If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Hepatic steatosis is characterized by the accumulation of lipid droplets (LDs), which are composed of a neutral lipid core surrounded by a phospholipid monolayer embedded with many proteins. Although the LD-associated proteome has been investigated in multiple tissues and organisms, the dynamic changes in the murine LD-associated proteome in response to obesity and hepatic steatosis have not been studied. We characterized the hepatic LD-associated proteome of C57BL/6J male mouse livers following high-fat feeding using isobaric tagging for relative and absolute quantification. Of the 1,520 proteins identified with a 5% local false discovery rate, we report a total of 48 proteins that were increased and 52 proteins that were decreased on LDs in response to high-fat feeding. Most notably, ribosomal and endoplasmic reticulum proteins were increased and extracellular and cytosolic proteins were decreased in response to high-fat feeding. Additionally, many proteins involved in fatty acid catabolism or xenobiotic metabolism were enriched in the LD fraction following high-fat feeding. In contrast, proteins involved in glucose metabolism and liver X receptor or retinoid X receptor activation were decreased on LDs of high-fat-fed mice. This study provides insights into unique biological functions of hepatic LDs under normal and steatotic conditions.
Non-alcoholic fatty liver disease (NAFLD), which is characterized by hepatic steatosis, is considered the hepatic component of metabolic syndrome. NAFLD is the most common chronic liver disease in Western nations and commonly occurs with comorbidities such as obesity and type 2 diabetes (
). Hepatic LDs are dynamic organelles that change in size and number in response to acute perturbations, such as fasting/feeding, and to chronic diseases, such as obesity. LDs play a role in protein quality control, protein storage, cell signaling, and viral replication, in addition to their major role as lipid metabolism mediators (
); however, changes in the LD proteome have not been quantified in a chronic animal model of steatosis. In the present study, we have quantified the hepatic LD-associated proteome and characterized its changes in response to diet-induced development of hepatic steatosis. The findings from this study further our understanding of LD biology and the dynamic nature of the LD-associated proteome.
MATERIALS AND METHODS
Animals and diets
Male C57BL/6J weaned mice (n = 3) were fed a 60% fat diet [high-fat diet (HFD), F3282; Bio-Serv] or a normal fat diet [normal diet (ND), F4031; Bio-Serv] for 9 weeks. The liver was excised from fed anesthetized mice at approximately 9:00 AM and a portion was reserved for microscopy prior to LD isolation. All procedures were approved by the University of Minnesota Institutional Animal Care and Use Committee.
LDs were isolated from freshly excised livers using the method of Zhang et al. (
) with the following modifications. The minced liver was resuspended in 4 ml buffer A [25 mM tricene (pH 7.6), 250 mM sucrose, and protease inhibitor cocktail] and dounce homogenized 30 times on ice. Cells were homogenized using a needle and syringe instead of a nitrogen bomb and then centrifuged for 10 min at 1,000 g. Following addition of 4 ml of buffer A, the postnuclear supernatant was divided into two ultracentrifuge tubes, overlaid with 3 ml buffer B [20 mM HEPES (pH 7.4), 100 mM KCl, 2 mM MgCl2, and protease inhibitor cocktail] and centrifuged at 300,000 g for 1 h at 4°C. The LD-containing band was transferred to a fresh tube and centrifuged at 20,000 g for 20 min at 4°C. The underlying liquid was carefully removed and the LD fraction was washed six times with 200 μl buffer B to remove copurifying membranes (
The LD precipitate was reconstituted with 65 μl of protein solubilization buffer [7 M urea, 2 M thiourea, 0.4 M triethylammonium bicarbonate (pH 8.5), 20% methanol, and 4 mM tris(2-carboxyethyl)phosphine]. The samples were bath sonicated for 2 min. The samples were then transferred to a pressure cycling technology tube with a 50 μl cap for the Barocycler NEP2320 (Pressure Biosciences, Inc.) and cycled between 35 kpsi for 30 s and 0 kpsi for 15 s for 40 cycles at 37°C. Two hundred millimoles of methyl methanethiosulfonate were added to a final concentration of 8 mM. Protein concentration was determined by Bradford assay.
