Most of the phospholipid biosynthetic genes of Saccharomyces cerevisiae are coordinately regulated in response to inositol and choline. Inositol affects the intracellular levels of phosphatidic acid (PA). Opi1p is a repressor of the phospholipid biosynthetic genes and specifically binds PA in the endoplasmic reticulum. In the presence of inositol, PA levels decrease, releasing Opi1p into the nucleus where it represses transcription. The opi1 mutant overproduces and excretes inositol into the growth medium in the absence of inositol and choline (Opi− phenotype). To better understand the mechanism of Opi1p repression, the viable yeast deletion set was screened to identify Opi− mutants. In total, 89 Opi− mutants were identified, of which 7 were previously known to have the Opi− phenotype. The Opi− mutant collection included genes with roles in phospholipid biosynthesis, transcription, protein processing/synthesis, and protein trafficking. Included in this set were all nonessential components of the NuA4 HAT complex and six proteins in the Rpd3p–Sin3p HDAC complex. It has previously been shown that defects in phosphatidylcholine synthesis (cho2 and opi3) yield the Opi− phenotype because of a buildup of PA. However, in this case the Opi− phenotype is conditional because PA can be shuttled through a salvage pathway (Kennedy pathway) by adding choline to the growth medium. Seven new mutants present in the Opi− collection (fun26, kex1, nup84, tps1, mrpl38, mrpl49, and opi10/yol032w) were also suppressed by choline, suggesting that these affect PC synthesis. Regulation in response to inositol is also coordinated with the unfolded protein response (UPR). Consistent with this, several Opi− mutants were found to affect the UPR (yhi9, ede1, and vps74).
TRANSCRIPTION of the phospholipid biosynthetic structural genes in Saccharomyces cerevisiae is regulated by inositol and choline (Paltauf et al. 1992; Greenberg and Lopes 1996; Henry and Patton-Vogt 1998; Carman and Henry 1999; Santiago and Mamoun 2003; Jesch et al. 2005). Gene expression is maximally repressed in the presence of inositol and choline and derepressed when they become limiting. This regulation requires the concerted effort of several transcription factors, including Ino2p, Ino4p, Opi1p, the Ume6p–Sin3p–Rpd3p histone deacetylase (HDAC) complex, the ISW2 and INO80 chromatin-remodeling complexes, and Mot1p (White et al. 1991; Ambroziak and Henry 1994; Nikoloff and Henry 1994; Jackson and Lopes 1996; Rundlett et al. 1996, 1998; Kadosh and Struhl 1997, 1998; Elkhaimi et al. 2000; Shen et al. 2000; Fazzio et al. 2001; Dasgupta et al. 2005). Ino2p and Ino4p belong to a family of bHLH regulatory proteins, which form a heterodimer that binds to a UASINO sequence to activate transcription of most phospholipid biosynthetic genes (e.g., INO1, CHO1, CHO2, and OPI3 in Figure 1) (Santiago and Mamoun 2003; Jesch et al. 2005). The Ume6p–Sin3p–Rpd3p HDAC complex, the ISW2 and INO80 chromatin-remodeling complexes, and Mot1p are global regulators that play a negative role in phospholipid biosynthetic gene expression (Jackson and Lopes 1996; Rundlett et al. 1996, 1998; Kadosh and Struhl 1997, 1998; Elkhaimi et al. 2000; Shen et al. 2000; Fazzio et al. 2001; Dasgupta et al. 2005). Opi1p was the first repressor found to specifically regulate the phospholipid biosynthetic pathway.
The OPI1 locus was identified in a screen for mutants that overproduce and excrete inositol (Opi) into the growth medium in the absence of inositol and choline (Opi− phenotype) (Greenberg et al. 1982). The Opi− phenotype correlates with constitutive overexpression of the INO1 gene (Hirsch and Henry 1986), which is responsible for de novo synthesis of inositol (Figure 1) (Culbertson and Henry 1975). Therefore it is not surprising that several other mutants defective in repression of INO1 also have an Opi− phenotype. These include mutations in the UME6, SIN3, and RPD3 global repressor genes (Elkhaimi et al. 2000). The link between regulation of INO1 gene expression and the Opi− phenotype is further supported by the existence of the dominant OPI5+ mutant allele (Swift and McGraw 1995). This mutant is a deletion of the INO1 promoter (renamed the INO1-100 allele), which yields constitutive derepressed expression of the INO1 gene (Swift and McGraw 1995).
Recently, our understanding of the role of Opi1p in the repression of phospholipid biosynthetic gene expression has improved dramatically. Opi1p is tethered in the endoplasmic reticulum (ER) by Scs2p, an integral membrane protein (Gavin et al. 2002; Kagiwada and Zen 2003; Loewen et al. 2003, 2004; Loewen and Levine 2005). Repression in response to inositol and choline is mediated by the level of phosphatidic acid (PA) (Figure 1). In the absence of inositol, PA levels are elevated and Opi1p binds PA (Loewen et al. 2004). When inositol is added, PA levels decrease, releasing Opi1p from the ER. Opi1p then translocates to the nucleus where it inhibits transcription by presumably interacting with the Ino2p transcriptional activator (Wagner et al. 2001; Gardenour et al. 2004; Heyken et al. 2005). The addition of choline by itself has little effect on PA levels. However, in combination with inositol, choline further reduces PA levels, resulting in additional repression (Henry and Patton-Vogt 1998). Not surprisingly, blocks in de novo phosphatidylcholine (PC) biosynthesis that elevate PA levels also yield an Opi− phenotype (Klig et al. 1988; Summers et al. 1988; McGraw and Henry 1989; Shen and Dowhan 1996). Thus, cds1, cho2, and opi3 mutants all have the Opi− phenotype (Figure 1). However, the Opi− phenotype of these mutants is conditional because it can be suppressed by adding choline (C) to the medium. Choline restores PC synthesis through a salvage pathway (Kennedy pathway), thereby alleviating the accumulation of PA caused by the block in the de novo PC pathway (Figure 1) (Henry and Patton-Vogt 1998).
