Fld1p, a functional homologue of human seipin, regulates the size of lipid droplets in yeast

Lipid droplets (LDs) are intracellular lipid inclusions consisting primarily of triacylglycerols (TAGs) and sterol esters (SEs; Murphy and Vance, 1999). Recent studies have proven that LDs are dynamic intracellular organelles that are functionally connected with other cellular compartments (for review see Martin and Parton, 2006). LDs from different cell types are bounded by a monolayer of phospholipids and characterized by unique surface proteins, which may contribute to the biogenesis, maturation, and stability of LDs (Martin and Parton 2006; Wolins et al., 2006). LDs are believed to originate from microdomains of the ER, where enzymes for the synthesis of neutral lipids reside. Newly synthesized neutral lipids are thought to accumulate between two leaflets of the ER bilayer before budding into the cytosol (Martin and Parton 2006; Wolins et al., 2006). With the assistance of microtubules, nascent LDs appear to fuse with each other until a certain size is reached (Bostrom et al., 2005). A recent study identified SNARE proteins as part of the core machinery for the fusion of LDs (Bostrom et al., 2007). Despite recent progress, many fundamental questions remain. For instance, how is the fusion process regulated, and what factors determine the final size of a mature LD?

In an effort to identify novel gene products that may play a role in LD formation, Nile red, a vital dye specific for intracellular LDs, was used to visually screen the entire collection of viable single-gene deletion mutants of the budding yeast Saccharomyces cerevisiae for abnormalities in the number and morphology of LDs (Greenspan et al., 1985). Because the number and morphology of LDs may vary depending on growth phases, wild-type cells and all mutants were grown overnight to stationary phase (OD₆₀₀ = ∼5) in this study immediately followed by Nile red staining and fluorescence microcopy. Wild-type cells at stationary phase showed 5.16 ± 2.18 LDs per cell on average (±SD; n = 200), and ∼80% of the cells displayed three to seven LDs (Fig. 1 A, a). To simplify the screening process, we arbitrarily categorized deletion strains with the majority (>80%) of cells accumulating on average less than three LDs as fld (few LDs) mutants and strains accumulating more than seven LDs as mld (many LDs) mutants. Among the mld mutants, strains containing >11 LDs were classified as strong mld mutants. We isolated 17 fld mutants and 116 mld mutants (Tables S1 and S2).

We focused on FLD1, which corresponds to a previously uncharacterized ORF, YLR404W, because its deletion not only affected the number of LDs but also gave rise to strikingly enlarged or aggregated LDs (Fig. 1). When grown in rich medium until stationary phase, wild-type cells usually display three to seven LDs under the microscope. The LDs were between 0.2 and 0.4 μm in diameter and were almost spherical in shape (Fig. 1 A, a). In contrast, LDs observed in fld1Δ cells were very irregular in terms of quantity, shape, and size. Up to 30% of the total population of fld1Δ cells contained one or a few supersized LDs that were spherical in shape and were about 0.5–1.5 μm in diameter (Fig. 1 A, b; arrow), which means that the volume of the largest LD of the fld1Δ cells was about 50 times that of the largest LD found in wild-type cells. About 60% of the fld1Δ population contained an amorphous aggregation of neutral lipids in addition to several small LDs (Fig. 1 A, c and d; arrowheads). The remaining ∼10% of the fld1Δ cells contained scattered and weakly stained LDs, which had diameters of <0.1 μm (Fig. 1 A, e). The phenotypic characteristics of fld1Δ were also observed in log-phase cells. When grown to log phase (OD₆₀₀ = 0.8), most of the wild-type cells contained two or three LDs that were slightly smaller than those of stationary-phase cells (Fig. 1 A, a′). The LDs in fld1Δ at log phase show similar morphology to cells grown to stationary phase except that the supersized LDs and the aggregation of neutral lipids were smaller and more weakly stained (Fig. 1 A, b′ and c′; arrow and arrowhead, respectively).