In-solution proteolytic digestion and iTRAQ® labeling
A 28 μg aliquot of each sample was transferred to a new 1.5 ml microfuge tube and brought to the same volume with protein solubilization buffer plus 8 mM methyl methanethiosulfonate. All samples were diluted 4-fold with 80% ultra-pure water; 20% methanol and trypsin (Promega) were added in a 1:35 ratio of trypsin to total protein. Samples were incubated overnight for 16 h at 37°C after which they were frozen at −80°C for 0.5 h and dried in a vacuum centrifuge. Subsequently, samples were cleaned with a 4 ml Extract Clean™ C18 SPE cartridge (Grace-Davidson) and eluates were vacuum dried and resuspended in 0.5 M triethylammonium bicarbonate (pH 8.5) to a final 1 μg/μl concentration. Twenty-six micrograms of each sample were labeled with isobaric tagging for relative and absolute quantification (iTRAQ)® 8-plex reagent (AB Sciex, Foster City, CA). After labeling, the samples were multiplexed together and dried in vacuo. The multiplexed sample was cleaned with a 4 ml Extract Clean™ C18 SPE cartridge, and the eluate was dried in vacuo.
Peptide LC fractionation and MS
The iTRAQ®-labeled sample was resuspended in buffer A [10 mM ammonium formate (pH 10) in 98:2 water:acetonitrile] and fractionated offline by high pH C18 reversed-phase chromatography. A MAGIC 2002 HPLC system (Michrom BioResources, Inc.) was used with a C18 Gemini-NX column [150 mm × 2 mm internal diameter, 5 μm particle, 110 Å pore size (Phenomenex)]. The flow rate was 150 μl/min with a gradient from 0 to 35% buffer B [10 mM ammonium formate (pH 10) in 10:90 water:acetonitrile] over 60 min, followed by 35–60% over 5 min. Fractions were collected every 2 min and UV absorbances were monitored at 215 nm and 280 nm. Peptide-containing fractions were divided into two equal numbered groups, “early” and “late”. The first early fraction was concatenated with the first late fraction, and so on (
). Concatenated samples were dried in vacuo, resuspended in load solvent (98:2:0.01, water:acetonitrile:formic acid), and 1.5 μg aliquots were run on a Velos Orbitrap mass spectrometer (Thermo Fisher Scientific, Inc.) as described previously, with the exception that the activation energy was 40 ms (
Database searching, protein identification, and quantitation
The mass spectrometer RAW data (ProteoWizard files) were converted to mzXML using MSconvert software and to MGF files using TINT RAW-to-MGF converter (http://github.com/jmchilton/tint). ProteinPilot 4.5 (AB Sciex) searches were performed against the NCBI reference sequence Mus musculus (taxon 10090; April 21, 2012 version) protein FASTA database with canonical and isoform sequences (85,763 proteins), to which a contaminant database (http://www.thegpm.org/crap/) was appended. Search parameters were as follows: cysteine MMTS; iTRAQ 8plex (peptide labeled); trypsin; instrument Orbi MS (1–3 ppm) Orbi MS/MS; biological modifications ID focus; thorough search effort; and false discovery rate analysis (with reversed database). Proteins were accepted if they were identified by two or more unique peptides and fell within the 5% local false discovery rate cutoff in all of the biological replicates. Relative quantification of each protein was performed using ProteinPilot 4.5 (AB Sciex) with bias correction. Each ND sample (tag 113–115) was used as a denominator and protein summary results were exported to Microsoft Excel. The HFD/ND ratios and the corresponding P values from each report were combined, yielding nine total HFD/ND ratios. A subset of proteins was generated with the following criteria: Protein Pilot P values <0.05 for six out of nine protein ratios and a log fold change of ≥0.55 or less than or equal to −0.55, and two times the standard deviation of the nine ratios was not greater than the average log fold change.
