The TGL2 Gene of Saccharomyces cerevisiae Encodes an Active Acylglycerol Lipase Located in the Mitochondria*

The Saccharomyces cerevisiae Tgl2 protein shows sequence homology to Pseudomonas triacylglycerol (TAG) lipases, but its role in the yeast lipid metabolism is not known. Using hemagglutinin-tagged Tgl2p purified from yeast, we report that this protein carries a significant lipolytic activity toward long-chain TAG. Importantly, mutant hemagglutinin-Tgl2pS144A, which contains alanine 144 in place of serine 144 in the lipase consensus sequence (G/A)XSXG exhibits no such activity. Although cellular TAG hydrolysis is reduced in the tgl2 deletion mutant, overproduction of Tgl2p in this mutant leads to an increase in TAG degradation in the presence of fatty acid synthesis inhibitor cerulenin, but that of Tgl2pS144A does not. This result demonstrates the lipolytic function of Tgl2p in yeast. Although other yeast TAG lipases are localized to lipid particles, Tgl2p is enriched in the mitochondria. The mitochondrial fraction purified from the TGL2-overexpressing yeast shows a strong lipolytic activity, which was absent in the tgl2 deletion mutant. Therefore, we conclude that Tgl2p is a functional lipase of the yeast mitochondria. By analyzing phenotypic effects of TGL2-deficient yeast, we also find that lipolysis-competent Tgl2p is required for the viability of cells treated with antimitotic drug. The addition of oleic acid, the product of Tgl2p-catalyzed lipolysis, fully complements the antimitotic drug sensitivity of the tgl2 null mutation. Thus, we propose that the mitochondrial Tgl2p-dependent lipolysis is crucial for the survival of cells under antimitotic drug treatment.

The Saccharomyces cerevisiae Tgl2 protein shows sequence homology to Pseudomonas triacylglycerol (TAG) lipases, but its role in the yeast lipid metabolism is not known. Using hemagglutinin-tagged Tgl2p purified from yeast, we report that this protein carries a significant lipolytic activity toward long-chain TAG. Importantly, mutant hemagglutinin-Tgl2p S144A , which contains alanine 144 in place of serine 144 in the lipase consensus sequence (G/A)XSXG exhibits no such activity. Although cellular TAG hydrolysis is reduced in the tgl2 deletion mutant, overproduction of Tgl2p in this mutant leads to an increase in TAG degradation in the presence of fatty acid synthesis inhibitor cerulenin, but that of Tgl2p S144A does not. This result demonstrates the lipolytic function of Tgl2p in yeast. Although other yeast TAG lipases are localized to lipid particles, Tgl2p is enriched in the mitochondria. The mitochondrial fraction purified from the TGL2-overexpressing yeast shows a strong lipolytic activity, which was absent in the tgl2 deletion mutant. Therefore, we conclude that Tgl2p is a functional lipase of the yeast mitochondria. By analyzing phenotypic effects of TGL2deficient yeast, we also find that lipolysis-competent Tgl2p is required for the viability of cells treated with antimitotic drug. The addition of oleic acid, the product of Tgl2p-catalyzed lipolysis, fully complements the antimitotic drug sensitivity of the tgl2 null mutation. Thus, we propose that the mitochondrial Tgl2p-dependent lipolysis is crucial for the survival of cells under antimitotic drug treatment.
Lipases or triacylglycerol (TAG) 3 hydrolases (EC 3.1.1.3) are ubiquitous enzymes found from bacteria to humans and catalyze the hydrolysis (and synthesis) of ester bonds in relatively long-chain acylglycerols, comprising a subclass of the esterases. Lipase-catalyzed hydrolysis of TAG primarily produces fatty acid and diacylglycerol (DAG). DAG can be further degraded to fatty acid, monoacylglycerol (MAG), and/or glycerol. Both TAG and its degradation products perform important functions in an organism, and the regulation of lipid metabolism is critical for growth and proliferation of all types of cells. Per-turbed lipid homeostasis has been linked to lipotoxic cell death pathways and various metabolic disorders (1,2).
The catalytic center of lipases consists of an S . . . D/E . . . H triad which is found in the active site of ␣/␤ hydrolase fold enzymes and serine proteases, although the order of the three residues is different (3)(4)(5). In particular, serine is essential for catalysis and participates, with aspartic acid/glutamic acid and histidine, in the charge relay system. Serine invariably occurs in a highly conserved sequence, GXSXG, in which the first glycine residue is replaced by alanine in some lipases, commonly in Bacillus lipases (3)(4)(5)(6)(7).
In the yeast Saccharomyces cerevisiae, TAG is mobilized by three lipases, Tgl3p, Tgl4p, and Tgl5p (8 -10). Interestingly, all three proteins are embedded in lipid particles, which store energy in the form of neutral lipids, predominantly TAGs and steryl esters. In addition to being simple storage sites for energy, lipid particles play diverse physiological roles, including communication and transfer of molecules through organellar associations (11). Because the S. cerevisiae lipid particles contain many enzymes involved in the metabolism of TAGs and steryl esters, it has now become clear that these particles actively participate in lipid metabolism (12).
