GGA (geranylgeranoic acid) is a natural polyprenoic acid, derivatives of which has been shown to prevent second primary hepatoma. GGA induces mitochondria-mediated PCD (programmed cell death), which may be relevant to cancer prevention. To gain further insights into GGA-induced PCD, autophagy processes were examined in human hepatoma-derived HuH-7 cells. Treatment of HuH-7/GFP (green fluorescent protein)–LC3 cells with GGA induced green fluorescent puncta in the cytoplasm within 30 min and their massive accumulation at 24 h. After 15 min of GGA treatment, a burst of mitochondrial superoxide production occurred and LC3β-I was appreciably converted into LC3β-II. GGA-induced early stages of autophagy were unequivocally confirmed by electron-microscopic observation of early/initial autophagic vacuoles. On the other hand, LC3β-II as well as p62/SQSTM1 (sequestosome 1) continuously accumulated and co-localized in the cytoplasmic puncta after GGA treatment. Furthermore, GGA treatment of HuH-7/mRFP (monomeric red fluorescent protein)–GFP–LC3 cells showed yellow fluorescent puncta, whereas glucose deprivation of the cells gave red fluorescent puncta. These results strongly suggest that GGA induces the initial phase of autophagy, but blocks the maturation process of autolysosomes or late stages of autophagy, insomuch that GGA provides substantial accumulation of autophagosomes under serum-starvation conditions in human hepatoma cells.
GGA [geranylgeranoic acid, a 20-carbon polyprenoic acid (all-trans 3,7,11,15-tetramethyl-2,4,6,10,14-hexadecapentaenoic acid)] and its derivatives were historically developed as synthetic ‘acyclic retinoids’ by screening their binding affinity to cellular retinoic-acid-binding protein  as well as to nuclear retinoid receptors , which exert transcriptional activation of some hepatocyte-specific genes in hepatoma cells . 4,5-DidehydroGGA, one of the most potent acyclic retinoids, has been reported to prevent chemical-induced and spontaneous hepatocarcinogenesis in animals [4,5]. Furthermore, efficacy of 4,5-didehydroGGA on prevention of second primary hepatoma was demonstrated in a placebo-controlled double blind randomized phase II clinical trial conducted in post-operative hepatoma patients with few side effects . Subsequently, it was revealed that oral administration of polyprenoic acid for 12 months significantly increased 5-year survival in these patients after radical therapy of primary hepatoma .
Although GGA shares characteristics of natural cyclic retinoids in vitro, GGA apparently differs from natural retinoids, such as all-trans and 9-cis retinoic acids, in the following respects: (i) GGA up-regulates albumin mRNA expression in the human hepatoma-derived cell lines HuH-7 and PLC/PRF-5, whereas all-trans retinoic acid down-regulates their expression ; (ii) GGA is much less toxic than natural retinoids in a human hepatoma cell line in the presence of FBS (fetal bovine serum) , and 1-year intake of 4,5-didehydroGGA (600 mg/day) caused no apparent side effects in the abovementioned clinical trial [6,7]; and finally, and most importantly, (iii) GGA induces cell death in hepatoma cell lines in the absence of FBS, whereas neither all-trans nor 9-cis retinoic acid did [8,9]. Previously, we found that hepatoma-preventive GGA is naturally present in various medicinal herbs that are commonly used in the prevention of liver diseases .
During GGA-induced cell death, HuH-7 cells displayed dissipated inner-membrane potential of mitochondria . This mitochondria-involved cell death showed characteristics of apoptosis such as chromatin condensation as revealed by Hoechst-staining; however, caspase inhibitors were unable to block GGA-induced cell death . These results suggest that GGA-induced cell death is not a typical apoptotic process, but might be a caspase-independent and non-necrotic cell death.
