Proteomic Study and Marker Protein Identification of Caenorhabditis elegans Lipid Droplets*

Lipid droplets (LDs) are a neutral lipid storage organelle that is conserved across almost all species. Many metabolic syndromes are directly linked to the over-storage of neutral lipids in LDs. The study of LDs in Caenorhabditis elegans (C. elegans) has been difficult because of the lack of specific LD marker proteins. Here we report the purification and proteomic analysis of C. elegans lipid droplets for the first time. We identified 306 proteins, 63% of these proteins were previously known to be LD-proteins, suggesting a similarity between mammalian and C. elegans LDs. Using morphological and biochemical analyses, we show that short-chain dehydrogenase, DHS-3 is almost exclusively localized on C. elegans LDs, indicating that it can be used as a LD marker protein in C. elegans. These results will facilitate further mechanistic studies of LDs in this powerful genetic system, C. elegans.

mechanisms of regulation. Identifying C. elegans LD-proteins by biochemical purification and proteomic studies is necessary to address this question. C. elegans LDs, which are free of LRO contamination, have previously been isolated using a density centrifugation approach (22), however, the quantity and purity of the isolated LDs were not tested for further proteomic studies. Therefore, it is critical for lipid metabolic study using C. elegans to purify high quality and quantity LDs, analyze their proteome, and identify their marker proteins.
In order to identify C. elegans LD-proteins, especially some specific marker proteins, we developed a purification method based on our previous work (12,17,22) to purify LDs from this organism. Proteomic analysis identified 306 proteins, including 193 proteins that were previously found to be present in the LD proteomes of other organisms. Through both in vivo imaging and in vitro biochemical approaches, we found that DHS-3 is localized and highly enriched on the surface of LDs, indicating that it can be used as a specific marker protein for LDs in C. elegans.

EXPERIMENTAL PROCEDURES
Materials-A Colloidal Blue stain kit was purchased from Invitrogen (Carlsbad, CA). PDH-E1␣ monoclonal antibody was from Mito-Sciences (Eugene, OR). RME1 and Dyn1 monoclonal antibodies were purchased from the Developmental Studies Hybridoma Bank (Japan). Na-K ϩ ATPase was from Upstate Biotechnology (Charlottesville, VA). Bip monoclonal antibody and caveolin polyclonal antibody were from BD Biosciences. Actin monoclonal antibody and Rab7 was from Santa Cruz Biotechnology (Santa Cruz, CA). Rab18 was from Calbiochem (San Diego, CA). Gelatin and tannic acid were from Sigma-Aldrich (Missouri, USA). Formvar was from BDH Chemicals. Ltd. (Poole, UK). Uranyl acetate, 25% glutaraldehyde solution (EM grade), and a low viscosity embedding media Spurr's Kit were all from Electron Microscopy Sciences (Hatfield, PA). Osmium tetraoxide (EM grade) was purchased from Nakalai Tesque Co. (Kyoto, Japan). Phosphotungstic acid was from Zhongjingkeyi Technology Co., Ltd. (Beijing, China).
Strains and Culture Conditions-The N2 Bristol strain and VS20 (pATGL-1::ATGL-1::GFP) were used in this study. Animal culture was conducted essentially the same as previously described (22). Briefly, E. coli strain OP50 was cultured in LB medium and seeded onto 9-cm and 15-cm Nematode Growth Medium plates. Synchronized L1 stage animals were then seeded onto the plates. Animal density was ϳ50 per cm 2 . Animals were harvested for LD isolation at the young adult stage (24 h post L4 stage).
Isolation of Lipid Droplets-LDs were isolated using a modified method previously described (12). First, about 4 ϫ 10 5 worms were harvested and washed three times with 50 ml PBS/0.001% Triton-X100. Worm pellets were then washed with 50 ml buffer A (25 mM Tricine, pH 7.6, 250 mM sucrose, and 0.2 mM phenylmethylsulfonyl fluoride), homogenized using a polytron (Cole-Parmer® Labgen™ 125 and 700 Tissue Homogenizers) in 10 ml buffer A, and centrifuged at 1000 ϫ g for 30 s. The supernatant was homogenized again by nitrogen cavitation (Ashcroft Duralife Pressure Gauge) at 500 psi for 15 min on ice, and then centrifuged at 1000 ϫ g for 10 min. Nine milliliters of this postnuclear supernatant (PNS) was loaded into an SW40 tube, and 3 ml buffer B (20 mM HEPES, pH 7.4, 100 mM KCl, and 2 mM MgCl 2 ) was overlaid on top. The tube was then centrifuged at 12,628 ϫ g for 1 h at 4°C. The LD fraction was carefully collected from the top layer and washed three times with 200 l buffer B.
