Vicenistatin induces early endosome-derived vacuole formation in mammalian cells

Yuko Nishiyama, Tomohiro Ohmichi, Sayaka Kazami, Hiroki Iwasaki, Kousuke Mano, Yoko Nagumo, Fumitaka Kudo, Sosaku Ichikawa, Yoshiharu Iwabuchi, Naoki Kanoh, Tadashi Eguchi, Hiroyuki Osada & Takeo Usui

Homotypic fusion of early endosomes is important for efficient protein trafficking and sorting. The key controller of this process is Rab5 which regulates several effectors and PtdInsPs levels, but whose mechanisms are largely unknown. Here, we report that vicenistatin, a natural product, enhanced homo- typic fusion of early endosomes and induced the for- mation of large vacuole-like structures in mammalian cells. Unlike YM201636, another early endosome vacuolating compound, vicenistatin did not inhibit PIKfyve activity in vitro but activated Rab5-PAS pathway in cells. Furthermore, vicenista- tin increased the membrane surface fluidity of cholesterol-containing liposomes in vitro, and choles- terol deprivation from the plasma membrane stimu- lated vicenistatin-induced vacuolation in cells. These results suggest that vicenistatin is a novel compound that induces the formation of vacuole-like structures by activating Rab5-PAS pathway and increasing membrane fluidity.The endosome system functions as an intracellular trafficking and sorting network that controls many cel- lular processes, including receptor signaling, recycling, and degradation of receptors. This system requires a series of vesicle fusion events to deliver internalized solute and membrane to the endosomal sorting com- partment and beyond. During the trafficking, the size of endosome vesicles is tightly regulated in the cells.

It has been proposed that small G-protein Rab5 controls early endosome docking and fusion1,2) by regulating several effectors, including EEA13,4) and PtdIns kinase/ phosphatases.5,6) The activity of Rab5 is regulated by the GTP/GDP status, and the function of the Rab5GTPase activity is probably to maintain a dynamic equilibrium between Rab5-GDP and Rab5-GTP.7) Phos- phoinositides (PtdInsPs) also play a key role in mem- brane fusion and fission.8) Rab5 interacts with PtdIns3Ks which generate PtdIns(3)P on the early endosome and the recruitment of a set of PtdIns(3)P- binding Rab5 effectors such as EEA13,9) and another PtdInsPs-modifying enzyme complex, PAS (PIKfyve- ArPIKfyve-Sac3).10) PIKfyve is a PtdIns(3)P 5-kinase and converts PtdIns(3)P to PtdIns(3,5)P2 on the early endosome membrane. Several lines of experimental evi- dence suggest that PtdIns(3,5)P2 coordinates fission and fusion events of endosome and that perturbations in PtdIns(3,5)P2 production impair several intracellular trafficking pathways11–13) and increase endosome fusion.14) These lines of evidence strongly suggest that the size of the early endosome is regulated by Rab5 activity and the balance between PtdIns(3)P and PtdIns (3,5)P2.

However, because many factors are involved in vesicle fusion and fission, it is also important to inves- tigate these other factors in the regulatory network.To investigate the functions of specific molecules orpathways, small molecules sometimes become a useful tool. Several bacterial toxins and compounds have been reported to induce the vacuole-like structure. Helicobacter pylori VacA cytotoxin,15) Vibrio cholerae hemolysin,16) and Clostridium perfringens epsilon- toxin17) induce vacuoles that are assumed to be late endosomes/lysosomes in mammalian cells. Because vacuolation is induced in a vacuolar-type ATPase (V-ATPases) and extracellular chloride ion-dependent manner, it is thought that these proteinous toxins form anion channels by self-oligomerization and stimulate the proton-pumping activity of the co-localized electrogenic V-ATPase.18–20) Two small molecules, cocaine and vacuolin-1, are also known to form late endosome- /lysosome-derived vacuoles, but the vacuolating mechanisms in these cases are largelyunknown.21,22) A PIKfyve inhibitor, YM201636, induces early endosome-derived vacuoles by disturbance of the balance of PtdIns(3)P and PtdIns(3,5)P2.23) Recently, it has revealed that heronamide C binds sphingolipids and induces early endosome-derived vacuoles.24,25)

