|Home | About | Journals | Submit | Contact Us | Français|
The requirement of Vps34p, the sole phosphatidylinositol (PI) 3-kinase in Saccharomyces cerevisiae, for protein sorting to the vacuole in yeast has exemplified the essential role for phosphoinositides, phosphorylated derivatives of PI, in membrane trafficking. To better understand mechanisms that regulate PI 3-phosphate [PI(3)P]-mediated signaling, the role of the yeast myotubularin-related PI(3)P phosphatase Ymr1p was investigated. We found that Ymr1p and the synaptojanin-like phosphatase Sjl3p function as key regulators of the localization and levels of PI(3)P. Our data indicated that the ymr1Δ sjl3Δ double mutant aberrantly accumulated PI(3)P and demonstrated a steady-state redistribution of this lipid that leads to enrichment on the vacuolar membrane. This resulted in vacuole protein sorting defects, vacuolar fragmentation, and the misregulation of PI(3)P-specific effectors. Triple deletion of YMR1, SJL2, and SJL3 was lethal, suggesting an essential requirement for phosphatase-mediated PI(3)P regulation. Consistent with this, growth was restored to a ymr1Δ sjl2Δ sjl3Δ triple mutant by a PI(3)P-targeted Sac1p domain chimera (GFP-Sac1ΔC-FYVEEEA1) that returned PI(3)P to levels comparable with wild-type cells. Together, this study demonstrated that Ymr1p, a myotubularin phosphatase family member, functions in the control of PI(3)P-dependent signaling and the maintenance of endosomal system integrity. In addition, this work defined an essential overlapping role for lipid phosphatases in the regulation of 3′ phosphoinositides in yeast.
Eukaryotic cells contain multiple membrane-bound organelles, each with a distinct composition and cellular function. To maintain the integrity and identity of each organelle, the cell selectively transports cargo molecules through an elaborate network of vesicle trafficking pathways to the appropriate target organelle (Bonifacino and Traub, 2003 ). Phosphoinositides, phosphorylated derivatives of phosphatidylinositol (PI), can be reversibly modified on the D-3, D-4, and D-5 positions of the inositol head-group (reviewed in Fruman et al., 1998 ; Hughes et al., 2000 ) and serve as key regulators of membrane trafficking events (Odorizzi et al., 2000 ; Simonsen et al., 2001 ). In particular, phosphoinositides generated by the action of PI 3-kinases have been the subject of much investigation and have been shown to play integral roles in various signal transduction and membrane trafficking pathways (Rameh and Cantley, 1999 ; Martin, 2001 ; Simonsen et al., 2001 ; Pelham, 2002 ). In Saccharomyces cerevisiae, the demonstration that the sole PI 3-kinase Vps34 was required for efficient sorting of proteins from the late Golgi to the vacuole exemplified the importance of phosphatidylinositol 3-phosphate [PI(3)P] in vesicular trafficking within the endosomal system (Schu et al., 1993 ; Stack et al., 1993 ). Accordingly, a number of PI(3)P-specific effector proteins in yeast and higher eukaryotes have been found to contain FYVE (for Fab1, YGL023, Vps27, and EEA1) domains or PX (for phagocyte NADPH oxidase) domains, which can specifically recognize PI(3)P (Burd et al., 1998 ; Corvera et al., 1999 ; Sato et al., 2001 ; Simonsen et al., 2001 ).
The signaling events initiated by the binding of effectors to phosphoinositides are also regulated by PI phosphatases. Phosphatases that hydrolyze the D-3 or D-4 position phosphate share a signature CX5R catalytic motif (where X is any amino acid), which is characteristic of Sac1p domain-containing phosphatases (reviewed in Hughes et al., 2000 ), as well as DSP (dual specificity tyrosine and serine/threoninie phosphatases) such as PTEN and myotubularin (reviewed in Nandurkar and Huysmans, 2002 ). Phosphoinositide 5-phosphatases such as synaptojanin or SHIP typically remove the D-5 phosphate from phosphatidylinositol 4,5-bisphosphate [PI(4,5)P2] or phosphatidylinositol 3,4,5-trisphosphate [PI(3,4,5)P3] (Majerus et al., 1999 ). S. cerevisiae encodes four Sac1p domain-containing proteins, Sac1p itself, Fig4p, and the synaptojanin-like phosphatases Sjl2p/Inp52p and Sjl3p/Inp53p (hitherto referred to as Sjl2p and Sjl3p) (Guo et al., 1999 ). In vitro studies on the substrate specificity of the Sac1p domain-containing proteins in yeast have suggested that Fig4p is specific for the D-5 position of phosphatidylinositol 3,5-bisphosphate [PI(3,5)P2] (Rudge et al., 2004 ), whereas the Sac1p domains of Sac1p, Sjl2p, and Sjl3p are more promiscuous and can convert PI(3)P, phosphatidylinositol 4-phosphate [PI(4)P], and phosphatidylinositol 5-phosphate [PI(5)P] as well as PI(3,5)P2 to PI. In addition, Sjl2p and Sjl3p contain a separate PI (4,5)P2 5-phosphatase domain. This combination of phosphatase activities allows Sjl2p and Sjl3p to convert all major phosphoinositides in yeast to PI. Consistent with their in vitro activities, studies on Sjl2p and Sjl3p have indicated that they provide redundant functions with other PI phosphatases that are dedicated to the hydrolysis of specific phosphoinositides in vivo. For example, yeast mutants lacking the Sac1p domain-containing phosphatases Sac1p, Sjl2p, and Sjl3p demonstrated an essential overlapping role for these proteins in the control of PI(4)P and Golgi integrity (Foti et al., 2001 ). In addition, Sjl1p, Sjl2p, and Sjl3p were shown to overlap in the regulation of PI(4,5)P2 and actin cytoskeleton organization through their concerted PI 5-phosphatase activities (Stolz et al., 1998 ; Stefan et al., 2002 ).
