Interaction between repressor Opi1p and ER membrane protein Scs2p facilitates transit of phosphatidic acid from the ER to mitochondria and is essential for INO1 gene expression in the presence of choline

In the yeast Saccharomyces cerevisiae, the Opi1p repressor controls the expression of INO1 via the Opi1p/Ino2p–Ino4p regulatory circuit. Inositol depletion favors Opi1p interaction with both Scs2p and phosphatidic acid at the endoplasmic reticulum (ER) membrane. Inositol supplementation, however, favors the translocation of Opi1p from the ER into the nucleus, where it interacts with the Ino2p–Ino4p complex, attenuating transcription of INO1. A strain devoid of Scs2p (scs2Δ) and a mutant, OPI1FFAT, lacking the ability to interact with Scs2p were utilized to examine the specific role(s) of the Opi1p–Scs2p interaction in the regulation of INO1 expression and overall lipid metabolism. Loss of the Opi1p–Scs2p interaction reduced INO1 expression and conferred inositol auxotrophy. Moreover, inositol depletion in strains lacking this interaction resulted in Opi1p being localized to sites of lipid droplet formation, coincident with increased synthesis of triacylglycerol. Supplementation of choline to inositol-depleted growth medium led to decreased TAG synthesis in all three strains. However, in strains lacking the Opi1p–Scs2p interaction, Opi1p remained in the nucleus, preventing expression of INO1. These data support the conclusion that a specific pool of phosphatidic acid, associated with lipid droplet formation in the perinuclear ER, is responsible for the initial rapid exit of Opi1p from the nucleus to the ER and is required for INO1 expression in the presence of choline. Moreover, the mitochondria-specific phospholipid, cardiolipin, was significantly reduced in both strains compromised for Opi1p–Scs2p interaction, indicating that this interaction is required for the transfer of phosphatidic acid from the ER to the mitochondria for cardiolipin synthesis.

In addition to elevated levels of PA, retention of Opi1p in the ER also requires its interaction with Scs2p (15). Scs2p, homolog of mammalian synaptobrevin-associated protein (VAMP), is a conserved integral ER protein and a component of a lipid-sensing complex (17)(18)(19). Proteins of the VAMP family serve as anchors to the ER for cytoplasmic proteins, including Opi1p, through a conserved motif known as FFAT, two phenylalanines (FF) in an Acidic Tract (19,20). Many proteins to which VAMP binds are targeted to other intracellular membranes. This implies that VAMP is an important bridge between the ER and a wide variety of organelles at membrane contact sites (21), serving to facilitate lipid transfer and communication between organelles (22)(23)(24). For example, in yeast, phosphatidylinositol 4-phosphate levels on the plasma membrane are regulated at membrane contact sites through interaction between Scs2p and Osh3p (25). In addition, deletion of SCS2 leads to a 50% loss in ER-plasma membrane contact sites (26). However, to date, no role for Opi1p in lipid transfer between ER and other membrane compartments has been reported.
Both scs2⌬ and opi1⌬ mutants exhibit complex, pleiotropic phenotypes, related to lipid metabolism and gene regulation (19,(27)(28)(29)(30). The yeast SCS2 gene was originally isolated as a high copy suppressor of choline sensitivity (39), a phenotype associated with an uncharacterized dominant mutation, CSE1 (31). Scs2p was subsequently identified as an integral membrane protein of the ER (17) and homolog of the mammalian VAMP (32). The scs2⌬ mutant exhibits a relatively weak Ino Ϫ phenotype at 30°C, which becomes stronger (more visible in plate tests) at growth temperatures of 34°C or higher (Fig. 2, A  and B). The scs2⌬ Ino Ϫ phenotype also becomes more stringent in the presence of exogenous choline (14) (Fig. 2, A and B) and is suppressed by mutations in the CDP-choline pathway for PC biosynthesis (17,18). The opi1⌬ mutation, in contrast, confers constitutive overexpression of INO1 and other UAS INO -containing genes (3)(4)(5)33), resulting in overproduction and excretion of inositol into the growth medium (Opi Ϫ phenotype). The opi1⌬ mutant also exhibits a Pet Ϫ phenotype (i.e. inability to survive complete loss of the mitochondrial genome) and produces very reduced levels of the mitochondrial lipid, CL (34). CL is synthesized in the inner mitochondrial membrane in a multistep pathway (Fig. 1), using PA synthesized in the ER and transferred to mitochondria via the ER-mitochondria encounter structure (ERMES) (35)(36)(37)(38).
A major goal of the current study was to determine which of the diverse regulatory and metabolic phenotypes conferred by the scs2⌬ and opi1⌬ mutations result from the loss of the interaction between Opi1p and Scs2p in the ER. Moreover, given the importance of the FFAT motif in facilitating interorganelle lipid transfer (21), we examined the specific function of the interaction of Opi1p and Scs2p in regulation of lipid metabolism. For this purpose, we used the BY4742 parent strain to mutate the FFAT domain in OPI1, within its genomic locus, to create an OPI1 ffat strain. A second goal was to determine the individual and relative roles of inositol and choline on both lipid metabolism and regulation of INO1. In each of three congenic strains, BY4742 (wild type), scs2⌬ (SJY39), and OPI1 ffat In the ER, PA serves as immediate precursor of CDP-DAG, precursor to PI and PS. PS is converted into PE by decarboxylation in the mitochondria or in the endosome compartment (accented yellow box). PC is synthesized by methylation of PE or from DAG and CDP-choline. DAG is derived from dephosphorylation of PA or deacylation of TAG. Acyl-CoA is localized in the cytosol (light yellow box) and serves as fatty acid pool for the synthesis of PA, TAG, and SE. PA is also transported from the ER to the mitochondria, where it generates CDP-DAG and CL. The positions of PA within the metabolic network are boxed, the positions of DAG are circled, and the positions at which inositol and choline enter in the metabolic pathway are boxed. The arrows represent routes of metabolic conversion. The names of the structural genes encoding enzymes catalyzing specific metabolic conversions are shown adjacent to the arrows.
(YCY47), we also tagged OPI1 or OPI1 ffat with GFP within the OPI1 genomic locus, to analyze and compare the kinetics of changes in lipid metabolism, Opi1p or Opi1 ffat p localization, and INO1 derepression in cells shifted from inositol-containing to inositol-free medium in the presence and/or absence of choline. We report here that the OPI1 ffat mutant exhibits relatively weak inositol auxotrophy (Ino Ϫ phenotype), which, similar to the phenotype of the scs2⌬ mutant (14,17), becomes stronger at higher growth temperatures and in the presence of choline. Moreover, Opi1p in the scs2⌬ strain and Opi1 ffat p in the OPI1 ffat strains, respectively, did not exit the nucleus following a shift to medium lacking inositol when choline was present, and under these conditions, INO1 failed to derepress. Thus, we conclude that these phenotypes are conferred by the loss of the Opi1p-Scs2p interaction in the ER. Moreover, we observed that loss of the Opi1p-Scs2p interaction in the ER results in reduced content of the mitochondrial lipid CL under all growth conditions tested, indicating that PA transfer from the ER to the mitochondria is impaired in the absence of this interaction. These results suggest a heretofore unrecognized role for Scs2p-Opi1p interaction in lipid transfer from ER to mitochondria.