These toxins and chemicals have been used for the analyses of membrane trafficking, membrane resealing, and func- tions of endosomes/lysosomes/exosomes.Here, we report a novel vacuolation inducer, vicenis- tatin (Fig. 1(A)), which is a 20-membered macrolactam antitumor compound isolated from the culture broth of Streptomyces halstedii HC34.26,27) This compound induces rapid formation of large, swollen structures derived from early endosomes by homotypic fusion combined with uncontrolled fusion of the inner and limiting membranes of these organelles. Vacuolation induced by vicenistatin was blocked by pretreatment with a V-ATPase inhibitor and by deprivation of chlo- ride ion from medium, suggesting that an influx of pro- tons and counter ions is required for vicenistatin- induced vacuolation. Unlike YM201636, vicenistatin exhibited no inhibitory effect on the PIKfyve activity in vitro. Instead, vicenistatin activated the Rab5-PAS pathway and increased the fluidity of the membrane surface. These results strongly suggest that vicenistatin is a novel vacuolation inducer.

Materials and methods
Cell, cultures, and reagents. 3Y1 cells (rat normal fibroblast), HeLa cells (human cervix epidermoid carci- noma), and HEK293T cells (human embryonic kidney cell line-derived cells) were cultured in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal calf serum (Nichirei Biosciences Inc., Tokyo, Japan) in a humidified atmosphere containing 5% CO2. Vicenis- tatin was purified from the culture broth of Strepto- myces halstedii HC34, as previously reported.27) Dioctanoyl PtdIns(3,5)P2 sodium salt (diC8-PI(3,5)P2) and phosphatidic acid sodium salt (diC8-PA) were syn- thesized based on the reported procedure28) with some modifications. The PIKfyve inhibitor YM201636 was purchased from Calbiochem-Merck (Darmstadt, Germany). The organelle-specific fluorescent reagents, acridine orange, human transferrin-Alexa568, ER Tracker, and Mito Tracker, were purchased from Invit- rogen Inc. (Carlsbad, CA) Anti-LBPA antibody was a gift from Professor T. Kobayashi (ASI RIKEN). The monoclonal antibodies against Golgi 58K pro- tein/formiminotransferase cyclodeaminase (FTCD) were products of Sigma (St. Louis, MO). Horseradish perox- idase-conjugated anti-mouse and anti-rabbit IgGs were purchased from Kirkegaard & Perry Laboratories Inc. (Gaithersburg, MD). Alexa488- and Alexa568-conjugated antibodies were purchased from Invitrogen. Egg phos- phatidylcholine (PC), egg phosphatidylethanolamine (PE), cholesterol, and the fluorescent probes, 1,6-diphenylhexa-1,3,5-triene (DPH) and 1-(4-trimethy- lammoniumphenyl)-6-phenyl-hexa-1,3,5-triene (TMA- DPH), were purchased from Avanti Polar Lipids (Alabaster, AL).Transfection and microscopic analyses.