Autophagic cell death is a third mode of cell death, or type II PCD (programmed cell death) , which kills cells under certain conditions generating a caspase-independent form of PCD. Nowadays, it is advocated that macroautophagy (autophagy) is active to prevent tumorigenesis . It is, however, well established that autophagy constitutes a cellular self-restructuring system involving degradation of intracellular components such as organelles, proteins and RNA. Therefore autophagy is recognized to be a cytological process for cell survival or cytoprotection, but not for cell death. Indeed, autophagy has been shown to be involved in embryogenesis, neonatal development and starvation .
Autophagy is classically elicited by nutrient deprivation, and much is known about starvation-induced autophagy. A series of ATGs (autophagy genes) have been identified . Autophagy is initiated by conjugation of ATG7, an E1-like enzyme, and ATG8/LC3 (microtubule-associated protein 1 light chain 3) and recruitment of PE (phosphatidylethanolamine)-labelled LC3 (LC3-II) to the nascent autophagosome membrane or phagophore. Upstream signals that regulate bulk-phase-starvation- or trophic-deprivation-induced autophagy include Vps34 (vacuolar protein sorting 34), a class III PI3K (phosphoinositide 3-kinase), complexed with Beclin1/ATG6 [15,16]. Reversible modification of LC3 with PE is crucial for maintenance of autophagy . In other words, recycling of LC3 from LC3-II to LC3-I by delipidation with ATG4 is part of the normal progression of autophagy and is important for fusion between autophagosome and lysosome .
As for passengers for the autophagosome vehicle, p62/SQSTM1 (sequestosome 1) is particularly interesting because this LC3- as well as ubiquitin-binding protein is accumulated in intracytoplasmic hyaline bodies frequently found in hepatocellular carcinoma cells . Recently, White's group  demonstrated the importance of p62/SQSTM1 in hepatocarcinogenesis, suggesting that p62/SQSTM1 up-regulation may contribute directly to tumorigenesis. Although several mechanical studies [2–4,8,9] have so far been carried out on GGA-induced cell death, little is known about the signalling pathway regulating autophagic responses during GGA-induced cell death in human hepatoma cells.
In the present study, we addressed the question of whether GGA induced autophagic responses in human hepatoma-derived HuH-7 cells. To answer this question, we observed GFP (green fluorescent protein)–LC3 puncta formation in HuH-7 cells stably expressing GFP–LC3 using a live-cell imaging technique to visualize the nascent autophagosome, and conducted Western blot analysis of LC3 to assess lipidation of LC3 as an early stage of autophagy. In addition, we investigated the late stage of GGA-induced autophagic responses in terms of fusion of autophagosome with lysosome and degradation of p62/SQSTM1 after GGA treatment.
The fluorescence probes MitoTracker® Red CMXRos, MitoSOX™ Red and Hoechst 33258 were obtained from Molecular Probes. Wortmannin, chloroquine diphosphate salt, pepstatin A and E-64d were obtained from Sigma–Aldrich. GGA was a gift from Kuraray (Okayama, Japan). Tiron (1,2-dihydroxy-3,5-benzene-disulfonic acid, disodium salt, monohydrate) was purchased from Dojin Laboratories.
Human hepatoma-derived HuH-7 cells were obtained from the RIKEN BioResource Center, Tsukuba, Japan and maintained in 25-cm2 flasks in DMEM (Dulbecco's modified Eagle's medium; Wako Pure Chemical Industries) containing 5% FBS (HyClone).
Plasmids and stable transfection
Using Lipofectamine™ 2000 transfection reagent (Invitrogen), stable cell clones [HuH-7/GFP–LC3 and HuH-7/mRFP (monomeric red fluorescent protein)–GFP–LC3] were established by means of transfection of HuH-7 cells with an EGFP (enhanced GFP)-tagged LC3 expression vector (pEGFP-LC3m, a gift from Dr T. Yoshimori, National Institute of Genetics, Mishima, Japan, and Dr N. Mizushima, Tokyo Medical and Dental University, Bunkyo Ku, Tokyo, Japan) or with an mRFP–EGFP tandem-tagged LC3 expression vector (ptfLC3; Addgene) respectively. The transfected cell clones were selected by resistance to G418 (50–150 μg/ml, Invitrogen) in six-well plates with a cylinder-cloning method or by visual recognition of specific live fluorescence under a fluorescence microscope.