Tgl3p, Tgl4p, and Tgl5p do not show homology to other lipases identified so far, except that all three carry the lipase consensus sequence (G/A)XSXG. Serine 315 in this sequence is essential for the catalytic activity of Tgl4p, the functional ortholog of mammalian adipose TAG lipase (10). Most recently, it has been reported that Tgl4p is activated by the cyclindependent kinase Cdk1/Cdc28 via phosphorylation at threonine 675 and serine 890 and that lipolysis catalyzed by the phosphorylated Tgl4p contributes to the early bud formation in late G 1 phase of the cell cycle (13). Tgl5p may act synergistically with Tgl4p and play a minor or regulatory role in TAG hydrolysis, because single deletion of TGL5 results in no reduction in TAG degradation in vivo in the presence of the fatty acid synthesis inhibitor cerulenin, whereas the tgl4⌬tgl5⌬ double mutant exhibits significant reduction in this activity compared with the tgl4⌬ single mutant (9). Tgl3p carries relatively high lipolytic activity among the three lipases, and together with Tgl4p, it constitutes the majority of lipolytic activity in lipid particles and whole cells (8 -10).
Meanwhile, a TAG lipase activity has been detected from the mitochondrial fraction of S. cerevisiae (14,15). None of the known yeast lipases can account for such mitochondrial activity, as evidenced by their subcellular localization to lipid particles. A mitochondrial protein carrying the lipolytic activity still needs to be identified.
The S. cerevisiae Tgl2p is considered to be a potential TAG lipase based on the sequence homology to Pseudomonas lipases (16). Initially, TGL2 has been isolated as a yeast gene capable of complementing the Escherichia coli DAG kinase mutation, which results in lethal accumulation of DAG in the presence of hydroquinone ␤-D-glucopyranoside arbutin (16). Because the expression of TGL2 in this mutant produces E. coli lysates harboring no DAG kinase activity but lipolytic activity toward relatively short-chain TAGs and DAGs, it is plausible that TGL2 suppresses the toxicity via the DAG hydrolytic activity of its gene product (16). However, no lipolytic activity of Tgl2p has been demonstrated in S. cerevisiae. The S. cerevisiae DAG kinase that catalyzes the formation of phosphatidate from DAG has recently been identified (17,18).
To unravel the biological function of Tgl2p, we conducted enzymatic analysis of the protein and phenotypic characterization of its gene deletion and overexpression strains. Here, we report the identification of Tgl2p as an active lipase of the yeast mitochondria based on the data obtained from subcellular fractionation and in vivo and in vitro experiments on lipid hydrolysis. By phenotypic analysis of TGL2-deficient yeast, we also demonstrate that Tgl2p-catalyzed lipolysis or its product oleic acid is required for the viability of cells treated with tubulintargeting agents.
Plasmids and Cloning-To generate plasmid pYNO4-HA-TGL2 expressing the hemagglutinin (HA)-tagged TGL2 under the control of the GAL1 promoter, a PCR product containing the TGL2 open reading frame with an EcoRI site before the start codon and a ClaI site after the stop codon was cloned into the EcoRI and ClaI sites of pYNO4. Plasmid pYNO4 was derived from pRS314 (CEN6, TRP1) by inserting into its KpnI and SacI sites a 1-kb cassette carrying the GAL1/10 promoter, triple HA repeats, and multiple cloning sites. The amino acid sequence of a peptide linked to the N terminus of Tgl2p is MVGYPYDVP-DYAGYPYDVPDYAGSYPYDVPDYAAQCGRSRTSGSPGL-QEFPEAWRKSIKK (residues corresponding to three copies of the HA epitope are underlined).
For construction of plasmid pYNO4-HA-TGL2 S144A expressing the mutant gene HA-TGL2 S144A from the GAL1 promoter, site-directed mutagenesis was performed using primers 5Ј-CGCGGGCCCAGCATAAAAAAAATGAAA-3Ј and 5Ј-ATCGGCAGTCTAGTCCCCCCATTGCGTGTGC-3Ј (the underlined nucleotide converts serine 144 (TCA) into alanine 144 (GCA)). A PCR fragment encompassing the first 454 bp of the mutated TGL2-coding region was used to replace the equivalent wild-type sequence in pYNO4-HA-TGL2. The resulting plasmid pYNO4-HA-TGL2 S144A was sequenced to verify the mutation.
