Defining the roles of FSP27 in lipid droplet formation and apoptosis

Transcription

1 The University of Toledo The University of Toledo Digital Repository Theses and Dissertations 2010 Defining the roles of FSP27 in lipid droplet formation and apoptosis Kun Liu Medical University of Ohio Follow this and additional works at: Recommended Citation Liu, Kun, "Defining the roles of FSP27 in lipid droplet formation and apoptosis" (2010). Theses and Dissertations This Dissertation is brought to you for free and open access by The University of Toledo Digital Repository. It has been accepted for inclusion in Theses and Dissertations by an authorized administrator of The University of Toledo Digital Repository. For more information, please see the repository's About page.

2 A Dissertation entitled Defining the Roles of FSP27 in Lipid Droplet Formation and Apoptosis by Kun Liu Submitted to the Graduate Faculty as partial fulfillment of the requirements for the Doctor of Philosophy Degree in Biomedical Sciences in Cancer Biology Dr. Cynthia Smas, Committee Chair Dr. William Maltese, Committee Member Dr. Kevin Pan, Committee Member Dr. Randall Ruch, Committee Member Dr. Dorothea Sawicki, Committee Member Dr. Xiaodong Wang, Committee Member Dr. Patricia Komuniecki, Dean College of Graduate Studies The University of Toledo August 2010

3

4 ABSTRACT The adipocyte-specific protein fat-specific protein 27 (FSP27) is one of the three cell death-inducing DNA fragmentation factor 45 (DFF45)-like effector (CIDE) proteins. The other two CIDE proteins are CIDEA and CIDEB. CIDE proteins consist of two domains. The CIDE-N domain shares a homology to the CIDE-N domain of DFF45, and the CIDE-C domain is highly conserved across the three family members (Inohara et al., 1998). FSP27 expression is restricted to adipose tissues. Expression of FSP27 is significantly induced during adipogenesis. Insulin, a key lipogenic hormone, enhances FSP27 transcript expression in adipocytes, whereas the lipolytic agent TNF-! diminishes its expression. The first known function for CIDEs was the promotion of apoptosis upon ectopic expression in mammalian cells. FSP27 leads to a dose- and time-dependent apoptosis in all non-adipose cell lines tested to date. This is mediated through the intrinsic apoptosis pathway involving the release of cytochrome c from the mitochondria, activation of caspases and cleavage of their substrates. The apoptosis induced by FSP27 can be inhibited by the pan-caspase inhibitor Z-VAD-FMK or in the presence of a dominant-negative caspase-9. In addition to its role in apoptosis, FSP27 is also demonstrated to be a lipid droplet associated protein whose expression enhances formation of enlarged lipid droplets and which is required for the unilocular lipid droplets typical of white adipocytes in vivo. Here I delineate relationships between lipid droplet localization and apoptotic function of FSP27. I demonstrate that ectopic expression of FSP27 induces enlarged lipid droplets in multiple human cell lines, indicating that its mechanism involves ubiquitously present, rather than adipocyte-specific, cellular machinery. Furthermore, promotion of lipid droplet formation in HeLa cells by culturing iii!

5 in exogenous oleic acid offsets FSP27-mediated apoptosis. Using transient cotransfections and analysis of lipid droplets in HeLa cells stably expressing FSP27, I show that the FSP27 protein does not protect lipid droplets from the action of ATGL lipase. Using domain mapping with egfp-fsp27 deletion constructs, I found that lipid droplet localization of FSP27 requires amino acids 173 to 191 of its CIDE-C domain, which is also essential for FSP27 to execute its apoptotic function. In addition, FSP27 interacts with another CIDE family protein, CIDEA. Overall, my findings demonstrate the function of the FSP27 CIDE-C domain and its subregions in apoptosis, lipid droplet localization and interaction with CIDEA. iv

6 ACKNOWLEDGEMENTS! I would like to express my sincere gratitude to my major advisor, Dr. Cynthia Smas, for the great opportunity to work in her lab. Her guidance, patience, and encouragement have been one of the major driving forces during my graduate school study. More importantly, it is she who has taught me how to become a qualified scientist with all the scientific thinking and research skills I have developed during these years. I would like to thank all my committee members: Dr. William Maltese, Dr. Kevin Pan, Dr. Randall Ruch, Dr Dorothea Sawicki, Dr. Xiaodong Wang and Dr. Han-Fei Ding for all the constructive advice they have offered. I would like to thank all the lab members: Shengli Zhou, Dr. Yu Wu, Dr. Ji Young Kim, David Rearick, Kristin Tillison, Dr. Junho Lee and Dr. Ruby Fernandez for all the help they offered and all the happy memories we have shared. I would like to thank the Department of Cancer Biology for creating such a wonderful scientific research atmosphere. Last but not least, I would like to thank my families back in China and here for all their endless support and care. v!

7 TABLE OF CONTENTS Abstract. iii Acknowlegements.. v Table of Contents.. vi Introduction... 1 Literature... 7 Manuscript One Assessment of fat specific protein 27 (FSP27) in the adipocyte lineage suggests a dual role for FSP27 in adipocyte metabolism and cell death Manuscript Two 127 Functional analysis of FSP27 protein regions for lipid droplet localization, caspasedependent apoptosis and dimerization with CIDEA Discussion and Summary Conclusions References. 210 vi

8 INTRODUCTION The increased prevalence of obesity is a top health threat to individuals in developed countries. The percentage of the US population with obesity has increased strikingly since World War II, due to life style changes combined with imbalanced energy intake and expenditure. Obesity is co-morbid with various life-threatening diseases, such as insulin resistance, type II diabetes, cardiovascular disease and some types of cancers (Kahn and Flier, 2000; Shulman, 2000; Unger, 2003). Therefore, obesity is without doubt already impeding the standard of living and economic development in multiple direct or indirect ways. Understanding the molecular mechanisms controlling the development of obesity is essential for designing effective therapies for this disease. When energy intake exceeds energy expenditure, surplus energy is stored in the form of triglycerides (TG), consisting of fatty acids esterified to glycerol. TG is packaged into lipid droplets within white adipocytes of white adipose tissue (WAT) for use as an energy reservoir (Brown, 2001; Martin and Parton, 2006). When energy is needed, such as during starvation, TG in lipid droplets undergoes a coordinated, hormonally controlled hydrolysis regulated by the camp-pka pathway. With the subsequent action of adipocyte cytosolic lipases on TG, fatty acids and glycerol are released (Ducharme and Bickel, 2008). Fatty acids can be used by tissues via "-oxidation to yield ATP. Complete oxidation of one gram of fatty acid gives rise to seven times the energy generated by the same mass of glycogen, which underscores the importance of TG as the energy storehouse. 1!

9 While white adipocytes serve as a depot to harbor extra energy, such storage also acts to compartmentalize fatty acids to prevent their exerting lipotoxic effects. The white adipocyte is unique among all cell types in that it is exclusively designed to safely store massive amounts of TG. In obesity, however, the storage capacity of white adipocytes appears to reach an upper limit. As storage capacity of white adipocytes in WAT becomes exceeded, elevated levels of non-esterified fatty acids rise in the circulation. This in turn can result in lipotoxicity as various other tissues take up and store lipids. Unlike WAT, such tissues are not designed to harbor excess lipid. Lipotoxicity contributes to the development of several diseases such as insulin resistance, hepatic steatosis and cardiovascular disease (Unger, 2003). Therefore, proper energy homeostasis relies on a finely tuned balance of TG storage and mobilization. Roughly 95% of the cellular space of the mature white adipocyte in vivo is occupied by its unilocular lipid droplet. Lipid droplets were originally regarded merely as inert cellular structures used for energy storage (Brown, 2001; Martin and Parton, 2006). In recent years, however, a growing body of data has led to the new realization that lipid droplets are highly dynamic organelles. Much research is now focused on achieving a thorough understanding of the molecular mechanisms of the generation, enlargement, and turnover of lipid droplets. Since white adipocyte lipid droplets are the central determinant in protection from lipotoxicity, the identification and functional study of adipocyte lipid droplet-associated proteins is of high importance. Apoptosis is a type of programmed cell death that has an enormous impact on tissue homeostasis, development, and oncogenesis (Meier et al., 2000; Taylor et al., 2008; Hotchkiss et al., 2009). Apoptosis can be triggered by a number of stimuli, such as DNA 2!

10 damage, elevated reactive oxygen species, deprivation of growth factors, and ligand activation of Fas receptor (Taylor et al., 2008; Hotchkiss et al., 2009). The two bestcharacterized apoptosis pathways are intrinsic and extrinsic. The intrinsic apoptosis pathway involves BCL-2 family proteins, and competition among them at the outer mitochondrial membrane determines the outcome of this pathway. Once the activity of pro-apoptotic BCL-2 proteins exceeds the activity of anti-apoptotic BCL-2 proteins, mitochondrial membrane integrity is compromised and results in release of cytochrome c from mitochondria. Assembly of cytochrome c with APAF-1 and pro-caspase-9 leads to the generation of active caspase-9, followed by induction of the caspase cascade for execution of cell death. The extrinsic apoptosis pathway is mediated through ligation of cell surface death receptors. This results in the formation of the death-inducing signaling complex (DISC) that enhances the self-proteolytic activity of pro-caspase-8. Active caspase-8 then cleaves downstream caspases to trigger apoptosis. Cross-talk can occur between these two apoptosis pathways. For example, activation of caspase-8 during the extrinsic pathway cleaves pro-apoptotic BID to generate a truncated form that translocates to mitochondria to induce the intrinsic apoptosis pathway (Li et al., 1998). FSP27 (Fat-Specific Protein 27) was originally discovered as a gene whose expression increased dramatically during in vitro TA1 preadipocyte adipogenesis (Danesch et al., 1992). It was subsequently found to be a member of the cell deathinducing DFF45-like effector (CIDE) protein family. The CIDE family consists of FSP27, CIDEA and CIDEB. An alternative name for FSP27 is CIDEC. CIDEs were initially identified based on their protein sequence homology to an N-terminal region of DFF40 and DFF45 (Inohara et al., 1998), termed the CIDE-N domain. DFF40 and 3!

11 DFF45 act to govern genomic DNA fragmentation during late-stage apoptosis. DFF40 is the active nuclease and DFF45 is its inhibitory partner protein. In addition, the three CIDE proteins also share a conserved CIDE-C domain in their C-terminal halves. Early research on the CIDE family mainly focused on their apoptotic activity. However, studies in the last several years, including gene knockout, have shown that CIDE family proteins are crucial modulators of lipid metabolism (Nishino et al., 2008; Puri et al., 2008a; Toh et al., 2008). FSP27 is an adipocyte specific lipid droplet-associated protein functioning in white adipocyte unilocular lipid droplet formation. FSP27 null mice have reduced WAT mass, are resistant to diet-induced obesity and have enhanced insulin sensitivity. FSP27 is therefore a unique protein with functions in both apoptosis and lipid droplet formation. I sought to further identify the mechanisms of FSP27 in apoptosis and lipid droplet formation, and to begin to address whether FSP27 could serve as a link between these two cellular processes. My study focused on FSP27 apoptotic mechanism, its lipid droplet localization, and its lipid droplet formation function. Work by others in the laboratory had shown that: 1.) In mice, FSP27 is detected only in adipocytes of WAT and brown adipose tissue (BAT). 2.) Use of multiple in vitro adipogenesis models determined that FSP27 is not expressed in preadipocytes and is dramatically induced during adipocyte differentiation. 3.) Treatment of 3T3-L1 adipocytes with the lipogenic hormone insulin significantly increases the transcript expression level of FSP27. 4.) On the other hand, treatment of 3T3-L1 adipocytes with the lipolytic agent TNF-! diminishes its expression. By employing lipid-loaded HeLa cells as a model, I showed that FSP27 is a lipid droplet- 4!

12 associated protein and that a subregion of its CIDE-C domain is responsible for its lipid droplet localization. I determined that ectopic expression of FSP27 is capable of triggering the enlargement of lipid droplets in multiple cell lines. This suggests that the cellular machinery utilized by FSP27 for lipid droplet formation and enlargement is ubiquitously expressed, rather than fat cell specific. At the same time, I also demonstrated that the two other CIDE family members, CIDEA and CIDEB, also had lipid droplet localization and could promote lipid droplet enlargement. I revealed that the key ratelimiting adipocyte cytosolic lipase for TG hydrolysis, adipose triglyceride lipase (ATGL), could offset the lipid droplet enlargement phenotype of FSP27. This indicates that the balance between FSP27 and adipocyte cytosolic lipase action is essential for homeostasis of lipid droplet size. Unlike protein targeting to other intracellular organelles, there is no known short linear consensus sequence for targeting to lipid droplets for any protein studied to date. My further investigations identified amino acids 173 to 191 within the CIDE-C domain of FSP27 as necessary for its lipid droplet localization. It is important to note, though, that this 19 amino acid sequence alone was not sufficient for lipid droplet localization when fused to egfp. Interestingly, the 19 amino acid region I identified that is key to FSP27 lipid droplet localization was recently reported to be the site of an FSP27 homozygous mutation in a partial lipodystrophy patient. This patient harbors a homozygous mutation at amino acid 186 which leads to generation of a form of FSP27 protein that is truncated within the CIDE-C domain. Ectopic expression studies revealed that this mutated protein was unable to localize to lipid droplets (Rubio et al., 2009). 5!

13 In parallel to my studies of FSP27 function in lipid droplets, I have investigated the apoptotic mechanism of FSP27 by employing transient transfection in mammalian cells. Ectopic expression of FSP27 provokes robust cell death in all cell lines I have analyzed to date (Kim et al., 2008; Liu et al., 2009). I next determined that cell death induced by FSP27 is apoptosis since it features genomic DNA fragmentation, activation of caspases- 3, -7, -9 and cleavage of their substrates such as PARP and!-fodrin. FSP27 induces apoptosis in a dose- and time-dependent manner as early as 18 h post-transfection. FSP27-mediated apoptosis can be blocked by addition of a pan-caspase inhibitor to cells at the time of transfection. Furthermore, based on the release of cytochrome c, activation of caspase-9, and cell death inhibition by dominant-negative caspase-9, I concluded that FSP27 induces the intrinsic apoptosis pathway. Curiously, I found that the region controlling the pro-apoptotic activity of FSP27 mapped to the same 19 amino acid sequence as that which I showed necessary for its lipid droplet targeting. The observation that the same protein region is utilized by FSP27 to regulate two distinct processes led us to look for a possible tie between these two events. Interestingly, I discovered that lipid droplet localization of FSP27 could significantly attenuate its proapoptotic activity. This was based on comparing the degree of cell death induced by FSP27 in HeLa cells incubated with or without exogenous fatty acids. This discovery may partially explain why lipid-laden white adipocytes, which express high levels of FSP27, are refractory to its pro-apoptotic effects. Overall, understanding the roles of FSP27 in regard to lipid droplet formation and apoptosis should lead to significant insights for the development of specific therapies for obesity. 6!

14 LITERATURE I. LIPID METABOLISM AND LIPID DROPLET BIOLOGY A. Role of Lipid Droplets in Human Disease The lipid droplet is a dynamic cytoplasmic organelle serving as the major site for storage of neutral lipids as TG and cholesterol esters (Brown, 2001; Martin and Parton, 2006; Ducharme and Bickel, 2008; Thiele and Spandl, 2008). Under normal conditions, nearly all types of cells generate small lipid droplets whose TG is utilized for their ongoing energy needs. However, lipid droplets are found most abundantly in adipocytes and also steroid-producing cells. White adipocytes in vivo are unique in that they contain a single large unilocular lipid droplet. As a reservoir for surplus energy, the white adipocyte serves to prevent excess levels of circulating non-esterified fatty acids in times of energy excess. Historically, the lipid droplet was considered merely as a simple static organelle for storage of surplus energy (Beckman, 2006). In recent years, increasing attention has been paid to the biogenesis, trafficking, and fusion of lipid droplets, and to the identification of their associated proteins (Martin and Parton, 2006; Ducharme and Bickel, 2008; Thiele and Spandl, 2008). A growing body of data clearly indicates that the lipid droplet is actively involved in a plethora of essential physiological processes. This includes finely tuned regulated uptake and release of energy by orchestration of lipase activity, membrane assembly, and steroid synthesis. In addition, products of lipid metabolites are essential signaling pathway intermediates. For instance, diglyceride (DG), a derivative of TG hydrolysis, is an important second messenger in activation of protein kinase C (PKC) 7!

15 (Brose and Rosenmund, 2002). Fatty acids can serve as ligands for peroxisome proliferator-activated receptor gamma (PPAR#), the master regulator of the adipogenic differentiation program (Tontonoz et al., 1994). Thus, on one hand, fatty acids are essential for cell and organism function. On the other hand, excessive fatty acids are clearly detrimental to human health. Even white adipocytes, which have the capacity to dramatically increase in size as TG content expands, cannot store the mass of TG present in marked obesity, leading to lipid overflow into non-adipose tissues and lipotoxicity. The lipid-induced dysfunction in the lean tissues is called lipotoxicity and lipid-induced programmed cell death is referred to as lipoapoptosis (Unger, 2003). Lipodystrophies are diseases hallmarked by reduced white adipose tissue mass. As such, lipodystrophic individuals have very limited TG storage capacity within WAT. As a consequence, such patients commonly evidence ectopic lipid accumulation in non-adipose tissues and its pursuant lipotoxic effects, similar to those manifested in obesity (Gary, 2004; Garg and Agarwal, 2009). Imbalanced lipid regulation is closely associated with the malfunctions of non-fat storage tissues and multiple metabolic diseases (FIG. 1), for example, obesity, insulin resistance and type II diabetes (Kahn and Flier, 2000; Shulman, 2000; Spiegelman and Flier, 2001). Lipolytic activities in WAT are coordinately regulated with respect to the concentration of fatty acids in plasma. Due to their very limited capacity to safely store lipid, chronic exposure to fatty acids is detrimental to virtually all non-adipose tissues. Efficient sequestration of TG in white adipocytes is therefore an essential process to prevent overload of fatty acids to circulation and peripheral tissues. In summary, understanding lipid droplet biology has true potential to lead to therapeutic inroads into treatment of life-threatening human diseases. 8!

16 Figure 1. Imbalanced Lipid Regulation is Associated with Multiple Diseases. 1. Key Target Organ Sites in Lipotoxicity Multiple cells and organs are affected by lipotoxicity, lipoapoptosis and imbalanced lipid homeostasis. These include: a.) Peripheral (muscle) insulin resistance. Elevation of plasma fatty acids is often associated with insulin resistance. Experiments conducted in human volunteers reveal that fatty acid can inhibit insulin-stimulated glucose uptake in a dose-dependent manner (Boden et al., 1994; Shulman, 2000). A molecular mechanism contributing to insulin resistance in human muscle cells involves decreased activity of glucose transporter type 4 9!