Cells were lysed in RIPA buffer (Pierce Biotech) and proteins were quantified using the Bradford assay (Bio-Rad Laboratories). Equal amounts (5–15 μg each) of protein were separated by SDS/PAGE. PVDF membranes (Bio-Rad Laboratories) were probed with rabbit polyclonal antibodies against LC3 (MBL), β-actin (Santa Cruz Biotechnology), ATG4B (APG4B; Abgent), ATG7 (APG7L; Abgent), p62 (SQSTM1; Abgent) and Beclin1 (Cell Signaling Technology). HRP (horseradish peroxidase)-labelled secondary antibodies were detected with ECL (enhanced chemiluminescence) plus (GE Healthcare).
HuH7/GFP–LC3 or HuH-7/mRFP–GFP–LC3 cells were cultured on glass-bottomed dishes (Matsunami) at a density of 6×104 cells/dish in a chamber unit (INUG2-ZIL; Tokai Hit), equipped with a Carl Zeiss LSM 510 inverted laser-scanning confocal fluorescence microscope. Live-cell images were scanned every 10 min after the addition of GGA for 48 h under either single-track or multi-track settings.
Mitochondria inner-membrane potential (ΔΨm)
At selected times after the addition of GGA, cells were rinsed with HBSS (Hanks balanced salt solution; Sigma–Aldrich) and incubated for 30 min with 100 nM MitoTracker® Red CMXRos (Molecular Probes) solution in FBS-free DMEM. After rinsing three times with HBSS, the cells were fixed for 30 min with 4% paraformaldehyde containing 2% sucrose in PBS(−), and were then rinsed three times with HBSS. Finally, the cells were mounted in ParmaFluor (Immunotech), covered with a glass coverslip, and observed under a Carl Zeiss Axiovert 200M fluorescence microscope equipped with a laser-scanning confocal fluorescence microscope system LSM 510. The NIH Image program (version 1.62) was used to measure the cellular intensity of red fluorescence as the integrated density per cell by summing the pixels in each cell with the background subtracted.
After overnight incubation with 10 μM GGA in the absence of FBS, HuH-7/GFP–LC3 cells on the glass-bottomed dish were rinsed with PBS(−) and fixed for 30 min with 4% formaldehyde containing 2% sucrose in PBS(−) and then rinsed three times with HBSS. Then, cells were permeated with 0.5% Triton X-100 and blocked with 10% pre-immune rabbit serum. Next, the cells were incubated at 4 °C overnight with a polyclonal anti-p62/SQSTM1 antibody (Abgent) followed by a 2.5 h incubation with an Alexa Fluor 568-conjugated goat anti-rabbit IgG (Molecular Probes). After rinsing with HBSS, the cells were mounted in ParmaFluor (Immunotech), covered with a glass coverslip, and co-localization of autofluorescence of GFP–LC3 and red fluorescence derived from anti-p62 was observed by sequential scanning under a Carl Zeiss Axiovert 200M microscope equipped with LSM 510.
Transmission electron microscopy
HuH-7/GFP–LC3 cells were cultured on glass-bottomed dishes, and after GFP–LC3 puncta were observed by live-cell imaging at 60 min of GGA treatment, the cells were fixed with 4% paraformaldehyde in PBS and then post-fixed with 2% osmium tetraoxide, dehydrated in ethanol, and embedded in Epon-812 resin (TAAB). Ultrathin sections of the cultured cells were cut at 60 nm using an ultramicrotome (LKB 250) and cells were stained with 7% uranyl acetate and 0.4% lead citrate. The stained sections were examined under a JEOL microscope, model JEM-1200EX, at an acceleration voltage of 100 kV.