To examine the ultrastructure of fld1Δ cells, we performed transmission EM (TEM) of the wild-type and fld1Δ strains. Cells were grown in rich medium until stationary phase and were subjected to TEM analysis. One typical cross section of a wild-type cell contained five LD profiles, which were round and about 0.2–0.4 μm in diameter (Fig. 1 B, a). The cross sections of fld1Δ mutant cells again showed three classes of LDs with distinct morphologies. Up to 30% of cells displayed one or a few supersized LDs per section, which were either round or oval (Fig. 1 B, b–d). Consistent with the result of fluorescence microscopy, the diameters of some LDs were up to 1.5 μm (Fig. 1 B, d). Aggregated LDs were found in ∼60% of the mutant population (Fig. 1 B, d–g). These aggregations were reminiscent of the amorphous neutral lipid clump observed under fluorescence microscopy (Fig. 1 A, c and d). The rest of the fld1Δ cells (∼10% of the total population) contained many tiny LDs, most of which had a diameter of <0.1 μm and were loosely scattered (Fig. 1 B, h and i).

Yeast cells undergo marked proliferation of LDs when they are grown in defined medium (synthetic complete [SC]) or in an oleate-based medium (YPO [YP plus oleate, no glucose]; Binns et al., 2006). As shown in Fig. 1 C (a–c), a great increase in the number of LDs was observed in wild-type cells grown in SC or YPO media. When fld1Δ cells were grown in YPD, the majority (∼70%) of the cells showed amorphous aggregations of many intermediate-sized LDs (Fig. 1 B, f and g). In contrast, >70% of the fld1Δ cells displayed only one or two supersized LDs when cultured in SC media (Fig. 1 C, e). Moreover, amorphous aggregation of many small LDs that were common in cells cultured in YPD media were only observed in about 10% of the cells grown in SC. Interestingly, when fld1Δ cells were cultured in YPO medium, >95% (191/200 cells examined) of the cells displayed amorphous aggregations of LDs without the supersized LDs (Fig. 1 C, f). However, when fld1Δ cells were cultured in YPDO medium (YPD + oleate), the large LDs appeared again together with the aggregation of smaller LDs (Fig. 1 C, g). Thus, we identified a commonly used growth medium (SC) that allowed us to easily distinguish wild-type and fld1Δ cells based on the presence or absence of supersized LDs (Fig. 1 C, compare b with e).

We measured the steady-state levels of TAG and SE as well as the rate of oleate incorporation into TAG and SE (Oelkers et al., 2002). For cells grown in YPD to log phase, the deletion of FLD1 caused about a doubling in the steady-state levels of both TAG and SE (Fig. S1 A). The rate of SE synthesis was also upregulated by 70% in fld1Δ deletion cells, but little difference in the rate of oleate incorporation into TAG was observed (Fig. S1 B). Similar patterns of changes were detected for cells grown in YPD to stationary phase or in SC to either log or stationary phase (unpublished data).

The expression of FLD1-GFP in fld1Δ cells restored the normal morphology of LDs (Fig. 2, A and B). We performed a series of subcellular fractionation experiments to analyze the cellular distribution of Fld1p. Cell extracts prepared from the fld1Δ strain expressing Fld1-GFP were fractionated by centrifugation at 13,000 g for 10 min, resulting in P13 pellet and S13 supernatant fractions that were probed with antibodies against GFP and Dpm1p, an ER marker. Both Fld1p and Dpm1p were found in the P13 fraction, which contains large membranous structures such as the vacuole, ER, and plasma membrane (Fig. 2 C). The same cell extracts were subjected to continuous sucrose density gradient analysis. 13 fractions were collected from top to bottom (1–13) and were probed for the presence of GFP and Dpm1p by immunoblotting. Dpm1p and Fld1p appeared to exist in the same density fractions (Fig. 2 D). Localization of Fld1-GFP was also examined in live cells by fluorescent microscopy, and Fld1-GFP was found in both perinuclear and peripheral ER (Fig. 2 E). Finally, immuno-EM was used to pinpoint the exact location of Fld1p. Fld1p-GFP was found to be associated with the cortical ER and the nuclear envelope, which is consistent with the putative ER localization observed by light microscopy. In addition, labeling was observed throughout the ER, including regions in contact with LDs (Fig. 2 F).

