Genome‐wide lethality screen identifies new PI4,5P2 effectors that regulate the actin cytoskeleton

Anjon Audhya, Robbie Loewith, Ainslie B Parsons, Lu Gao, Mitsuaki Tabuchi, Huilin Zhou, Charles Boone, Michael N Hall, Scott D Emr

Author Affiliations

  1. Anjon Audhya1,,
  2. Robbie Loewith2,
  3. Ainslie B Parsons3,4,
  4. Lu Gao5,6,
  5. Mitsuaki Tabuchi1,
  6. Huilin Zhou5,6,
  7. Charles Boone3,4,
  8. Michael N Hall2 and
  9. Scott D Emr*,1
  1. 1 Department of Cellular and Molecular Medicine, The Howard Hughes Medical Institute, University of California, San Diego School of Medicine, La Jolla, CA, USA
  2. 2 Division of Biochemistry, Biozentrum, University of Basel, Basel, Switzerland
  3. 3 Banting and Best Department of Medical Research, University of Toronto, Toronto, Ontario, Canada
  4. 4 Department of Molecular and Medical Genetics, University of Toronto, Toronto, Ontario, Canada
  5. 5 Ludwig Institute for Cancer Research, La Jolla, CA, USA
  6. 6 Department of Cellular and Molecular Medicine, University of California, San Diego, La Jolla, CA, USA
  1. *Corresponding author. Division of Cellular and Molecular Medicine, HHMI, UCSD School of Medicine, Cellular and Molecular Medicine Bldg, Rm 318, 9500 Gilman Drive, 3rd Floor, La Jolla, CA 92093‐0668, USA. Tel.: +1 858 534 6462; Fax: +1 858 534 6414; E-mail: semr{at}
  • Present address: Ludwig Institute for Cancer Research, La Jolla, CA 92093, USA

View Full Text


To further understand the roles played by the essential phosphoinositide PI4,5P2, we have used a synthetic lethal analysis, which systematically combined the mss4ts mutation, partially defective in PI4P 5‐kinase activity, with each of approximately 4700 deletion mutations. This genomic screening technique uncovered numerous new candidate effectors and regulators of PI4,5P2 in yeast. In particular, we identified Slm1 (Yil105c), a previously uncharacterized PI4,5P2 binding protein. Like Mss4, Slm1 and its homolog Slm2 (Ynl047c) were required for actin cytoskeleton polarization and viability. Co‐immunoprecipitation experiments revealed that Slm1 interacts with a component of TORC2, a Tor2 kinase‐containing complex, which also regulates the actin cytoskeleton. Consistent with these findings, phosphorylation of Slm1 and Slm2 was dependent on TORC2 protein kinase activity, both in vivo and in vitro, and Slm1 localization required both PI4,5P2 and functional TORC2. Together, these data suggest that Slm1 and Slm2 function downstream of PI4,5P2 and the TORC2 kinase pathway to control actin cytoskeleton organization.


Over the past decade, derivatives of the lipid phosphatidylinositol, collectively known as phosphoinositides, have emerged as essential regulatory molecules involved in a diverse spectrum of cellular processes (Fruman et al, 1998; Takenawa and Itoh, 2001). Multiple characterized effectors of phosphoinositides harbor specific domains important for lipid recognition. Pleckstrin homology (PH) domains from several proteins have been shown to interact with phosphoinositides, sometimes showing a clear preference for a single lipid (Lemmon and Ferguson, 2000). The PH domain from phospholipase C‐δ1 (PLCδ1) specifically binds PI4,5P2 and its soluble head group IP3 in vitro (Cifuentes et al, 1993), and when fused to GFP, can be used as an in vivo probe for this lipid (Stauffer et al, 1998). These data suggest that proteins containing PH domains or other phosphoinositide interacting motifs (i.e. FYVE, PX, ENTH, etc.) are recruited and/or activated by binding phosphoinositides on specific membrane compartments. Through the action of a set of well‐conserved and tightly regulated phosphoinositide kinases, phosphatases, and lipases, phosphoinositides can efficiently modulate downstream events mediated by their effectors in response to a variety of stimuli. The yeast Saccharomyces cerevisiae has proven to be a useful model system for the study of phosphoinositides, showing essential roles for PI3P in endosomal membrane trafficking, PI4P in Golgi secretion, and PI4,5P2 in endocytosis and actin cytoskeleton organization (Odorizzi et al, 2000).

Generation of PI4,5P2 is catalyzed by a single phosphoinositide kinase in yeast, Mss4. Using temperature‐sensitive alleles of MSS4, multiple groups have demonstrated that PI4,5P2 generated at the plasma membrane is required for actin cytoskeleton organization in yeast (Desrivières et al, 1998; Homma et al, 1998). Correspondingly, studies using several different biological systems indicate the existence of numerous PI4,5P2 effector proteins involved in cytoskeletal dynamics, including gelsolin, cofilin, the Arp2/3 activator WASP, ERM proteins, and exchange factors for small Rho‐type GTPases (Takenawa and Itoh, 2001). In particular, Mss4‐generated PI4,5P2 has been shown to recruit/activate the Rho1 GTPase exchange factor Rom2, which is required for Rho1 activation and maintenance of actin polarization (Audhya and Emr, 2002). However, the mechanism(s) behind how PI4,5P2 coordinates the recruitment/activation of its many effectors remains unclear.

In addition to Mss4, multiple other factors have been implicated in the regulation of actin cytoskeleton organization, several of which affect Rho1 activation. Previous studies have shown that Tor2 also functions upstream of Rho1, and tor2ts mutant cells exhibit a defect in actin organization similar to mss4ts cells (Schmidt et al, 1997). TOR, the target of rapamycin, is a conserved phosphatidylinositol kinase‐related protein kinase that controls cell growth in response to nutrient conditions. Yeast cells express two TOR homologs, Tor1 and Tor2, which can participate in two distinct complexes, TORC1 and TORC2 (Loewith et al, 2002). TORC1 is rapamycin sensitive, contains either Tor1 or Tor2, and controls translation initiation, ribosome biogenesis, transcription, and other growth‐related readouts. In contrast, TORC2 is rapamycin insensitive, contains Tor2 but not Tor1, and controls polarization of the actin cytoskeleton (Loewith et al, 2002). In addition to Tor2, TORC2 also contains five additional subunits: Avo1, Avo2, Avo3, Bit61, and Lst8 (Loewith et al, 2002; Wedaman et al, 2003; Reinke et al, 2004). Interestingly, tor2ts mutations can be suppressed by overexpression of Mss4, suggesting an interaction between the Tor2 protein kinase and the Mss4 lipid kinase (Helliwell et al, 1998).