The condensed nuclei were stained with 500 nM Hoechst 33258 (Molecular Probes) and viewed under an inverted fluorescence microscope (IX70; Olympus). Cellular superoxide was detected by fluorescence microscopy of cells incubated for 10 min with 5 μM MitoSOX™ Red . Viable cell counting was performed using the Trypan Blue dye-exclusion method (Sigma–Aldrich).
Autophagic cell death in HuH-7 cells induced by GGA
Treatment of HuH-7 cells with 10 μM GGA in the absence of FBS caused a time-dependent condensation of chromatin (Figure 1A and Supplementary Figure S1 at http://www.BiochemJ.org/bj/440/bj4400063add.htm). Ethanol-treated control cells showed no significant condensation in 24 h (Supplementary Figure S1). The cells died within 24 h after GGA treatment as judged by the Trypan Blue dye-exclusion method. This cell death was dose-dependent and exhibited an LD50 of approximately 8 μM GGA (Figure 1B).
GGA induces chromatin condensation and cell death in HuH-7 cells
Although depletion of FBS for 24 h did not apparently affect the intracellular uniform distribution of GFP–LC3 vague fluorescence in ethanol-treated HuH-7/GFP–LC3 cells (Figures 2A and 2B), Figures 2(C) and 2(D) clearly show that, 24 h after addition of 10 μM GGA in the absence of FBS, cells presented typical and discrete GFP–LC3 puncta formation in their cytoplasmic space. The GGA-induced increase in GFP–LC3 puncta formation was already detectable 30 min after GGA addition by time-lapsed live-cell imaging (Figure 3), and afterwards the puncta increased in size until 24 h. Even after 48 h, several large-size puncta still remained in dead cells.
Accumulation of autophagic vacuoles induced by GGA treatment in HuH-7/GFP–LC3 cells
Time-lapsed observation of GFP-labelled autophagosomes in live-cell images after GGA treatment
Induction of an early stage of autophagy by GGA
As shown in Figure 4(A), 5% FBS in medium prevented the GGA-induced puncta formation of GFP–LC3. Wortmannin, an inhibitor of autophagy initiation, also efficiently blocked the GGA-induced increase in GFP–LC3 puncta formation in HuH-7/GFP–LC3 cells.
Early/initial autophagic vacuoles induced by GGA treatment in HuH-7 cells
Next, we examined endogenous LC3β expression by immunoblotting to study the molecular conversion of LC3β-I into LC3β-II in HuH-7 cells after GGA treatment. Either deprivation of FBS from the medium or amino-acid-depleted medium in the absence of FBS did not cause a significant increase in LC3β-II in 1 h without GGA (Figure 4B, lanes 1–3), whereas treatment with 10 μM GGA significantly enhanced PE-conjugation of LC3β either in amino-acid-rich or amino-acid-free medium without FBS (Figure 4B, lanes 4 and 5). Wortmannin prevented GGA-induced accumulation of LC3β-II in a dose-dependent manner (Figure 4C), consistent with the live-cell imaging data of HuH-7/GFP–LC3 cells. The GGA-increased formation of autophagosomes revealed by cytological and biochemical findings was further confirmed by transmission electron microscopy. Typical early/initial autophagic vacuoles  frequently appeared in GGA-treated cells (Figure 4D, white arrows in the upper panel) in which puncta formation of GFP–LC3 had been detected at 1 h after GGA addition prior to paraformaldehyde fixation, whereas early and late autophagosomes and mature autolysosomes (black arrow in the lower panel) were spontaneously found in control cells at 1 h after deprivation of FBS.
Dysfunction of mitochondria induced by GGA
We previously reported that GGA-triggered generation of ROS (reactive oxygen species) was essential in GGA-induced cell death . Hence we measured intracellular superoxide production to assess whether GGA-triggered generation of superoxide was indispensable for GGA-induced accumulation of LC3β-II.
It is shown in Figures 5(A) and 5(B) that GGA induced hypergeneration of superoxide in mitochondria at 15 min, when accumulation of GFP–LC3 puncta was not yet observed in live-cell imaging. The autophagy inhibitor wortmannin also blocked GGA-triggered generation of superoxide, implying that GGA-triggered generation of superoxide might be an upstream signal to induce the formation of GFP–LC3 puncta and conversion of LC3β-I into LC3β-II.