The scs2⌬ and OPI1 ffat mutations confer inositol auxotrophy that is strengthened in the presence of choline and at higher growth temperatures
In this study, we sought to investigate the root cause of the choline-sensitive inositol auxotrophy phenotype of scs2⌬. The scs2⌬ mutant exhibits reduced growth in the absence of inositol (Ino Ϫ phenotype), which is most evident at higher growth temperatures (Fig. 2, A and B) (18,19,40) and in the presence of exogenous choline (14) (Fig. 3). The scs2⌬ Ino Ϫ phenotype is also suppressed by mutations in the CDP-choline pathway for PC biosynthesis (18), indicating that incorporation of choline into phospholipids via this pathway is involved in the Ino Ϫ phenotype of scs2⌬ (Fig. 3).
Because a functional FFAT domain is necessary for interaction of Opi1p with Scs2p in the ER, we asked whether an OPI1 ffat strain exhibits phenotypes similar to those observed in the scs2⌬ strain. We reasoned that shared phenotypes in these two strains indicate a functional role for this important interaction. To test this hypothesis, we used the BY4742 parent strain to mutate the FFAT domain in OPI1, within its genomic locus, thereby creating the OPI1 ffat strain. The OPI1 ffat (YCY47) and scs2⌬ (SJY39) strains were then tested and compared in standard plate assays for their ability to grow in the absence of inositol at 30 and 34°C in the presence and/or absence of choline. The OPI1 ffat and scs2⌬ strains both exhibit similar (weak) inositol auxotrophy at 30°C. (Compare the stringent Ino Ϫ phenotype of ino1⌬ with the weak Ino Ϫ phenotypes of scs2⌬ and OPI1 ffat at 30°C; Fig. 2A). The Ino Ϫ phenotype of both scs2⌬ and OPI1 ffat is strengthened at 34°C and further enhanced when choline is present at 34°C (Fig. 2B). Because scs2⌬ and OPI1 ffat SCS2 share only one defect in common, namely the lack of a functional Scs2p-Opi1p interaction in the ER, we conclude that loss of this interaction is responsible for these growth phenotypes. Thus, interaction of Opi1p with Scs2p in the ER is required, in addition to its interaction with PA, for optimal growth in the absence of inositol, especially at higher growth temperatures and in the presence of choline.
We also deleted PCT1, encoding choline-phosphate cytidylyltransferase (41), in the OPI1 ffat and scs2⌬ strains, thereby blocking incorporation of choline via the CDP-choline (Kennedy) pathway for PC synthesis (42) (Fig. 1). The pct1⌬ mutation suppressed choline-sensitive inositol auxotrophy phenotype in both OPI1 ffat and scs2⌬ strains (Fig. 3). Thus, choline has to enter the CDP-choline pathway for PC synthesis to influence the Ino Ϫ phenotype of the scs2⌬ and OPI1 ffat strains. PC synthesized in the ER via CDP-choline pathway utilizes diacylglycerol (DAG), which also serves as the immediate precursor to triacylglycerol (TAG), a major constituent of lipid droplets (Fig. 1). In addition, we constructed and tested a diploid strain, OPI1 ffat /OPI1, to determine whether the OPI1 ffat mutation was dominant or recessive in terms of its growth in the presence and absence of inositol and choline at 30 and 34°C. The diploid strain, OPI1 ffat /OPI1, exhibited a slight growth reduction on I Ϫ C Ϫ and I Ϫ C ϩ medium (data not shown), a phenotype intermediate between that observed in the wild-type and OPI1 ffat strains, indicating that the OPI1 ffat mutation is semidominant with respect to its sensitivity to choline in the absence of inositol. This result implies that both gene products are expressed  Figure 2. Inositol auxotrophy (Ino ؊ phenotype) of the OPI1 ffat and scs2⌬ mutants at different temperatures. A suspension of BY4742 (wild type), YCY47 (OPI1 ffat ), and SJY39 (scs2⌬) cells at a concentration of 1.0 A 600 nm /ml and four subsequent 1:10 serial dilutions of strains were spotted on I ϩ C Ϫ , I Ϫ C Ϫ , I ϩ C ϩ , and I Ϫ C ϩ plates and allowed to grow for 2 days at 30°C (A) or 34°C (B). SJY425 (ino1⌬) serves as a control for the Ino Ϫ phenotype.

Significance of the Opi1p-Scs2p interaction in yeast lipid metabolism
from the native promoter of OPI1 and contribute proportionately to the phenotype of the diploid.

The effects of the OPI1 ffat and scs2⌬ mutations on INO1 expression and Opi1p localization following a shift to inositol-free medium in the absence of choline
Manipulation of PA levels following withdrawal of inositol and its subsequent readdition in the presence or absence of choline provides a powerful method for analyzing the relative effects of PA on Opi1p function (15,43). To this end, we analyzed and compared the relative timing of changes in lipid metabolism, Opi1p localization, and INO1 expression in the wild-type (YCY3), OPI1 ffat (YCY5), and scs2⌬ (YCY7) strains under these growth conditions.
As expected, before the shift to inositol-free medium in the absence of choline (i.e. from I ϩ C Ϫ to I Ϫ C Ϫ medium) at 30°C, INO1 was fully repressed in the wild-type strain (repressed level set arbitrarily at 1 unit of expression) (Fig. 4A). Under these conditions, Opi1p-GFP was localized exclusively to the nucleus (Fig. 4B, 0 h, left panel). At 1 h following the shift to I Ϫ C Ϫ medium, Opi1p-GFP, in the wild-type strain, had largely exited the nucleus and translocated, primarily to the perinuclear ER region, with a slight residual pool of fluorescence remaining visible within the nucleus (Fig. 4B, 1 h, Figure 3. The pct1⌬ mutation suppresses the Ino ؊ phenotype of the OPI1 ffat strain at all temperatures and in the presence of choline. Suspension of BY4742 (wild type), YCY47 (OPI1 ffat ), LGY169 (pct1⌬), and LGY541 (OPI1 ffat pct1⌬) cells at a concentration of 1.0 A 600 nm /ml and four subsequent 1:10 serial dilutions of strains were spotted on I ϩ C Ϫ , I Ϫ C Ϫ , I ϩ C ϩ , and I Ϫ C ϩ plates and allowed to grow for 2 days at 30°C (A) or 34°C (B). (OPI1 ffat ), and YCY7 (scs2⌬) expressing genomic Opi1p-GFP were diluted to A 600 nm ϭ 0.2 in I ϩ C Ϫ medium and allowed to grow to mid-logarithmic phase at 30°C. Cells were harvested by centrifugation and washed and resuspended in I Ϫ medium, followed by incubation for 3 h at the same temperature. Inositol was added back after 3 h of inositol starvation. Samples were taken at 0, 1, 2, and 3 h of inositol starvation and 30 min after adding back inositol. Total RNA was isolated and analyzed by RT-PCR as described under "Experimental procedures." Solid triangles, wild type; solid squares, OPI1 ffat ; solid crosses, scs2⌬. B, Opi1p-GFP localization over the same time course. Cells were imaged by fluorescence microscopy. A representative z-section is chosen for each image. Scale bar, 5 m. White arrows, Opi1p-GFP associated with distinctive puncta.

