Overexpression of Yeast PAM1 Gene Permits Survival without Protein Phosphatase 2A and Induces a Filamentous Phenotype*

Protein phosphatase 2A (PP2A) is an essential enzyme which is present in all eukaryotic cells. PP2A has been implicated in regulating various metabolic processes and also in the control of cell cycle progression. In the yeast Saccharomyces cerevisiae, the catalytic subunit of PP2A is encoded by two duplicated genes, PPH21 and PPH22. A third related gene, PPHS, also contributes Borne PP2A activity. We have used a yeast strain in which a single functional PP2A gene is expressed from a regu- lated promoter to screen for high copy number suppres-sors of PP2A depletion. A new gene was cloned, PAM1 (PP2A multicopy suppressor), which in high copy nun-ber can bypass the need for a PP2A catalytic subunit. The PAM1 gene encodes a hydrophilic 93-kDa protein that contains two coiled coil motifs and has a highly basic C-terminal tail. High level overexpression of PAM1 inhibits growth and induces a filamentous phenotype. Many cellular processes are regulated by protein kinases that phosphorylate target proteins on serine and threonine residues (Hunter, 1987). However, eukaryotic cells also a number of serindthreonine-specific protein phosphatases that reverse the action kinases. Based chemical properties, enzymes classified type 1,2A, 2B, and 2C protein phosphatases Sequence analysis the catalytic subunits of the type 1,2A,

indicate this fact.
The abbreviations used are: PP, protein phosphatase; bp, base pailis). PPH21 and PPH22 (Ronne et al., 1991). It is not known if the Pph3 protein interacts with Tpd3 or Cdc55, nor is it clear that Pph3 has the same in vivo function as PP2A. Yet another PP2Arelated enzyme in budding yeast is PPHl/SIT4, which is essential in some yeast strains but dispensable in others . A SIT4 homolog in fission yeast, ppel, seems to be involved in regulating piml, a homolog of the human RCCl protein (Matsumoto and Beach, 1993;Shimanuki et al., 1993). The sequence similarity between PP1 and PP2A is reflected in overlapping substrate specificities. Thus, both enzymes can dephosphorylate a number of proteins involved in different metabolic processes. However, the two enzymes differ in intracellular location. PP1 is largely particle bound, being associated with glycogen particles, actomyosin complexes, and ribosomes (Cohen, 1989). It is therefore thought that PP1 regulates glycogen metabolism, muscle contractility, and protein synthesis, conclusions which in part are supported by biochemical and genetic data (Cohen, 1989). Moreover, PP1 is enriched in the nucleus (Okhura et al., 1989). In contrast, PP2A occurs as a soluble complex in the cytosol and is excluded from the nucleus (Kinoshita et al., 1993). It is thought to regulate cytosolic processes such as glycolysis, gluconeogenesis, and amino acid catabolism (Cohen, 1989). However, there is evidence that PP2A also contributes to the control of glycogen metabolism (Penf et al., 1991), and a minor role for PP2Ain other processes that are regulated by PP1 cannot be excluded (Cohen, 1989).
Genetic data suggest that PP1 and PP2A are involved in regulating cell cycle progression. PP1 is required for completion of mitosis in fungal and insect cells (reviewed by Sneddon and Stark, 1991). PP2Ais a negative regulator of the cdc2 kinase in amphibian and mammalian cells (FBlix et al., 1990;Yamashita et al., 1990;Lee et al., 19911, and modulation of PPPA activity plays an important role in transformation by the polyoma and SV40 viruses (reviewed by Mumby and Walter, 1991). Moreover, PPPA inhibits cyclin degradation in amphibian cells (Lorca et al., 19911, and the PP2A-related Sit4 enzyme is required for cyclin accumulation in yeast (Fernandez-Sarabia et al., 1992). In addition to its proposed roles in metabolic regulation and cell cycle control, there is also some evidence that PP2A regulates transcription in mammalian cells (Alberts et al., 1993;Wadzinski et al., 1993) and that it is involved in tissue pattern formation in Drosophila (Uemura et al., 1993).