Microscopy and immunoblotting
Formalin-fixed paraffin-embedded liver sections were processed for immunohistochemistry using standard methodologies. Slides were examined using a Nikon A1 spectral confocal microscope and NIS Elements imaging software. The nucleus was stained with DAPI and perilipin 2 (PLIN2) was immunolabeled with chicken polyclonal primary antibody (Abcam, Cambridge, MA; ab37516) and Rhodamine Red-X-conjugated donkey anti-chicken secondary antibody (Jackson ImmunoResearch Laboratories, West Grove, PA). Acyl-CoA synthetase 1 (ACSL1) was immunolabeled with goat primary antibody (Santa Cruz Biotechnology) and Alexa-Fluor 647-conjugated donkey anti-goat secondary antibody (Jackson ImmunoResearch Laboratories). Similarly, carnitine palmitoyltransferase 2 (CPT2), glucose phosphate isomerase (GPI), malic enzyme (ME), and ribosomal protein large P0 (RPLP0) were labeled with specific primary antibodies (One World Lab, San Diego, CA) and Alexa-Fluor 488 anti-rabbit secondary antibody (Cell Signaling). Voltage-dependent anion channel was labeled with a specific primary antibody (Cell Signaling) and Alexa-Fluor 488 anti-rabbit secondary antibody. Images were obtained with a 60× magnification under oil immersion. For immunoblotting, equal amounts of protein, verified by Ponceau staining, were separated by SDS-PAGE and transferred onto polyvinylidene difluoride membranes for Western blot analysis. Primary antibodies used were as follows: anti-ATP-citrate lyase (ATPCL), anti-binding immunoglobulin protein (BIP), anti-cyclooxygenase IV (COX IV) (Cell Signaling Technology), anti-β-actin (LI-COR Biosciences), anti- PLIN2 (kindly provided by Dr. Andrew Greenberg), anti-early endosomal antigen 1 (Bethel Laboratories), and anti-catalase and anti-anexin-2 (One World Lab).
For electron microscopy, livers of mice fed ND or HFD were perfused with 2.5% glutaraldehyde in 0.1 M sodium cacodylate. Tissues were subsequently washed, sectioned, and stained using standard methods (
). Sections were observed under a JEOL 1200 EX II transmission electron microscope (JEOL Ltd., Tokyo, Japan). Images were obtained using a Veleta 2K×2K camera with iTEM software (Olympus SIS, Munster, Germany). For quantitative analysis, images from 20 cells were randomly selected from each mouse (two mice per dietary group). LDs, mitochondria, and their physical interactions were subsequently quantified.
Ingenuity Pathway Analysis (IPA) (QIAGEN) was used to assign enriched biological function and canonical pathway categories. The analysis settings were restricted to experimentally observed and hepatoma cell lines or liver. Suspension trapping (STRAP) was used for cellular component Gene Ontology (GO) term protein annotation (
To identify the hepatic LD-associated proteome, LDs were isolated from freshly excised livers of C57BL/6J mice that had been fed either an ND or HFD for 9 weeks. Hematoxylin and eosin-stained liver sections (Fig. 1A) show that the HFD-fed mice accumulated more LDs than mice fed the ND, as expected. Postnuclear supernatant and LD fractions were evaluated for purity based on the presence of common copurifying organelles using Western blotting of marker proteins (Fig. 1B, supplementary Fig. 1). These data show that the LD fraction is highly enriched for the LD marker, PLIN2, while we were unable to detect common endoplasmic reticulum (ER) (BIP), mitochondrial (COX IV), cytosolic (ATPCL), plasma membrane (annexin-2), and endosome (early endosomal antigen 1) contaminating proteins. However, the peroxisomal marker protein, catalase, was not depleted in the LD fraction. Although the LD fraction is largely free of contaminant proteins, the sucrose centrifugation method to isolate LD does yield some contaminant proteins, as has been reported in other LD proteomic studies (
Following LC-MS/MS analysis, we identified 1,520 LD-associated proteins (supplementary Table 1). The proteins identified through this method are proteins that are associated with the LD fraction and are not necessarily bona-fide LD proteins confirmed to reside primarily or exclusively on the LD and, thus, will be referred to as LD-associated proteins. For comparison, a list of the LD-associated proteins identified by previous proteomic studies of mammalian cells and tissues was compiled (
). To decrease the likelihood of similar isoforms preventing a protein match, we used the UniRef50, which merges proteins of similar sequence to one UniRef entry. We compared the UniRef50 and gene symbol identifiers between the previously identified LD proteins and our protein list (
). From this analysis, we determined that 671 out of 1,520 proteins identified have been reported in previous mammalian LD proteomics studies. In addition, 72 and 95% of the proteins found in previous studies on human (
) hepatic LDs, respectively, were identified in our analysis (supplementary Table 1). Using confocal microscopy, we show that three of these proteins, RPLP0, GPI, and ME1, show partial colocalization with the LD marker protein, PLIN2, in mouse liver tissue sections (supplementary Fig. 2). It should be noted that these proteins do not primarily localize to LDs and the small amount of colocalization would suggest that these, and perhaps many other proteins identified, are not abundant LD-associated proteins.