To initiate an unbiased investigation of the physiological roles of PI4,5P2, we conducted a genome‐wide synthetic lethal screen using a strain expressing a functionally impaired, temperature‐sensitive allele of MSS4. We identified two previously uncharacterized proteins, Slm1 and Slm2, which are downstream effectors of both Mss4 and TORC2 and are required for organization of the actin cytoskeleton. These findings suggest an integration of lipid kinase function with TORC2 protein kinase regulation, both necessary for spatial and temporal regulation of actin cytoskeleton dynamics.


Synthetic genetic array analysis reveals an Mss4 network of genetic interactions

MSS4 encodes an essential PI4P 5‐kinase in yeast. To extend our understanding of Mss4 regulation and activity, we utilized a method, termed synthetic genetic array (SGA) analysis, for systematic construction and characterization of double‐mutant strains. An mss4ts mutation was combined with each of approximately 4700 deletion mutations at the permissive growth temperature for mss4ts cells. Under these conditions, mss4ts cells generate approximately 40% reduced levels of PI4,5P2 as compared to wild‐type cells (Stefan et al, 2002), slightly compromising but not inactivating PI4,5P2‐dependent cellular functions. We reasoned that inviable or slow‐growing double‐mutant progeny could identify a functional interaction between Mss4 and the product of the wild‐type version of the deleted gene. Such proteins could represent regulators of the PI4P 5‐kinase or targets of its product, PI4,5P2. Previous work indicated that mutations in MSS4 are lethal in combination with mutations that affect activation of the Rho1 GTPase (Audhya and Emr, 2002). In agreement with these studies, the SGA analysis identified Rom2, the exchange factor for Rho1. Since Rom2 requires PI4,5P2 synthesis for its appropriate localization (Audhya and Emr, 2002), the SGA analysis should identify potential novel PI4,5P2 effectors.

In total, almost 80 double‐deletion combinations were initially isolated in the screen that resulted in a synthetic growth defect. Of these, approximately half were confirmed using an alternative method (see Supplementary Table I). Multiple genes that were Synthetic Lethal with Mss4 (SLM) encoded proteins that have been implicated in actin cytoskeleton organization and/or cell polarity including Cap1, Cap2, Avo2, Spa2, Bem4, Boi2, Myo5, Cyk3, Ilm1, Van1, and Api2. Additionally, multiple components of the GimC complex, a chaperone believed to be important for normal folding of actin monomers (Siegers et al, 1999), were isolated in this screen. The identification of genes encoding proteins involved in a number of other processes including cellular stress response, ribosomal biosynthesis, protein/lipid modification, vesicle trafficking, and mitochondrial function suggests that Mss4 is involved in multiple biological pathways. Taken together, the results from the SGA analysis confirmed an essential function for Mss4 in cytoskeletal organization, but also indicated that Mss4/PI4,5P2 likely plays other roles in maintaining cell structure and viability (Figure 1A).

Figure 1.

SGA analysis using the mss4ts mutation indicates diverse roles for PI4,5P2. (A) The genes that are required for normal growth of mss4ts cells are represented as nodes. Each node is color coded based on its functional classification as defined by published literature. (B) Cartoon showing the domains of Slm1 and Slm2. Both proteins contain PH domains that interact with phosphoinositides. Additionally, Slm1 and Slm2 share a highly conserved domain, the Slm domain, which has been highlighted.

Of particular interest is the SGA analysis that revealed a genetic interaction between MSS4 and YIL105c, a gene encoding a protein that contains a PH domain (Figure 1B), which has been shown to interact weakly with phosphoinositides in vitro (Yu et al, 2004). Moreover, plasma membrane localization of the PH domain from Yil105c requires Mss4‐dependent PI4,5P2 production (Yu et al, 2004). Consistent with studies of the Yil105c PH domain, full‐length Yil105c also interacted with multiple phosphoinositides in liposome binding assays (Supplementary Figure 1).

The genetic interaction between the mss4ts mutation and yil105cΔ was confirmed in the SEY6210 background, where mss4tsyil105cΔ double‐mutant cells exhibited a potent growth defect at 31.5°C in contrast to the near‐normal growth of both single‐mutant strains at the same temperature (Figure 2A). Analysis of phosphoinositide levels in an mss4ts mutant and in mss4tsyil105cΔ double‐mutant cells at 31.5°C revealed comparable levels of phosphoinositides (data not shown), suggesting that Yil105c is not upstream of Mss4 and is consistent with the idea that PI4,5P2 may play a role in the recruitment/activation of Yil105c. Given its genetic interaction with the mss4ts mutation, we have named this gene SLM1 for Synthetic Lethal with Mss4.

Figure 2.

Slm1 and Slm2 perform an essential function downstream of Mss4. (A) slm1Δ, mss4ts and slm1Δmss4ts cells were grown at 26 or 31.5°C for 60 h. (B) slm1Δ and slm2Δ cells were mated, and the resulting diploids were sporulated and dissected. Tetrads (1–6) are shown, yielding four viable spores (indicative of two slm1Δ colonies and two slm2Δ colonies), three viable spores (indicative of one wild‐type colony, one slm1Δ colony, one slm2Δ colony, and one inviable slm1Δslm2Δ colony), or two viable spores (indicative of two wild‐type colonies and two inviable slm1Δslm2Δ colonies). The arrows highlight inviable slm1Δslm2Δ colonies. (C) TCA‐precipitated extracts from cells expressing SLM1‐GFP or SLM2‐GFP were analyzed by SDS–PAGE followed by Western blot using α‐GFP antibodies to detect Slm1 and Slm2 or α‐G6PDH antibodies as a loading control.