The existence of morphologically distinct LDs within fld1Δ suggests enhanced fusion activities of LDs: the small, discrete LDs may represent the newly synthesized LDs, which tend to aggregate before eventually fusing into a supersized LD. To test this hypothesis, wild-type and fld1Δ cells were cultured in SC medium until midlog phase (OD₆₀₀ = ∼1.0), stained with Nile red, and observed for the fusion of LDs by fluorescent microscopy. Cells in which two or several LDs lay close together were targeted. We examined 200 cases of adjacent LDs each for mutant and wild-type cells and monitored every case for 1 min. The criteria we used to define fusion were described previously (Bostrom et al., 2007). No fusion events were observed in wild-type cells, but fusion was detectable for about 10% (19/200) of all mutant cases. As shown in Fig. 3 A, two closely positioned LDs appeared to completely fuse within a span of 2 s. The size of the newly formed LD appeared to be the combined size of the two parent LDs.

To further demonstrate enhanced LD fusion in fld1Δ cells, we isolated LDs from wild-type and mutant cells both carrying Tgl3p-GFP. Tgl3p is a triglyceride lipase that localizes to the surface of LDs (Fig. 3 B; Athenstaedt and Daum, 2003). Purified LDs from both wild-type and fld1Δ cells were left in PBS buffer and examined by microscopy before and after 60 min. Whereas LDs from wild-type cells remained scattered and unchanged in size, LDs from fld1Δ cells formed aggregates or fused into huge lipid inclusions reminiscent of the supersized LDs observed in live fld1Δ cells (Fig. 3, C and D). The lipid inclusions observed in vitro are much larger than the supersized LDs in live cells (∼5 μm vs. ∼1 μm in diameter), suggesting that additional fusion events occurred in vitro.

Remote homology detection techniques were used to look for a mammalian homologue to Fld1p. Human seipin, a protein associated with Berardinelli-Seip congenital lipodystrophy (BSCL) and motorneuron disorders, was identified (Agarwal and Garg, 2004). Seipin shows weak sequence conservations to Fld1p, but they both contain two predicted transmembrane domains, and their predicted secondary structures are also very similar (Fig. 4 A). Human seipin is encoded by the BSCL2 gene, which gives rise to at least three different mRNAs and two peptides with 398 or 462 amino acids (Lundin et al., 2006). The region that covers the first 280 amino acids of human seipin is 88% identical among rat, mouse, chimpanzee, and human homologues (Agarwal and Garg, 2004). This is interesting because all of the sequence conservations between yeast Fld1p (285 amino acids) and seipin fall within this region. To test whether seipin is a functional homologue of Fld1p, full-length human and mouse seipin homologues were expressed in fld1Δ cells grown in SC media. The average diameter of the LDs is 1.27 ± 0.19 μm (n = 117) without human seipin and 0.43 ± 0.05 μm (n = 106) with seipin (Fig. 4 B). We also examined the effects of point mutations in seipin that are implicated in lipodystrophy (A212P) and motoneuron disorders (N88S and S90L). The expression of N88S and S90L but not A212P rescued the defects in LD morphology (Fig. 4 C). This is not surprising because A212P is considered a loss-of-function mutation, whereas N88S and S90L may represent gain-of-function mutations (Windpassinger et al., 2004). Lastly, expression of the highly conserved 280–amino acid region of seipin rescued the defects in LD morphology (Fig. 4 C). Together, these results strongly suggest that human seipin represents the functional homologue of Fld1p.

Congenital generalized lipodystrophy (CGL; or BSCL) is an autosomal recessive disorder that is characterized by the almost complete absence of adipose tissue and severe insulin resistance (Agarwal and Garg, 2004). Genome-wide linkage analysis identified two loci for CGL: CGL type 1 (CGL1) is caused by mutations in the AGPAT2 (1-acylglycerol-3-phosphate-O-acyl transferase 2) gene, and CGL2 is caused by mutations in BSCL2, which encodes seipin (Magre et al., 2001; Agarwal et al., 2002). AGPAT2 catalyzes the formation of phosphatidic acid (PA), but knocking down AGPAT2 led to elevated levels of several phospholipid species, including PA, and to a delay in the activation of key transcription factors for adipogenesis such as C/EBPβ and PPARγ (Gale et al., 2006). Therefore, AGPAT2 controls adipogenesis through modulation of the synthesis of phospholipids. In contrast, little is known about the role of seipin in adipogenesis and lipodystrophy. Because mutations in AGPAT2 and BSCL2 cause similar clinical manifestations, we wondered whether aberrant phospholipid metabolism may underlie the cellular defects for both conditions. Lipid species from wild-type and fld1Δ whole cell extracts were analyzed by electrospray ionization tandem mass spectrometry. The level of PA increased slightly in fld1Δ cells (unpublished data). Interestingly, there is a shift from long-chain (18:1) to medium/short-chain (16:0, 14:0, and 12:0) fatty acid incorporation into all major phospholipids as a result of the deletion of FLD1 (Fig. 5 and Fig. S3).