Yeast also contains a single representative of the myotubularin family of phosphoinositide 3-phosphatases, Ymr1p (encoded by the YJR110W open reading frame) (Laporte et al., 1998 ). Myotubularins are grouped together as a family based on a conserved core functional domain organization consisting of an N-terminal PH-like GRAM domain (for glucosyltransferase, Rab-like GTPase activator and Myotubularin), a RID domain (for Rac-induced recruitment domain), a DSP domain (dual specificity tyrosine and serine/threonine phosphatase), and a SID domain (SET-protein interaction domain) (Laporte et al., 2003 ). The catalytically active myotubularins all exhibit specific activity against the d-3 position of PI(3)P and PI(3,5)P2, leading to the generation of PI and PI(5)P, respectively (Schaletzky et al., 2003 ). Among the 14 known human myotubularin-related phosphatases, three are mutated in severe genetic diseases: myotubular myopathy, and two distinct forms of Charcot-Marie-Tooth neuropathy (Nandurkar and Huysmans, 2002 ; Laporte et al., 2003 ). Although it has been speculated that myotubularins may regulate of PI(3)P-mediated endosome functions, the precise cellular roles of these proteins remains unknown (Laporte et al., 2003 ).
In this study, we sought to elucidate the cellular role of the yeast myotubularin-related phosphatase Ymr1p. Consistent with a previous study (Taylor et al., 2000 ), we found that Ymr1p is a phosphoinositide 3-phosphatase in vivo. However, deletion of YMR1 yielded no obvious phenotypes. Because previous work has indicated functional overlap between phosphoinositide phosphatases in yeast (Foti et al., 2001 ; Gary et al., 2002 ; Stefan et al., 2002 ), we took a genetic approach to identify those that could compensate for the absence of Ymr1p activity. We identified a specific genetic interaction between YMR1 and SJL3, because ymr1Δ sjl3Δ double mutant cells displayed an ~2.5-fold increase in cellular levels of PI(3)P, which aberrantly accumulated on the vacuolar membrane. Together, this study indicated that Ymr1p, a myotubularin family member, has a direct role in the regulation of PI(3)P-dependent signaling pathways that are important for the maintenance of endosome integrity. In addition, this work uncovered a novel requirement for phosphatase-mediated PI(3)P regulation in yeast; an essential function shared by Ymr1p, Sjl2p, and Sjl3p.
Enzymes used for recombinant DNA techniques were purchased from commercial sources and used as per manufacturers' recommendations. Methods were performed essentially as described previously (Sambrook et al., 1989 ). Sources for growth media for yeast and bacterial strains have been described previously (Gaynor et al., 1994 ). Standard yeast genetic methods were used throughout the study (Sherman et al., 1979 ). All oligos used in this study are available upon request.
All yeast strains used in this study are based on the parent strain SEY6210 (Table 1) (Robinson et al., 1988 ). The ymr1Δ strain was generated using a HIS3 deletion cassette strategy similar to that described previously (Audhya et al., 2000 ). A strain expressing Atg24 (Cvt13, Snx4)-GFP was created using a polymerase chain reaction (PCR) approach that has been described previously (Longtine et al., 1998 ). All gene deletions and epitope tag integrations were confirmed by genomic PCR. To generate all multiple mutant strains, the appropriate parent strains were mated and then sporulated. Tetrads were dissected, and germinated spores harboring the appropriate markers were selected. All disruptions were confirmed by genomic PCR analysis. To create the temperature-sensitive BPY13 strain, plasmid pRS416sjl2-8ts was transformed into BPY03 diploid cells, which were subsequently sporulated, and Ura+ prototrophic ymr1Δ sjl2Δ sjl3Δ triple mutants were identified by PCR analysis. The same strategy was used to create BPY14. The BPY97 strain was generated by transforming BPY13 with a LEU2-marked GFP-SAC1ΔC-FYVEEEA1 plasmid and subsequently selecting for transformants that grew on plates containing 5-fluorooroic acid (5-FOA). Generation of the ymr1ts strain YTS1 is described below.
The YMR1gene (YJR110W) was amplified by PCR by using the Pfu Turbo DNA polymerase (Stratagene, La Jolla, CA) from yeast strain SEY6210 with oligos that added a NotI restriction site to the 5′ end of the fragment, and a SacI site to the 3′ end. The resulting ~2.75-kb NotI-SacI fragment was then ligated into the LEU2-marked plasmid pRS415 (Sikorski and Hieter, 1989 ) (plasmid pBP02), and the insert was confirmed by DNA sequence analysis. Subsequently, the NotI-SacI fragment of pBP02 was subcloned into the URA3-marked plasmid pRS416 (plasmid pBP03) and the high copy number LEU2-marked plasmid pRS425 (plasmid pBP29). Plasmids containing the lipid-targeted GFP-SAC1ΔC chimeric genes were constructed as follows. PCR was used to amplify the SAC1 coding region from codons 1–522 lacking the transmembrane anchor of Sac1p (Foti et al., 2001 ) by using oligos that added a unique BglII site at the 5′ end and a unique SalI site at the 3′ end. This fragment was then digested accordingly and ligated into plasmid pGO35 (Odorizzi et al., 1998 ) to create plasmid GFP-SAC1ΔC. This plasmid was then used to create the lipid-targeted Sac1p domain constructs. For GFP-SAC1ΔC-2 × PHPLCd, the PI(4,5)P2-binding plectrin homology (PH) domain of PLCδ (Kavran et al., 1998 ) was amplified with oligos that added a SalI site (PH1), or a BamHI site (PH2) to the 5′ end of the sequence, and a BamHI site (PH1), or a KpnI site (PH2) at the 3′ end of the fragment. These PCR fragments were then digested with the appropriate enzymes and ligated into the GFP-SAC1ΔC plasmid cleaved with SalI and KpnI. For GFP-SAC1ΔC-PHFAPP1, the PI(4)P-binding PH domain of FAPP1 (Dowler et al., 2000 ) was amplified with oligos that added a 5′ SalI site and a 3′ KpnI site; the resulting PCR fragment was then digested and ligated into the GFP-SAC1ΔC plasmid as described above. The same strategy was used to generate the GFP-SAC1ΔC-FYVEEEA1 plasmid. LEU2-marked versions of the GFP-SAC1ΔC-FYVEEEA1 and GFP-SAC1ΔC-2 × PHPLCd were generated by subcloning the NaeI-SpeI fragment from these plasmids into plasmid pRS425.