Purification of HA-tagged Tgl2p-Yeast cells (strain YHY058d2; tgl2⌬) harboring pYNO4, pYNO4-HA-TGL2, or pYNO4-HA-TGL2 S144A were grown at 30°C overnight in synthetic glucose medium lacking tryptophan (0.7% yeast nitrogen base without amino acids, 2% glucose, 0.5% casamino acid, and 20 g/ml adenine and uracil). Cultures were harvested, washed twice with water, and grown at 30°C in 100 ml of synthetic galactose medium lacking tryptophan to exponential phase (ϳ4 ϫ 10 7 cells/ml). Whole yeast lysates were prepared by freeze-thawing cells (ϳ4 ϫ 10 9 cells) in liquid nitrogen (four times), followed by vortexing in the presence of glass beads, and clarified by centrifugation (30 min, 15,700 ϫ g). The HA-Tgl2 and HA-Tgl2 S144A fusion proteins were identified by immunoblot analysis with an anti-HA antibody (catalog number sc-7392 horseradish peroxidase; Santa Cruz Biotechnology, Inc. (Santa Cruz, CA)). To isolate HA fusion proteins, cell lysates were incubated with 50 l of anti-HA affinity matrix (clone 3F10, Roche Applied Science) at 4°C overnight. The beads were washed extensively as described before (21). After washing with 1 ml of 20 mM sodium phosphate buffer (pH 7.4), the beads were resuspended with 200 l of the same buffer and used for a lipase assay. For protein analysis, proteins bound to the anti-HA affinity beads were eluted by the addition of 1% SDS-containing TE buffer (10 mM Tris-HCl, pH 8.0, 1 mM EDTA) and 15 min of boiling, separated on an SDS-polyacrylamide gel, and visualized by Sypro Ruby staining (Molecular Probes, Inc., Eugene, OR). HA-Tgl2p and HA-Tgl2p S144A levels were quantified using the fluorescence gel imaging system (VersaDoc 5000 MP, Bio-Rad). Bovine serum albumin was used as a standard.
Lipase Assay in Vitro-Lipase assay buffer was prepared by mixing 150 l of 50 mM Tris-HCl (pH 8.0) and 20 l of a bovine serum albumin solution (20 mg/ml). Lipids (TAG, DAG, MAG, cardiolipin, or phosphatidylcholine) were added to a final concentration of 2-4 mM. The mixture was sonicated at 37°C until the solution became cloudy (Ն4 min). After the addition of 50 l of 200 mM MgCl 2 , the mixture was prewarmed at 37°C (total volume, 220 l). Lipase reaction was initiated by mixing 220 l of prewarmed solution and 200 l of protein preparation (mitochondrial fractions or anti-HA affinity-purified proteins). An aliquot was taken out at the start (time 0) to determine the initial lipid concentration in the reaction mixture. After incubation for 1 h at 37°C, lipids were immediately separated by TLC. If necessary, lipids were extracted by the addition of 1 ml of chloroform before TLC separation.
Lipid Analysis-Routine analysis of lipid samples (TAG, DAG, or MAG) was done as described previously (22) with modifications. Lipids were applied as 2-l spots on Silica Gel 60 G plates (catalog number 5721, Merck) and separated by TLC in the solvent mixture of hexane/diethyl ether/acetic acid (50: 50:1, v/v/v). Lipid spots were identified under a UV lamp after spraying a primuline solution (5 mg in 100 ml of acetone/water (80:20, v/v)) or 50 mM CuCl 2 solution. Analysis of phospholipids (cardiolipin or phosphatidylcholine) was done by following a published procedure (23). Lipase activity was measured by comparing the amount of lipid degraded after 1 h of incubation relative to the amount at time 0. Known amounts of lipid were run on the TLC plates to obtain the linear standard curve, and the amount of each lipid spot after 1 h of incubation was determined from the curve. The fluorescence imaging system was employed to integrate lipid spot intensities. TAGs (triolein, tripalmitin, tricaprylin, and tributyrin), DAGs (diolein, dipalmitin, and dicaprylin), MAGs (monoolein, monopalmitin, and monocaprylin), and phospholipids (cardiolipin and phosphatidylcholine) were purchased from Sigma.
Analysis of TAG Degradation in Vivo-To obtain samples grown in the presence or absence of cerulenin, tgl2⌬ cells (strain YHY058d2) and isogenic wild-type cells were grown in synthetic glucose medium at 30°C overnight and inoculated to A 600 nm ϭ 2 in the fresh medium supplemented with 10 g/ml cerulenin (dissolved in ethanol) or ethanol alone. Meanwhile, tgl2⌬ cells harboring pYNO4, pYNO4-HA-TGL2, or pYNO4-HA-TGL2 S144A were grown in synthetic glucose medium lacking tryptophan at 30°C overnight and inoculated to A 600 nm ϭ 3 in tryptophan-free, synthetic galactose medium supplemented with 10 g/ml cerulenin or ethanol. Yeast growth was monitored by measuring the absorbance at 600 nm. At the times indicated, 10-ml aliquots were pelleted, washed with distilled H 2 O, instantly frozen in liquid nitrogen, and stored at Ϫ70°C. Cells were resuspended in 0.5 ml of sorbitol buffer (40 mM potassium phosphate, pH 6.5, 0.5 mM MgCl 2 , 1.2 M sorbitol) and incubated with 1 l of ␤-mercaptoethanol and 30 l of 5 mg/ml zymolyase 100T (catalog number 120493, Seikagaku Corp.) for 15 min at 30°C. After spheroplast disruption, lipids of whole yeast lysates were extracted by the procedure of Folch et al. (24) and analyzed as described above. TAGs were quantified by optical scanning at 325 nm with triolein as a standard.