Significance of the Opi1p-Scs2p interaction in yeast lipid metabolism
interval, consistent with the partial exit of Opi1p-GFP from the nucleus (Fig. 4B, 1 h, left panel), INO1 expression increased about 50-fold (Fig. 4A). Within the interval from 1 to 2 h following the shift to I Ϫ C Ϫ medium, expression of INO1 increased from 50-fold to about 375-fold in the wild-type strain (Fig. 4A). After 2 h in I Ϫ C Ϫ medium, the rate of increase in INO1 expression leveled off, rising slightly above 410-fold ( Fig. 4A). At this time point, a significant pool of Opi1p-GFP was still visible in the perinuclear ER region (Fig. 4B, 2 h, left panel).
In contrast to the relatively rapid initial increase in INO1 expression in the wild-type strain, INO1 expression in the OPI1 ffat and scs2⌬ strains increased by only about 15-fold during the first hour following the shift to I Ϫ C Ϫ medium (Fig. 4A). After the shift to inositol-free medium in the absence of choline (Fig. 4), the nuclear pool of Opi1 ffat p-GFP had become less intense by 2 h but was still visible in most cells (Fig. 4B, 2 h, middle panel). After 2 h in I Ϫ C Ϫ medium, Opi1p ffat -GFP in the OPI1 ffat strain was associated with distinctive "puncta" (about 3-6/cell), and the nuclear pool of Opi1 ffat p-GFP had become less intense but was still visible in most cells (Fig. 4B, 2 h, middle panel). At this time, INO1 expression in the OPI1 ffat strain had increased by 210-fold, still significantly lagging the 375-fold increase seen in wild type (Fig. 4A). However, within 3 h following the shift to I Ϫ C Ϫ medium, INO1 expression in OPI1 ffat had increased 320-fold, and signal from Opi1 ffat p-GFP had reappeared to some extent in the nucleus, with fewer distinctive puncta (1-2/cell) remaining visible (Fig. 4B, 3

h, middle panel).
In the scs2⌬ strain, Opi1p-GFP fluorescence was still visible in the nucleus at 1 h following the shift to inositol-free medium (Fig. 4B, 1 h, right panel). At this time, one or two puncta, similar in appearance to those associated with Opi1 ffat p-GFP in the OPI1 ffat strain, were visible adjacent to the nucleus or plasma membrane in the scs2⌬ strain (Fig. 4B). By 2 h following the shift to inositol-free medium, Opi1p-GFP in the scs2⌬ strain had translocated more prominently to puncta adjacent to nucleus and plasma membrane. However, unlike either wild type or OPI1 ffat (Fig. 4B, 2 h, left and middle panels), a diffuse pattern of Opi1p-GFP also persisted throughout the cell in the scs2⌬ strain, possibly reflecting a reduction in cortical ER associated with the scs2⌬ mutation, as reported by Loewen et al. (29). At 3 h, a diffuse pattern of Opi1p-GFP localization also remained visible in the nuclear region in the scs2⌬ strain.
However, despite differences between the scs2⌬ and wildtype strains in the pattern of Opi1p-GFP localization (Fig. 4B) following the shift to I Ϫ C Ϫ medium, overall expression of INO1 in the scs2⌬ strain reached levels only slightly lower than those seen in wild type (Fig. 4A). By 2 h following the shift to I Ϫ C Ϫ medium, INO1 expression had increased by 325-fold in the scs2⌬ strain, a level close to the 375-fold increase observed in the wild-type control during the same interval. Within 3 h following the shift to I Ϫ C Ϫ medium, INO1 expression had increased by 375-fold in the scs2⌬ strain, a level that is not significantly different in comparison with the 410-fold increase observed in the wild-type strain (Fig. 4A). In contrast, INO1 expression in the OPI1 ffat strain remained somewhat lower in comparison with the other two strains at each time point following the shift to inositol-free medium (Fig. 4A). After 3 h following the shift to I Ϫ C Ϫ medium, inositol was added back to each of the cultures. INO1 expression in all three strains decreased within about 30 min to values similar to those observed in these same strains between 0 and 1 h following the shift from I ϩ C Ϫ to I Ϫ C Ϫ medium (Fig. 4A). Translocation of Opi1-GFP or Opi1 ffat p-GFP back into the nuclei of the respective strains was observed within this same 30-min interval (Fig. 4B).

The puncta associated with Opi1 ffat p-GFP, following a shift to medium lacking both inositol and choline, are ER-bound lipid droplets
As described above, within 2 h following a shift of the OPI1 ffat strain to I Ϫ C Ϫ medium, a fraction of the fluorescence of Opi1 ffat p-GFP was associated with distinctive puncta, adjacent to the perinuclear ER (Fig. 4B, 2 h, middle panel). In addition to its role as precursor to PI and other membrane forming phospholipids, PA also serves as precursor, via DAG, in the synthesis of TAG (Fig. 1), a major constituent of lipid droplets. The appearance of Opi1 ffat p-GFP-associated puncta in cells deprived of inositol is consistent with the report of Han et al. (44) showing that Opi1 ffat p associates with areas of nascent lipid droplet formation. Therefore, given the morphology of the Opi1 ffat p-GFP-labeled puncta and their proximity to the perinuclear ER, we suspected that these puncta represent a pool of PA associated with nascent lipid droplets. To test this hypothesis, we created cells co-expressing Tgl4p-RFP, a lipid droplet marker (45). We observed a significant overlap of Tgl4p-RFPand Opi1 ffat p-GFP-associated puncta following a 2-h shift to I Ϫ C Ϫ medium (Fig. 5A, right), supporting the identification of the puncta as lipid droplets. After 3 h of inositol starvation, Opi1 ffat p-GFP had translocated in part back to the nucleoplasm. Following inositol addition, Opi1 ffat p-GFP relocated completely to the nucleus within 30 min (Fig. 4B). To further verify and quantify Opi1 ffat p co-localization with lipid droplets, we examined the localization of GFP-Opi1 ffat p, expressed from a centromeric plasmid, in cells also expressing Erg6p-RFP, a protein that is abundant in yeast lipid droplets (46,47). Co-localization of Erg6p and Opi1p was quantified using Pearson's correlation coefficient (PCC) (48) in 30 cells that exhibited clear expression of the plasmid versions of GFP-Opi1p or GFP-Opi1 ffat p. The co-localization of GFP-Opi1p and RFP-Egr6p, both before and 2 h after the shift from I ϩ C Ϫ to I Ϫ C Ϫ medium, was not significant (PCC ϭ 0.24 and 0.21, respectively) ( Fig.  5B). In contrast, in the strain co-expressing GFP-Opi1 ffat p and RFP-Erg6p, PCC was 0.27 before the shift and 0.61 by 2 h following the shift to I Ϫ C Ϫ medium (Fig. 5B), indicating a significant increase in Opi1 ffat p co-localization with lipid droplets at 2 h following the shift to I Ϫ C Ϫ medium. These observations are consistent with previous studies indicating that a pool of PA in the ER and/or lipid droplets (49), used for synthesis of DAG, precursor to TAG (Fig. 1), competes with the pool of PA used in the synthesis of CDP-DAG, precursor to PI and other membrane phospholipids. Thus, in cells grown in the presence of inositol, increased PI synthesis is correlated with decreased TAG accumulation (4,43,50,51).
To further test this hypothesis, translocation of Opi1p-GFP and Opi1 ffat p-GFP was also examined in a dga1⌬ lro1⌬ are1⌬ are2⌬ quadruple mutant (Table 1), which is unable to make

Significance of the Opi1p-Scs2p interaction in yeast lipid metabolism
lipid droplets (52). A plasmid version of GFP-Opi1p or GFP-Opi1 ffat p (as described under "Experimental procedures"; see Table 2) was transformed into the wild-type strain and the quadruple mutant. In both strains, GFP-Opi1p translocated to ER membranes within 2 h following transfer to I Ϫ C Ϫ medium (Fig. 5C, top), following a pattern and kinetics similar to that observed using the genomic Opi1p-GFP construct (Fig. 4B, 2 Figure 5. Opi1p ffat -GFP transiently colocalizes with lipid droplets following a shift to medium lacking inositol and choline. Overnight cultures grown in I ϩ C Ϫ medium at 30°C were diluted to A 600 nm ϭ 0.2 and allowed to grow to mid-logarithmic phase in I ϩ C Ϫ medium at 30°C. Cells were harvested by centrifugation, washed and resuspended in I Ϫ C Ϫ medium, incubated for 2 h, and analyzed by fluorescence microscopy. A, Tgl4-RFP, carried on a plasmid, was co-expressed with genomic Opi1p-GFP in wild type (top), or with genomic Opi1p ffat -GFP in OPI1 ffat (bottom). B, the CMY564 strain expressing genomic RFP-Erg6p was transformed with plasmids expressing either GFP-Opi1p or GFP-Opi1p ffat . Colocalization of Tgl4-RFP with genomic Opi1p-GFP or Opi1p ffat -GFP, respectively, was assessed using PCC (n ϭ 30) between wavelength 528 nm (GFP) and 617 nm (RFP). PCC Ͼ 0.5 indicates significant correlation. C, the wild-type and dga1⌬lro1⌬are1⌬are2⌬ quadruple mutant (4⌬) strains were transformed with plasmids expressing either GFP-Opi1p (top) or GFP-Opi1p ffat (bottom). Cells were grown and collected as indicated above and analyzed by fluorescence microscopy. A representative z-section was chosen for each image. Scale bars, 5 m. Error bars, S.D.

