Mammalian Cell Morphology and Endocytic Membrane Homeostasis Require Enzymatically Active Phosphoinositide 5-Kinase PIKfyve*

The dual specificity mammalian enzyme PIKfyve phos-phorylates in vitro position D -5 in phosphatidylinositol (PtdIns) and PtdIns 3-P, itself or exogenous protein sub-strates. Here we have addressed the crucial questions for the identity of the lipid products and the role of PIKfyve enzymatic activity in mammalian cells. CHO, HEK293, and COS cells expressing PIKfyve WT at high levels and > 90% efficiencies increased selectively the intracellular PtdIns 3,5-P 2 production by 30–55%. In these cell types PtdIns 5-P was undetectable. A kinase-deficient point mutant, PIKfyve K1831E , transiently transfected into these or other cells elicited a dramatic dominant phenotype. Subsequent to a dilation of the PIKfyve-containing vesicles, PIKfyve K1831E -expressing cells progressively accumu-lated multiple swollen lucent vacuoles of endosomal ori-gin, first in the perinuclear cytoplasm and then toward the cell periphery. Despite their drastically altered morphology, the PIKfyve K1831E -expressing cells were viable and functionally active, evidenced by several criteria. This phenotype was completely reversed by introducing PIKfyve WT into the PIKfyve K1831E -transfected cells. Dis-ruptions of the localization signal in the PIKfyve kinase-deficient mutant yielded a PIKfyve K1831E D fyve protein, in-competent

Despite continuous inward and outward membrane flow, eukaryotic cells maintain their morphology, endomembrane homeostasis, and composition of the intracellular organelles by coordinated interactions between multiple molecular elements located on different membranes (for a recent review, see Ref. 1). The mechanisms and molecular factors integrating these interactions are just beginning to be understood. It has become increasingly clear that phosphatidylinositol (PtdIns) 1 and its derivatives, phosphorylated separately or in all possible combinations on D-3, D-4, or D-5 positions of the inositol head group (called collectively phosphoinositides, PI), are key elements in both the constitutive and regulated membrane traffic (for recent reviews, see Refs. [2][3][4][5][6][7][8]. This notion is supported by studies in yeast and mammalian cells demonstrating defects in membrane trafficking events upon inactivation of the enzymes responsible for PI biosynthesis and vice versa, activation of transport steps (for example GLUT4-vesicle dynamics) upon overproduction of the PI-generating enzyme activities. A major group of the proteins implicated in the generation of vesicle carriers or regulation of their fusion with the target membrane are PI-binding molecules implying that site-and time-specific PI production and PI-protein interactions are crucial for membrane trafficking events.
Recent studies in Saccharomyces cerevisiae suggest a distinct function of PtdIns 3,5-P 2 intracellular levels in yeast membrane trafficking (9 -11). The major phenotypic characteristics resulting from inactivation of yeast fab1, whose gene product is responsible for the intracellular PtdIns 3,5-P 2 production, include severe growth defect and extremely enlarged vacuoles that occupy the majority of the cell (9,12). Despite these severe defects, however, all transport pathways to the vacuole in Fab1p-deficient cells appear intact (11). Emr and collaborators (11) therefore suggest that Fab1p kinase and PtdIns 3,5-P 2 function to maintain vacuolar size and membrane homeostasis by regulating recycling/turnover of membranes from the yeast vacuolar surface to earlier compartments.