Purification of Yeast Mitochondria-Crude mitochondrial fractions were prepared by the published method (25) and loaded onto a sucrose gradient to remove the contaminated endoplasmic reticulum and vacuole (26). After centrifugation at 134,000 ϫ g in a Beckman SW41 Ti swinging bucket rotor for 1 h at 2°C, pure mitochondria were recovered from the 32%/ 60% interface of the gradient. The protein concentration was adjusted to 10 mg/ml after pelleting pure mitochondrial fraction at 10,000 ϫ g.

RESULTS
TAG Lipase Activity of Tgl2p-To determine the lipolytic activity of Tgl2p, HA-tagged Tgl2p was purified close to homogeneity from a tgl2⌬ strain overexpressing this fusion protein.
The anti-HA affinity-purified fraction exclusively contained HA-Tgl2p (ϳ42 kDa) and its degradation product (ϳ27 kDa), except for a few minor proteins that were also present in cells harboring the vector control (Fig. 1A). This affinity preparation showed a significant level of TAG lipase activity, 600 Ϯ 60 nmol mg Ϫ1 min Ϫ1 , when the assay was performed at pH 8 and 37°C using 3.6 mM triolein (C18:1) as substrate (Fig. 1B). At this concentration, we detected only DAG and fatty acid on TLC plates. When the concentration of triolein was low (below 0.7 mM), DAG was further hydrolyzed to MAG and fatty acid.
Of importance, the enzymatic activity was completely lost after mutating serine 144 in the AXSXG sequence of Tgl2p to alanine 144 (Fig. 1B). Nonetheless, the S144A mutation had no effect on the expression profile of Tgl2p as well as the growth rate of yeast cells. When the mutant Tgl2p S144A was epitopetagged, overexpressed, and purified like the wild type, the pattern of co-purified proteins and the amount of Tgl2p S144A were almost identical to those of HA-Tgl2p (Fig. 1A, lane 3). Thus, the observed lipolytic activity was indeed due to Tgl2p, not any other protein tightly associated with it, and serine 144 of the lipase active site was essential for this activity. It appears that small epitope (HA) tagging does not interfere with the catalytic activity of Tgl2p.
In Vivo TAG Degradation by Tgl2p-To examine whether Tgl2p functions as a TAG lipase in vivo, degradation of cellular TAGs was monitored in the presence or absence of cerulenin. Cerulenin inhibits the de novo synthesis of fatty acids and thus induces TAG hydrolysis (28). With cerulenin treatment, TAG hydrolysis was reduced in tgl2⌬ cells in comparison with the wild-type cells but not completely abolished as expected from the presence of other yeast lipases (Fig. 2B). Furthermore, overexpression of the HA-TGL2 gene in the tgl2⌬ mutant resulted in an increase in TAG degradation in the presence of cerulenin (Fig. 2D). On the contrary, no alteration of the degradation rate resulted from elevated expression of the mutated gene (HA-TGL2 S144A ) under the same condition (Fig. 2D). Anti-HA immunoblot analysis confirmed that the relative amounts of HA-Tgl2p and HA-Tgl2p S144A were almost identical in our protein extracts with or without cerulenin treatment (data not shown). Significant change in the growth rate was not observed in the presence of cerulenin, although tgl2⌬ cells grew slightly slower than wild-type cells when glucose was used as the carbon source (Fig. 2, A and C). These data show that Tgl2p is required for degradation of cellular TAGs, and serine 144 of the protein is indeed essential for TAG hydrolysis, confirming the lipolytic function of Tgl2p in yeast.
Catalytic Properties of the Tgl2 Lipase-To characterize the Tgl2 lipase, we carried out enzymatic analysis of Tgl2p under various reaction conditions (Fig. 3). First, the dependence on the acyl chain length of substrate molecule was determined with four different TAGs, tributyrin (C4:0), tricaprylin (C8:0), tripalmitin (C16:0), and triolein (C18:1, ⌬9). Tgl2p purified by means of anti-HA affinity preferred tributyrin as substrate, showing a specific activity of 3.2 Ϯ 0.2 mol mg Ϫ1 min Ϫ1 . When the activity toward tributyrin was set at 100%, the activity relative to tricaprylin, tripalmitin, and triolein was 72, 6, and 19%, respectively (Fig. 3A). The TAG hydrolytic activity of Tgl2p increased as the chain length of saturated fatty acids decreased. Tgl2p was more active toward unsaturated fatty acid C18:1 containing triolein than saturated fatty acid C16:0 carry-ing tripalmitin. This tendency is reminiscent of a lipase purified from rat liver mitochondria (29). Although lipases use TAGs with long acyl chains (Ͼ10 carbon atoms) as substrates, they also can perfectly hydrolyze short-chain TAGs (30). Because   Tgl2p showed optimal activity toward tributyrin, further enzymatic analyses were performed using tributyrin as substrate.