17 (GLUT4). In normal conditions, insulin binding to its receptor leads to tyrosine phosphorylation of insulin receptor substrate-1/-2 (IRS-1/-2). This activates downstream PI-3 kinase pathways and ultimately facilitates translocation of GLUT4 to the plasma membrane and other responses to stimulate glucose uptake. However, in the situation of increased delivery of fatty acid to muscle cells, protein kinase C$ (PKC$) is activated, which in turn increases serine phosphorylation of IRS-1/-2. Activity of IRS-1/-2 depends on its tyrosine phosphorylation status. Hence, serine phosphorylation of IRS-1/-2 results in lowered PI-3 kinase activity and subsequent reduced overall glucose uptake via GLUT4 (Gerald, 2000). b.) Pancreatic " cell apoptosis. Multiple studies showed that fatty acids are also involved in inducing apoptosis in cultured rat pancreatic " cells. This appears to be due to the proapoptotic function of ceramide. Ceramide triggers an increase in the production of inducible nitric oxide synthase (inos). A ceramide synthetase inhibitor completely blocked the apoptosis provoked by fatty acids (Shimabukuro et al., 1998). c.) Lipodystrophy. Patients with lipodystrophy contain almost no adipocytes; as a result, most of their lipids are stored in other, non-adipose, tissues. These patients evidence severe insulin resistance, hepatic steatosis, diabetes and cardiac disorders in early adulthood (Garg, 2004; Garg and Agarwal, 2009). d.) Atherosclerosis. Imbalanced lipid droplet homeostasis also contributes to the development of atherosclerosis (Mozaffarian et al., 2006). Foam cells are macrophages 10

18 that have accumulated a large amount of low-density lipoproteins (LDL), a carrier of cholesterol in the blood. Such cells harbor overly large lipid droplets. Aggregation of cholesterol-enriched foam cells forms atherosclerotic plaques (Becker et al., 2010). e.) Non-alcoholic fatty liver disease (NAFLD). This disease, which often stems from viral infection, features accumulation of large numbers of lipid droplets in hepatocytes via the process of steatosis (Marceau et al., 1999; Musso et al., 2009). NAFLD is frequently accompanied by hepatic insulin resistance and elevated levels of circulating fatty acids. This leads to a de novo lipogenesis and impaired very low-density lipoprotein (VLDL) export. This ultimately results in accumulation of significantly increased amounts of lipid droplets in liver (Parekh and Anania, 2007). f.) Cancer cachexia. Lack of WAT is also a common symptom of cancer patients with cachexia (Lelli et al., 2003). This is due to the penetration of macrophages into white adipose tissue which produce the lipolytic agent TNF-!. TNF-! stimulates lipolysis leading to enhanced TG hydrolysis. B. Lipid Droplet Structure Mature white adipocytes are striking in morphology in that they contain an enormous unilocular lipid droplet, which occupies nearly 95% of the intracellular space. This type of lipid droplet is not normally found in any other cell types. Lipid droplets contain neutral lipids in their core, predominantly TGs and cholesteryl esters. By using cryoelectron microscopy, it has been shown that the lipid droplet core is surrounded by a 11

19 phospholipid monolayer (Tauchi-Sato et al., 2002, also see FIG. 2). This is also unique as all other intracellular organelles are surrounded by a phospholipid bilayer. In recent years, proteomic methods using fat cells or lipid-loaded non-fat cells as source material have determined that lipid droplets contain a vast array of proteins, such as the wellstudied perilipin, ADRP and TIP47, so-called PAT protein family members (Brasaemle, 2007). Though the physiological roles of most lipid droplet-associated proteins remain to be determined, studies of a handful of such proteins to date reveal that they are key to all aspects of lipid droplet function such, as maintaining structural integrity or controlling the process of lipolysis. Figure 2. Morphology of a Lipid Droplet. Lipid droplets consist of a hydrophobic core surrounded by a monolayer of phospholipids with various associated proteins. Note: Relative sizes of lipid core, phospholipids and associated proteins are not shown to scale. 12

20 C. Biogenesis of Lipid Droplets 1. Fatty Acid Synthesis Fatty acid synthesis takes place in cytosol, and in humans this occurs primarily in the liver. The precursor for fatty acid synthesis is acetyl-coa, which is then activated by acetyl-coa carboxylase (ACC), leading to the generation of malonyl-coa. In mammalian cells, the formation of fatty acids takes place in a series of reactions that are catalyzed by fatty acid synthase (FAS), a complex of two identical units. Each unit contains two subunits, the substrate entry and condensation subunit includes acetyl transferase, condensing enzyme and malonyl transferase. The reduction unit consists of dehydratase, enoyl reductase, "-ketoacyl reductase and acetyl transferase (Berg et al., 2002). 2. Packaging of Triglycerides into Lipid Droplets Fatty acids are initially ingested as dietary components in the form of TG. They can not be directly absorbed by the small intestine in this form. Bile acids, a derivative of cholesterol, play a critical role in TG absorption by promoting their emulsification. This facilitates hydrolysis of TG into MG (monoglyceride) and free fatty acids via the activity of pancreatic lipase. MG and fatty acids retain their association with bile acids and complex with other lipids to form micelles. Micelles then diffuse to the brush border of small intestinal enterocytes, where most fatty acids are taken up by fusion of micelles with the enterocyte plasma membrane. In some cases, fatty acid uptake by enterocytes can be mediated by specific membrane transport proteins. Upon uptake by enterocytes, fatty acids and MG are transported into the endoplasmic reticulum (ER) of these cells and 13

21 used to synthesize TG. TG is packaged with cholesterol, lipoproteins and other lipids into chylomicrons. Chylomicrons are released into the lymphatic system and subsequently enter the circulation. Fatty acids synthesized in the liver are converted to TG and transported to blood in the form of VLDL, a type of lipoprotein made by the liver. Once chylomicrons and VLDL reach the surface of adipocytes, membrane-associated adipocyte lipoprotein lipase hydrolyzes TG into fatty acids and MG. These are then transported into adipocytes followed by resynthesis of TG and packaging into the lipid droplet (Berg et al., 2002). 3. Generation of Lipid Droplets From the ER The prevailing view is that the origin of the lipid droplet is the ER. It is thought to involve a transient accumulation of neutral lipids between the luminal and cytoplasmic leaflets of the ER membrane to form a disk-like structure. This then bulges into the cytoplasm and buds from the ER to form a nascent lipid droplet (Martin and Parton, 2006; Ploegh, 2007; Thiele and Spandl, 2008). This model is supported by multiple experimental observations and is consistent with the fact that lipid droplets are surrounded by a phospholipid monolayer. Substantial numbers of reports indicate that lipid droplets are in close apposition or have a certain level of continuity with the ER membrane in mammalian cells (Blanchette et al., 1995; Ozeki et al., 2005; Martin et al., 2005). Moreover, there are reports that lipid droplets are associated with ER-like cisternal structures (Martin et al., 2005; Robenek et al., 2006). In S. cerevisiae, large numbers of lipid droplets can be observed to be associated with the ER (Szymanski et al., 2007). Further support for this model comes from proteins identified to be localized to both ER 14

22 and lipid droplets, such as calnexin (Sato et al., 2006). In addition, acyl- CoA:diacylglycerol acyltransferase (DGAT2), an enzyme for the final reaction of TG synthesis, is also localized in ER (Wakimoto et al., 2003). Nonetheless, there are still some observations that cannot be explained by this model of lipid droplet biogenesis. For instance, the fatty acid composition of the phospholipid monolayer of the lipid droplet differs from that of the ER (Tauchi-Sato et al., 2002). Overall, even though there are a large number of observations supporting the current ER-derivation model for lipid droplet formation, the molecular mechanisms involved remain poorly defined. Recently, a genome-wide RNAi screen in Drosophila S2 cells indicated that 1.5% of all Drosophila genes function in biogenesis and regulation of lipid droplets (Beller et al., 2008; Guo et al., 2008). In this study, lipid droplet phenotypes upon gene knockdown were classified into five categories based on lipid droplet number, size and intracellular localization pattern (Guo et al., 2008). Genes affecting lipid droplet phenotype included those for regulation of vesicular transport, cytoskeleton reorganization, translational machinery, transcriptional regulators, and subunits of the proteasome. For instance, sirna or pharmacological inhibition of the COPI-mediated trafficking pathway led to a lower rate of lipolysis. This was due to a decreased ability of ATGL, a key TG lipase, to associate with lipid droplets (Beller et al., 2008). Overall, such screens suggest that fine tuned coordination of multiple proteins participating in various cellular processes determines the formation and function of lipid droplets. D. Lipid Droplet Associated Proteins 1.) PAT Proteins 15

23 PAT proteins were the first family of lipid droplet-associated proteins identified. Among all lipid droplet-associated proteins, PAT proteins are the most well-characterized and have served as a long-standing prototype for research on lipid droplet-associated proteins. PAT proteins have a region of amino acid similarity at their N-terminal half, which is termed the PAT domain (FIG. 3). The PAT protein perilipin is the first lipid droplet-associated protein described and was identified through cdna cloning of transcripts abundantly expressed in adipocytes (Greenberg et al., 1991). The PAT family name is based on the first letter of each of the three founding family members: perilipin, ADRP and TIP47 (Brasaemle 2007). The PAT family was later extended to five members with the addition of S3-12 and OXPAT. Herein, I will mainly discuss the cellular function of perilipin, ADRP and TIP47. Identification of consensus amino acids sequences that are specific for targeting proteins to the lipid droplet has been of particular interest to lipid droplet research. However, so far, little information has been generated regarding this. For instance, even though all PAT proteins localize to lipid droplets, the PAT domain is dispensable for their lipid droplet localization. Rather, their localization requires multiple regions such as three hydrophobic stretches in the case of perilipin (Garcia et al., 2003). The perilipin ortholog LSD1 of Dictyostelium, and LSD1 and LSD2 of Drosophila maintain the ability to localize to lipid droplets in mammalian cells, indicating a conservation of lipid droplet-localization signals for PAT proteins (Miura et al., 2002). 16

24 Figure 3. PAT Protein Family. (Modified from Brasaemle, 2009). a. Perilipin Perilipin is the first identified lipid droplet-associated protein (Greenberg et al., 1991). It has high expression level, around 0.25% to 0.5% of total fat cell protein. It is highly enriched in white and brown adipocytes, and can also be detected at a low level in steroidogenic cells of the adrenal cortex, testes and ovaries (Greenberg et al., 1991; Servetnick et al., 1995). Perilipin is essential for regulating both basal- and hormonallystimulated lipolysis. Overexpression of perilipin A, the predominant isoform of perilipin, increases TG storage in 3T3-L1 preadipocytes due to the inhibition of TG hydrolysis (Brasaemle et al., 2000a). Under basal conditions, perilipin is regarded as a structural barrier around the lipid droplet that prevents the free access of cytosolic lipases to TG, and thereby exerts an anti-lipolytic function. During starvation, catecholamine-mediated signals through "-adrenergic receptors result in hormone-stimulated lipolysis. In this 17

25 setting, perilipin is phosphorylated by protein kinase A. Phosphorylated perilipin serves as a docking site for HSL to facilitate the maximal hydrolysis of reserved TGs to fulfill energy requirements (Sztalryd et al., 2003). To analyze the physiological role of perilipin, perilipin knockout mice were generated by two individual groups (Martinez-Botas et al., 2000; Tansey et al., 2001). Both showed that perilipin -/- mice had reduced WAT mass and were resistant to dietinduced obesity. In addition, adipocytes isolated from these mice demonstrated elevated basal lipolysis compared to wild-type mice. Generation of perilipin knockout mice also provided additional insight into the contribution of perilipin in overall lipid droplet homeostasis, in that there was a profound decrease in hormone-stimulated lipolysis. Since perilipin is a key factor regulating lipolysis, it is therefore not surprising that it is a downstream target for regulation by many factors key to lipid homeostasis. For example, the lipolytic agent TNF-! can reduce its mrna expression (Souza et al., 2003), whereas the lipogenic transcription factor PPAR# stimulates its gene expression (Dalen et al., 2004). b. ADRP The PAT family member adipose-differentiation-related protein (ADRP) was first identified as an adipocyte enriched transcript of unknown function. It was subsequently identified as a lipid droplet-associated protein based on its localization at milk lipid globule membranes (Heid et al., 1996). Unlike perilipin, ADRP is not adipocyte enriched; it is expressed in various cell lines and tissues (Brasaemle et al., 1997). ADRP localizes exclusively to lipid droplets (Brasaemle et al., 1997; Heid et al., 1998). This may be due 18

26 to its instability in the cytoplasm where its rapid proteasome-mediated degradation has been described (Xu et al., 2005). Overexpression of ADRP in COS, NIH-3T3, and Swiss- 3T3 cells increases lipid droplet formation (Gao and Serrero, 1999; Imamura et al., 2002), and knockdown of ADRP expression in THP-1 macrophages reduces number and size of lipid droplets (Larigauderie et al., 2006). Interestingly, adipocyte lipid droplets in perilipin -/- mice show increased levels of ADRP. This suggests that ADRP may be able to occupy the same niche on the lipid droplet surface as perilipin. Since perilipin null mice nonetheless evidence higher levels of basal lipolysis, despite this lipid droplet enrichment of ADRP, this means that the increased presence of ADRP at lipid droplets does not substitute for perilipin function in this animal model (Tansey et al., 2001). c. TIP47 Tail-interacting protein of 47 kda (TIP47) was first identified as a participant in transporting mannose 6-phosphate receptors (MPR) from endosomes to the trans Golgi network. This was discovered as the result of a yeast-two hybrid screen using the cytoplasmic domain of MPR as bait (Díaz and Pfeffer, 1998). TIP47 was later proved to be a lipid droplet-associated protein (Wolins et al., 2001). TIP47 is expressed in adipocytes and various other cells and has been reported to have both a cytoplasmic and lipid droplet localization. With the addition of excess exogenous fatty acids, TIP47 is rapidly recruited from cytosol to lipid droplets. During the process of lipid droplet maturation in adipocytes, association of TIP47 at small lipid droplets is gradually replaced by ADRP and perilipin as lipid droplets increase in size (Wolins et al., 2005). Study of the crystal structure of TIP47 suggested that its C-terminal region can form a 19

27 four-helix bundle that is similar to the LDL receptor binding domain of apoliprotein E, a protein that can be localized to both lipid droplet and cytosol. This may explain the double cellular localization of TIP47 (Hickenbottom et al., 2004). 2.) Proteomic Identification of Lipid Droplet Protein Complexity The past few years have seen a dramatic increase in the number of proteins identified as localized to lipid droplets or as being recruitable to lipid droplets under conditions of increased free fatty acid supply. Largely through proteomic studies, numerous lipid droplet-associated proteins have been identified (Brasaemle et al, 2004; Liu et al., 2004; Fujimoto et al., 2004; Cermelli et al., 2006). Vast types of proteins have been reported to be associated with lipid droplets. These include enzymatic and regulatory components of lipid metabolism and vesicular trafficking machinery. The identification of trafficking proteins in lipid droplets was the first clue that they are highly dynamic organelles. Advances in mass spectroscopy have led to highly sensitive protein identification with respect to lipid droplet proteomics. However, the purity of lipid droplet preparations, whose isolation is typically based on physical methods for organelle separation, continues to be a prime concern in respect to proteomic data. Therefore, in order to distinguish bona fide lipid droplet-associated proteins from artifacts, careful case-by-case assessment needs to be conducted for each protein. For example, at least 18 Rab family members have been found in lipid droplets by proteomic studies. However, for all but one of these Rab proteins, their localization to lipid droplets has yet to be confirmed by morphological observation or functional analysis. To date, Rab18 is the only such protein proven to be a 20

28 lipid droplet-associated (Ozeki et al., 2005; Martin et al., 2005). In contrast, several other Rab proteins, previously found in lipid droplets by proteomic studies, have failed to be validated for lipid droplet localization by morphological observation or functional analysis (Ozeki et al., 2005; Martin et al., 2005). Various others of the 18 Rab proteins still require full investigation. There is also a possibility that certain Rab proteins can associate with lipid droplets only under specific circumstances, such as during lipolysis or lipogenesis. Nonetheless, even if some of these Rab proteins do not ultimately prove to be lipid droplet-associated, the fact that lipid droplets contain trafficking proteins remains a profound conceptual advance. Caveolins are also a group of proteins observable in lipid droplets. These proteins normally reside at caveolae, a type of lipid raft of the plasma membrane, or at the Golgi apparatus. Their localization can be shifted to lipid droplets by treatment with exogenous fatty acids such as oleic acid. This process is reversible as caveolins move back to their original location once extra fatty acid is removed (Pol et al., 2004). Another interesting study uncovered an unexpected role(s) for lipid droplets during Drosophila embryogenesis (Cermelli et al., 2006). It showed that lipid droplets can transiently store large amount of histones H2A, H2Av and H2B in early embryos and transfer them to the nucleus throughout the process of embryogenesis. By this means, the lipid droplet serves as a shelter to prevent the release of free histones when they are not needed or would be otherwise detrimental. A theory has been proposed that lipid droplets may generally serve as a cellular sequestration sites for controlling inactivation of proteins, prevention of toxic protein accumulation and for the localized delivery of signaling molecules (Welte, 2007). 21

29 E. Enlargement of Lipid Droplets The current model is that large lipid droplets result from fusion of small lipid droplets. There are several studies that have begun to unravel roles of specific proteins that influence the size of lipid droplets. However, the means by which these proteins execute their functions and the mechanisms whereby the size of lipid droplets increases from tiny nascent to mature remain to be elucidated. It appears that the growth of lipid droplets results from fusion of nascent lipid droplets, since incubation with triacsin C, a potent inhibitor of fatty acid synthase (FAS), does not affect the enlargement of lipid droplets once small lipid droplets are already present (Boström et al., 2005). It is also widely observed that small lipid droplets tend to aggregate together, which could facilitate their close association and subsequent fusion. It has been shown that increase in the size of lipid droplets is dependent on motordriven transport along microtubules (Boström et al., 2005). Using lipid-loaded NIH 3T3 cells, enlargement of lipid droplets was impaired in cells treated with nocodazole to depolymerize microtubules, or vanadate to block ATPase activity of the motor protein dynein. Furthermore, dynein was found to co-immunoprecipitate with ADRP, a wellcharacterized lipid droplet-associated protein. This suggested that lipid droplet growth involves interactions of lipid droplet proteins with motor proteins and microtubules. Such interactions could help lipid droplets become closely associated with each other to facilitate subsequent fusion. Lipid storage droplet-2 (LSD2), an ortholog of perilipin in Drosophila, was demonstrated to be critical to the movement of LDs in Drosophila embryos. In this study, although lipid droplets in embryos that lacked LSD2 could still be 22