PIKfyve (phosphoinositide kinase for five position containing a fyve finger) appears to be the mammalian ortholog of Fab1p lipid kinase that rescues the vacuolar defects in the ⌬fab1 yeast strain (13,14). In vitro, PIKfyve lipid kinase synthesizes Pt-dIns 5-P and PtdIns 3,5-P 2 in a wortmannin-resistant fashion (ID 50 , 600 nM) (15). Recent studies indicate that PIKfyve is also a protein kinase that likely acts in intact cells to modulate PIKfyve lipid kinase activity and/or specificity by autophosphorylation (16). PIKfyve partitions between the soluble and membrane-bound intracellular pools: the membrane-bound populations being visualized as distinct vesicles largely positive for late endosomal markers, but not for protein residents of earlier compartments in the endocytic pathway (17). This characteristic intracellular localization is most likely conferred by PIKfyve's FYVE finger, a PtdIns 3-P-binding protein module found in other mammalian proteins as a major localization determinant for the endosomal membranes, enriched in PtdIns 3-P (7,18). However, despite this intensive characterization, the functions of PIKfyve enzymatic activity as well as the identity of its lipid or protein products in the context of live mammalian cells remain unknown. In this study we report the essential role of PIKfyve enzymatic activity in cell morphology and intracellular membrane integrity by characterization of a PIKfyve kinase-dead point mutant. Mammalian cells expressing PIKfyve K1831E display a dominant phenotype characterized by progressive, but reversible, PIKfyve-vesicle dilation and vacuolation of membranes with endocytic origin. Because high levels of PIKfyve WT increase selectively the endogenous PtdIns 3,5-P 2 production in these cells, we conclude that PIKfyve enzymatic activity, likely through PtdIns 3,5-P 2 production, plays an essential role to maintain cell morphology by regulating late endocytic membrane homeostasis.
Generation of pEGFP-HA-PIKfyve S , or the HA-tagged PIKfyve S (PIKfyve S ⌬CH , PIKfyve S ⌬fyve , and PIKfyve S K1831E ) cDNA constructs in pCMV5 vector was described previously (15)(16)(17). EGFP-HA-PIKfyve ⌬CH (⌬560 -1231) and ⌬fyve(⌬177-198) truncated mutants were generated by replacing the full-length PIKfyve S released as a XbaI fragment from pEGFP-HA-PIKfyve S cDNA, with the corresponding fragments from pCMV5-HA-PIKfyve S ⌬CH and pCMV5-HA-PIKfyve S ⌬fyve , respectively. pEGFP-HA-PIKfyve S K1831E⌬fyve and pEGFP-HA-PIKfyve S K1831E were generated by replacing the KpnI/SalI fragment in pEGFP-HA-PIKfyve S ⌬fyve and the XbaI/SalI insert in pEGFP-HA-PIKfyve S with the corresponding fragments from pCMV5-HA-PIKfyve S K1831E cDNA. Transient Transfection, Fluorescence Microscopy, and Western Blotting-COS-7 cells, seeded on 22 ϫ 22-mm coverslips were transfected with the cDNA constructs indicated in the figure legends by Lipo-fectAMINE (Life Technologies, Inc.). Cells were then processed for fluorescence microscopy at the indicated posttransfection time or further treated as described below. Cells transfected with HA-PIKfyve constructs in pCMV5 were detected with anti-HA monoclonal or polyclonal antibodies and Texas red-coupled goat anti-mouse or CY3-coupled goat anti-rabbit IgG, respectively, following fixation (4% formaldehyde) and washing as described previously (17). Cells transfected with pEGFP constructs were detected by the GFP fluorescence signal following cell fixation or directly in live cells. HEK293 cells were transfected with PIKfyve constructs in an adenoviral plasmid (see below). Coverslips were mounted on slides using the Slow Fade Antifade Kit (Molecular Probes). Fluorescence analyses were performed with a confocal microscope (Zeiss LSM 310) using a 63/1.4 oil or 40/0.75 water immersion lense and a standard green fluorescence filter (for GFP). The levels of overexpressed EGFP-PIKfyve constructs were compared with that of the endogenous PIKfyve by Western blotting with anti-PIKfyve antibodies as described previously (17).
Preparation of Recombinant Adenovirus and Cell Infection-Recombinant adenoviruses, expressing HA-tagged PIKfyve S and GFP (AdPIKfyve), or GFP alone (AdEmpty), were generated by the AdEasy system (kindly provided by Dr. B. Vogelstein, Ref. 19). Briefly, HA-PIKfyve S was first cloned into pAdTrack-CMV shuttle vector, engineered with two separate CMV promoters for expression of GFP, and for expression of PIKfyve S , respectively. For this purpose the N-terminal part of PIKfyve S , released by BglII-KpnI from pEGFP-HA-PIKfyve S , and the KpnI-SalI C-terminal part taken from pCMV5-HA-PIKfyve S , were ligated into a BglII-SalI-digested shuttle vector. The resultant shuttle construct or an "empty" shuttle vector was linearized with PmeI and co-transformed with pAdEasy-1 adenoviral backbone plasmid into Escherichia coli BJ5183 cells. Selected recombinants were confirmed by restriction mapping, linearized with PacI, and used to transfect a HEK293 adenovirus packaging cell line with LipofectAMINE. Two weeks posttransfection the cells were harvested. The viruses were extracted by freeze-thaw and subsequent centrifugation. This viral extract was used for further viral propagation in HEK293 cells. Viral stocks were purified by ultracentrifugation in two discontinuous CsCl 2 gradients and subsequent passage through a Nap 10 column (Sephadex G25, Amersham Pharmacia Biotech). Purified viral stocks were titrated, and the lowest dilution resulting in 100% infection of HEK293 cells 18 h postinfection (monitored by the GFP signals) was defined as multiplicity of infection ϭ 1. The adenovirus-mediated overexpression and enzymatic activity of PIKfyve were confirmed by Western blotting and lipid kinase assays, performed under previously specified conditions (15).