To determine the temperature optimum, tributyrin hydrolysis was measured at 25, 37, 55, and 70°C (Fig. 3B). Although the activity reached a maximum at 37°C, which was set at 100%, Tgl2p was still active at high temperatures, retaining 40 and 39% activity at 55 and 70°C, respectively. Lipases are known to be stable at high temperatures. At 25°C, Tgl2p showed 50% activity. Regarding the pH optimum, Tgl2p was most active at pH 8, whereas it maintained more than 40% activity at pH 6 (53%) and pH 10 (42%) (Fig. 3C).
Next, the dependence on the divalent metal ion was monitored by replacing Mg 2ϩ in the reaction mixture with Ca 2ϩ , Cu 2ϩ , and Zn 2ϩ . In comparison with the MgCl 2 -containing sample, the tributyrin hydrolytic activity slightly increased by ϳ20% in the presence of CaCl 2 (Fig. 3D). In the cases of CuCl 2 and ZnCl 2 , we were unable to measure lipolytic activity because precipitates were formed immediately upon the addition of these salts. Intriguingly, there was little need for divalent cations in the Tgl2 lipase activity. The addition of EDTA up to 50 mM did not abolish this activity (Fig. 3D).
Mitochondrial Localization and Substrate Specificity of Tgl2p-Subcellular location of Tgl2p is of interest because all three yeast lipases, Tgl3p, Tgl4p, and Tgl5p, are localized to the same organelle, the lipid particle. Using anti-HA immunofluorescence microscopy, we localized Tgl2p to the mitochondria after overexpressing the gene from the GAL1 promoter. It appeared that Tgl2p co-localized with the mitochondrial DNA as shown by the overlap of HA-Tgl2p staining with DNA staining excluding the nuclear DNA (Fig. 4A). The staining pattern of the mutant HA-Tgl2p S144A was identical to that of HA-Tgl2p (data not shown). We were unable to localize Tgl2p expressed from the endogenous TGL2 promoter, most likely due to its low abundance (data not shown). Recently, Tgl2p has been localized to the mitochondria using large scale cloning under the control of the moderately strong TEF1 promoter, green fluorescent protein tagging, and high resolution microscopy (31).
The mitochondrial localization of Tgl2p was confirmed by immunoblot analysis of subcellular fractions as shown in Fig.  4B. We isolated pure mitochondria with high yield from the crude mitochondrial fraction through a three-step sucrose gradient. In addition to the abundant Tgl2p obtained by galactose induction, we were able to detect the low level Tgl2p expressed from its native promoter in the pure mitochondria (Fig. 4B, lane  3). Small amounts of Tgl2p were also present in the cytosol and microsome when the protein was highly produced from the GAL1 promoter (Fig. 4B, lanes 4 and 5). This appears to be an artifact of overproduction because Tgl2p in its endogenous level was exclusively recovered with pure mitochondria just like the mitochondrial marker protein porin (Fig. 4B, lane 3). The S144A mutation did not cause mislocalization of Tgl2p (data not shown). In a separate experiment, we confirmed that Tgl2p was not present in the lipid particle or vacuole (data not shown).