The phosphatase inactive GFP-Sac1ΔC-FYVEEEA1 chimera was generated by QuikChange site-directed mutagenesis (Stratagene), substituting the active site cysteine residue for serine (GFP-Sac1ΔC(C392S)-FYVEEEA1) (Guo et al., 1999 ). The subsequent plasmid (pBP33) was confirmed by DNA sequence analysis, and impairment of catalytic activity was confirmed by analysis of in vivo phosphoinositide metabolism of yeast cells harboring pBP33. A strain expressing a temperature conditional allele of YMR1 was generated in the following manner. YMR1 sequences from plasmid pBP03 were amplified by PCR under error-prone conditions as described previously (Muhlrad et al., 1992 ). The mutagenized PCR fragment was then cotransformed with SphI-PacI-gapped pBP02 into BPY13 cells. Leu+ prototrophic transformants were then selected and screened for their ability to grow on 5-FOA at 26°C, but not at 38°C. From >6000 transformants, 10 putative ymr1ts plasmids were recovered into Escherichia coli and retransformed into BPY13 to confirm the temperature conditional growth phenotype on 5-FOA. A strain displaying temperature conditional lethality at 34°C was selected for further studies (YTS1).
In vivo phosphoinositide analysis was done essentially as described previously (Rudge et al., 2004 ), except that cells were labeled for 1 h after a 20-min preincubation at the appropriate temperature. Processed samples were resuspended in sterile water, and 10.5 × 106 cpm of deacylated [3H]glycero-phosphoinositols were analyzed.
Cell labeling and immunoprecipitations were performed as described previously (Gaynor et al., 1994 ) with minor modifications. Briefly, mid-log phase cultures were concentrated to 1–2 OD600 units/ml and labeled with Tran 35S label (PerkinElmer Life and Analytical Sciences, Boston, MA) for 10 min in YNB containing 100 μg/ml bovine serum albumin. Cells were then chased with 5 mM methionine, 2 mM cysteine, and 0.2% yeast extract for the indicated times; proteins were precipitated with 9% trichloroacetic acid. Where required, cells were preincubated at the appropriate temperature for 20 min before the addition of label. To assay for secreted carboxypeptidase Y (CPY), cells were converted to spheroplasts after pulse-chase by adding an ice-cold 2× buffer containing 50 mM Tris (pH 7.5), 2 M sorbitol, 40 mM NaF, 40 mM NaN3, and 10 mM dithiothreitol for 10 min. Zymolyase T100 (15 μg/OD600 unit; Seikagaku Kogyo, Tokyo, Japan) was then added to the cell suspension and incubated at 30°C for 30 min. Cells were then centrifuged at 6000 rpm for 5 min, and the supernatants were harvested away from the cell pellets and both were precipitated with 9% trichloroacetic acid. In all cases, trichloroacetic acid-precipitated protein pellets were washed twice with ice cold acetone, dried in vacuo, and processed for immunoprecipitation as described previously (Gaynor et al., 1994 ). Immunoprecipitated proteins were resuspended in sample buffer for resolution by SDS-PAGE, and gels were developed by autoradiography.
Labeling with N-[3-triethylammoniumpropyl]-4-[p-diethylaminophenyl-hexatrienyl] pyridinium dibromide (FM4-64); (Molecular Probes, Eugene, OR) was done essentially as described previously (Vida and Emr, 1995 ). All fluorescent images were obtained using an Axiovert S1002TV inverted fluorescent microscope (Carl Zeiss, Thornwood, NY) and processed using a Delta Vision deconvolution system (Applied Precision, Seattle, WA) and Photoshop 5.5 software (Adobe Systems, Mountain View, CA). All observations are based on the examination of at least 100 cells, and representative fields are shown. In all cases, the relative magnification is the same within a given figure.
The yeast YMR1 gene product is an ortholog of the myotubularin family of PI(3)P phosphatases (Figure 1A) (Laporte et al., 1998 ). As an in vitro PI(3)P-specific phosphatase (Taylor et al., 2000 ), we reasoned that Ymr1p might function in the regulation of PI(3)P signaling in vivo. To test this, we first assayed PI(3)P levels in ymr1Δ cells and in cells overexpressing YMR1 from a 2μ LEU2-marked plasmid. We found that total cellular levels of PI(3)P were relatively unchanged in ymr1Δ cells compared with wild type; however, there was a subtle increase in PI(3,5)P2 and PI(4,5)P2 levels (Figure 1B). In contrast to the ymr1Δ mutant, overexpression of YMR1 led to an ~25% decrease in PI(3)P levels compared with cells harboring an empty vector (Figure 1B), whereas other phosphoinositides remained relatively unaffected. These results were consistent with Ymr1p having specific activity against PI(3)P in vivo, as was reported for recombinant Ymr1p in vitro (Taylor et al., 2000 ). Next, we examined vacuolar protein sorting by [35S]methionine pulse-chase immunoprecipitation analysis in ymr1Δ cells and in YMR1-overexpressing cells by assaying the maturation of the vacuolar hydrolase CPY, which is delivered to the vacuole via a PI(3)P-dependent pathway (Schu et al., 1993 ). Compared with wild-type cultures, the ymr1Δ mutant had no obvious defects in the maturation of CPY (Figure 1C). In contrast, overexpression of YMR1 in wild-type cells caused a partial defect in CPY sorting in that roughly 15–20% was secreted in the Golgi-modified P2 precursor form (Figure 1C). These results suggested that YMR1 overexpression depletes a pool of PI(3)P that is important for regulating CPY sorting from the Golgi to the vacuole and implied a role for Ymr1p in the regulation of PI(3)P-dependent protein sorting pathways in vivo.