We identified seipin (Fld1p) as a novel regulator of the cellular dynamics of LDs. The deletion of FLD1 causes increased levels of neutral lipids, clustering of LDs, and formation of enlarged (supersized) LDs. Increased neutral lipids often lead to an increase in the number but not the size of LDs (Fig. 1, C, b and c; and Fig. S2). Therefore, the appearance of supersized LDs in fld1Δ cells is unlikely to be caused by an increase in neutral lipids. Rather, the physical property of the surface of LDs might have been altered in fld1Δ cells as a result of changes in phospholipids, which may facilitate the clustering and fusion of LDs. In support of this, LDs isolated from fld1Δ cells can aggregate and fuse without the supply of ATP and cytoplasmic proteins (Fig. 3 C).

What is the molecular function of Fld1p or seipin? We favor a role of Fld1p in phospholipid/fatty acid metabolism for the following reasons: (1) a shift from long to medium/short acyl chains was detected in fld1Δ cells (Fig. 5); (2) LDs isolated from fld1Δ cells can aggregate and fuse without the supply of ATP and cytoplasmic proteins, suggesting a role for phospholipids (Fig. 3); (3) the other gene that is associated with CGL is APGAT2, a major enzyme in phospholipid metabolism; and (4) lipin, another well-known protein that is associated with lipodystrophy in mice, is a phosphatidate phosphatase (Han et al., 2006). Deletion of SPO7 or NEM1, two genes involved in the activation of lipin, also caused aberrant LD morphology (Table S1 and our unpublished data).

Results described herein show for the first time that seipin and its homologues modulate the formation and especially fusion of the LDs. Our data also suggest that changes in phospholipid metabolism may be the unifying theme for both CGL1 and CGL2. Understanding the molecular function of seipin will provide mechanistic insights into LD formation and adipogenesis.

Both wild-type BY4741 (MATa his3Δ1 leu2Δ0 met15Δ0 ura3Δ0) and single deletion mutants, including the ylr404wΔ strain (MATa his3Δ1 leu2Δ0 met15Δ0 ura3Δ0 YLR404W∷KanMX4), were obtained from the European Saccharomyces cerevisiae Archives for Functional Analysis collection (Winzeler et al., 1999).

Nile red, oleic acid, and Brij58 were purchased from Sigma-Aldrich; glutaraldehyde and osmium tetroxide were obtained from Electron Microscopy Sciences. DAPI, anti-Dpm1 mouse monoclonal, and anti-GFP rabbit polyclonal were purchased from Invitrogen. [4-¹⁴C]cholesterol, Ficoll PM400, and [9, 10(n)-³H]oleic acid were purchased from GE Healthcare.

Unless otherwise stated, yeast cells were grown with rotary shaking at 30°C in liquid YPD medium (1% yeast extract, 2% bacto peptone, and 2% dextrose). Alternatively, cells were grown in SC medium (0.67% nitrogen base and 2% dextrose with all amino acids supplemented), YPO medium (1% yeast extract, 2% bacto peptone, and 0.1% oleic acid in 1% Brij58), or in YPDO medium (1% yeast extract, 2% bacto peptone, 2% dextrose, and 0.1% oleic acid in 1% Brij58). Plasmid-carrying strains were grown on synthetic medium with appropriate selection as described previously (Kaiser et al., 1994).

Fluorescence imaging was performed on a microscope (DMLB; Leica) with a fluorescent lamp (ebq 100; Curtis). Samples were viewed using a 100× NA 1.30 oil immersion objective lens (Leica). Images were taken with a digital camera (DFC480; Leica) and FW4000 software (Leica).

Nile red (Sigma-Aldrich) is a specific and excellent vital dye for intracellular LDs. A 450–490-nm bandpass excitation filter, a 510-nm dichromatic mirror, and a 515-nm longpass emission filter (filter cube I3; Leica) were chosen to view LDs.