As described above, the mitochondrial fraction of S. cerevisiae carries lipolytic activity (14,15). The mitochondrial lipolytic activity was not detectable in tgl2⌬ cells, but expression of HA-TGL2 from the native TGL2 promoter resulted in a specific activity of 0.4 Ϯ 0.1 nmol mg Ϫ1 min Ϫ1 (Table 1). We were unable to recognize HA-Tgl2p produced from its endogenous promoter by visualizing mitochondrial proteins with Sypro Ruby staining even after anti-HA affinity isolation (Fig. 5, lanes  (YHY058d2; tgl2⌬) harboring the pYNO4-HA-TGL2 plasmid was grown to mid-log phase at 30°C. Fixed cells were stained for HA-Tgl2p with anti-HA antibody and fluorescein isothiocyanate-conjugated anti-mouse IgG by indirect immunofluorescence. Nuclear DNA (large object marked by an arrow) and mitochondrial DNA (small dots) were visualized by staining with 4Ј,6-diamidino-2Ј-phenylindole (DAPI). Bar, 4 m. B, subcellular fractionation. YHY058d2 strain expressing HA-TGL2 either from the TGL2 promoter (endogenous level) or from the GAL1 promoter (overproduced level) was used to prepare cellular fractions for immunoblot analysis. After removing cell debris and nuclei from the homogenized spheroplasts by centrifugation, the supernatant was centrifuged at 12,000 ϫ g to pellet the crude mitochondrial fraction. The pellet was loaded onto a sucrose density gradient. The supernatant was further centrifuged at 100,000 ϫ g to separate cytosolic fraction (supernatant; lane 5) and microsomal fraction (pellet; lane 4). Pure mitochondria were recovered from the 32%/60% interface of the gradient (lane 3). Interfaces of 15%/23% (lane 1) and 23%/32% (lane 2) sucrose were also analyzed. The blots were probed with anti-HA and anti-porin antibodies, respectively. A highly sensitive enhanced chemiluminescent substrate, West Dura substrate (Pierce), was used to detect antibody-conjugated horseradish peroxide on immunoblots. Affinity-purified from mitochondrial fraction 13,000 Ϯ 700 33,000 TGL2 S144A overexpression Affinity-purified from mitochondrial fraction ND HA-Tgl2p was seen only from immunoblots using highly sensitive enhanced chemiluminescence, which could detect femtogram amounts of antigen (e.g. see Fig. 4B). Overexpression of HA-TGL2 led to ϳ50-fold amplification of the mitochondrial lipolytic activity, whereas that of HA-TGL2 S144A showed no increase in the activity ( Table 1). As shown in Fig. 5, HA-Tgl2p was significantly overproduced and accumulated in the mitochondria following galactose induction. To our surprise, the lipolytic activity (13,000 Ϯ 700 nmol mg Ϫ1 min Ϫ1 ) of HA-Tgl2p purified from the mitochondrial fraction was enhanced by a factor of ϳ22 in comparison with that (600 Ϯ 60 nmol mg Ϫ1 min Ϫ1 ) purified from the whole cell lysate (Fig. 1B and Table 1). Mutant Tgl2p S144A isolated from the mitochondria was completely inactive (Table 1).
Although the mitochondrial Tgl2p exhibits a strong TAG hydrolytic activity toward long-chain TAG, this enzyme may have other substrates because TAGs are enriched in lipid particles and not in mitochondria. It is possible that mitochondrial phospholipids, such as cardiolipin, may be a physiological substrate for the Tgl2 lipase. Thus, we assayed various lipids for substrate specificity of the Tgl2 lipase purified from the mitochondrial fraction (Table 2). Notably, Tgl2p was incapable of hydrolyzing cardiolipin or phosphatidylcholine. Of the acylglycerols tested, Tgl2 lipase showed strong lipolytic activity toward DAGs as well as TAGs and not toward MAGs. Similarly to the Tgl2 lipase purified from whole cell lysates, the enzyme isolated from the mitochondrial fraction was very active toward short chain TAGs and DAGs in addition to triolein and diolein, although it was completely inactive toward tripalmitin and dipalmitin. It has been reported that both pancreatic lipase and adipose tissue lipase hydrolyze DAGs better than TAGs but show little activity against MAGs (32,33). None of these lipids were hydrolyzed by mutant Tgl2p S144A isolated from the mitochondria (data not shown). Taken together, our results imply that Tgl2p purified from the mitochondria possesses a strong lipase activity.
Phenotypic Effects of the tgl2⌬ Mutation-The single null mutant tgl3⌬, tgl4⌬, or tgl5⌬, the double mutant tgl4⌬tgl5⌬, and even the triple mutant tgl3⌬tgl4⌬tgl5⌬ grow like wild type (9). Similarly, the TGL2 gene is neither essential nor important for cell growth (Fig. 2A). Deletion of TGL2 resulted in no growth defect in rich (YPD) or minimal medium at 16, 30, and 37°C. No cold or temperature sensitivity was observed for tgl2⌬ cells. Also, yeast growth on galactose or glycerol was not affected by the deletion (data not shown).
In a search for phenotypes associated with TGL2-deficient yeast, we found that the tgl2⌬ mutant was more sensitive than the wild type to the anti-microtubule drug, benomyl (Fig. 6A). This phenotype was also confirmed by using another anti-microtubule drug, nocodazole (data not shown). As a control, tgl1⌬ cells defective for steryl ester hydrolysis were also tested for the benomyl sensitivity. Tgl1p produces sterols and fatty acids by mobilizing steryl esters in the lipid particle (34). Similarly to the wild-type cells, tgl1⌬ cells were not hypersensitive to benomyl (Fig. 6A). The benomyl-sensitive phenotype of the tgl2⌬ mutation was rescued by expressing HA-TGL2 from the endogenous promoter. No suppression was observed by introducing the lipolysis-defective HA-TGL2 S144A gene (Fig. 6B). Apparently, Tgl2 lipase is not required for the microtubule assembly or stability, because tgl2⌬ cells exhibit no appreciable changes in the microtubule-based cytoskeleton or cold-sensitive growth. 4 Because microtubules depolymerize at low temperature, mutants having unstable microtubule structure would be expected to grow more slowly than the wild type below 16 -17°C.