Treatments of Transfected Cells-COS-7 cells transfected with pCMV5 PIKfyve K1831E cDNAs were infected 24 h posttransfection with recombinant adenoviruses expressing PIKfyve and GFP, or control GFP. Twenty-four to 48 h postinfection the cells were fixed, processed for fluorescence microscopy, and observed under a 63/1.4 oil immersion lens. In some experiments COS-7 cells transfected with pEGFP-HA-PIKfyve WT were treated 24 h posttransfection with NHCl 4 (20 mM) or wortmannin (800 nM) for 30 min at 37°C and then examined live by confocal microscopy with a 40/0.75 water immersion lens.
Labeling of Cellular Phospholipids with 32 P and HPLC Analysis-COS-7 cells seeded on 60-mm dishes were transfected with pCMV5-HA-PIKfyve WT or only pCMV5 cDNAs by electroporation (Bio-Rad Gene Pulser II electroporator), a technique reaching Ͼ90% cell transfection efficiency (20). CHO-T or HEK293 cells (60-mm dishes) were infected with adenoviruses expressing PIKfyve WT and GFP, or only GFP at multiplicity of infection ϭ 1. Fluorescence microscopy of the infected cells indicated ϳ100% efficiency 24 h posttransduction. Forty-eight hours posttransfection or 24 h postinfection, the cells were washed in phosphate-free Dulbecco's modified Eagle's medium and then labeled for 3 h at 37°C in phosphate-free Dulbecco's modified Eagle's medium, containing 0.5% bovine serum albumin, 2 mM sodium pyruvate, and 0.6 mCi/ml of [ 32 P]orthophosphate (PerkinElmer Life Sciences) as described previously (21). 3T3-L1 adipocytes differentiated as described previously (13,17) were labeled in a similar way on day 10 of the differentiation program. Cell monolayers were washed with ice-cold phosphate-buffered saline and scraped with CH 3 OH/1 M HCl (1:1) in the presence of 5 mM EDTA and 5 mM tetrabutylammonium sulfate, as a reagent increasing PI extraction (22). Extracted radiolabeled lipids were deacylated and analyzed by HPLC with 3 H-labeled GroPIns 3-P, GroPIns 4-P, GroPIns 5-P, and GroPIns 4,5-P 2 internal or 32 P-labeled GroPIns 3,4-P 2 and GroPIns 3,5-P 2 external standards under previously specified conditions (15).