30 observed to undergo random bidirectional movement, they failed to demonstrate the highly directional net movement typical of lipid droplets in this setting (Welte et al., 2005). The mechanism appears to involve association of LSD2 with other proteins controlling lipid droplet movement in flies such as Klarsicht (KLAR), a crucial protein in the regulation of bidirectional transport of lipid droplets (Welte et al., 1998). KLAR, a protein with novel sequence, is unique to flies; other types of proteins may function similar to KLAR in higher organisms. Given that soluble NSF attachment protein receptors (SNARE) proteins are key components of membrane fusion machinery (Jahn and Scheller, 2006), Olofsson s group postulated that these proteins may also be associated with lipid droplets. To isolate lipid droplets, they used immuno-affinity purification targeting the well-characterized lipid droplet-associated protein, ADRP. Western blot analysis of this lipid droplet preparation revealed the presence of N-ethylmaleimide-sensitive factor (NSF),!-synaptosomalassociated protein (!-SNAP), synaptosomal-associated protein 23 (SNAP23), syntaxin-5 and vesicle-associated membrane protein (VAMP). Further characterization via sirna delivery or microinjection of dominant-negative constructs led to the conclusion that these SNARE proteins could mediate the fusion of lipid droplets (Boström et al., 2007). This intriguing finding indicates that lipid droplets can utilize universal membrane fusion machinery to increase in size, despite the fact that they have a phospholipid monolayer rather than a bilayer. Genetic approaches in yeast have been used to identify genes affecting lipid droplet size as well as other phenotypes. This model offers the advantage of the limited number of genes encoded by the yeast genome combined with the availability of single-gene 23

31 knockout yeast collections for every gene, allowing for whole genome analysis. Two groups discovered that the yeast gene Fld1p, whose mammalian ortholog is Berardinelli- Seip congenital lipodystrophy 2 (BSCL2)/Seipin, was critical for controlling the size of lipid droplets (Szymanski et al., 2007; Fei et al., 2008). Deletion of F1d1p significantly increased the size and reduced the number of lipid droplets, compared to the small, discrete lipid droplets observed in the wild type BY4741 yeast strain. Introduction of two human Seipin isoforms enabled yeast Seipin mutants to display normal lipid droplet morphology. Interestingly, mutations in Seipin cause a form of human lipodystrophy, Berardinelli-Seip congenital lipodystrophy 2 (Garg, 2004). These patients lack subcutaneous fat"!#$%!insulin resistant, and display muscular hypertrophy (Friguls et al., 2009). There are some intriguing observations that the complement of lipid dropletassociated proteins can change in a defined manner during lipid droplet enlargement. For example, during the adipocyte differentiation process of 3T3-L1 preadipocytes, PAT family members S3-12 and TIP47 were observed in nascent lipid droplets, however these were subsequently replaced by two other PAT family proteins, ADRP and perilipin, during the growth in size of these lipid droplets (Wolins et al., 2005). In HepG2 cells, the content of Rab18 and ADRP on lipid droplets appears inversely correlated, since overexpression of Rab18 decreased the association of ADRP with lipid droplets (Ozeki et al., 2005). In this regard, it appears likely that distinct sets of temporally-defined lipid droplet-associated proteins could serve to facilitate processes such as lipid droplet fusion or to maintain the structural integrity of lipid droplets during their maturation. 24

32 F. Degradation of Lipid Droplets Degradation of TG stored in lipid droplets is important for energy supply in situations such as starvation and the energy demands of intense exercise. It is conducted via lipolysis, the process by which lipids are hydrolyzed. TGs are hydrolyzed into glycerol and fatty acids that are then released into the circulation for uptake by other tissues, such as skeletal muscle, to generate ATP through "-oxidation within mitochondria. Glycerol and free fatty acids also serve as substrates for gluconeogenesis and ketogenesis in the liver (Berg et al., 2002; Ducharme et al., 2008). Lipolysis in adipocytes is mediated via the camp-dependent protein kinase A (PKA) pathway. When energy is needed, catecholamine hormones such as epinephrine and norepinephrine bind and activate cell-surface "-adrenergic receptors, a class of G protein-coupled receptors mainly expressed in adipocytes (Fève et al., 1991). The resultant receptor conformational change leads to activation of adenylate cyclase, which in turn results in increased production of camp. camp activates PKA, which subsequently governs the actions of two key cytosolic hormone responsive lipases in adipocytes, namely, hormone sensitive lipase (HSL) and adipocyte triglyceride lipase (ATGL). 1. Adipocyte Cytosolic Lipases Lipid droplet TG lipolysis is carried out through the concerted of three cytosolic lipases (FIG. 4): ATGL, HSL and monoglyceride lipase (MGL). These lipases govern basal lipolysis in most, if not all, cells and also hormone-stimulated lipolysis in fat cells. 25

33 In the adipocyte, response to starvation or energy demand governed by "-adrenergic signals, involves the coordinate regulation of ATGL and HSL for maximal lipolysis. MGL activity appears to be constitutive. Figure 4. Steps in TG Hydrolysis by Lipases at Adipocyte Lipid Droplets. a. Hormone Sensitive Lipase HSL was the first hormone-stimulated lipase described in adipocytes. HSL was first cloned from rat adipocytes (Holm et al., 1988), and has strong expression in adipocytes with lower expression in other cells and tissues. Because in vitro biochemical studies showed that HSL could readily carry out hydrolysis of TG and DG (Cook et al., 1981; Fredrikson et al., 1981; Osuga et al., 2000), HSL and MGL were regarded to be the only lipases needed by adipocytes for the complete hydrolysis of TG to fatty acid and glycerol. Initial studies of HSL action led to a model for hormone-regulated activity of HSL via its 26

34 phosphorylation status. When hormone-stimulated lipolysis occurs, PKA phosphorylates three serine residues in HSL as well as serine sites in perilipin. This results in the translocation of HSL from cytosol to its perilipin docking site on the lipid droplet surface and maximizes HSL enzymatic activity (Egan et al., 1992; Anthonsen et al., 1998; Brasaemle et al., 2000b; Clifford et al., 2000; Sztalryd et al., 2003; Carmen and Víctor, 2006). As will be discussed, this model has subsequently been updated with the incorporation of the actions of ATGL and comparative gene identification-58 (CGI-58). For many years, HSL was believed to be the only hormone-sensitive lipase for TG hydrolysis. Surprisingly, HSL -/- mice did not demonstrate metabolic features researchers initially expected to result from the predicted impaired TG hydrolysis (Osuga et al., 2000). The mass of WAT showed no change in HSL -/- compared to wild type, although some increase was observed for adipocyte size. In contrast to expectations, HSL -/- mice were not obese or cold-sensitive, which suggested no major defects in TG hydrolysis. More importantly, HSL -/- WAT still maintained 40% of the TG hydrolysis activity compared to wild type. Since HSL -/- mice showed significantly increased accumulation of DG in adipose and other tissues, this indicated that its primary substrate in vivo was DG, not TG (Haemmerle et al., 2002). Taken together, these observations suggested that an as yet unidentified lipase governed hydrolysis of TG to DG in adipocytes and in other cells in vivo. Further studies led to the discovery that this lipase was ATGL. b. ATGL In 2004, three papers were published at nearly the same time that described a newly identified gene termed adipose triglyceride lipase (ATGL) (also named Desnutrin, 27

35 Calcium-independent phospholipase A2%, and PNPLA2). These studies demonstrated it was ATGL, not HSL, that was responsible for the hormone-responsive hydrolysis of TG to DG, and that this was the rate-limiting step in lipolysis (Jenkins et al., 2004; Villena et al., 2004; Zimmermann et al, 2004). In murine tissues, ATGL is expressed predominantly in BAT and WAT and to a lesser extent in testis, heart, kidney, thymus, cardiac and skeletal muscle (Villena et al., 2004; Zimmermann et al, 2004). Compared to wild-type mice, obese ob/ob and diabetic db/db mouse express decreased levels of ATGL transcript (Villena et al., 2004), which points towards its role in TG mobilization. ATGL expression is rapidly induced during the process of differentiation from preadipocytes to adipocytes. In vitro biochemical studies indicate that ATGL is a highly active lipase for the hydrolysis of TG to DG, which is then a substrate for the subsequent action of HSL (Zimmermann et al, 2004). Following initial biochemical assessment, genetic inactivation of ATGL was carried out, which validated its TG hydrolysis function in vivo (Haemmerle et al, 2006). Because ATGL is responsible for the first step of TG hydrolysis to release fatty acids, ATGL -/- mice show defects in lipolysis. These mice had increased whole body fat with a striking elevated level of lipids in heart. As the heart relies heavily on fatty acid "-oxidation for energy, ATGL -/- mice evidenced premature death due to cardiac dysfunction. Due to lack of sufficient fatty acid availability, these animals also had defective thermogenesis with increased cold-sensitivity. As fatty acids were less available for energy generation by "- oxidation to ATP, this resulted in the use of alternative sources of energy such as glucose. ATGL -/- mice showed increased glucose use, enhanced glucose tolerance and elevated 28

36 insulin sensitivity. These studies have proved that ATGL is an essential enzyme regulating energy homeostasis in vivo. Unlike HSL, whose activity depends on its phosphorylation and docking to perilipin for its lipid droplet localization, the level of lipid droplet-associated ATGL is not altered by "-adrenergic receptor stimulation. Rather, ATGL needs the presence of its cofactor CGI-58 for full activity. Under basal condition (i.e. in the absence of "-adrenergic stimulation), CGI-58 interacts with perilipin at the lipid droplet and is not complexed with ATGL. Upon hormone stimulation and phosphorylation of perilipin, CGI-58 dissociates from perilipin to become cytosolic, and then interacts with ATGL at the surface of lipid droplets (Fig. 5). CGI-58 association with ATGL, leads to an increase of ATGL activity by 20-fold. CGI-58 is frequently mutated in patients with Chanarin- Dorfman syndrome, a disease featured by excessive accumulation of TG in multiple tissues. Mutant CGI-58 protein present in Chanarin-Dorfman syndrome patients fails to activate ATGL, which underscores significant role of CGI-58 in ATGL-mediated lipolysis (Lass et al., 2006). Human patients with ATGL mutations show mild myopathy. All of them harbor a truncated form of ATGL that is unable to properly interact with CGI-58 (Fischer et al., 2007; Kobayashi et al., 2008). 29

37 Figure 5. Adipocyte Cytosolic Lipases Are Important Regulators for Both Basaland Hormone-Stimulated Lipolysis. Upper: Basal lipolysis is conducted at a low rate. Bottom: Hormone-stimulated lipolysis. ATGL, CGI-58 and HSL work coordinately to control the rate of TG hydrolysis during hormone-stimulated lipolysis. II. APOPTOSIS A. Overview of Apoptosis In Greek, apoptosis is translated as leaves falling from a tree (Hotchkiss et al., 2009). It was first described based on key morphological changes during this type of cell death (Kerr et al., 1972). Apoptosis is one of the two fundamental types of programmed 30

38 cell death, together with autophagy-associated cell death (Hotchkiss et al., 2009). Cells undergoing apoptosis are characterized by cell shrinkage, plasma membrane blebbing, nuclear condensation and eventually the formation of apoptotic bodies. There are two major well-characterized pathways. The intrinsic pathway is mediated through mitochondrial activity, while the extrinsic pathway depends on cell death receptors. Given its bearing on the fates of cells, apoptosis must be tightly controlled, with dysregulation of this program underlying the pathological basis of various diseases. For example, inefficient apoptosis leads to development of cancer (Hanahan and Weinberg, 2000; Lowe et al., 2004) and autoimmune disorders (Strasser et al., 2009). On the other hand, unrestrained apoptosis results in neurodegenerative diseases, for example (Yuan and Yankner, 2000). Hence, numerous endeavors have been made towards identifying key components of various apoptotic pathways; this was one of the most intensive research areas in the 1990s. Early research on apoptosis was conducted extensively in the nematode Caenorhabditis elegans, due to the countable number of cells in the adult body and the short lifespan of this organism (Sulston and Horvitz, 1977; Ellis and Horvitz, 1986; Meier et al., 2000). With this system, various key proteins regulating apoptosis were identified (Miura et al., 1993; Yuan et al., 1993). For instance, egg laying abnormal-1 (EGL-1), cell death abnormality-3 (CED-3), CED-4, and CED-9. Loss-of-function analyses determined that any of EGL-1 (Conradt and Horvitz, 1998), CED-3 (Miura et al., 1993) and CED-4 (Yuan and Horvitz, 1992) triggered cell death, whereas CED-9 (Hengartner et al., 1992) promoted cell survival. Since apoptosis is an evolutionarily conserved process, researchers then turned to identification and study of the mammalian 31

39 counterparts of these proteins (Adams and Cory, 1998; Meier et al., 2000). Subsequently, BCL-2 associated X protein (BAX) (Oltvai et al., 1993), B-Cell CLL/Lymphoma 2 (BCL- 2) (Hengartner and Horvitz, 1994), Apoptotic Protease-Activating Factor (APAF-1) (Zou et al., 1997), and caspase-9 (Li et al., 1997) were identified as mammalian orthologs of EGL-1, CED-9, CED-4, and CED-3, respectively. However, the apoptotic mechanisms in mammals were far more complex than in nematodes, with a large number of factors involved. B. Caspases and Their Key Substrates Caspases are a family of proteases that have an essential cysteine in their active site and cleave their substrates at an aspartic acid residue. To date, twelve caspases have been identified in human (Taylor et al., 2008). Many of these are pivotal in regulating the progression of apoptosis. Some caspases appear not to participate in apoptosis but are involved in other cellular processes, such as caspase-1 in inflammation (Cerretti et al., 1992). Based on their roles during apoptosis, caspases can be further subdivided into two categories. Initiator caspases, including caspase-8 and -9, are mainly responsible for cleavage and activation of the effector caspases, caspase-3, -6 and -7. Effector caspases execute their roles by cleaving their specific substrates, which leads directly to the morphological and biochemical changes of apoptosis. Genetic knockout of individual caspases in mice showed that, unlike loss-offunction of CED-3 in C. elegans where a complete inhibition of apoptosis was observed, none of the caspase knockout mice displayed phenotypes of absolute prevention of apoptosis (Yuan, 2006; Degterev and Yuan, 2008). This could be due, at least in part, to 32

40 the presence of redundancies in the mammalian apoptosis system that are needed in order to tightly regulate this type of programmed cell death. Under some circumstances it appears that certain caspases may compensate for one another. For instance, withdrawal of nerve growth factor from cultures of caspase-2 null neurons led to an increase of the activity of caspase-9 compared to wild type neurons in the same setting (Troy et al., 2001). While genetic knockout of certain individual caspases do not show severe phenotypes, deletion of others in mice do generate severe developmental defects. For instance, deletion of caspase-8 is embryonically lethal and mice die from abnormal heart muscle development (Zheng et al., 1999). Under normal conditions, caspases exist in cells as inactive precursors with almost no protease activity. These inert caspases contain three domains: an N-terminal prodomain, a large subunit (P20), and a small subunit (P10). The latter two are also present in the respective mature caspase. Precursors are converted to active forms via proteolytic cleavage between these domains to generate a heterodimer consisting of two large subunits and two small subunits (Earnshaw et al., 1999). Caspases can be activated through three different mechanisms. 1). Cleavage by upstream caspases, such as caspase- 3, -6 and -7. 2). By the low intrinsic proteolytic activity of the precursor caspase itself when it is present at high local concentration, such as for the case of caspase-9. 3). With the assistance of a co-factor, for example, the requirement of APAF-1 and cytochrome c for activation of caspase-9 (Taylor et al., 2008). To date, nearly 400 substrates have been uncovered to be caspase targets (Thornberry and Lazebnik, 1998; Hengartner, 2000; Taylor et al., 2008; Lüthi and Martin, 2007). However, the significance of many of these cleavage events still unclear. 33

41 Nonetheless, a large body of evidence has accumulated to indicate that the destruction of many key substrates directly leads to biochemical and morphological changes during apoptosis to ultimately result in the disassembly of cells. 1. Examples of major well-studied caspase substrates during apoptosis a). DFF40 and DFF45. The biochemical hallmark of apoptosis is the appearance of genomic DNA fragmentation (Wyllie, 1980). The DNase responsible for this is DFF40. Normally, DFF40 is associated with its inhibitor, DFF45, to block its nuclease activity. During apoptosis, DFF45 becomes a target for caspase-3. Cleavage of DFF45 unleashes DFF40 to execute its nuclease activity (Liu et al., 1997; Enari et al., 1998). b). Poly-ADP-Ribose Polymerase (PARP). PARP is a DNA-binding enzyme that catalyses poly-adp-ribose ligation during DNA repair under moderate DNA damage. During apoptosis, the ability of PARP to repair DNA damage is abolished after its cleavage by caspase-3 (Bouchard et al., 2003). c). Fodrin. Fodrin is a ubiquitously expressed membrane-associated protein. It consists of!- and "-subunits and is important for maintaining normal membrane structure. It is a major target of caspases during apoptosis. Fodrin cleavage contributes to the morphological changes and destruction of cellular integrity (Wang et al., 1998). 34

42 C. BCL-2 Family The major role of BCL-2 family members is to regulate mitochondrial integrity within the intrinsic apoptosis pathway. This family was initially named after its founding member, the proto-oncogene BCL-2 (Tsujimoto et al., 1984). It was subsequently expanded and members in this family further categorized based on their anti- or proapoptotic functions (Reed, 1997; Gross et al., 1999; Hotchkiss et al., 2009). Proteins of the BCL-2 family contain from one to four conserved BCL-2 homology (BH) domains. Examples of anti-apoptotic BCL-2 family proteins are BCL-2 and BCL-X L. Examples of pro-apoptotic BCL-2 family proteins are BAX and BAK. Based on the number of BH domains, one subclass of pro-apoptotic BCL-2 proteins was identified as BH3-only proteins, with examples being BAD, BID and PUMA. Prior to death signal stimuli, anti-apoptotic BCL-2 members are often found residing in the mitochondrial and ER membrane, whereas pro-apoptotic members localize in the cytosol or associate with the cytoskeleton (Reed, 1997; Gross et al., 1999). After a death signal, BCL-2 family proteins undergo a series of changes including dimerization, posttranslational modification, and proteolytic cleavage that regulates their activity. The balance between pro- and anti-apoptotic BCL-2 family proteins and their cross-talk determines the ultimate fate of the cell. For example, a large portion of BAX resides in the cytosol as a monomer, but once triggered by death signals, it can form a homodimer or heterodimer with BCL-2 (Oltvai et al., 1993; Wolter et al., 1997; Gross et al., 1998). BAD can be phosphorylated by AKT upon cell stimulation by pro-survival signals. Phosphorylated BAD can bind with the adaptor protein in the cytosol to abrogate its pro-apoptotic activity (Zha et al., 1996; Datta et al., 1997). 35

43 D. Two Major Apoptosis Pathways Generally, apoptosis can be mediated by the intrinsic apoptosis pathway at the mitochondria through the action of BCL-2 family members, or by the extrinsic apoptosis pathway through the death receptors-mediated pathway (FIG. 6). Apoptosis can also proceed through other pathways. For example, free cholesterol can initiate the unfolded protein response through ER-mediated apoptosis, where it activates expression of C/EBP homologous protein (CHOP). CHOP is a transcription factor that down-regulates expression of the anti-apoptotic BCL-2 (Puthalakath et al., 2007) and depletes endogenous calcium stores (Feng et al., 2003). Granzyme B, a component of cytotoxic lymphocyte granules, can trigger apoptosis in the absence of caspase activity by directly cleaving DFF45 to initiate DNA fragmentation (Thomas et al., 2000). 36

