C. elegans: A Model for Exploring the Genetics of Fat Storage

Summary

To gain insights into the genetic cascades that regulate fat biology, we evaluated C. elegans as an appropriate model organism. We generated worms that lack two transcription factors, SREBP and C/EBP, crucial for formation of mammalian fat. Worms deficient in either of these genes displayed a lipid-depleted phenotype–pale, skinny, larval-arrested worms that lack fat stores. On the basis of this phenotype, we used a reverse genetic screen to identify several additional genes that play a role in worm lipid storage. Two of the genes encode components of the mitochondrial respiratory chain (MRC). When the MRC was inhibited chemically in worms or in a mammalian adipocyte model, fat accumulation was markedly reduced. A third encodes lpd-3, whose homolog is also required for fat storage in a mammalian model. These data suggest that C. elegans is a genetically tractable model to study the mechanisms that underlie the biology of fat-storing tissues.

Introduction

An organism’s ability to regulate the production, storage, and release of energy is crucial for health and survival. A major source of energy is stored as fat, which is required for the life cycle of many organisms (Campbell and Dhand, 2000; Spiegelman and Flier, 2001). Unfortunately, abnormalities in fat accumulation produce pathological states. Lipodystrophy, too little fat, produces severe metabolic disturbances, including hypertriglyceridemia and early onset type II diabetes mellitus (Garg, 2000). The converse disease state, obesity, increases the risk of heart disease, type II diabetes, hypertension, and other diseases (Must et al., 1999). The alarming worldwide increase in obesity has intensified the search to identify genes that control the development, differentiation, and function of fat-storing tissues. While some key regulators of these pathways were identified with biochemistry and cell culture systems (reviewed in Rosen et al., 2000), an appropriate invertebrate genetic model system might hasten the discovery of new genes important in fat biology.

Model organisms are a powerful resource for the discovery of genes critical to human health and disease: greater than 60% of human disease genes have invertebrate orthologs (Hodgkin et al., 1998; Jasny, 2000). As C. elegans is an excellent model for many biological processes, it seemed plausible that it might also prove to be a powerful system for analyzing the mechanisms of fat storage. There are several clues that these processes might be related in worms and mammals, including data demonstrating that neuropeptide, serotonergic, and insulin signaling pathways regulate fat storage in both nematodes and mammals (Kimura, 1997; Leibowitz and Alexander, 1998; Sze et al., 2000; Wang et al., 1998). For example, serotonin 5-HT2C receptor knockout mice and worms lacking serotonin signaling accumulate excess fat (Sze et al., 2000; Nonogaki et al., 1998). Insulin signaling is a primary mechanism for controlling metabolic physiology not only in humans, but also in worms (Kimura, 1997). In addition, worms have homologs of many mammalian lipogenic and lipolytic enzymes (C. elegans Sequencing Consortium, 1998). Worms also contain homologs of the transcription factors C/EBP and SREBP (see below), which regulate mammalian adipocyte differentiation (Rosen et al., 2000). C. elegans therefore contains homologs of genes encompassing a wide range of components of the mammalian fat regulatory cascade, including hormonal regulators, differentiation factors, and biosynthetic enzymes.

There are some potential problems with using worms as a model for fat biology. In worms, the intestine, an endodermal derivative, is the major site of energy stores, which are contained in several types of gut granules; some contain protein and carbohydrate, while others are lipid-laden, “fat granules” (Wood, 1988). In contrast, mammalian fat is thought to be a mesodermal derivative. Furthermore, worm intestinal cells are multifunctional and are not dedicated adipocytes, as in mammals. However, the important aspect of a worm model is that the genetic (i.e., molecular) control must be conserved between worm fat granules and mammalian fat.

Here we report our analysis of C. elegans as a system for studying the genetics of fat. We first characterized C. elegans homologs of two mammalian genes, SREBP and C/EBP, that control mammalian adipogenesis (Rosen et al., 2000). The worm SREBP homolog was expressed in the endodermal fat-storing cells, and RNA-mediated interference (RNAi) targeting either SREBP or C/EBP resulted in pale, skinny larval-arrested worms. These worms lacked fat stores and had reduced expression of lipogenic enzymes, suggesting a conserved role for SREBP and C/EBP from nematodes to mammals. With a reverse genetic approach, we also identified several new genes that gave a similar, fat-depleted phenotype upon RNAi. The genes identified in our reverse genetic screen represent a variety of cellular processes, suggesting that this approach might identify novel pathways that regulate fat storage.

If the C. elegans model is robust, some of the genes we identified in worms should also play roles in mammalian fat biology. To test that notion, we chose three of the genes found in our worm screen and characterized them in greater detail. Two encode components of the mitochondrial respiratory chain (MRC); worms grown in the presence of chemical inhibitors of MRC function arrested at early larval stages and had reduced fat, similar to the RNAi phenotype. We extended the data to a mammalian adipocyte model, 3T3-L1 tissue culture cells, and found that, just as in worms, lipid accumulation in 3T3-L1 cells was blocked when the cells were cultured in the same chemical inhibitors. The third gene, lpd-3, encodes a large, novel protein that, like SREBP, is expressed in the worm intestine, the site of fat storage. The mammalian lpd-3 homolog is expressed dynamically in murine and human fat. Notably, murine lpd-3 is required for lipid accumulation in the 3T3-L1 mammalian cell system. Taken together, these data suggest that C. elegans is an appropriate model organism to study fat biology; with the model, we identified genes that function in diverse processes, and all three genes that we have tested to date are required for fat storage not only in worms, but also in mammalian systems.