Significance of the Opi1p-Scs2p interaction in yeast lipid metabolism
the plasmid-borne GFP-Opi1 ffat p to I Ϫ C Ϫ medium, fluorescence was also associated with puncta. However, in the dga1⌬lro1⌬are1⌬are2⌬ strain, Opi1 ffat p was mainly localized in the nucleus and the perinuclear region, and no puncta were observed by 2 h following the shift to I Ϫ C Ϫ medium (Fig. 5C, bottom right), indicating that formation of the puncta is most likely related to lipid droplet formation.

Following a shift to inositol-free medium in the presence of choline, INO1 fails to derepress in the OPI1 ffat and scs2⌬ strains
In wild-type cells grown in medium containing both inositol and choline and shifted to medium lacking inositol but containing choline (i.e. a shift from I ϩ C ϩ to I Ϫ C ϩ medium; Fig. 6), derepression of INO1 was delayed by about 30 -60 min in comparison with the same cells shifted to inositol-free medium in the absence of choline (i.e. from I ϩ C Ϫ to I Ϫ C Ϫ medium; Fig.  4A). Within 2 h following the shift from I ϩ C ϩ to I Ϫ C ϩ medium, Opi1p-GFP in the wild-type strain had exited the nucleus and was distinctly observed in the perinuclear ER (Fig. 6B). At this time point, INO1 expression had increased by about 100-fold (Fig. 4A), and by 3 h INO1 expression had increased by about 400-fold (Fig. 6A), an increase comparable with that seen in this same strain after the shift from I ϩ C Ϫ to I Ϫ C Ϫ for 3 h (Fig. 4A).
In marked contrast to the behavior of Opi1p-GFP in the wildtype strain, Opi1 ffat p-GFP in the OPI1 ffat strain completely failed to exit the nucleus following a shift to inositol-free medium in the presence of choline (i.e. from I ϩ C ϩ to I Ϫ C ϩ medium; Fig. 6B). Significantly, the Opi1 ffat p-GFP-labeled puncta, observed in association with lipid droplets following the shift of OPI1 ffat from I ϩ C Ϫ to I Ϫ C Ϫ medium (Fig. 4B, 2 h, middle panel), were not observed after the shift from I ϩ C ϩ to I Ϫ C ϩ medium (Fig. 6B). Similar to Opi1 ffat p-GFP in the OPI1 ffat strain, Opi1p-GFP in scs2⌬ also failed to exit the nucleus fol-lowing the shift from I ϩ C ϩ to I Ϫ C ϩ medium. Moreover, the Opi1p-GFP-associated puncta, seen in this strain after the shift from I ϩ C Ϫ to I Ϫ C Ϫ medium, were also not observed (data not shown). Furthermore, INO1 failed to derepress in both the OPI1 ffat and scs2⌬ strains following the shift to I Ϫ C ϩ medium (Fig. 6A). These observations suggest that a pool of PA, associated with the synthesis of DAG, precursor to TAG in nascent lipid droplets (44), is sufficient to attract Opi1 ffat p or Opi1p from the nuclei of the OPI1 ffat and scs2⌬ strains, respectively, in the absence of inositol, but only when choline is also absent (compare Fig. 4B with Fig. 6B). This result is consistent with the hypothesis that synthesis of PC via the CDP-choline pathway ( Fig. 1) competes for a pool of DAG derived from PA in the ER, a pool that, in the absence of exogenous choline, is available for increased production of TAG ( Fig. 1) and lipid droplet formation (Fig. 4B).

Overall cellular PA levels increase in all three strains following a shift from I ؉ medium to I ؊ medium, whether choline is present or not, but do not correlate with INO1 expression in the OPI1 ffat and scs2⌬ strains when choline is present
To determine whether the lack of the FFAT domain in Opi1p affects PA content, we performed lipid analysis under all growth conditions described above. PA levels in the wild-type strain increased in a comparable fashion following a shift from I ϩ to I Ϫ medium, whether choline was present or not (Fig. 7, compare A with B and C with D). There was also no significant difference in the levels of PA in the wild-type strain growing in I ϩ C Ϫ medium versus I ϩ C ϩ medium (Fig. 7, compare A with C). After the shift to inositol-free medium in the absence or presence of choline (i.e. from I ϩ C Ϫ to I Ϫ C Ϫ medium or from I ϩ C ϩ to I Ϫ C ϩ medium), PA levels rose significantly in the wild-type strain. However, the kinetics of the increase in PA levels were somewhat affected by the presence of choline, initially spiking

Significance of the Opi1p-Scs2p interaction in yeast lipid metabolism
higher at 2 h following the shift to I Ϫ C ϩ (Fig. 7D) and then dropping down by 3 h (Fig. 7E) to a level comparable with that seen at 2 h after the shift to I Ϫ C Ϫ medium (Fig. 7B). At 2 h after the shift to I Ϫ C ϩ medium, INO1 expression in wild type was significantly lower (Fig. 6A) than the level seen by 2 h following the shift to I Ϫ C Ϫ medium (Fig. 4A). However, by 3 h following the shift to I Ϫ C ϩ medium (Fig. 6A), INO1 expression in wild type had reached levels as high as or higher than those observed in this strain after the shift to I Ϫ C Ϫ medium (Fig. 4). This transient contrast between the kinetics of rising PA levels and INO1 derepression, in the presence versus the absence of choline, suggests that the rising pool of PA, accompanying the removal of inositol, is not as rapidly available for interaction with Opi1p as it is when choline is absent, an issue to be taken up under "Discussion." In contrast to both the wild-type and scs2⌬ strains, PA levels in the OPI1 ffat strain were significantly lower when growing in medium lacking choline, whether inositol was present or not (compare PA levels in OPI1 ffat in comparison with wild type and scs2⌬; Fig. 7, A and B). This observation is consistent with the somewhat reduced level of INO1 expression observed in this strain in comparison with wild type after the shift to I Ϫ C Ϫ medium (Fig. 4A). Nevertheless, in each of the three strains, PA levels increased about 2-fold following the shift from I ϩ C Ϫ medium to I Ϫ C Ϫ medium (Fig. 7). Furthermore, in each of the three strains, in the absence of choline, the increase in PA that followed the shift from I ϩ C Ϫ to I Ϫ C Ϫ medium was correlated with the exit of Opi1p-GFP or Opi1p ffat -GFP, respectively, from the nucleus (Fig. 4B) and with derepression of INO1 (Fig.  4A). Strikingly, however, after a shift to inositol-free medium in the presence of choline, Opi1p-GFP and Opi1p ffat -GFP in the OPI1 ffat and scs2⌬ strains, respectively, failed to exit the nucleus, and INO1 failed to derepress (Fig. 6). Significantly, Overnight cultures grown in I ϩ C Ϫ or I ϩ C ϩ medium at 30°C in the presence of 32 P as described under "Experimental procedures" were diluted to A 600 nm ϭ 0.2 in I ϩ C Ϫ or I ϩ C ϩ medium at 30°C maintaining label constant. Cells were allowed to grow to mid-logarithmic phase, and samples (time 0) were collected for lipid analysis (A and C). The remaining cultures were filtered and resuspended in I Ϫ C Ϫ medium or I Ϫ C ϩ medium maintaining label constant. Cultures growing in I Ϫ C Ϫ were allowed to continue growing for an additional2h(B), whereas I Ϫ C ϩ cultures were allowed to grow for an additional 2 and 3 h (D and E, respectively). At these time points, cells were collected for lipid analysis. Data are expressed as cpm of radiolabel 32 P incorporated into total phospholipids per OD unit in the cell culture. Data are expressed as mean Ϯ S.D. (error bars) (n ϭ 3). **, p Ͻ 0.005; *, p Ͻ 0.05.