Partial loss of PPPA activity in fission yeast is reported to cause premature mitosis, while complete loss is lethal (Kinoshita et al., 1990;1993). Moreover, a disruption of the major PP2A gene in fission yeast, ppa2, partially suppressed a thermosensitive mutation in the cdc25 tyrosine phosphatase (Kinoshita et al., 1993). The latter protein is required to activate the cdc2 kinase at the start of mitosis, and it was therefore proposed that PPSA may inhibit cdc2 through negative control of cdc25 activity (Kinoshita et al., 1993). PP2A is essential also in budding yeast (Sneddon and Strak, 1990;. However, PP2A-depleted cells do not arrest in a distinct phase 3429 Yeast PAM1 Gene Suppresses Loss of Protein Phosphatase 2A All strains except H365 and X3271-1C were congenic to W303-lA and therefore also have the SUC2 &2-1 canl-100 his3-11,15 leu23,112 trpl-1 urd-1 markers (Thomas and Rothstein, 1989 1992 Mortimer andHawthorne. 1973 of the cell cycle. Instead, they acquire an abnormal bud shape, suggesting a role for PP2A in cellular morphogenesis . "he two regulatory subunits of PP2A, Cdc55 and Tpd3, are not essential, but loss of either protein causes a cold-sensitive cell cycle arrest that seems to involve inhibition of cytokinesis (Healy et al., 1991;van Zyl et al., 1992). We have found that simultaneous loss of PPH21, PPH22, and PPH3 is lethal in yeast, which shows that some P E A catalytic activity is required for growth under normal conditions . We have now used this lethality to screen for genes which in high copy number can rescue cells that lack PP2A. One such gene, PAMl, is described below. It encodes a 93-kDa hydrophilic protein with two coiled coil motifs and a highly basic C terminus. Overexpression of PAMl inhibits growth and induces a filamentous phenotype.

EXPERIMENTAL PROCEDURES
Yeast Strain-The yeast strains are listed in Table I. The paml-61::LEUZ disruption was made by cloning the LEU2 HpaI-Sal1 fragment between the PfEMI and BglII sites in PAMl, andpaml-&?::LEU2 by cloning the same fragment between the NruI and 3' HpaI sites. H109 has a 660-bp EcoRI-Sau3AI fragment carrying the GAL1 promoter inserted between the BglII site and 3' HincII site in the LEU2 5' region.
It was made using a LEU2 promoter substitution system that we developed for cloning of heterologous transcription factors in yeast (Ellerstrijm et al., 1992). Briefly, a promoter is cloned between the two BglII sites in pHR35 (see below), in place of the LEU2 upstream activating sequence (Martinez-Arias et al., 1984). The plasmid is then cut with KpnI to target integration to the LEU2 gene and transformed into strain U457. Integration evmts (popins) are selected on uracil-less plates. The resulting strain has two tandem copies of the LEU2 promoter which are. separated by the URA3 marker. One copy is the wild type chromosomal promoter, and one copy is the mutant promoter from the plasmid. Direct repeat recombination events that lose the URA3 marker and one copy of the promoter (popouts) are then selected on 5-fluoro-orotic acid plates (Boeke et al., 1984). Two types of popouts occur: those that restore the wild type LEU2 gene, and those that leave a mutant promoter on the chromosome. The two events are easily distinguished in U457, which carries the SUP53-a amber suppressor tRNA gene on the fragment that is deleted in the mutant promoter. Correct popouts can therefore be identified through loss of SUP53-a mediated suppression of the trpl-1 amber allele.