RESULTS AND DISCUSSION
To determine the function of PIKfyve enzymatic activity in mammalian cells we have used a catalytically deficient HAtagged form of PIKfyve, bearing a single point mutation at Lys 1831 proposed as a candidate ATP-binding site (16). Lack of both lipid and protein kinase activity for this PIKfyve K1831E point mutant was demonstrated in vitro by the complete inability of the HA immunoprecipitates derived from lysates of PIKfyve K1831E -transfected COS cells to synthesize PtdIns 5-P and PtdIns 3,5-P 2 or to autophosphorylate (16). Transient transfection in COS cells with HA-PIKfyve K1831E cDNA engineered in mammalian expression vectors pCMV5 or pEGFP resulted in a drastic dominant phenotype, readily seen upon observation of live cells under a light microscope. Approximately 15 h posttransfection, large vacuoles appeared as round structures, initially at the perinuclear region and 24 h posttransfections, throughout the whole cytoplasm with a tendency to increase in size and decrease in number within the observation period (72 h posttransfections). In some cells expressing PIKfyve K1831E at higher levels, two to three giant vacuoles could be seen 2 days posttransfection, most likely having arisen as a result of fusion, reaching diameters of 5-10 m and comprising the majority of the cell volume. The intracellular appearance of PIKfyve K1831E also underwent a dramatic change. The first alterations, which actually preceded the formation of the large vacuoles were associated with dilation of the PIKfyve K1831E -containing vesicles as revealed by fluorescence signals associated with anti-HA antibodies (Fig. 1) or GFP (data not shown). With the progression of cell vacuolation, the fluorescence signals detected PIKfyve K1831E -containing dilated vesicles distributed between the empty vacuoles or in some instances on the limiting membrane of the vacuoles typically at the perinuclear region (Fig. 1). The vacuoles at the cell periphery were typically negative for PIKfyve K1831E . The lumen of the vacuoles was free of PIKfyve K1831E and appeared as an empty, lucent space in phase contrast images from a confocal microscope (Fig. 1). The cell vacuolation was striking and unequivocal and was detected even in cells for which the HA immunofluorescence or GFP fluorescence signals of expressed PIKfyve K1831E were only slightly above the background level. Western blotting with anti-PIKfyve antibodies of lysates from transiently transfected COS cells showed equal intensities for the endogenous PIKfyve (ϳ200 kDa) and GFP-PIKfyve K1831E (230 kDa) bands (data not shown). Given the transfection efficiency in these experiments (ϳ40%) and the variability of overexpression in individual cells, these results support the notion that even at a low ratio ( ϳ 1:1) of the mutant to the endogenous PIKfyve, the PIKfyve K1831E point mutant induces the appearance of dilated PIKfyve-containing vesicles and abnormal vacuoles. Parallel immunofluorescence and phase contrast microscopy in COS cells transiently expressing HA-PIKfyve WT revealed no morphological changes (Fig. 1). Consistent with our previous observations (17), a distinct peripheral vesicular pattern of the fluorescence staining associated with expressed PIKfyve WT , along with a diffuse staining could be observed (Fig. 1). Because PIKfyve WT as well as PIKfyve K1831E largely co-localize with the late endosomal marker CI-MPR, but not with marker proteins for Golgi, early/recycling endosomes or end lysosomes (Fig. 2, Ref. 17, and this study, data not shown), it is conceivable that the PIKfyve K1831E -induced vacuoles originate from the late endosomal compartment.
The dramatic and unequivocal dominant effect of PIKfyve K1831E in inducing cell vacuolation was observed in other cell types. Thus, transient transfection of HIRcB or 3T3-L1 fibroblasts with the PIKfyve K1831E cDNA in either pCMV5 or pEGFP vectors resulted in the appearance of the characteristic phenotype of multiple empty vacuoles in the PIKfyve K1831E -expressing cells (data not shown). Next, transient transfection of HEK293 cells with PIKfyve K1831E engineered in pAdTrack CMV vector with an independent expression of GFP as a transfection reporter protein induced the characteristic dominant phenotype. Multiple lucent vacuoles were observed by phase contrast only in the cells transfected with PIKfyve K1831E but not with PIKfyve WT (Fig. 1, c and f). When viewed by the fluorescence signals of GFP, the vacuolated compartments were seen as "holes" (Fig. 1c). Although the PIKfyve K1831E -transfected cells displayed a drastically altered morphology, they were functionally active within the observation period (72 h posttransfections), as evidenced by their ability to exclude trypan blue dye or propidium iodide, to adhere to tissue culture plates subsequent to trypsinization, to endocytose materials, such as dextran (data not shown) or adenoviruses (see below), and to selectively reverse the phenotype (see below). Together these results demonstrate that expression of a kinase-dead PIKfyve K1831E point mutant induces a drastic dominant phenotype, associated with abnormal cell morphology and endocytic membrane vacuolation, and suggest a key function of the products of PIKfyve enzymatic activity in maintaining cell architecture and endomembrane homeostasis.
