Complementation Analysis in PtdInsPKinase-deficient Yeast Mutants Demonstrates ThatSchizosaccharomyces pombe and Murine Fab1p Homologues Are Phosphatidylinositol 3-Phosphate 5-Kinases*

Phosphatidylinositol 3,5-bisphosphate (PtdIns(3,5)P 2) is widespread in eukaryotic cells. In Saccharomyces cerevisiae,PtdIns(3,5)P 2 synthesis is catalyzed by the PtdIns3P 5-kinase Fab1p, and loss of this activity results in vacuolar morphological defects, indicating that PtdIns(3,5)P 2 is essential for vacuole homeostasis. We have therefore suggested that all Fab1p homologues may be PtdIns3P 5-kinases involved in membrane trafficking. It is unclear which phosphatidylinositol phosphate kinases (PIPkins) are responsible for PtdIns(3,5)P 2synthesis in higher eukaryotes. To clarify how PtdIns(3,5)P 2 is synthesized in mammalian and other cells, we determined whether yeast and mammalian Fab1p homologues or mammalian Type I PIPkins (PtdIns4P 5-kinases) make PtdIns(3,5)P 2 in vivo. The recently cloned murine (p235) and Schizosaccharomyces pombe FAB1homologues both restored basal PtdIns(3,5)P 2synthesis in Δfab1 cells and made PtdIns(3,5)P 2 in vitro. Only p235 corrected the growth and vacuolar defects of fab1 S. cerevisiae. A mammalian Type I PIPkin supported no PtdIns(3,5)P 2 synthesis. Thus, FAB1and its homologues constitute a distinct class of Type III PIPkins dedicated to PtdIns(3,5)P 2 synthesis. The differential abilities of p235 and of SpFab1p to complement the phenotypic defects of Δfab1 cells suggests that interaction(s) with other protein factors may be important for spatial and/or temporal regulation of PtdIns(3,5)P 2synthesis. These results also suggest that p235 may regulate a step in membrane trafficking in mammalian cells that is analogous to its function in yeast.


Phosphatidylinositol 3,5-bisphosphate (PtdIns(3,5)P 2 )
is widespread in eukaryotic cells. In Saccharomyces cerevisiae, PtdIns(3,5)P 2 synthesis is catalyzed by the PtdIns3P 5-kinase Fab1p, and loss of this activity results in vacuolar morphological defects, indicating that PtdIns(3,5)P 2 is essential for vacuole homeostasis. We have therefore suggested that all Fab1p homologues may be PtdIns3P 5-kinases involved in membrane trafficking. It is unclear which phosphatidylinositol phosphate kinases (PIPkins) are responsible for PtdIns(3,5)P 2 synthesis in higher eukaryotes. To clarify how PtdIns(3,5)P 2 is synthesized in mammalian and other cells, we determined whether yeast and mammalian Fab1p homologues or mammalian Type I PIPkins (PtdIns4P 5-kinases) make PtdIns(3,5)P 2 in vivo. The recently cloned murine (p235) and Schizosaccharomyces pombe FAB1 homologues both restored basal PtdIns(3,5)P 2 synthesis in ⌬fab1 cells and made PtdIns(3,5)P 2 in vitro. Only p235 corrected the growth and vacuolar defects of fab1 S. cerevisiae. A mammalian Type I PIPkin supported no PtdIns(3,5)P 2 synthesis. Thus, FAB1 and its homologues constitute a distinct class of Type III PIPkins dedicated to PtdIns(3,5)P 2 synthesis. The differential abilities of p235 and of SpFab1p to complement the phenotypic defects of ⌬fab1 cells suggests that interaction(s) with other protein factors may be important for spatial and/or temporal regulation of PtdIns(3,5)P 2 synthesis. These results also suggest that p235 may regulate a step in membrane trafficking in mammalian cells that is analogous to its function in yeast.
In eukaryotes, phosphoinositides play important roles in sev-eral cell functions. In particular, they have been implicated in the membrane trafficking events by which membranes and proteins are sorted into vesicles and targeted to various cell compartments (1). Golgi-to-vacuole trafficking of membranes and proteins in Saccharomyces cerevisiae has provided a very informative system for the study of phosphoinositide-dependent membrane trafficking, because it is not essential for growth and so lends itself to mutational analysis. The phosphatidylinositol 3-phosphate (PtdIns3P) 1 that is made by the phosphatidylinositol kinase Vps34p is essential for the targeting of proteins from the Golgi to the yeast prevacuolar compartment and also for the prevacuolar compartment-to-vacuole trafficking step (2).