Plasmids-The URA3 LEU2-d 2-pn plasmid pHR81 and the genomic library made in this vector have been described (Nehlin et al., 1989). The LEU2 promoter substitution vector pHR35 was made in three steps. First, a 1920-bp PstI-EcoRI fragment of YEpl3 with the 5' half of the LEU2 gene and adjacent upstream DNAwas cloned between thePstI and EcoRI sites of pUC18, generating pHR29. This plasmid was cleaved with HincII, ligated to BglII linkers, cleaved with BglII, and religated. The resulting plasmid, pHR33, has a deletion of 780 bp spanning the LEU2 upstream activating sequence (Martinez-Arias et al., 19841, the sup53 gene, and the 6 repeat 5' to LEU2. In its place, pHR33 has two tandem copies of the BglII linker. The URA3 Hind111 fragment was then cloned into the Hind111 site of pHR33, to generate pHR35. Plasmids pHGZ23, pHGZ27, and pHGZ28 were made by cloning different inserts into the BarnHI site of pHR81 (Nehlin et al., 1989). Thus, pHGZ23 contains a 3300-bp NdeI fragment carrying the PAMl gene (Fig. 2), pHGZ27 a 2700-bp SpeI fragment carrying the PPH3 gene (Ronne et al., 19911, and pHGZ28 a 2200-bpXhoI-ma1 fragment carrying thePPH21 gene (Ronne et al., 1991). Plasmid pHGZ45 and pHGZ46 were generated from pHGZ23 and pPAMl by ClaI+SalI digestion and religation, thereby removing the LEU2-d marker from these plasmids. Plasmid pHGZ38, in which PAM1 is expressed from the GALl:TPK2 hybrid promoter, was made by cloning a 4500-bp NruI-SpeI fragment carrying PAMl into the BamHI site of pJN92 .
Other Methods-The methods used for yeast genetics, molecular cloning, and Northern blotting have been described (Nehlin et al., 1989;Nehlin and Ronne, 1990). Attempts to induce filamentous growth in W303 congenic diploids using the published protocol (Gimeno et al., 1992) were not successful. After further testing, we found that a low agar concentration facilitates filamentous growth. Thus, if the amount of agar is reduced from 2 to 1% the protocol works well also with standard laboratory strains. Nucleotide sequences were determined on both strands after subcloning in pUC118 and pUC119 (Vieira and Messing, 1987). For photography, yeast cells were pelleted, resuspended in 1 M sorbitol, fixed in ethanol, and stained with 4',6-diamidino-2-phenylindole (Lundin et al., 1991). ALeitz Ortholux 2 microscope was used to take the pictures, using filter system A and a 63/1.30 combined fluorescence and phase contrast objective.

RESULTS
Cloning of the PAMl Gene-In an attempt to analyze the cellular function of P E A , we have screened the yeast genome for genes that in high copy number can suppress the growth defect caused by loss of the PPPA catalytic subunit. For this, we used the previously described yeast strain H328 . This strain is disrupted for PPH21, one of two duplicated genes encoding PPSA, whereas the other gene, PPH22, is expressed from the galactose-induced GAL1 promoter. H328 is also disrupted for PPH3, a third more distantly related gene which provides some residual PPPA activity in cells lacking both PPH21 and PPH22 (Rohe et al., 1991). H328 is unable to grow on other carbon sources than galactose, since it is dependent on expression of PPH22 from the GAL1 promoter to provide the PP2A activity required for survival.
H328 cells were transformed with a yeast genomic library made in the high copy number vedor pHRSl (Nehlin et al., 1989). "ransformants were selected on uracil-less galactose plates and then replicated to glucose and raffinose plates. Colonies that could grow on glucose or raffinose were tested for co-segregational loss of the plasmid and the ability to grow without galactose. Plasmids were rescued from the cells and their ability to suppress PP2A depletion was confirmed by retransformation into H328. In addition to multiple copies of the PPH genes, we cloned one new gene which we call PAMl for PP2A Multicopy suppressor. The degree of suppression provided by PAM1 is much weaker than that obtained with PPH21 or PPH22, but comparable to that obtained with plasmids containing the PPH3 gene (Fig. L4).
Genetic Mapping of the PAMl Locus-The PAMl gene was located to chromosome IV in a Southern blot of yeast DNA separated on a contour-clamped homogenous electric field gel (Chu et al., 1986). We proceeded to map the gene by tetrad analysis, using yeast strains in which the PAMl locus was genetically marked by one-step gene disruptions (see below). We found that the PAMl gene is on the right arm of chromosome IV, 43 cM from pet14 and 61 centiMorgans from trp4 (Table 11). The sequence containing different plasmids were tested for growth on glucose, as previously described (41). B , co-segregational loss of a PAMl plasmid and the ability of pph21 pph22 pph3 cells to grow well on 2% glucose. Triple disrupted cells containing the plasmid were streak purified on 8% glucose. Single colonies were picked to an 8% glucose plate and replicated to different media. The third colony from the left has lost both the URA3 marker on the PAM1 plasmid and the ability to grow well on 2% glucose.