44 Figure 6. Intrinsic and Extrinsic Apoptosis Pathways. (Modified from Hotchkiss et al., 2009). 1. The Intrinsic Apoptosis Pathway The intrinsic apoptosis pathway occurs as a response to DNA damage, elevated reactive oxygen species, and deprivation of growth factors (Hotchkiss et al., 2009). This pathway resembles apoptosis as described in C. elegans. BCL-2 family members are the 37

45 major regulators controlling this apoptosis pathway. Battle between pro-apoptotic and anti-apoptotic BCL-2 family members at the outer mitochondrial membrane tightly controls the intrinsic apoptosis pathway (Gross et al., 1999). There are two mechanisms for action of BCL-2 proteins in regulation of mitochondria-mediated apoptosis. 1). They can interact with pre-existent outer-membrane channel proteins to affect mitochondrial membrane potential as in the case of the voltage-dependent anion channel (VDAC). For example, use of a system based on reconstitution of VDAC in liposomes showed that pro-apoptotic BAX and BAK could trigger the opening of this channel, whereas antiapoptotic BCL-X L leads to closing of this channel (Shimizu et al., 1999). 2). Oligomerization of BCL-2 family proteins themselves results in channel formation in the outer mitochondrial membrane. The first clue regarding this model was the resolution of a crystal structure for BCL-X L. This suggested that the!-helical arrangement of BCL-X L resembled the pore-forming subunit of diphtheria toxin (Muchmore et al., 1996). It was later found that BCL-X L (Minn et al., 1997), BCL-2 (Schendel et al., 1997), and BAX (Antonsson et al., 1997) each had the ability to form such a channel, and that this could cause disruption of mitochondrial membrane potential. Research from Xiaodong Wang s group further characterized the downstream events after the battle among BCL-2 families members at the outer mitochondrial membrane. By using a cell-free system based on cytosol from HeLa cells to examine factors required for the generation of biochemical markers of apoptosis, they successfully isolated cytochrome c, initially designated as APAF-2, as one of the most important proteins released after permeabilization of the mitochondrial membrane (Liu et al., 1996). Cytochrome c and APAF-1, together with procaspase-9, form the apoptosome which 38

46 leads to activation of caspase-9 (Li et al., 1997; Zou et al., 1997, 1999). Cytochrome c is not the only important protein released from mitochondria during apoptosis, other factors include second mitochondria-derived activator of caspase (Smac) (Du et al., 2000; Wu et al., 2000). Smac s ability to promote caspase-9 activation rests upon its binding to endogenous caspases inhibitors, known as the inhibitor of apoptosis proteins (IAPs) to block their anti-apoptotic activity. One thing we need to bear in mind is that while in many cases, these two apoptotic pathways are generally independent of each other, crosstalk does happen. One of the most critical message proteins connecting these pathways is BID, a BH3 only BCL-2 family protein. Once the extrinsic apoptosis pathway is activated, BID can be cleaved by caspase-8, which generates a truncated BID (tbid). tbid then translocates to mitochondria to induce the intrinsic apoptosis pathway (Li et al., 1998). Therefore, tbid is an important mediator to manage the cross-talk between intrinsic and extrinsic apoptosis pathways. 2. The Extrinsic Apoptosis Pathway Mammalian cells also possess an extrinsic apoptosis pathway, which can be elicited by Fas ligand and TNF-! (Itoh et al., 1991). Binding of these ligands to death receptors including CD95, DR3, TNF-R1, and two TRAIL receptors leads to the assembly of the death-inducing signaling complex (DISC). DISC is composed of the death receptor, its activator Fas-associated protein with death domain (FADD), procaspase-8 and its regulator c-flip. Interaction of death receptors and FADD is mediated through their death domain (DD) (Chinnaiyan et al., 1995; Kischkel et al., 1995). Assembly of DISC 39

47 induces a conformational change of procaspase-8, which leads to the cleavage of procaspase-8 through self-proteolytic activity (Hotchkiss et al., 2009; Strasser et al., 2009). Active caspase-8 is then used to cleave downstream effector caspases to induce the apoptotic cascade. Cellular flice inhibitory protein (c-flip), a unit in DISC, is also crucial in controlling caspase-8 activity. c-flip structurally resembles caspase-8 but without enzymatic activity (Irmler et al., 1997; Inohara et al., 1997). At high concentration, it serves as an inhibitor of caspase-8 by reducing its proteolytic activity, whereas at low concentration it triggers caspase-8 activity (Maedler et al., 2002). III. FSP27 IN LIPID DROPLET FORMATION AND APOPTOSIS A. Overview of CIDE Family FSP27, also called CIDEC, stands for fat-specific protein 27. This designation is based on its exclusive expression in adipose tissue and its molecular weight of 27 kda. It is one of three members of the Cell Death Inducing DFF45-Like Effector (CIDE) protein family, which also includes CIDEA and CIDEB. This protein family was first discovered based on a region of shared amino acid sequence homologous to the CIDE-N domain of DFF45. The three CIDE proteins also share a conserved CIDE-C domain (Inohara et al., 1998), a domain not present in any other protein (FIG. 7). In addition to sequence similarities, CIDE family proteins share several other features: 1.) They all promote apoptosis upon ectopic expression in mammalian cells; 2.) Each has the ability to localize to lipid droplets; and 3.) Knockout mouse models and other studies have revealed each is a critical regulator of energy homeostasis. The CIDE-N domain of CIDE family is wellconserved from lower organisms to mammals, but the CIDE-C domain is only present in 40

48 vertebrates (Wu et al., 2008). It might be speculated that during evolution, the CIDE-C domain was developed to control energy balance, which requires sophisticated regulation in higher organisms. Figure 7. Schematic Structure of Human CIDE Family Members. The CIDE family consists of three members, FSP27/CIDEC, CIDE-A and CIDE-B. They all contain a CIDE-N domain (Red) similar to DFF45 (DNA-Fragmentation Factor 45) and also a highly conserved CIDE-C domain. Blue box indicates region I found to be important for the lipid droplet localization ability and apoptotic activity. Each CIDE gene shows highly tissue-restricted expression in mice and humans. In murine tissues, FSP27 is expressed exclusively in WAT and BAT and is nearly undetectable in all other tissues examined (Kim et al., 2008). In humans FSP27 is also 41

49 predominantly expressed in WAT with trace levels in colon, heart, lung and skeletal muscle, when assessed by real-time PCR (Magnusson et al., 2008). FSP27 expression has not yet been examined for human BAT. CIDEA is expressed exclusively in BAT of mice (Zhou et al., 2003; Kim et al., 2008) and in WAT in human (Nordström et al., 2005); CIDEA expression in human BAT has not been examined. Intriguingly, FSP27 and CIDEA expression could also be detected in fatty liver versus the normal state which is negative for FSP27 (Kim et al., 2008; Toh et al., 2008). Unlike CIDEA and FSP27, CIDEB is not found in adipose tissue (Kim et al, 2008). In mice, CIDEB is expressed predominantly in liver and is also readily detected in small intestine and kidney (Kim et al, 2008). In humans, CIDEB also shows high enrichment to liver (Inohara et al., 1998). Overall, the highly restricted expression patterns of CIDEs suggested that these genes might have unique tissue-specific roles. The first known function demonstrated for CIDEs was promotion of apoptosis upon their ectopic expression in mammalian cells (Inohara et al., 1998; Chen et al., 2000; Erdtmann et al., 2003; Liang et al., 2003; Kim et al., 2008; Liu et al., 2009). In addition to their apoptotic activity, CIDE family members were rather recently demonstrated to be lipid droplet-associated proteins whose ectopic expression results in formation of lipid droplets when exogenous fatty acid is also present culture media. FSP27, in particular, has been demonstrated to be required for formation of the unilocular lipid droplet typical of white adipocytes in vivo (Nishino et al., 2008; Puri et al., 2008a; Toh et al., 2008; Liu et al., 2009). Recently, there are also reports from one group showing that CIDE family members also show localization to the ER (Qi et al, 2008; Ye et al., 2009; Nian et al., 2010). Given that the lipid droplet is believed to derive from ER, this is not unexpected. 42

50 The relative levels CIDE protein in the ER to that in lipid droplets has only been examined for CIDEB. Western blot analysis of subcellular fractions reveals that CIDEB is highly enriched in lipid droplets as compared to level in ER (Ye et al., 2009). It is currently thought that lipid droplets are the major localization site for all CIDE proteins. Although each of the three CIDE proteins were reported to have mitochondrial localization, it has since been acknowledged this was incorrect and attributed to artifacts of cell morphology occurring during later stages of CIDE-induced apoptosis (Chen et al., 2000; Zhou et al., 2003). B. Apoptosis by CIDE Family Members Since the CIDE family was first identified based on the presence of a conserved CIDE-N domain homology to DFF45, early focus was mainly on their pro-apoptotic activity. CIDEB was the first of the three CIDE proteins characterized in detail in respect to its apoptotic activity. CIDEB was shown to be a pro-apoptotic protein when ectopically expressed in COS and 293T cells. It was determined that a subregion of its CIDE-C domain was critical to mediate its apoptotic activity and self-dimerization (Chen at al., 2000) Interestingly, through a yeast two hybrid screening using a human liver cdna library, CIDEB was discovered to be an interaction partner of hepatitis C virus (HCV) nonstructural 2 (NS2) protein (Erdtmann et al., 2003). This interaction was mediated via the CIDE-C domain of CIDEB. Interaction of NS2 with CIDEB significantly attenuated the pro-apoptotic activity of CIDEB. CIDEA is also a proapoptotic protein, shown to induce robust apoptosis when expressed in 293T cells. Its pro-apoptotic activity maps to its CIDE-C domain (Inohara et al., 1998). 43

51 Like the other two CIDE family members, FSP27 is pro-apoptotic when expressed ectopically as observed by morphological change, caspase activation and DNA fragmentation (Liang et al., 2003; Keller et al., 2008; Kim et al., 2008; Liu et al., 2009). As presented here, my studies have resulted in FSP27 currently being the most wellcharacterized of any CIDE protein in regard to apoptosis. The pro-apoptotic activity of FSP27 is mainly manifested the through the intrinsic apoptosis pathway. Ectopic expression of FSP27 by transient transfection of mammalian cells resulted in release of cytochrome c from mitochondria. This involved generation of active caspase-9. FSP27-induced cell death could be blocked by co-transfection of a dominant-negative caspase-9 and by a pan-caspase inhibitor (Liu et al., 2009). We further determined that the FSP27 CIDE-C domain is nearly as effective as full-length FSP27 in its ability to induce the intrinsic apoptosis pathway. This apoptosis required amino acids 173 to 191 of the FSP27 CIDE-C domain. However, at this time, I have yet to determine the upstream trigger of FSP27 action. C. CIDE Proteins and Lipid Metabolism FSP27 was first identified as a protein whose expression is induced dramatically during adipocyte differentiation of the TA1 preadipocyte cell culture (Danesch et al., 1992). It was later categorized as one member of CIDE family (Inohara et al., 1998). In the last several years, FSP27 and two other CIDE proteins have been demonstrated to have major roles in energy metabolism. This has been achieved through biochemical, cell culture and genetic studies of CIDE knockout mice as well limited investigation in humans. 44

52 1. CIDEA In vitro studies showed that CIDEA localizes to lipid droplets in lipid-loaded HeLa cells (Liu et al., 2009) and that ectopic expression of CIDEA led to an increase in lipid droplet size in COS and 3T3-L1 preadipocytes (Puri et al., 2008a). CIDEA knockout mice were lean and resistant to high-fat diet-induced obesity which was reported to be a consequence of enhanced thermogenesis and elevated lipolysis in BAT (Zhou et al., 2003). Enhanced thermogenesis was explained by data showing that CIDEA coimmunoprecipitated with UCP1. Interaction of CIDEA with UCP1 was postulated to suppress UCP1 uncoupling action. However, these results need to be re-evaluated since CIDEA was later proved to be a protein localized to lipid droplets and ER, but not to mitochondria where UCP1 resides. A follow-up study from Li and colleagues provided a revised molecular mechanism for the lean phenotype of CIDEA -/- mice (Qi et al, 2008). AMPK has been demonstrated to be an important sensor for energy homeostasis. It can inhibit fatty acid synthesis by phosphorylation of acetyl-coa carboxylase (ACC) (Kahn et al., 2005). Li and colleagues demonstrated that CIDEA interacted directly with the "-subunit of AMP-activated protein kinase (AMPK) at the ER. As CIDEA is a very short-lived protein subject to rapid proteasome-mediated degradation, interaction of CIDEA and AMPK results in AMPK undergoing proteasome-mediated degradation. Therefore, the mechanism whereby CIDEA -/- mice resist high-fat diet-induced obesity is currently attributed to an elevated level of AMPK, leading to enhanced energy expenditure. Although both CIDEA and FSP27 knockout mice demonstrate a lean phenotype, it has not yet been reported whether 45

53 FSP27 can also affect AMPK level or activity. 2. CIDEB In vitro studies revealed that CIDEB localized in lipid droplets in lipid-loaded HeLa cells (Liu et al., 2009). CIDEB -/- mice were resistant to high-fat diet-induced obesity and had increased insulin sensitivity. CIDEB is not expressed in adipose tissue but is highly enriched in liver, indicating the effects of CIDEB knockout extent beyond its normal liver expression site. CIDEB deficiency prevented the liver steatosis that is normally observed in WT mice on a high-fat diet (Li et al., 2007). A follow-up study by the same group observed that CIDEB deficiency caused a defect in VLDL secretion from liver as a result of failed lipid incorporation into VLDL, which led to blocked VLDL maturation. They also showed that CIDEB interacted with apolipoprotein B, and that defects of lipid insertion into VLDL were due to lack of this interaction in CIDEB -/- mice (Ye et al., 2009). Given that CIDEB is also expressed in intestine, it is also a possible that CIDEB can affect TG packaging in intestine, and subsequent energy availability from diet. This may account for the lean phenotype of CIDEB -/- mice and their resistance to high-fat diet induced obesity. 3. FSP27 a. FSP27 in Lipid Droplet Formation and Function 1. Localization of FSP27 to Lipid Droplets The first hint that FSP27 may be a lipid droplet-associated protein came from a proteomic study designed to identify lipid droplet-associated proteins in basal and 46

54 lipolytically-stimulated 3T3-L1 adipocytes (Brasaemle et al, 2004). However, due to the previous, now clearly incorrect, reports indicating mitochondrial localization for the two other CIDE family members (Chen et al., 2000; Zhou et al., 2003), it was assumed detection of FSP27 in the lipid droplet proteome was due to mitochondrial contamination of the lipid droplet preparations used in this study. Thus, for several years thereafter the potential for FSP27 to function in lipid droplet biology was overlooked. FSP27 was subsequently found to be a protein that regulates lipid droplet formation during an RNAi screen in cultured adipocytes (Puri et al., 2008b). Ectopic expression of an FSP27-GFP construct in 3T3-L1 or 293T cells showed that it surrounded lipid droplets, overlapping with the signal of the well-characterized lipid droplet-associated protein, perilipin (Puri et al., 2007; Keller et al., 2008). This lipid droplet-associated localization was also verified for endogenous FSP27 in 3T3-L1 cells. I further characterized the region of FSP27 that is critical for lipid droplet localization using oleic acid loaded HeLa cells. I demonstrated that amino acids 173 to 191 of the CIDE-C domain of FSP27 was not only required for its pro-apoptotic activity, but also for the lipid droplet targeting of FSP27 (Liu et al., 2009). Interestingly, like FSP27, there is also a similar subregion in the CIDE-C domain of CIDEB that is important for its lipid droplet localization and apoptotic activity (Chen et al., 2000; Ye et al., 2009). I also discovered that the addition of oleic acid to culture media significantly attenuated the pro-apoptotic activity of FSP27 in HeLa cells (Liu et al., 2009). This suggests that by promoting its lipid droplet localization, such sequestration could shield FSP27 from triggering apoptosis. This could be a clue as to why normally lipid-laden 47

55 adipocytes can express high levels of FSP27 but are not responsive to its apoptotic effects. 2. Role of FSP27 in Lipid Droplet Enlargement Not only does FSP27 localize to lipid droplets, but it also controls lipid droplet size. I showed that ectopic expression of FSP27 led to formation of enlarged lipid droplets across all cell lines examined, under culture conditions of oleic acid supplementation (Liu et al., 2009). This was also observed by others for several cell lines (Keller et al., 2008; Puri et al., 2008). For example, FSP27 overexpression led to enlargement of the multilocular lipid droplets of in vitro differentiated brown adipocytes. In regard to effects of FSP27 in preadipocytes and their differentiation, MacDougald showed that ectopic expression of FSP27 in 3T3-L1 preadipocytes resulted in lipid droplet formation. In the preadipocyte/adipocyte lineage, expression of FSP27 normally only occurs during the process of adipogenesis. MacDougald determined whether ectopic expression of FSP27 in preadipocytes, could alone lead to induction of the adipogenic program in such cells (Keller et al., 2008). However, there was no increase in the expression of typical white adipocyte marker genes such as FABP4, PPAR! and CEBP" in FSP27-expressing 3T3- L1 preadipocytes (Keller et al., 2008). Together, these data indicate that the mechanism(s) employed by FSP27 to promote lipid droplet formation did not require adipocyte specific proteins, nor was the expression of FSP27 per se sufficient to promote adipogenesis. Knockdown studies of FSP27 further confirmed its key role in regulating lipid droplet size. Transient knockdown of FSP27 expression in 3T3-L1 adipocytes in vitro led 48

56 to the formation of multiple smaller lipid droplets reminiscent of brown adipocytes. Curiously, these cells also contained increased numbers of mitochondria, typical of brown adipocytes. As such, several brown adipocyte markers were examined in these cells, for example uncoupling protein 1 (UCP1), but induction of brown adipocyte transcripts were not observed (Keller et al., 2008). Therefore, in this in vitro model, lack of FSP27 resulted in more, smaller lipid droplets as well as apparent mitochondrial biogenesis, but did not appear to alter the fundamental identity of cells, i.e. transdifferentiation of white adipocytes to brown. Studies in FSP27 -/- knockout mice also showed smaller lipid droplets and increased mitochondria in WAT. Unlike wild type in vivo white adipocytes, which are typified by a large unilocular lipid droplet, FSP27 -/- white adipocytes contained multiple smaller lipid droplets and increased mitochondria. This was consistent with the in vitro finding that FSP27 is essential to maintaining lipid droplet size (Puri et al., 2007). Intriguingly, WAT derived from FSP27 -/- mice displayed a brownish color similar to BAT. Furthermore, in this animal model, there was a drastic increase in the expression of UCP1, peroxisome proliferator-activated receptor # coactivator-1! (PGC1") and deiodinase, iodothyronine, type II (DIO2) in WAT. These genes are normally specifically enriched to BAT versus all other tissues (Gesta et al., 2007). Thus, unlike the in vitro FSP27 knockdown studies, data from FSP27 -/- WAT suggested a degree of transcriptional transdifferentiation. This implied that loss of FSP27 might promote certain signaling pathways resulting in WAT acquiring properties that are fundamental to a brown adipocyte identity. Retinoblastoma (RB), retinoblastoma-like 1 (p107) and nuclear receptor interacting protein 1 (RIP140), are transcription factors or co-repressors 49