Since cells deleted for SLM1 alone failed to display any significant phenotype, we performed a database search to detect potential homologs that may serve a redundant function. This analysis led to the identification of YNL047c, which encodes a protein that is 53% identical to Slm1. Due to this strong similarity, we have named this gene SLM2 (Figure 1B). Slm2 also contains a PH domain that binds phosphoinositides, with a preference for PI4,5P2 when expressed as a GFP fusion in cells (Yu et al, 2004). Deletion of SLM2 did not yield any noticeable effects on cell growth, but in combination with deletion of SLM1, cells were no longer viable (Figure 2B). Together, these data indicate that Slm1 and Slm2 serve an essential function in yeast, likely downstream of Mss4/PI4,5P2.

Although Slm1 and Slm2 appear to be redundant in function, only SLM1 was uncovered in the SGA analysis using the mss4ts mutation. Consistent with this observation, mss4tsslm2Δ double‐mutant cells did not exhibit a significant synthetic growth defect in contrast to mss4tsslm1Δ double‐mutant cells. These data suggest that Slm1 may play a more dominant role in cells as compared to Slm2. To test directly the expression levels of Slm1 and Slm2, we measured the cellular levels of each protein by first constructing strains that expressed chromosomally GFP‐tagged forms of each gene. Both fusion proteins were found to be functional, as determined by the normal growth of slm1ΔSLM2‐GFP and SLM1‐GFPslm2Δ cells. Western blot analysis revealed that levels of Slm1 protein were approximately 10‐fold higher than those of Slm2 (Figure 2C), confirming that Slm1 is more abundant and likely plays a more significant role than Slm2 in cells. Consistent with these data, Slm1‐GFP was clearly visible at endogenous levels, while Slm2‐GFP fluorescence was more difficult to detect (see below).

Slm1 and Slm2 are new PI4,5P2 effectors

In contrast to the confluent plasma membrane localization of the PH domains from both Slm1 and Slm2 (Yu et al, 2004), localization of each full‐length protein showed they were enriched in punctate structures distributed at the cell periphery (Figure 3A and Supplementary Figure 2). The Slm1 patches were not polarized, as is the case with cortical actin patches. Time‐lapse video microscopy indicated only occasional movement of Slm1 on the mother cell cortex, while Slm1 patches were more dynamic on the daughter cell cortex (Supplementary Movie 1). Additionally, Slm1 patches were not affected by latrunculin A, a drug that causes actin depolymerization, indicating that Slm1 localization was independent of F‐actin (Figure 3A). Since the PH domains from both proteins showed localizations dependent on Mss4 activity, localization of Slm1‐GFP was analyzed in mss4ts cells. Following shift to the nonpermissive temperature, Slm1 appeared more soluble in mss4ts cells as compared to wild‐type cells treated identically, although a portion of Slm1 could still associate with cortical punctate structures (Figure 3B). These data suggest that Slm1 is at least in part recruited to the plasma membrane by PI4,5P2 and then potentially stabilized by other protein–protein or protein–lipid interactions.

Figure 3.

Localization of Slm1 is dependent on PI4,5P2 but not F‐actin. (A) Cells expressing SLM1‐GFP and ABP1‐RFP were visualized by fluorescence microscopy in the presence or absence of latrunculin A. (B) mss4ts cells expressing SLM1‐GFP were visualized by fluorescence microscopy following a 60 min shift to 37°C. (C) slm1Δslm2Δ cells expressing slm1KA‐GFP were visualized by fluorescence microscopy at 26°C.

To further investigate a role for PI4,5P2 in the recruitment of Slm1 to the plasma membrane, we generated a mutant form of Slm1 (slm1KA), which harbors point mutations in critical lysine resides (K483A, K487A) in the PH domain that were previously shown to be required for phosphoinositide binding (Yu et al, 2004). slm1Δslm2Δ double‐mutant cells expressing slm1KA were viable but temperature sensitive for growth, indicating an essential requirement for Slm1 phosphoinositide binding under conditions of stress. Strikingly, even at permissive temperatures, slm1KA‐GFP was largely mislocalized, accumulating mostly in the cytoplasm, with only a weak ability to associate with the plasma membrane, similar to mss4ts cells expressing wild‐type Slm1‐GFP (Figure 3C). Western blot analysis revealed that this form of Slm1 was stable, both at permissive and nonpermissive temperatures, indicating that cytoplasmic accumulation of GFP fluorescence was not due to clipping of the GFP fluorophore. Together, these data suggest a role for Mss4‐generated PI4,5P2 in the recruitment of Slm1 to the plasma membrane, and also indicate the existence of other factor(s) that contribute to Slm1 localization.

Slm1 and Slm2 are essential for normal actin cytoskeleton organization

To gain further insight into the function of Slm1 and Slm2, we constructed temperature‐sensitive forms of SLM1 using error‐prone PCR and plasmid shuffle techniques. Western blot analysis indicated that each form of Slm1 was stable, but three different patterns of Slm1 localization emerged from the screen (Table I). One group of temperature‐sensitive mutants localized similarly to wild‐type Slm1 at both permissive and nonpermissive temperatures (class A). In contrast, a second group of mutants showed normal localization at 26°C, but accumulated in the cytoplasm following shift to nonpermissive temperature (class B). Finally, a third group of mutants showed strong cytoplasmic fluorescence independent of temperature, similar to slm1KA described above (class C). Representative members from each class of mutants were chosen for sequencing. Both class A and B mutants harbored mutations between amino acids 192 and 436, a region we refer to as the Slm domain, due to its high conservation between Slm1 and Slm2 (Figure 1B). However, only class C mutants contained mutations in the PH domain (amino acids 469–581). Although mutations in the Slm domain could affect the localization of Slm1 (class B), its association with PI4,5P2 as determined by liposome binding assays was not significantly perturbed (Supplementary Figure 1). These data further indicate the importance of the Slm1 PH domain for localization, and also highlight an additional region of Slm1 as being important for both function and maintenance of Slm1 at the plasma membrane.