This functional sequence was recently extended by the demonstration that the yeast FAB1 gene, the protein product of which is necessary for membrane efflux from the vacuole, encodes a PtdIns3P 5-kinase (3,4). Phenotypes caused by mutations in FAB1 (temperature-sensitive growth and massive enlargement of the yeast vacuole, which also fails to acidify) had been identified earlier, but it had been predicted that the Fab1p protein encoded by FAB1 (which we shall term ScFab1p) would be a PtdIns4P 5-kinase (5). Because PtdIns(3,5)P 2 has been found in all eukaryote cells so far examined (6,7), we predicted that ScFab1p would be the first member of a novel family of PtdInsP kinases (PIPkins) that would be dedicated to PtdIns(3,5)P 2 synthesis: the PtdIns3P 5-kinases of other organisms would be their Fab1p homologues (3). We suggested that an appropriate general name for this group of enzymes would be "Type III PIPkins," to distinguish them from Type I PIPkins (which are PtdIns4P 5-kinases) and Type II PIPkins (which are PtdIns5P 4-kinases) (for a recent review, see Ref. 8).
Taken alone, however, such enzyme assays in vitro do not unambiguously define the biological substrate specificities of inositol lipid kinases. In order to determine what lipids the Fab1p-like kinases make in vivo, we therefore determined in what ways the expression of a mammalian Type I␤ PIPkin, the murine Fab1p-like protein p235, or the Schizosaccharomyces pombe Fab1p (SpFab1p) would change the endogenously synthesized phosphoinositide complements of yeasts carrying mutants of MSS4 (which encodes PtdIns4P 5-kinase) and of yeasts in which the FAB1 gene was knocked out. The conditionallethal mss4-1 mutants that we used have a reduced PtdIns(4,5)P 2 content (our data); such mutants show defects in their actin cytoskeleton (11,12), and ⌬fab1 cells make no PtdIns(3,5)P 2 (3,4). The utility of this type of approach has already been demonstrated by the finding that expression of a Type I PIPkin, but not of a Type II PIPkin, corrects the temperature-sensitive growth defect of mss4-1 mutant yeast (12 (13).
Strains and Plasmids-The plasmids and S. cerevisiae strains used in this study are listed in Table I. Rich YPD and inositol-free synthetic complete (SC) media were prepared according to Sherman (19). SC medium containing 2% (w/v) raffinose contained a full complement of amino acids (3). Strain SDY500 was constructed by mating the 616-3-2 (fab1-1) strain with SEY6210 and sporulating the resulting diploids to recover haploid segregants. The complete upstream and coding sequence of the ScFAB1 gene, as a NotI-XhoI restriction fragment, was excised from pEMY105 and ligated into the NotI-Xho I sites of the single-copy pRS416 vector. The resulting plasmid was designated pFABCEN. Plasmid pRM19 was constructed by ligating the entire SpFAB1 ORF (amplified using Pwo polymerase from genomic S. pombe DNA using primers that inserted BamHI restriction sites at the 5Ј and 3Ј ends) into the BglII site of pYO2144. The SpFAB1 ORF was cloned into the BamHI restriction site of pEG-KT to generate plasmid pRM22. The p235s splice variant, which lacks 11 amino acids from a region upstream of the FYVE domain (9) was used for all characterization studies. The 5Ј and 3Ј ends of the p235s ORF were polymerase chain reaction-tagged with XbaI and SalI restriction sites, respectively and ligated with XbaI-XhoI-restricted pVT102U to produce plasmid pRM24. For in vitro characterization, XbaI-SalI-tagged p235s was cloned into the XbaI and SalI sites of pEG-KT, producing pRM23. All polymerase chain reaction-amplified clones were confirmed by automated DNA sequencing.