57 of a plethora of nuclear receptors; their expression can negatively regulate BAT differentiation. These three transcription factors were all down-regulated in FSP27 -/- WAT, providing more support for the transdifferentiation hypothesis. It is curious that the effects of lack of FSP27 clearly differed for in vivo compared to in vitro settings. In vivo, knockout of FSP27 led to expression of brown adipocyte associated genes, suggesting a white to brown adipocyte transdifferentiation. However, a similar phenotype was not observed for in vitro cell culture studies. This might be attributed to the much more complex cell-to-cell or cell-to-extracellular matrix interactions in vivo, which can not be reflected by cell culture, or degree of knockdown achieved. If indeed, lack of FSP27 promotes a brown adipocyte-like phenotype and/or true transdifferentiation in white adipocytes, understanding the details of FSP27 function would be anticipated to have a dramatic impact on the development of approaches to control fat storage and obesity. This is because unlike white adipocytes, the major function of brown adipocytes is to promote thermogenesis which results in energy consumption rather than storage (Gesta et al., 2007). b. Role of FSP27 in Energy Metabolism 1. Lipolysis Knockout or knockdown of FSP27 leads to the morphological change of striking lipid droplet fragmentation. Hence, physically, this can increase the net lipid droplet surface to volume ratio in respect to TG stores. This alone would be predicted to increase lipase access per volume of TG. Thus, for adipocytes, the size of lipid droplets is closely correlated to their lipolytic state. The lipolytic state of lipid droplets is also further 50

58 impacted by differential protein associations that occur as the result of signals of hormone-stimulated lipolysis, such as HSL docking with perilipin at the lipid droplet surface. Thus, for lipid droplets morphology and function are closely tied. This is exemplified by the observation that knockdown of FSP27 impacts TG lipolysis rates. The roles of FSP27 in basal and hormone stimulated lipolysis have been investigated by using white adipocytes from FSP27 -/- mice, or in vitro differentiated 3T3-L1 and HW white adipocytes (Puri et al., 2007; Keller et al., 2008; Nishino et al., 2008; Toh et al., 2008). Analysis of white adipocytes from two independently generated FSP27 -/- mouse models revealed increased basal lipolysis in null compared to wild type (Nishino et al., 2008; Toh et al., 2008). For hormone-stimulated lipolysis, different results were found for each mouse model. One group showed decreased lipolysis for FSP27 null (Nishino et al., 2008), while no significant change was observed by the other investigators (Toh et al., 2008). This difference in hormone-stimulated lipolysis could be due to the different dose of isoproterenol used for activation of "-adrenergic receptors, with effects shown for 10 &M (Nishino et al., 2008) but not for 1 &M (Toh et al., 2008). Data from in vitro knockdown of FSP27 in white adipocytes is also somewhat contradictory for effects on lipolysis. In HW adipocytes, knockdown of FSP27 led to an increase in both basal and hormone-stimulated lipolysis (Nishino et al., 2008). A similar increase was found by Czech for FSP27 knockdown in 3T3-L1 adipocytes (Puri et al., 2007). However, knockdown of FSP27 in 3T3-L1 adipocytes by MacDougald failed to show effects on either basal or hormone-stimulated lipolysis (Keller et al., 2008). The different data from these in vitro studies may due to the residual level of FSP27 after gene knockdown or the approach employed. MacDougald established permanent 3T3-L1 51

59 cell culture knockdown for FSP27, while the others used a transient transfection of sirna for knockdown. However, taken together, the majority of the in vivo and in vitro data indicate that decreasing FSP27 level enhances basal lipolysis. This suggests that one mechanism whereby FSP27 acts to enlarge lipid droplets could be by protection from basal lipolysis. Although FSP27 and perilipin have been both shown to colocalize at the lipid droplet surface, evidence to date indicates that the mechanism utilized by FSP27 to impact lipolysis is different from that used by perilipin. Knockout of FSP27 led to formation of multilocular lipid droplets in white adipocytes, while perilipin null mice still have unilocular lipid droplets (Nishino et al., 2008). Furthermore, as previously described, the effects on depletion of perilipin on basal and hormone-stimulated lipolysis are clearly distinct from that for FSP27. Overall, further investigation is needed to determine the roles of FSP27 in both basal and hormone-stimulated lipolysis. 2. Integration of Energy Metabolism by FSP27 Manipulation of FSP27 level not only leads to changes in the size and number of lipid droplets, but also has consequences in regard to whole-body energy homeostasis. As previously mentioned, FSP27 -/- mice showed reduced WAT, with white adipocytes taking on features of brown adipocytes. These mice demonstrated enhanced insulin sensitivity and were resistant to diet-induced obesity. Furthermore, this anti-obesity effect of FSP27 null has also been observed on background of the obese ob/ob mouse model of genetic obesity (Toh et al., 2008). In most cases, the predicted consequence of reduction of WAT mass is increased fatty acids in circulation. This is the result of a reduced capacity to store fatty acid as TG 52

60 in the face of limited or reduced WAT mass. This can subsequently result in insulin resistance and presence of ectopic accumulation of lipids in other tissues such as in liver and heart, and other aspects of lipotoxicity. The phenotype of FSP27 -/- mice is unique in that despite a drastically reduced WAT mass, these mice nonetheless show enhanced insulin sensitivity. This was likely due to an increased level of "-oxidation within white adipocytes due to the transdifferentiation and acquirement of brown adipocyte-like features in FSP27 -/- mice. For example, these white adipocytes generated more mitochondria which showed expression of the thermogenic protein UCP1. This enabled FSP27 -/- mice to release energy as heat rather generate ATP during "-oxidation of fatty acids. The WAT of FSP27 -/- mice evidenced elevated expression of nuclear respiratory factor 1 (NRF1), mitochondrial transcription factor A (mttfa), cytochrome c oxidase subunit IV (COXIV), and many other proteins whose function is integral to mitochondrial biogenesis, fatty acid oxidation and electron chain transport. Consequently, FSP27 -/- mice showed an increased metabolic rate that led to considerable energy consumption, which could decrease the amount of fatty acid released into circulation. This spared such animals the lipotoxicity that usually ensues as a result of diminished mass of WAT. There was no evidence of ectopic lipid accumulation in any non-adipose organs (Nishino et al., 2008; Toh et al., 2008). FSP27 -/- mice had increased glucose uptake and improved insulin sensitivity compared to wild type mice. This also likely stemmed from altered metabolism as a result of decreased fatty acid availability. Western blot analysis of WAT from FSP27 -/- and wild type mice showed significantly increased expression of glucose transporter type 4 (GLUT4), V-Akt murine thymoma viral oncogene homolog 2 (AKT2), and enhanced 53

61 tyrosine phosphorylation of insulin receptor substrate 1 (IRS-1) in FSP27 -/- WAT, all of which play important roles in regulating insulin sensitivity (Kahn and Flier, 2000). Surprisingly, BAT in FSP27 -/- mice contained larger, unilocular, lipid droplets, rather than the typical multilocular lipid droplets (Nishino et al., 2008; Toh et al., 2008). Their BAT also evidenced lower mitochondrial activity and a reduced level of mitochondrial marker proteins. However, the mechanisms underlying these effects on BAT for FSP27 -/- still need to be elucidated. It is important to recall that FSP27 is also normally expressed in murine BAT (Kim et al., 2008). It is not yet clear, therefore, which of these BAT effects are due to lack of FSP27 in BAT per se, or a secondary consequence of lack of FSP27 in WAT. There is the possibility that enhanced thermogenic activity of WAT may inhibit this activity in BAT. Another possibility is that as white adipocytes of FSP27 -/- can only store a very reduced amount of TG, the excess TG is stored in brown adipocytes. c. FSP27 and Metabolic Diseases 1.) FSP27 in Human Lipodystrophy Lipodystrophy is a disease characterized by adipose tissue deficiency (Garg, 2004). Very recently a report suggested that a mutation in FSP27 could cause a form of human lipodystrophy (Rubio et al., 2009). This individual had a homozygous mutation of FSP27 within its CIDE-C domain which led to a premature stop codon. This is predicted to produce an FSP27 protein truncated midway through its CIDE-C domain. Whether or not this patient harbored this truncated FSP27 was not determined. However, a genetically engineered version of this mutant truncated form of FSP27 could not localize to lipid 54

62 droplets upon ectopic expression in vitro. No other FSP27 mutations were found in an additional 168 lipodystrophy patients. Nonetheless, this is still an exciting discovery in that my studies have shown that the CIDE-C domain is absolutely required for the lipid droplet localization of FSP27 (Liu et al., 2009), and this patient has a mutation within the 19 amino acid region I determined as critical for lipid droplet localization of FSP27 (Liu et al., 2009). 2.) FSP27 in Human Obesity and Insulin Sensitivity Consistent with the role of FSP27 in directing the formation of liver steatosis in rodents (Matsusue et al., 2008), data from extremely obese patients indicated that hepatic FSP27 expression level was positively correlated with body weight, and there was a profound decrease of FSP27 expression level, but not CIDEA and CIDEB, in the liver of these subjects one year after gastric bypass surgery (Hall et al., 2010). FSP27 enhances fatty acid incorporation into LDs, and appears to protect TG from basal lipolysis. Therefore, it might be predicted that individuals with increased FSP27 levels in WAT would evidence decreased levels of free fatty acids in their circulation and thereby not suffer the same degree of lipotoxic as those with lower FSP27 levels. In humans, increased levels of circulating free fatty acids is frequently positively correlated with insulin resistance. Along these lines, Puri et al (2008a) found that the level of FSP27 transcript expressed in WAT, as well as that for CIDEA, was positively correlated with insulin sensitivity in obese human patients. This suggests that individuals with increased WAT levels of FSP27 may be more effective in maintaining TG storage in WAT, perhaps due to the lipid droplet enlargement function of FSP27. 55

63 3.) FSP27 in Fatty Liver Disease FSP27 is not normally expressed in liver, but its expression emerges under conditions of high energy excess as liver steatosis develops. The mechanism for this has been addressed using genetically manipulated mouse models. FSP27 is expressed at high levels in liver of ob/ob mice (Kim et al., 2008; Matsusue et al., 2008). PPAR#, a master regulator of adipogenesis (Evans et al., 2004) has a critical role in the development of hepatic steatosis in ob/ob mice (Matsusue et al., 2003). Further investigation indicated that FSP27 is a direct transcriptional target of PPAR# in this process. Tissue-specific knockout of PPAR! in ob/ob liver nearly abolished the expression of FSP27 transcript levels (Matsusue et al., 2008). Overexpression of FSP27 in PPAR! -/- hepatocytes increased their TG both in vitro and in vivo, whereas knockdown of FSP27 in ob/ob hepatocytes in vivo by shrna lowered their TG level and reduced the number and size of fatty vacuoles in PPAR! -/- ob/ob liver. These data provide convincing evidence that FSP27 is a pivotal PPAR#-regulated mediator for the development of hepatic steatosis. d. Regulation of FSP27 Expression 1. Transcriptional Regulation of FSP27 Expression As just described, studies in the liver steatosis setting showed that FSP27 is a direct target of PPAR#. In these studies, it was shown that PPAR# was able to bind a peroxisome proliferator response element (PPRE) in the promoter of FSP27, which was confirmed by Chromatin immunoprecipitation (ChIP) assay. These data are consistent 56

64 with a previous finding that PPAR# could transactivate the FSP27 promoter (Yu et al., 2003). In respect to regulation of FSP27 in white adipocytes, depletion of PPAR! led to a more than 90% corresponding decrease of FSP27 mrna expression in 3T3-L1 adipocytes (Puri et al., 2008a). In white adipocytes, FSP27 expression is also regulated by CCAAT/enhancer binding protein (C/EBP) (Danesch et al., 1992), a family of key transcription regulators for adipogenesis (Farmer, 2006). This was shown by interaction of C/EBP! at a response element in the FSP27 promoter (Danesch et al., 1992). Kim et al showed that FSP27 mrna is positively regulated by insulin, a key lipogenic hormone, and that this may involve the PI-3 kinase pathway. Exposure of 3T3-L1 adipocytes to the PI-3 kinase inhibitor, LY294002, greatly attenuated the positive regulation of FSP27 transcript by insulin. On the other hand, the lipolytic agent TNF-! reduced FSP27 transcript expression levels in 3T3-L1 adipocytes (Kim et al., 2008). A very recent study further delineated the relationship of insulin and FSP27 in lipid droplet formation (Ito et al., 2010). Knocking down FSP27 in human adipocytes completely abrogated the lipid droplet-enlargement effects of insulin typically seen. 2. Post-Transcriptional Regulation of FSP27 Expression As in the case for CIDEA (Chan et al., 2007), it has recently been shown that FSP27 protein expression can be controlled at the post-translational level through the ubiquitinmediated proteasome degradation pathway. This was addressed for endogenous FSP27 protein in 3T3-L1 adipocytes as well as through use of ectopic transfection of FSP27 in 293T cells (Nian et al., 2010). The addition of the proteasome inhibitors MG132 or epoxomicin delayed turnover of FSP27. Further investigation showed that three lysine 57

65 residues in the C-terminal region of FSP27 were critical for its stability. Site directed mutagenesis of these to alanine increased the half-life of FSP27 from ~30 minutes to more than 2 hours, and decreased the level of poly-ubiquitinated FSP27 protein. Oleic acid supplementation of cultures provides an external free fatty acid source for the synthesis of TG and therefore promotes lipid droplet formation. Interestingly, when oleic acid was added to culture medium, FSP27 protein half-life was significantly prolonged, with no transcriptional effect noted. This suggests that FSP27 association at lipid droplets protects it from ubiquitin-mediated degradation. An additional piece of evidence supporting this idea is that a dramatic increase in the half-life of FSP27 was found in cells treated with the "-adrenergic receptor agonist isoproterenol (Nian et al., 2010). 3T3-L1 adipocytes treated with isoproterenol undergo a rapid fission of their lipid droplets to produce many smaller ones. The increase in the overall surface area of lipid droplets in these isoproterenol treated cells presumably provided more sites for FSP27 association. The lipid droplet environment likely resulted in a more effective protection of FSP27 from proteosome-mediated degradation. IV. Summary In summary, the research on the function of FSP27 reported by others and by us clearly indicates that FSP27 is an important protein with dual functions in regulating lipid droplet metabolism and apoptosis. 58

66 MANUSCRIPT #1 Assessment of Fat Specific Protein 27 (FSP27) in the Adipocyte Lineage Suggests a Dual Role for FSP27 in Adipocyte Metabolism and Cell Death Ji Young Kim +, Kun Liu +, Shengli Zhou, Kristin Tillison, Yu Wu and Cynthia M. Smas* Department of Biochemistry and Cancer Biology, University of Toledo Health Science Campus, Toledo, OH USA Am J Physiol Endocrinol Metab 294: E , These authors contributed equally to the work. Running Head: Expression of FSP27 in the Adipocyte Lineage Please address reprint requests to: Cynthia M. Smas, D.Sc. (* Corresponding author) Department of Biochemistry and Cancer Biology, University of Toledo Health Science Campus, Toledo, OH USA Phone: FAX: cynthia.smas@utoledo.edu 59

67 Abstract Fat Specific Protein 27 (FSP27)/Cidec was initially identified by its upregulation in TA1 adipogenesis and is one of three CIDE family proapoptotic proteins. Ectopic expression of CIDEs promotes apoptosis of mammalian cells. On the other hand, FSP27 has very recently been illustrated to regulate lipid droplet size and promote lipid storage in adipocytes. Regulation of endogenous FSP27 expression is unknown. We assessed FSP27 transcript level in the well-characterized 3T3-L1 in vitro adipocyte differentiation model and found its emergence parallels the adipocyte-enriched transcripts afabp and SCD1. Furthermore, FSP27 is a differentiation-dependent transcript in adipogenesis of primary rodent and human preadipocytes and in brown adipogenesis. FSP27 transcript is inversely regulated by TNF! and insulin; consistent with an antilipolytic function. It is nearly abolished with a 4 h exposure of 3T3-L1 adipocytes to 10 ng/ml TNF!, while treatment with 100 nm insulin increased FSP27 transcript 8-fold; LY blocked this response indicating involvement of PI3-kinase signals. Northern blot analysis of murine tissues indicated exclusive expression of FSP27 in WAT and BAT; however a dramatic upregulation occurred in liver of ob/ob mice. Ectopic expression of murine FSP27 in 293T cells and in 3T3-L1 preadipocytes led to the appearance of key apoptotic hallmarks and cell death. However, despite the upregulation for FSP27 in adipogenesis we failed to detect DNA laddering indicative of apoptosis in 3T3-L1 adipocytes. This suggests that adipogenesis is accompanied by decreased susceptibility to the proapoptotic effects of FSP27. Overall, our findings support roles for FSP27 in cell death and in adipocyte function. 60

68 Key Words: Insulin, TNFa, adipogenesis, FSP27, CIDE, apoptosis, 3T3-L1 61

69 Introduction The primary metabolic role of white adipocytes is the storage of excess energy as triglyceride and its mobilization to meet energy needs. White adipose tissue (WAT) is also an endocrine organ that synthesizes and secretes of a number of soluble factors, some of which are adipocyte-derived such as leptin, resistin and TNF! (4, 9, 19, 23). Mature adipocytes arise via differentiation of adipocyte precursors present in adipose tissue (3, 12, 16, 17, 32). For the past several decades, in vitro preadipocyte cell lines, primarily 3T3-L1, have been extensively used to define genes central to the adipocyte phenotype and adipogenesis (15, 17). Adipogenesis is accompanied by upregulation for genes encoding proteins critical for lipogenesis, lipolysis, lipid transport, insulin sensitivity, and hormone signaling and other adipocyte functions (41). A variety of in vitro and in vivo studies have determined that peroxisome proliferator-activated receptor # (PPAR#), a member of the ligand activated steroid hormone receptor family, is a master transcriptional regulator of the adipogenic program (14, 33, 43-46, 60, 61). The important contribution of the CCAAT/enhancer-binding protein (C/EBP) family of transcriptional regulators to adipogenesis has also been firmly established (2, 42, 55, 60, 61). Fat Specific Protein 27 (FSP27) was cloned by Ringold and coworkers in 1984 via a differential screening approach to identify cdnas with differentiation-dependent upregulation during in vitro adipogenesis of murine TA1 preadipocytes (6) and in 1992 they demonstrated that the FSP27 promoter binds C/EBP!, confers adipocyte differentiation-dependent expression on a heterologous chloramphenicol acetyl transferase (CAT) reporter gene, and was repressed by TNF! (11, 59). Until recently, 62