We reasoned that if the above statement is correct then increases of PIKfyve enzymatic activity in the PIKfyve K1831Etransfected cells should reverse the cell vacuolation phenotype. To test this we introduced PIKfyve WT by adenovirus-mediated gene delivery in COS cells 24 h posttransfection with PIKfyve K1831E cDNA. Infected cells were monitored over a period of 16 -90 h by the fluorescence signals of GFP expressed in parallel with HA-PIKfyve WT but under an independent promoter. The vacuoles diminished in diameter ϳ16 h posttransduction and completely disappeared ϳ36 h posttransduction only in those PIKfyve K1831E -expressing cells for which infections with the PIKfyve WT adenovirus could be documented (Fig. 3). Conversely, the PIKfyve K1831E -expressing cells that remained noninfected with the PIKfyve WT adenovirus displayed the characteristic vacuolar phenotype until the end of the observation period (Fig. 3). Concordantly, a recombinant adenovirus encoding only GFP was ineffective in rescuing the phenotype in the PIKfyve K1831E -transfected cells within the observation period; the swollen vacuoles remained unchanged and were seen as empty "ghosts" (Fig. 3). These data demonstrate that a selective expression of PIKfyve WT rescues the abnormal phenotype induced by the PIKfyve K1831E , indicating the crucial role of the PIKfyve enzymatic activity for the morphology of the vacuolating compartments.
PIKfyve has a strong in vitro enzymatic activity to generate PtdIns 3,5-P 2 and PtdIns 5-P (15), both of which are now detected in several mammalian cell types (23)(24)(25)(26). To determine the identity of the lipid products generated in intact cells, for which the vacuolar phenotype was documented, we have examined the effect of PIKfyve WT overexpression on cellular phosphoinositide levels. COS-7, HEK293, and CHO-T infected with adenovirus expressing PIKfyve WT , or transfected by electroporation at an efficiency of infection/transfection Ͼ90%, were subsequently labeled with [ 32 P]orthophosphate and the extracted radioactive lipids analyzed by HPLC. Heterologous expression of PIKfyve WT in these cell types increased the [ 32 P]PtdIns 3,5-P 2 production by 30 -55% (Table I). In some of the cell types a decrease of PtdIns 3-P could be documented (Table I). However this was not always the case, as seen in COS cells (Table  I). Expressed PIKfyve WT had no significant effect on the radiolabeled levels of PtdIns 4,5-P 2 , PtdIns 3,4-P 2 , or PtdIns 4-P in the cell types studied (Table I). Surprisingly, in COS-7, HEK293, and CHO-T cells we were unable to detect any amounts of [ 32 P]PtdIns 5-P at basal conditions or under PIKfyve heterologous expression, similarly to previous observations in resting yeast cells or upon expression of PIKfyve WT in ⌬fab1 strain (9,14). Because our HPLC separation system readily detects substantial levels of endogenous PtdIns 5-P in [ 32 P]orthophosphate-labeled basal 3T3-L1 adipocytes, exceeding 4-fold the radioactive levels of PtdIns 3-P (Ref. 24 and data not shown), our inability to detect 32 P accumulation into PtdIns 5-P in the above cell types should indicate low levels or absence of this species, rather than a detection problem. Thus, it appears that under conditions of in vivo metabolic labeling, PtdIns 5-P is produced only in certain mammalian cells such as NIH-3T3 fibroblasts (23), 3T3-L1 adipocytes (24), and possibly others, for which a role of the PIKfyve enzymatic activity in their biosynthesis could be anticipated. This result indicates that although PIKfyve has the potential to generate both PtdIns 5-P and PtdIns 3,5-P 2 (15), only the latter is synthesized in COS and HEK293 cells, in which the dominant phenotype is restored upon PIKfyve WT expression. Together, the results are consistent with the notion that a selective increase of PtdIns 3,5-P 2 production in HEK293 and COS cells is sufficient to reverse the cell vacuolation induced by PIKfyve K1831E . This conclusion is in agreement with observations in ⌬fab1 yeast strain where expressed PIKfyve WT complements the phenotypic defects together with the basal PtdIns 3,5-P 2 concentra-tions (14). Noteworthy, the radiolabeled cellular PtdIns 3,5-P 2 remained largely unaltered upon heterologous expression of PIKfyve K1831E in COS cells (ϳ40% transfection efficiency). One explanation of this result is that the PIKfyve action as a lipid kinase is unrelated to the phenotype observed. However, based on similarity with the phenotypic changes associated with loss of the PtdIns 3,5-P 2 pool in yeast (9), we favor the concept that the PIKfyve K1831E -induced dominant negative effect is associated with inactivation of minor populations of PIKfyve enzymatic activity, most likely localized at specific key intracellular sites.