Genetic mapping data for the PAMl locus
The tetrads are from a cross of H365 to X3271-1C (Table I) analysis (see below) revealed that the 3' end of PAMl is located 2 kb 5' to the CHLP gene (Kouprina et al., 1993b). This gene, which is also known as CTFl7 (Spencer et al., 1990), is involved in mitotic chromosome segregation. It is also known to be located on the right arm of chromosome IV, in the vicinity of the SUP2 gene (Kuoprina et al., 1993a). This position is in good agreement with our mapping of the PAMl locus.
Sequence of the PAM1 Gene-The restriction map of the pPAMl plasmid is shown in Fig. 2., and the nucleotide sequence of the PAMl gene and its encoded protein in Fig. 3. Deletion mapping revealed that the suppressing activity is mediated by an open reading frame of 830 codons. A plasmid lacking the last 150 codons of PAMl was still partially active as a suppressor of PP2A depletion. To prove that the activity is mediated by the encoded protein rather than by the plasmid DNA, we also made a frame-shift mutation by filling in the A d 1 site in the 5' part of PAMl. This plasmid was without shown are the deletions made by one-step gene disruptions, and the activity, which confirms that suppression is mediated by the Paml protein. There are no sites in the PAMl promoter that match any of the known consensus sequences recognized by yeast transcription factors (Verdier, 1990). However, some regulatory elements seem to be present, since deletions in the promoter caused a partial loss of the suppressor activity (Fig. 2).
Structure of the Paml Protein-The PAMl open reading frame predicts a hydrophilic protein with an apparent moiecular mass of 93 kDa. A computer search of the EMBL data bank (Release 34) and the PIR and SWISS protein data bases (Releases 35 and 24) using the FASTA program (Pearson and Lipman, 1988) did not reveal any strong similarity to previously known proteins. However, a number of entries with optimized FASTA scores between 90 and 110 were found, including dystrophin, the Drosophila BicD protein, laminins, myosins, and various intermediate filament proteins. These proteins all contain coiled coil motifs (Cohen and Parry, 1986;Lupas et aZ., 1991), and the sequence similarities are due to the presence of two such motifs in the Paml protein. The two coiled coil motifs are located in the middle of the protein and are separated by 80 amino acid residues. The BicD protein (Wharton and Struhl, 19891, which showed the highest local similarity in a FASTA comparison to Paml, has a similar pair of coiled coil motifs separated by a spacer region (Fig. 4). BicD is a maternal factor which is involved in establishing anterior-posterior polarity in the Drosophila oocyte. However, the two proteins do not have the same overall organization, since BicD contains several other coiled coil motifs that have no counterparts in Paml. The second motif in Paml has an unusual structure: it contains two staggered leucine heptad repeats and is highly charged (18 of 34 residues). Between the two coiled coil motifs, there is a possible target site for the CAMP-dependent protein kinase (Krebs and Beavo, 1979). "he second motif is followed by a polyglutamine tract. Such tracts are found in many proteins and have been proposed to act as flexible hinges that separate different domains (Wootton and Drummond, 1989). A third notable feature in the Paml sequence is its highly basic C terminus. Of the 16 C-terminal residues in Paml, 10 are basic, and there is a stretch of 5 consecutive lysines.
Disruption of the PAMl Gene-TO investigate the function of   PAMl, we carried out one-step disruptions of the gene (Rothstein, 1983). Two disruptions were tested paml-61 which deletes the central part of the protein including the two coiled coil motifs and paml-62 which deletes the start codon and most of the open reading frame. The resulting yeast strains were viable and showed no obvious defects in vegetative growth, mating, or in the ability to use different carbon sources. Nor was thermotolerance, survival of nitrogen starvation, or the ability ta accumulate glycogen as determined from iodine staining markedly affected. Diploids that are homozygous for a paml disruption sporulate normally and are capable of induced pseudohyphal growth.