View this table:
Table 1. Three classes of slm1ts alleles each show different localizations

Previous studies have indicated a role for PI4,5P2 in the maintenance of actin cytoskeleton organization and cell wall integrity, at least in part through activation of the Rho1–Pkc1 pathway. To determine if Slm1 and Slm2 also function as downstream effectors of Mss4 in the maintenance of actin polarization, we examined actin localization in all three classes of slm1tsslm2Δ cells. At the permissive temperature, both wild‐type and slm1tsslm2Δ cells with relatively small buds showed normal polarization of the actin cytoskeleton, with cortical actin patches restricted to daughter cells and actin cables extending through mother cells. In contrast to wild‐type cells, at the nonpermissive temperature slm1tsslm2Δ cells exhibited a complete depolarization of the actin cytoskeleton, with actin patches randomly distributed throughout both mother and daughter cells and an apparent loss of actin cables (Figure 4A). Additionally, extended incubation of slm1tsslm2Δ cells at 37°C (greater than 3 h) resulted in cell lysis. These phenotypes were identical to those observed in mss4ts cells placed at the nonpermissive temperature.

Figure 4.

Slm1 and Slm2 control actin cytoskeleton organization and polarized secretion. (A) Wild‐type and slm1tsslm2Δ cells expressing either an activated allele of PKC1 or an empty vector were shifted to 37°C for 2 h, fixed and stained with rhodamine–phalloidin, and visualized by fluorescence microscopy. (B) Wild‐type and slm1tsslm2Δ cells expressing ABP140‐GFP were shifted to 37°C for 1 h and visualized by fluorescence microscopy. The arrows indicate long actin cables found in wild‐type cells. (C) slm1Δslm2Δ double‐mutant and slm1Δslm2Δsac7Δ triple‐mutant cells carrying pRS416SLM1 were grown on 5‐FOA for 4 days at 26°C. (D) Wild‐type and slm1tsslm2Δ cells expressing GFP‐SEC4 were shifted to 37°C for 1 h and visualized by fluorescence microscopy. Arrows indicate Sec4 localization to the bud or mislocalization to the mother cell cortex. (E) Wild‐type and slm1tsslm2Δ cells expressing GFP‐SNC1 were shifted to 37°C for 1 h and visualized by fluorescence microscopy.

Since actin cables were rather poorly preserved following fixation, we directly examined actin cable organization in living cells by visualizing a GFP fusion to the actin cable binding protein Abp140 (Yang and Pon, 2002). Although long actin cables oriented along the mother–bud axis were clearly visible in slm1tsslm2Δ cells at 26°C, they were absent following a 60‐min shift to the nonpermissive temperature (Figure 4B). Under identical conditions, wild‐type cells maintained actin cable orientation independent of temperature. Together, these data indicate a requirement for Slm1 and Slm2 in both actin patch polarization and actin cable stability/orientation.

Slm1 and Slm2 are important for polarization of the secretory pathway

To further investigate the functions of Slm1 and Slm2, we performed a gene dosage‐dependent suppression screen using class A slm1tsslm2Δ cells to identify proteins that when overexpressed could suppress the temperature sensitivity of the mutant. Analysis of 40 000 transformants allowed the isolation of five genes, each of which could reproducibly suppress the growth defect of slm1tsslm2Δ cells: SLM1, SLM2, PKC1, MSS4, and SEC4, a Rab‐type GTPase essential for protein secretion. Moreover, overexpression of these genes also suppressed the actin defects characteristic of slm1tsslm2Δ cells at 37°C, suggesting that increased signaling through the Rho1–Pkc1 cell integrity pathway could bypass the requirement for Slm1 and Slm2 (Figure 4A and Table II). Consistent with this idea, expression of PKC1R398P, encoding a constitutively activated form of Pkc1, allowed growth of cells completely lacking SLM1 and SLM2 (slm1Δslm2Δ) at 26°C. However, at elevated temperatures, slm1Δslm2Δ cells harboring PKC1R398P were no longer viable, indicating the existence of other factors necessary for bypass of Slm1 and Slm2. In contrast to the rescue of slm1Δslm2Δ cells by PKC1R398P, expression of constitutively activated alleles of BCK1 or MKK1, both downstream effectors of Pkc1, failed to suppress the phenotypes of slm1tsslm2Δ or slm1Δslm2Δ cells (Table II).

View this table:
Table 2. Genetic interactions in slm1tsslm2Δ cells

Since Pkc1 is an effector of Rho1, we tested whether elevated Rho1‐GTP levels were sufficient to bypass the requirement for Slm1 and Slm2 for cell viability at elevated temperature. Deletion of SAC7, a gene that encodes a Rho1 guanine nucleotide activating protein, has been shown to elevate levels of cellular Rho1‐GTP (Schmidt et al, 1997). Strikingly, in contrast to slm1Δslm2Δ double‐mutant cells, slm1Δslm2Δsac7Δ triple‐mutant cells were viable both at 26 and 38°C, indicating that Slm1 and Slm2 are dispensable under conditions where Rho1 is independently activated (Figure 4C).

Previous work has demonstrated a requirement for the Sec4‐dependent secretory pathway in transport of Rho1 (Abe et al, 2003). Interestingly, inactivation of Sec4 causes a defect in actin cytoskeleton organization (Mulholland et al, 1997), possibly due to a defect in transport of Rho1 (or other machinery necessary for actin organization) to sites of polarized growth. Since overexpression of Sec4 suppressed the growth defect of slm1tsslm2Δ cells, we examined whether Sec4 localization was perturbed in the absence of Slm1 and Slm2 activity. At 26°C, Sec4 localized to sites of polarized growth in slm1tsslm2Δ cells, similar to that observed in wild‐type cells. However, following a 60‐min shift to the nonpermissive temperature, Sec4 was mislocalized in greater than 75% of slm1tsslm2Δ cells, accumulating in the cytoplasm as well as in unpolarized patches decorating the entire cell periphery (Figure 4D). Since Sec4 transport requires polarized actin cables (Pruyne et al, 1998), these data are consistent with a primary defect in actin cytoskeleton organization when Slm1 and Slm2 function is compromised. Similar to the defect in Sec4 localization, examination of another secretory vesicle marker, the v‐SNARE Snc1, showed that it also became unpolarized in slm1tsslm2Δ cells at elevated temperature (Figure 4E). Increased signaling through the Rho1–Pkc1 pathway or overexpression of Sec4, which may elevate Rho1–Pkc1 signaling through enhanced delivery of Rho1 to the cortex, returned Snc1 to sites of polarized growth in slm1tsslm2Δ cells at elevated temperature (Table II). Together, these data indicate that appropriate Sec4 localization to sites of polarized growth becomes limiting in the absence of Slm1 and Slm2 activity, likely due to a primary defect in actin organization. Furthermore, Sec4 mislocalization potentially exacerbates the defect in actin polarization observed in slm1tsslm2Δ cells by virtue of its role in actin organization. Consistent with this idea, combining the temperature‐sensitive sec4‐8 mutation with the slm1tsslm2Δ mutations resulted in synthetic lethality at 28°C, a normally permissive temperature for sec4‐8 or slm1tsslm2Δ cells individually (Table II).

