The yeast lipin orthologue Pah1p is important for biogenesis of lipid droplets

Decrease in droplets but not neutral lipids in pah1Δ cells

We previously identified 59 yeast gene disruption strains with altered lipid droplet morphology based on staining with the neutral lipid dye BODIPY 493/503 (Szymanski et al., 2007). In two of these strains, nem1Δ and spo7Δ, BODIPY-stained rings were apparent in a minor fraction (∼30%) of these cells (Fig. 1 B). These rings resembled the ER, and indeed, they colocalized with an ER marker, CFP-HDEL (pseudocolored red in the merge images for better visualization; Fig. 1 B).

The corresponding proteins Nem1p and Spo7p form a complex in the ER (Siniossoglou et al., 1998) and recruit yeast lipin, Pah1p (Santos-Rosa et al., 2005), to convert phosphatidate (PA) to DAG. To determine whether the ring phenotype was related to activation of Pah1p, we generated the deletion strain, pah1Δ, which was missing in the library we originally screened. In this strain, BODIPY-staining rings were apparent in every cell and again colocalized with the ER marker. Strikingly, there were only 37% of lipid droplets in this strain compared with wild type (Fig. 1, B and C). Droplets were more heterogeneous in size in pah1Δ than in wild type.

Cell TAG is known to decrease in pah1Δ because PAH1 encodes the main PA phosphatase that normally provides its precursor, DAG (Han et al., 2006). To determine whether the decrease in lipid droplets in pah1Δ can be attributed to an overall decrease in neutral lipids, we measured both TAG and StE in log-phase cells, where the decrease in TAG is not as severe as in stationary phase (Han et al., 2006). (All fluorescence experiments were also performed in log phase.) As expected, there was a decrease in TAG in pah1Δ cells of comparable magnitude to the decrease of lipid droplets, but there was a compensatory increase in StE so that the sum of these two neutral lipid classes, which normally are packaged into droplets, was at least as high in the mutant (Fig. 1 D). Therefore, the decrease in lipid droplets cannot be attributed to a decrease in overall neutral lipid.

Neutral lipid accumulation in the ER in pah1Δ

The decrease in lipid droplet number without a concomitant decrease in total neutral lipid suggests that there is a defect in droplet biogenesis in pah1Δ. If this were the case, neutral lipids should accumulate in the ER or other membranes in this strain. We first considered that the BODIPY-stained ER rings in the strain reflected neutral lipid accumulation. However, the ER proliferates in pah1Δ as a consequence of derepression of phospholipid biosynthesis, resulting in nuclear envelope deformation and the appearance of ER stacks (Santos-Rosa et al., 2005). To determine whether the BODIPY-stained rings in pah1Δ might be a result of altered ER morphology rather than neutral lipid accumulation, we stained cells that were disrupted in both PAH1 and the four acyltransferases that generate TAG and StE (Oelkers et al., 2002; Petschnigg et al., 2009). BODIPY weakly stained the cytosol in the acyltransferase-deficient “quad” mutant (Fig. 2 A, 4KO). In contrast, membrane rings were indeed observed when pah1Δ was additionally knocked out, generating a “quint” mutant (Fig. 2 A, 5KO). (The intensities of staining among strains can be directly compared because the absolute intensity range for all images of an experiment was fixed in microscope software.) We conclude that BODIPY 493/503 can stain membrane stacks even devoid of neutral lipids as well as it stains lipid droplets in wild type; thus, it cannot be used to detect membrane neutral lipids in pah1Δ.

However, both Nile red and Oil red O better differentiated neutral lipids from membrane stacks. Nile red intensely stained both membranes and droplets in pah1Δ; however, the dye also stained membranes in some cells of the quint mutant, regardless of the emission wavelength used (Fig. 2 A and not depicted; Wolinski and Kohlwein, 2008). In contrast, Oil red O, which is traditionally used to label neutral lipid in plants and animals (Ramírez-Zacarías et al., 1992), yielded strong patchy staining in pah1Δ that was quite different from the punctate appearance of droplets in wild type. In contrast, Oil red O did not stain membranes of the quint mutant at all. Although this result suggested neutral lipid staining in membranes in the pah1Δ strain, we could not visualize ER labeled with CFP-HDEL in Oil red O–stained cells to assess colocalization because isopropanol, which destroyed ER morphology, was required for Oil red O staining.