Opi1p function is also regulated by post-translational modifications. Opi1p is phosphorylated at Ser26 by protein kinase C (Sreenivas et al. 2001), at Ser31 and Ser251 by protein kinase A (Sreenivas and Carman 2003), and at Ser10 by casein kinase II (Chang and Carman 2006). Mutations at each of these phosphorylation sites show that phosphorylation of Opi1p plays a role in regulation of target gene expression in response to inositol.
Recently, it has been shown that under derepressing conditions the INO1 promoter is recruited to the nuclear membrane where Ino4p and Ino2p activate its transcription (Brickner and Walter 2004). Under repressing conditions, Opi1p translocates to the nucleus and presumably interacts with Ino2p (Wagner et al. 2001; Gardenour et al. 2004; Heyken et al. 2005). The interaction with Ino2p is suggested by in vitro experiments (Wagner et al. 2001; Heyken et al. 2005) and the existence of dominant INO2c mutations that yield an Opi− phenotype (Gardenour et al. 2004; Heyken et al. 2005).
Phospholipid biosynthesis is coordinated with the unfolded protein response (UPR) (Cox et al. 1997; Chang et al. 2002; Jesch et al. 2005). The UPR is initiated in the ER in response to the accumulation of unfolded proteins (Patil and Walter 2001; Schroder and Kaufman 2005). The UPR can also be induced by depletion of inositol (Cox et al. 1997; Chang et al. 2002). Upon UPR induction, Ire1p is activated, initiating splicing of HAC1 mRNA (Sidrauski and Walter 1997). The properly spliced HAC1 transcript produces the Hac1p basic leucine zipper transcription factor that binds to the UPR element (UPRE) of target genes such as KAR2 but also regulates UASINO-containing promoters by counteracting the function of Opi1p (Cox and Walter 1996).
Recent studies show that screening the viable yeast deletion set (VYDS) (Winzeler et al. 1999; Giaever et al. 2002) can yield valuable insight into well-studied processes such as regulation in response to phosphate concentration (Huang and O'Shea 2005). To further understand the repression of phospholipid biosynthesis, the VYDS was screened to identify mutants that display the Opi− phenotype. These mutants were characterized with respect to the severity of the phenotype, dependence on de novo PC biosynthesis, and defects in the UPR.
MATERIALS AND METHODS
Yeast strains, media, and growth conditions:
Yeast strains used in this study include the VYDS in the BY4742 background (MATα his3Δ1 leu2Δ0 lys2Δ0 ura3Δ0) (Winzeler et al. 1999; Giaever et al. 2002) and BRS1005 (AID-1) (MATa/MATα, ade1/ade1, ino1-13/ino1-13) (White et al. 1991). Yeast cultures were grown at 30° in complete synthetic medium (Kelly and Greenberg 1990) containing 2% glucose (w/v) but lacking inositol and choline (I−C−). Where indicated, I−C− medium was supplemented with 75 μm inositol (I+) ± 1 mm choline (C+), 1 mm ethanolamine (E+), 1 mm monomethylethanolamine (MME+), or 1 mm dimethylethanolamine (DME+). X-gal medium was prepared as previously described (Hudak et al. 1994) and supplemented with tunicamycin (Tm) (1 μg/ml) to induce the UPR (Cox and Walter 1996).
Plasmid pJC104 containing four copies of the UPRE upstream of the CYC1-lacZ reporter was a gift from Peter Walter (Howard Hughes Medical Institute, University of California, San Francisco) (Cox and Walter 1996). Plasmid pJH330 has been previously described (Elkhaimi et al. 2000) and contains 543 bp of the INO1 promoter and 132 codons of the INO1 ORF fused in frame to the lacZ reporter in YEp357R (Myers et al. 1986). The OPI1-cat reporter plasmid (pMK162) has been described (Kaadige and Lopes 2003). A 7.2-kb SstII–ClaI fragment containing the OPI1-cat reporter (525 bp of the OPI1 promoter) was excised from pMK162 and used to transform relevant yeast strains. This results in single-copy integration at the GAL4 locus. Yeast transformations were performed using the YEASTMAKER transformation system 2 (BD Biosciences Clontech, Mountain View, CA).
Opi− phenotype plate assay:
The Opi− phenotype was assessed by a cross-feeding assay using an inositol auxotrophic strain. The original assay (Greenberg et al. 1982; McGee et al. 1994) was used to screen the VYDS. Briefly, the ino1 homozygous diploid tester strain BRS1005 was spread on I−C− medium. Strains from the VYDS were pinned at a density of 48 strains per plate. Inositol excretion by the VYDS strains results in a ring of growth of the BRS1005 tester strain.
A second round of screening tested each of the Opi− mutants individually. Each Opi− mutant strain was streaked horizontally on a plate containing I−C− medium. After 3 days of growth at 30°, the tester strain BRS1005 was streaked perpendicularly to the Opi− strain. Inositol excretion results in growth of the BRS1005 tester strain (McGee et al. 1994).