Results

C. elegans Mobilize Fat upon Starvation

During fasting, mammals mobilize fat stores to fulfill their energy requirements. To determine whether C. elegans also utilize fat upon starvation, we removed worms from their food at various stages of life. By six hours of starvation, both larval and adult worms were pale and lacked the dark gut granules present in well-fed animals (Figure 1A). Of note, fruitfly larvae lacking fat stores also appear pale and clear, so the starved worms might also have decreased fat. To evaluate this possibility, we fixed and stained starved and well-fed worms with Sudan black, a fat-specific dye (Ogg et al., 1997). While well-fed worms had prominent Sudan black staining, starved worms displayed reduced Sudan black staining (Figure 1B). This suggests that worms, like mammals, mobilize fat stores upon physiologic demand.

Figure 1. C. elegans Mobilize Fat Stores upon Starvation.

(A) Larval and adult worms were grown on plates with food (well-fed) or on plates without food for 8 hr (starved) and then microscopically examined and photographed. Solid arrows indicate the dark gut granules present in the well-fed worms. Double arrows indicate the absence of these granules.

(B) Well-fed and starved adult worms were fixed and stained with Sudan black. The solid arrow indicates fat labeled with Sudan black (Ogg et al., 1997). The double arrow indicates the absence of Sudan black staining in the starved worms.

Worm SREBP Is Expressed in the Intestine

To explore the possibility that the mechanisms that regulate fat biology are conserved, we examined the C. elegans genome to determine whether it contained homologs of genes known to be essential for mammalian adipocyte differentiation. Tissue culture and biochemical studies had identified three transcription factors–PPARγ, C/EBP, and SREBP-1 (ADD1)–as necessary and sufficient for mammalian adipogenesis (Rosen et al., 2000). If C. elegans is an appropriate model organism, these or related genes should be required for worm fat formation. Through database analysis, we identified SREBP-1 (Y47D38.7; 35% similarity) and C/EBP (C48E7.3; 25% similarity) homologs. PPARγ, a nuclear hormone receptor (NHR), had many similar genes in the worm database. While mammals have less than 50 NHRs, the worm genome contains >200 NHRs, making the identification of a potential PPARγ homolog difficult (Mangelsdorf and Evans, 1995; Sluder et al., 1999). If SREBP or C/EBP are involved cell autonomously in C. elegans fat accumulation, then they should be expressed in the intestine, the primary site of fat storage in the worm. To determine Y47D38.7 (SREBP) expression, we constructed an SREBP∷GFP fusion and generated transgenic SREBP∷GFP worms, which showed strong GFP expression, exclusively in the intestine (Figure 2A).

Figure 2. SREBP and C/EBP Homologs Are Required for C. elegans Fat Storage.

(A) Worm SREBP (Y47D38.7) is expressed in the intestine. Transgenic worms expressing an SREBP∷GFP translational fusion were examined by bright-field and GFP fluorescence microscopy.

(B) RNAi against either Y47D38.7 (SREBP) or C48E7.3 (C/EBP) results in lipid-depleted worms. dsRNA was injected into the gonads of young adult hermaphrodites, and the progeny were examined. The RNAi worms arrested as larvae, and DIC microscopy showed that they were pale and skinny compared with wild-type worms (worms are at the same developmental stage). In wild-type worms, intestinal fat granules fluoresced brightly after soaking in Nile red, a lipid-specific dye (Greenspan et al., 1985). Worms treated with lpd-1 (SREBP) or lpd-2 (C/EBP) dsRNA displayed no Nile red fluorescence.

RNAi against Worm SREBP and C/EBP Results in Larval-Arrested Worms that Lack Fat

The expression pattern of Y47D38.7 (SREBP) was consistent with a possible role for the gene in formation or accumulation of fat in worms. To elucidate the endogenous roles of Y47D38.7 (SREBP) and C48E7.3 (C/EBP), we performed RNAi to block their function (Fire et al., 1991, 1998). RNAi against either gene produced a similar phenotype:pale, skinny worms that arrest as early larvae (Figure 2B). We also observed the pale phenotype in the starved worms that lacked fat stores (Figure 1). Both the SREBP and the C/EBP RNAi worms appeared pale because the dark, lipid-laden fat granules normally present in the intestines of wild-type worms were absent (Figure 2B). Although the RNAi worms did not contain fat granules, other gut granules, including a population of gut granules that are birefringent under dark-field optics (Leung et al., 1999; Peters et al., 1999), were present (data not shown). This suggests that at least some intestinal function was normal.

To test the worms for fat accumulation, we stained them with Sudan black. We found that the C/EBP or SREBP RNAi worms, like the starved worms, did not stain with Sudan black (data not shown). We also evaluated fat storage with Nile red, a vital dye that specifically stains fat (Greenspan et al., 1985). This dye has the distinct advantage that it can fluorescently stain fat in living worms, providing a powerful tool for genetic screening. Consistent with the pale phenotype and lack of Sudan black staining, the RNAi worms, in contrast to wild-type worms, did not fluoresce when treated with Nile red, indicating an absence of fat (Figure 2B). That is, in the absence of the homologs of either SREBP or C/EBP, worms display a “lipid-depleted” phenotype: pale, skinny worms that lack fat. As a reflection of that, we chose lpd-1 (lipid depleted-1) as the genetic name for Y47D38.7 (SREBP) and lpd-2 as the genetic name for C48E7.3 (C/EBP).