TCTCAMACGGTACCTCAMAGTTCCTCAGATCMTATTACGCMCCATCACCMTTCAGACCMCT~GCTACCAGTGATMCCCTGCT S Q N C T S K V P P I N I T Q P S P I Q T N F A T S D N P A G C T G T M T M A A C T A~C A C C T T C T G M G A T A C A G m f A C C T C A V I -K L G T P S E D T V S A A A T A N N I S T H G D E S R K E D V K E K K K K . K F S F F G K R K K M A G A C G A C G T T M C C~~~T T C A G~C T T T U j M T A C G C G A G C C C T T G T C T T T C C A A T C C A A A T M T C C G T T T T T G T T M T C A T A T G T T A T A r m A T t T A T T M T G A C C A
The lack of a detectable gene disruption phenotype in yeast is frequently due to gene duplications (Toda et al., 1987;. We therefore proceeded to investigate whether a second gene closely related to PAMl is present in the yeast genome, using low stringency hybridizations with a PAMl probe. To increase our ability to detect a weak signal, we used DNA from the paml-62 strain H405, in which the entire probe has been deleted. We were unable to detect any signal under these conditions (data not shown). This argues against the presence of a second gene closely related to PAMl but does not rule out the existence of more distantly related genes.
Expression of the PAMl mRIVA-One possible reason for the apparent lack of a disruption phenotype would be that PAMl is expressed only under certain conditions, being without function during normal vegetative growth. We therefore proceeded to study the expression of PAMl in Northern blots with mRNA from cells grown on different carbon sources. We found that PAM1 is expressed at a fairly high level in vegetative cells (Fig.  5). Moreover, while some differences in the amount of mRNA could be seen, PAMl was clearly expressed on all carbon sources tested. This suggests that PAMl could be a housekeeping gene which is expressed under most conditions, a conclusion which would be consistent with the absence of known regulatory sequence motifs in the PAMl promoter.
PAMl Does Not Activate the GALl Promoter-A possible mechanism for suppression would be that PAMl allows expression of the GALl promoter on other carbon sources than galactose. To investigate this possibility, we tested the PAMl gene in two genetic systems. First, we used yeast strain H172, in which the TPK2 gene is expressed from the GALl promoter (Nehlin et al ., 1992). This strain is conditional lethal for growth on galactose, since expression of TPK2 from the GALl promoter rapidly kills the cell. We found that the two PAMl plasmids pPAMl and pHGZ23 failed to prevent growth of H172 on other carbon sources than galactose (data not shown). This indicates that the GALl promoter is not improperly turned on under these conditions.

P O~I L N K G D S E U F I R K D E N T A L T R V D D L Q N S X A L BicD G T I K I C I D I P F S ElIlH L NIEIL K K ( L I E K --P I L I E S M E S
In a second and more sensitive test, we used yeast strain H109 (Table I). This strain has the chromosomal LEU2 gene expressed from the GALl promoter, and is therefore dependent on galactose for growth in the absence of leucine. Due to the complete absence of background growth, very low levels of expression can be detected in this system. To permit us to assay LEU2 expression, we used two PAMl plasmids, pHGZ45 and pHGZ46, that were obtained from pHGZ23 and pPAMl by removing the LEU2-d marker on these plasmids. We found that these PAMl plasmids failed to permit any growth of H109 in the absence of galactose and leucine (data not shown). This shows that the GALl promoter is still tightly regulated in the presence of the PAMl plasmids. It should be noted that this assay would select for cells with a high plasmid copy number, if this was necessary to derepress the GALl promoter. However, prolonged incubation of the plates failed to generate prototrophic papillae, which would have appeared if such selection took place.