Transfec- tion of wild-type and Rab5a mutants (Rab5aS34N) intoFig. 1. Vicenistatin induced the formation of early endosome-derived vacuole-like structures. (A) Structure of vicenistatin. (B) 3Y1 cells were treated with 300 nM vicenistatin for 2 h. Scale bar = 30 μm. (C) Time-lapse analysis of the formation of vacuole-like structures in 3Y1 cells. 3Y1 cells were treated with 300 nM vicenistatin and observed at the indicated time points. The arrows indicate the site where two vacuole-like struc- tures were fused. The arrowheads indicate the site where the large vacuoles were broken down to small vacuoles.HeLa cells, and of FLAG-tagged PIKfyve into HEK293T cells, was performed using PerFectin (Genlantis, San Diego, CA) as recommended by the manufacturer. Differential interference contrast (DIC) and fluorescence images were photographed with Leica AF6000 systems (Leica Microsystems, Wetzlar, Germany) using a 40 × (dry, numerical aperture, 1.3) or a 63 × (oil, numerical aperture, 1.3) objective lens. Images were captured with a cooled, charge-coupled device (CCD) camera, Leica DFC350FX. In time-lapse analysis, vicenistatin-treated cells were observed with a DeltaVision system (Applied Precision, Issaquah, WA) including an Olympus IX70 inverted microscope (Olym- pus, Tokyo, Japan), with a 60 × (numerical aperture, 1.4) objective lens and an Olympus MI-IBC stage-heating device (Olympus). Images were captured with a cooled CCD camera from Princeton Instruments (Trenton, NJ). About 180-360 optical images were captured at 5-s inter- vals, and out-of-focus light was removed by iterative deconvolution on a Silicon Graphics (Mountain View, CA) Iris workstation. For quantification of vacuolated cells, we judged the cell with a few large vacuoles or several small vacuoles as vacuolated cell.

In vitro PIKfyve kinase assay. FLAG-PIKfyve activity was measured using immunoprecipitated PIK- fyve derived from transfected HEK293T cells. Thirty- six h after transfection, HEK293T cells were lysed and sonicated in ice-cold RIPA buffer (50 mM Tris–HCl, pH 8.0, 150 mM NaCl, Nonidet P-40, 0.5% sodium deoxycholate, 1 mM phenylmethylsulfonyl fluoride, 5 mM benzamidine, protease inhibitor cocktail, and phosphatase inhibitor cocktail). Cell lysates were pre- cleared by centrifugation (20,000 × g, 15 min, 4 °C) and FLAG-PIKfyve was immunoprecipitated with anti- M2 antibody (Sigma). Immunoprecipitations were car- ried out for 16 h at 4 °C with protein A/G-Agarose (Santa Cruz Biotechnology Inc.) added in the final 1 h of incubation. Immunoprecipitates were washed five times with RIPA buffer and once with kinase assay buffer (25 mM HEPES, pH 7.4, 2.5 mM MgCl2,2.5 mM MnCl2, 5 mM β-glycerophosphate, and 1%phosphatase inhibitor cocktail). Immunoprecipitated proteins were separated by SDS-PAGE, transferred to a polyvinylidene fluoride microporous membrane (Millipore, Billerica, MA), blocked with 5% (w/v) non- fat milk, probed with anti-PIKfyve antibody and horse- radish peroxidase-conjugated anti-rabbit IgG antibody, and detected by enhanced chemi-luminescence using Chemi-Lumi One (Nacalai Tesque). For the kinase assay, 50 μM ATP, [γ-32P]ATP (462.5 kBq), and 100 μM PtdIns (from soybean; Nacalai Tesque Inc.) were added to washed PIKfyve immunoprecipitates in the presence of each compound and incubated for 15 min at 37 °C. The lipids were extracted and ana- lyzed by silica gel thin-layer chromatography (TLC) using an acidic solvent system (Chloroform:Methanol: Acetone:Glacial acetic acid:water = 7:5:2:2:2 (v/v)).