Significance of the Opi1p-Scs2p interaction in yeast lipid metabolism
the failure of INO1 derepression in these two strains under these conditions occurred despite rising PA levels, comparable with or higher than those observed in the wild-type strain after the shift to I Ϫ C ϩ medium (compare PA levels in all three strains; Fig. 7, B, D, and E). Thus, when choline is present, direct interaction of Opi1p and Scs2p in the ER, in addition to rising cellular PA levels, is essential for INO1 expression.

CL levels are significantly reduced in the OPI1 ffat and scs2⌬ strains
In yeast, PA serves as precursor to two separate pools of CDP-DAG. In the ER, Cds1p catalyzes the production of CDP-DAG used in the synthesis of PI and PS (Fig. 1). PA used in the synthesis of CDP-DAG in the inner mitochondrial membrane is synthesized in the ER and must be transferred to the outer mitochondrial membrane (53-55) (Fig. 1), primarily via the ERMES (35)(36)(37)(38). From the outer mitochondrial membrane, PA must be transferred to the inner mitochondrial membrane by a process requiring the Ups1p-Mdm35 complex (36,53,56). Once transferred to the inner mitochondrial membrane, PA serves as precursor to CDP-DAG, catalyzed by the mitochondrial CDP-DAG synthase, Tam41p ( Fig. 1) (54). CDP-DAG is then converted to phosphatidylglycerol phosphate in the mitochondria. Phosphatidylglycerol phosphate is then dephosphorylated to form phosphatidylglycerol, the immediate precursor of CL (54).
Despite the significant increase in the wild-type strain in the level of PA, CL content was not greatly affected by the shift from I ϩ C Ϫ to I Ϫ C Ϫ medium (in Fig. 7, compare relative changes in PA and CL levels in A with those in B). Moreover, in comparison with wild type, PA levels in the OPI1 ffat strain were somewhat reduced, whereas those in scs2⌬ were elevated, both before and after the shift to I Ϫ C Ϫ medium (Fig. 7, A and B). However, CL levels were greatly reduced in both OPI1 ffat and scs2⌬, in comparison with wild type, both before and after the shift from I ϩ C Ϫ to I Ϫ C Ϫ medium (Fig. 7, compare data in A with data in B). Because low CL content was reported in the opi1⌬ strain in combination with a Pet Ϫ phenotype (34), we assessed growth of the scs2⌬ and OPI1 ffat strains, in comparison with wild type, on YPD plates supplemented with ethidium bromide, for the Pet Ϫ phenotype. Growth of all three strains was comparable under these conditions (data not shown). Thus, basic mitochondrial function was not compromised by the changes in lipid metabolism observed in scs2⌬ and OPI1 ffat strains. Consistent with these observations, the scs2⌬ mutant was not identified among the mutations conferring the Pet Ϫ phenotype in a genome-wide screen conducted by Dunn et al. (57).
Whereas the shift of the wild-type strain from I ϩ C Ϫ to I Ϫ C Ϫ medium had little impact on CL levels (Fig. 7, compare A with  B), the presence of choline together with inositol (i.e. I ϩ C ϩ versus I ϩ C Ϫ medium) resulted in a significant decrease in CL in comparison with inositol alone (Fig. 7, compare CL levels in wild type in A with those in C). However, CL levels in the OPI1 ffat and scs2⌬ strains were significantly lower than those seen in the wild-type strain under these same conditions (i.e. in I ϩ C ϩ medium; Fig. 7C). Following the shift from I ϩ C ϩ to I Ϫ C ϩ medium, CL level in wild type decreased by an additional 3-fold in the first 2 h (Fig. 7, compare C with D) but recovered somewhat during the interval from 2 to 3 h (Fig. 7E). As described above, in contrast, the level of PA, precursor to CL, increased significantly in the wild-type strain after the shift from I ϩ C ϩ to I Ϫ C ϩ medium (Fig. 7, compare C with D and E). Thus, the continuing low level of CL after the shift to I Ϫ C ϩ medium suggests that a specific pool of PA, accessible for transport from the ER to mitochondria in wild-type cells, is specifically impacted in the presence of exogenous choline.

TAG levels increased in all three strains after a shift to inositolfree medium in the absence of choline; however, when inositol and choline were both present, neutral lipid composition was affected in distinctly different ways in each of the three strains
TAG levels rose significantly in all three strains after the shift to inositol-free medium in the absence of choline (i.e. I ϩ C Ϫ medium to I Ϫ C Ϫ medium; Fig. 8, compare A with B). However, DAG, the immediate precursor of TAG (Fig. 1), showed relatively little change in any of the three strains following a shift from I ϩ C Ϫ medium to I Ϫ C Ϫ medium (Fig. 8, compare DAG levels in A with those in B). Indeed, in the wild-type strain, shifting from I ϩ C Ϫ medium to I Ϫ C Ϫ medium had relatively little effect on any single neutral lipid category other than TAG (Fig. 8, A and B), as reported previously (43,50). In contrast, in the OPI1 ffat and scs2⌬ strains, free fatty acids (FFA) were reduced, in comparison with wild type, both before and after the shift from I ϩ C Ϫ medium to I Ϫ C Ϫ medium, and steryl ester (SE) levels in these two strains were reduced only after the shift to I Ϫ C Ϫ medium. Free sterols were also significantly reduced in the OPI1 ffat SCS2 strain in comparison with OPI1 SCS2 both before and after the shift from I ϩ C Ϫ medium to I Ϫ C Ϫ medium (Fig. 8, compare A with B).
However, in the presence of both choline and inositol (I ϩ C ϩ medium; Fig. 8C), as compared with inositol alone (I ϩ C Ϫ medium; Fig. 8A), the wild-type strain exhibited reductions of 50% or more in essentially all neutral lipids. In stark contrast to wild type, in the scs2⌬ strain, every category of neutral lipid, except free sterols, was elevated in I ϩ C ϩ medium (Fig. 8C), as compared with I ϩ C Ϫ medium (Fig. 8A). In the OPI1 ffat strain, the levels of essentially all of the individual neutral lipids growing in I ϩ C ϩ medium were intermediate between the other two strains. However, neutral lipid levels in the OPI1 ffat strain were more similar to those seen in wild type than in scs2⌬. We conclude that the higher levels of FFA, TAG, and SE observed in the scs2⌬ strain growing in I ϩ C ϩ medium are attributable to functions of Scs2p beyond its interaction with Opi1p in the ER.
After the shift to inositol-free medium in the presence of choline (i.e. from I ϩ C ϩ to I Ϫ C ϩ medium), essentially all categories of neutral lipids in the wild-type strain increased (Fig. 8, compare C with D) while remaining generally lower, especially with respect to TAG, SE, and FFA, than the levels seen in this same strain after the shift from I ϩ C Ϫ medium to I Ϫ C Ϫ medium (Fig. 8, compare data in B with data in D). However, DAG, immediate precursor to both PC via the CDP-choline pathway and TAG, a major constituent of lipid droplets (Fig. 1), is an exception. DAG levels were significantly lower in the wild-type strain growing in I ϩ C ϩ versus I ϩ C Ϫ medium (Fig. 8, compare A with C) but increased about 2-fold after the shift from I ϩ C ϩ to Fig. 8, C and D). In contrast, no significant change in DAG levels occurred in the wild-type strain following the shift to inositol-free medium in the absence of choline (i.e. from I ϩ C Ϫ to I Ϫ C Ϫ medium; Fig. 8, compare data in A with data in B). DAG levels in the OPI1 ffat and scs2⌬ strains, which were higher in I ϩ C ϩ medium than those seen in wild type, also did not change significantly after the shift from to I Ϫ C ϩ medium.