Radiolabeling of Yeast Mutants-Labeling and analyses of yeast strains employed techniques described in detail by Dove and Michell (20). As appropriate, cells were subjected to hyperosmotic shock by the addition of 1.8 M NaCl to a final concentration of 0.9 M. For comparison between strains, the levels of radioactivity were corrected by standardizing the counts against glycerophosphoinositol (GroPIns), glycerophosphoinositol 3-phosphate (GroPIns3P), and glycerophosphoinositol 4-phosphate (GroPIns4P), as appropriate. Experiments reported (performed in duplicate) are representative of at least two independent experiments that gave similar results.
Fluorescent Labeling with FM4 -64 -Yeast cultures were grown for 24 -48 h in the appropriate selective medium and labeled with FM4 -64 according to Bonangelino et al. (22). Labeled cells were examined by epifluorescence using the rhodamine filter.
Lipid Kinase Assays-Phosphoinositide kinase assays on recombinant Fab1p, SpFab1p and p235 proteins were performed according to Cooke et al. (3) (and the linked Supplementary Information), except that GST fusion protein-loaded beads were prewashed only once with 1 ml of kinase buffer before assaying.
HPLC Resolution of GroPInsP n s-GroPInsP n s were resolved on a Partisphere 5 m SAX column (Whatman) using gradient 1 (23) or gradient 2 (24). For separating all known isomers of GroPInsP, the novel SD001 gradient based on Buffers A (water) and B (1.

RESULTS
Functional Complementation of ScFAB1 Inactivation-We tested whether expression of the presumptive murine Fab1plike protein (protein p235 of Shisheva et al. (9); termed PIKfyve in the Mouse Genome DataBase (MGI 1335106) (49)), which is the only mammalian Fab1p homologue so far reported, or of SpFab1p (the Fab1p homologue of S. pombe) would correct some or all of the defects caused by deletion or mutation of ScFAB1 in S. cerevisiae. In pFABCEN cells, which express a single copy of the ScFAB1 gene behind its own promoter,
Expression of p235 rescued several of the phenotypic defects of fab1 mutants. p235-expressing fab1-1 cells grew at the restrictive temperature (Fig. 2b), and p235 suppressed the vacuolar defects of ⌬fab1 cells to a substantial degree (Fig. 2a). Moreover, p235 restored the ability of ⌬fab1 cells, which make no PtdIns(3,5)P 2 (3), to make a basal concentration of PtdIns(3,5)P 2 similar to that present in wild-type cells (Fig. 1,  right; Fig. 2c, center panel). However, hyperosmotic stress (0.9 M NaCl for 1-10 min) provoked no increase in PtdIns(3,5)P 2 synthesis in the p235-expressing cells (Fig. 1, right; Fig. 2c, right panel): this type of stress provokes rapid PtdIns(3,5)P 2 synthesis in S. cerevisiae and S. pombe (6) and also in fab1 cells that overexpress exogenous ScFab1p (3). The negative result obtained with p235 is consistent with the fact that none of the mammalian cell lines we have tested have shown enhanced PtdIns(3,5)P 2 synthesis when hyperosmotically stressed (6). 2 Under some experimental conditions in vitro, p235 synthesizes a phosphoinositide that appears to be PtdIns5P (9,49). However, we never detected any PtdIns5P in yeast lipids that were analyzed by an HPLC method that resolves deacylated PtdIns3P, PtdIns4P and PtdIns5P, even in the p235-expressing cells (Fig. 2c, left panel). Moreover in vitro, recombinant p235 phosphorylated PtdIns3P much more readily than PtdIns or PtdIns4P under the experimental conditions under which we previously defined the PtdIns3P 5-kinase activity of ScFab1p (Fig. 3a). When PtdInsP 2 spots from such PtdIns3P kinase assays were deacylated, the resulting GroPInsP 2 co-chromatographed with GroPIns(3,5)P 2 , confirming that PtdIns(3,5)P 2 was synthesized (Fig. 3c). p235 preferentially synthesized PtdIns(3,5)P 2 , even when the only PtdIns3P present was a contaminant present in commercially available PtdIns and b, FAB1 homologues were tested for their ability to complement temperature-sensitive fab1-1 mutants at the restrictive temperature. Strains expressing p235 on a multicopy plasmid (pRM24) or FAB1 on a single-copy plasmid (pFABCEN) were streaked on SC-Ura plates and incubated at 37°C for 3 days; fab1-1 transformed with empty pRS416 and pVT102U plasmids served as controls. c, anion-exchange HPLC analysis of GroPInsP n s in a p235-expressing fab1::LEU2 strain. Cells were labeled at 24°C in inositol-free SC-Ura medium containing 10 Ci⅐ml Ϫ1 [ 3 H]inositol to 10 6 cells⅐ml Ϫ1 . As indicated, cells were subjected to hyperosmotic stress with 0.9 M NaCl for 10 min. Following lysis, lipids were deacylated and resolved by HPLC gradient 1 (as described in Ref. 23)  PtdIns4P preparations. We conclude that p235 serves solely as a PtdIns3P 5-kinase in vivo and that its main activity in kinase assays in vitro is the 5-phosphorylation of PtdIns3P.