Previous studies on yeast phosphoinositide phosphatase mutants have indicated that Sac1p domain-containing proteins can provide redundant functions (Foti et al., 2001 ; Gary et al., 2002 ). Therefore, we tested whether Sac1p, Fig4p, Sjl1p, Sjl2p, or Sjl3p could compensate for Ymr1p function by generating a set of double mutants that were deleted of each corresponding open reading frame in combination with ymr1Δ. These double phosphatase mutant cells were labeled with myo-[2-3H]inositol, and the resulting deacylated glycero-phosphoinositides were analyzed by high-performance liquid chromatography (HPLC). We found that only ymr1Δ sjl3Δ double mutant cells displayed a greater than twofold elevation in PI(3)P levels over that of either single mutant (Figure 2A and Table 2). These findings indicated that Ymr1p and Sjl3p cooperate in the regulation of PI(3)P levels in yeast.
We next determined the intracellular distribution of PI(3)P in ymr1Δ sjl3Δ double mutant cells by using a PI(3)P-specific fluorescent reporter, GFP-FYVEEEA1 (Burd and Emr, 1998 ). Wild-type cells expressing this reporter exhibited a primarily punctate perivacuolar staining pattern representative of endosomal compartments (Figure 2B) (Burd and Emr, 1998 ). In contrast, the ymr1Δ sjl3Δ mutant displayed a dramatic enrichment of the green fluorescent protein (GFP) signal on vacuolar membranes, indicating that at steady state, PI(3)P accumulated on this organelle (Figure 2B). The vacuolar membrane staining by the GFP-FYVEEEA1 chimera in ymr1Δ sjl3Δ cells was accompanied by a reduction in perivacuolar punctate staining (Figure 2B). This could suggest that PI(3)P is depleted from endosomes in ymr1Δ sjl3Δ cells or that perhaps the integrity of these compartments is compromised in the double mutant. Interestingly, although PI(3)P accumulated on the vacuole where the PI(3)P 5-kinase Fab1p is thought to be active (Odorizzi et al., 1998 ; Rudge et al., 2004 ), cellular PI(3,5)P2 levels remained relatively unaffected in the ymr1Δ sjl3Δ mutant (Figure 2A and Table 2). This suggested that elevation in local PI(3)P concentration was not sufficient to drive Fab1p activity or perhaps that enhanced PI(3)P 5-kinase activity is met with elevated activity of a PI(3,5)P2 phosphatase. Consistent with this, Fab1p has been found to be tightly regulated by several accessory factors such as Vac7p, Vac14p, and the Sac1p-domain–containing phosphatase Fig4p (Bonangelino et al., 2002 ; Dove et al., 2002 ; Gary et al., 2002 ; Rudge et al., 2004 ).
During the course of these PI(3)P localization studies, it became apparent that the misregulation of PI(3)P in ymr1Δ sjl3Δ mutant cells was associated with a fragmented vacuole phenotype not unlike that observed in class B vps mutants (Horazdovsky et al., 1997 ). Therefore, ymr1Δ sjl3Δ mutant cells were labeled with FM4-64, a lipophilic dye that is endocytosed and trafficked to the vacuolar membrane (Vida and Emr, 1995 ), and observed by fluorescence microscopy. Compared with wild-type cultures labeled under the same conditions, the ymr1Δ sjl3Δ mutant displayed a multiple or fragmented vacuole phenotype (Figure 2B; see below), suggesting a role for Ymr1p and Sjl3p in the maintenance of vacuolar homeostasis.
We next examined whether PI(3)P regulation by Ymr1p and Sjl3p was important for directing vesicle-mediated protein trafficking in the endosomal system. ymr1Δ sjl3Δ double mutant cells were pulse labeled with [35S]methionine and assayed for their ability to sort and mature the vacuolar hydrolases CPY and carboxypeptidase S (CPS). Interestingly, the ymr1Δ sjl3Δ mutant displayed a significant defect in CPY sorting in that roughly 30–40% was secreted in the Golgi-modified P2 form (Figure 3A). In addition to missorting of CPY, there was a delay in the processing of CPS (Figure 3A) that correlated to a defect in multivesicular body (MVB) sorting (see below). Therefore, to gain insight into the mechanisms underlying these vacuole protein sorting defects in the ymr1D sjl3Δ double mutant, the localization and function of several PI(3)P-specific effectors that play important roles in directing endosomal sorting were examined.
The retromer complex functions in the recycling of the CPY receptor Vps10p from endosomes back to the Golgi (Cereghino et al., 1995 ; Cooper and Stevens, 1996 ; Seaman et al., 1998 ). Two retromer complex subunits, Vps5p and Vps17p, contain PX domains and are PI(3)P-specific effectors (Burda et al., 2002 ). Therefore, to test whether the CPY sorting defect in the ymr1Δ sjl3Δ mutant could be due at least in part to the misregulation of these Vps10p sorting factors, the subcellular distribution of Vps5-GFP and Vps17-GFP was determined. Wild-type cells expressing Vps17-GFP displayed a perivacuolar punctate staining pattern consistent with previously published results (Burda et al., 2002 ). In contrast, in ymr1Δ sjl3Δ double mutant cells, Vps17-GFP was diffuse in the cytoplasm with a partial enrichment on FM4-64–positive fragmented vacuoles (Figure 3B). Similar results also were obtained from cells expressing a Vps5-GFP fusion (our unpublished data). Thus, accumulation of PI(3)P in ymr1Δ sjl3Δ cells likely caused the mislocalization of these retromer complex components. Because mutants defective for Vps5p or Vps17p function also displayed a fragmented vacuole phenotype (Horazdovsky et al., 1997 ), it remains tempting to speculate that defects in their regulation might in part account for vacuolar morphology defects observed in the ymr1Δ sjl3Δ double mutant.