The above statement suggests that mutant PIKfyve K1831E , if mislocalized in cells, will be inefficient in inducing a dominantnegative effect. Both the endogenous and heterologously expressed HA-or GFP-tagged PIKfyve WT proteins display a characteristic discrete punctate staining, associated largely with the membranes of the late endosomal system (Ref. 17 and see above). Because expressed PIKfyve protein fragments bearing the FYVE finger exhibit a similar vesicular pattern, while the FYVE finger-deleted protein fragments did not (13), we assigned the FYVE finger in PIKfyve as a necessary determinant for intracellular localization, similarly to the observations with other FYVE finger-containing mammalian proteins (5,7). This assumption was confirmed in experiments in which we have characterized the localization pattern of the PIKfyve ⌬fyve truncated mutant versus PIKfyve WT or other PIKfyve mutants, harboring deletions in different regions of the molecule. As illustrated in Fig. 4, when expressed in COS cells, PIKfyve WT and PIKfyve ⌬CH produced the characteristic peripheral vesicular puncta scattered throughout the cytoplasm, while the protein with a FYVE finger truncation produced a dramatically different staining pattern consisting of mainly perinuclear and diffuse appearance; scattered puncta were practically undetected. Introduction of the kinase-dead point mutation into otherwise enzymatically active PIKfyve ⌬fyve (16) was typically without phenotype. The absence of endomembrane vacuolation in COS cells expressing PIKfyve K1831E⌬fyve is depicted in Fig. 4. Immunofluorescence detection of PIKfyve K1831E⌬fyve confirmed its mistargeting to the cell perinuclear region. Together, these results indicate that membrane attachment to late endosomes through the FYVE finger is required for the dominant-negative phenotype of PIKfyve K1831E .
It should be emphasized that a deletion of the chaperonin-like domain abrogated the in vitro lipid kinase activity associated with expressed PIKfyve ⌬CH (Ref. 15 and this study, data not shown). The fact that PIKfyve ⌬CH failed to induce a dominant phenotype in COS cells, despite its correct intracellular targeting (Fig. 4), could be taken as an indication that possibly slight PIKfyve enzymatic activity at the right site in the cellular context could prevent cell vacuolation. This notion is consistent with studies in yeast demonstrating that replenishment of as low as 10% of the PtdIns 3,5-P 2 pool reverses the vacuolar phenotype in fab1-deficient strains (9). An alternative interpretation could view the chaperonin-like domain as essential in recruiting/assembly of additional molecular elements, whose coordinated action in the absence of PtdIns 3,5-P 2 induces cell vacuolation.
Progressive and reversible mammalian cell vacuolation can be induced by different pharmacological treatments including weak bases (27), inhibition of the PI 3-kinase family members by wortmannin and LY294002 (28 -30), intoxication with the Helicobacter pylori VacA cytotoxin (31,32), or under other conditions (for a recent review, see Ref. 33). Similarly to the cell vacuolation induced by PIKfyve K1831E , the cytoplasmic vacuoles observed in the above experimental paradigms have been shown to originate from swollen vesicles of the postendosomal/ late endocytic pathway (28 -31). While the molecular mechanism(s) responsible for the vacuolar formation in these instances is largely obscure, we rationalized that if the vacuolation phenomenon depends on the PIKfyve pathway and membrane-localized PtdIns 3,5-P 2 production, then an intracellular overproduction of PIKfyve WT should prevent cell vacuolation induced by weak bases. Conversely, overproduced PIKfyve WT should be inefficient for rescue of the wortmannininduced vacuolation due to a direct dependence of the PtdIns 3,5-P 2 biosynthesis on cellular PtdIns 3-P levels (25), shown to undergo a dramatic depletion (70%) upon acute cell treatment with wortmannin (28,34,35). Consistent with our prediction, heterologously expressed PIKfyve WT prevented the COS cell vacuolation induced by short treatment with NH 4 Cl (20 mM; Fig. 5, a and b). By contrast, PIKfyve WT -transfected cells vacuolated just as well as the nontransfected cells upon wortmannin treatment (800 nM; Fig. 5, c and d). These results suggest that cell vacuolation, at least in the experimental paradigms examined here, is likely induced by common mechanisms, operating probably through PtdIns 3,5-P 2 production. With the reservation regarding the broad specificity in wortmannin inhibition, it is worth emphasizing that the most dramatic cell vacuolation is produced by high wortmannin concentrations (5-10 M; Ref. 28 and this study, data not shown), doses that cause a complete inhibition of not only the PI 3-kinase activity but that of PIKfyve (15) as well.