PAMl Suppresses Complete Loss of PP2A-To verify that the PAMl plasmid is a bona fide suppressor of PP2A depletion, we proceeded to test it in a strain that is disrupted for all threePPH genes, rather than having one of them expressed from the GALl promoter. Such triple disrupted strains that are completely deficient for PP2A activity can be constructed by genetic crosses, but they fail to grow under normal conditions. To solve this prob-lem, we took advantage of a recent observation by Thompson-Jaeger et al. (1991). They found that the severe growth defect of bcyl cells, which have a deregulated CAMP-dependent kinase, is partially relieved on high glucose media (8 instead of 2%). Moreover, a bcyl snfl double disruption, which normally is lethal, could grow on such media. PP2A-depleted cells are phenotypically similar to bcyl and snfl cells ( b n n e et al., 19911, and we therefore reasoned that they, too, might grow better on high glucose plates. This was indeed the case: triple disrupted pph21 pph22 pphd spores germinate under these conditions. The triple disrupted cells grow poorly on 8% glucose, extremely poorly on 2% glucose, and fail to grow on all other carbon sources. Like pph21 pph22 double disrupted cells, they acquire suppressor mutations at a high frequency (Ronne et al., 1991).
We proceeded to dissect tetrads from pphd homozygous, pph21 pph22 heterozygous diploids containing either the PAMl plasmid pHGZ23, or the control vector pHR81. Tetrads were dissected on 8% glucose plates, and cells containing plasmids were identified by the presence of the URA3 marker. We found that triple disrupted spores that contained the PAMl plasmid grew markedly better on 2% glucose plates than those without plasmid. The control vector had no such effect. Moreover, when pph21 pph22 pphd cells containing the PAMl plasmid were streak purified on 8% glucose plates, the ability to grow well on 2% glucose co-segregated with the URA3 marker on the plasmid (Fig. m). This confirms that the enhanced growth of these cells on 2% glucose is due to the PAMl plasmid and not to unlinked suppressor mutations.

High Level Overproduction of Paml Inhibits Growth-We
proceeded to test the effect of high level expression of the PAMl gene. This can be achieved by growing the cells in the absence of leucine, since the pHR81 vector has a defective LEU2-d marker that requires a very high copy number (Erhart and Hollenberg, 1983; Nehlin and b n n e , 1990). We found that the PAMl gene inhibits growth under these conditions. The growth inhibiting activity was mapped to the PAM1 open reading frame, using the same deletions and frame shifts that we used to map the PP2A suppressor activity (Fig. 2). However, the growth inhibition was more sensitive to some of the deletions.
Thus, a deletion of the 150 C-terminal codons abolished growth inhibition, as did a deletion down to base -382 in the promoter. A further deletion down to base -143 partially restored growth inhibition. This suggests that a high level of expression is necessary to achieve growth inhibition. This is consistent with the fact that moderate overexpression of PAMl, which suppresses PP2A depletion, does not inhibit growth.

Paml Overproducing Cells Acquire a Filamentous Pheno-
type-We next wanted to examine the terminal phenotype of cells that cease to grow due to P a m l overproduction. A problem with LEU2-d selection is that most cells will arrest due to leucine starvation when they are transferred to leucine-less media, since they have too few copies of the plasmid. These cells outnumber those that have enough plasmids to support growth, but instead cease to grow due to overexpression of the insert. To avoid this problem, we first cloned PAMl into the LEU2-d vector pJN92 (Ronne et al., 1991), where the insert is expressed from the galactose-induced GALl promoter. Cells containing this plasmid were then grown in leucine-less glycerol-lactate media, to select a high plasmid copy number in all cells . Finally, expression of PAMl was induced by adding 2% galactose to the culture, and samples of the cells were removed for microscopy at regular intervals.