To investigate whether the mislocalization of Vps5p and Vps17p results in Vps10p trafficking defects, we expressed a chromosomally tagged VPS10-GFP chimera (Burda et al., 2002 ) in ymr1Δ sjl3Δ mutant cells to visualize the subcellular localization of Vps10p. Consistent with previously published results (Burda et al., 2002 ), wild-type cells displayed a punctate cytoplasmic staining pattern typical of Golgi and/or endosomal structures (Figure 3C). The ymr1Δ sjl3Δ mutant also displayed a cytoplasmic punctate staining pattern; however, these Vps10-GFP–containing structures were qualitatively different from the wild type in that they typically seemed larger, and many colocalized with FM4-64 on vacuolar compartments (Figure 3C). Although these results indicated that Vps10-GFP was mislocalized in the ymr1Δ sjl3Δ mutant, this staining pattern does not phenocopy retromer loss of function mutants, in which case Vps10-GFP is found almost exclusively on the vacuole membrane (Seaman et al., 1997 ; Burda et al., 2002 ). Consistent with an incomplete defect in CPY sorting in the ymr1Δ sjl3Δ double mutant, this would suggest that retromer function was only partially compromised in these cells.
Proper sorting of Vps10p in the endosomal system is required to maintain its stability and function (reviewed in Pfeffer, 2001 ). Therefore, we next examined the stability of Vps10p by [35S]methionine pulse-chase immunoprecipitation assays as an alternative means for examining endosomal sorting function in ymr1Δ sjl3Δ mutant cells. As expected from previous work (Cereghino et al., 1995 ), in wild-type cells Vps10p remained stable throughout the 2-h time course (Figure 3D). In addition to the full-length protein observed in wild-type cells, after 30 min, the ymr1Δ sjl3Δ mutant also displayed a Vps10p breakdown product (Figure 3D). This roughly 20-kDa shift in mobility was consistent with Vps10p being clipped from its transmembrane anchor and cytoplasmic domain. In addition to clipping, there was nearly a complete loss of Vps10p by the end of the time course in ymr1Δ sjl3Δ cells, indicating a strong defect in endosome function (Figure 3D). This defect in Vps10p stability correlated with the CPY sorting defect, and was a specific property of the ymr1Δ sjl3Δ double mutant, because Vps10p stability was not altered in the ymr1Δ or the sjl3Δ single mutants (Figure 3D). Defects in Vps10p stability are typical of class E vps mutants that have altered endosome morphology and function (Cereghino et al., 1995 ). Although ymr1Δ sjl3Δ cells do not exhibit an obvious class E compartment, it is possible that instability of Vps10p results from a combination of partial defects in retromer function and in the MVB sorting pathway, which is ablated in class E vps mutants (reviewed in Katzmann et al., 2002 ). Therefore, we next determined whether MVB sorting was affected by the accumulation of PI(3)P in the ymr1Δ sjl3Δ mutant.
The MVB sorting pathway depends on the proper localization and function of Vps27p. Vps27p is recruited to the endosome via its FYVE domain, which binds to PI(3)P. Subsequent to PI(3)P binding, Vps27p recruits/activates the ESCRT (for endosomal sorting complex required for transport) machinery required for the selection of MVB cargoes (Katzmann et al., 2003 ). In wild-type cells, this MVB sorting process occurs at the endosome where PI(3)P is enriched. Mutants that are blocked in the MVB sorting pathway typically accumulate an aberrant endosome-like class E compartment and have severe defects in endosome function (Rieder et al., 1996 ; Babst et al., 1997 ). Therefore, we expressed GFP fusions of Vps27p (Katzmann et al., 2003 ) or the MVB cargo protein CPS (Odorizzi et al., 1998 ) in ymr1Δ sjl3Δ cells. Consistent with previously published results, wild-type cells localized GFP-Vps27 to punctate perivacuolar endosomal structures (Figure 4A) (Katzmann et al., 2003 ). In ymr1Δ sjl3Δ cells, there was a redistribution of GFP-Vps27 to the surface of the vacuole (Figure 4A). The mislocalization of Vps27p correlated with a defect in MVB sorting of GFP-CPS in ymr1Δ sjl3Δ cells. Wild-type cells properly sorted GFP-CPS to the lumen of the vacuole; however, ymr1Δ sjl3Δ cells showed enrichment of GFP-CPS on the limiting membrane of the vacuole (Figure 4B). Thus, the delay in CPS maturation that was observed in ymr1Δ sjl3Δ cells (Figure 3A) was most likely due to the mislocalization of Vps27p and subsequent defects in MVB sorting.
The CVT pathway in yeast is also PI(3)P-dependent (Kametaka et al., 1998 ) and uses the PX domain-containing sorting nexins Atg24p (Cvt13p, Snx4p) and Atg20p (Cvt20p, Snx42p) (Nice et al., 2002 ). Therefore, we determined whether regulation of the CVT pathway was affected in the ymr1Δ sjl3Δ double mutant by assaying the maturation of the CVT cargo protein aminopeptidase 1 (Ape1p) by [35S]methionine pulse-chase immunoprecipitation analysis. Compared with wild-type cells, the ymr1Δ sjl3Δ mutant displayed a strong defect in Ape1p processing, with the precursor form remaining stable throughout the course of the experiment (Figure 5A). We next generated a chromosomal Atg24-GFP fusion to test whether Atg24p was mislocalized in the ymr1Δ sjl3Δ double mutant. In wild-type cells, Atg24-GFP was found on punctate structures (Figure 5B) that are thought to represent preautophagosomal structures (PAS) (Nice et al., 2002 ). In contrast to this, the ymr1Δ sjl3Δ double mutant redistributed Atg24-GFP to the cytoplasm (Figure 5B). Because this could result from a defect in PAS formation, the subcellular localization of two PAS markers, Cvt9 (Atg11)-GFP and Aut7 (Atg8)-GFP, also was analyzed (Kim et al., 2001 ). The ymr1Δ sjl3Δ mutant displayed punctate staining for each PAS marker protein that was indistinguishable from the wild type (our unpublished data). Therefore, it is likely that Atg24p, in addition to binding of PI(3)P via its PX domain, requires other factors for stable membrane association that may not be present on the surface of vacuoles where PI(3)P accumulates in ymr1Δ sjl3Δ cells.