SREBP Deletion Mutants Do Not Contain Fat

To eliminate the possibility that the lpd phenotype might have been secondary to nonspecific RNAi effects, we isolated an lpd-1 mutant, lpd-1(gf1) by screening a deletion library (Jansen et al., 1997). The mutant contained a 4 kb deletion that eliminated most of the coding region, including the DNA binding domain, so the mutant should behave like a null. The phenotype of the lpd-1 deletion mutant was indistinguishable from that generated with RNAi: larval-arrested, pale, skinny worms that did not contain fat (data not shown); therefore, the lpd-1 RNAi worms accurately phenocopied a genetic loss-of-function mutant.

Next, we examined the lpd-1(gf1) deletion mutants by electron microscopy (Figure 3A). The overall ultrastructure appeared normal, and the worms contained the appropriate complement of muscle and hypodermal cells. The intestinal lumen of the lpd-1(gf1) deletion mutants was intact, and the intestinal cells contained well-organized microvilli. However, the intestinal cells of the mutant worms appeared to lack the dark storage granules that are present in wild-type worms (Figure 3A). Taken together, these data suggest that the intestinal cells of the lpd-1(gf1) mutants have relatively normal morphology but have a specific defect in fat storage, leading to the lipid-depleted phenotype.

Figure 3. Analysis of Lipid-Depleted Worms.

(A) Electron microscopy on ultrathin transverse sections of lpd-1 (SREBP) deletion mutants demonstrated that the worm intestinal cells were relatively normal, except that they lacked dark gut granules.

(B) lpd-1 and lpd-2 RNAi worms eat and defecate at rates similar to wild-type worms. The number of pumps/minute and seconds/defecation cycle was confirmed twice for each worm (n = 15). To test for intestinal patency, worms were fed bacteria mixed with blue latex beads (Sigma) and then examined at 400 X.

(C) The expression of several lipogenic enzymes was downregulated in lpd-1 and lpd-2 RNAi worms. Feeding RNAi was performed (Fraser et al., 2000) with an empty vector, HT115 (con), or with the indicated DNA insert, lpd-1 or lpd-2, and the worms were harvested for RT-PCR. Totals of 0.5 µg and 1 µg of cDNA template were used to ensure that the PCR was semiquantitative.

SREBP and C/EBP RNAi Worms Eat and Defecate at Wild-Type Rates

It was possible that the lipid-depleted (lpd) phenotype was secondary to some mechanical problem in the alimentary canal; for example, the RNAi worms may be unable to ingest or pass food as effectively as wild-type worms. To analyze this possibility, we examined the pharyngeal pumping (“eating”) and defecation rates of the RNAi worms. Wild-type, lpd-1, and lpd-2 RNAi worms had similar average pumping and average defecation rates (Figure 3B) (Wood, 1988). In contrast, eat-2(ad465) worms pumped at a much slower rate and had an extended defecation cycle compared with wild-type worms (Figure 3B), as described (Raizen et al., 1995; Thomas, 1990). To test intestinal patency, we determined whether the RNAi worms could ingest and excrete 0.8 µm latex beads. Like wild-type worms, lpd-1 and lpd-2 RNAi worms contained beads throughout the length of the intestine, and both were able to defecate the latex beads, indicating that the intestine was functional and unobstructed (Figure 3B). These data are consistent with the idea that the lpd phenotype was not due to defective feeding, intestinal transit, or defecation.

Worm SREBP and C/EBP Regulate Lipogenic Gene Expression

To attempt to explore the mechanisms that underlie the lpd phenotype observed when either C/EBP or SREBP function was reduced, we examined the expression levels of candidate downstream genes. In mammals, C/EBP and SREBP regulate the expression of a number of lipogenic enzymes, including acetyl coenzyme A carboxylase (ACC), ATP-citrate lyase (ACL), fatty acid synthase (FAS), glycerol 3-phosphate acyltransferase (G3PA), and malic enzyme (Christy et al., 1989; Ericsson et al., 1997; Kim and Spiegelman, 1996; Rosen et al., 2000; Shimano et al., 1999). A search of the C. elegans genome identified candidate homologs of each. To determine whether lpd-1 and lpd-2regulate the expression of these lipogenic enzymes in the worm, we isolated RNA from wild-type worms and lpd-1 and lpd-2 RNAi worms. We first determined whether the RNAi had indeed decreased the expression of the targeted gene. For both lpd-1 and lpd-2, the RNAi reduced the expression of the respective gene, demonstrating the efficacy of the RNAi (Figure 3C). lpd-2 RNAi also altered the expression levels of lpd-1, but the converse was not true (Figure 3C), supporting the notion that, in worms, SREBP functions downstream of C/EBP. This epistatic relationship was also recently proposed for mammals (Rosen et al., 2002). Next, we analyzed the expression of the lipogenic enzymes (ACC, ACL, FAS, G3PA, and malic enzyme) and found that their expression levels were decreased to varying extents when either lpd-1 or lpd-2 function was reduced (Figure 3C). As a specificity control, we also examined the expression of an intestine-specific Na⁺H⁺ transporter, B0495.4 (P. Bauer et al., WormBase submission), as well as an intestine-specific actin, act-5 (Jim Waddle, personal communication); both remained unchanged in the RNAi worms (Figure 3C). With these results taken together, the molecular analysis suggested that SREBP and C/EBP regulate lipogenic genes not only in mammals, but also in C. elegans.