To record the process of LD fusion in the fld1Δ mutant, 3 μl of late log-phase cells (OD₆₀₀ = ∼1) stained with Nile red were spotted on a slide and covered with a coverslip. Under the microscope, cells in which two or several LDs lay close together were targeted. Images were collected at 2-s intervals.

GFP signal was visualized with a 470/40-nm bandpass excitation filter, a 500-nm dichromatic mirror, and a 525/50-nm bandpass emission filter (filter cube GFP; Leica).

Cells fixed with 3.7% formaldehyde for 15 min were stained with 5 μg/ml DAPI. DAPI fluorescence was observed with a 340–380-nm bandpass excitation filter, a 400-nm dichromatic mirror, and a 425-nm longpass emission filter (filter cube A; Leica).

Cells were grown in rich medium until stationary phase, harvested, fixed with 2.5% glutaraldehyde, and postfixed with 2% (wt/vol) osmium tetroxide. The samples were subsequently dehydrated in a series of graded ethanol and embedded in Spurr's resin. 80-nm ultrathin sections were stained with uranyl acetate and lead citrate and examined under an electron microscope (JEM-1230; JEOL).

Immuno-EM of ultrathin cryosections was performed essentially as described previously (Martin et al., 2005). In brief, fld1Δ cells transformed with FLD1-GFP were grown in SC media to late log phase and fixed in 2% PFA/0.2% glutaraldehyde in 0.1 M PHEM buffer (60 mM Pipes, 25 mM Hepes, 2 mM MgCl₂, and 10 mM EGTA), pH 6.9, for 1 h at room temperature. Cells were embedded in 10% gelatin, cryoprotected using polyvinyl pyrrolidone–sucrose, and snap frozen onto specimen holders in liquid N₂. Ultracryomicrotomy was performed by a slight modification of the Tokuyasu technique, and sections were picked up with a 1:1 mixture of 2.3 M sucrose and 2% methyl cellulose. Grids were viewed using a transmission electron microscope (model 1010; JEOL).

A 1.4-kb fragment containing the native promoter and the coding sequence before the stop codon of YLR404W was amplified by PCR using a 5′ primer (5′-HindIII-TTACCATGCACGTTGTCG) and a 3′ primer (5′-BamHI-GCTATGTTTCTTGGATT). This fragment was then subcloned into the HindIII- and BamHI-cleaved YCplac111-GFP plasmid, in which the GFP coding sequence was inserted between the BamHI and EcoRI restriction sites. A sequence encoding yeast codon bias GFP (yEGFP; a gift from B. Winsor, Université Louis Pasteur, Strasbourg, France) was used for all GFP constructs.

A 2.7-kb fragment containing the natural promoter and the coding sequence before the stop codon of TGL3 was amplified by PCR using a 5′ primer (5′-ACGGC-HindIII-TCTGTT) and a 3′ primer (5′-BamHI-CCTACTCCGTCTTGCTCTT). This fragment was then subcloned into the HindIII- and BamHI-cleaved YCplac111-GFP plasmid, in which the GFP coding sequence was inserted between the BamHI and EcoRI restriction sites.

The coding sequences of human seipin and mouse seipin were amplified by PCR using a 5′ primer (5′-SphI- ATGGTCAACGACCCTCCAGTACCTGC) and a 3′ primer (5′-BamHI-TCAGGAACTAGAGCAGGTGGGGCGCTG) from human cDNA BC012140, a 5′ primer (5′-SphI-ATGTCTACAGAAAAGGTAGACCAAAAGG) and a 3′ primer (5′-BamHI-TCAGGAACTAGAGCAGGTGGGGCGCTG) from human cDNA AF052149, and a 5′ primer (5′-SphI-ATGATACATCAAAGAAGAGAAGCTGG) and a 3′ primer (5′-XbaI-TCAGGAACTGGAGCAGGTCGGGCGTTG) from mouse cDNA BC061689. The gene sequence encoding the first 280 amino acids of seipin was amplified using a 5′ primer (5′-SphI-ATGGTCAACGACCCTCCAGTACCTGC) and a 3′ primer (5′-BamHI-TCAGGAATTGTCTCTTTTTCGGATG) from human cDNA BC012140. The fragments were subcloned into SphI- and BamHI-cleaved or SphI- and XbaI-cleaved YCplac111-MET3 plasmid in which the MET3 promoter was inserted between HindIII and SphI restriction sites.