β-Galactosidase and chloramphenicol acetyltransferase (Cat) assays were performed as described previously (Lopes et al. 1991; Kaadige and Lopes 2006). Units of β-galactosidase activity = (OD420/minute/milligram total protein) × 1000. Units of Cat activity = percentage of conversion of 14C-chloramphenicol/microgram total protein/time (in hours) (Griggs and Johnston 1991).
A genomewide screen identified 89 Opi− mutants:
Three versions of the Opi test have been reported (Greenberg et al. 1982; Klig et al. 1988; McGee et al. 1994). The original Opi− screen was performed by replica plating yeast, mutagenized with ethyl methane sulfonate, onto synthetic medium lacking inositol that had been spread with a lawn of the AID-1 indicator strain. AID-1 is an inositol auxotroph (ino1-13) that also harbors the ade1 marker, which allows identification of inositol-excreting mutants by a red halo of growth around the mutant strain. Four complementation groups were originally identified, OPI1–OPI4 (Greenberg et al. 1982). While the OPI1 and OPI3 genes have been studied extensively, the opi2 and opi4 mutants were not further investigated. Subsequent genetic screens identified several other Opi− mutants, including cdg1 (CDS1 gene) (Klig et al. 1988), cho2 (Summers et al. 1988), ume6, rpd3, sin3 (Elkhaimi et al. 2000), and reg1 (Ouyang et al. 1999).
The VYDS mutant collection was screened for the Opi− phenotype using the original version of the Opi− test. The VYDS collection was pinned onto plates lacking inositol and choline that were spread with a suspension of the BRS1005 (AID-1) indicator strain. It was necessary to pin the collection at a density of 48 mutants per plate to create sufficient separation between strains. This screen identified 218 Opi− mutants. However, in some cases, the red halo of growth covered several adjacent mutant strains, obscuring which strains actually displayed the Opi− phenotype. Thus, a second screen was conducted using a more recent adaptation of the Opi test (McGee et al. 1994). Each of the 218 Opi− mutants were individually streaked onto plates with medium lacking inositol and choline and incubated at 30° for 3 days. A suspension of the BRS1005 tester strain was streaked perpendicularly to the mutant and incubated for an additional 3 days. This second screen reduced the number of Opi− mutants to 89 (Table 1).
It is important to note that inositol auxotrophic strains starved for inositol die, a phenomenon termed “inositol-less death” (Henry et al. 1977). Consequently, spreading the BRS1005 strain on medium lacking inositol as was done here, and in the original screen (Greenberg et al. 1982), could have affected its viability. However, there is a window of opportunity, lasting approximately one generation in liquid medium, where cells survive and continue to divide normally (Henry et al. 1977). Nevertheless, reduced viability of the BRS1005 indicator strain may have precluded identification of mutants with a weak Opi− phenotype.
As expected, the VYDS collection screen identified opi1, cho2, opi3, ume6, rpd3, sin3, and reg1 (Table 1). Several previously isolated Opi− mutants were not identified for predictable reasons. CDS1 is involved in synthesizing cytidine diphosphate–diacylglycerol (CDP–DAG) from PA (Figure 1) and was originally defined by a point mutation (cdg1) that displayed the Opi− phenotype (Klig et al. 1988; Shen and Dowhan 1996). However, the cds1 null allele is inviable (Shen et al. 1996) and therefore not present in the VYDS. CHO1 converts CDP–DAG and serine to phosphatidylserine (PS) (Figure 1) and displays the Opi− phenotype under the conditions tested here, but the cho1 mutant was not present in the VYDS. PSD1 (Clancey et al. 1993; Trotter et al. 1993) and PSD2 (Trotter et al. 1995) are involved with decarboxylating PS to phosphatidylethanolamine (PE) (Figure 1). These genes encode redundant functions and therefore only the double mutant would be expected to yield the Opi− phenotype. MOT1 and SMP2 have also recently been shown to be negative regulators of INO1 expression (Dasgupta et al. 2005; Santos-Rosa et al. 2005), suggesting that they may have an Opi− phenotype. However, MOT1 is an essential gene and the smp2 mutant was not present in the VYDS.
The Opi− mutant collection identified several biological functions:
The VYDS contains 4827 strains. Of the 4827 deletion strains, 89 (∼1.8%) displayed the Opi− phenotype (Table 1). Of the 89 Opi− phenotype mutants, ∼90% of the genes have been assigned to a particular biological process (http://db.yeastgenome.org/cgi-bin/go/gotermmapper) (Figure 2). We used the GoTermMapper program to determine if any of the functional categories of genes were overrepresented in the Opi− mutant set. There was an overrepresentation of Opi− mutants in vesicle-mediated transport (30% observed, 4% expected), transport (53% observed, 15% expected), transcription (15% observed, 7% expected), protein modification (11% observed, 7% expected), protein catabolism (12% observed, 3% expected), organelle organization and biogenesis (37% observed, 14% expected), and lipid metabolism (7% observed, 3% expected) (Figure 2).