We next performed functional annotation of the hepatic LD-associated proteome. The top five enriched biological functions categories of LD-associated proteins, as determined by IPA, are listed in Fig. 2A. Within the functional categories of lipid metabolism, molecular transport, and small molecule biochemistry, the most highly enriched biological function annotations were related to lipid metabolism, as expected, and are as follows: 1) concentration of TG, acylglycerol, and lipid; 2) transport of molecules, fatty acid, and palmitic acid; 3) secretion of molecules and lipid; and 4) oxidation of lipid and fatty acid. Similar to the IPA analysis, the major enriched molecular functions by DAVID GO term annotation are oxidation-reduction, generation of precursor metabolites and energy, and protein localization (data not shown).
We also performed a canonical pathway analysis using IPA and show the top 30 enriched canonical pathways (Fig. 2B). The most significantly enriched pathways are eIF2 signaling, mitochondrial dysfunction, fatty acid β-oxidation I, and clathrin-mediated endocytosis signaling. Because our dataset identified more LD-associated proteins than previous mammalian LD proteomic datasets, we were unsure whether the same pattern of organelle interactions would be observed. We used a GO slim assignment software to assign cellular component GO terms to both the novel and previously identified proteins from our dataset (
). The proportions of the cellular components are nearly identical to previous studies (Fig. 2C), suggesting that the increased number of proteins we identified was not due to new organelle associations.
After filtering for statistical significance, 48 proteins were increased and 52 proteins decreased in response to high-fat feeding and development of hepatic steatosis (Table 1). Of the proteins that were increased, eight were also increased on LDs in humans with NAFLD (Table 1, supplementary Table 1) (
). IPA was used to identify biological function category enrichment (Fig. 3A). The increased functional categories of endocrine system disorders and energy production are consistent with the insulin resistance and altered lipid metabolism that characterize high-fat-fed mice. Canonical pathway analysis using IPA (Fig. 3B, Table 2) revealed that the most significantly enriched pathways for the increased proteins are those involved in fatty acid metabolism. Numerous LD-associated proteins were also decreased in pathways involving liver X and retinoic acid receptor activation, coagulation and prothrombin activation pathways, and glucose metabolism and clathrin-mediated endocytosis. As in the entire dataset, we used a GO slim assignment software to assign cellular component GO terms to the regulated proteins (Fig. 3C, Table 3) (
Two of the most robustly induced LD-associated proteins in response to the HFD were long chain ACSL1 and CPT2. ACSL1 has been identified to activate long chain fatty acids and selectively channel them to mitochondrial β-oxidation (
). CPT2 is an inner mitochondrial membrane protein important for fatty acid transport and activation. Using confocal microscopy, we found both ACSL1 and CPT2 to have more colocalization with the LD marker protein, PLIN2, in response to high-fat feeding (Fig. 4). Despite increased colocalization on LDs in response to high-fat feeding, the staining patterns of ACSL1 and CPT2 were very different. Only a small amount of ACSL1 was found on LDs and it was generally confined to one region of a given LD. In contrast, much more CPT2 was found on LDs and it surrounded LDs similar to PLIN2 or other bona-fide LD proteins (Fig. 4). As a control, we stained mouse liver sections with a mitochondrial protein that was not identified in our proteomics screen. Voltage-dependent anion channel, an inner mitochondrial membrane protein, stained mitochondria in the cell but did not show colocalization with LD marker protein PLIN2 (supplementary