RNAi Pilot Screen Identifies Additional lpd Genes

Next, we attempted to identify other genes that regulate fat storage, by scoring for the lpd phenotype. We were aided in this effort by large-scale RNAi screens in which various broad categories of mutant phenotypes, including larval arrest/lethal, were reported (Fraser et al., 2000; Maeda et al., 2001). As both the lpd-1- and lpd-2-deficient worms arrested as larvae, we performed RNAi, followed by Nile red staining, of ∼80 genes that were reported in two large-scale RNAi screens as having larval phenotypes (Fraser et al., 2000; Maeda et al., 2001). We found that 8 of the 80 genes we tested resulted in the lpd phenotype upon RNAi. These genes represented a diverse array of cellular processes, and none of the genes were fatty acid biosynthetic enzymes (Table 1). This suggests that the pathways and mechanisms that regulate formation and storage of fat are broad and varied and extends the current view of the process. Notably, seven of the eight genes we identified have mammalian homologs. We chose to extend the analysis of three, lpd-3, lpd-4, and lpd-5, of those seven genes.

Table 1.

Genes Required for Fat Storage Identified by Pilot Screen

Worm Gene	WormBase	Gene Name/Function	Mammalian Homolog
lpd-3	Y47G6A.23	novel	+
lpd-4	F26E4.9	cytochrome c oxidase subunit Vb	+
lpd-5	ZK973.10	NADH-upiquinone oxidoreductase 18 kDa subunit	+
lpd-6	K09H9.6	Peter Pan homolog	+
lpd-7	R13A5.12	Pescadillo homolog	+
lpd-8	R10H10.1	nitrogen-fixing domain/RNA binding domain	+
mac-1	Y48C3A.6	AAA-ATPase	+
lpd-9	T21C9.5	novel	−

MRC Function Is Required for Fat Accumulation in Worms and in a Mammalian Cell Culture Model

ZK973.10 (lpd-5) and F26E4.9 (lpd-4) encode components of the mitochondrial respiratory chain (MRC), which couples electron transport to ATP production (Lehninger, 1982). ZK973.10 (lpd-5) encodes the MRC complex I NADH-ubiquinone oxidoreductase 18 kDa subunit. F26E4.9 (lpd-4) encodes cytochrome c oxidase Vb (COX Vb), a subunit of complex IV of the MRC. To further examine the role of the MRC in worm fat storage, we cultured L4 stage larvae on control plates or plates containing either rotenone, a specific complex I inhibitor, or NaN₃, a specific complex IV inhibitor. The F1 progeny cultured on control plates were examined and found to be healthy in appearance, with dark gut granules in their intestines that stained brightly with Nile red (Figure 4A). Many of the F1 progeny of the worms on the inhibitor plates were pale and skinny and had reduced Nile red staining (Figure 4A). EM analysis of the MRC-inhibited worms demonstrated that they are morphologically normal, with developed muscle, hypodermal, and intestinal cells (data not shown). Taken together, these results indicate that reducing MRC function in C. elegans specifically alters fat storage.

Figure 4. Inhibiting MRC Function Blocks Lipid Accumulation in Worms and 3T3-L1 Cells.

(A) Worms cultured on the indicated chemical MRC inhibitor were stained with Nile red and examined by DIC and fluorescence microscopy.

(B) Murine 3T3-L1 cells were induced to become adipocytes as described (MacDougald and Lane, 1995). At day 2 postinduction, 5 µM rotenone, a complex I inhibitor, or 5 mM NaN₃, a complex IV inhibitor, was added to the media. Fresh media and inhibitors were added every other day. At day 8 postinduction, cells were fixed and then stained with oil red O as described (Green and Kehinde, 1975). Induced cells (IND) accumulate lipid (red). Uninduced (UN) cells and inhibitor-treated, induced cells accumulate much less lipid.

(C) 3T3-L1 cells, treated as in (B), were harvested with Trizol, treated with DNase I, and analyzed, on day 3 of induction, for expression of the indicated adipogenic markers by semiquantitative RT-PCR. The –RT samples had no reverse transcriptase included in the reaction.