Chloride ion-depletion and synthetic diC8-PI(3,5)P2-pre- treatment experiments. In chloride ion-depletion experi- ments,29) cells were treated with phosphate-bufferedsaline (PBS)/BSA pH 7.4 (135 mM NaCl, 6.5 mM Na2HPO4, 3.5 mM KCl, 1.5 mM KH2PO4, 5 mM glu- cose, 0.5 mM CaCl2, 0.5 mM MgCl2, 0.02% BSA, pH7.4) or with PBS/BSA/gluconate (135 mM Na-glu- conate, 6.5 mM Na2HPO4, 3.5 mM K-gluconate,1.5 mM KH2PO4, 5 mM glucose, 0.5 mM Ca-(glu-conate)2, 0.5 mM Mg-(gluconate)2, 0.02% BSA, pH 7.4) at 37 °C for 2 h. In the synthetic diC8-PI(3,5)P2-pretreat- ment experiments, 3Y1 cells were treated with 1 mM diC8-PA/PBS or diC8-PI(3,5)P2/PBS for 10 min and fol- lowed with 1 μM vicenistatin at 37 °C for 2 h.Measurement of physical properties of liposomes. Unilamellar liposomes consisted of PC, PE, and choles- terol were prepared by the extrusion technique. In brief, 10 mL of chloroform solution containing PC, PE, and cholesterol (7:3:1 M ratio) was added to a 100-mL round-bottom flask, followed by the removal of chloro- form under reduced pressure with a rotary evaporator (Type N-1; Eyela, Tokyo, Japan). The lipid film was dried with a vacuum freeze-drier (FDU-1200; Eyela) overnight. The lipid was dispersed in aqueous solution by adding 10 mL PBS, followed by vortexing. After 10 rounds of freezing in liquid nitrogen and thawing in water, the vesicle suspension was extruded through a doubly stacked polycarbonate membrane with a mean pore diameter of 100 nm using the LIPEX Extruder (Northern Lipid Inc., Vancouver, Canada).The hydrodynamic diameter and zeta potential ofliposomes containing vicenistatin were measured with a Zetasizer Nano ZS (Malvern Instruments, Ltd., Worces- tershire, UK). The average of three measurements of the sample was determined. To measure membrane flu- idity, liposomes containing or not containing vicenista- tin were mixed with the fluorescent probe DPH or TMA-DPH for 1 h, and then fluorescent anisotropy was measured using a spectrofluorometer with polariz- ing plates (FP-6500; Jasco, Tokyo, Japan) at 37 °C.

Results
It has been shown that vicenistatin exhibits an antitu- mor activity both in vitro and in vivo.26) During investi- gation of the antitumor mechanism, we noticed that vicenistatin induces vacuole-like structures in several mammalian cell lines (Fig. 1(B)). In particular, large vacuoles occupying over half of the cell were formed in normal rat fibroblast 3Y1 cells at 300 nM within 2 h of treatment, but not at higher concentrations (Fig. 1(B), Supporting information Fig. S1). Time-lapse analysis revealed that these large vacuoles were formed by repetitive homotypic fusion of small vesicles (Fig. 1(C) arrows). Interestingly, some large vacuoles were broken down to small vacuoles (Fig. 1(C), arrow- head). The effects of vicenistatin on vacuole formation seem to be irreversible, because vicenistatin-induced vacuoles were not disappeared even 18 h after vicenis- tatin removal (Supporting information Fig. S2). Fur- thermore, the concentrations required for vacuolation on 3Y1 cells were lower than that for cytotoxicity (IC50 value was about 1 μM), suggesting thatvacuolation activity and cytotoxicity by vicenistatin are induced by the independent mechanisms.To reveal the organelle from which these vacuoles were derived, we stained vicenistatin-induced vacuoles with several organelle markers (Fig. 2(A)). The orga- nelle-specific fluorescent probes, or marker protein/lipid for ER (ER Tracker), mitochondria (Mito Tracker), Golgi (Golgi 58K protein/FTCD), and late endosomes (LBPA), were not colocalized with the membrane of vacuoles. In striking contrast to these markers, EEA1, an early endosome marker, was colocalized with the membrane of most of the vicenistatin-induced vac- uoles.2)