GGA induces a burst of superoxide in mitochondria
Simultaneous treatment with tiron (0.1 mM in medium), a superoxide scavenger, slightly delayed GGA-induced conversion of LC3β-I into LC3β-II, although treatment with tiron alone somewhat up-regulated LC3β-II (Figure 5C). Figure 5(C) also shows that wortmannin delayed GGA-induced conversion of LC3β-I into LC3β-II, which is consistent with the preventive effect of the drug against the GGA-induced mitochondrial burst of superoxide (Figures 5A and 5B).
A 2 h treatment with GGA diminished the intensity of ΔΨm-dependent fluorescence of MitoTracker® Red in a concentration-dependent manner (Figure 6A), whereas ΔΨm-intact mitochondria remained in the perinuclear space after GGA treatment, which was supported by electron microscopic findings (Supplementary Figure S2 at http://www.BiochemJ.org/bj/440/bj4400063add.htm). GGA-induced impairment of mitochondrial function occurred after 2 h of GGA treatment (Figure 6B).
GGA induces dissipation of ΔΨm in HuH-7 cells
Fate of autophagosomes after GGA treatment
We next performed a time-course study of the effect of GGA on the expression of several autophagy-related genes. As shown in Figure 7(A), the amount of LC3β-II started to increase at 15 min and gradually became more abundant up to 1 h after the addition of GGA. Its cellular content was maintained at higher levels until 24 h. In the same experiment, GGA also induced increased cellular levels of Beclin1 and ATG4B at 30 min, and then after 1 h gradually decreased the cellular levels of ATG4B below the original levels at 24 h, whereas the cellular levels of ATG7 remained constant during the experiment (Supplementary Figure S3 at http://www.BiochemJ.org/bj/440/bj4400063add.htm). On the other hand, cellular levels of p62/SQSTM1 steadily increased from 30 min to 24 h after the addition of GGA (Figure 7A). GGA-induced accumulation of p62/SQSTM1 was confirmed by immunofluorescence, which showed its co-localization with GFP–LC3 (Figure 7B), indicating that p62/SQSTM1 was in autophagosomes and remained intact after treatment with GGA for 24 h.
Massive accumulation of LC3β-II and p62/SQSTM1 in GGA-induced autophagy
Impairment in the late stages of autophagy
In mammalian cells, LC3β-II located on the outer surface of autophagosomes is removed prior to fusion with lysosomes . It is believed that in mammalian cells this might be used to prevent premature fusion with lysosomes, whereas LC3β-II located in the inner membrane is trapped inside autophagosomes and remains detectable until degraded by hydrolases . In this context, it is reasonable to speculate that a sizeable accumulation of LC3β-II after treatment with GGA for 8 h may imply an incomplete or impeded autophagy process. Hence we utilized the inhibitors of lysosomal proteases pepstatin A and E-64d to block lysosomal turnover of LC3β-II . Although addition of the lysosomal protease inhibitors did not cause any significant accumulation of LC3β-II in the presence of FBS, the inhibitors increased the cellular level of LC3β-II in the absence of FBS to a comparable level after GGA treatment (Figure 8), suggesting that serum starvation alone might cause autophagy in HuH-7 cells and the inhibitors might impair the processing of LC3β-II. In the presence of GGA, pepstatin A plus E-64d did cause some additional accumulation of LC3β-II (Figure 8B).