Mutations in genes that function in the MRC have been linked to abnormalities in fat accumulation in humans, including lipodystrophy (Brinkman et al., 1999; Zaera et al., 2001), suggesting that the lpd-4 and lpd-5 genes identified in C. elegans might also be important in mammalian fat storage. To investigate the potential role of the MRC in mammalian fat accumulation, we turned to a murine adipocyte model system–3T3-LI tissue culture cells. Under standard culture conditions, these cells behave like fibroblasts, but, in the presence of a defined hormonal induction mix, the cells differentiate into adipocytes that accumulate lipid and express appropriate fate-specific markers, including C/EBP, PPARγ, SREBP, aP2, and LPL (Cornelius et al., 1994; Green and Kehinde, 1975). If the functions of COX Vb and NADH-ubiquinone oxidoreductase were conserved from worms to mammals, then blocking their function in 3T3-L1 cells should decrease lipid accumulation. To test that notion, we grew 3T3-L1 cells to confluence, induced adipogenesis with the appropriate culture media (MacDougald and Lane, 1995), and then added rotenone to inhibit complex I or NaN₃ to inhibit complex IV. By day 6 postinduction, a large percentage of the cells cultured in induction media without inhibitor had differentiated into adipocytes, as evidenced by morphological changes and the accumulation of lipid droplets, which stained red with the lipophilic dye oil red O (Figure 4B) and fluoresced brightly when stained with Nile red (data not shown). However, induced cells cultured in the presence of either MRC inhibitor contained much less lipid (Figure 4B). To determine whether MRC inhibitor treatment blocked cellular differentiation or lipid accumulation, we assessed the expression of fat cell differentiation markers. We found that inhibitor-treated cells expressed cell fate markers, as did induced controls (Figure 4C and data not shown). In some experiments, the level of PPARγ, C/EBPα, and SREBP-1 expression was less in inhibitor-treated cells than in induced, untreated cells, which may be secondary to feedback regulation (Rosen et al., 2000). Taken together, these data support the notion that MRC function, in worms and mammals, is important in lipid accumulation.

lpd-3 Is Expressed in the Worm Intestine and in Mammalian Fat

Y47G6A.23 (lpd-3) is predicted to encode a protein with an open reading frame of 1590 amino acids that contains a presumptive signal sequence and a putative coiled-coil domain. On the basis of the lipid-depleted phenotype of Y47G6A.23 dsRNA animals, we named this gene lpd-3. lpd-3∷GFP transgenic worms, like lpd-1 (SREBP)∷GFP worms, displayed prominent GFP expression in the intestine, the site of fat storage, consistent with a cell-autonomous effect, producing the lipid-depleted phenotype (Figure 5A). Database analysis detected related ESTs in both flies (32% similarity) and humans (32% similarity). We then examined the distribution of human lpd-3 (hlpd-3) in 16 human tissues and found that the hlpd-3 transcript is most highly expressed in brain, testis, and adipose tissue (Figure 5B). To extend the lpd-3 data to mouse model systems, we cloned a fragment of mouse lpd-3 (mlpd-3) based upon the human sequence and designed mouse-specific primers. RT-PCR with these primers demonstrated that mlpd-3 is dynamically regulated in two murine systems. First, we analyzed the expression of mlpd-3 in the 3T3-L1 model and, as controls, also assessed levels of PPARγ and pref-1, which mark distinct stages of 3T3-L1 adipocyte differentiation (Sul et al., 1998; Tontonoz et al., 1994). We found that PPARγ is expressed at very low levels in uninduced cells and is markedly upregulated during induction, while pref-1 had the opposite pattern of expression, consistent with the literature (Figure 5C). mlpd-3 had a low level of expression in the uninduced cells, and this level increased during adipogenesis in a manner similar to PPARγ (Figure 5C).

Figure 5. lpd-3 Is Expressed in the Worm In testine and in Mammalian Fat.

(A) lpd-3 is expressed in the intestine. Transgenic worms expressing an lpd-3∷GFP translational fusion were examined by bright-field and GFP fluorescence microscopy.

(B) The human homolog of lpd-3 is expressed in fat, brain, and testis. PCR with human lpd-3 primers was performed with a panel of cDNA from 16 different human tissues as template. S9 ribosomal PCR is a loading control.

(C) Mouse lpd-3 is expressed at higher levels in induced 3T3-L1 (IND) than in uninduced (UN) 3T3-L1 cells. 3T3-L1 cells were grown with (IND) or without (UN) induction media and then harvested for RT-PCR. G3PDH is a loading control. PPARγ is upregulated in induced cells, while Pref-1 is downregulated.

(D) Mouse lpd-3 is expressed at higher levels in the mouse embryonic day 14.5 fat enlagen than in adult fat. PPARγ and adipsin, a marker of terminal differentiation, are expressed at higher levels in adult fat. The –RT samples had no reverse transcriptase included in the reaction.

To directly examine endogenous gene expression, we harvested RNA from adult white fat pads and from the presumptive fat pad at embryonic day 14.5 and analyzed the expression of mlpd-3, PPARγ, and adipsin. We found that adipsin was not expressed at E14.5 but was strongly expressed in adult fat, as expected of a gene that marks terminally differentiated fat (Figure 5D) (Cook et al., 1987). PPARγ also had highest levels of expression in the adult fat, but, unlike adipsin, PPARγ also had a modicum of expression in the embryonic fat enlagen at E14.5 (Figure 5D). Notably, mlpd-3 was highly expressed in the embryonic fat, with lower levels in the mature, fully differentiated fat (Figure 5D). Therefore, lpd-3 is expressed in fat storing tissues from worms to humans, and its expression is highest in developing fat, potentially presaging PPARγ expression.