The results presented in this study indicate that the kinase-   (30). Inhibition of the retrograde transport out of lysosomes is proposed to yield the enlarged vacuoles documented in ⌬fab1 yeast strains, in which PtdIns 3,5-P 2 production is fully suppressed (9). An increase of the endosomal membrane permeability to anions and accumulation of osmotically active species, that cause an osmotic imbalance of late endosomes with subsequent vacuole formation, have been proposed for VacA-induced cell vacuolation (36). Based on the results presented herein and the above considerations we propose the following model to explain the PIKfyve K1831E -induced dilation of late endosomal membranes and cell vacuolation. Lack of PIKfyve enzymatic activity and, most likely, the PtdIns 3,5-P 2 production at specific sites within late endosomes/multivesicular bodies changes membrane composition and induces a dilation of PIKfyve endocytic compartment, due to a block in membrane outward/inward budding or an increase in its fusogenic activity. This event, alone or in conjunction with the loss of the PtdIns 3,5-P 2 pool and the subsequent changes in endocytic membrane composition, induces alterations in endomembrane ion permeability that lead to accumulation of ions, increases in osmotic pressure, and a subsequent vacuole formation. It is worth emphasizing that while this model predicts an essential function of the PtdIns 3,5-P 2 products of PIKfyve, a contribution of the protein products phosphorylated by PIKfyve protein kinase activity (16) could not be ruled out. Whether this model is correct requires testing of different aspects of the proposed steps. In any case the present results demonstrate for the first time the essential role of PIKfyve enzymatic activity in maintaining mammalian cell morphology and endocytic membrane homeostasis. The studies also add functional similarity to the structural homology between PIKfyve and yeast Fab1p, indicating functional conservation in these two proteins.
FIG. 4. Membrane attachment through an intact FYVE finger is required for the PIKfyve K1831E -induced cell vacuolation. COS cells transfected with the indicated constructs in pEGFP-HA vector were fixed and processed for fluorescence as described under "Experimental Procedures." A punctate pattern is seen in PIKfyve WT -and PIKfyve ⌬CH -transfected cells and diffuse staining upon elimination of the FYVE finger in PIKfyve ⌬fyve and PIKfyve K1831E⌬fyve . Absence of vacuolation phenotype for PIKfyve K1831E ⌬fyve is shown in the phase contrast image. Bar, 10 m.

FIG. 5. Protective effect of overexpressed PIKfyve WT against cell vacuolation induced by NH 4
؉ but not by wortmannin. COS-7 cells transfected with pEGFP-HA-PIKfyve WT cDNA were treated 24 h posttransfection with NH 4 Cl (20 mM) or wortmannin (800 nM) for 30 min at 37°C and then immediately observed live in a confocal microscope for the expression of GFP-PIKfyve WT by the GFP fluorescence signals (b and d) or under phase contrast for the vacuolation phenotype (a and c) as described under "Experimental Procedures." NH 4 ϩ treatment induced marked vacuolation in 82% Ϯ 8 (mean Ϯ S.E.) of the nontransfected cells and in 11% Ϯ 3 of the PIKfyve WT -expressing cells. Wortmannin treatment induced vacuolation in Ͼ95% of the nontransfected cells and in 85 Ϯ 8% of the PIKfyve WT -expressing cells. The quantitation is based on four separate experiments counting 10 randomly selected fields per experiment. Shown are: a cell expressing PIKfyve WT that fails to be vacuolated by NH 4 Cl treatment in contrast to the nontransfected control cell (a and b) and a vacuolation phenotype in a PIKfyve WT -expressing cell, which is similar to that in the nontransfected cells upon wortmannin treatment (c and d). Bar, 10 m.