Accumulation of LC3β-II is induced by lysosomal protease inhibitors as well as GGA
Finally, to dissect the autolysosome maturation process by live-cell imaging with a tandem fluorescent-tagged LC3 probe, we established a stable clone of HuH-7/mRFP–GFP–LC3, which expressed mRFP–GFP tfLC3 (tandem fluorescent-tagged LC3). This recombinant protein shows a GFP and mRFP signal before the fusion with lysosomes, and exhibits only the mRFP signal after fusion with lysosomes, because the GFP fluorescence easily fades away in acidic conditions, such as in lysosomes . As shown in Figure 9, the cells showed red fluorescent puncta and dissipation of green fluorscence after 24 h of glucose deprivation, because the green fluorescence of GFP faded away under the acidic conditions in lysosomal environment, whereas control cells displayed red and yellow fluorescence signals scattered as smaller puncta. However, in the presence of chloroquine, a basic lysosomotropic agent, 24 h of glucose deprivation induced the accumulation of yellow fluorescent puncta (Figure 9), indicating that chloroquine-impaired fusion of autophagosomes with lysosomes, or converted lysosomal environment from acidic into basic. Indeed, LC3β-II was accumulated in chloroquine-treated cells in glucose-deprivation-induced autophagy (Supplementary Figure S4 at http://www.BiochemJ.org/bj/440/bj4400063add.htm). As was the case with glucose deprivation plus chloroquine, GGA treatment increased yellow fluorescent puncta. Taken together with the data that a lysosomal protease inhibitor induced yellow puncta in HuH-7/mRFP–GFP–LC3 cells (Supplementary Figure S5 at http://www.BiochemJ.org/bj/440/bj4400063add.htm), the experiment with the tfLC3 probe strongly suggests that GGA-induced autophagy was incomplete.
Dissection of the autophagosome maturation process by tfLC3 by live-cell imaging
We have clearly demonstrated in the present study that autophagic events became detectable in human hepatoma-derived HuH-7 cells immediately after treatment with GGA, a cancer-preventive diterpenoid that has been repeatedly shown to induce mitochondria-mediated caspase-independent cell death in these cells [8,9]. GGA-induced autophagic events included immediate appearance of wortmannin-sensitive puncta formation of GFP–LC3, a time-dependent increase in its number and size in the cytoplasmic space, accumulation of LC3β-II molecular species, and frequent electron microscopic detection of early/initial autophagic vacuoles, all of which are phenomena linked to the induction phase of autophagy and nascent autophagosome . Prior to discussion about a possible molecular link between GGA-induced autophagic events and GGA-induced cell death, it may be important to illustrate what happened to autophagy in GGA-treated HuH-7 cells.
The earliest event induced by GGA treatment found in the present study was a burst of mitochondrial superoxide (Figure 5), which is consistent with previous findings that dihydroethidine was immediately oxidized 15 min after addition of GGA in two guinea-pig cell lines . The second event, an increment in cellular LC3β-II levels, occurred at 15–30 min (Figure 7), and up-regulation of the cellular Beclin1 level was observed at 30 min (Supplementary Figure S3). Then, puncta formation of GFP–LC3 was detected at 30–60 min on live-cell imaging (Figure 3). GGA-induced dissipation of ΔΨm became evident after 2 h of GGA treatment (Figure 6), which is also consistent with the previous findings in HuH-7  and 104C1  cells. After 8 h of GGA treatment, the cells showed condensed chromatin (Figure 1 and Supplementary Figure S1), one of the cytological signs accompanying both type I (apoptotic) and type II (autophagic) cell death .