We further examined whether benomyl sensitivity was suppressed by the addition of the fatty acid, the product of lipolysis,   1 and 4), pRS314 -796-HA-TGL2 (lanes 2 and 5), or pYNO4-HA-TGL2 (lanes 3 and 6) was analyzed by SDS-PAGE and stained with Sypro Ruby. ϳ250 g of protein was loaded in lanes 1-3 to see the mitochondrial protein pattern. In lanes 4 -6, HA-Tgl2p was immunoprecipitated from ϳ250 g of mitochondrial preparation using anti-HA affinity matrix. The asterisk indicates the position of a potential Tgl2p-interactor, Mir1p, which exists only in the mitochondrial fraction.

TABLE 2 Substrate specificity of the mitochondrial Tgl2 lipase
HA-Tgl2p affinity-purified from mitochondrial fraction (prepared from tgl2⌬ strain overexpressing HA-TGL2) was tested for substrate specificity. The final concentration of all lipids was 3.6 mM. Details of the assay conditions are given under "Experimental Procedures." Values (mean Ϯ S.D.) are calculated from 3-7 assays using 2-3 separate mitochondrial preparations. ND, not detected. to the medium. Of the four major fatty acids of yeast, palmitic (C16:0), palmitoleic (C16:1, ⌬9), stearic (C18:0), and oleic (C18:1, ⌬9) acid, only oleic acid fully complemented the benomyl-sensitive phenotype of TGL2-deficient yeast (Fig. 6A). Butyric acid (C4:0) was also tested because tributyrin served as the optimal substrate for Tgl2 lipase in vitro. It did not suppress the benomyl sensitivity of the tgl2⌬ null mutation on the YPD plate (Fig. 6A). Additionally, caprylic acid (C8:0) failed to suppress the benomyl sensitivity (data not shown). Both wild-type and tgl2⌬ cells grew equally well on medium containing each fatty acid in the absence of benomyl with the exception of palmitoleic acid (Fig. 6A). The addition of palmitoleic acid to the YPD medium was toxic to yeast cells. Of significance, oleic acid rescued the benomyl sensitivity of yeast cells defective for Tgl2pdependent lipolysis in a concentration-dependent manner (Fig.  6B). The addition of 1 mM oleic acid was ineffective for suppression of drug sensitivity, whereas 2 mM oleic acid fully suppressed the defect. These data indicate that lipolytic activity of Tgl2p is required for its physiological role, and small epitope (HA) tagging does not interfere with this function. More specifically, our data imply that Tgl2 lipase improves the viability of benomyl-treated tgl2⌬ cells via oleic acid production, and the limiting level of this unsaturated fatty acid is responsible for the detrimental consequence of perturbing microtubule assembly in these cells.
Notably, valuable information on the substrate specificity of Tgl2 lipase in vivo can be obtained from the present fatty acid suppression data, because the benomyl sensitivity of TGL2-deficient yeast reflects a defect in Tgl2p-dependent lipolysis. It is noteworthy that the major fatty acid distribution in cellular TAGs isolated from the triple mutant tgl3⌬tgl4⌬tgl5⌬ exhibits an increase in C14:0, C16:0, C16:1, and C26:0 fatty acids but a decrease in C18:0 and C18:1 species (9). This suggests that besides Tgl3p, Tgl4p, and Tgl5p, other lipases capable of hydrolyzing C18:0 and C18:1 containing TAGs do exist in S. cerevisiae. It appears that Tgl2 lipase is highly specific toward a C18:1 containing TAG and DAG ( Fig. 6 and Table 2).

DISCUSSION
We present several lines of evidence that TGL2 encodes a functional lipase located in the S. cerevisiae mitochondria. First, HA-Tgl2p purified close to homogeneity from a tgl2⌬ strain overproducing this fusion protein exhibits high TAG hydrolytic activity toward emulsified triolein. The S144A mutation in the lipase consensus motif (G/A)XSXG abolishes this activity. Second, deletion of TGL2 results in an increase in the cellular TAG content. Although overexpression of HA-TGL2 leads to an increase in TAG degradation in the presence of fatty acid synthesis inhibitor cerulenin, no significant change is observed by elevating the mutant HA-TGL2 S144A gene expression. Finally, Tgl2p is recovered with pure mitochondria upon cellular fractionation, and Tgl2p-specific lipolytic activity is present in isolated mitochondria. Like serine 315 of Tgl4p, serine 144 of Tgl2p is essential for catalytic activity as evidenced by the loss of lipolytic activity both in vivo and in vitro when the residue is mutated to alanine. Mutant Tgl2p S144A having no catalytic activity is synthesized at levels comparable with the native protein and correctly localized to the mitochondria. In summary, we provide evidence that Tgl2p plays a direct role in yeast TAG mobilization for the first time.