Protein Preparation and Western Blotting-Proteins were extracted and analyzed using Western blotting by a method described in our previous study (16).
Mass Spectrometry Analysis-The protocol used was the same as previously described (39). LD-proteins were subjected to reduce with 10 mM dithiotreitol by incubating at 56°C for 1 h and then alkylated for 45 min by 55 mM iodoacetamide in the dark. Proteins were then incubated with 10 l trypsin solution (10 ng/l in 25 mM ammonium bicarbonate) for 30 min on ice. After removing excess enzyme solution, 30ϳ40 l 25 mM ammonium bicarbonate was added, and digestion was allowed to proceed at 37°C overnight. Five percent formic acid was added to stop the digestion reaction, which was then vortexed and centrifuged. A C 18 trap column was used to add the peptide solution, eluted and then subjected to nano-LC-ESI-LTQ Orbitrap XL MS/MS analysis. The Orbitrap mass spectrometer was operated under data-dependent mode and was set at an initial 300ϳ1800 Da MS scan range. All MS/MS data were searched against the WormBase database Wormpep218, which was released on August 22, 2010, and contains 24761 protein sequences. The BioWorks (3.31 sp1) Sequest search parameters were as follows: enzyme: trypsin; precursor ion mass tolerance: 2.0 Da; and fragment ion mass tolerance: 1.0 Da. The variable modification was set to oxidation of methionine (Met ϩ15.99 Da). The fixed modification was set to carboxyamidomethylation of cysteine (Cys ϩ57.02 Da). The search results were filtered with Xcorr versus Charge values of, Xcorr (ϩ2) Ͼ 2.5, and Xcorr (ϩ3) Ͼ 3.5. Two times peptide FDR: FDR 1 ϭ 0.434%, FDR 2 ϭ 0.426%. peptide mass accuracy Ͻ 5ppm, SP score Ͼ 500, RSp Ͻ 5. distinct peptidesՆ2, misscleavages 2.
Transmission Electron Microscopy-The purity of LDs was examined by transmission electron microscopy. Negative staining, whole mount EM, and ultra-thin sectioning methods were used. For negative staining, the sample of purified LDs was loaded onto a carboncoated, Formvar-covered copper grid and stained for 1 min by 0.5% neutral phosphotungstic acid. To view the isolated LDs by whole mount EM, a carbon-coated, Formvar-covered copper grid was placed onto a drop of isolated LD suspension. The grid was then placed onto a drop of 2.5% glutaraldehyde solution (0.1 M PBS, pH 7.2) for 10 min and subsequently onto a drop of 2% osmium tetraoxide solution (0.1 M PBS, pH 7.2) for 10 min to fix the LDs. After fixation, LDs were stained with 0.1% tannic acid for 10 min and 2% uranyl acetate for 10 min. After each step, the grid was washed with deionized water three times, 1 min each time. For ultra-thin sectioning of purified LDs, the LD sample was first embedded in 10% gelatin (0.1 M PBS, pH 7.2). After solidification, the sample was cut into small blocks and prefixed in 2.5% glutaraldehyde (0.1 M PBS, pH 7.2) for 12 h at 4°C and post-fixed in 2% osmium tetraoxide for 24 h at 4°C. The sample was then dehydrated in an ascending series of ethanol concentrations at room temperature and embedded in Spurr. Sections of a thickness of 70 nm were prepared with a Leica EM UC6 Ultramicrotome (Leica Germany) and loaded onto Formvar-covered copper grids. Grids were then stained with 2% uranyl acetate for 15 min at room temperature before viewing with a FEI Tecnai 20 (FEI Co., Netherlands) electron microscope.
Nile Red Staining of C. elegans-Fixed Nile Red staining of C. elegans was performed as previously described (24). Approximately 200ϳ1000 nematodes were suspended in 1 ml of water. Fifty microliters of freshly prepared 10% paraformaldehyde solution was added, mixed, and worms were rocked for 1 h at room temperature. Worms were allowed to settle, and then the paraformaldehyde solution was replaced by 1 ml of 1 g/ml Nile Red in M9 and incubated for 15ϳ30 min at room temperature, with occasional gentle agitation. After most of the staining solution was removed, the fixed worms were mounted onto 2% agarose pads for microscopic observation and photography.