Phosphorylation of Slm1 and Slm2 requires TORC2 protein kinase activity

High‐throughput two‐hybrid analysis has previously suggested that Slm1 and Slm2 have a common set of interactors (Uetz et al, 2000; Ito et al, 2001). To test directly whether Slm1 functions as part of a complex, we used a tandem affinity purification (TAP) method to isolate Slm1 and any tightly associated proteins. From this analysis, we found only one other copurifying protein by mass spectrometry following stringent washing conditions, Slm2 (Figure 5A). Interestingly, we identified several electrophoretic species of both Slm1 and Slm2 in our purification. Western blot analysis directed against the S‐tag epitope fused to Slm1 (and used during its purification) revealed that only two of the four bands identified as being Slm1 by mass spectrometry contained the S tag. These data indicate that Slm1 can self‐associate in vivo. The existence of multiple forms of Slm1 and Slm2 suggested that both proteins undergo post‐translational modification. To test this directly, we immunoprecipitated newly synthesized Slm1‐GFP and Slm2‐GFP after briefly labeling cells with 35S and chasing with unlabeled cysteine and methionine. After 10 min of labeling, we detected single bands corresponding to Slm1 and Slm2, but following a brief period of chase, we observed a shift in the molecular weight of Slm1, which migrated as two distinct, higher molecular weight species, and Slm2, which migrated as a single, higher molecular weight species (Figure 5B). Due to its higher level of expression, we focused on Slm1 to determine if the gel shift corresponded to phosphorylation.

Figure 5.

Slm1 and Slm2 are phosphoproteins that form a tightly associated complex in vivo. (A) Silver‐stained SDS–PAGE gel showing the Slm1/Slm2 complex following TAP. Slm1‐S refers to Slm1 fused to the S tag used during purification. Slm1‐S* is a breakdown product of Slm1‐S as detected by Western blot analysis using antibodies directed against the S tag. (B, D) Cells expressing SLM1‐GFP or SLM2‐GFP were metabolically labeled for 10 min with 35S‐labeled cysteine and methionine and chased for the indicated time. Extracts were immunoprecipitated with anti‐GFP antibodies, incubated in the presence or absence of shrimp alkaline phosphatase (SAP), and subjected to SDS–PAGE analysis followed by autoradiography. The asterisks indicate the phosphorylated forms of Slm1. (C) Cells expressing SLM1‐GFP were metabolically labeled for the indicated time with 32P‐labeled orthophosphate. Extracts were immunoprecipitated with anti‐GFP antibodies, followed by SDS–PAGE analysis and autoradiography.

Cells expressing Slm1‐GFP were labeled with 32P, and extracts were immunoprecipitated with GFP antibodies. This analysis indicated that Slm1 was phosphorylated on at least two sites in vivo (Figure 5C). Additionally, treatment of 35S‐labeled cell extracts with phosphatase eliminated the gel mobility shift, confirming that the shift in Slm1 gel mobility was specifically due to phosphorylation (Figure 5D). Similar studies with Slm2 confirmed that it is also phosphorylated (data not shown).

Since the Rho1–Pkc1 pathway responds to heat stress by first causing depolarization of the actin cytoskeleton followed by a period of repolarization (Delley and Hall, 1999), we also analyzed the effect of elevated temperature on the phosphorylation of Slm1. Heat shock significantly reduced the phosphorylation of Slm1 normally observed at low temperature. However, Slm1 phosphorylation recovered at later time points (Figure 6A), which correlated with the timing of actin cytoskeleton repolarization, again suggesting a link between Slm1 activity and organization of the actin cytoskeleton.

Figure 6.

Slm1 associates with the TORC2 component Avo2 and depends on TORC2 for appropriate phosphorylation. (A) Wild‐type cells expressing SLM1‐GFP were metabolically labeled for 10 min with 35S‐labeled cysteine and methionine, chased for 30 min and shifted to 37°C for the indicated time. Extracts were immunoprecipitated with anti‐GFP antibodies and subjected to SDS–PAGE analysis followed by autoradiography. (B) Extracts from cells expressing either AVO2‐MYC alone or AVO2‐MYC and SLM1‐GFP were immunoprecipitated using α‐GFP antibodies under varying salt conditions. Total Avo2 protein in the extract is shown below. (C) Wild‐type or avo2Δ cells expressing SLM1‐GFP were metabolically labeled as in Figure 5A. Below each gel is the percentage of dually phosphorylated Slm1‐GFP as determined by densitometry. (D) Wild‐type or avo2Δ cells expressing SLM1‐GFP were metabolically labeled as in Figure 5C. (E, F) tor2ts and mss4ts cells expressing SLM1‐GFP were metabolically labeled as in (A). (G) slm1Δslm2Δ cells expressing SLM1‐GFP or slm1KA‐GFP were treated as in Figure 5B. (H) Wild‐type cells or cells grown for 16 h in glucose‐containing media and expressing TOR2 under the control of an integrated GAL1‐10 promoter (to repress expression of genomic TOR2) were metabolically labeled for 10 min with 35S‐labeled cysteine and methionine and chased for the indicated time in the presence or absence of rapamycin (200 ng/ml). Extracts were treated as in Figure 5A.