To gain better evidence for ER accumulation of neutral lipids in pah1Δ, we cultured the strains in medium with oleic acid to stimulate neutral lipid synthesis and subjected them to electron microscopy. As expected, wild-type cells contained large lipid droplets, which were absent in the quad mutant (Fig. 2 B). The proliferated ER was clearly seen in pah1Δ. Significantly, the membrane stacks contained many electron-transparent inclusions consistent with neutral lipid in this compartment. Sometimes, larger inclusions were seen within the proliferated ER, as if tiny lipid droplets had coalesced (Fig. S1). The bilayer structure could not be clearly visualized adjacent to the inclusions, and it was impossible to verify that the inclusions were between leaflets of the bilayer. Nevertheless, these membrane inclusions were entirely absent in the proliferated ER of the quint strain.

To biochemically verify membrane accumulation of neutral lipids in pah1Δ, we separated membranes and lipid droplets by ultracentrifugation from postnuclear supernatants of wild-type and pah1Δ cells cultured in oleic acid. In wild-type cells, 10.2% of TAG was found in the membrane fraction (Fig. 2 C). In contrast, 22.0% of TAG was found in membranes in pah1Δ cells, although the absolute amount of TAG in membranes was actually decreased 28% in the mutant. Thus, the membrane inclusions in pah1Δ were likely not predominantly TAG. In contrast, there was 5.2 times more membrane-associated StE in pah1Δ compared with wild type. Based on the large decrease in lipid droplet number with little change in total neutral lipids, membrane staining with Nile red and Oil red O, transparent inclusions seen by electron microscopy, and accumulation of StEs in pah1Δ membranes, we conclude that Pah1p is somehow important for assembly of droplets.

It has previously been shown that PA phosphatase activity of Pah1p depends on the intact N-terminal lipin homology domain and haloacid dehalogenase-like domain of the protein (Fig. 3 A; Han et al., 2006, 2007). To determine whether the low droplet phenotype of pah1Δ is caused by the absence of catalytic activity by itself, three point mutants were generated that are known to disrupt activity, and the pah1Δ strain was transformed with plasmids encoding intact or mutated Pah1p. Although the wild-type copy of Pah1p was able to revert both the ring (Fig. 3 B) and droplet phenotype (Fig. 3 C) to wild type, none of the three point mutants were able to complement the strain. Similarly, only the wild-type allele was able to reverse the accumulation of StE in isolated membrane fractions (Fig. 2 C). We conclude that the PA phosphatase activity of Pah1p is essential for its effect on droplet biogenesis.

PAH1 and StEs in combination are essential for droplet formation

Although total neutral lipid is not decreased in pah1Δ, we next asked whether Pah1p was particularly important for packaging TAG versus StE. To address this, we knocked out PAH1 in cells that either lacked the genes for the two DAG acyltransferases (DGA1 and LRO1) or for the two steryl acyltransferases (ARE1 and ARE2; Yang et al., 1996; Kohlwein, 2010). As reported previously (Garbarino et al., 2009), droplets are formed in the absence of either TAG or StE (Fig. 4 A), although numbers are reduced, indicating that neither TAG nor StE formation is essential for droplet formation. There was a small (26%) but highly reproducible decrease in droplet number when PAH1 was disrupted in dga1Δlro1Δ cells, even though there was no decrease in total StE in this strain (Fig. 4, B and C; also see Fig. S2 and Tables S2 and S3 for distribution of acyl chains). If expressed in terms of cell volume (disruption of PAH1 results in a calculated 96% increase in volume; compare cell sizes in Fig. 4 A), the decrease in droplet number is 51%. The effect of disrupting PAH1 in the are1Δare2Δ background is much more extreme: no droplets at all were detected in pah1Δare1Δare2Δ, even though the cells contained about as much TAG as wild type. In cells cultured in medium containing oleate, the large fatty acid load resulted in some droplet formation in the pah1Δare1Δare2Δ strain, but droplets were far fewer in number even compared with pah1Δ (Fig. 4, D and E). Perhaps the presence of droplets in this strain growing in oleic acid medium is caused by reaching saturation for storing neutral lipids in the ER bilayer. We conclude that in the absence of StE, Pah1p is required to form droplets containing TAG in the absence of exogenous fatty acids, although it also facilitates the packaging of StE in the absence of TAG.