Confocal Microscopy-Purified LDs were stained with Nile Red and viewed with an Olympus Fluoview1000 as described previously (39).
Confocal microscopy images of fixed worms were obtained using a Spinning Disk Confocal Microscope (Zeiss). The figures were produced using a 100 ϫ 1.45 numerical aperture oil objective (Olympus PLAN APO) and an electron-multiplying charge-coupled device (EMCCD) camera (Andor iXon DV-897 BV). The fluorescence signals were analyzed by Image J software (NIH).
For photobleaching experiment, the worms were anesthetized in 0.01% tetramisole hydrochloride in PBS for 30 min and mounted on 2% agarose pads. We identified the area of interest on the Olympus Flu-oview1000, brought it to the desired focus, defined a region of interest for the photobleach, and chose photobleaching conditions so that after photobleaching the fluorescent signal of the photobleached area decreases to background intensity levels. The bleaching conditions require a 100ϳ1000-fold increase in laser power for 20 bleach iterations (roughly 2 s) with 100% transmission of a 488 nm laser. After bleaching, the image of the selected area was captured every 2 min for 10 min.
Thin Layer Chromatography (TLC)-Lipids were extracted and analyzed using TLC by a method described in our previous study (17).

Isolation of Lipid
Droplets from C. elegans-Using our newly established method, we successfully isolated LD fraction in large quantity on top of a density gradient (Fig. 1A). The LD fraction was then washed three times to remove coisolated cytosolic proteins and other cellular organelles. To test its lipid composition, we extracted total lipids from the LD fraction by chloroform and acetone, and then separated different lipid species by TLC. TLC results showed that triacyglycerol (TAG) was highly enriched in the fraction (Fig. 1B). We also measured the size of isolated LDs by a Delsa Nano C particle analyzer. As shown in Fig. 1C, the size of LDs was distributed in the range of 50 nm to 3,000 nm. All these characterizations such as density, lipid composition, and size were agreed with previous isolated lipid droplets from other sources (1,17,26,27).
Verification of the Purity of Isolated Lipid Droplets-To further test the purity of isolated LDs, transmission electron microscopy (TEM) images were taken. The results showed that all the isolated LDs appeared as spherical structures and, importantly, that membrane debris and other cellular organelles were barely detected ( Fig. 2A, upper panels). With Nile red staining and DIC imaging, purified LDs appeared to be a bead structure ( Fig. 2A, lower panels), similar to purified LDs from CHO K2 cells (12,17). We then analyzed the LD-protein profile using gel electrophoresis. The protein pattern of isolated LDs was significantly different from the other three cellular fractions (Fig. 2B). In addition, three independent isolations were analyzed and their protein profiles were almost identical, suggesting the reproducibility and quality of the purification method (supplemental Fig. S1). Immunoblotting was used to further verify the purity of the LDs (Fig. 2C). Equal amounts of proteins from LD, total membrane (TM), cytosol (Cyto), and PNS fractions were separated by SDS-PAGE and detected by the antibodies indicated. Fig. 2C shows clearly that the isolated LDs had little contamination with mitochondria (PDH-E1␣), endosomes (RME), ER (Bip), plasma membranes (Na-K ϩ ATPase), cellular membranes (Dyn), or cytosol (actin). In the LD fraction we observed the enrichment of Rab 18 and Rab 7, proteins previously identified (12,17) in LDs isolated from CHO cells (Fig. 2C). Proteomic and Bioinformatic Analyses of Lipid Droplet Proteins-Having verified the purity of isolated LDs by biochemical and morphological examinations, we extracted proteins from the purified LDs and subjected them to nano-LC-ESI-LTQ Orbitrap XL MS/MS analysis. From two independent proteomic studies, we identified 306 proteins that were presented in both analyses (Table I and Fig. 3A), and Table S1 shows the top 50 abundant proteins. It was found that 63% of these proteins had been identified in previous LD proteomic studies (Fig. 3B). When this proteome was compared with one previous proteomic study that was done using Chinese hamster ovary cells (CHO K2) (16), 28 of 125 LD proteins of CHO cells were overlapped with LD-associated proteins from C. elegans (Fig. 3C). We classified these proteins into nine groups by function or localization: lipid metabolism, other metabolism, transcription, ribosome, membrane trafficking, chaperone, signal transduction, cytoskeleton, and proteins with unknown function (Fig. 3D). The enrichment (41%) of proteins involved in metabolism further indicates the similarity of C. elegans LDs to other eukaryotic LDs.