These results strongly suggested that the source of the vacuole-like structures induced by vicenistatin was the early endosomes.It has been reported that the vacuolation of late endosomes/lysosomes induced by bacterial toxins requires both V-ATPase activity and extracellular chlo- ride ions.29,30) Therefore, we next examined the effects of V-ATPase inhibitor and of chloride ion deprivation on vicenistatin-induced vacuolation (Fig. 2(B) and (C)). The effect of bafilomycin A1 (BafA1), a potent inhibi- tor of V-ATPase, was confirmed by the staining with acridine orange, a weak base fluorescent reagent thataccumulates in acidic organelles (Fig. 2(B)). In the absence of BafA1 pretreatment, many small signals of acridine orange were observed around nuclei, indicating the presence of acidic vesicles, including early/late endosomes and lysosomes. Under this condition, the vicenistatin-induced vacuoles were weakly stained with acridine orange. In contrast, all the acridine orange-pos- itive vesicles had disappeared in both the control and vicenistatin-treated cells by pretreatment with 10 nM BafA1. Furthermore, no vicenistatin-induced vacuoles were observed in the BafA1-pretreated cells. To investi- gate the effects of chloride ion, we used gluconate/ BSA buffer in which the contents of chloride ion in PBS/BSA was exchanged to that of gluconate ion (Fig. 2(C)). Vicenistatin efficiently induced the vacuola- tion in 3Y1 cells incubated with PBS/BSA, but drasti- cally decreased the vacuolation efficiency in cells incubated with gluconate/BSA buffer. Unlike the bacte- ria vacuolating toxin VacA, the chloride channel inhibi- tors did not inhibit the vicenistatin-induced vacuolation (Supporting information Fig. S3).31,32) These results strongly suggested that the vacuolation induced by vicenistatin requires proton influx in early endosomes and chloride ions utilized as counter ions for protons.

Fig. 2. Vicenistatin-induced vacuoles were derived from early endosomes. (A) 3Y1 cells were treated with 300 nM vicenistatin for 2 h and stained with organelle-specific markers. Early endosome, late endosome, ER, Golgi, and mitochondria were stained with anti-EEA1 antibody, anti- LBPA antibody, ER Tracker, anti-Golgi 58K protein/formiminotransferase cyclodeaminase antibody, and Mito Tracker, respectively. (B) 3Y1 cells were treated with or without 10 nM bafilomycin A1 (BafA1) for 2 h and then treated with 300 nM vicenistatin for an additional 2 h. Acidic orga- nelles were stained with acridine orange for 15 min and photographed. (C) 3Y1 cells were treated with gluconate/BSA buffer (–Cl− buffer) in which contents of chloride ion in PBS/BSA (+Cl− buffer) were exchanged to that of gluconate ion for 2 h and then treated with 300 nM vicenista- tin for further 2 h.

Vicenistatin-induced vacuolation via the activation of Rab5-PAS pathway
It has been thought that the Rab5-PAS pathway is important for homotypic fusion of early endosomes. In particular, the deprivation of PtdIns(3,5)P2 by the inac- tivation of PIKfyve causes swelling of an endocytic vacuole-like structure.23) Because EEA1 is also concen- trated on the surface of a subset of vacuole-like struc- tures induced by PIKfyve inhibitor YM201636, we investigated the relationships between PIKfyve activity and vicenistatin-induced vacuolation. HEK293T cells were cotransfected with EGFP and FLAG-PIKfyve and then treated with vicenistatin. Vacuoles formed in EGFP transfected cells efficiently, but their formation was suppressed in EGFP and FLAG-PIKfyve cotrans- fected cells (Fig. 3(A)). Furthermore, pretreatment with dioctanoyl PtdIns(3,5)P2 (diC8-PI(3,5)P2), an analog of the product of PIKfyve, but not with dioctanoyl phos- phatidic acid (diC8-PA), suppressed vicenistatin-induced vacuolation (Fig. 3(B)). These results strongly sug- gested that vicenistatin suppresses PtdIns(3,5)P2 pro- duction in cells. However, vicenistatin did not inhibit PIKfyve kinase activity in vitro (Fig. 3(C)). We next investigated the effects of Rab5, which regulates the recruitment of the PAS complex. To determine whether or not Rab5 activation is required for vicenistatin-in- duced vacuolation, we transfected the EGFP-fused wild-type Rab5a (Rab5aWT) or GTP binding-deficient mutant Rab5a (Rab5S34N) into HeLa cells and observed the effects on vicenistatin-induced vacuolation. Vacuole formation was efficiently induced by vicenistatin in the transfectants of EGFP alone but not in the EGFP- Rab5aS34N-transfected cells (Fig. 3(D), Supporting information Fig. S5). Taken together, these results sug- gest that vicenistatin induces vacuolation by reducing the amount of PtdIns(3,5)P2 through the activation of the Rab5-PAS pathway.