Significance of the Opi1p-Scs2p interaction in yeast lipid metabolism
However, following the shift from I ϩ C ϩ to I Ϫ C ϩ medium, both TAG and SE levels underwent significant reductions in the scs2⌬ strain, changes that were far more dramatic than those observed in OPI1 ffat or wild type (Fig. 8, compare data in C with data in D). The level of FFA was also significantly elevated in scs2⌬, in comparison with the other two strains, both before and after the shift from I ϩ C ϩ to I Ϫ C ϩ medium (Fig. 8, compare data in C with data in D). Again, these changes are presumably attributable to functions of Scs2p beyond those controlled by its interaction with Opi1p in the ER.

Effects of inositol and choline on phospholipid composition
Under steady-state growth conditions in the presence of inositol and absence of choline (I ϩ C Ϫ medium), all three strains exhibited comparable levels of CDP-DAG, PS, and PE (Fig. 9A). PI and PC levels were also comparable in the wild-type and OPI1 ffat strains in I ϩ C Ϫ medium. However, in the scs2⌬ strain, PI and PC levels were about 20 and 30% lower, respectively, in comparison with the levels observed in the other two strains (Fig. 9A). Following the shift to inositol-free medium in the absence of choline (I ϩ C Ϫ to I Ϫ C Ϫ medium), levels of PI in all three strains declined significantly, as reported previously for wild-type strains (50), reaching levels equivalent to about 20% of the PI level seen in the wild-type strain before the shift to I Ϫ C Ϫ medium (Fig. 9B). The level of PS, precursor to PE (Fig. 1), was comparable in all three strains in I ϩ C Ϫ medium and increased slightly in all three strains after the shift to I Ϫ C Ϫ medium (Fig. 9, compare data in A with data in B). However, a significant reduction in PE content was observed in all three strains when growing in the presence of choline, independent of inositol supplementation (compare PE levels in cells grown in the absence of choline (Fig. 9, A and B) with levels in cells grown in its presence (Fig. 9, C-E)) PE is synthesized from PS via two distinct pathways, localized to different cellular compartments, catalyzed either by Psd2p in the endosome or by Psd1p in the mitochondria (58) (Fig. 1). The Psd1p protein precursor is synthesized on cytoplasmic ribosomes and is processed in the mitochondria (59). Transcription of PSD1 is regulated by the Ino4p, Ino2p, and Opi1p transcription factors in response to inositol availability, whereas PSD2 is not (7).
The level of PC increased significantly in the wild-type strain (Fig. 9, A and B) when shifted to inositol-free medium in the absence of choline (i.e. from I ϩ C Ϫ to I Ϫ C Ϫ medium), consistent with previous studies (50). However, when shifted to inositolfree medium in the presence of choline (i.e. from I ϩ C ϩ to I Ϫ C ϩ medium; Fig. 9, compare data in C with data in D and E), PC levels in wild type changed minimally if at all. PC levels were significantly lower in the scs2⌬ strain than in the other two strains, both before and after the shift to I Ϫ C Ϫ medium (Fig. 9,  A and B). However, PC in the scs2⌬ strain, when growing in I ϩ C ϩ medium, was also significantly higher than in I ϩ C Ϫ medium and was comparable with the level observed in the wild-type strain growing in I ϩ C ϩ medium (Fig. 9A).

Discussion
In the present study, we compared changes in lipid metabolism and gene regulation in the wild-type, scs2⌬, and OPI1 ffat strains to determine the individual and combined effects of exogenous inositol and choline on lipid metabolism, Opi1p localization, and INO1 expression. Fig. 10 provides a visual summary of the major findings and the significance of this work. One striking outcome of this study was the discovery that the mitochondrial lipid, CL, is significantly reduced in the OPI1 ffat and scs2⌬ strains in comparison with the wild-type strain (Fig. 7). These observations, coupled with the earlier report by Luévano-Martinez et al. (34) that the opi1⌬ mutant exhibits similarly low CL levels, indicate that the reductions in CL content in these strains are most likely due to the loss of the Opi1p-Scs2p interaction in the ER, the one interaction these two proteins share in common. Moreover, we report that the scs2⌬ and OPI1 ffat mutations both confer weak Ino Ϫ phenotypes at 30°C, phenotypes that are strengthened at higher growth temperatures and in the presence of choline (Fig. 2). In contrast, the opi1⌬ mutant exhibits unregulated high-level constitutive expression of INO1 and other UAS INO -containing genes and excretes excess inositol into the growth medium (4,33,60). Thus, the opi1⌬ mutant shares no reported metabolic or regulatory phenotype with scs2⌬ other than low CL levels. To support CL biosynthesis, PA synthesized in the ER must be transferred to the outer mitochondrial membrane and transferred to the inner mitochondrial membrane to be converted to CDP-DAG by Tam41p, the mitochondrial CDP-DAG synthase (35,53,54) (Fig. 1). Optimal synthesis of CL requires transfer of PA from the ER to the outer mitochondrial membrane via the ERMES complex (38). The mitochondrial pathway for PE synthesis, similarly, requires the transit of PS from the ER to the mitochondria (55), and deletion of several genes encoding ERMES subunits results in reduced levels of both CL and PE (38). For example, deletion of the MMM1 gene, encoding an ERMES subunit, containing a conserved "synaptotagmin-like mitochondrial lipid-binding" (SMP) domain (61), is associated with significantly reduced levels of both CL and PE in the mitochondria (38). However, deletion of MDM34, encoding an outer mitochondrial membrane protein, is associated with reduced levels of CL but does not affect PE levels (38).
The reductions in CL content that we observed in scs2⌬ and OPI1 ffat strains relative to wild type (Fig. 7) are proportionately comparable with the reductions in CL reported in the mitochondrial lipids of mutants carrying deletions of ERMES subunits (38). Thus, the reductions in CL synthesis that we observed in the scs2⌬ and OPI1 ffat strains are consistent with the hypothesis that the interaction of Opi1p with Scs2p facilitates transfer of PA from the ER to the mitochondria. Moreover, simultaneous disruption of both ERMES and vCLAMP