Site-directed mutagenesis was performed using the QuikChange II XL kit (Stratagene) according to the manufacturer's instructions.

Lipid extraction from lyophilized yeast cells was performed as previously described (Zhang et al., 2003). In brief, cells were grown in the appropriate medium until the required growth phase (determined by OD₆₀₀), harvested, washed twice with 0.5% Nonidet P-40 and once with deionized H₂O, and lyophilized. The dried cell pellets were resuspended in 50 μl lyticase (Sigma-Aldrich) solution (1,700 U/ml in 10% glycerol) and incubated at 37°C for 15 min at −70°C for 1 h and at 37°C for 15 min. Lipids were extracted with hexane and blown dry. Quantitation of neutral lipids was performed as described previously by Zweytick et al. (2000) with modifications. Samples were dissolved in 100 μl chloroform/methanol (2:1; vol/vol) and applied to Silica gel 60 F₂₅₄ plates (Merck), and chromatograms were developed in hexane/diethyl ether/acetic acid (85:15:1) with triolein and cholesteryl ester as the standard. For quantitation of SE and TAG, plates were dipped into methanolic MnCl₂ solution (0.63 g MnCl₂-4H₂O, 60 ml water, 60 ml methanol, and 4 ml of concentrated sulfuric acid), dried, and heated at 120°C for 15 min. Densitometric scanning was performed at 500 nm with a scanner (thin-layer chromatography; CAMAG). For each assay, at least three independent tests were performed; average and SD were calculated.

Incorporation of [³H]oleic acid into SE and TAG was performed as described previously (Zhang et al., 2003). 5 ml of cells at OD₆₀₀ = ∼0.6–0.8 were pulsed with 5 μCi [³H]oleic acid at 30°C for 30 min with shaking. Lipid extraction and thin-layer chromatography were performed as mentioned in the previous paragraph except that the plates were stained with iodine vapor. Incorporation of label into neutral lipids was determined after scintillation counting (Beckman Coulter) and normalization to a [¹⁴C]cholesterol internal standard and cell dry weight. For each assay, at least three independent tests were performed; average and SD were calculated.

A high performance liquid chromatography system (Agilent) coupled with a triple quadrupole/ion trap mass spectrometer (4000Qtrap; Applied Biosystems) was used for quantitation of individual lipids. Samples were introduced into the mass spectrometer by loop injections with chloroform/methanol/300 mM piperidine (1:1:0.1) as a mobile phase at a flow of 200 μl/min (Shui et al., 2007). Based on product ion and precursor ion analysis of head groups and fatty aclys, two comprehensive sets of multiple reaction monitoring (MRM) transitions were set up for quantitative analysis of various lipids (Guan et al., 2006), and results are expressed as normalized intensities.

Fatty acyl profiles of polar lipids were acquired using a micromass spectrometer (Micromass Q-TOF; Waters Corp.) in the negative electrospray ionization ion mode. Piperidine (final concentration of 15 μM) was added into lipid extracts to enhance ionization (Leber et al., 1994). Samples were infused into the mass spectrometer at a flow rate of 15 μl/min. High collision energy (55 V) was applied to fragment polar lipid parent ions to generate fatty acyl profiles. The operation parameters for the mass spectrometer are a sample cone temperature of 250°C, a sample cone voltage of 70 V, and collision energy of 55 V. Mass spectra were recorded from mass/charge ratios of 140–900.

Precursor ion or neutral loss scans of phospholipid headgroup fragments (Han and Gross, 2005) were used to obtain information on yeast phospholipid compositions. Based on this information, a comprehensive list of MRM transitions was set up to follow fatty acyl compositions of these lipids (parent→fatty acyl fragment transitions). The signal intensity of each MRM value was normalized using the following equation:

Cells were harvested by centrifugation at 3,000 g for 5 min, washed once with double-distilled H₂O, suspended to 0.5 g of wet weight/ml in 0.1 M Tris-Cl, pH 9.4, and 10 mM DTT, and incubated at 30°C for 10 min. Afterward, they were washed once with 1.2 M sorbitol and suspended in 1.2 M sorbitol and 20 mM K₃PO₄, pH 7.4, to 0.15 g of wet weight/ml. Zymose 20T was added to 2 mg/g of wet weight cells. The suspension was incubated at 30°C with gentle shaking for 20–60 min. Spheroplasts were harvested by centrifugation for 5 min at 1,500 g, washed twice with 1.2 M sorbitol, and suspended in the appropriate buffer for homogenization.