Vicenistatin increased the membrane surface fluidity
Recently, it was reported that heronamide C, another vacuolating compound, binds to sphingolipid.24,25) Because vicenistatin is an amphiphilic compound like heronamide C, there is a possibility that vicenistatin binds to lipid of the plasma membrane and changes the physical properties of membrane. To test this idea, we investigated the effects of vicenistatin on the hydrody- namic diameter, zeta potential, and membrane fluidity in an in vitro liposome system. The membrane con- stituents of early endosomes are thought to resemble those of the plasma membrane.33) Therefore, we pre- pared liposomes consisted of PC, PE, and cholesterol (7:3:1 M ratio, w/ cholesterol) with or without vicenis- tatin. Cholesterol-depleted liposomes (PC:PE = 7:3) with or without vicenistatin were also prepared to investigate cholesterol’s effect on the membrane proper- ties. After the extrusion of liposome through a 100-nm pore membrane followed by mixing with fluorescent probes, DPH, or TMA-DPH, the hydrodynamic diame- ter, zeta potential, and polarities were measured. No differences in the hydrodynamic diameter or zeta potential were observed in the presence or absence of vicenistatin (Supporting information Fig. S4), but the fluidity (1/P, reciprocal of polarity) of the cholesterol-free liposome membrane increased drasti- cally in the presence of vicenistatin (Fig. 4(A) and (B), -Chol). More importantly, vicenistatin also increased the 1/PTMA-DPH, but not 1/PDPH, of liposomes contain- ing cholesterol (Fig. 4(A) and (B), +Chol). Because DPH is located in the hydrophobic membrane interior, whereas TMA-DPH is usually present at the lipid/water interface due to the charged moiety, these results sug- gest that vicenistatin increases the fluidity of the mem- brane surface of cholesterol-containing liposomes.

Membrane fusion requires the transient disorder of lipid integrity to form a membrane curvature and to mix with donor and acceptor vesicles. Therefore, the increase in membrane surface fluidity by vicenistatin is a good candidate for a factor involved in vacuolation. Because cholesterol increases the rigidity of the mem- brane (Fig. 4(A) and (B)), the decrease in the choles- terol content might enhance vicenistatin-induced vacuolation. Therefore, we next investigated the effects of cholesterol deprivation on vicenistatin-induced vac- uolation. Because cyclodextrin (CD) and its analogs have been widely used to control the amount of cholesterol in the plasma membrane,34) we used methyl-β-cyclodextrin (MβCD) to reduce the choles- terol in the plasma membrane. After cholesterol was reduced by treatment with 5 mM MβCD for 1 h, MβCD was removed by washing with PBS, and the cells were then treated with vicenistatin for 2 h. As shown in Fig. 3(C), 100 nM vicenistatin could not induce the vacuole-like structure in control cells that were not treated with MβCD (left column). However, vicenistatin did induce the vacuole-like structure forma- tion in cholesterol-deprived cells even at 100 nM (middle column). The population of vacuolated cells was up to about 41.9%. Furthermore, the population was suppressed to 19.4% by cholesterol incorporation into the plasma membrane by the cholesterol/MβCD complex (Fig. 4(C), right column). These results strongly suggest that cholesterol content is also an important factor for determining the sensitivity of vicenistatin-induced vacuolation in cells.