To determine which kinase(s) phosphorylated Slm1, we first analyzed results from high‐throughput two‐hybrid analyses, which suggested that both Slm1 and Slm2 may interact with Avo2, a component of the Tor2 protein kinase‐containing TORC2 complex, required for proper actin cytoskeleton organization (Loewith et al, 2002). Although TAP of Slm1 failed to indicate an interaction with Avo2, immunoprecipitation of Slm1‐GFP from cells expressing Avo2‐Myc showed that Avo2 could associate with Slm1. This interaction was extremely sensitive to salt conditions used during TAP, suggesting that Slm1 and Avo2 only weakly associate (Figure 6B). While mutations in the Slm domain of Slm1 failed to perturb this interaction, mutations in the PH domain of Slm1 resulted in a reduced association between Slm1KA and Avo2 (Supplementary Figure 3). However, the weakened interaction may be an indirect consequence of Slm1KA mislocalization. Additionally, SGA analysis of the mss4ts mutation revealed a genetic interaction with AVO2, which was confirmed in the SEY6210 background (Figure 1A and Supplementary Table I). Consistent with a role for TORC2 in phosphorylation of Slm1, avo2Δ cells exhibited a kinetic defect in Slm1 phosphorylation (Figure 6C and D). Moreover, tor2ts cells failed to show recovery of Slm1 phosphorylation at the nonpermissive temperature, indicating that Tor2 function is required for this process (Figure 6E). Since PI4,5P2 also plays a role in Slm1 recruitment/activation, we determined the effect of Mss4 inactivation on Slm1 phosphorylation. As in tor2ts cells, Slm1 phosphorylation failed to recover in the absence of Mss4 activity, suggesting that PI4,5P2 and Tor2 are both required for localization and phosphorylation of Slm1 (Figure 6F). Consistent with these data, the mutant form of Slm1 harboring mutations in its PH domain was not normally phosphorylated, likely due to its inability to target efficiently to the plasma membrane (Figure 6G).

Tor2 has previously been demonstrated to function in two different protein complexes, one required for the response to nutrient availability (TORC1), while the other regulates actin cytoskeleton organization (TORC2). Studies have also indicated that TORC1 is uniquely sensitive to the effects of the drug rapamycin (Loewith et al, 2002). To determine if Slm1 phosphorylation requires TORC1, we treated cells with rapamycin and examined phosphorylation of Slm1. Following a 30 min pretreatment with rapamycin, we found that Slm1 phosphorylation was unperturbed, suggesting that TORC2, but not TORC1, multiply phosphorylates Slm1 (Figure 6H). Consistent with these data, cells specifically repressed for TOR2 expression showed a defect in Slm1 phosphorylation (Figure 6H).

TORC2 phosphorylates Slm1 and Slm2 in vitro and is partially required for Slm1 localization

To investigate further a role for TORC2 in phosphorylation of Slm1 and Slm2, recombinant forms of Slm1 and Slm2 were purified from bacteria and subjected to an in vitro kinase assay. In the presence of purified TORC2, both GST‐Slm1 and GST‐Slm2, but not GST alone, were phosphorylated, similar to a previously characterized mammalian TOR substrate, 4EBP1 (Figure 7A). However, purified TORC2 containing a kinase‐dead form of Tor2 failed to phosphorylate Slm1 or Slm2, suggesting that a contaminant from the purification was not responsible for the observed kinase activity.

Figure 7.

Slm1 phosphorylation is TORC2 dependent in vitro. (A) SW80‐1D cells expressing TOR2 under the control of an integrated GAL1‐10 promoter and TAP‐tagged AVO2 were transformed with either vector, (HA)3TOR2 (TOR2) or (HA)3TOR2D2998E (TOR2KD). After growth for 16 h in glucose‐containing media (to repress expression of genomic TOR2), TORC2 (via Avo2‐TAP) was precipitated with IgG‐Sepharose beads and tested for Slm1, Slm2, and 4EBP1 kinase activity. Shown are autoradiographs (Phos‐Slm1, Phos‐Slm2, Phos‐4EBP1) and the corresponding Coomassie blue‐stained region of the SDS–PAGE gel (Slm1, Slm2, 4EBP1). The bottom panel is a Western blot showing the amount of HA‐tagged TOR2/TOR2KD added to the respective kinase assays. (B) Wild‐type or tor2ts cells expressing slm1KA‐GFP were visualized by fluorescence microscopy following a 90 min shift to 37°C.

Finally, to investigate a function for Slm1 phosphorylation by TORC2, we tested whether the localization of Slm1 depends on Tor2 activity. Since our studies have shown a partial role for PI4,5P2 in the localization of Slm1, we examined whether the remaining plasma membrane association of Slm1 was dependent on Tor2. Therefore, wild‐type and tor2ts cells expressing a mutant form of Slm1 that cannot interact with PI4,5P2 (slm1KA‐GFP) were visualized following a 90 min shift to 37°C. In contrast to wild‐type cells, tor2ts cells failed to exhibit any localization of slm1KA‐GFP to the cell periphery. Cells lacking other plasma membrane‐associated protein kinase activities, including pkc1ts, bck1Δ, slt2Δ, tpk1Δtpk2tstpk3Δ, ypk1tsypk2Δ, and pkh1tspkh2Δ cells, did not affect slm1KA‐GFP localization (Figure 7B and data not shown). Together, these data suggest that Slm1 and Slm2 function downstream of TORC2 in signaling to the actin cytoskeleton, and represent the first substrates of TORC2 protein kinase activity.