Expression of SpFab1p also restored the ability of ⌬fab1 cells to make a basal concentration of PtdIns(3,5)P 2 similar to that present in wild-type cells (ϳ5% of the concentration of PtdIns(4,5)P 2 ) and about half of that present in pFABCEN cells (Fig. 1, right; Fig. 4a, left panel). SpFab1p-expressing cells showed a modest increase in PtdIns(3,5)P 2 synthesis in response to hyperosmotic stress (Fig. 4a, right panel). There was an approximately 2-fold elevation in PtdIns(3,5)P 2 concentration, as compared with a 14-fold increase in cells expressing ScFAB1 from a single-copy plasmid (Fig. 1, right). This behavior is similar to that of strains carrying the fab1-2 temperature-sensitive allele of ScFAB1, which at restrictive temperatures show only a 2-fold stimulation in PtdIns(3,5)P 2 in response to hyperosmotic shock (3). In contrast to the results with p235, expression of SpFab1p neither complemented the temperature-sensitive growth phenotype of fab1-1 cells (data not shown) nor corrected the vacuolar morphology of fab1::LEU2 cells (Fig. 4b). However, when recombinant SpFab1p was assayed for lipid kinase activity in vitro, it was an active PtdInsP kinase that showed a marked preference for PtdIns3P as a substrate (Fig. 3a). Deacylation of the product and co-chromatography with a GroPIns(3,5)P 2 standard confirmed that the PtdInsP 2 synthesized was PtdIns(3,5)P 2 (Fig. 3b).
PIPkin-I␤ Restores PtdIns(4,5)P 2 Synthesis in mss4 Mutants-We examined the effect of expressing a Type I PIPkin and the Fab1p proteins in strain YOC808, which is wild-type for the FAB1 gene but harbors the temperature-sensitive MSS4 allele mss4-1. YOC808 cells fail to grow at 37°C, at least partly because of defects in their actin cytoskeleton, and this fault is overcome by overexpression of a mammalian Type I PIPkin (12). To confirm that the mss4-1 lesion limits PtdIns-(4,5)P 2 synthesis, the PtdIns(4,5)P 2 content of YOC808 cells was analyzed at 23°C and after 2 h at the restrictive temperature (38°C). Even at the lower temperature, the mss4-1 cells contained about one-quarter of the [ 3 H]PtdIns(4,5)P 2 of wildtype cells (Fig. 5, a versus b; Fig. 6, left). When the temperature was raised, the wild-type cells increased their PtdIns(4,5)P 2 complement but PtdIns(4,5)P 2 declined further in the mss4-1 cells (Fig. 6, left). Expression of Type I␤ PIPkin in these cells restored their PtdIns(4,5)P 2 content to a level somewhat higher than that of wild-type cells, confirming that the Type I␤ PIPkin is an active PtdIns4P 5-kinase in yeast (Fig. 5c). Expression of a mammalian Type II PIPkin, described as a PtdIns5P 4-kinase (25), had no effect on the PtdIns(4,5)P 2 complement (not shown). We have never detected PtdIns5P in yeast (see above), so this was to be expected. Overexpression of ScFab1p, SpFab1p or p235 also had no major effect on the PtdIns(4,5)P 2 content of mss4-1 cells (Fig. 5, d-f, respectively; Fig. 6, left), indicating that these enzymes make no contribution to PtdIns(4,5)P 2 synthesis.

