprivation and ER stress induce Sesn2 expression through ATF4

ATF4 functions downstream of PERK to activate this branch of ER-stress-induced target gene expression. Indeed, ATF4 knockout blocked Tu-induced Sesn2 expression (Fig. 4a). As PPD-induced Sesn2 expression was inhibited by inactivation of ATF4 but not PERK, we tested the possibility that PPD may induce another pathway to activate ATF4. Amino acid deprivation is known to activate ATF4²¹. Indeed, deprivation of amino acid Gln induced Sesn2 expression, which was blocked by ATF4 knockout (Fig. 4b). Furthermore, LC3-II accumulation induced by Gln deprivation was also inhibited by inactivation of ATF4 or Sesn2 (Fig. 4c, d). These results show that ATF4 is required for amino acid deprivation and ER stress-induced Sesn2 expression as well.

GCN2 and PERK redundantly regulate PPD-induced Sesn2 expression

As GCN2 can upregulate ATF4 in response to amino acid deprivation, we generated GCN2-knockout cells with CRIPSR/Cas9 approach to characterize the effect of inactivating GCN2 on PPD-induced Sesn2 expression (Fig. 4e). Interestingly, inactivation of GCN2 inhibited Gln-deprivation-induced Sesn2, CHOP, or ATF3 expression (Fig. 4f) but did not inhibit PPD-induced Sesn2, CHOP, or ATF3 expression (Fig. 4g). These results suggest that inactivation of GCN2 alone is not sufficient to block for PPD-induced Sesn2 expression either.

To test the possibility that PPD activates both GCN2 and PERK redundantly to upregulate ATF4, we generated two independent GCN2 and PERK DKO cell lines from the GCN2 single knockout cells shown in Fig. 4e. As expected, both DKO cells reduced PERK protein (Fig. 5a). Importantly, both DKO cells significantly reduced PPD-induced expression of Sesn2, CHOP, and ATF3 (Fig. 5b–d). These results suggest that PPD can activate ATF4 and Sesn2 expression through either GCN2 or PERK, which functions redundantly downstream of PPD to upregulate ATF4 (Fig. 5e).

Our results showed that Sesn2 is strongly induced by ginseng metabolite PPD and is required for PPD to activate AMPK and induce autophagy (Fig. 5e). Expression of Sesn2 can potentially be induced by a variety of stress signals. Interestingly, although inactivating GCN2 or PERK alone can block amino acid deprivation or ER stress-induced Sesn2 expression, respectively, inactivating either GCN2 or PERK failed to block PPD-induced Sesn2 expression. In contrast, inactivation of both GCN2 and PERK together blocked PPD-induced Sesn2 expression. These results suggest that PPD treatment activate both GCN2 by causing amino acid deprivation and PERK by inducing ER stress (Fig. 5e).

As PPD significantly affected expression of genes involved in amino acid transport, metabolism, and synthesis of aa-tRNA (Table S1), it is likely to be that PPD treatment caused deficiency in the level of certain amino acid and certain aa-tRNA, which induced GCN2 activation. In addition, PPD also significantly inhibited expression of genes involved in fatty acid and cholesterol biosynthesis, and interfered with normal lipid metabolism, which may induce ER stress and PERK activation directly through causing lipid disequilibrium^47,48. Consistent with this, we found that PPD-induced ER stress was not significantly affected by inhibition of new protein synthesis²⁹.

It is increasingly clear that activation of AMPK, induction of autophagy, and inhibition of mTORC1 are associated with increased metabolic health. Metabolic syndrome is a disorder in the balance between energy supply, utilization, and storage, and is linked with obesity, insulin resistance, development of type 2 diabetes, and increased risks of developing cardiovascular disease and many types of cancers^49,50,51. Extensive evidence suggest that dysregulation of AMPK is associated with these metabolic disorders, and that modulation of AMPK activation, either through excise or by pharmacological AMPK activators, could be used to treat or prevent these diseases⁵². AMPK activation reprograms cellular metabolism, induces autophagy, and promotes mitochondria homeostasis⁵³. Therefore, the observed activation of AMPK by ginseng metabolite PPD raise the possibility that ginseng may have beneficial effects to promote metabolic health. This is consistent with finding from a systematic review of randomized controlled trials that showed ginseng exhibited promising results for improving glucose metabolism²⁴. In addition, studies of several ginsenosides in animal models have also showed that these compounds can promote metabolic health and modulate glucose and fat metabolism^54,55,56. Taken together, the finding that ginseng metabolite PPD induces Sesn2, AMPK activation, and autophagy provides mechanistic insights into the beneficial effects of ginseng to promote metabolic health.

1. 1.

   He, C. & Klionsky, D. J. Regulation mechanisms and signaling pathways of autophagy. Annu. Rev. Genet.43, 67–93 ().

   CASArticle Google Scholar

2. 2.

   Nakatogawa, H., Suzuki, K., Kamada, Y. & Ohsumi, Y. Dynamics and diversity in autophagy mechanisms: lessons from yeast. Nat. Rev. Mol. Cell Biol.10, – ().

   CASArticle Google Scholar

3. 3.

   Levine, B. & Kroemer, G. Autophagy in the pathogenesis of disease. Cell, 27–42 ().

   CASArticle Google Scholar

4. 4.

   Mizushima, N. & Levine, B. Autophagy in mammalian development and differentiation. Nat. Cell Biol.12, – ().

   CASArticle Google Scholar