The VYDS screen also identified two transcription factor complexes. The screen identified seven components of the Rpd3p–Sin3p HDAC. Deletions of RPD3, SIN3, and UME6 genes have been previously shown to yield an Opi− phenotype (Elkhaimi et al. 2000). The VYDS screen identified four more components of the HDAC complex: Pho23p (Loewith et al. 2001), Eaf3p (Reid et al. 2004; Gavin et al. 2006), Dep1p, and Rxt2p gene (Carrozza et al. 2005) (Table 1, Figure 3A). However, the HDAC complex has recently been shown to include six other factors not uncovered in the Opi− screen. These include Ash1p, Ume1p, Sap30p, Cti6p, Sds3p, and Rxt3p (Ho et al. 2002; Reid et al. 2004; Carrozza et al. 2005; Gavin et al. 2006). Thus, we tested mutants in the genes that encode each of these factors independently. This revealed that the ume1 and sap30 mutant strains also yielded the Opi− phenotype (Figure 3B) while the remaining four mutant strains did not (data not shown). The weak Opi− phenotype of the ume1 and sap30 mutant strains may have been the reason that these were not identified originally (Figure 3B). These two additional Opi− mutants increased the total number of Opi− mutants to 91.
In addition to the HDAC complex mutants, the screen also identified all of the nonessential components of the NuA4 histone acetyltransferase (HAT) complex: EAF3, EAF5, EAF7, VID21, and YAF9 (Table 1; Figure 4, A and B). This is curious since the NuA4 complex has HAT activity, which is responsible for increased gene expression. Thus, mutations in this complex should yield reduced expression of the target genes. One possible explanation for the Opi1 phenotype is that the NuA4 complex may act indirectly on INO1 expression by increasing OPI1 expression. This was a reasonable assumption since altered expression of OPI1 has been shown to affect target gene expression (Wagner et al. 1999). To test this possibility, we assayed the effect of several NuA4 component mutants on expression of an OPI1-cat reporter (Figure 4C). As reported, the expression of the OPI1-cat reporter was repressed twofold by the addition of inositol and choline (Jiranek et al. 1998; Wagner et al. 1999; Kaadige and Lopes 2003, 2006). However, the NuA4 component mutants did not yield reduced expression of the OPI1-cat reporter. Instead, they yielded elevated expression under repressing conditions (Figure 4C). That is, they have the same effect on OPI1-cat as they have on INO1 expression. Thus, the phenotype of the NuA4 mutants is not due to an indirect effect of reduced OPI1 expression.
The trafficking category included a large number of vacuolar protein sorting (vps) mutants, including the four components of the AP-3 adaptor protein complex, encoded by the APL5, APL6, APM3, and APS3 genes (Odorizzi et al. 1998; Bowers and Stevens 2005) (Table 1). The AP-3 adaptor complex is involved in clathrin-independent transport of proteins (such as alkaline phosphatase, Vam3p, and Yck3p) from the late Golgi to the vacuole (Odorizzi et al. 1998; Sun et al. 2004; Bowers and Stevens 2005). The product of the VPS41 gene associates with the AP-3 complex and is required for the formation of AP-3 vesicles from the late Golgi (Darsow et al. 2001; Bowers and Stevens 2005). The vps41 mutant was also identified in the VYDS screen (Table 1).
The Opi− mutants affected regulation of INO1-lacZ expression:
The Opi− phenotype of previously defined Opi− mutants correlates with a defect in the repression of INO1 or INO1-lacZ expression (Summers et al. 1988; McGraw and Henry 1989; White et al. 1991; Shen and Dowhan 1997; Ouyang et al. 1999; Elkhaimi et al. 2000). However, the mechanism for inositol excretion remains unknown; thus it was reasonable to predict that some mutants from the VYDS screen might affect inositol excretion but not repression of INO1 expression. To test for this, an INO1-lacZ reporter (pJH330) was transformed into the Opi− mutant strains. Transformants were assayed on X-gal plates under derepressing (I−C−) and repressing (I+C+) conditions.
Indeed, the results revealed that a significant number of the Opi− mutants did not affect repression of INO1-lacZ expression (Table 2). As expected, the wild-type transformants grew as white colonies on X-gal I+C+ medium (repressing). In contrast, the opi1 strain grew as dark-blue colonies, indicating that INO1-lacZ was constitutively expressed. The 89 Opi− mutants were plated in triplicate on X-gal I+C+ medium and categorized according to severity of the phenotype (Table 2). Dark blue indicates a strong phenotype similar to that of an opi1 mutant strain. Medium blue suggests an intermediate effect on the INO1-lacZ reporter, and light blue/white indicates weak or no effect on repression of INO1-lacZ. The assay identified 24 Opi− mutants that grew as dark blue colonies, 35 with a medium-blue phenotype, and 30 displayed a light-blue/white phenotype (Table 2).
To quantify the effect on INO1-lacZ expression, the INO1-lacZ transformants were grown in I−C− (derepressing) and I+C+ (repressing) media and assayed for β-galactosidase activity (Figure 5, A and B, respectively; see also supplemental data at http://www.genetics.org/supplemental/). In I−C− conditions, the wild-type strain yielded 475 units of β-gal activity (yellow bar in Figure 5A). Most of the Opi− mutants displayed INO1-lacZ expression levels comparable to the wild-type strain (Figure 5A). Thirteen mutants yielded levels of expression that were more than twofold the level of the wild-type strain while 17 mutant strains yielded levels of expression that were <50% of wild type. In I+C+ conditions, the wild-type strain yielded 28 units of β-gal activity (yellow bar in Figure 5B). The strains that grew as dark-blue colonies on X-gal I+C+ medium (Table 2) generally yielded more β-gal activity than the wild-type strain (red bars in Figure 5B). The Opi− mutants that grew as medium-blue colonies on X-gal I+C+ medium (Table 2) yielded intermediate β-gal activity levels (blue bars in Figure 5B). Likewise, the Opi− mutants that grew as light-blue/white colonies on X-gal I+C+ medium (Table 2) tended to yield decreased β-gal activity levels (open bars in Figure 5B). In addition, the two HDAC mutants not identified in the VYDS screen were also assayed and found to be defective in repression (Figure 3C) .