Type I PIPkins Do Not Make PtdIns(3,5)P 2 in Vivo and Do Not Phenotypically Rescue ⌬fab1
Cells-The fab1::LEU2 strain, which contains an inactivated FAB1 allele, has a wildtype MSS4 gene and hence a normal PtdIns(4,5)P 2 complement. However, it makes no PtdIns(3,5)P 2 either under basal conditions or when osmotically stressed (Fig. 1, right; Fig. 7a, left  panel). These cells have enlarged vacuoles that occupy much of the cell and fail to correctly acidify (3)(4)(5). When expressing a mammalian Type I␤ PIPkin, they still made no PtdIns(3,5)P 2 (even when hyperosmotically stressed; Fig. 1 and Fig. 7a, right  panel), and their vacuolar phenotype remained abnormal (Fig.  7b). Moreover, the Type I PIPkin did not correct the temperaturesensitive growth phenotype of fab1-1 cells (data not shown).

ScFab1p and Its Homologues Are Very Similar Both in Overall Organization and in Amino Acid Sequence, Thus Defining a Family of Closely Related Proteins-Having suggested that
ScFab1p and its orthologues in other organisms will make up a "Type III" family of PIPkins that specifically 5-phosphorylate PtdIns3P (3), we compared the molecular organizations and amino acid sequences of the available full-length ScFab1p-like proteins. The sequences of five close relatives of ScFab1p have been reported so far: from S. pombe (SpFab1p; the two overlapping clones AL023534 and AL021838), Caenorhabditis elegans (CeFab1p; AL023817), Drosophila melanogaster (DmFab1p; AL035311), mouse (p235; AF102777), and Arabidopsis thaliana (AtFab1p; AL035525). Fig. 8a compares the overall domain arrangements of these proteins, and Fig. 8, b and c, shows detailed alignments of their FYVE and kinase domains, respectively. All of the sequences include a C-terminal PtdInsP kinase domain, an N-terminal FYVE zinc finger domain, and a central domain with similarities to a conserved sequence motif present in Cct1p and its homologues (26): sequence conservation is most striking in the FYVE and kinase domains. FYVE domains, at least some of which specifically bind the ScFab1p substrate PtdIns3P (27)(28)(29)(30), contain the highly conserved signature motif (R/K)(R/K)H-HCR (see Ref. 31) (Fig. 8b, asterisks) within a region rich in Cys and His residues. The Cct1p-like domain resembles a domain present in chaperone complexes implicated in actin and tubulin folding (26). The kinase domains of all the Type III PIPkins identified so far include three invariant catalytic residues that are shared with protein kinases, phosphoinositide 3-kinases, and other PIPkins (32)(33)(34)(35)(36). In ScFab1p, these are Lys-2059, FIG. 3. Substrate specificities of Fab1p homologues. a, TLC analysis of the products of lipid kinase assays using GST-fusion proteins of Fab1p, SpFab1p, and p235 (as indicated above the autoradiograph). Assays were as described under "Experimental Procedures," using the substrates indicated below the autoradiogram. The positions of migration of known products of kinase assays using GST-Fab1p are shown on the right. b and c, confirmation that PtdIns(3,5)P 2 is the product of the kinase reaction when PtdIns3P is used as a substrate with SpFab1p (b) and with p235 (c). The spots corresponding to the major product of these reactions were recovered, deacylated, and analyzed as described previously (20). 32 P-Labeled products were co-chromatographed with 3 H-labeled standards (broken curves). The slowermigrating spot from the products of the assay of p235 kinase activity, with PtdIns as substrate, was also analyzed (not shown). It was PtdIns(3,5)P 2 , presumably formed from PtdIns3P, that contaminated the PtdIns substrate.
Asp-2196, and Asp-2216. In protein kinase A, the Lys residue equivalent to Lys-2059 of ScFab1p interacts with the ␣-phosphate group of ATP (37) and is essential for kinase activity (32). Substitution of Lys-2059 with Met abolishes ScFab1p PtdIns3P 5-kinase activity (data not shown).