To determine the relationships between these 306 proteins, we analyzed our data with STRING 9.0 using default parameters (28). supplemental Fig. S2 summarizes the network of predicted associations for these proteins. Some of these proteins showed close relationships and formed two main clusters. The densest cluster (upper cluster) consists of 53 ribosome proteins. Identification of ribosome proteins on LDs has also been reported in some early LD proteomic studies in humans (12), yeast (10), and Drosophila (18). This finding suggests that LDs may provide a surface for ribosomes where certain proteins can be efficiently synthesized. The other cluster (lower cluster) mainly consists of metabolic proteins that are involved in processes such as lipid metabolism, energy generation, and nucleotide biosynthesis. These results are consistent with early reports that LDs are not only centers for lipid synthesis, storage and metabolism, but also provide extra membrane surface area for protein synthesis and intracellular reactions.
Identification of DHS-3 as a C. elegans Lipid Droplet Marker Protein-Because dehydrogenases are found in most of previous LD proteomes, we constructed a transgenic line that expresses the fusion protein DHS-3::GFP under the control of a daf-22 promoter in gut epithelial cells. Interestingly, DHS-3::GFP formed ring structures in vivo, typical of the localization pattern of PLIN proteins (Fig. 4A). To validate this result biochemically, we probed the distribution of the DHS-3::GFP protein in the transgenic line using a GFP antibody and an anti-DHS-3 antibody that was generated in our lab. The Western blotting result (Fig. 4B) shows that the two antibodies recognized a single band of molecular weight ϳ63 kDa, verifying the specificity of the DHS-3 antibody and GFPfusion protein expression. Using the DHS-3 antibody, we detected DHS-3 in the purified LD fraction but not in the total membrane (TM), cytosol (Cyto), or PNS fractions of wild-type C. elegans with equal protein loading (Fig. 4C). To further analyze the enrichment of DHS-3 in the LD fraction, we tested an excess of the PNS fraction (10 ϫ and 50ϫ) using the DHS-3 antibody (Fig. 4D). A very weak DHS-3 signal was detected in the PNS fraction at 10-fold excess loading (Fig.  4D). Although 50-fold excess protein loading, the DHS-3 signal of PNS fraction was still much lower than the signal of LD (Fig. 4D). In addition, to determine if DHS-3 is a dynamic protein on LDs, selected regions of the worm were photobleached and then the fluorescent signal in those regions was detected in every 2 min, up to 10 min after bleaching (Fig. 4E).   (19,46) the same LD (arrowhead-pointed region). After 20 min DHS-3::GFP signal was still not detected in the bleached region (supplemental Fig. S3). The notable position shift of LDs indicates that these LDs were still active during these 20 min. These results demonstrate that DHS-3 can be used as a protein marker for C. elegans LDs. Verification of Lipid Droplet Proteins-To further confirm DHS-3 as a LD resident protein, we then constructed a transgenic line expressing the fusion protein pDHS-3::DHS-3::GFP under the control of its own promoter and observed that similar to pDAF22::DHS-3::GFP, pDHS-3::DHS-3::GFP also formed ring structures in vivo (Fig. 5A). To determine that these ring structures were LDs, purified LDs from DHS-3::GFP strain were analyzed using fluorescence and DIC microscopy (Fig. 5B). Almost all purified LDs detected by DIC were over-lapped with fluorescence images. A few of DIC-detected LDs were not surrounded by GFP signals (Fig. 5Bb, arrows), suggesting that not all LDs contain DHS-3. In addition, LDs in DHS-3::GFP strain were labeled using Nile Red fixed staining and merged with DHS-3::GFP fluorescence signals (Fig. 5C). Most Nile Red signals were surrounded with DHS-3::GFP (Fig.  5Cc). These experiments proved that DHS-3::GFP ring structures are LDs.