70 FSP27 was regarded soley as an adipocyte marker gene of unknown function. However, during revision of this manuscript, a study of FSP27 localization and function in 3T3-L1 adipocytes reported that FSP27 is a lipid-droplet associated protein that promotes triglyceride deposition in adipocytes by inhibiting lipolysis (39); as such it appears to a major new modulator of lipid droplet function that is required for optimal triglyceride storage in adipocytes (39). Knockdown of FSP27 in 3T3-L1 adipocytes was reported to result in the fragmentation of large lipid droplets to produce many small lipid droplets and ectopic expression of FSP27 promoted lipid droplet formation when assessed in 3T3- L1 preadipocytes, COS cells and CHO cells (39). A comprehensive assessment of the expression and the regulation of FSP27 in adipogenesis remains to be investigated. While not yet extensively examined, the human transcript appears to have a somewhat wider expression pattern(29). The protein sequence and domain structure of FSP27 place it in the Cell death Inducing DFF45-like Effector (CIDE) protein family (20); this family is comprised of CIDEA, CIDEB and FSP27 (also known as CIDEC and CIDE-3). While ectopic expression studies have demonstrated that CIDE protein expression is sufficient to induce apoptotic cell death (19), whether CIDEs are indeed an integral or necessary part of the apoptotic death machinery remains to be determined. CIDEs share protein sequence similarities with DNA Fragmentation Factor 45 (DFF45) in their N-terminal CIDE-N domain (19). Each of the three CIDE proteins also evidence a region of C-terminal shared homology, the CIDE-C domain. Data supports a current model whereby CIDE proteins exert a proapoptotic effect via modulation of the actions of the DFF40/DFF45 DNA fragmentation factor complex via CIDE protein sequestration of the inhibitory subunit 63

71 DFF45 (21, 31) this leads to the activation of DFF40 nuclease, the major nuclease responsible for the DNA fragmentation that is a classic hallmark of apoptotic cell death (20, 30). The full range and mechanisms of CIDE protein action in apoptosis remain less than fully addressed, however it appears that in the case of CIDEB, it is the CIDE-C domain that is necessary and sufficient for mitochondrial localization and apoptosis (8), when assessed by ectopic expression studies. To date, data on FSP27 function in apoptosis is limited to the observation that ectopic expression of human FSP27 in 293T cells and CHO cells, promotes cell death as evidenced by cell morphology and DNA fragmentation (29). The mechanisms underlying this apoptotic response to FSP27 expression are not known. Ectopic expression studies of an EGFP-hFSP27 expression construct in COS cells has revealed that the protein, as has been described for CIDEA and CIDEB, localizes to mitochondria (29). In contrast, a recent study failed to find mitochondrial localization but demonstrated presence of FSP27 protein at adipocyte lipid droplets (39). This observation is consistent with the previous findings of lipid droplet association for FSP27 during an analysis of the adipocyte lipid droplet proteome (5), although in the latter case this was initially attributed to possible mitochondrial contamination of lipid droplet preparations. In addition to the recently reported role of FSP27 in lipid droplet function, CIDEA and CIDEB have been firmly established to have a metabolic role in healthy cells. Studies in null mice indicates that CIDEA has a central function in the normal physiology of BAT, where it is key to regulating thermogenesis and energy expenditure via its impact on the activity of the mitochondrial uncoupling protein 1 (UCP1) (28, 30). Protein-protein interaction of CIDEA with UCP1 has been demonstrated in ectopic 64

72 expression studies by coimmunoprecipitation, it is postulated that CIDEA acts to inhibit UCP1 uncoupling activity, thereby resulting in leanness of CIDEA null mice (28, 30). Studies also indicate a role for CIDEA in human energy balance (10, 18, 36) where CIDEA depletion in cultured human adipocytes has been demonstrated to stimulate lipolysis (36). CIDEB, which is highly expressed in liver, has recently been found to be an important regulator of lipid metabolism in liver (27). CIDEB null mice evidence lower levels of plasma triglycerides, free fatty acids, and are resistant to high-fat diet-induced obesity and liver steatosis; these mice also had increased insulin sensitivity, enhanced rates of whole body metabolism and hepatic fatty acid oxidation, and decreased lipogenesis (27). Thus CIDEs appear to be a family of proteins that have dual functions in apoptosis and in the physiology of normal healthy cells. Our studies herein provide new details on the regulation of the endogenous FSP27 transcript in the adipocyte lineage and in obesity and support the notion that FSP27 has a dual function(s) in adipocyte metabolism and in cellular apoptosis. Materials and Methods Cell Culture and Adipocyte Differentiation 3T3-L1 cells were obtained from American Type Culture Collection, Manassas, VA and propagated in DMEM supplemented with 10% calf serum. For differentiation, 3T3-L1 cells were treated at two days post-confluence with DMEM supplemented with 10% FCS in the presence of the adipogenic inducers 0.5 mm methylisobutylxanthine (MIX) and 1 µm dexamethasone for 48 h. Adipogenic agents were then removed and 65

73 growth of cultures continued in DMEM containing 10% FCS. At five days post-induction of differentiation, adipocyte conversion had occurred in approximately 90% of the cells, as judged by lipid accumulation and cell morphology. For differentiation of brown preadipocytes obtained from C.R. Kahn (Joslin Diabetes Foundation, Harvard Medical School, Boston, MA), cells were cultured to confluence in DMEM with 10% FCS, 20 nm insulin and 1 nm triiodotyronine (differentiation medium per Kahn and colleagues (24)). Confluent cells were incubated in differentiation medium that was supplemented with 0.5 mm MIX, 0.5 µm dexamethasone, and mm indomethacin for 48 h. Following this period nearly 100% of cells showed adipogenic conversion at which time culture medium was replaced with differentiation medium and was replenished every two days thereafter. For culture and differentiation of primary white adipocytes, WAT collected from male Sprague Dawley rats was digested with 1 mg/ml of type I collagenase for 40 min with shaking at 37 C. Following digestion, material was filtered through a 300 micron pore size nylon mesh (Sefar America Inc, Depew, NY) and filtrate centrifuged at 2,000 rpm for 5 min. Floating adipocyte fraction was removed and the pellet of stromalvascular cells was resuspended in DMEM containing 10% FCS and plated. Upon confluence cells were either harvested or subjected to differentiation media consisting of DMEM containing 10% FCS, 0.1 µm dexamethasone, 0.25 mm MIX, and 17 nm insulin for three days, at which time differentiation media was removed and cell cultures were maintained in DMEM containing 10% FCS and 17 nm insulin. Human preadipocyte RNA and adipocyte RNA were purchased from Zen-Bio Inc. (Research Triangle Park, NC). 66

74 TNF!, Insulin, and Inhibitor Treatment of 3T3-L1 Adipocytes For treatments of 3T3-L1 adipocytes with TNF!, cells were incubated with TNF! for the indicated dose and times in DMEM supplemented with 10% FCS. For studies of regulation by insulin, 3T3-L1 adipocytes were cultured for 16 h in serum-free DMEM with 0.5% BSA and media was then changed to DMEM containing 0.5% BSA supplemented with insulin, as indicated. For studies using inhibitors to assess signaling mechanisms in 3T3-L1 adipocytes, cells were treated with 50 µm PD98059, 50 µm LY294002, or 1 µm rapamycin (Sigma-Aldrich, St. Louis, MO), or DMSO vehicle. For inhibitor studies, in the case of TNF!, experiments were carried out in normal serumcontaining culture conditions with a 1 h pretreatment with the indicated inhibitor or with DMSO vehicle, followed by 16 h culture with or without 10 ng/ml TNF!. For studies of the effects of pharmacological inhibitors on insulin signaling, cells were first serumstarved for 6 h, pretreated with inhibitor or DMSO vehicle for 1h, and treated with 100 nm insulin or untreated for 16 h. Studies were carried out in either duplicate or triplicate, in wholly independently conducted experiments. RNA Preparation and Transcript Expression Analysis For studies of transcript expression in murine tissues, 8 wk old C57BL/6 or ob/ob male mice were utilized. All animal procedures were carried out with approval from Medical University of Ohio Animal Care and Use Committee. RNA was purified using TriZol reagent (Invitrogen Corp.) according to manufacturer s instruction. For Northern 67

75 blot analysis, 5 mg of RNA was fractionated in 1% agarose-formaldehyde gels in MOPS buffer and transferred to Hybond-N membrane (GE Healthcare, Piscataway, NJ). 32 P- labeled probes for use in Northern blot analysis were synthesized using a random-priming kit (Promega, Corp.). Blots were subject to 1 h hybridization in ExpressHyb solution (BD Biosciences Clontech, Palo Alto, CA). Following high stringency washing, membranes were exposed at -80 C to Kodak Biomax film with a Kodak Biomax intensifying screen or to a phosphorimager screen and read using a Typhoon 8600 PhosphorImager (GE Healthcare). In some instances blots were rehybridized with a probe for 36B4 transcript, which encodes the acidic ribosomal phosphoprotein PO, a commonly employed internal control (26). For real-time PCR, total RNA was isolated as above, and subjected to DNase I treatment and clean-up (Qiagen). Reverse transcription was performed with SuperScript II RNase H-reverse transcriptase (Invitrogen Corp.) and an oligo(dt)-22 primer. Transcript levels of were assessed by SYBR green-based real-time PCR conducted with an ABI 7500 Real Time PCR System (Applied Biosystems, Foster City, CA). Reaction conditions were 1X SYBR Green PCR Master Mix (Applied Biosystems), 100 nm each forward and reverse primers, and 50 ng of cdna. PCR was carried out over 40 cycles of 95 C for 15 sec, 60 C for 30 sec, and 72 C for 34 sec with an initial cycle of 50 C for 2 min and 95 C for 10 min. Primers used were: FSP27, 5'- CAGAAGCCAACTAAGAAGATCG-3' and 5'-TGTAGCAGTGCAGGTCATAG-3' and for GAPDH, 5 -AACAGCCTCAAGATCATCAGC-3 and 5 - GGATGATGTTCTGGAGAGCC-3. Expression was normalized against respective 68

76 GAPDH transcript level and fold differences calculated. Statistical analyses were conducted on triplicates using single factor ANOVA. Cellular Fractionation of Adipose Tissue To assess expression of FSP27 in adipose tissue components in vivo, tissue was fractionated into adipocyte and stromal-vascular fractions. For this, WAT and BAT was removed from male C57BL/6 mice, rinsed three times in sterile PBS, and minced with scissors. Tissue was transferred to a 50 ml sterile tube with 15 ml of HBSS containing 0.2 mg/ml of type II collagenase (Sigma-Aldrich). After digestion for 40 min at 37 C with constant agitation, material was filtered through a 300-µm pore size nylon mesh. Filtrate was collected into sterile 50 ml centrifuge tubes and centrifuged at 2,000 rpm for 5 min. The floating adipocyte cell fraction was then lysed in 10 volumes of TriZol reagent and the stromal-vascular pellet fraction lysed in 2 ml TriZol reagent. RNA was extracted and analyzed by Northern blot analysis as described above. 69

77 Constructs for Mammalian Expression Full-length expression constructs for murine (GenBank BC099676) and human FSP27 (GenBank BC016851) in the vector CMVSport6 were obtained as an I.M.A.G.E. clone from ATCC and the full sequence verified. To prepare the HA-FSP27-FLAG expression construct, PCR amplification was carried with human FSP27 as template to amplify the open reading frame of FSP27. A FLAG epitope tag was incorporated into the 3' PCR primer and KpnI and BamHI restriction enzyme sites were incorporated, respectively, into the 5' and 3' PCR primer pairs to facilitate directional subcloning into an N'-terminal epitope tag CMV expression vector (kindly provided by Dr. W. Maltese, University of Toledo Health Science Campus). For egfp fusion protein studies, the open reading frame of human FSP27, minus the initiator methionine, was PCR-amplified and subcloned into the pegfp-c1 vector (Clontech), resulting in an expression construct wherein the egfp protein is present N-terminal to the FSP27 coding region. The inserts, cloning junctions and epitope tag regions of all constructs were fully sequence verified. Cell Death Assay To assess the effects of ectopic expression of mfsp27 on cell viability, 293T cells or 3T3-L1 preadipocytes were co-transfected with an mfsp27 expression construct or empty vector in combination with a "-galactosidase (LacZ) expression construct. The mass of DNA(s) utilized is presented in the respective figure legend. Transfections were done in triplicate using Lipofectamine 2000 (Invitrogen Corp.) and numbers of LacZtransfected cells assessed at 48 h post-transfection via "-galactosidase staining. For this, 70

78 cells were fixed at room temperature in 0.5% glutaraldehyde. Following two PBS washes, cells were incubated in staining solution (2 mm MgCl 2, 5 mm K 3 Fe (CN) 6, 5 mm K 4 Fe(CN) 6, 1 mg/ml 5-bromo-4-chloro-3-indolyl-"-D-galactopyranoside (X-gal) in PBS) and incubated at 37 C for 4 h. After incubation, blue cells per microscopic field (200X) were enumerated with 10 independently and randomly chosen fields analyzed per transfection. This protocol has been used in previous studies of apoptosis (13) and was kindly provided by Dr. Han Fei Ding, University of Toledo Health Science Campus. This assay serves as an indirect and visual measurement of cell death. The "- galactosidase expression construct serves as a reporter to mark transfected cells, which are also co-transfected with a test "effector" plasmid, in this case FSP27 or an empty vector pcdna3.1 control. Depending on the cell type under study, cells which have undergone cell death are lost from the cultures and are thus not counted among the LacZ+ blue cells; or appear as very small round LacZ+ blue apoptotic bodies. Comparison of the numbers of LacZ+ blue cells in those cultures transfected with empty vector vs. those transfected with the "effector" plasmid (i.e. FSP27) allows for the detection of cell death attributable to the "effector" plasmid. Single factor ANOVA was used for statistical assessments. Microscopic fields of cells in the studies were observed and photographed using an Olympus IX70 fluorescence microscope and Spot Advanced Software version (Diagnostic Instruments Inc.). All images shown accurately represent the original data, however in some instances brightness and contrast were adjusted to allow for better visualization of details. For DNA fragmentation assays, genomic DNA was prepared from 3T3-L1 preadipocytes, 3T3-L1 adipocytes, or transfected 293T cells. For the latter, 2 µg of the 71

79 murine FSP27 expression construct or empty vector was transfected in 293T cells using Lipofectamine DNA was prepared with an Apoptotic DNA-Ladder Kit (Roche Diagnostics, Nutley, NJ) exactly per manufacturer directions or by manual preparation using standard methods. For the latter, cells were collected from the media and culture plates and subjected to low speed centrifugation. The pellet was resuspended in cell lysis buffer (0.5% SDS, 20 mm EDTA, 5 mm Tris HCl ph 8.0) and lysates were incubated on ice for 20 min. Insoluble material was removed by centrifugation and the supernatant extracted with phenol/chloroform. DNA was ethanol precipitated and the pellet resuspended in water and subject to RNase digestion. DNA was assessed by fractionation on 1.2% agarose gels, stained, visualized under short wave UV illumination and photographed. Western Blot Analysis 293T cells were transfected with 2 µg of FSP27 or empty vector and harvested at 48 h post-transfection by lysis in TNN(+) buffer (10 mm Tris ph 8.0, 120 mm NaCl, 0.5% NP-40, 1 mm EDTA supplemented with a protease inhibitor cocktail). Lysates were incubated on ice for 30 min with intermittent vortexing, supernatant collected via centrifugation, and protein content determined (Bio-Rad, Hercules, CA). For Western blot analysis, 30 µg of protein extract was fractionated on SDS-PAGE, followed by electroblotting onto PVDF membrane with M Tris/0.192 M glycine transfer buffer supplemented with 20% methanol. Membranes were blocked for 1 h in 5% non-fat milk in PBS containing 0.5% Tween 20 (PBS-T) followed by either 1 h incubation at room 72

80 temperature, or overnight at 4 o C, with a 1:2,000 dilution of antibody to full length PARP (Affinity BioReagents) or cleaved PARP,!-Fodrin, cleaved!-fodrin (Cell Signaling Technology), or a 1:10,000 dilution of "-tubulin (Covance Research Products). Polyclonal FSP27 antibody was used at 1:1000 and was produced by arrangement with ProSci, using a full length mouse FSP27-TrpE fusion protein as immunogen. Secondary antibody was HRP-conjugated goat anti-rabbit (Bio-Rad), used at a 1:2000 dilution, and washes were conducted in PBS-T. Signal was detected by ECL Plus enhanced chemiluminescence (GE Healthcare) and exposure to X-ray film. Localization Studies Confocal imaging studies employing MitoTracker Red CMX Ros (Invitrogen) were conducted on live cells by plating 3T3-L1 preadipocytes or COS cells onto laminin coated MatTek glass bottom dishes (MatTek Corporation, Ashland MA) at a density of 3 X10 5 cells per 35 mm diameter dish. For studies using DSRed2-Mito (Clontech) COS cells were plated onto laminin-coated coverslips and observed after methanol fixation. Cells were transfected using Lipofectamine 2000 with 2 µg of either egfp empty vector or egfp-fsp27 construct (for MitoTracker studies) or with 1 µg of egfp empty vector or egfp-fsp27 construct in combination with 1 µg of pdsred2-mito expression construct (Clontech). MitoTracker staining was according to manufacturer instruction at a concentration of 50 nm for 45 min at 37 o C. Cells were observed at ~20 h post transfection. Confocal studies were performed using resources of the Advanced Microscopy and Imaging Center at the University of Toledo Health Science Campus. Images were captured using a Leica TCS SP5 broadband confocal microscope (Leica, 73

81 Mannheim, Germany) equipped with Argon-488 and diode pumped solid state-561 laser sources and 63.0x 1.40 N.A. oil immersion objective. A series of optical Z sections, 0.5&M in thickness and totaling 5-6 &M, were collected and visualized as projection images using Leica LAS software. Laser intensities and microscope settings between samples were maintained constant. In the case of 3T3-L1 preadipocytes transfected with either egfp empty vector or egfp-fsp27 construct and used for observation of apoptotic morphology, cells were plated at a density of 5 X10 4 per well of six well plates. Nuclei of live cells were stained at 24 h post-transfection with 1 mg/ml Hoechst dye for 20 minutes at 37 o C and observed and photographed using an Olympus IX70 fluorescence microscope and Spot Advanced Software version (Diagnostic Instruments Inc.). All images shown accurately represent the original data, however in some instances brightness and contrast were adjusted to allow for better visualization of details. Results Differentiation-Dependent Expression of FSP27 in Multiple Models of Adipogenesis Despite the fact that Ringold and coworkers initially identified FSP27 over two decades ago, information on the expression and regulation of endogenous FSP27 transcript is minimal. FSP27 was originally identified as a transcript upregulated during in vitro adipose conversion of TA1 cells (11). They also demonstrated that a 2.5 kb fragment of the FSP27 promoter evidenced adipocyte differentiation-dependent expression. This was ascribed in part to functional C/EBP! binding sites in the FSP27 74