In general, there was a good correlation between the plate phenotype and the quantitative β-gal assay. However, this need not have been the case since INO1 expression is also growth phase regulated (Lamping et al. 1995; Jiranek et al. 1998; Robinson et al. 2000). Thus, those mutants whose X-gal plate and β-gal activity phenotypes do not correlate may reflect altered growth-phase regulation. Nevertheless, the data clearly showed that some of the Opi− mutants did not appear to affect INO1-lacZ repression. In some cases (cho2 and opi3), the absence of a defect in INO1-lacZ repression was predictable because defects in PC biosynthesis yield an Opi− phenotype that can be suppressed by the addition of choline (Summers et al. 1988; McGraw and Henry 1989). Thus, some of the other Opi− mutants may also affect PC biosynthesis and be similarly suppressed by choline. This was found to be the case and is described below.
Seven new Opi− mutants were suppressed by growth in choline:
PC synthesis de novo proceeds from PA to CDP–DAG to PC and requires three sequential methylation reactions (Figure 1). As stated above, PA levels control the translocation of Opi1p from the ER into the nucleus (Loewen et al. 2004). Thus, blocks in de novo synthesis of PC yield an Opi− phenotype due to elevated PA levels. However, the Opi− phenotype of these mutants can be suppressed by adding the appropriate supplement (E, MME, DME, or C) that restores PC synthesis through the Kennedy (salvage) pathway (Summers et al. 1988; McGraw and Henry 1989) (Figure 1). Thus, a cho2 mutant has an Opi− phenotype in I− medium and the phenotype is retained in the presence of E but suppressed by the addition of MME, DME, and C (Table 3 and Figure 6) (Summers et al. 1988).
To determine if any of the 89 Opi− mutants might also affect PC biosynthesis, the Opi− test was performed on medium lacking inositol but containing choline. The Opi− phenotype of 9 mutants was suppressed by the addition of choline (Table 3). These mutants were retested on medium lacking inositol but supplemented with E, MME, or DME (Figure 6) to determine if they affected a specific biochemical step in de novo PC biosynthesis. The addition of E, MME, and DME suppressed the Opi− phenotype of the mrpl38, mrpl49, and yol032w mutant strains (Table 3). As expected, MME and DME suppressed the Opi− phenotype of the cho2 mutant strain (Table 3 and Figure 6) (Summers et al. 1988) but also suppressed the fun26, kex1, and nup84 mutant strains (Table 3 and Figure 6). As previously reported, the opi3 mutant was suppressed by the addition of DME (McGraw and Henry 1989) (Table 3 and Figure 6). The addition of DME to an opi3 mutant suppresses the Opi− phenotype because it allows for the synthesis of high levels of phosphatidyldimethylethanolamine (PDME) (McGraw and Henry 1989). This is apparently sufficient to relieve the buildup of PA in the opi3 mutant. The tps1 mutant was also suppressed by the addition of DME (Table 3). Thus, seven new Opi− mutants likely affect de novo synthesis of PC.
The Opi− plate assay is obligatorily performed in medium lacking inositol. However, the ability of PC branch precursors (E, MME, DME, and C) to suppress the Opi− phenotype can also be observed in repressing conditions. For example, blocks in de novo PC biosynthesis are defective in inositol-mediated repression of the INO1 gene and its protein product; however, this defect is suppressed by addition of PC branch precursors (Letts and Henry 1985; Summers et al. 1988; McGraw and Henry 1989; Shen and Dowhan 1997). To determine if this property applied to any of the new Opi− mutants, we chose two mutants that affect PC biosynthesis (fun26 and yol032w) and tested them for defects in INO1-lacZ regulation (Figure 7). A wild-type strain and cho2, opi3, fun26, and yol032w mutant strains containing an INO1-lacZ reporter (pJH330) were assayed for β-galactosidase activity. Strains were grown in the absence (I−) and presence (I+) of inositol ± E (E+), MME (MME+), DME (DME+), or C (C+). As expected, in the wild-type strain, INO1-lacZ expression was derepressed in all five media lacking inositol (Figure 7A). The presence of inositol alone reduced INO1-lacZ expression, which was further repressed when supplemented with E, MME, DME, or C (Figure 7A). These results were consistent with those previously reported for regulation of INO1, INO1-lacZ, and Ino1p levels (Hirsch and Henry 1986; Summers et al. 1988; McGraw and Henry 1989; Lopes et al. 1991).