Although ScFab1p was initially designated as a likely Type I PIPkin (38), the kinase domains of all of the Type III PIPkins proteins lack an "insert" region that is present in the Type I and Type II PIPkins (38,39). At least in the recombinant Type II PIPkin, this insert is structurally disordered (35). A second region of divergent sequence in the kinase domains of the PIPkins is again disordered in the Type II PIPkin (35). By analogy with protein kinases (32,40,41), this may be a substrate-binding activation loop that is involved in defining the substrate specificity of the Type II PIPkin (35). The relevant sequence in the Type III PIPkins (Thr-2200 to Gly-2238 in ScFab1p) is indicated by the plus signs in Fig. 8c. Alongside that region of the sequence comparison of the different Type III PIPkins, we also show the relevant parts of a mammalian Type I PIPkin, of Mss4p (functionally equivalent to a Type I PIPkin  in S. cerevisiae), and of a mammalian Type II PIPkin. The sequences of this "activation loop" are very different in each PIPkin family but are well conserved throughout any one family. For the Type III PIPkins, the consensus sequence is T(F/ Y)T(W/L)DKKLE(S/T/M)WVKXXG(I/L)(V/L)G: this motif has a well maintained pattern of hydrophobicity and of charge (including three conserved basic residues) in all Type III PIPkins. This motif may be involved in defining PtdIns3P as the substrate for 5-phosphorylation. DISCUSSION Recent studies suggest that mammalian Type I PIPkins can act as PtdIns3P 5-kinases in vitro (10,25,42). On the basis of these experiments, Tolias et al. (10) have suggested that the Type I PIPkins are responsible for the synthesis of PtdIns(3,5)P 2 that has been observed to occur in mammalian cells (6,7). We directly addressed this substrate specificity issue by expressing the murine Type I␤ PIPkin in ⌬fab1 MSS4 and FAB1 mss4-1 yeast. Our data show that the murine Type I␤ PIPkin is expressed in yeast in a functional state because this enzyme restores wild-type levels of PtdIns(4,5)P 2 to FAB1 mss4-1 yeast as well as complementing the mss4-1 phenotype (12). In contrast, the murine Type I␤ PIPkin is unable to synthesize any PtdIns(3,5)P 2 when expressed in ⌬fab1 MSS4 yeast and failed to complement the ⌬fab1 phenotype. This suggests that the ability of Type I PIPkins to synthesize PtdIns(3,5)P 2 in vitro is not a biologically relevant activity: although there appears to be a modest increase in FIG. 5. ScFab1p, SpFab1p, and p235 do not enhance PtdIns(4,5)P 2 levels in mss4-1 cells. mss4-1 strains expressing the mammalian Type I␤ PIPkin, ScFab1p, SpFab1p, and p235 were grown at 23°C in selective inositol-free SC medium containing 10 Ci⅐ml Ϫ1 [ 3 H]inositol. Cultures were grown to a cell density of 1 ϫ 10 6 cells⅐ml Ϫ1 and then split into four samples. Two samples were grown at 23°C for a further 2 h, and the other two were grown at the restrictive temperature for 2 h, after which their lipids were extracted and analyzed. HPLC chromatograms are shown of the GroPInsP 2 s at 23°C of wild-type (YPH499) (a), mss4-1 (YOC808) (b), mss4-1 expressing Type I␤ PIPkin (c), mss4-1 expressing ScFab1p (d), mss4-1 expressing SpFab1p (e), and mss4-1 expressing p235 (f).
PtdIns(3,5)P 2 levels in FAB1 mss4-1 cells expressing the murine Type I␤ PIPkin relative to FAB1 mss4-1 cells containing an empty vector, this ability to increase the steady-state levels of PtdIns(3,5)P 2 requires that the cells also express a wild-type FAB1 gene, i.e. no increase occurs in ⌬fab1 MSS4 cells. The increase in the steady-state levels of PtdIns(3,5)P 2 in the FAB1 mss4-1 cells is therefore unlikely to result from direct synthesis of PtdIns(3,5)P 2 by the murine Type 1␤ PIPkin.