Next we wanted to see whether GGA enhanced the early phase of autophagy processing. Beclin1 is a central regulatory node engaged to initiate autophagic responses to diverse stimuli . GGA-induced up-regulation of Beclin1 (Supplementary Figure S3) provides firm evidence of an initial autophagic response. Furthermore, the initial burst of mitochondrial superoxide seemed essential for accumulation of GFP–LC3 puncta and LC3β-II because wortmannin inhibited accumulation of GFP–LC3 puncta and LC3β-II, as well as hyperproduction of superoxide. Although we are at present unaware how wortmannin blocked the mitochondrial burst of superoxide induced by GGA, Cohen et al.  have reported that wortmannin prevented mitochondrial ROS production induced by a δ-opioid agonist. Wortmannin is known to block new LC3 puncta generation, whereas pre-existing ones undergo lysosomal degradation . Therefore the finding that wortmannin prevented GGA-induced accumulation of GFP–LC3 puncta (Figure 4) indicates that the site of action of GGA may reside upstream of the wortmannin action site at least during treatment for 1 h. As for the importance of ROS production for induction of autophagy, Chen et al.  reported similar findings that 1.0 mM tiron prevented 2-mercaptoethanol-induced conversion of LC3β-I into LC3β-II in HEK (human embryonic kidney)-293 cells. Taking into account the enhanced macroautophagy in superoxide dismutase 1 mutant mice , it is reasonable to speculate that GGA-induced hyperproduction of superoxide may be essentially involved in triggering of autophagy and autophagosome processing. Indeed, the present study also demonstrated that 0.1 mM tiron, a specific scavenger of superoxide, to some extent delayed the GGA-induced conversion of LC3β-I into LC3β-II (Figure 5). In this context, we may be able to state that GGA enhanced the early phase of autophagy or the GGA-induced burst of mitochondrial superoxide triggered early events of autophagy . This is partially supported by the autophagic flux assay , that GGA caused some additional accumulation of LC3β-II in the presence of the lysosomal protease inhibitors pepstatin A and E-64d, and the lysosomal protease inhibitors induced the further accumulation of LC3β-II in GGA-treated cells (Figure 8). In those events, deprivation of FBS may also trigger early events in autophagy in FBS-independent HuH-7 cells because no accumulation of LC3β-II was detected by treatment with pepstatin A plus E-64d in the presence of 5% FBS (Figure 7). Several reports have supported that serum deprivation induced autophagy, by revealing inhibiting components in serum such as steroid hormone  and insulin .
We were also interested in observing any effects of GGA on the late phase of autophagy or the maturation process of autolysosomes. As mentioned above, it is reasonable to speculate that sizeable accumulation of LC3β-II and GFP–LC3 puncta after 8 h of GGA treatment may imply incomplete or impeded autophagy processing. In this regard, it is of note that p62/SQSTM1 started to increase its cellular content at 30 min and accumulated at 24 h in a punctate form, although other autophagy-related proteins, such as Beclin1 and ATG4B, gradually decreased after a transient increase at 1 h after GGA treatment (Supplementary Figure S3). These findings strongly indicate a reduction in autophagy performance levels after 1 h of GGA treatment. Previous evidence suggests that accumulation of p62 represents a convenient in vivo marker for impaired autophagy . Furthermore, structural analysis revealed that the intracellular level of p62 is tightly regulated by autophagy through a direct interaction of LC3 with p62 . Therefore in most of these cases an accumulation of p62 is accompanied by a decrease in LC3β-II protein levels. However, simultaneous accumulation of LC3β-II and p62 was inconsistently induced in HuH-7 cells by GGA treatment (Figure 7). In other words, even though LC3β-II protein was accumulated by GGA treatment, simultaneous accumulation and co-localization of p62 strongly imply a defect in its autophagic degradation. Indeed, a combination of pepstatin A and E-64d, membrane-permeant inhibitors against lysosomal proteases, increased cellular levels of LC3β-II at 1 h to the same extent as GGA (Figure 8), implying that GGA may also play an inhibiting role in autophagy progression because the inhibitors prevented maturation process of autophagy (Supplementary Figure S5). To dissect the autolysosome maturation process, a novel reporter protein, tfLC3, is useful ; if there is a significant population of red-only signal, the autophagosomes are normally matured to autolysosomes (where GFP will be relatively unstable). If most puncta exhibit both red and green signals (i.e. yellow signal in the merged image), autophagy is impaired prior to fusion with lysosomes. By using this novel probe we were able to demonstrate GGA-induced incomplete autophagy with distinct glucose-deprivationinduced complete autophagy (Figure 9). In terms of incomplete autophagy, there is also another possibility that excess amounts of ROS may interfere in the fusion process of autophagosomes with lysosomes. As mentioned above, reversible modification of LC3β with PE is crucial for maintenance of autophagy . Recycling of LC3β from LC3β-II to LC3β-I by delipidation with ATG4 has been described in the normal progression of autophagy . Previously, ATG4, a cysteine endoproteinase, has been reported to be highly sensitive to oxidative stress . Once a cysteine residue located near the ATG4 catalytic site is oxidized, complete formation of autophagosomes is prevented . Therefore GGA-induced hyperproduction of superoxide may inhibit ATG4 activity for recycling of LC3β. Indeed, cellular levels of ATG4B protein started to decrease at 1 h after GGA treatment (Supplementary Figure S3), which may contribute to GGA-induced incomplete autophagy.