The Sac1p domain-containing phosphatases, in particular the synaptojanin-like proteins Sjl2p and Sjl3p, were previously shown to significantly overlap in substrate specificity (Guo et al., 1999 ) as well as cellular function (Stolz et al., 1998 ; Foti et al., 2001 ; Stefan et al., 2002 ). Therefore, to address a possible role for Sjl2p in the regulation of PI(3)P, we sought to construct a ymr1Δ sjl2Δ sjl3Δ triple mutant. Interestingly, this mutant combination yielded a synthetic lethal phenotype (Table 3). Because Sjl2p and Sjl3p both contain a 5-phosphatase activity as well as Sac1p domain activity (Stefan et al., 2002 ), we took advantage of this synthetic lethal phenotype to determine which activity of these phosphatases was required to support viability of the ymr1Δ sjl2Δ sjl3Δ triple mutant. To do this, a diploid strain that was homozygous for ymr1Δ and heterozygous for sjl2Δ and sjl3Δ was transformed with a URA3-marked plasmid carrying a temperature conditional sjl2ts-8 allele (Stefan et al., 2002 ) and subsequently sporulated. A ymr1Δ sjl2Δ sjl3Δ triple mutant harboring the sjl2ts-8 plasmid was isolated (ymr1Δ sjl2ts sjl3Δ) and transformed with LEU2-marked sjl2-C446S or sjl2-D850S plasmids, which impair the Sac1p domain activity or the 5-phosphatase activity of Sjl2p, respectively (Stefan et al., 2002 ). The sjl2-C446S plasmid failed to rescue lethality of the ymr1Δ sjl2ts sjl3Δ triple mutant at 38°C, whereas the sjl2-D850S plasmid fully restored growth to the strain (Figure 6), indicating that Sac1p domain activity was required to support viability. These results suggested that maintenance of PI(3)P regulation is an essential function that is disrupted in the ymr1Δ sjl2Δ sjl3Δ triple mutant.
To test this, we took advantage of the ymr1Δ sjl2ts sjl3Δ triple mutant and determined the effects of inactivating Sjl2p on cellular phosphoinositide levels. In ymr1Δ sjl2ts sjl3Δ cells at permissive temperature, there was a decrease in PI(4,5)P2 levels, whereas PI(3)P, PI(4)P, and PI(3,5)P2 levels remained nearly the same compared with the ymr1Δ sjl3Δ mutant [compare ymr1Δ sjl3Δ from Table 2 and ymr1Δ sjl2Δ sjl3Δ (sjl2-8ts) from Table 3]. On shift to the restrictive temperature, the ymr1Δ sjl2ts sjl3Δ mutant displayed further accumulation of 3′ phosphoinositides, including a roughly 20% increase in PI(3)P, and an eightfold increase in PI(3,5)P2 compared with the permissive temperature (Table 3). This indicated that Ymr1p, Sjl2p, and Sjl3p overlap in the regulation of PI(3)P. Together, these results strongly implicated the misregulation and/or accumulation of 3′ phosphoinositides in the synthetic lethality of the ymr1Δ sjl2Δ sjl3Δ mutant.
These results suggested that deficient phosphoinositide 3-phosphatase activity allows PI(3)P to become toxic in ymr1Δ sjl2Δ sjl3Δ triple mutant cells. To test this, we constructed several Sac1p domain chimeras in which the phosphatase domain of Sac1p was fused to various lipid targeting domains that have known phosphoinositide binding specificities (see below). To facilitate the detection of these constructs in vivo, each was constructed as an N-terminal GFP fusion. By this approach cytoplasmic Sac1p domain activity could be targeted to PI(4,5)P2 on the plasma membrane (GFP-Sac1ΔC-2 × PHPLCγ), to PI(4)P on Golgi compartments (GFP-Sac1ΔC-PHFAPP1), to PI(3)P on endosomes and the vacuole (GFP-Sac1ΔC-FYVEEEA1), or left diffusely throughout the cytoplasm (GFP-Sac1ΔC) (Figure 7) (Foti et al., 2001 ; Stefan et al., 2002 ). Each of these chimeras localized to the intended target membrane by fluorescence microscopy, and Western blot analysis using GFP-specific antibodies (Santa Cruz Biotechnology, Santa Cruz, CA) showed no obvious differences in the expression levels or stability of these chimeras (Figure 7; our unpublished data). To facilitate the detection of subtle effects of these Sac1p domain chimeras on the growth properties of the ymr1Δ sjl2Δ sjl3Δ triple mutant, we generated a ymr1ts allele that could not support viability above 34°C (see MATERIALS AND METHODS; our unpublished data). Next, each Sac1p domain chimera was expressed in ymr1ts sjl2Δ sjl3Δ triple mutant cells, and its ability to promote viability at the restrictive temperature was determined. Interestingly, only the PI(3)P-targeted GFP-Sac1ΔC-FYVEEEA1 fusion was able to promote growth at restrictive temperature (Figure 8A). To confirm that rescue of the temperature conditional lethality was dependent on phosphatase activity, the conserved active site cysteine residue of the GFP-Sac1ΔC-FYVEEEA1 chimera was mutated to serine; a substitution that ablates the activity of Sac1p (Guo et al., 1999 ). Although this mutant version, GFP-Sac1ΔCC392S-FYVEEEA1 was expressed and localized comparably to the phosphatase-active chimera, it failed to rescue the ymr1ts sjl2Δ sjl3Δ mutant, even at 34°C (Figure 8A; our unpublished data). This indicated that phosphatase activity of the PI(3)P-targeted chimera was required to promote growth of the mutant at restrictive temperature. We next tested whether the GFP-Sac1ΔC-FYVE chimera could complement the synthetic lethality of the ymr1Δ sjl2Δ sjl3Δ triple mutant. To do this, a LEU2-marked version of the GFP-SAC1ΔC-FYVEEEA1 plasmid was constructed and transformed into ymr1Δ sjl2ts sjl3Δ cells and their ability to grow on media containing 5-FOA, which selects against cells containing the URA3-marked sjl2ts plasmid, was determined. Interestingly, cells containing the GFP-Sac1ΔC-FYVEEEA1 chimera grew on 5-FOA, whereas control cells with an empty vector or the plasma membrane targeted GFP-Sac1ΔC-PHPLCδ chimera did not (Figure 8B; our unpublished data). The ability of these cells to grow on 5-FOA indicated that the GFP-Sac1ΔC-FYVEEEA1 chimera was sufficient to support viability of the ymr1Δ sjl2Δ sjl3Δ mutant.