Tgl2p is distinct from other yeast lipases with respect to subcellular localization. Previously, Schousboe (14, 15) has described a mitochondrial lipase activity from S. cerevisiae. This activity constituted ϳ46% of the total intracellular lipolytic activity and seemed to have no requirement for free Mg 2ϩ or Ca 2ϩ . It appears that Tgl2p also has little requirement for free Mg 2ϩ and Ca 2ϩ as shown in Fig. 3D. The reported lipase activity of intact mitochondria was ϳ6.7 nmol mg Ϫ1 min Ϫ1 at a TAG concentration of 9.0 mM, pH 7.5, and 30°C (15). The mitochondrial fraction prepared from cells expressing endogenous levels of TGL2 showed a specific activity of 0.4 Ϯ 0.1 nmol mg Ϫ1 min Ϫ1 at pH 8 and 37°C, using 3.6 mM triolein (Table 1). Thus, it is not clear at present whether Tgl2p is the sole lipase located in the mitochondria, although the activity in this organelle was undetectable in the absence of TGL2. Also, the two activities cannot be compared directly because the triolein concentration and other assay conditions are not identical. Nonetheless, Tgl2p possesses a strong lipase activity when purified from the mitochondria (Tables 1 and 2). To our knowledge, it is the first lipase identified in the yeast mitochondria.
An important clue to the physiological role of Tgl2p may be further provided by the current phenotypic analysis of the null mutation. The benomyl and nocodazole sensitivity of the TGL2-deficient yeast combined with physical interaction between Tgl2p and ␣-tubulin suggests that this protein may participate in the microtubule-mediated function (Figs. 6 and  7). Significantly, lipolysis catalyzed by Tgl2p is crucial for this role because in contrast to expression of the wild-type TGL2 gene, the benomyl sensitivity of the null mutation was not rescued by expression of the mutant TGL2 S144A gene encoding inactive lipase (Fig. 6B). Anti-microtubule drugs like vinblastine and nocodazole specifically inhibit palmitoylation of tubulin, which has been implicated in hydrophobic interaction between microtubules and various intracellular membranes (35). However, because the addition of palmitate failed to complement the benomyl-sensitive phenotype of the tgl2⌬ null mutation (Fig. 6), it is implausible that a probable decrease in the endogenous palmitate level, as expected from reduced TAG degradation in tgl2⌬ cells, has made the deletion mutant hypersensitive to tubulin-targeting agents. In line with this, Tgl2p showed no lipolytic activity toward tripalmitin or dipalmitin in vitro ( Table 2).
In S. cerevisiae, all effective microtubule poisons cause depolymerization of microtubules and mitotic arrest, and mutants defective in the process of nuclear migration or chromosome segregation are sensitive to these agents (36). Nuclear positioning and microtubule morphology in tgl2⌬ cells are indistinguishable from those in wild-type cells, ruling out the involvement of Tgl2p in nuclear migration. 4 According to our preliminary data, Tgl2p is not directly involved in the process of chromosome segregation because tgl2⌬ cells are not blocked by benomyl or nocodazole at mitosis and segregate tester minichromosome normally.
Previously, it has been shown that S. cerevisiae cells with a mutation in the MDM2/OLE1 gene encoding the ⌬9 fatty acid desaturase display defective mitochondrial movement and aberrant mitochondrial distribution (37). Not surprisingly, a product of the desaturase, oleic acid, performs an essential role in mitochondrial movement and inheritance (37). It is almost tempting to propose that Tgl2 lipase serves as a specific link between mitochondria and microtubules to establish proper localization of mitochondria in the cell. Unfortunately, movement and distribution of mitochondria inside the cell and during cell division require interaction with actin cytoskeleton and not microtubule-based structure in budding yeast (38). No appreciable effects on mitochondrial function or morphology are detected in the tgl2⌬ mutant, also arguing against this role. 4 Alternatively, Tgl2 lipase may prevent death of cells arrested in mitosis by elevating the concentration of oleic acid in the mitochondria. We propose that lipolysis catalyzed by the mitochondrial Tgl2 lipase or its likely product oleic acid may become crucial for the survival of cells under stress-inducing conditions, including anti-microtubule drug treatment. Of course, Tgl2p-dependent lipolysis has no clear effects on budding yeast except for a modest increase in TAG content under normal conditions. TGL2-deficient yeast grows even better than the wild-type in the absence of stress. In support of our hypothesis, there are many reports suggesting a relationship between the distribution of fatty acids and some stresses. For instance, the composition of cellular fatty acids changes when S. cerevisiae cells are under acetaldehyde stress, showing an increase in C18:0 and C18:1 (oleic acid) levels and a decrease in C12:0, C14:0, C16:0, and C16:1 levels (39). In particular, the content of oleic acid is remarkably increased in these cells. Supplementation of oleic acid into the medium partially complements growth defects in acetaldehyde-sensitive mutants and ole1⌬ cells, whereas it barely affects the growth rate of wild-type cells under the same stress (39). Future work on Tgl2p could uncover yet undefined function of TAG lipases. Further studies on the yeast lipases will be valuable in unraveling the molecular basis of lipid metabolism in cell physiology and hopefully provide clues on the treatment of lipid-associated disorders.