The Opi− phenotype of the cho2 and fun26 mutants was suppressed by addition of MME, DME, and C (Figure 6 and Table 3). Consistent with this plate phenotype, expression of INO1-lacZ was derepressed in all five I− media as well as in I+ and I+E+ media (Figure 7, B and D). As predicted from the plate phenotype, repression of INO1-lacZ expression was restored in I+ media supplemented with MME, DME, and C (Figure 7, B and D). The Opi− phenotype of the opi3 mutant was suppressed by the addition of DME and C (Figure 6 and Table 2). Consistent with this pattern, INO1-lacZ expression was derepressed in all five I− media as well as in I+, I+E+, and I+MME+ media (Figure 7C). As expected, repression of INO1-lacZ was restored in I+DME+ and I+C+ media (Figure 7C). The Opi− phenotype of yol032w mutant was suppressed by the addition of E (Table 3). INO1-lacZ expression was derepressed in all five I− media at a level approximately twofold greater than that observed in the wild-type strain (compare Figure 7E to 7A) and partially repressed in I+ media. Complete repression was restored in I+E+, I+MME+, I+DME+, and I+C+ media (Figure 7E). Thus, in general the pattern of regulation of INO1-lacZ expression correlates with the Opi− phenotype suppression pattern.
Opi− mutants affected the UPR:
Inositol depletion and accumulation of unfolded proteins in the ER affect expression of both phospholipid biosynthetic genes and UPR target genes (Travers et al. 2000; Jesch et al. 2005). This is due to the fact that both responses are coordinated through the IRE1 and HAC1 gene products. Previous reports restrict the activity of Opi1p to the inositol response (Cox et al. 1997). Microarray studies show that Opi1p is not required for inositol-mediated regulation of UPR target genes (Jesch et al. 2005) but does affect a small set of genes regulated in response to accumulation of unfolded proteins (Travers et al. 2000). Thus, it was reasonable to assume that some of the Opi− mutant genes might affect the UPR. To test this, the expression of an UPRE-driven lacZ reporter was examined in the Opi− mutant strains.
The 89 Opi− mutant strains were transformed with pJC104 (Cox and Walter 1996) to determine their response to UPR activation. Plasmid pJC104 contained four repeats of the UPRE fused to lacZ. Wild-type and hac1 mutant strains were used as controls and an opi1 mutant strain was also tested. Transformants were grown on X-gal I+C+ medium in the absence (Tm−) and presence of Tm (Tm+; induces UPR). The wild-type strain grew dark blue on Tm+ medium after an overnight incubation. As expected the hac1 mutant strain grew white on Tm+ medium. Several other mutant strains grew as white colonies (including vps74) and some as light-blue colonies (including opi1, yhi9, and ede1). The complete data set will be reported in a separate publication.
To quantify the effect of the Opi− mutants on the UPR, the wild-type strain (BY4742) and the opi1, yhi9, hac1, vps74, and ede1 mutant strains were assayed for β-galactosidase activity. Transformants were grown at 30° to midlogarithmic phase in I+C+ medium and shifted to I+C+, I−C−, or I+C+Tm+. Following the shift, cells were allowed to grow for 1 and 3 hr at which times samples were collected.
In general, the most dramatic effects were observed 3 hr after the shift (compare Figure 8, A and B). In the wild-type strain, β-galactosidase activity increased 72-fold after the 3-hr shift in I+C+Tm+ and 12-fold after the 3-hr shift in I−C− (Figure 8B). This is consistent with previously published results (Cox et al. 1997; Chang et al. 2002). Also as expected, the hac1 strain was defective in responding to inositol depletion and Tm addition (Cox and Walter 1996; Brickner and Walter 2004) (Figure 8). Consistent with its plate phenotype, the effect of vps74 resembled that of hac1 (Figure 8B). In contrast to microarray studies (Travers et al. 2000; Jesch et al. 2005), opi1 did appear to play a role in activation of UPRE in response to Tm and inositol depletion (Figure 8). The opi1 mutant was able to initiate the response to both conditions (Figure 8A) but was not able to reach the wild-type level of induction (Figure 8B). The yhi9 mutant was completely defective in the response to inositol and displayed a delayed and abated response to Tm (2.4-fold induction in Tm+) (Figure 8B). The ede1 mutant exhibited a normal early response to Tm but an abated 3-hr response to Tm (6.5-fold induction); however, the response to inositol depletion was normal (Figure 8). These data show that the Opi− mutants are likely to provide significant insight into understanding the response to accumulation of unfolded proteins in the ER.
Recent findings have significantly advanced our understanding of the mechanism of Opi1p repression (Loewen et al. 2003, 2004; Gardenour et al. 2004; Heyken et al. 2005; Jesch et al. 2005; Loewen and Levine 2005). However, many questions remain unanswered. A powerful and effective way to deal with this is to screen the VYDS (Huang and O'Shea 2005). Thus, we identified 89 genes from the VYDS that display the Opi− phenotype. These 89 genes affect multiple biological processes (Figure 2), including six that are involved in phospholipid biosynthesis, the five nonessential members of the NuA4 complex, and seven that are the members of the Rpd3p HDAC complex. The screen also identified 43 mutants that alter protein trafficking and, notably, the four members of the AP-3 complex (Table 1). The screen successfully identified all predictable Opi− mutants. These include cho2, opi1, opi3, ume6, rpd3, sin3, and reg1 (Table 1). As previously reported, mutations in the Ume6p–Sin3p–Rpd3p HDAC displayed the Opi− phenotype (Elkhaimi et al. 2000). The Opi− phenotype was most severe in the ume6 strain and least severe in the rpd3 strain (Figure 3) (Elkhaimi et al. 2000). More recently, it has been shown that Pho23p, Dep1p, Rxt2p, and Eaf3p are also associated with the Rpd3p HDAC (Loewith et al. 2001; Carrozza et al. 2005; Gavin et al. 2006) and the VYDS screen found that mutants in each of these genes display the Opi− phenotype (Figure 3) and misregulate INO1-lacZ expression (Figure 5B). However, the screen did not identify two other members of this complex (ume1 and sap30) that turned out to have a weak Opi− phenotype (Figure 3B) and four other members of the complex do not appear to yield an Opi− phenotype.