We and others have previously shown that regulation of PI4,5P2 synthesis at the plasma membrane is critical for regulation of the actin cytoskeleton, at least in part through activation of the Rho1 GTPase (Desrivières et al 1998; Homma et al, 1998; Audhya and Emr, 2002). Using an unbiased approach to identify additional factors involved in PI4,5P2 signaling, we have identified two new PI4,5P2 effectors, Slm1 and Slm2, which also appear to be critical for regulation of actin cytoskeleton organization. Furthermore, our screen showed that deletion of AVO2, which encodes a nonessential component of the Tor2 protein kinase‐containing TOR complex 2, is lethal in combination with the mss4ts mutation, suggesting that TORC2 and Mss4 function together in regulating actin organization. Consistent with this hypothesis, overexpression of MSS4 suppresses a tor2 mutation (Helliwell et al, 1998), and both Slm1 and Slm2 are substrates of the TORC2 kinase. Together, our data suggest the existence of multiple PI4,5P2 effectors in yeast, whose actions can be coordinated through additional signaling events to regulate polarity of the actin cytoskeleton. Importantly, Slm1 and Slm2 are the first effectors identified that require both TORC2‐dependent phosphorylation and Mss4‐dependent recruitment to control the actin cytoskeleton, possibly through activation of Rho1. We speculate that Mss4 functions as a sensor for cell stress, such as heat shock, generating variable levels of PI4,5P2 to alter actin polarization and cell wall integrity, while other inputs, such as TORC2 kinase regulation, further modulate the magnitude/duration of these changes.

Multiple roles for PI4,5P2 revealed by synthetic genetic array analysis

Previous studies have established that PI4,5P2 plays essential roles in both actin cytoskeleton organization and endocytosis, and several effectors of PI4,5P2 have been identified that are important for these processes (Takenawa and Itoh, 2001). Using an unbiased, genome‐wide‐based approach in yeast, we have identified several new genes that may be involved in regulating PI4,5P2 synthesis or that function downstream of PI4,5P2 (Figure 8). When combined with the mss4ts mutation, several components of two protein chaperone complexes, the Hsp70–Hsp90 complex and the GIMC complex (Siegers et al, 1999), became essential for cell viability, suggesting a role for PI4,5P2 in the response to heat shock. Consistent with this interpretation, studies have indicated that PI4,5P2 levels transiently increase following heat shock, potentially required for preventing protein misfolding through activation of a chaperone complex (Desrivières et al, 1998; Audhya et al, 2000). However, further studies are required to determine whether PI4,5P2 influences chaperone function during heat stress.

Figure 8.

Regulation of actin cytoskeleton organization by PI4,5P2. Genetic interactions identified in the Mss4 SGA analysis indicate multiple roles for Mss4‐mediated PI4,5P2 production in actin organization. In particular, PI4,5P2 recruitment/activation of the Slm1–Slm2 complex is required for maintenance of actin cytoskeleton polarity. Additional factors including TORC2 protein kinase activity further modulate and specify Slm1–Slm2 function in actin organization.

Additionally, the SGA analysis uncovered a number of proteins directly involved in actin organization and cell polarity. Of particular interest was the identification of three proteins that harbor PH domains, Rom2, Boi2, and Slm1. Rom2, an exchange factor for the Rho1 GTPase, has been shown to bind PI4,5P2 and function downstream of Mss4 in activation of the Rho1–Pkc1 pathway (Audhya and Emr, 2002). Similarly, the PH domain of Boi2, a protein important for polarized growth and bud formation, has been shown to interact with phosphoinositides in vitro (Yu et al, 2004). Additionally, a functionally redundant homolog of Boi2, Boi1, has been shown to interact with PI4,5P2 via its PH domain, and this lipid–protein interaction in part mediates its normal localization and is essential in the absence of Boi2 (Hallett et al, 2002).

Beyond the identification of known PI4,5P2 binding proteins, SGA analysis also identified a new PI4,5P2 effector, Slm1. Together with its homolog and binding partner Slm2, these proteins are essential for maintaining proper actin cytoskeleton organization and cell wall integrity, possibly through regulation of the Rho1 GTPase (Figure 8). However, we were unable to identify a direct interaction between Slm1 or Slm2 and known regulators of Rho1, including its exchange factors (Tus1, Rom1, and Rom2) or its activating proteins (Sac7, Bem2, and Bem3). In an alternative model, Slm1 and Slm2 may function in parallel to Rho1 in regulation of actin organization. Genome‐wide two‐hybrid analysis suggests that Slm2 interacts with both subunits of calcineurin, a Ca2+/calmodulin‐dependent protein phosphatase required during environmental stress, which may function in actin organization (Uetz et al, 2000; Ito et al, 2001; Cyert, 2003). Further studies will be necessary to determine the extent to which calcineurin may function in Slm1/Slm2 signaling.

Coordination of Mss4 and the TORC2 protein kinase in actin organization

Previous studies in organisms ranging from insects to humans have indicated a relationship between growth factor‐stimulated phosphoinositide lipid signaling and nutrient‐dependent TOR protein kinase signaling in the regulation of cell growth (Jacinto and Hall, 2003). We have found that this interaction may also extend to yeast, where PI4,5P2 and Tor2 both function to control the actin cytoskeleton. Yeast express two forms of TOR: TORC1, which contains either Tor1 or Tor2, mediates rapamycin‐sensitive growth in response to nutrients, while TORC2, which contains Tor2 but not Tor1, controls polarization of the actin cytoskeleton. However, substrates of TORC2 that affect actin organization have remained elusive. Our findings support a role for Slm1 and Slm2 in this capacity. Additionally, the physical interaction between Slm1 and Avo2, the only nonessential component of either TORC complex, suggests a role for this component of TORC2 in the recruitment of effector proteins. In the absence of Avo2, PI4,5P2 is likely to be sufficient for recruitment of Slm1 and Slm2 to sites of TORC2 activity. However, loss of multiple inputs important for TORC2‐dependent phosphorylation cannot be tolerated, as demonstrated by the lethality exhibited by mss4tsavo2Δ double‐mutant cells. Interestingly, Rom2, an exchange factor for Rho1, has also been identified as an effector of PI4,5P2 and potentially Tor2, suggesting that integration of phosphoinositide‐dependent recruitment and regulation by TOR may not be limited to Slm1 and Slm2.