Interestingly, we found that Paml overproduction induces a filamentous phenotype which resembles the pseudohyphae that are formed during nitrogen starvation (Gimeno et al., 1992). Thus, branched filaments are formed, which contain a mixture of elongated and spheroidal cells. This phenotype ap- pears after 8 h in galactose and is most prominent after 20 h (Fig. 6). The filamentous phenotype is seen only with the Pamloverproducing plasmid. Cells containing the vector pJN92 instead continue to grow and eventually arrest as single unbudded cells (Fig. 6). Frequently, a branched chain of cells emanates from a single large mother cell. It should be noted that individual cells within a filament are well separated and that each cell has its own single nucleus, as shown by DAPI staining (Fig. 6). This phenotype is distinct from that of cdc4 cells, which form multiple buds without nuclei, and from cdc3, cdcl0, cdcll, and cdcld cells, which form elongated buds with multiple nuclei (Hartwell, 1971a(Hartwell, , 1971b. The phenotype is also distinct from that of cdc55 and tpd3 mutants, which are deficient for the regulatory subunits of PP2A (Healy et al., 1991;van Zyl et al., 1992). These mutants form elongated multinucleate structures with partial constrictions suggesting a defect in cytokinesis, and the nuclei are more diffuse than in wild type cells. In Paml-overproducing cells, there are no partial constrictions, and the nuclei appear normal.

Yeast PAM1 Gene Suppresses
We also examined the effect of P a m l overproduction in haploid cells. Pseudohyphal growth was originally thought to be 3ss of Protein Phosphatase 2A specific for diploids (Gimeno et al., 1992), but has since been shown to occur also in haploids (Wright et al., 1993). We found that Paml overproduction induces a filamentous phenotype in haploids, but the effect is less pronounced than in diploids (Fig.  6). Finally, we tested whether PAMl overexpression in a diploid can induce pseudohyphal growth on agar plates in the absence of nitrogen starvation. We found that overproduction of PAMl under these conditions produced a cellular phenotype resembling that seen in liquid culture. However, there was no directed outgrowth of filaments from the colonies, nor was there invasive growth into the agar (data not shown). We conclude that while overproduction of Paml can induce a cellular morphology which resembles that found in pseudohyphae, it is not sufficient to promote a directed outgrowth of filaments into the surrounding media.

DISCUSSION
We have cloned a gene, PAMl, which in high copy number can suppress the loss of PP2A. There are several possible mechanisms of suppression that have to be considered. A protein phosphatase could suppress the PPPAdeficiency directly through its enzyme activity. An example of this is suppression by the PPH3 gene (Fig. 1). Paml shows no sequence similarity to known protein phosphatases, so it is unlikely that Paml is a phosphatase. However, it is possible that Paml could activate a PP2A-related phosphatase, such as Sit4 or Pph3, which in turn suppresses the PPSAdeficiency. This could involve control of gene expression or regulation at the protein level of enzyme activity, specificity, or intracellular targeting. P a m l shows no similarity to the PPPA regulatory subunits or to other proteins implicated in phosphatase function (Luke et al., 1991;Sutton et al., 1991;Wilson et al., 1991). The C terminus of Paml contains a stretch of polylysine (Fig. 3). Polylysine stimulates PP2A activity against some substrates more than 200-fold (Pelech and Cohen, 1985), and it has been proposed that such basic peptides could mimic a physiological activator of PP2A (8). However, the fact that a Paml protein lacking its 150 C-terminal residues is still partially active as a suppressor (Fig. 2) argues against a crucial role for the polylysine stretch. The fact that Paml functions in apph3 strain shows that suppression does not involve activation of Pph3. The other PP2A-related enzyme in yeast, Sit4, is essential in the W303-1A background , and we were therefore unable to test its role in suppression.
Alternatively, Paml could inhibit a kinase whose activity is lethal unless balanced by PP2A. An obvious candidate would be the CAMP-dependent kinase, since overexpression of this kinase is lethal in yeast (Nehlin et al., 1989). However, PAMl does not suppress the galactose-dependent lethality of the GAL1:TPID fusion, nor does it modify the pleiotropic phenotype of bcyl cells, which have a deregulated kinase (data not shown). This suggests that Paml does not inhibit the CAMP-dependent kinase. It is still possible that Paml could act by inhibiting another protein kinase which remains to be identified.