Discussion
Endosome system plays pivotal functions in signal transduction; therefore, its regulation of trafficking and maturation are tightly regulated in the cells. In this study, we showed that the antitumor compound vicenis- tatin induced the large vacuole formation in cells by disturbing the membrane trafficking and fusion pro- cesses. Vicenistatin-induced vacuolation was dependent on the activity of V-ATPase and the presence of extra- cellular chloride ions (Fig. 2). Together with the results that vacuoles induced by vicenistatin were weakly acid- ified, these results indicate that efficient proton incorpo- ration is required for vicenistatin-induced vacuolation. Vacuoles induced by vicenistatin were surrounded with EEA1, a maker protein of early endosomes (Fig. 2(A)), suggesting that the vacuoles were derived from the early endosomes. Exogenous overexpression of PIKfyve and supply of PtdIns(3,5)P2 analog sup- pressed vicenistatin-induced vacuole formation (Fig. 3(A) and (B)), suggesting that cellular amount of Fig. 3. Rab5-PAS pathway is involved in vicenistatin-induced vacuolation. (A) PIKfyve overexpression decreased vicenistatin-induced vacuole formation. pEGFP-C3 and pcDNA3-FLAG-PIKfyve (PIKfyve) co-transfected HEK293T cells were treated with 300 nM vicenistatin for 2 h, and the number of vacuolated cells was calculated. Open and closed bars indicate the vacuolated cell population in DMSO- and vicenistatin-treated cells, respectively. Statistical analyses were done by unpaired Student’s t-test. (B) Pretreatment of dioctanoyl PI(3,5)P2 (diC8-PI(3,5)P2), but not dioctanoyl phosphatidic acid (diC8-PA), suppressed vicenistatin-induced vacuolation in 3Y1 cells. After 10-min treatment of 1 mM diC8-PA in PBS or diC8-PI(3,5)P2 in PBS, 1 μM vicenistatin was treated for 2 h and counted vacuolated cells (totally above 200 cells each sample). Values are expressed as mean ± deviation of the percentage of vacuolated cells from two independent experiments.

Structures of sodium salts of diC8-PA and diC8-PI(3,5)P2 are shown. (C) Vicenistatin is not the PIKfyve kinase inhibitor. FLAG-PIKfyve activity was measured using immunoprecipitated PIKfyve derived from non-transfected or transfected HEK293T cells. 100 μM PtdIns and γ-[32P]ATP were used as substrates. Kinase reaction was performed in the presence of 10 μM YM600123 or vicenistatin at 37 °C, 30 min. YM600123 inhibited PtdIns(5)P production, but vicenistatin did not. Values are expressed as mean ± s.d. from four independent experiments. (D) Dominant-negative mutant of Rab5 (Rab5S34N) repressed vicenis- tatin-induced vacuole formation. pEGFP-C3 or pEGFP-C3-Rab5S34N was transfected in HeLa cells and treated with 300 nM vicenistatin for 2 h. The number of vacuolated cells was calculated. Values are expressed as mean ± s.d. from three independent experiments. Statistical analyses were done by unpaired Student’s t-test. PtdIns(3,5)P2 is reduced in vicenistatin-treated cells. Since vicenistatin showed no inhibitory effect on PIK- fyve kinase activity in vitro (Fig. 3(C)), it is suggested that vicenistatin reduces the amount of PtdIns(3,5)P2 through the activation of the Rab5-PAS pathway. Indeed, exogenous expression of Rab5aS34N, a domi- nant-negative mutant of Rab5, suppressed the vicenista- tin-induced vacuolation (Fig. 3(D), Supporting information Fig. S5). These results strongly suggested that vicenistatin induces vacuolation via Rab5 activa- tion.

The mechanism of the suppression of PIKfyve activity by Rab5 activation remains to be revealed. One possibility is the constitutive activation of Rab5 results in the suppression of recruitment of PAS complex on early endosomes. Many factors including EEA1 and PIKfyve are recruited to early endosome through their PI3P-binding motif. Continuous activation of Rab5 leads the stable binding with EEA1, which might competitively inhibit the PAS complex binding on early endosomes. The other possibility is the constitutive activation of Rab5 results in the stable complex forma- tion with PtdIns3Ks and accumulates PtdIns(3)P on the membrane of early endosomes. In this case, even if PIKfyve is recruited to and produces PtdIns(3,5)P2 on early endosome, the ratio of PtdIns(3,5)P2/PtdIns(3)P is too low to inhibit the homotypic fusion. Further inves- tigation is required to reveal the mechanism of Rab5 activation-induced fusion of early endosomes. To obtain information about the other factors required for vacuolation by vicenistatin, we focused on the structural properties of vicenistatin and investigated the effects of cholesterol on the vicenistatin-induced vacuolation. In vitro analyses suggested that vicenista- tin increased the fluidity of the membrane surface of Fig. 4. Effects of cholesterol on vicenistatin-induced vacuolation and membrane fluidity. (A) Effects of vicenistatin on membrane surface fluidity. 25 μM liposomes containing 1.25 μM vicenistatin and/or cholesterol (Chol) were mixed with a fluorescent probe, TMA-DPH, for 1 h, and polarization was measured. Values are expressed as mean ± s.d. from three independent experiments. Statistical analyses were done by unpaired Student’s t-test. (B)