Significance of the Opi1p-Scs2p interaction in yeast lipid metabolism
(vacuole and mitochondrial patch) resulted in a higher reduction of CL synthesis than disruption of ERMES alone (62). Thus, the residual synthesis of CL, which we observed in both the scs2⌬ and OPI1 ffat strains, compared with wild type (Fig. 7), could be the result of compensatory vCLAMP facilitation of PA transfer to mitochondria from the vacuole. Regardless, the low CL levels detected in the scs2⌬ and OPI1 ffat strains strongly support the hypothesis that the interaction of Opi1p and Scs2p facilitates optimal transfer of PA from the ER to mitochondria. However, when growing in the presence of choline, CL levels were reduced in all strains, and the relative reduction in CL levels in the presence of choline was proportionally much greater in the wild-type strain than in the OPI1 ffat and scs2⌬ strains, in which CL levels were already greatly reduced due to the loss of the Opi1p-Scs2p interaction (Fig. 7). As we discuss below, the presence of exogenous choline exerts distinctly different effects on specific pools of PA in the ER. The presence of choline alone has a significant impact on TAG content. The substantial reduction in the level of TAG observed in the wild-type strain growing in the presence of choline is consistent with the hypothesis that PC synthesis via the CDP-choline pathway competes directly for a pool of DAG that serves as precursor to TAG synthesis. Also, in contrast to the dramatic increase in PI levels observed in the presence of inositol (Fig. 9), choline is associated with only a modest increase in PC levels in the wild-type strain, regardless of inositol supplementation. These relatively small changes in PC levels in the wild-type strain growing in the presence of choline are consistent with previous reports that choline induces turnover of PC by deacylation, via the Nte1p phospholipase B (63)(64)(65)(66). However, in both the wild-type and OPI1 ffat strains, supplementation with both choline and inositol was associated with a reduction in most neutral lipids, especially TAG. In contrast, most neutral lipids were markedly elevated in the scs2⌬ strain in I ϩ C ϩ medium, as compared with the other two strains as well as with its own neutral lipid composition in I ϩ C Ϫ medium. These data are indicative of significant additional perturbations of neutral lipid metabolism related to the total loss of Scs2p function.
A major issue that remains to be discussed is the root cause of the choline-sensitive inositol auxotrophy of the OPI1 ffat and scs2⌬ strains. Both strains exhibit Ino Ϫ phenotypes that are more evident both in the presence of choline and at the higher growth temperature of 34°C (Fig. 2). This phenotype is essentially identical to the phenotype of "choline-sensitive inositol auxotrophy" as originally described in association with the CSE mutant used in the isolation of the SCS2 gene as a high copy suppressor (31). Retention of the Opi1p repressor in the ER requires its interaction with PA in the ER, and this interaction is

Significance of the Opi1p-Scs2p interaction in yeast lipid metabolism
essential both for expression of INO1 and growth of wild-type cells in the absence of inositol (15). Opi1 ffat p-GFP in the OPI1 ffat strain and Opi1p-GFP in the scs2⌬ strains, respectively, both retain the ability to interact with PA in the ER. However, neither the OPI1 ffat nor the scs2⌬ strain has the capacity for direct interaction between Opi1p and Scs2p in the ER. Importantly, after the shift to inositol-free medium in the absence of choline, Opi1 ffat p-GFP in the OPI1 ffat strain exited the nucleus and localized to distinctive puncta associated with synthesis of TAG in lipid droplet formation. Moreover, the level of INO1 derepression in the OPI1 ffat strain following the shift from I ϩ C Ϫ to I Ϫ C Ϫ medium was only slightly lower than that supported by Opi1p in the wild-type strain under the same conditions. Thus, we conclude that the pool of PA associated with lipid droplet formation in the ER, which is created by diversion of PA from PI to increasing TAG synthesis after the shift from I ϩ C Ϫ to I Ϫ C Ϫ medium, is sufficiently robust to serve as a signal for the exit of Opi1 ffat p from the nucleus. However, when choline is present in the absence of inositol, increased PC synthesis competes directly for the pool of DAG derived from PA, a pool of DAG in the ER that also serves as the immediate precursor for TAG. Thus, the presence of choline results in both reduced TAG levels and reduced lipid droplet formation. Under these conditions, the Opi1 ffat p in the OPI1 ffat strain is unable to interact with sites of lipid droplet formation in the perinuclear ER and remains in the nucleus, preventing INO1 expression. This, indeed, is the root cause of the choline-sensitive inositol auxotrophy shown in Fig. 2.

Strains
Yeast strains used are listed in Table 1. The parent strain BY4742 (MAT␣, his3⌬1, leu2⌬0, lys2⌬0, ura3⌬0) derived from S288C (67) and mutants derived from BY4742 were used. All strains were maintained on YPD plates (1% yeast extract, 2% bactopeptone, 2% glucose, and 2% agar). Mutations in the FFAT motif (OPI1 ffat ) (residues 200 -203 mutated from EFFD to ALLA) within the genomic OPI1 locus were created using the delitto perfetto method (68). The mutated sites are as described by Loewen et al. (15). Deletion mutant strains for PCT1 were generated in wild-type, scs2⌬, and OPI1 ffat strains by PCR-mediated gene replacement, as described previously (69). The plasmid pRS315 was used as a template to generate a PCR fragment for the PCT1 gene disruption. The entire open reading frame of the PCT1 gene was replaced with the LEU2 marker gene. Leucine prototrophs were screened by colony PCR to verify integration at the correct genetic locus.

Construction of strains expressing Opi1p-GFP and Opi1p ffat -GFP fusion proteins tagged at the C terminus of the OPI1 genomic locus
Fusions of GFP to the C termini of wild-type, scs2⌬, and OPI1 ffat strains were constructed by PCR-mediated gene integration at the genomic OPI1 locus using the template plasmid pFA6a-GFP (S65T)-kanMX6 (a gift from M. Longtine) (69). The insertion of GFP was confirmed by PCR, and expression of genomic Opi1p-GFP and its variants was confirmed by Western blotting.

Analysis of Ino ؊ phenotypes on solid media
The presence of choline has been shown to increase the severity of Ino Ϫ phenotypes (i.e. reduce residual growth in the absence of inositol) in a number of mutant strains in plating assays, including scs2⌬ (14). Chromatographic assessment of residual choline, performed in our laboratory on samples of agar sourced from several vendors, indicated that all of these agar samples contained varying trace levels of choline (data not shown). However, in our analysis, agar sourced from Sigma contained the lowest residual trace of choline and was therefore used exclusively in all plate assays for assessing Ino Ϫ phenotypes in this study. Yeast strains were grown to mid-logarithmic phase in synthetic complete medium containing 75 M inositol (I ϩ medium), harvested, washed with sterile distilled water, and resuspended in sterile distilled water at a concentration of 1.0 A 600 nm /ml. 10-Fold serial dilutions were then spotted onto plates containing 0 M inositol (I Ϫ ) or 75 M inositol (I ϩ ) with or without 1 mM choline (C ϩ ) and incubated at the indicated temperature of 30 or 34°C for 3 days.

Protocol for assessing the Pet ؊ phenotype
The opi1⌬ mutation was shown to confer the Pet Ϫ phenotype (34), namely the inability to tolerate the loss of the mitochondrial genome (57). Accordingly, we tested the OPI1 ffat strain for ethidium bromide sensitivity, a phenotype that is associated with complete loss of the mitochondrial genome, following the procedure used by Luévano-Martinez et al. (34). In brief, strains were cultured in YPD supplemented with 25 g/ml filtered sterilized ethidium bromide at 30°C to an A 600 nm ϭ 0.5. At this optical density, samples were subjected to serial dilution in sterile distilled water and spotted onto YPD and YPD plus 25 g/ml ethidium bromide agar plates. Strains were grown for 2 days on the plates, whereupon the plates were examined for growth conditions under which the Pet Ϫ strains fail to grow.

Protocol for growth in liquid medium and shifting cells from medium containing inositol to medium lacking inositol
All studies on cells grown in liquid medium were conducted at 30°C. As described by Gaspar et al. (43), cells were pregrown overnight in medium containing 75 M inositol with or without 1 mM choline (i.e. I ϩ C ϩ or I ϩ C Ϫ medium). The following day, cultures were diluted back to A 600 ϭ 0.2 in the same medium at the same temperature and allowed to grow to mid-logarithmic growth phase, A 600 ϭ 0.5-0.6, in I ϩ C Ϫ or I ϩ C ϩ medium, respectively. Cells were then collected by filtration, washed with I Ϫ C Ϫ or I Ϫ C ϩ medium prewarmed to 30°C, and resuspended in I Ϫ C Ϫ or I Ϫ C ϩ medium at the same temperature. Samples were harvested by filtration or centrifugation immediately after 0, 1, 2, and 3 h of growth. Inositol was then reintroduced to the cell cultures at a concentration of 75 M, and cells were harvested 30 min after inositol readdition. Harvested cells were flash frozen on dry ice after harvesting and stored at Ϫ80°C for RNA extraction. Changes in lipid metabolism and gene expression, as described below, were also measured over the same interval in a strain carrying Opi1-GFP (wild type; YCY3) and compared with its parent, BY4742 (Table 1), expressing untagged Opi1p. The level of expression of INO1 at Significance of the Opi1p-Scs2p interaction in yeast lipid metabolism each time point was not statistically different in the BY4742 strain, in comparison with the YCY3 strain at any time point before or after a shift from I ϩ C Ϫ to I Ϫ C Ϫ medium or from I ϩ C ϩ to I Ϫ C ϩ medium or following the readdition of inositol (data not shown). On this basis, the YCY3 strain was used as the "wild-type" control for analysis of gene expression, Opi1p localization, and lipid metabolism in response to inositol and choline supplementation.