Some of the Opi− mutants identified in the screen overlap other ORFs. In those situations, the disruption of one of the two genes may be the cause of the Opi− phenotype. In some cases, the screen identified both genes that overlap. For example, the screen identified vid21 and the overlapping ydr360w (opi7), as well as vps61 and rgp1. Examination of the INO1-lacZ plate phenotypes and the β-galactosidase activity of the vid21/ydr360w mutant pair revealed different phenotypes. The vid21 mutant has a medium-blue-plate phenotype and β-galactosidase activity of 823 units in I−C− and 97 units in I+C+ compared to the ydr360w mutant, which has a light-blue-plate phenotype and β-galactosidase activity of 351 units in I−C− and 4 units in I+C+ (Table 2 and Figure 5). YDR360W (OPI7) is listed as a dubious gene in the Saccharomyces Genome Database (http://www.yeastgenome.org/). Thus, while mutation of either gene may yield an Opi− phenotype, a more likely explanation for the Opi− phenotype of the ydr360w mutant is that it partially affects expression of VID21 because it only partially overlaps VID21. The vps61 and rgp1 mutants both have a medium-blue-plate phenotype and comparable β-galactosidase activities in I−C− (494 and 382 units, respectively) and I+C+ (26 and 20 units, respectively) media. This suggests that only one of the two mutants actually yields the Opi− phenotype. APL6 and TNA1 transcribe toward each other; however, the apl6 mutant shows a medium-blue-plate phenotype with β-galactosidase activities of 731 units and 28 units (I−C− and I+C+, respectively) while the tna1 mutant shows a dark-blue-plate phenotype with β-galactosidase activities of 87 units and 149 units (I−C− and I+C+, respectively). Thus, the different phenotypes suggest that both mutants have the Opi− phenotype. In a few instances, only one of the two overlapping genes was identified. For example, the screen identified ydl096c (opi6), ykr035c (opi8), ylr338w (opi9), and ypr044c (opi11) but not the overlapping pmt1 (ydl095w), did2 (ykr035w-a), vrp1 (ylr337c), and rpl43a (ypr043w), respectively. In each of these cases, the screen identified a dubious gene. However, because each of these mutants had a relatively weak Opi− phenotype, we screened mutants in the corresponding overlapping genes and found them to also have a weak Opi− phenotype. In another case, the screen identified lbd16 and ycl006c. The YCL006C ORF has since been removed from the database (http://www.yeastgenome.org/); however, it was originally located immediately upstream of and divergent from the LBD16 gene. Thus, the ycl006c mutant probably removes the promoter of the LDB16 gene. In every one of these cases, additional experiments will be needed, and are ongoing, to determine which mutant is responsible for the phenotype.
Curiously, a subset of the Opi− mutants did not overexpress the INO1-lacZ reporter under repressing conditions (Figure 5B). Transcriptional regulation of INO1 is dependent on growth phase (Lamping et al. 1995). INO1 transcript levels increase during exponential growth in wild-type strains grown in the absence of inositol and drop precipitously when cells enter stationary phase. In an opi1Δ strain, INO1 transcript levels are elevated during all growth stages (Jiranek et al. 1998). INO1 transcription is repressed in the presence of inositol during all phases of growth (Lamping et al. 1995). However, in the opi1Δ strain, INO1 expression remains elevated at all growth phases in the presence of inositol (Jiranek et al. 1998). Thus, it is possible that some of the Opi− mutants affect growth-phase regulation, yielding the Opi− phenotype and dark-blue colonies in the X-gal plate assay (Table 2) without demonstrating a phenotype in the β-gal assay, which reflects a single time point in the growth of the culture (Figure 5).
Yet another explanation for the Opi− mutants that do not affect INO1-lacZ expression is that they may affect inositol excretion directly. The mechanism for inositol transport is understood and requires two membrane-associated transporters encoded by the ITR1 and ITR2 genes (Nikawa et al. 1991). Mutations in the INO2, INO4, and OPI1 genes alter the transcription of the ITR1 gene (Lai and McGraw 1994). An itr1 mutant has been shown to have a defect in INO1 repression (Lai and McGraw 1994). Consistent with this, the VYDS screen identified the itr1 mutant (Table 1). Unlike inositol uptake, the mechanism for inositol excretion is unknown. Thus, some of the Opi− mutants may help to elucidate this mechanism.
Several membrane trafficking pathways exist in yeast, including secretory, endocytic, vacuolar protein sorting, and multivesicular body sorting (Deamilli et al. 1996; Odorizzi et al. 1998, 2000; Lee et al. 2004; Bowers and Stevens 2005). The VYDS identified several vps mutants that are known to affect these pathways. While some of these vps mutants affected the regulation of INO1-lacZ, many had no effect (Table 2 and Figure 5). The screen also identified all four components (Apl5p, Apl6p, Apm3p, and Aps3p) of the AP-3 complex that is involved in protein transport to the vacuole (Odorizzi et al. 1998). However, mutants of the AP-3 complex did not affect regulation of INO1-lacZ (Table 2 and Figure 5). It is tempting to speculate that the Opi− phenotype of the vps and AP-3 complex mutants could be indirect and due to misregulation of intracellular stores of inositol. However, it is also possible that some of these mutants have direct effects since phospholipids are known to be integrally involved in the regulation of membrane trafficking (McGee et al. 1994; Deamilli et al. 1996; Kagiwada et al. 1996; Odorizzi et al. 2000; Yanagisawa et al. 2002).