Fig. 3). To further explore LD-mitochondria interactions, we examined electron microscopy micrographs of normal and high-fat-fed mouse livers. We quantified the number of mitochondria and LDs and calculated the percent of mitochondria interacting with LDs and the percent of LDs interacting with mitochondria. These data revealed that approximately 40% of LDs were in contact with mitochondria and 5% of mitochondria were in contact with LDs, but there was no change in either of these parameters between the livers of mice fed the ND or HFD (supplementary Fig. 4). Consistent with these data, we also did not observe an enrichment of mitochondria proteins in our GO analysis (Fig. 3C). Taken together, these data support the proteomic analysis showing that proteins involved in lipid catabolism are enriched on LDs in response to high-fat feeding and hepatic steatosis, and suggest that they are not the result of increased organelle interactions.
The nature of the interaction between primarily non-LD-localized proteins and the LD is unclear in most cases. The LD comes in contact with other organelles, so it is possible that other organelles, membrane fragments, and/or proteins copurify with the LD fraction. It is also possible that some of these proteins do indeed reside on the LD, but in a very transient or low stoichiometric basis. To represent the variety of proteins identified in the LD-associated protein dataset, we built a network using Cytoscape 3.0.1 with the ClueGO plugin (Fig. 5). Consistent with these LD-organelle interactions, many of the proteins in our dataset are annotated with multiple cellular components and/or biological functions, as shown by the many connecting edges. Nodes, which are GO terms, were selected to represent the variety of organelle interactions.
These data provide a quantitative analysis of changes in the hepatic LD-associated proteome in response to a model of diet-induced hepatic steatosis. Our analysis identified many proteins that have been previously found in LD proteomic analysis. We also found a greater number of proteins than previous hepatic LD proteomic analyses, which may in part be due to the higher level of detection of the mass spectrometer used in the current study or different LD isolation or analytical methodologies. A limitation of any proteomics study, including the current one, is determining whether proteins identified in our analysis are bona-fide LD proteins or contaminants. We did not detect contamination, via Western blotting, of any organelle except peroxisomes. However, Western blotting is much less sensitive than MS. As an example, we found the organelle markers ATP-CL and β-actin in our MS analysis of LD-associated proteins, but we were unable to detect them in the LD fraction by Western blotting. Thus, it is difficult to distinguish between bona-fide LD proteins and contaminant proteins, as is the case in any proteomic study. Additionally, many LD-associated proteins may be associated with LDs, but this only represents a small amount of the total intracellular distribution. Indeed, we found that several LD-associated proteins colocalized with PLIN2, but most of the proteins resided primarily in other intracellular locations. An additional caveat of these studies is that although PLIN2 is the most abundant hepatic LD protein, it is not present on all hepatic LDs and, therefore, is an imperfect LD marker (
) recently used a protein correlation profiles strategy to identify high confidence LD proteins from Drosophila SH2 cells. Given that this study identified a similar total number of proteins to our study (1,481 vs. 1,520), we determined the mammalian orthologs of the entire Drosophila LD protein dataset and compared them to our dataset (
) to our dataset; seven of the Drosophila proteins had no mouse ortholog by our analysis. Of the 111 proteins they identified as bona-fide LD proteins, we found 60 homologs in our proteomic analysis.