mlpd-3 Is Required for Mammalian Fat Accumulation

The worm studies coupled with the mammalian expression data support the notion that lpd-3 might function during mammalian fat formation. To investigate that potential role, we exploited the 3T3-L1 adipocyte culture system for tests of lpd-3 necessity. As lpd-3 does not contain well-established structural domains, it was not possible to rationally design dominant-negative forms to block lpd-3 function. However, RNAi was recently shown to be an effective loss-of-function methodology in some mammalian cell lines (Caplan et al., 2001; Elbashir et al., 2001), although no data has been reported for 3T3-L1 cells. As the process of 3T3-L1 adipogenesis takes many days, we needed to stably inhibit the function of lpd-3. Of note, a new approach to RNAi for mammalian cells, termed short-hairpin RNAi (shRNA), relies on the expression of a short-hairpin RNA from a DNA-based vector (Brummelkamp et al., 2002; Yu et al., 2002), which should provide the ability to generate stable clones that express an shRNA to the gene of interest. We obtained an shRNA vector, mU6-Pro (Yu et al., 2002), in which the shRNA is transcribed from the mU6 RNA polymerase III promoter, and we cloned the neomycin gene, under control of the PGK promoter, into the vector in order to generate stable cell lines.

To take advantage of the shRNA approach, we constructed a vector (neo-mU6-mlpd3) containing a hairpin structure that was designed to decrease expression of mlpd-3. Then, we transfected the vector into 3T3-L1 cells and selected stable lines with G418. In a pilot study, we found that the majority of the neo-mU6-lpd-3 shRNA stable clones had lower levels of lpd-3 expression (data not shown). Next, we induced the lpd-3 shRNA 3T3-L1 cells to undergo adipogenesis and found that the clones that contained low levels of the lpd-3 transcript accumulated much less fat than those cells with wild-type lpd-3 levels. On the basis of these results, we decided to undertake a more detailed analysis that included a positive sh RNA control, for which we selected PPARγ, which is required for 3T3-L1 adipogenesis. We generated the appropriate plasmid, neo-mU6-PPARγ, and simultaneously selected many new stable clones containing either neo-mU6-mlpd-3 or neo-mU6-PPARγ. To assess the efficacy of the RNAi effect, we evaluated the levels of expression of the salient genes with RT-PCR. Again, slightly more than 50% of the clones of each type demonstrated significant reduction of expression of the appropriate gene (data not shown); the cells that did not have a lowered level of the targeted gene could then serve as controls.

To evaluate the potential of the stable lines to accumulate fat, we grew five PPARγ shRNA lines and five lpd-3 shRNA lines to confluence and induced them to become adipocytes. After induction, two of the PPARγ shRNA cell lines accumulated lipid, while three did not (Figure 6A). We then assessed PPARγ expression and found that those cell lines that contained fat had high levels of PPARγ, indicating an ineffective knockdown. All three lines that lacked lipid stores had reduced PPARγ expression (Figure 6B). Strikingly similar results were obtained with the mlpd-3 shRNA cell lines (Figures 6C and 6D). That is, when mlpd-3 expression was reduced, lipid accumulation was abrogated, but those clones that retained mlpd-3 expression stored fat normally (Figures 6C and 6D). We have undertaken similar studies with many of our shRNA cell lines, and one with a low level of mlpd-3 did accumulate lipid (data not shown). We found similar inconsistent results with a PPARγ shRNA stable clone. Next, we analyzed the induced cell lines for expression of markers of adipocyte differentiation. As expected, the expression of the markers was markedly reduced in the PPARγ shRNA lines that had low levels of PPARγ expression (Figure 6E). However, mlpd-3 expression was not altered by reduction of PPARγ (Figure 6E), which suggests that mlpd-3 expression is not regulated by PPARγ. We also found that levels of the cell fate markers were reduced in the mlpd-3 shRNA cells lacking the normal complement of mlpd-3 (Figure 6E). Of note, PPARγ expression was reduced in the mlpd-3-deficient lines, suggesting that its transcriptional regulation is dependent on mlpd-3 (Figure 6E). That is, these data are consistent with the notion that mlpd-3 functions upstream of PPARγ; however, it is possible that a feedback network underlies these findings (Rosen et al., 2000). Taken together, these data demonstrate that lpd-3 is required for fat accumulation in a mammalian model.

Figure 6. lpd-3 Is Required for Mammalian Lipid Accumulation.

(A) 3T3-L1 cells containing the PPARγ shRNA cassette were induced to undergo adipogenesis as described (Green and Kehinde, 1975). After 10 days of induction, lipid accumulation was assessed by visual inspection. While PPARγ shRNA stable lines 5 and 10 accumulated many lipid droplets, PPARγ shRNA stable lines 3, 4, and 6 did not accumulate lipid upon induction.

(B) Cells shown in (A) were harvested and processed for RT-PCR analysis. Of note, the inability to accumulate lipid correlated with an shRNA-dependent decrease in PPARγ expression.

(C) 3T3-L1 cells that contained the lpd-3 shRNA cassette were induced and evaluated for lipid accumulation. While lpd-3 shRNA stable line 1 accumulated lipid droplets, lpd-3 shRNA stable lines 4, 6, 8, and 9 did not.

(D) RT-PCR of cells shown in (C) demonstrated that the lines that did not accumulate lipid after induction were depleted in lpd-3 expression.