Because the Type I PIPkins do not synthesize PtdIns(3,5)P 2 in vivo, the identity of the mammalian PtdIns3P 5-kinase appeared unresolved. The similarity of a cDNA encoding p235, a PIPkin homologue, to the yeast FAB1 gene suggested to us that the protein product of this gene might fulfil this function. Other studies have shown that this enzyme can act as a PtdIns  Asterisks identify the signature FYVE motif. c, alignment of the catalytic domains. ϩ, residues proposed to form an activation loop for substrate binding; †, residues thought to be involved in ATP binding (Lys-2059, Asp-2196, and Asp-2216 in ScFab1p). The sequences of this activation loop are very different in each of the three PIPkin families but are well conserved throughout any one family. To illustrate this, the relevant sequences of the catalytic domains of Mss4p (GenBank TM accession number P38994) and of one representative each of the mammalian Type I and II PIPkin families (U78575 and U85245, respectively) are shown below the six aligned Fab1p-related sequences. 5-kinase in vitro (9,49), but our results indicate that p235 does not serve as a PtdIns 5-kinase in vivo in yeast: neither p235 nor any other Type III PIPkin caused any detectable PtdIns5P synthesis. We suspect that the p235-catalyzed 5-phosphorylation of PtdIns that is seen in vitro does not reflect its normal biological function. Such observations re-emphasize the conclusion that attempts to define the substrate specificities of phospholipid-directed enzymes solely on the basis of enzymological assays in vitro, which do not reproduce the conditions of substrate presentation that prevail in vivo, must be interpreted with caution. Complementation studies similar to those described here may in future prove valuable for analyzing the substrate specificities of other lipid-metabolizing enzymes.
In addition to restoring basal PtdIns(3,5)P 2 levels, p235 rescued the temperature-sensitive growth phenotype and the vacuolar defects of fab1-1 and ⌬fab1 cells, respectively. The complementation of these phenotypic defects suggests that p235 (but not SpFab1p) can functionally interact with the proteins that regulate the spatial and/or temporal synthesis of PtdIns(3,5)P 2 : expression of both enzymes in ⌬fab1 yeast results in a similar basal level of PtdIns(3,5)P 2 synthesis. These results make it likely that Fab1p-like proteins fulfill similar roles both in making PtdIns(3,5)P 2 and in regulating membrane traffic (probably in the Golgi/lysosome/endosome continuum) in most or all eukaryotes, including mammals. This would be consonant with the fact that synthesis of PtdIns3P, the substrate of the Type III PIPkins, by Vps34p and other Type III phosphoinositide 3-kinases is essential for protein trafficking to vacuolar/lysosomal compartments both in yeast (43)(44)(45) and in mammalian cells (46 -48).
The absence of a hyperosmotic stress-induced increase in PtdIns(3,5)P 2 in p235-expressing cells suggests that this enzyme is not involved in the stress response in mammalian cells and also demonstrates that basal PtdIns(3,5)P 2 synthesis is sufficient to rescue all the known phenotypic defects of ⌬fab1 yeast. In contrast, SpFab1p partly restored the hyperosmotic activation of PtdIns(3,5)P 2 synthesis, but it neither complemented the temperature-sensitive growth of fab1-1 cells nor restored the morphology of ⌬fab1 cells, suggesting that an additional regulatory interaction (separate from its PtdIns3P 5-kinase activity and involved in the response of the cell to stress) is conserved between S. cerevisiae and S. pombe: the different regulatory properties seen with SpFab1p and p235 might merely have reflected a difference in their expression levels or rates of degradation. However, although expression of FAB1 from a single-copy plasmid was sufficient to complement the phenotypic defects of ⌬fab1 and fab1-1 cells (see Fig. 2, a  and b), expression of the SpFAB1 ORF from constitutively active promoters of the high copy plasmids pYO2144 (Table I) and pVT102U (data not shown) and from the galactose regulated promoter of pEGKT in 2% (w/v) galactose (data not shown) failed to rescue these defects. Thus, because both enzymes restored basal PtdIns(3,5)P 2 synthesis to fab1 cells and complemented at least one of the other functions of ScFab1p, this trivial explanation seems much less likely than a real difference in the functional properties of the two enzymes.
In support of our experimental data, the results of the sequence comparison establish that the six known proteins of the Fab1p family that includes the experimentally untested proteins of Arabidopsis, Caenorhabditis, and Drosophila are very closely related, consistent with the idea that they fulfil the same function in each of their host organisms. This conclusion, combined with the earlier demonstrations of the specific PtdIns3P 5-kinase activity of ScFab1p (3), makes it clear that ScFab1p and its homologues do indeed make up a new family of Type III PIPkins that synthesize PtdIns(3,5)P 2 : we suggest that this could be abbreviated to PIPkin-III, with an appropriate prefix to indicate species (e.g. ScPIPkin-III for Fab1p).