Spheroplasts were lysed in ice-cold lysis buffer II (50 mM Tris-HCl, pH 7.5, 1 mM EDTA, and 0.2 M sorbitol) containing protease inhibitors. Cell lysates were cleared at 500 g for 5 min, resulting in supernatant S5, which was then subjected to centrifugation at 13,000 g for 15 min, resulting in supernatant S13 and pellet P13. A continuous sucrose density gradient fractionation procedure was used as described previously (Kölling and Hollenberg, 1994) except that only 13 fractions were taken from top to bottom. In brief, cells were lysed in STED10 (10% wt/wt sucrose, 10 mM Tris, pH 7.4, 1 mM EDTA, and 1 mM DTT) using a glass bead beater and were spun at 500 g for 5 min to remove cell debris. 950 μl of cleared cell extract was loaded onto a sucrose density prepared as follows: in 13-ml centrifuge tubes (SW41; Beckman Coulter), 3.8 ml STED53 (53% sucrose), STED35 (35% sucrose), and STED20 (20% sucrose) was loaded on top of each other. Centrifugation was performed at 30,000 rpm for 13–17 h for membranes to reach their equilibrium density. 950-μl fractions were collected from the top of the gradient. Dpm1p was used as an ER marker.

LDs were purified as described by Leber et al. (1994) with minor modifications. In brief, 0.15 g/ml spheroplasts were suspended in breaking buffer (10 mM Tris-Cl, pH 6.9, 0.2 mM EDTA, and 12% [wt/wt] Ficoll 400) and chilled on ice. Protease inhibitors were added in buffers as follows: 1 mM PMSF, 2 μg/ml aprotinin, 0.5 μg/ml leupeptine, and 1 μg/ml pepstatin A. Spheroplasts were homogenized with a Dounce homogenizer applying 10–20 strokes using a loose-fitting pestle, and the homogenate was transferred into 13.5-ml centrifuge tubes (Ultra-Clear; Beckman Coulter) and overlaid with an equal volume of breaking buffer. Centrifugation was performed for 60 min at 28,000 rpm in an SW41 swing bucket rotor (Beckman Coulter). A floating layer that consists mainly of LDs and vacuoles was collected and suspended gently in breaking buffer using a Dounce homogenizer with a loose-fitting pestle. The suspension was again transferred into 13.5-ml tubes, overlaid with an equal volume of 10 mM Tris-Cl, pH 7.4, 0.2 mM EDTA, and 8% (wt/wt) Ficoll 400, and centrifuged for 60 min at 28,000 rpm in an SW41 swing bucket rotor. The floating layer was recovered, suspended gently in 10 mM Tris-Cl, 0.2 mM EDTA, 0.6 M sorbitol, and 8% (wt/wt) Ficoll 400 in 13.5-ml tubes, and overlaid with 10 mM Tris-Cl, 0.2 mM EDTA, and 0.25 M sorbitol. After centrifugation at 28,000 rpm for 30 min, the floating layer consists of highly enriched LDs that are separated from vacuoles.

Tables S1 and S2 summarize the genes identified in the genome-wide screen for fld strains and mld strains, respectively. Fig. S1 presents neutral lipid analysis of wild-type and ylr404wΔ (fld1Δ) strains. Fig. S2 shows that mutants (mld) with increased levels of neutral lipids do not contain enlarged LDs. Fig. S3 is an analysis of phosphatidylethanolamine and phosphatidylcholine, which is continued from Fig. 5.

We thank Prof. Sepp Kohlwein for helpful discussions.

This work is jointly supported by research grants from the Biomedical Research Council of Singapore and the University of New South Wales to H. Yang as well as grants from the National Health and Medical Reasearch Council of Australia to R.G. Parton. The Institute for Molecular Bioscience is a special research center of the Australian Research Council. M.R. Wenk was supported by the Office of Life Sciences and the Academic Research Fund of the National University of Singapore and by the Biomedical Research Council of Singapore. L. Kuerschner is supported by a postdoctoral fellowship from the Deutsche Forschungsgemeinschaft.