82 promoter, although the direct transcriptional activation of the FSP27 promoter by C/EBP! was not investigated (11). TA1 cells are a rarely-utilized preadipocyte cell line established by Ringold and coworkers via treatment of the murine 10T1/2 mouse embryo fibroblast cell line with the demethylating agent 5-azacytidine (7, 25); this model system for in vitro adipogenesis was used solely by Ringold and coworkers (6). TA1 cells are therefore not well-characterized, in contrast to the many key molecular events of adipogenesis that have been delineated in the 3T3-L1 in vitro adipogenesis model. Thus we deemed it important to define the expression of FSP27 transcript in the widely used and well-characterized 3T3-L1 model. As shown by the Northern blot analysis in Figure 1A, FSP27 transcript is not detected in 3T3-L1 preadipocytes and is readily detected starting at 2 days post-induction and continues to increase throughout the 5 day time point, which represents mature 3T3-L1 adipocytes. Real-time PCR analysis indicated a 5.7 X fold increase (p<0.001) in FSP27 transcript level in day 7 adipocytes vs. preadipocytes. Adipocyte conversion is validated by expression of the adipocyte marker genes stearoyl Co-A desaturase 1 (SCD1), adipocyte fatty acid binding protein (afabp), adipose tissue triglyceride lipase (ATGL) and PPAR#. We next examined expression of FSP27 transcript in several additional models of adipogenesis. We found that FSP27 transcript is not detected in rat primary preadipocytes and that expression of FSP27 transcript emerges upon their adipocyte conversion (Figure 1B), paralleling that of the adipocyte marker transcript ATGL. The regulation of expression of FSP27 during brown adipogenesis is unknown, although we (Figure 6) and others (64) have detected its expression in BAT. To examine this we used a permanent brown preadipocyte cell line established by Kahn and coworkers that was derived from 75

83 neonatal BAT. In this cell culture model, brown adipocyte conversion is evidenced by emergence of the brown adipocyte-specific transcript UCP1, and by multilocular lipid accumulation and cell morphology (data not shown). As shown in Figure 1C, FSP27 transcript is not found in brown preadipocytes and is detected at both 3 and 8 days post induction of adipocyte conversion, with particularly enriched signal found at day 3. Quantitation of FSP27 in brown preadipocytes (day 0) and brown adipocytes (day 8) by real time PCR indicated a 57-fold increase (p<0.001) in FSP27 transcript level during brown adipogenesis. We also determined expression of FSP27 in adipose tissue cell types derived from an in vivo source. For this whole murine WAT and BAT was fractionated into adipocytes and the non-adipocyte stromal-vascular fraction. Figure 1D indicates that FSP27 transcript is not detected in stromal-vascular cells, which are the source of in vivo preadipoctyes, but is readily detected in the adipocyte cell population of WAT and BAT. We also addressed adipocyte differentiation dependent expression of FSP27 transcript in in vitro differentiation of primary human preadipocytes. Figure 1E shows that while the FSP27 transcript is not detected in human preadipocytes it is readily detected in differentiated human adipocytes. Thus by assessing multiple adipocyte model systems, we conclude that upregulation of FSP27 transcript is integral to the molecular definition of white and brown adipocytes. Puri and coworkers recently indicated a similar adipocyte enriched expression for FSP27 transcript, although in that case data was not shown (39). The full regulatory pathways governing adipocyte expression of FSP27 are not currently known. However, as discussed above, C/EBP! action is implicated in that it has been demonstrated to physically bind at FSP27 promoter elements (11). At the time that FSP27 was under study by Ringold and colleagues, PPAR# had yet to be discovered. However, a 76

84 microarray profiling study of a PPAR#1 transgenic mouse model indicates FSP27 might also be a target of PPAR# transactivation (63). These mice were found to express elevated levels of a number of adipocyte transcripts in liver, including FSP27. TNF! Downregulates FSP27 Transcript in 3T3-L1 Adipocytes A previous report indicated that downregulation of an FSP27 promoter-cat reporter construct occurred at 3 h of TNF! treatment of TA1 adipocytes, and this was ascribed as likely due to TNF!-mediated downregulation of C/EBP! (59). However, we are not aware of reports on regulation of endogenous FSP27 transcript level by exogenous agents such as cytokines or hormones. As our ultimate goal is to fully understand the regulation and function of FSP27 in the adipocyte lineage, we assessed the regulation of FSP27 transcript in 3T3-L1 adipocytes by treatment with insulin and TNF!, two agents closely linked to normal adipocyte metabolism and adipocyte pathophysiology. Comprehensive oligonucleotide microarray assessments of the global transcriptional response of 3T3-L1 adipocytes to TNF! have revealed that this cytokine has global effects on 3T3-L1 adipocyte gene expression (47, 48). It has long been known that TNF! treatment of preadipocytes inhibits adipogenic conversion (54, 62), exposure of adipocytes to TNF! stimulates lipolysis (53) and promotes a dedifferentiated adipocyte phenotype (62). These effects have been ascribed, at least in part to the TNF!-mediated transcriptional downregulation of the master adipocyte transcription factor, PPAR# (62) and C/EBP! (59). 77

85 Figure 2A reveals the regulation of the FSP27 transcript with a 24 h exposure of 3T3-L1 adipocytes to a range of TNF! concentrations. These studies were conducted in FCS-containing media, culture conditions typically employed to demonstrate the TNF!- mediated effects adipocyte dedifferentiation of 3T3-L1 cells. TNF! treatment was highly effective at downregulating FSP27 transcript level and did so at the lowest TNF! concentration tested, 0.01 ng/ml. It was similarly effective 10 ng/ml, a concentration that is typically used for studies of TNF! effects in adipocytes; real-time PCR revealed this treatment reduced FSP27 transcript to ~12% of the level in control untreated cells (p<0.001). Also shown is the effect of TNF! on three other adipocyte-enriched genes, ATGL, resistin, and SCD1. Here, distinctions are noted in the concentration-dependent decrease among these transcripts, in that the 0.01 ng/ml treatment effectively reduced FSP27 and resistin transcript levels but was largely ineffective in regard to transcripts for ATGL and SCD1. Figure 2B shows a Northern blot analysis for temporal assessment of FSP27 transcript levels in untreated 3T3-L1 adipocytes (0 h) through 48 h of 10 ng/ml TNF! treatment. A reduction of FSP27 transcript level is clearly noted at 4 h of TNF! exposure; this decrease is sustained through the final time point examined, 48 h. Realtime PCR analysis indicated that TNF! treatment reduced FSP27 to 10% of that in control cells (p<0.001) when measured at the 24 h time point. The temporal effects of TNF! on FSP27 transcript level are similar to that for the master adipogenic transcriptional regulator PPAR# and for two of the three other adipocyte-enriched genes we assessed, ATGL and resistin. A third adipocyte-enriched transcript, SCD1, also evidences downregulation by the TNF!, albeit with a somewhat slower temporal effect. 78

86 To investigate the intracellular signaling pathways underlying downregulation of FSP27 transcript by TNF!, we pretreated 3T3-L1 adipocytes with specific pharmacological inhibitors and assessed FSP27 transcript level in the absence or presence of TNF!. The results are shown in the Northern blot analysis in Figure 3A and the accompanying graphical representation in Figure 3B. Neither a pretreatment with the p44/42 MAP kinase inhibitor PD98059 nor the p70s6 kinase inhibitor rapamycin affected levels of endogenous FSP27 transcript, nor did either of these inhibitors attenuate the TNF!-mediated decrease of FSP27 transcript level. In the case of LY294002, a PI3-kinase inhibitor, treatment with this inhibitor alone resulted in a dramatic reduction of FSP27 transcript level to 15% of the level of the DMSO vehicletreated control. Although the regulation of endogenous FSP27 we note upon treatment with LY alone makes an assessment of its effect in the presence of TNF! difficult to ascertain. However LY did not block the effects of TNF! on FSP27 transcript level, and if anything resulted in a small additional inhibition. Moreover, the observation that LY alone led to a marked reduction of endogenous FSP27 transcript in serum-containing culture conditions suggested that a component of FCS, whose signaling mechanism is affected by LY294002, was responsible for the noted attenuation of expression of the FSP27 transcript level. FSP27 Transcript Level is Under Tight Regulation by Insulin and Involves PI3- Kinase Signals 79

87 To specifically address the ability of insulin to regulate FSP27 transcript we utilized in vitro studies with 3T3-L1 adipocytes. Given that serum components might interfere with our assessments, these studies were carried out in serum-free conditions. Figure 4A shows the Northern blot analysis for the FSP27 transcript under culture conditions of no insulin through 200 nm insulin. An increase in FSP27 transcript level is observed at the lowest insulin concentration tested, 0.1 nm, with a similar magnitude of increase noted for each of the insulin concentrations tested. The Northern blot in Figure 4B reveals the temporal nature of the response and indicates that insulin markedly increases FSP27 transcript level within 24 h while no change in FSP27 transcript was evident in time-matched serum-free cultures. We next addressed the mode of insulin action in upregulation of FSP27 transcript by pretreatment with DMSO (vehicle), PD98059, LY294002, or rapamycin. For these studies, 3T3-L1 adipocytes were cultured under serum-free conditions with the indicated inhibitors only, or cultured with insulin in the presence of the indicated inhibitors. As shown by the Northern blot analysis and the accompanying graphical representation of this data, (Figures 5A and 5B), a barely detected signal for FSP27 transcript was found under serum-free culture conditions in the absence exogenous insulin. Insulin stimulated FSP27 transcript level by ~8-fold over that of DMSO vehicle-treated control cultures. The magnitude of this increase was not affected by PD98059 or rapamycin pretreatment. That these two agents were unable to block the insulin upregulation of FSP27 transcript suggests that neither p44/42 MAP kinase nor p70s6 kinase signals function in the insulin responsiveness of the FSP27 gene. In marked contrast, pretreatment of adipocytes with LY reduced the magnitude of the insulin-stimulation of FSP27 transcript level to 80

88 14% of that of insulin-treated control cultures. We were concerned that the barely detectable level of FSP27 transcript present in serum-free insulin-free conditions may have limited our ability to assess the effects of inhibitors alone on FSP27 transcript by quantitative Northern blot. However real-time PCR analysis of the effects of LY under these conditions indicated that LY resulted in an 80% decrease in FSP27 transcript levels (p<0.001), similar to the 88% decrease found by quantitative Northern blot analysis. Together, these observations demonstrate that insulin stimulation of FSP27 transcript expression involves PI3-kinase, an intracellular mediator whose role in insulin signal transduction is well established. Expression of FSP27 in Wild Type and obese Murine Tissues We next examined the tissue distribution of FSP27 transcript in a wide panel of murine tissues by Northern blot analysis. Figure 6A demonstrates that the FSP27 transcript is highly enriched in murine WAT and also demonstrates readily detectable expression in BAT. Adipose tissue enrichment of FSP27 was also recently confirmed by others, although the data for such was not shown (39). Also shown in Figure 6A is the murine tissue distribution for the other two CIDE family members. As previously reported, Cidea is highly enriched in murine BAT. In regard to murine Cideb, we detect expression in liver as has been previously reported for the human transcript (20), and also demonstrate expression of Cideb in murine intestine and kidney. Figure 6B shows assessment of FSP27 transcript level in various murine tissues by real-time PCR and indicates that WAT expresses ~7.5-fold higher levels (p<0.001) than does BAT. Signal is 81

89 also detected in several other tissues, albeit at level more than 50-fold below (p<0.001) that found in WAT. In contrast to the fat-specific nature of FSP27 transcript expression in mice, very limited studies to date on the cell and tissue expression of human FSP27 have led to the suggestion that it is not particularly restricted to human adipose tissue (29). We next addressed if FSP27 transcript showed dysregulated expression in obese tissues using the murine ob/ob obesity model; such mice are obese due a mutation in the leptin gene. The Northern blot analysis in Figure 7A indicates that WAT from wild type and ob/ob mice show a similar level of FSP27 transcript. However upon examination of liver of wild type and ob/ob mice for FSP27 transcript, a dramatic increase was found for ob/ob liver (Figure 7B). Real time PCR analysis revealed a 49-fold enrichment (p<0.001) in liver of ob/ob mice, compared to wild type C57BL/6 mice. The only data of which we are aware that may link FSP27 to liver dysfunction is a report by Reddy and coworkers on adipocyte-specific gene expression and adipogenic hepatic steatosis in the mouse liver due to PPAR#1 overexpression on a PPAR! null background (63). Upregulation of a number of adipocyte enriched genes occurred in liver of these mice, with an 11-fold increase reported for FSP27 transcript (63). At this time, however, it is not possible to define the distinctions between the effects of upregulation of FSP27 and hepatic steatosis and the effects of the myriad of other genes also increased in the liver of these transgenic mice. We also detected upregulation of Cidea transcript in ob/ob liver (Figure 7C). No difference in FSP27 or Cidea transcript level was detected for wild type vs. ob/ob kidney, nor did we detect Cidea in either wild type or ob/ob adipose tissue (data not shown). The upregulation of Cidea in ob/ob liver is in line with the finding by Kopchick and coworkers of enhanced expression of Cidea in the liver of aged or type 2 diabetic mice 82

90 exhibiting steatosis (22). It has also been reported that Cidea is markedly induced in liver via activation of PPAR# and PPAR! transcription factors, which interact with their cognate PPREs in the Cidea 5'-flanking region (58). Murine FSP27 is Proapoptotic and Leads to Cleavage of PARP and!-fodrin. While human FSP27 has been determined to promote apoptosis upon ectopic expression in 293T and CHO cells, as evidenced by cell death and DNA fragmentation (29), to our knowledge murine FSP27 has not been examined for apoptotic effects. We thus first assessed the ability of murine FSP27 to promote apoptosis in 293T cells, a frequently used cell line for apoptosis studies. Figure 8A shows the result of a DNA fragmentation assay where DNA-laddering is evident in FSP27-transfected cells at 18, 21, 24, and 48 h post-transfection compared to lack of detectable DNA laddering for empty vector transfectants at the 48 h time point. This demonstrates for the first time that, as has been reported for human FSP27, that murine FSP27 is a proapoptotic molecule. As far as we are aware, apoptotic end points other than cell death and DNA fragmentation have not been described in regard to human FSP27 or murine FSP27 action. We therefore used Western blot analysis to examine whether apoptosis mediated by murine FSP27 resulted in the appearance of the cleaved forms of PARP and!-fodrin, two welldescribed proteins that are targets for caspase-mediated cleavage during the apoptotic cascade. Figure 8B shows, for the first time, that FSP27 apoptosis leads to generation of cleaved PARP and!-fodrin which are present at 15 h and later time points, whereas minimal cleavage product(s) is present in empty vector transfectants. Given that PARP 83

91 and!-fodrin are targets for caspase-mediated cleavage, these data suggest that the proapoptotic effects of FSP27 are exerted, at least in part, through caspase-dependent mechanism(s). We also examined the time and dose response of cell death using an assay wherein transfected cells are marked blue due to cotransfection of a LacZ expression construct along with either empty vector or a murine FSP27 expression construct. As described in the Materials and Methods section, in this assay loss of blue cells from the culture serves as an indirect indicator of cell death. Samples were collected for assessment of cell death at indicated time points post-transfection and reveal that the proapoptotic effect of FSP27 can be clearly seen with the naked eye at 48 h and microscopic evaluation showed a clearly discernable effect at 18 h (Figure 8C, and data not shown). Figure 8D illustrates the dose response of cell death wherein a maximum of ~80% cell death is observed at the highest mass of FSP27 plasmid tested, and cell death diminishes with transfection of decreasing mass of the FSP27 expression construct; "- galactosidase stained culture dishes representing maximal dose of FSP27 (left) and empty vector controls (right) are shown beneath the graph. We also sought to compare the level of FSP27 expression that we expressed in transient transfection assays and which we had demonstrated to promote apoptosis, with the level of FSP27 normally found in white adipocytes. To do so we used RT-PCR to determine FSP27 transcript level for 293T cells at various time points post-transfection with an FSP27 expression constructs and compared it with that present in WAT and in 3T3-L1 cells that had been subjected to the adipocyte differentiation protocol, Figure 8E. Transfection studies were conducted under the same conditions utilized for Figure 8A-C. In order to be able to compare FSP27 transcript expression level in the transfected cells 84

92 within the 293T cell population with that in adipocytes, we conducted parallel transfection of 293T cells with a Lac-Z expression construct. "-galactosidase staining of these parallel cultures indicated a 70% transfection efficiency. In regard to the percentage of adipocytes in either the WAT sample or the differentiated 3T3-L1 cell sample, it has been reported that the percentage of adipocytes in whole WAT is between one-third and two-thirds (1), thus we estimated that ~50% of the cells in the WAT sample were fat cells. Adipocytes make up ~85% of the differentiated 3T3-L1 cultures. The plotted data for FSP27 transcript level, after correction for the percent transfection of the 293T cell population and the percentage of adipocytes in either the WAT or differentiated 3T3-L1 cell populations, is shown in Figure 8E. At no time from 4 h through 24 h posttransfection does the level of ectopically expressed FSP27 transcript appreciably exceed that detected in adipocytes from WAT or 3T3-L1 adipocytes. Our data thus supports the conclusion that compared to the FSP27 transcript level that is normally expressed by adipocytes, ectopic expression of FSP27 in 293T cells, at all post-transfection time points tested, is similar to that found in WAT; while FSP27 clearly leads to cell death in the former case. One caveat to this is the possibility that those transfected cells that express extremely high levels of FSP27 transcript have rapidly died off, and are thus no longer within the population of cells we have assessed. FSP27 Promotes Preadipocyte Cell Death We reasoned that since white adipocytes express a high level of FSP27, that perhaps if a role of this molecule in these cells were mediating adipocyte apoptosis, we 85

93 would see an increase in basal apoptotic rate in adipocytes versus preadipocytes. Figure 9A indicates that neither 3T3-L1 preadipocytes nor 3T3-L1 adipocytes evidence detectable apoptosis under normal culture conditions, as demonstrated by the lack of DNA laddering. Thus the induction of expression of FSP27 that occurs during adipogenesis does not, in that context, result in increased cellular apoptosis. This is generally in line with the observations that adipocytes are relatively resistant to apoptosis; at least when studied under the specific conditions of growth factor withdrawal (52). To confirm that mature 3T3-L1 adipocytes indeed express FSP27 protein while preadipocytes do not, we conducted western blot analysis. As shown in Figure 9B, a dramatic induction of FSP27 protein expression occurs in mature adipocytes, with a major protein species in good agreement with the predicted molecular mass of 27.3 kd for FSP27. We also note two minor, faster-migrating species of the protein. It is possible that the smallest of these may be a product of alternate splicing. Human FSP27 has been reported to be present as two transcripts; the smaller of which generates a protein consistent with that we note (28). However our search for identification of alternate FSP27 transcripts from murine adipocytes did not indicate the presence of shorter form(s) of the transcript (data not shown). Interestingly, during our studies on the FSP27 protein we have found that the size of the FSP27 protein species detected varied when the protein possessed an N-terminal vs a C-terminal epitope tag. To further address this observation, we created an expression construct for FSP27 that contained two distinct epitope tags, an N-terminal HA tag, and a C-terminal FLAG tag. Figure 9C shows that use of the anti-ha antibody to detect the N-terminal HA tag results in a single protein species consistent in mass with that predicted for the FSP27 open reading frame. In contrast, use of the anti- 86