The canonical pathway of fatty acid β-oxidation was enriched in our screen along with many proteins involved in lipid metabolism. We identified enzymes of both peroxisomal and mitochondrial fatty acid oxidation. Tight interaction between LDs and the peroxisome has been previously reported (
). Consistent with these data, we detected the peroxisomal marker protein, catalase, in our LD fraction both by MS and Western blot (supplementary Fig. 1, supplementary Table 1). We identified the mouse homologs for seven of the eight yeast proteins involved in peroxisomal β-oxidation that were identified in a previous LD screen in yeast (
). Consistent with this tight interaction, the most highly enriched GO annotation cluster by DAVID analysis was the mitochondrial part (data not shown).
In addition to mitochondria and peroxisomes, many organelles are in close proximity to LDs. ER proteins also commonly copurify with the LD fraction, which may be due to the fact that the ER, the reported origin of LDs, surrounds and/or protrudes into most, if not all, LDs (
) put forward the hypothesis that LDs are reservoirs for protein sequestration. The sequestration could be used to prevent highly abundant proteins from aggregating, to store unfolded proteins, to hold proteins that are targeted for degradation, or to inactivate proteins. It is possible that the abundance of vesicle trafficking proteins that localize to the LD bring other organelles into close contact with the LD. Proteins associated with vesicle trafficking, such as RAB GTPases and synaptosome-associated proteins, have been commonly reported on LDs (
Pathways involving fatty acid catabolism were highly increased in the LD proteome in response to high-fat feeding. Isoform 2 of dynamin-like 120 kDa protein (OPA1), which was increased on LDs following high-fat feeding, has been shown to regulate protein kinase A-mediated lipolysis (
). Consistent with increased lipolysis, monoglyceride lipase (MGLL) was also enriched in LDs from mice fed the HFD. ACSL1 has been shown to promote the activation of fatty acids destined for mitochondrial oxidation in numerous tissues (
). The mitochondrial L-carnitine shuttle pathway is enriched due to the increase of CPT2, which converts medium and long chain acylcarnitines into acyl-CoAs that can be used for β-oxidation. Very long chain acyl-CoA dehydrogenase (ACADVL), which catalyzes the first step of β-oxidation in the mitochondria, was also enriched in LDs in response to the HFD. In support of this data, ACADVL, ACSL1, MGLL, and CPT2 were also found in higher abundance on hepatic LDs in human subjects with NAFLD, suggesting that these proteins are conserved in models of NAFLD (
). However, our studies do not show increased LD-mitochondria interactions in response to high-fat feeding. Clearly, much remains to be learned regarding the factors governing the oxidation of hydrolyzed fatty acids in the liver and how this changes during the development of steatosis.
We also observed an increase in many ribosomal and ER proteins on hepatic LDs in high-fat-fed mice. Ribosomal proteins have been previously observed on LDs in leukocytes via electron microscopy (
) also identified pyruvate kinase and phosphoglycerate mutase as hepatic LD-associated proteins and found that numerous proteins involved in carbohydrate metabolism were also decreased on LDs in response to acute high-fat feeding. Glycolytic enzymes are known to interact with caveolae (
), suggesting that alterations in the LD-associated proteome could directly influence carbohydrate metabolism.
It is not immediately clear why extracellular proteins would localize intracellularly to LDs or why their localization would be affected by hepatic steatosis; however, many of the identified extracellular proteins have links to hepatic steatosis, obesity, and diabetes (
). Additionally, it is not clear whether these extracellular proteins are in the secretory pathway or are being endocytosed. However, in support of the latter, proteins involved in endosomal trafficking have been previously identified on LDs (
). The decreased presence of secreted proteins accounts for the reduction of numerous canonical pathways in response to high-fat feeding, including liver X and retinoic acid receptor activation, coagulation system acute phase response signaling, and prothrombin activation pathways.
In summary, our analysis has more comprehensively defined the hepatic LD-associated proteome and, for the first time, quantified the dynamic regulation of LD-associated proteins during chronic diet-induced hepatic steatosis. These data provide novel insights into the complexity of LD biology and point toward unique functions of LDs beyond simple lipid storage organelles.