The NuA4 HAT complex is associated with global and targeted acetylation of H4 tails in vivo (Doyon and Coté 2004). This complex consists of 13 proteins (Carrozza et al. 2003; Utley and Coté 2003; Doyon and Coté 2004). Of these, Tra1p, Epl1p, Esa11p, Eaf2p, Yng2p, Arp4p, Act1p, and Eaf6p are essential while Eaf3p, Eaf5p, Eaf7p, Vid21p, and Yaf9p are nonessential. All of the nonessential genes were identified as Opi− mutants (Table 1, Figure 4). The NuA4 complex is primarily associated with acetylation of histone H4, which relaxes chromatin and allows transcription to take place. Therefore, mutants in NuA4 would be expected to reduce INO1 expression. One model to explain why the NuA4 complex mutants affected repression is that NuA4 may function indirectly via regulation of the OPI1 gene. However, we found that OPI1-cat expression was not reduced in four different strains containing NuA4 mutants (Figure 4C). Consequently, NuA4 may regulate INO1 indirectly through another transcription factor or it may also regulate INO1 directly via a different chromatin-modifying activity.
Of the 89 Opi− mutants, 9 (7 new mutants) were suppressed by the addition of choline, suggesting that they are involved in PC biosynthesis. These included 3 mutants that were suppressed by the addition of E (mrpl38, mrpl49, and yol032w), 4 that were suppressed by MME (cho2, fun26, nup84, and kex1), and 2 that were suppressed by DME (opi3 and tps1) (Table 3 and Figure 6). Because these mutants have phenotypes similar to those of cho2 and opi3 mutants, it is reasonable to predict that they play different roles in the biochemical steps of PC biosynthesis. It is well established that CHO2 is responsible for the methylation of PE to phosphatidylmonomethylethanolamine (PMME) and OPI3 is responsible for the final two methylation steps from PMME to PDME to PC (Kodaki et al. 1987; Summers et al. 1988; McGraw and Henry 1989). In most cases, the annotation of the 7 new conditional Opi− mutants does not provide obvious explanations for the Opi− phenotype. MRPL38 and MRPL49 encode mitochondrial ribosomal proteins of the large subunit that are involved in protein biosynthesis (Graack and Wittmann-Liebold 1998). YOLO32W (OPI10) is of unknown function and KEX1 is required for protein processing (Latchinian-Sadek and Thomas 1993). FUN26 is of unknown function; however, it has been shown to be an intracellular nucleoside transporter with broad specificity (Vickers et al. 2000). However, the nup84 Opi− phenotype may be explained by some recent observations. NUP84 encodes a subunit of the nuclear pore complex (NPC) that has recently been shown to be required for activation of transcription via Rap1p, Gcr1p, and Gcr2p by recruiting these factors to the NPC (Menon et al. 2005). Expression from the OPI3 (PEM2) promoter has been shown to be dependent on Rap1p (Grf1p) (Kodaki et al. 1991). Therefore, it is conceivable that expression of the CHO2 gene is dependent on Rap1p or on some other transcription factor that is recruited to the Nup84 NPC. Thus, while it is not yet possible to explain the PC biosynthesis-dependent Opi− phenotype mutants, it is likely that these mutants may be affecting the export, processing, or function of the known PC biosynthesis enzymes.
The inositol depletion signal and the UPR-signaling pathway converge via the activation of Ire1p. Activation of Ire1p leads to the splicing of the HAC1 transcript, producing a functional Hac1p that is able to bind to target genes containing the UPRE (Figure 9). Hac1p has been shown to antagonize Opi1p from binding to the Ino2p/Ino4p heterodimer (Brickner and Walter 2004), thereby allowing induction of the phospholipid biosynthetic genes. However, Opi1p was not reported to play a role in the UPR. Contrary to published reports (Travers et al. 2000), this study shows that Opi1p appears to act as an activator of the UPR (Figures 8 and 9). This effect could be due to a direct effect on UPRE-regulated genes or an indirect effect through Hac1p (Figure 9). Some of the other Opi− mutants were also found to affect the UPR (Figure 8). For example, the vps74 mutant phenotype resembles the hac1 (and ire1) mutant phenotype. This suggests that VPS74 is required for the function/expression of either HAC1 or IRE1 (Figure 9). Likewise, yhi9 and ede1 resemble the opi1 mutant (Figures 8 and 9), suggesting that they may affect Opi1p function. Further analysis of the effects of the Opi− mutants in the UPR may therefore reveal how these mutants affect regulation of both the UPR target genes and the phospholipid biosynthetic genes.
We thank Meng Chen, Linan Chen, Ying He, and Niketa Jani for critical reading of the manuscript. We thank Peter Walter (University of California at San Francisco) for providing the UPRE-lacZ reporter. The authors also thank Miriam Greenberg (Wayne State University) for developing the Opi− plate screen. This work was supported by a grant from the National Science Foundation to J.M.L. (MCB 0415511). L.C.H. was the recipient of the William A. Turner Jr. Memorial Foundation Scholarship and R.P.B. was the recipient of the Nicolette Therese Keller Endowed Fellowship in Biological Sciences.
Communicating editor: B. J. Andrews
- Received February 21, 2006.
- Accepted March 28, 2006.
- Copyright © 2006 by the Genetics Society of America