Finally, it is possible that Paml does not affect protein phosphorylation, but instead acts directly on a system that is regulated by PPBA, preventing the lethal effect of PP2A depletion. A problem with testing this hypothesis is that we do not know why PP2A is required for growth. In addition to its proposed roles in metabolic regulation (Cohen, 1989) and cell cycle control (FClix et al., 1990;Lee et al., 1990;Kinoshita et al., 1990;19931, PP2A has also been implicated in control of cell shape  and gene expression (Alberts, 1993;Wadzinski et al., 1993). If the essential function of PP2A is to inhibit the cdc2 kinase, one would expect a mitotic catastrophe to occur in PP2A-depleted cells (Russell and Nurse, 1987). However, PP2A-depleted budding yeast cells do not arrest in mitosis or in any other specific stage of the cell cycle (Ronne et al., 1991).
Similarly, depletion of PP2A in fission yeast by disruption of ppa2 and exposure to okadaic acid failed to cause a mitotic catastrophe (Kinoshita et al., 1993). Instead, binucleate cells accumulated, suggesting a defect in cytokinesis.
Interestingly, we found that the need for PP2A is partially relieved by high amounts of glucose (Fig. 1B). This raises the possibility that the essential function of PP2A could be in the control of carbon metabolism. PP2A regulates several glycolytic and gluconeogenic enzymes (Cohen, 19891, and the PP2A inhibitor okadaic acid rapidly stimulates gluconeogenesis in mammalian cells (Haystead et al., 1989). It is conceivable that loss of PP2A could activate gluconeogenesis also in yeast, which would cause futile cycling of intermediates in cells grown on glucose. Such a condition is expected to be partially relieved by high amounts of glucose. However, it is also possible that the survival of PPPA-depleted cells on high glucose is due to a generally enhanced growth under these conditions. The aberrant bud shape of PP2A-depleted cells and the elongated shape of PPZA overproducing cells suggests a possible role for PPSA in controlling cell shape (Fkmne et al., 1991). In this context, it is interesting that Paml shows some limited similarity to cytoskeletal coiled coil proteins. The similarity to BicD (Fig. 4) is particularly intriguing, since this protein is involved in establishing cell polarity in the Drosophila oocyte (Wharton and Struhl, 1989). The fact that Paml overexpression causes a filamentous phenotype (Fig. 6) could indicate a similar role for Paml in the control of cell polarity in yeast. However, it should be emphasized that Paml differs from classical coiled coil proteins such as myosin and BicD in that it contains only two coiled coil motifs.
The filamentous phenotype induced by Paml resembles the pseudohyphae that are formed during nitrogen starvation. Pseudohyphal growth is stimulated by the RAS2"l19 mutation, which activates the CAMP-dependent kinase (Gimeno et al., 1992). Since PP2A dephosphorylates some known targets for this kinase (Cohen, 19891, one might expect PP2A depletion to produce a similar effect. However, PP2A-depleted cells do not acquire a filamentous phenotype (Ronne et al., 1991). In contrast, cells that overexpress PP2A have an elongated shape resembling pseudohyphal cells. Like pseudohyphal growth, this effect is also more pronounced in diploids . Cell elongation is the first visible step in pseudohyphal growth (Gimeno et al., 1992), and it is possible that the elongation induced by PPPA could reflect a partial activation of this process. This would suggest that it is under negative control by phosphorylation, which is surprising in view of its being induced by RAS2va119. However, some protein kinases in yeast have opposing regulatory functions (Denis and Audino, 1991), which makes it difficult to predict the effect of protein phosphorylation on a given process. Moreover, the pleiotropic effects of RAS2va119 do not necessarily reflect the normal function of the RASICAMP pathway, which is in fact dispensable for most processes that are induced by RAS2va119 (Cameron et al. ,1988).
Paml is not required for pseudohyphal growth, but it induces a filamentous phenotype when overexpressed. This suggests that overproduction of Paml interferes with the regulatory pathway(s) that control cell shape and budding. The fact that Paml overproduction failed to cause invasive growth on agar plates further suggests that pseudohyphal growth requires more than a simple change in cell shape and budding patterns. It is conceivable that a nutrient-sensing mechanism similar to chemotaxis also is involved in pseudohyphal growth, which is not active in Paml-overproducing cells. In any case, further studies of Paml may shed light on the control of cell shape in yeast, and the role of PP2A in this process.