Our analysis has successfully identified two novel proteins that function as essential regulators of actin polarity downstream of PI4,5P2. We speculate that under conditions of cell stress such as heat shock, elevated PI4,5P2 levels result in the increased recruitment and/or activation of effector proteins including Slm1, Slm2, and Rom2, all necessary for regulation of actin cytoskeleton organization. Moreover, to ensure both rapid and specific responses to cell stress, additional factors are necessary. In the case of PI4,5P2‐dependent Rom2 recruitment/activation to the cell cortex, the integral plasma membrane cell surface sensors, Wsc1 and Mid2, appear to serve this function (Audhya and Emr, 2002). Correspondingly, we have found that PI4,5P2‐dependent Slm1 and Slm2 recruitment requires the parallel activity of the TORC2 protein kinase complex, potentially also dependent on conditions of nutrient availability. Interestingly, following heat shock, the levels of Slm1 and Slm2 phosphorylation were diminished, likely due to reduced TORC2 activity and/or elevated protein phosphatase activity. This decreased level of Slm1 and Slm2 phosphorylation may be important for appropriate depolarization of the actin cytoskeleton and secretion observed following cell stress. Strikingly, phosphorylation of Slm1 and Slm2 recovered in parallel to actin repolarization, suggesting a link between Slm1 and Slm2 activity and actin cytoskeleton organization. Collectively, these data indicate that PI4,5P2 functions together with both positive and negative control elements in a cellular stress response system necessary for organizing the actin cytoskeleton.

Materials and methods

Strains and media

Enzymes used for recombinant DNA techniques were used as recommended by the manufacturer. Standard recombinant DNA techniques and yeast genetics methods were performed, and growth media used have been described elsewhere (Maniatis et al, 1982; Gaynor et al, 1994). S. cerevisiae strains used in this study are listed in Table III. All gene disruptions and temperature‐sensitive mutant strains were generated similarly, as described previously (Audhya and Emr, 2002). Integrated epitope tagging was performed as described previously (Longtine et al, 1998).

View this table:
Table 3. S. cerevisiae strains used in this study

Synthetic genetic array analysis and yeast plasmids

SGA analysis was performed as described previously (Tong et al, 2001). Strains expressing temperature conditional alleles of slm1 were generated similarly as described using error‐prone PCR (Audhya et al, 2000). All point mutations were created using site‐directed mutagenesis.

Metabolic labeling and immunoprecipitation

Cell labeling and immunoprecipitations were performed as described previously (Gaynor et al, 1994; Audhya et al, 2000, 2002). For 32P labeling, cells were grown in low phosphate media to early log phase and labeled with 1 mCi 32P‐labeled orthophosphate in the absence of phosphate. Treatment with shrimp alkaline phosphatase and immunoprecipitation with antiserum against GFP have been described previously (Audhya and Emr, 2003).

Fluorescence microscopy

All fluorescence images were observed using a Zeiss Axiovert S1002TV fluorescent microscope and subsequently processed using a Delta Vision deconvolution system. For actin localization, cells were fixed and stained with rhodamine–phalloidin (Molecular Probes) as described previously (Audhya et al, 2000). To depolymerize F‐actin, latrunculin A was added from a 10 mM DMSO stock to a final concentration of 200 μM for 30 min.

Tandem affinity purification and mass spectrometric analysis

TAP of Slm1 was performed as described previously (Cheeseman et al, 2002). Proteins in the SDS gel were stained by silver and protein bands were excised and digested by trypsin overnight. Peptides were extracted and analyzed by microcapillary liquid chromotography and tandem mass spectrometry. An in‐house MS system consisting of an HPLC and LCQ‐ion trap mass spectrometer (Thermo Finnigan) was used. All identified peptides were further manually inspected and verified.

TORC2 purification and in vitro kinase assay

SW80‐1D cells (5 l) expressing either vector alone, HA‐TOR2, or a kinase‐inactive form of HA‐TOR denoted HA‐TOR2KD (Jiang and Broach, 1999) pregrown to saturation in synthetic raffinose/glycerol medium were additionally grown at 30°C for 16 h in media containing glucose. Cells were lysed as described previously (Loewith et al, 2002). Pooled lysates were cleared, normalized to ∼50 ml and ∼375 mg protein and passed over 125 μl of Sepharose CL‐4B (Sigma) previously equilibrated in lysis buffer. To the flow through was added 150 μl IgG‐Sepharose (Amersham Bioscience) for 2 h. IgG beads were then incubated with either ∼5 μg GST, ∼2 μg GST‐Slm2, ∼2 μg GST‐Slm1, or 2 μg PHAS‐I (Stratagene) in 50 μl of lysis buffer with 20% glycerol. In all, 6 μl of 10 × buffer (40 mM MnCl2, 100 mM dithiothreitol, 10 × Roche protease inhibitor cocktail–EDTA, 100 mM NaN3, 100 mM NaF, 100 mM p‐nitrophenylphosphate, 100 mM β‐glycerophosphate) was added before the reaction was started with the addition of 4 μl of ATP mix (1.2 mM ATP, 2.5 μCi/μl [γ32P]ATP (3000 Ci/mmol) in kinase assay buffer). Samples were then subjected to SDS–PAGE.

Supplementary data

Supplementary data are available at The EMBO Journal Online.

Supplementary Information

Supplementary Figure 1 [emboj7600384-sup-0001.tiff]

Supplementary Figure 2 [emboj7600384-sup-0002.tiff]

Supplementary Figure 3 [emboj7600384-sup-0003.tiff]

Supplementary Table 1 [emboj7600384-sup-0004.doc]

Supplementary Movie 1 [emboj7600384-sup-0005.mpg]


We are grateful to Drs Arshad Desai, Mark Lemmon, and Beth Weaver for many helpful suggestions on this manuscript. Also, we thank Stephan Wullschleger for strains and Wolfgang Oppliger and Iain Cheeseman for technical assistance. AA is currently supported by a Helen Hay Whitney postdoctoral fellowship at the Ludwig Institute for Cancer Research. RL was the recipient of an EMBO long‐term fellowship, and ABP holds a Natural Sciences and Engineering Research Council of Canada graduate student fellowship. This work was partially supported by grant CA58689 from the National Institutes of Health (to SDE), grants from the Canton of Basel and the Swiss National Science Foundation (to MNH), and grants from the Canadian Institute of Health Research and Genome Ontario (to CB). SDE is an investigator of the Howard Hughes Medical Institute.


View Abstract