Evidence that DAG is important for droplet formation

Thus far, our data suggest that the DAG generated by Pah1p is important for droplet formation independent of its role as a precursor of TAG. This is consistent with a recent study showing that manipulation of DAG levels in mammalian cells can drastically affect droplet formation (Skinner et al., 2009). If DAG is required for the formation of droplets, overproduction of DAG might be able to bypass the role of Pah1p. To test this idea, we forced more DAG formation by eliminating the backward reaction to PA in a DAG kinase knockout strain, dgk1Δ. DGK1 has recently been shown to be important for providing PA for phospholipid synthesis during the lag phase (Fakas et al., 2011). We confirmed an increase in DAG and decrease in PA in this strain (Fig. 5 C), as has been previously reported (Han et al., 2008; Fakas et al., 2011). As predicted, the dgk1Δ strain alone had significantly more lipid droplets than wild type (Fig. 5, A and B), which were often in a “pearls on a string” (defined as four or more lipid droplets in a row spaced ≤0.5 µm apart) on the perinuclear ER (Fig. 5 A, arrows). The increase in droplets completely bypassed the role of PAH1 as shown in the pah1Δdgk1Δ strain. (In the absence of PAH1, DAG can be generated by another Mg²⁺-dependent PA phosphatase activity whose identity has not yet been established [Han et al., 2006].) Even if TAG synthesis was completely blocked, the knockout of DGK1 resulted in more droplets (Fig. 5 B, compare droplet number between dga1Δlro1Δ and dgk1Δdga1Δlro1Δ) and an enhancement of the pearls-on-a-string phenotype (Fig. 5 D). These data again demonstrate that the production of DAG itself is probably important for droplet formation in yeast.

The Pah1p phosphatase Nem1p is close to droplets

DAG should be produced at the site of droplet formation for it to be active because otherwise it would rapidly diffuse laterally in the membrane. Pah1p is dephosphorylated on the ER by the membranous Nem1p–Spo7p complex; Nem1p is the catalytic subunit of the activating phosphatase (Santos-Rosa et al., 2005), which results in membrane localization of Pah1p, where it is active (Karanasios et al., 2010; Choi et al., 2011). Both Spo7p-GFP and GFP-Nem1p localize to the ER when exogenously expressed (Siniossoglou et al., 1998), whereas GFP knocked in behind the NEM1 reading frame revealed puncta of uncertain origin (Huh et al., 2003). To more closely examine the localization of Nem1p, we tagged the chromosomal copy of NEM1, keeping the endogenous promoter intact, with mCherry. We confirmed that Nem1p forms puncta, and they colocalized with the ER marker CFP-HDEL (Fig. 6 A, bottom left inset). (As expected, GFP-tagged Spo7p colocalized in puncta with dtTomato-tagged Nem1p [Fig. S3].) Interestingly, only one Nem1p-mCherry punctum was usually found per cell (viewing through z sections); only 3% of cells had two puncta, and none had more. Most of the puncta were in close proximity to BODIPY-stained droplets (Fig. 6 A). By scoring the proximity of Nem1p-mCherry to droplets in three dimensions, 77% of them appeared to be in contact with droplets, whereas only 13% could clearly be resolved from droplets (Fig. 6 B). The proximity to droplets was clearly visualized in a computer-aided projection of a field of cells (Fig. 6 A, right; and Video 1). 10 out of 12 Nem1p-mCherry puncta within the field were within 0.65 µm of a droplet, the limit of resolution in this method.

To observe the colocalization of Nem1p with droplets in another way, we subjected membranes derived from a postnuclear supernatant of cells expressing Nem1p-mCherry in the genome to fluorescence microscopy. When grown on glucose medium, droplets were small, and some remained attached to the ER even after centrifugation; these can be visualized as puncta with BODIPY in this preparation (Fig. 7). Most of these spots also coincided with mCherry staining. In contrast, when membranes containing the CFP-HDEL marker were analyzed, there was no colocalization with lipid droplets. The proximity of Nem1p to droplets suggests that Pah1p docks onto the Nem1p–Spo7p complex, where it provides DAG (and later TAG) during droplet production. In yeast, we have observed rapid lateral movement of droplets on the surface of the ER, but never dissociation (Video 2). Perhaps the ∼20% of Nem1p puncta that are clearly separated from droplets are cycling to a new droplet after one is fully formed. In summary, our data strongly suggest that DAG is produced by Pah1p at the site of droplet formation, where it plays an important role in biogenesis of cytoplasmic lipid droplets.

Reagents

Restriction enzymes, DNA ligases, and other reagents for DNA manipulation were purchased from New England Biolabs, Inc., Roche, and Bio-Rad Laboratories. Neutral lipid standards were obtained from Nu-Chek Prep, Inc. Zymolyase 100T was purchased from US Biologicals, and hygromycin was bought from A.G. Scientific. All other chemical reagents were obtained from Sigma-Aldrich or Thermo Fisher Scientific.