(E) RT-PCR analysis on the induced PPARγ shRNA and lpd-3 shRNA cell lines (A and C) demonstrated that cells with low levels of either lpd-3 or PPARγ did not express either C/EBPα, an adipogenic gene, or LPL and adipsin, markers of terminally differentiated fat. While the PPARγ shRNA lines lacking PPARγ expressed lpd-3, the lpd-3 shRNA cells with reduced lpd-3 levels (4, 6, 8, and 9) did not express PPARγ. HPRT is a loading control. The –RT samples had no reverse transcriptase included in the reaction.

Discussion

The ability to regulate fat storage is a fundamental process that is required for the life cycle of many multicellular organisms, and abnormalities in fat storage underlie important human diseases (Campbell and Dhand, 2000). Invertebrates are powerful tools for gene discovery, yet they have not been fully exploited in the study of fat development or function. In this report, we have encouraging data suggesting that C. elegans intestinal fat granules might provide an invertebrate model amenable to the study of fat. Database analysis identified C. elegans genes that were homologous to SREBP and C/EBP, transcription factors that are required for mammalian adipogenesis. Blocking the function of these genes resulted in lipid-depleted worms that arrest as larvae, are pale and skinny, and lack fat. That is, the same genes that are required for mammalian adipocyte differentiation also appear essential for fat accumulation in worms. In addition, SREBP and C/EBP regulate the expression of the same lipogenic enzymes in worms and mammals. These data are consistent with the notion that worm fat granules may serve as a model for mammalian fat storage. Considering the early state of the studies, it is not yet clear whether the pathways in worms model fat storage in adipocytes or in other mammalian cells, such as hepatocytes, that can store fat. Taken together, our data suggest a conservation of molecules that regulate fat biology across a wide evolutionary distance.

One goal of developing a model system is that further studies might lead to a better understanding of the molecular mechanisms that control adipocyte formation and function and may identify new therapeutic targets, providing tangible benefits to man. Analysis of lipid-depleted mutants and identification of the responsible genes may increase our knowledge of nematode, and, potentially, mammalian, fat storage. Our pilot screen for lipid-depleted worms identified a surprisingly diverse array of genes representing various aspects of cellular functions. It is notable that none of the genes we isolated are in the biosynthetic pathway; rather, they encompass proteins of extremely diverse structures and mechanisms (RNA binding proteins, MRC genes, AAA-ATP-ases, novel genes, etc.). This suggests a high complexity to this important biological problem. Although the isolation of mitochondrial genes as important in energy storage can potentially be rationalized on the basis of our current understanding of mitochondrial function, surprisingly few of the genes that we identified in the pilot screen had a preexisting link to fat biology. One of the genes we identified–mac-1–had been shown, just as we found, to have a larval arrest phenotype, but the preliminary studies did not include any analysis of fat (Wu et al., 1999). On the basis of our pilot screen, we estimate that approximately 5%–10% of larval arrest/ lethal worms will display the lpd phenotype; if this is so, then we should be able to identify approximately 25–75 lpd genes. This suggests that a full-scale C. elegans lpd screen will identify many new genes that regulate fat biology.

One key feature of a model system is its ability to uncover genes that are important in the analogous process in humans. The finding that seven of the eight genes we identified in the pilot screen had clear mammalian homologs suggests that the worm model is robust and increases the likelihood that we will be able to make connections to mammalian fat biology. In that regard, we attempted to extend the data from worms to mammals with three of the newly identified lpd genes. Surprisingly, all three that we tested were required in mammalian fat accumulation, just as in worms. This strongly suggests that other genes identified in worm fat screens will play a role in mammalian fat biology. The three lpd genes that we selected to examine in the mammalian system represent two ends of the spectrum of genes that we anticipate finding in the C. elegans screens. The MRC genes represent known and characterized genes that have in some way been suggested to have roles in energy storage, while lpd-3 was unknown, had no discernable functional domains, and had no known function.

COX Vb (lpd-4), one of the MRC genes that we identified, is a component of MRC complex IV. Recent data implicate reduced complex IV activity as pathogenic for two types of human fat abnormalities: HAART-related lipodystrophy and multiple symmetrical lipomatosis (Brinkman et al., 1999; Zaera et al., 2001). Of note, MRC genes are regulated in a PPARγ-dependent fashion (Mueller et al., 2002). We found that blocking MRC function, with two distinct chemical inhibitors, decreased fat accumulation in worms and in a mammalian adipogenic cell line. This demonstrates the key role that the mitochondria play in worm and mammalian fat accumulation. One goal of studying fat biology in worms is to identify proteins that might be therapeutic targets for obesity, diabetes, or other metabolic diseases. Inhibiting mitochondrial function in humans, for weight loss or as a diabetes treatment, is of potential concern because of presumed pleiotropic effects. However, worms were able to complete embryogenesis and developed to mid-larval stages when mitochondrial function was inhibited, so it is possible that a therapeutic window may exist for such a therapy.