Analysis of in vivo phosphoinositides in ymr1ts sjl2Δ sjl3Δ cells expressing the GFP-Sac1ΔC-FYVEEEA1 chimera at 38°C revealed that the most striking differences, compared with cells harboring an empty vector, were a roughly 70% decrease in PI(3)P levels and a roughly 60% decrease in PI(3,5)P2 levels, rendering the levels of both of these 3′ phosphoinositides comparable with YMR1-rescued control cells (Figure 8C). Collectively, these results argued that phosphatase-mediated PI(3)P regulation was an essential function, most likely at post-Golgi compartments, and supported the idea that 3′ phosphoinositide accumulation is toxic to cells, resulting in the synthetic lethal phenotype of the ymr1Δ sjl2Δ sjl3Δ mutant.
In this study, we have elucidated a role for the yeast myotubularin-related phosphatase Ymr1p in the regulation of PI(3)P-mediated vacuolar protein sorting. Our data argue that Ymr1p is a PI(3)P-specific phosphatase in vivo, and in conjunction with the synaptojanin-like phosphatase Sjl3p, controls the cellular levels and the subcellular distribution of PI(3)P. The misregulation of PI(3)P in ymr1Δ sjl3Δ cells impacts the localization and function of several PI(3)P effectors that play key roles in cargo sorting and vesicle trafficking in the endosomal system, which culminates in the dysfunction of endosomes as prevacuolar sorting compartments. Moreover, further deletion of SJL2 in the ymr1Δ sjl3Δ double mutant yields a synthetic lethal phenotype that directly correlates with further defects in the regulation of 3′ phosphoinositides. Together, this work provides the first experimental evidence that although PI(3)P is not an essential phosphoinositide (Schu et al., 1993 ), the accumulation and loss of spatial control of PI(3)P lead to membrane trafficking defects and ultimately to a toxic effect that results in the loss of viability.
The myotubularin family constitutes a large group of evolutionarily conserved phosphatases within the tyrosine/dual-specificity phosphatase superfamily (reviewed in Laporte et al., 2003 ). The catalytically active human myotubularins that have been tested (Mtm1, Mtmr2, Mtmr3, Mtmr4, Mtmr6, and Mtmr7) are all specific 3-phosphatases for PI(3)P and PI(3,5)P2 in vitro, generating PI and PI(5)P, respectively (Taylor et al., 2000 ; Walker et al., 2001 ; Tronchere et al., 2003 ), of which PI(5)P seems to be an allosteric activator of phosphatase activity (Schaletzky et al., 2003 ). In contrast to these mammalian myotubularins, our data suggest that Ymr1p, the sole myotubularin member in yeast, does not use PI(3,5)P2 as a substrate in vivo. This is supported by the observations that overexpression of YMR1 does not decrease the level of PI(3,5)P2 in otherwise wild-type cells (Figure 1C), even upon hyperosmotic stress (Parrish and Emr, unpublished results), which stimulates PI(3,5)P2 synthesis (Cooke et al., 1998 ). In addition, PI(5)P, the product of PI(3,5)P2 catalysis by myotubularins (Schaletzky et al., 2003 ; Tronchere et al., 2003 ), is not among the normal complement of phosphoinositides identified in yeast and was undetectable in all of our HPLC analyses.
The finding that deletion of YMR1 did not affect vacuole protein sorting (Figure 1C; Taylor et al., 2000 ) led us to predict that Sac1p domain-containing lipid phosphatases could compensate for the activity of Ymr1p in the context of the ymr1Δ mutant. This was indeed the case, because the deletion of SJL3 in the ymr1Δ mutant caused a roughly 2.5-fold increase in cellular PI(3)P levels and the redistribution of this lipid onto the surface of vacuoles. Consistent with Ymr1p and Sjl3p functioning in a similar pathway, we found that like YMR1, overexpression of SJL3 also significantly lowered PI(3)P levels and conferred defects in CPY sorting in otherwise wild-type cells (our unpublished data). Defects in PI(3)P regulation correlated with endosome dysfunction, fragmented vacuole morphology, and defects in PI(3)P-mediated vacuole protein sorting pathways in the ymr1Δ sjl3Δ double mutant. These phenotypes were apparently due to pleiotropic defects in the regulation of PI(3)P-specific effectors that carry out cargo sorting and vesicle trafficking. Together, these results indicate that PI(3)P-directed phosphatase activity of Ymr1p and Sjl3p maintains endosome function through the control of several key PI(3)P effectors. For Sjl3p, this function in PI(3)P metabolism and endosome maintenance may be in addition to its previously characterized role in the regulation of a clathrin/AP1 mediated “slow delivery” pathway of cargoes between the trans-Golgi and early endosomes, a function that requires both catalytic activities of Sjl3p (Ha et al., 2003 ).