Yeast strains and growth conditions

All yeast strains and plasmids used in this study are listed in Table S1. Strains SCY1703, SCY1998, and SCY328 were gifts from S. Sturley (Columbia University, New York, NY). To generate strain BY4742-KS001, we used the cassette within pGH317 provided by G.S. Han and G. Carman (Rutgers University, New Brunswick, NJ), and we followed their procedure to generate a PAH1 knockout from the cassette (Han et al., 2006) in BY4742. In brief, the cassette consisted of the URA3 gene flanked by ∼400 bp of upstream and downstream PAH1 sequence.

Cells were cultured at 30°C in 2% glucose or in oleic acid (Fig. 2, B and C; and Fig. 4, D and E) as previously described (Hettema et al., 2000). All cultures were harvested during log-phase growth.

DNA manipulations

All plasmids used in this study are listed in Table S1. Standard recombinant DNA techniques were used. Bacterial strain DH10β (Invitrogen) was used for all bacterial transformations. Site-directed mutations to plasmids were performed using the QuikChange site-directed mutagenesis kit (Agilent Technologies). PAH1 knockout in OAS001, OAS002, and OAS003 strains was performed by homologous recombination using a DNA cassette containing ∼500 bp of PAH1 sequence (from the 5′ and 3′ untranslated region flanking the PAH1 ORF) ligated to opposite ends of a TRP1 sequence. The BY4742-SL001 strain was generated by homologous recombination with the use of a DNA cassette containing mCherry followed by the PGK1 3′ terminator and URA3 to the 3′ end of the NEM1 ORF. BY4741-Spo7p-GFP-Nem1p-tdTomato was generated similarly from BY4741-Spo7p-GFP except that the cassette contained tdTomato instead of mCherry and the hygromycin-resistance marker (a gift from B. Tu, University of Texas Southwestern, Dallas, TX) instead of URA3; selection of transformants was performed on rich plates containing 300 µg/ml hygromycin. Yeast strains were transformed by the lithium acetate method (Ito et al., 1983). DNA manipulations resulting in changes to coding sequence as well as changes to genome were verified by sequence performed by the McDermott Center for Human Growth and Development.

Fluorescence and electron microscopy

All microscopy was performed at room temperature. For the acquisition of static images, strains were stained with BODIPY 493/503 (Invitrogen; Szymanski et al., 2007) to visualize lipid droplets and transformed (Figs. 1 B, 3 B, and 7) with pRS315-PGK-CFP-HDEL to visualize the ER. Cells in log phase were removed from growth medium, concentrated by centrifugation (3,000 g for 5 min at room temperature), and suspended on microscope slides in growth medium containing 1% agar (Kron, 2002). All images were acquired in Slidebook (version 4.1.0.3; Intelligent Imaging Innovations) using a microscope (Axiovert 200M; Carl Zeiss, Inc.) with a 100× 1.3 NA oil immersion objective lens (plan-Neofluar) equipped with a digital camera (Sensican; Cooke). BODIPY images were acquired using the fluorescein isothiocyanate filter set, CFP with the CFP filter set, and mCherry with the CY3 filter set (Chroma Technology Corp.). Z stacks were obtained at 0.3-µm spacing, and Slidebook commands of deconvolution (nearest neighbor option) were used to build images through cells. For Fig. 2 A and Fig. 4 (A and D), the range of pixel intensities used for the control strain to clearly visualize droplets was fixed and used for the other strains in the series so the images could be directly compared.

To observe movement of lipid droplets on the ER membrane, BY4742-DB001 expressing the ER marker GFP-HDEL was grown in glucose to log phase and subjected to spinning disk confocal microscopy using a spinning disk confocal system (Ultraview ERS; PerkinElmer) set up on a microscope (Axiovert 200M) with a camera (Orca; Hamamatsu Photonics) in the University of Texas Southwestern Live Imaging Core. Lipid droplets and ER were followed for 4.5 min by saving images over 32 planes (0.2 µm between planes), cycling every 5 s. To generate the QuickTime videos, data were collected over five continuous planes (0.2 µm between them) every 4–5 s and projected into one image per time point.

To visualize Nile red–stained cells, samples were processed and observed as previously published (Wolinski and Kohlwein, 2008). In brief, cells were fixed in 2% vol/vol formaldehyde, incubated for 1 min at room temperature, and washed twice in 50 mM Tris buffer, pH 7.5, before observation. Oil red O–stained cells were processed and observed as previously described (Binns et al., 2006). In brief, cells were washed with H₂O and resuspended in freshly filtered Oil red O solution (three parts of 1% stock in isopropanol and two parts H₂O). After 10 min of stain, cells were washed twice and observed. Three-dimensional reconstruction (Fig. 6 A, right) was performed using Imaris software (Bitplane AG).