lpd-3 encodes a novel protein that is conserved from worms to man. Worm lpd-3 is expressed in the intestine, the major site of fat storage, and the putative mammalian homolog is expressed in mouse and human fat. The expression analysis showed that (1) lpd-3 appears to be present in relatively high levels in three human tissues–brain, testis, and fat, (2) lpd-3 is regulated during 3T3-L1 adipogenesis, and (3) lpd-3 is present at higher levels in developing, rather than mature, fat. To examine the function of lpd-3 in a mammalian system, we performed necessity tests in the 3T3-L1 adipocyte model with an RNAi technique termed short-hairpin RNA (shRNA) (Brummelkamp et al., 2002; Yu et al., 2002). We generated stable 3T3-L1 lines containing shRNA directed against either PPARγ or mlpd-3 and found that greater than 50% of the lines had markedly reduced levels of the targeted gene, which correlates with other reported shRNA results (Brummelkamp et al., 2002). Of note, shRNA lines that were deficient in either PPARγ or mlpd-3 lacked lipid accumulation. However, we found that a small minority of shRNA clones directed against either gene did not completely behave as predicted, which may be expected from a new technique. Nonetheless, the trends were clear: PPARγ and mlpd-3 are required in 3T3-L1 cells for lipid accumulation. Molecular analysis of the shRNA clones revealed a potentially exciting insight; we found that cells lacking the normal amount of PPARγ maintained lpd-3 expression, while cells with reduced lpd-3 levels were deficient in PPARγ expression. The cells expressing lpd-3 shRNA also lacked C/EBPα expression. These data suggest that lpd-3 acts upstream of PPARγ. Since C/EBPα is thought to act upstream of PPARγ and is at the top of the transcriptional cascade (Rosen et al., 2002), its absence in mlpd-3 deficient cells is consistent with the notion that mlpd-3 functions quite proximally in the pathway. However, mammalian adipogenesis involves a series of positive- and negative-feedback loops (Rosen et al., 2000), so it is formally possible that mlpd-3 acts later than the data might indicate.

The control of fat storage is a fundamental problem that has important public health ramifications. The data provided here support the notion that C. elegans may indeed be a useful genetic model system to identify and study new genes involved in mammalian adipocyte differentiation and fat storage. Although we have elected to study a specific and directed worm fat phenotype, the lpd larval phenotype, C. elegans can also be further exploited to understand other aspects of fat biology. For example, one could screen to identify worms with too much fat or for other adult fat phenotypes. Through these approaches, the molecular basis that underlies this important biological area can be better elucidated.

Experimental Procedures

Sudan Black and Nile Red Staining

For Sudan black staining, well-fed and starved (removed from food for 8 hr) worms were washed in M9 and fixed in 1% paraformaldehyde in M9. The worms were then subjected to three freeze-thaw cycles and dehydrated through an ethanol series. The worms were stained overnight in a 50% saturated solution of Sudan black in 70% ethanol, rehydrated, and photographed. For Nile red staining, worms were grown on plates with 1 ng/ml Nile red and photographed with epifluorescence (rhodamine channel).

GFP Fusions

Translational GFP fusions for Y47D38.7 and Y47G6A.23 were made by SOEing PCR (Ho et al., 1989). Primer sequences are available upon request.

Electron Microscopy

Wild-type and lpd-1 mutant worms were collected, fixed in 4% paraformaldehyde by microwave fixation, and stained with osmium tetroxide. Ultrathin transverse sections were examined by electron microscopy.

Worm RNAi

For injection RNAi, primers were designed to contain either a T3 or T7 polymerase binding site. PCR products were amplified from cDNA or genomic DNA isolated from wild-type (N2) worms. The PCR products were purified on Qiagen columns and used as templates in a Megascript (Ambion) RNA transcription reaction. The dsRNA was injected at a concentration of 1 µg/µl into the gonads of young adult hermaphrodites, and the progeny were examined. The chromosome 1 feeding RNAi library was obtained from Dr. J. Ahringer (Fraser et al., 2000) through the UK HGMP Resource Centre, grown, and plated. The next day, L4 stage worms were put on the plates and allowed to lay eggs for 2 days before being removed. F1 progeny were examined for a phenotype.

RT-PCR

Total RNA from worms or cells was extracted with Trizol (Invitrogen) and DNase I treated, and then semiquantitative RT-PCR was done as described (LeSueur and Graff, 1999; Peters et al., 1999). Human RNA and cDNA was obtained from Clontech. Primer sequences are available upon request.

Cell Culture

3T3-L1 cells were grown in six-well dishes in DMEM plus 10% FCS plus pen/strep and induced to form adipocytes as described (MacDougald and Lane, 1995). Cells were fixed in 3.7% formaldehyde for 2 min and stained with oil red O.

MRC Inhibitors

Five millimolar NaNa₃ or 5 uM rotenone were added to the 3T3-L1 cells during induction. An equal volume of water (carrier control for NaN₃) or DMSO (carrier control for rotenone) was added to the control wells. Fresh NaN₃ or rotenone was added every other day when the induction media was changed. L4 worms were placed on plates containing either 70 µM NaNa₃ or 2.5 µM rotenone and allowed to lay eggs for 36–48 hr. The adult worms were then removed, and the F1 progeny were examined.

shRNA

Hairpin sequences targeting mlpd-3 and PPARγ were ordered from IDT (sequences available upon request) and cloned into mU6-Pro (Yu et al., 2002). A PGK-neo fragment was then cloned into both mU6-lpd-3 and mU6-PPARγ. 3T3-L1 cells were transfected with either the neo-mU6-lpd-3 or neo-mU6-PPARγ plasmid, and stable clones were selected in 400 µg/ml G418. Knockdown of the targeted message was determined by RT-PCR.