More experiments were carried out to verify this LD proteomic study in C. elegans. Because ATGL was previously found in LDs of mammalian cells (12) and C. elegans (21,29), it was chosen to do so. First, ATGL was indeed identified in the proteomics (Table I). Second, fluorescence study represented that pATGL::ATGL::GFP was colocalized with LDs stained by Lipid Tox (Fig. 5D). In addition, an unknown function protein F22F7.1 and an apolipoprotein Vit-2 that were also identified by the proteomics were further verified. Western blotting experiments also showed that F22F7.1 was very enriched in LDs and Vit-2 was partially associated with LDs (Fig. 2C). DISCUSSION We reported the purification and proteomic characterization of C. elegans LDs. Because of lack of LD marker proteins in C. elegans, we had to test the purity of our LD preparations by several other approaches. First, we demonstrated that the LD preparation was floated on top of gradient and highly enriched in TAG but not other lipid species. Second, our EM results showed that isolated LDs contained minimal contamination of other cellular organelles. Third, the protein pattern of LDs in SDS-PAGE was markedly different with TM, Cyto, and PNS. Fourth, using antibodies to detect specific marker proteins for each fraction, we detected only minimal mitochondrial, endosomal, ER, plasma membrane, and cytosolic contamination. Taken together, our results indicate that our LD preparations were relatively pure and suitable for proteomic analysis.
Results from our proteomics study suggest that C. elegans LDs and other eukaryotic LDs are conserved in that they share a large number of homologous proteins involved in lipid metabolism, membrane trafficking, and signal transduction. Of particular interest are the acyl-CoA synthetase (ACS) family, the homologue of mammalian adipose triglyceride lipase ATGL-1, and the short chain fatty acid dehydrogenases (DHS) family. C. elegans ACS-20 and ACS-22 are implicated in the incorporation of very long chain fatty acids into sphingomyelin, which may be linked to cuticle barrier formation (30). DHS-28, a peroxisomal dehydrogenase, has been shown to regulate long chain fatty acid ␤-oxidation and LD size (21). These proteins have also been identified in the LD proteome of other species (Table I). Another interesting category of LD-proteins shared between C. elegans and other species is proteins involved in trafficking and transport. We identified Rab1, 7, 8, 10, 11, and 18 in the C. elegans LD proteome. The small GTPase Rab18 localizes to LDs and promotes their close apposition to rough ER in human and mouse cells (2,31). Rab10 and Rab35 have been shown to positively regulate LD size in Drosophila S2 cells (32). Vit-1, 2, 3, 4, 5, and 6 were also present (Table I). Vit proteins are apolipoprotein homologs believed to transport lipids to growing oocytes during yolk deposition (33). Western blotting analysis showed that Vit-2 was only partially associated with LDs and the main signal was in total membrane fraction (Fig. 2C), suggesting there are at least two types of lipid storage structures in C. elegans, Vit containing lipoprotein particles with higher density and lipid droplets with lower density. Their functions and relationship are worth to be investigated.
It is not clear at present why there are so many mitochondrial metabolism proteins (e.g. mitochondrial respiratory chain proteins) in our LD proteome (Table I). Similar results have been reported before in LD proteomes for other organisms. It has been proposed that mitochondria, peroxisomes, ER membranes, and endosomes may be physically associated with LDs (5,34) or may even be continuous with LDs in terms of phospholipid membrane topology (35). Recently, we conducted an interactomic study on interaction between LDs and mitochondria and identified several pairs of proteins that may be involved in this interaction (36).
Our purification and proteomic study has uncovered many bona fide LD-proteins. Using morphological and biochemical analyses, we also identified DHS-3 as a good LD marker protein. These results prove the benefits of purifying LDs and using proteomic approaches over commonly-used genetic approaches to study C. elegans LDs. More importantly, this work provides a tool for establishing in vitro assays for the study of LDs in C. elegans.
As a direct LD detection, our approach may also be used to assess whether the Raman signals of live C. elegans captured by Coherent anti-Stokes Raman Spectroscopy and Stimulated Raman Spectroscopy come from LDs or other types of organelles (37)(38)(39). For example, the level of colocalization between LD marker proteins and the Raman signals from Coherent anti-Stokes Raman Spectroscopy and Stimulated Raman Spectroscopy may determine whether these signals can reliably act as a proxy for TAG levels in live C. elegans. Furthermore, an in vitro assay using isolated LDs and other cellular organelles may help to resolve the issue of the specificity of the Raman signals that are detected by Coherent anti-Stokes Raman Spectroscopy and Stimulated Raman Spectroscopy.