For electron microscopy, yeast cells were grown in 0.1% oleate medium and processed in the University of Texas Southwestern Molecular and Cellular Imaging Facility as previously described (Wright, 2000; Binns et al., 2006). In brief, cells were fixed in potassium permangenate, dehydrated, and postfixed in uranylacetate before sectioning.

Lipid analysis

Lipid extraction and one-dimensional TLC analysis were performed as previously described (Binns et al., 2006) with few modifications. Yeast cell pellets were treated with Zymolyase 100T at 4 mg/1,000 OD₆₀₀ for 40 min at 30°C. Spheroplasts were weighed and subjected to hot isopropanol and chloroform extraction. Thin layer plates were charred with 3% cupric acetate and 8% phosphoric acid solution, and lipid bands were digitally scanned and analyzed using ImageJ (v.1.41; National Institutes of Health).

Samples subjected to mass spectrometric analysis were spiked with an internal standard mixture composed of 2 µg tri-TAG (15:0), 4 µg tri-TAG (21:0), 2 µg cholesteryl ester (13:0), and 4 µg cholesteryl ester (19:0). Samples were suspended in 0.4 ml chloroform and fractionated by silica gel column chromatography (6 ml Supelco Discovery DSC-Si and 500 mg of solid-phase extraction cartridges). Columns were preconditioned with hexane, and lipids were separated by sequential elution with solvents (hexane/diethyl ether [4:1, vol/vol], hexane/diethyl ether [1:1, vol/vol], methanol, and finally chloroform; Laffargue et al., 2007). Solvents were evaporated off with N₂ before suspending neutral lipid in 0.1 ml chloroform. Neutral lipid fractions were diluted into chloroform-methanol [1:1, vol/vol] plus 10 mM aqueous ammonium acetate. Samples were introduced by direct infusion at a flow rate of 750 µl/h into the electrospray source of liquid chromatography tandem mass spectrometry (Waters Micromass Quattro Ultima Triple Quadrupole). TAGs and StEs were quantified in full-scan mode. Neutral loss scans for TAGs and StE were used separately to verify acyl chain compositions and head group. The collision energies, with argon in the collision cell, were 20 V for TAGs and StE. Full and neutral loss/precursor scans were acquired in continuum mode and averaged over 60 and 30 s, respectively. Conditions for acquisition were all under a positive capillary voltage of 3.50 kV, cone gas of 80 ml/h, desolvation gas of 200 ml/h, 80°C source temperature, and 200°C desolvation temperature. The background of each spectrum was subtracted using a solvent blank, the peak areas integrated, isotopic overlap and response correction factors applied, and peaks quantified against their respective internal standards using in-house Visual Basic macros developed in Microsoft Excel.

Cell lysis and organelle fractionation

Yeast microsomes were prepared as described previously (Brodsky et al., 1993) with few modifications. The postnuclear supernatant in yeast lysis buffer (20 mM Hepes, pH 7.4, 200 mM sorbitol, 50 mM KOAc, 2 mM EDTA, and protease inhibitor cocktail: 0.2 mM 4-[2-aminoethyl]benzenesulfonyl fluoride hydrochloride, 1.2 µg/ml pepstatin A, 1.2 µg/ml leupeptin, 12 µg/ml p-toluenesulfonyl-l-arginine methyl ester, 12 µg/ml tosyl-lysyl-chloromethylketone, 12 µg/ml p-nitroenzoyl-l-arginine methyl ester, and 12 µg/ml soybean trypsin inhibitor) was loaded into a tube (SW60; Beckman Coulter), layered with a 1-ml Hepes cushion (20 mM Hepes, pH 7.4, 100 mM KCl, and 2 mM MgCl₂), and centrifuged at 38,000 rpm for 1 h at 4°C in a rotor (SW60). Floating lipid droplet and membrane pellets were collected and taken through the lipid analysis procedure as described in the previous section. For Fig. 7, membrane pellets of the required yeast strain were resuspended gently in 0.5 ml lysis buffer and stained for 10 min with BODIPY 493/503 before observation by fluorescence microscopy.

Online supplemental material

The online supplemental material contains a table of strains and plasmids used in this study (Table S1), mass spectroscopic analysis of TAG chain length and StE sterol subtypes in various strains (Tables S2 and S3 and Fig. S2), electron micrographs showing neutral lipid inclusions in pah1Δ (Fig. S1), localization of tagged Pah1p and colocalization of tagged Nem1p and Spo7p (Fig. S3), animation of the proximity of Nem1p with lipid droplets (Video 1), and movement of droplets on ER membranes (Video 2). Online supplemental material is available at http://www.jcb.org/cgi/content/full/jcb.201010111/DC1.