GGA is a natural diterpenoid found in several medicinal herbs . Although a variety of food components including vitamins, curcumin, resveratrol and genistein have been shown to stimulate autophagy vacuolization , there has been scant evidence that certain nutrients prevent cells from autophagy, except for an old and interesting finding that ascorbic acid (vitamin C) inhibited 3-methyladenine-sensitive lysosomal autophagic degradation of ferritin . In this case, vitamin C protects cells from cellular toxicity of iron overload by retarding ferritin degradation via its autophagic incorporation into lysosomes.
Finally, we are ready to discuss a possible link between GGA-induced autophagic events and GGA-induced cell death. It is reasonable to speculate that GGA-induced impairment of the progression and maturation of autophagy in HuH-7 cells in the absence of FBS may result in energy depletion and/or excessive accumulation of damaged mitochondria, which finally may result in caspase-independent cell death. Indeed, a recent report by Rossi et al.  described that desmethylclomipramine, an active metabolite of clomipramine that is a therapeutic drug for psychiatric disorders, induced obstruction of autophagic flux, including accumulation of p62/SQSTM1, and consequently increased the cytotoxic effect of chemotherapeutic agents. Furthermore, p62 accumulation resulted in hyperactivation of NRF2 (nuclear factor erythroid 2-related factor 2), a stress-responsive transcription factor, indicating unexpected roles of p62-selective autophagy in controlling transcription of cellular defence enzyme genes . Although genetic defects or dysfunction of autophagy have been implicated in tumorigenesis of several human cancers , induction of incomplete autophagy by GGA, which logically requires autophagy-related genes, may provide a promising strategy in the prevention or treatment of cancers, as well as other lifestyle-related diseases . Therefore a knockdown-type experiment is definitely required in order to unequivocally demonstrate that GGA-induced incomplete autophagy is involved in cell death.
In conclusion, GGA, a cancer-preventive agent, induced wortmannin-sensitive autophagosome accumulation in human hepatoma-derived HuH-7 cells and promoted impairment of autophagy progression, which might be linked to caspase-independent cell death.
Dulbecco's modified Eagle's medium
fetal bovine serum
green fluorescent protein
Hanks balanced salt solution
microtubule-associated protein 1 light chain 3
monomeric red fluorescent protein
programmed cell death
reactive oxygen species
tandem fluorescent-tagged LC3
Kyoko Okamoto and Yoshihiro Shidoji conceived the study and designed the experiments. Kyoko Okamoto performed most of the experiments, Yoko Sakimoto did the HuH-7/mRFP–GFP–LC3 experiment, and Katsuyuki Imai and Haruki Senoo did the microscopic analysis. Kyoko Okamoto and Yoshihiro Shidoji interpreted the data and wrote the paper.
We thank Dr T. Yoshimori and Dr N. Mizushima for providing the pEGFP-LC3m plasmid and Mr T. Yonekura for his help with the live-cell imaging. We also thank Dr A.M. Jetten (National Institute of Environmental Health Sciences, National Institutes of Health, Triangle Park, NC, U.S.A.) for critical reading and helpful comments prior to the submission of the paper.
This work was supported, in part, by a grant-in-aid from the Japan Society for the Promotion of Science [grant number 19590230] and a research grant B from the University of Nagasaki.