Sjl2p and Sjl3p each possess two distinct phosphoinositide phosphatase activities, a Sac1p domain and a PI(4,5)P2 5-phosphatase domain, which renders them capable of converting all common phosphoinositides in yeast to PI (reviewed in Hughes et al., 2000 ). Although each likely has a primary role in the regulation of distinct phosphoinositide signaling pathways, it is not surprising that Sjl2p and Sjl3p significantly overlap in function with each other, and with other lipid phosphatases that are dedicated to the hydrolysis of specific phosphoinositides. Consistent with this, sjl2Δ sjl3Δ double mutants are viable and demonstrate only modest phenotypes, including slowed growth rate, cell wall thickening, and defects in vacuole morphology and actin cytoskeleton organization (Srinivasan et al., 1997 ; Stolz et al., 1998 ). Despite these phenotypes, sjl2Δ sjl3Δ double mutants do not display significant increases in cellular levels of PI(3)P, PI(4)P, or PI(4,5)P2 (Srinivasan et al., 1997 ; Stolz et al., 1998 ) and show only a small increase in PI(3,5)P2 (Guo et al., 1999 ), indicating that other phosphatases are sufficient to maintain the normal levels of these phosphoinositides. For example, in sjl2Δ sjl3Δ cells, further deletion of SAC1 or SJL1 results in lethality, which is attributable to defects in PI(4)P regulation (Foti et al., 2001 ) or PI(4,5)P2 regulation (Srinivasan et al., 1997 ; Stolz et al., 1998 ; Stefan et al., 2002 ), respectively.
Our work presented here reveals a further essential requirement for phosphatase-mediated PI(3)P regulation through the elucidation of Ymr1p as a PI(3)P-specific phosphatase that supports viability in sjl2Δ sjl3Δ double mutant cells. Several observations are consistent with an essential overlapping role for Ymr1p, Sjl2p, and Sjl3p in PI(3)P regulation. First, Ymr1p is apparently sufficient to control PI(3)P signaling and promote proper endosomal sorting. Second, further elimination of the Sac1p domain activity of Sjl2p in ymr1Δ sjl3Δ cells results in lethality. Third, our studies on the ymr1Δ sjl3Δ double mutant combined with the analysis of phosphoinositide metabolism in ymr1Δ sjl2ts sjl3Δ cells suggests that synthetic lethality is most likely due to exacerbated defects in 3′ phosphoinositide regulation. Finally, soluble Sac1p domain activity was able to rescue ymr1Δ sjl2Δ sjl3Δ cells, but only if it was targeted to PI(3)P.
An intriguing question that arises from this study revolves around the essential process(es) that is disrupted upon the loss of phosphatase-mediated PI(3)P regulation. Because PI(3)P is not an essential phosphoinositide (Schu et al., 1993 ), it seems reasonable that 3′ phosphoinositide accumulation might exert toxic effects by simultaneously interfering with the regulation of multiple phosphoinositide-dependent signaling pathways, potentially including those that regulate actin organization (reviewed in Yin and Janmey, 2003 ), cytokinesis (Casamayor and Snyder, 2003 ), or cell growth and survival (reviewed in Heinisch et al., 1999 ), culminating in the inability of ymr1Δ sjl2Δ sjl3Δ cells to grow. Accordingly, further work will be required to determine the extent to which phosphatase-mediated regulation of 3′ phosphoinositides and the integrity of the endosomal sorting system impacts the control of other phosphoinositide signaling systems.
In yeast, previous studies have indicated essential roles for the regulation of PI(4)P (Foti et al., 2001 ) and PI(4,5)P2 (Stolz et al., 1998 ; Stefan et al., 2002 ). In this study, we have elucidated an essential role for the regulation of cellular levels and the subcellular distribution of PI(3)P. To our knowledge, these findings represent the first clear demonstration that a myotubularin family member regulates PI(3)P-mediated endosomal sorting events. Thus, as has been speculated (Laporte et al., 2003 ), it is likely that myotubularins antagonize type III PI 3-kinase (hVps34) signaling in higher eukaryotes as well. This may have important implications for the ontogeny of severe human genetic diseases such as myotubular myopathy and Marie-Charcot-Tooth neuropathy, which are caused by mutations in myotubularin family members.
We are indebted to Steven Padilla, Perla Arcaira, Joshua Kaufman, Jane Atienza, and James Cregg for excellent technical support through the course of the study. We thank Drs. James Konopka, Jennifer Lin, Anne Savitt, and Greg Taylor for advice and critical comments on the work. We also thank members of the Emr laboratory, both past and present, especially Drs. Patrice Burda, Srimonti Sarkar, Simon Rudge, Vicki Sciorra, Takeshi Noda, and Mitsuaki Tabuchi, for generosity with strains and reagents and invaluable discussions and suggestions regarding the manuscript. C.J.S. is a postdoctoral research associate of the Howard Hughes Medical Institute. W.R.P. is supported by National Research Service Award postdoctoral research training grant 5 T32 AI07036-25 from the National Institutes of Health/National Institute of Allergy and Infectious Diseases. S.D.E. is an investigator of the Howard Hughes Medical Institute.
Article published online ahead of print. Mol. Biol. Cell 10.1091/mbc.E04-03-0209. Article and publication date are available at www.molbiolcell.org/cgi/doi/10.1091/mbc.E04-03-0209.
Abbreviations used: Ape1, aminopeptidase 1; CPS, carboxypeptidase S; CPY, carboxypeptidase Y; CVT, cytoplasm-to-vacuole; 5-FOA, 5-fluorooroic acid; GFP, green fluorescent protein; HPLC, high performance liquid chromatography; MVB, multivesicular body; PCR, polymerase chain reaction; PI, phosphatidylinositol; PI(3)P, phosphatidylinositol 3-phosphate; PI(4)P, phosphatidylinositol 4-phosphate; PI(3,5)P2, phosphatidylinositol 3,5-bisphosphate; PI(4,5)P2, phosphatidylinositol 4,5-bisphosphate; vps, vacuole protein sorting.