RNA isolation and RT-PCR analysis
Strains were pregrown as described above in I ϩ C Ϫ or I ϩ C ϩ liquid medium at 30°C and then shifted to I Ϫ C Ϫ or I Ϫ C ϩ medium, respectively, maintaining constant growth temperature. Total RNA was isolated using the RNeasy minikit, including a DNA digestion with an RNase-free DNase set (both from Qiagen). 1 g of RNA was transcribed into cDNA using oligo(dT) 12 Non-template control (10 ng of RNA) and non-reaction control (RNase-free water) were routinely performed. The thermal program for the PCR included stage 1 (95°C for 10 min), stage 2 (95°C for 0.5 min and 60°C for 1 min for a total of 40 cycles), and stage 3 (hold at 4°C). Relative quantitation was done using the ⌬⌬Ct method (see the StepOnePlus TM user manual, Applied Biosystems). The ⌬⌬Ct represents the change in mRNA expression after ACT1 normalization relative to the wild-type control calculated as follows: 2Ϫ(Gene Ct x Ϫ ACT1 Ct x ) Ϫ (Gene Ct cr Ϫ ACT1 Ct cr ), where Gene represents the mRNA under study (INO1), x refers to the strain from which the mRNA to be tested was derived (i.e. wild type or mutant), and cr refers to the control mRNA, the value of which was derived from the level of mRNA in the BY4742 parent strain, pregrown as described above in I ϩ medium at 30°C, shifted to IϪ medium at the same temperature at time 0 h. The Ct (cycle threshold) is defined as the number of cycles required for the fluorescent signal to cross the threshold (i.e. to exceed background level). Each RT-PCR experiment was performed in triplicate.

Fluorescence microscopy
Wild type, scs2⌬, and OPI1 ffat cells, expressing genomic or plasmid versions of GFP-tagged OPI1 or OPI1 ffat , were grown overnight, as described above, at 30°C in I ϩ C Ϫ or I ϩ C ϩ medium. Before microscopy, overnight cultures were diluted to A 600 ϭ 0.2 and allowed to continue to grow to mid-logarithmic phase in I ϩ C Ϫ or I ϩ C ϩ medium at 30°C. After reaching A 600 ϭ 0.5-0.6, cultures were shifted, as described above, to I Ϫ C Ϫ or I Ϫ C ϩ medium, respectively, and allowed to grow for 2 h (A 600 ϭ 0.8 -1.0) for single time point observations or to 4 -5 h (A 600 ϭ 1.5-2) for time course observations. Cells were then concentrated to OD ϭ 12.5 by centrifugation, and 3.5-l samples of concentrated cultures were subjected to deconvolution fluorescence microscopy using a Deltaversion RT microscopy system (Applied Precision, LLC). Cells were viewed using a X71 Olympus microscope equipped with a PlanApo 100ϫ objective (1.35 numeric aperture, Olympus), FITC, and rhodamine filters and a Cool Snap HQ digital camera (Photometrics). GFP images were acquired with the FITC filter set, and RFP/mCherry images were acquired with the RD-TR-PE filter set. Five or six z-sections from each strain were inspected. The acquired images were deconvolved using soft-WoRX version 3.5.0 software (Applied Precision, LLC). The colocalization of GFP-Opi1p or GFP-Opi1p ffat with the lipid droplet marker RFP-Erg6p (Table 1) was analyzed with PCC (48) in soft-WoRX version 3.5.0. The number of prominent punctate structures associated with Opi1p ffat -GFP was manually counted in 50 cells by scanning through the z-sections.

Phospholipid composition assessed by [ 32 P]orthophosphate steady-state labeling
Changes in the composition of cellular phospholipids were determined over a time course of 3 h following a shift of actively growing cells from medium containing inositol to medium lacking inositol, followed by subsequent reintroduction of inositol into the medium. For this purpose, cells were grown overnight, as described above, in I ϩ C Ϫ or I ϩ C ϩ medium at 30°C in the presence of 10 Ci/ml [ 32 P]orthophosphate. The following day, cultures were diluted to A 600 nm ϭ 0.1 in I ϩ C Ϫ or I ϩ C ϩ medium at 30°C maintaining label at 10 Ci/ml [ 32 P] orthophosphate and allowed to grow to mid-logarithmic phase (A 600 ϭ 0.5-0.6). At this cell density, each culture was divided in half. One-half of each culture was filtered, washed with prewarmed medium containing inositol, and resuspended in I ϩ C Ϫ or I ϩ C ϩ medium at 30°C, maintaining label at 10 Ci/ml [ 32 P]orthophosphate. The other half was also filtered and then washed with prewarmed medium lacking inositol and resuspended in I Ϫ C Ϫ or I Ϫ C ϩ medium at 30°C maintaining label at 10 Ci/ml [ 32 P]orthophosphate. Samples from each culture were taken at 2 or 3 h following the shift. Labeled lipids were extracted as described by Gaspar et al. (50). The individual phospholipid species were resolved by two-dimensional thin layer chromatography (50).
For the assessment of CL content, lipids were labeled with [ 32 P]orthophosphate and extracted, as described above, and the lipid extract was analyzed by one-dimensional thin layer chromatography according to the method developed by Vaden et al. (70). Briefly, Sigma-Aldrich Silicagel on TLC plates (layer thickness 250 m) were dipped in 1.8% boric acid prepared in 100% ethanol, dried for 5 min, and baked for 15 min at 100°C. Phospholipids were separated using the solvent system chloroform/ ethanol/water/triethylamine (30/35/7/35, v/v/v/v) for at least 2 h. Phospholipid identity was based on the mobility of known Significance of the Opi1p-Scs2p interaction in yeast lipid metabolism standards and quantified on a STORM 860 PhosphorImager (Amersham Biosciences).

Neutral lipid composition assessed by steady-state labeling with [ 14 C]acetate
Cells were grown as described above for 32 P steady-state labeling, except that they were labeled in the presence of 2 Ci/ml [1-14 C]acetate (specific activity, 57 mCi/mmol) to steady state in I ϩ C Ϫ or I ϩ C ϩ medium at 30°C and then shifted to I Ϫ C Ϫ or I Ϫ C ϩ at 30°C medium respectively, maintaining label, as described above, at 2 Ci/ml [1-14 C]acetate. Changes in neutral lipid composition were monitored over a set time course of 2 or 3 h after the shift to medium lacking inositol, followed by inositol readdition. Samples were taken at 0 and 2 h following the shift to I Ϫ C Ϫ or I Ϫ C ϩ medium. 5-ml samples were mixed with 0.5 ml of 50% trichloroacetic acid and allowed to stand on ice for 20 min. Lipids were extracted and analyzed as described by Gaspar et al. (50). Labeled lipids on the chromatograms were quantified on a STORM 860 PhosphorImager (Amersham Biosciences), and metabolites were identified as described previously (50).