Effects of vicenistatin on the fluidity of the membrane interior. 25 μM liposomes containing 1.25 μM vicenistatin and/or cholesterol (Chol) were mixed with a fluorescent probe, DPH, for 1 h, and the polarization was measured. Values are expressed as mean ± s.d. from three independent experiments. Statistical analyses were done by unpaired Student’s t-test. (C) Effects of cholesterol content on vicenistatin-induced vacuolation. The cholesterol was removed from the plasma membrane with 5 mM MβCD for 1 h and then treated with 100 nM vicenistatin (control and MβCD, respectively). To incor- porate cholesterol back into the cells, the cholesterol/MβCD complex (0.25 and 2.5 mM, respectively) was incubated with the cholesterol-removed 3Y1 cells for 1 h (MβCD → cholesterol), followed by 100 nM vicenistatin treatment. After 2 h of incubation with vicenistatin, the cells were fixed and observed. The percentage in each photograph indicates the vacuolated cell populations of representative experiment cholesterol-containing liposomes (Fig. 4(A)). Probably more importantly, vicenistatin increased the fluidity of not only the surface but also the interior of the mem- brane in cholesterol-free liposomes (Fig. 4(A) and (B)). Because it is well known that cholesterol contributes to membrane rigidity and that membrane fusion requires the transient disorder of lipid integrity to form mem- brane curvature and to mix with donor and acceptor vesicles, these results prompted us to consider that the other factor required for efficient vacuolation may be fluidity of the membrane and that the cholesterol func- tions as a restriction factor for vicenistatin-induced vac- uolation. Indeed, the vacuolation activity of vicenistatin is dependent on the cholesterol contents of the plasma membrane, i.e., the reduction of cholesterol contents in the plasma membrane by MβCD increased the vicenis- tatin sensitivity (Fig. 4(C)).

In conclusion, we found that an antitumor com- pound, vicenistatin, induced the formation of early endosome-derived vacuole-like structures in cells by activating Rab5 and by increasing the fluidity of the membrane surface. The mechanism by which vicenista- tin activates Rab5-PAS pathway and the relationships between Rab5 activation and membrane disorder remain to be resolved. Recently, it has reported that theonellamides bind to cholesterol and induce phase separation in lipid membranes.35) Since theonellamides activate Rho1, a Rab5-like small G-protein,36) there is a possibility that the changes of membrane property by vicenistatin activate Rab5-PAS pathway. Further inves- tigation is required, but vicenistatin is a useful bioprobe for investigating the mechanisms of homotypic early endosome fusion and the function of cholesterol in membrane integrity.

Acknowledgments
We thank Professor T. Kobayashi (RIKEN) and Pro- fessor H. Ikeda (Kitasato Univ.) for giving us the anti- LBPA antibody and bafilomycin A1. We also thank Ms. Kaoru Ohyama and Mr. Yuki Ohizumi (Univ. Tsukuba) for technical suggestions in the preparation of liposomes and the measurement of their properties.

Disclosure statement
No potential conflict of interest was reported by the authors.

Funding
This work was supported in part by a Japan Society for the Pro- motion of Science (JSPS KAKENHI) [Grant number 20380065]; Grant-in-Aid for Scientific Research on the Innovative Area “Chemi- cal Biology of Natural Products” from the YM201636 Ministry of Education, Culture, Sports, Science and Technology, Japan [Grant number 23102013]; Naito Foundation Subsidy for Promotion of Specific Research Projects.