94 FLAG antibody to detect the C-terminal FLAG tag results in full length FSP27 protein as well as several distinct smaller protein species. Future work will more fully address the nature and function of the full length and multiple shorter forms of the FSP27 protein. We do not at this time know what regions shorter forms of FSP27 may represent. However, since the smallest of these are ~ half the mass of full length FSP27, we hypothesize that this might correspond to the Cide-C domain; in studies of Cideb, it is the Cide-C domain that has been illustrated to be necessary and sufficient for apoptosis (7). These observations imply that the various FSP27 protein species might represent different functional forms. We next investigated the ability of preadipocytes to undergo FSP27-mediated cell death using ectopic expression of FSP27 in 3T3-L1 preadipoctyes wherein empty vector or FSP27 transfected cells are visualized as blue via cotransfection of a LacZ expression construct, as previously employed for 293T cells (Figure 8). Since compared to 293T cells, 3T3-L1 preadipocytes exhibit markedly reduced transfection efficiency, for these studies the mass of LacZ expression construct utilized was increased. Here the ratio of FSP27 to LacZ expression construct was only 3:1, rather than the small amount of tracer LacZ plasmid used with 293T cells in Figure 8. Figure 10A shows that relative to empty vector transfectants, 3T3-L1 preadipocytes transfected with an FSP27 expression construct evidenced a statistically significant ~20% decrease (p <0.05) in cell number. When these same transfection conditions were applied to 293T cells, they showed a similar ~20% decrease (p <0.05) in cell number, indicating that FSP27 is equally effective at promoting cell death in 3T3-L1 preadipocytes as for 293T cells. Figure 10B illustrates the cell death effect of FSP27 in 3T3-L1 preadipocytes by "-galactosidase 87

95 staining assay that utilized a 6:1 ratio of FSP27 to LacZ expression construct. A representative microscopic field of a culture dish is shown; transfection efficiency was too low to allow visualization of "-galactosidase staining by macroscopic view of culture dishes. We have also observed the cell death effect in CHO cells, where the majority of "-galactosidase positive cells appear as small apoptotic bodies in the FSP27 transfected cultures vs. the normal morphological appearance of cells transfected with empty vector. Using an egfp-taggged FSP27 protein, Figure 10C illustrates that cells harboring this fusion protein are clearly undergoing apoptosis, illustrated by membrane blebbing and loss of cellular integrity (upper panel) and by nuclear condensation as detected by staining with Hoechst dye (lower panel). The cellular morphology and nuclear staining intensity of empty egfp vector control transfectants was indistinguishable from nontransfected cells (data not shown). When we examined egfp-fsp27 transfected COS cells or 3T3-L1 preadipocytes at time points later than ~30 h post-transfection, we had difficulty finding any egfp-positive cells in the transfected population; those that were present had a small circular morphology indicative of late-stage apoptosis or were round, floating, and presumably dead cells (data not shown). Localization Studies of FSP27 Protein Our data clearly indicate that misexpression of FSP27 in cells other than adipocytes can trigger apoptosis accompanied by several key apoptotic hallmarks. The precise mechanism of this apoptotic action is not yet known; however the reported localization of FSP27, as well as Cidea and Cideb, to mitochondria would appear to 88

96 indicate that FSP27 may initiate apoptosis via mitochondrial signals. While this manuscript was in revision, it was reported by Puri and coworkers that FSP27 is a lipid droplet associated molecule; these authors also indicated lack of mitochondrial localization for FSP27 (39), in contrast to an earlier report (29). Both of these studies utilized the same type of egfp fusion proteins and MitoTracker probes for their analysis, with the latter assessing human FSP27 and the former murine FSP27. Since the localization of FSP27 is key to its functional consequences in both adipocyte physiology and apoptosis, we had also carried out studies on the localization of human egfp-fsp27 in COS cells and 3T3-L1 preadipocytes by confocal microscopy (Supplemental Figure 1). An extremely low transfection efficiency of 3T3-L1 adipocytes and the ineffectiveness of the FSP27 antibody for immunocytochemistry precluded a definitive study on FSP27 protein localization in adipocytes in our hands. However our observations in COS and 3T3-L1 preadipocytes indicated only an extremely limited degree of co-localization of FSP27 with MitoTracker Red; this is particularly apparent in the individual 0.5 micron Z- sections. All MitoTracker probes are to some extent dependent on intact mitochondrial membrane potential for fluorescence intensity. Given that the effects of FSP27 on mitochondrial membrane potential is not known, we therefore also assessed localization by transfection of an organelle specific fluorophore that is not known to be affected by mitochondrial membrane potential. The pdsred-2 protein is directed to specifically localize to mitochondria via the addition of a mitochondrial targeting signal; this method also failed to indicate mitochondrial localization (Supplemental Figure 2). However we have routinely observed that expression of FSP27 in COS cells leads to disruption of the 89

97 mitochondria from a fine spaghetti-like filaments that are normally present in these cells, to a disordered appearance (Supplementary Figure 3). This suggests the possibility of a role for mitochondria in the proapoptotic action of FSP27, as has been indicated in studies of Cideb (8). Discussion The regulation we describe for FSP27 transcript is wholly consistent with the recent report that in adipocytes FSP27 acts to support energy storage. FSP27 protein has been reported to be localized surrounding adipocyte lipid droplets in a pattern reminiscent of that for the well-characterized lipid droplet protein perilipin (39). Our observations that FSP27 transcript is upregulated by insulin and diminished by TNF!, suggests that not only the function of the protein per se, but also that a coordinated transcript regulation is also involved in the anti-lipolytic and pro-energy storage effects of FSP27. On the other hand, the regulation we find for the FSP27 transcript in adipocytes is at apparent odds with the role of TNF! and insulin in cellular apoptosis and cell survival. TNF! is a well-described proapoptotic cytokine in a number of cell types including adipocytes (29, 35, 37, 38, 40, 50). With the caveat that many of the control points in apoptotic signaling are at the level of protein processing and interaction, one might nonetheless anticipate that if the central role of FSP27 were in the stimulation/mediation of apoptosis, then pro- and anti-apoptotic agents would modulate the FSP27 transcript level accordingly. In this regard, TNF! would be predicted to increase, rather than decrease, FSP27 level. Importantly, we have discovered that FSP27 transcript is a target 90

98 for rapid and robust upregulation by insulin. Surprisingly, given the central role of insulin in adipocyte function, there are relatively few studies that have systematically addressed the effects of insulin on global adipocyte gene expression (34, 49). To our knowledge, we are the first to present data describing insulin upregulation of the FSP27 transcript. Insulin is generally a pro-survival factor and has been demonstrated to inhibit adipocyte apoptosis (37, 40). As such, insulin might be predicted to decrease expression of proapoptotic genes, but we observe a dramatic upregulation of FSP27 transcript by insulin. As insulin is well established to promote lipid storage in fat cells, upregulation FSP27 by insulin may be a type of protective mechanism whereby the cell is assured it has adequate levels of this particular lipid droplet protein to sufficiently handle increased lipid storage that occurs as a consequence of the pro-lipogenic effects of insulin on adipocytes. Puri and coworkers recently showed that knockdown of FSP27 in adipocytes was reported to lead to fragmentation of large lipid droplets into abundant small droplets, and that ectopic expression of FSP27 in 3T3-L1 preadipocytes, as well as in COS and CHO cells, leads to the accumulation of lipid droplets (39). While we have not assessed lipid droplet accumulation as the result of ectopic FSP27 expression, we have demonstrated that ectopic FSP27 promotes apoptosis. It remains to be investigated whether the ectopic accumulation of lipid that is mediated by FSP27 is a step in the apoptotic mechanism for FSP27; however the lipotoxic and apoptotic effect of lipids has been clearly documented in a number of physiological settings (56, 57). This also raises the interesting point of what might occur when FSP27 is upregulated in cells that lack the appropriate lipid droplet milieu found in the adipocyte; our data would support the notion that such cells 91

99 might be prone to apoptotic cell death. That we do not observe an increase in basal DNA fragmentation upon adipocyte differentiation, despite the dramatic upregulation of FSP27 that occurs in adipogenesis, leads to the hypothesis that perhaps various protein-protein interactions that occur in the adipocyte lineage might function to keep the pro-apoptotic activity of FSP27 in check. In light of the recent study indicating that FSP27 is a lipid droplet associated protein (39), one might postulate that some of these interactions occur at the lipid droplet. Overall, our studies on the localization of FSP27 in transfected COS cells are in line with those reported by Puri and coworkers (39), in that we do not find convincing evidence of mitochondrial localization. In the discussion section of that report, the authors indicated that in light of their data indicating that FSP27 is a lipid droplet protein rather than a mitochondrial protein, they thought it would be interesting to further explore the current working hypothesis for Cidea action, namely the ability of Cidea to apparently inhibit the action of the mitochondrial uncoupling protein UCP1 (thought to be the basis for the lean phenotype of Cidea null mice); might Cidea also be a lipid droplet associated protein? We would agree that this needs further exploration, particularly since we have observed by y east two-hybrid and biochemical methods that FSP27 and Cidea interact (data not shown), which would support the notion that these proteins might share a similar intracellular localization. Until relatively recently it was thought that adipose number in man was constant from young adulthood onward. Over the last several decades with the remarkable progress in the study of adipogenesis, it is now accepted that adipocyte differentiation is an ongoing process throughout the lifespan (17, 51). More recently, it has become clear 92

100 that in addition to adipocyte hypertrophy and hyperplasia, fat cell mass can also be impacted by adipocyte or preadipocyte apoptosis (52). The balance between differentiation, growth, and cell death of the adipocyte lineage is therefore key to determining whether the number of fat cells an individual has is sufficient for optimal health, or a burden that leads to diabetes, heart disease and associated morbidities. It is intriguing to speculate that under certain conditions wherein adipocyte apoptosis is required, cellular mechanisms may exist for the activation of cell death-promoting activities of FSP27 in adipocytes. Likewise, if pathways for the activation of the cell death-promoting aspects of FSP27 function in mature adipocytes were discovered, this might provide a therapeutic intervention point at which to target reduction of fat cell mass and combat obesity. 93

101 References 1. Ailhaud, G and Hauner, H Development of White Adipose Tissue. In: Handbook of Obesity, Etiology and Pathophysiology, edited by Bray, GA and Bouchard, C: Marcel Dekker, Birkenmeier EH, Gwynn B, Howard S, Jerry J, Gordon JI, Landschulz WH, and McKnight SL. Tissue-specific expression, developmental regulation, and genetic mapping of the gene encoding CCAAT/enhancer binding protein. Genes Dev 3: , Boone C, Mourot J, Gregoire F, and Remacle C. The adipose conversion process: regulation by extracellular and intracellular factors. Reprod Nutr Dev 40: , Bradley RL, Cleveland KA, and Cheatham B. The adipocyte as a secretory organ: mechanisms of vesicle transport and secretory pathways. Recent Prog Horm Res 56: , Brasaemle DL, Dolios G, Shapiro L, and Wang R. Proteomic Analysis of Proteins Associated with Lipid Droplets of Basal and Lipolytically Stimulated 3T3-L1 Adipocytes. J Biol Chem 279: Chapman AB, Knight DM, Dieckmann BS, and Ringold GM. Analysis of gene expression during differentiation of adipogenic cells in culture and hormonal control of the developmental program. J Biol Chem 259: ,

102 7. Chapman AB, Knight DM, and Ringold GM. Glucocorticoid regulation of adipocyte differentiation: hormonal triggering of the developmental program and induction of a differentiation-dependent gene. J Cell Biol 101: , Chen Z, Guo K, Toh SY, Zhou Z, and Li P. Mitochondria localization and dimerization are required for CIDE-B to induce apoptosis. J Biol Chem 275: , Coppack SW. Pro-inflammatory cytokines and adipose tissue. Proc Nutr Soc 60: , Dahlman I, Kaaman M, Jiao H, Kere J, Laakso M, and Arner P. The CIDEA gene V115F polymorphism is associated with obesity in Swedish subjects. Diabetes 54: , Danesch U, Hoeck W, and Ringold GM. Cloning and transcriptional regulation of a novel adipocyte-specific gene, FSP27. CAAT-enhancer-binding protein (C/EBP) and C/EBP-like proteins interact with sequences required for differentiation-dependent expression. J Biol Chem 267: , Darlington GJ, Ross SE, and MacDougald OA. The role of C/EBP genes in adipocyte differentiation. J Biol Chem 273: , Ding HF, Lin YL, McGill G, Juo P, Zhu H, Blenis J, Yuan J, and Fisher DE. Essential role for caspase-8 in transcription-independent apoptosis triggered by p53. J Biol Chem 275: ,

103 14. Forman BM, Tontonoz P, Chen J, Brun RP, Spiegelman BM, and Evans RM. 15-Deoxy-delta 12, 14-prostaglandin J2 is a ligand for the adipocyte determination factor PPAR gamma. Cell 83: , Green H and Kehinde O. An established preadipose cell line and its differentiation in culture. II. Factors affecting the adipose conversion. Cell 5: , Gregoire FM. Adipocyte differentiation: from fibroblast to endocrine cell. Exp Biol Med (Maywood) 226: , Gregoire FM, Smas CM, and Sul HS. Understanding adipocyte differentiation. Physiol Rev 78: , Gummesson A, Jernas M, Svensson PA, Larsson I, Glad CA, Schele E, Gripeteg L, Sjoholm K, Lystig TC, Sjostrom L, Carlsson B, Fagerberg B, and Carlsson LM. Relations of Adipose Tissue Cell Death-Inducing DFFA-like Effector A Gene Expression to Basal Metabolic Rate, Energy Restriction and Obesity: Populationbased and Dietary Intervention Studies. J Clin Endocrinol Metab, Havel PJ. Control of energy homeostasis and insulin action by adipocyte hormones: leptin, acylation stimulating protein, and adiponectin. Curr Opin Lipidol 13: , Inohara N, Koseki T, Chen S, Wu X, and Nunez G. CIDE, a novel family of cell death activators with homology to the 45 kda subunit of the DNA fragmentation factor. Embo J 17: ,

104 21. Inohara N and Nunez G. Genes with homology to DFF/CIDEs found in Drosophila melanogaster. Cell Death Differ 6: , Kelder B, Boyce K, Kriete A, Clark R, Berryman DE, Nagatomi S, List EO, Braughler M, and Kopchick JJ. CIDE-A is expressed in liver of old mice and in type 2 diabetic mouse liver exhibiting steatosis. Comp Hepatol 6: 4, Kim S and Moustaid-Moussa N. Secretory, endocrine and autocrine/paracrine function of the adipocyte. J Nutr 130: 3110S-3115S., Klein J, Fasshauer M, Klein HH, Benito M, and Kahn CR. Novel adipocyte lines from brown fat: a model system for the study of differentiation, energy metabolism, and insulin action. Bioessays 24: , Knight DM, Chapman AB, Navre M, Drinkwater L, Bruno JJ, and Ringold GM. Requirements for triggering of adipocyte differentiation by glucocorticoids and indomethacin. Mol Endocrinol 1: , Laborda J. 36B4 cdna used as an estradiol-independent mrna control is the cdna for human acidic ribosomal phosphoprotein PO. Nucleic Acids Res 19: 3998., Li JZ, Ye J, Xue B, Qi J, Zhang J, Zhou Z, Li Q, Wen Z, and Li P. Cideb regulates diet-induced obesity, liver steatosis, and insulin sensitivity by controlling lipogenesis and fatty acid oxidation. Diabetes 56: , Li P. Cidea, brown fat and obesity. Mech Ageing Dev 125: ,

105 29. Liang L, Zhao M, Xu Z, Yokoyama KK, and Li T. Molecular cloning and characterization of CIDE-3, a novel member of the cell-death-inducing DNAfragmentation-factor (DFF45)-like effector family. Biochem J 370: , Lin SC and Li P. CIDE-A, a novel link between brown adipose tissue and obesity. Trends Mol Med 10: , Lugovskoy AA, Zhou P, Chou JJ, McCarty JS, Li P, and Wagner G. Solution structure of the CIDE-N domain of CIDE-B and a model for CIDE-N/CIDE-N interactions in the DNA fragmentation pathway of apoptosis. Cell 99: , MacDougald OA and Mandrup S. Adipogenesis: forces that tip the scales. Trends Endocrinol Metab 13: 5-11., Mueller E, Drori S, Aiyer A, Yie J, Sarraf P, Chen H, Hauser S, Rosen ED, Ge K, Roeder RG, and Spiegelman BM. Genetic analysis of adipogenesis through peroxisome proliferator-activated receptor gamma isoforms. J Biol Chem 277: Mulligan C, Rochford J, Denyer G, Stephens R, Yeo G, Freeman T, Siddle K, and O'Rahilly S. Microarray analysis of insulin and insulin-like growth factor-1 (IGF-1) receptor signaling reveals the selective up-regulation of the mitogen heparin-binding EGF-like growth factor by IGF-1. J Biol Chem 277: Niesler CU, Siddle K, and Prins JB. Human preadipocytes display a depotspecific susceptibility to apoptosis. Diabetes 47: ,

106 36. Nordstrom EA, Ryden M, Backlund EC, Dahlman I, Kaaman M, Blomqvist L, Cannon B, Nedergaard J, and Arner P. A human-specific role of cell deathinducing DFFA (DNA fragmentation factor-alpha)-like effector A (CIDEA) in adipocyte lipolysis and obesity. Diabetes 54: , Prins JB, Niesler CU, Winterford CM, Bright NA, Siddle K, O'Rahilly S, Walker NI, and Cameron DP. Tumor necrosis factor-alpha induces apoptosis of human adipose cells. Diabetes 46: , Prins JB and O'Rahilly S. Regulation of adipose cell number in man. Clin Sci (Lond) 92: 3-11., Puri V, Konda S, Ranjit S, Aouadi M, Chawla A, Chouinard M, Chakladar A, and Czech MP. Fat specific protein 27: A novel lipid droplet protein that enhances triglyceride storage. J Biol Chem, Qian H, Hausman DB, Compton MM, Martin RJ, Della-Fera MA, Hartzell DL, and Baile CA. TNFalpha induces and insulin inhibits caspase 3-dependent adipocyte apoptosis. Biochem Biophys Res Commun 284: , Rajala MW and Scherer PE. Minireview: The adipocyte--at the crossroads of energy homeostasis, inflammation, and atherosclerosis. Endocrinology 144: , Rosen ED, Hsu CH, Wang X, Sakai S, Freeman MW, Gonzalez FJ, and Spiegelman BM. C/EBPalpha induces adipogenesis through PPARgamma: a unified pathway. Genes Dev 16: ,

107 43. Rosen ED, Sarraf P, Troy AE, Bradwin G, Moore K, Milstone DS, Spiegelman BM, and Mortensen RM. PPAR gamma is required for the differentiation of adipose tissue in vivo and in vitro. Mol Cell 4: , Rosen ED and Spiegelman BM. Molecular regulation of adipogenesis. Annu Rev Cell Dev Biol 16: , Rosen ED and Spiegelman BM. PPARgamma: a nuclear regulator of metabolism, differentiation, and cell growth. J Biol Chem 276: Rosen ED, Walkey CJ, Puigserver P, and Spiegelman BM. Transcriptional regulation of adipogenesis. Genes Dev 14: , Ruan H, Hacohen N, Golub TR, Van Parijs L, and Lodish HF. Tumor necrosis factor-alpha suppresses adipocyte-specific genes and activates expression of preadipocyte genes in 3T3-L1 adipocytes: nuclear factor-kappab activation by TNFalpha is obligatory. Diabetes 51: , Ruan H, Miles PD, Ladd CM, Ross K, Golub TR, Olefsky JM, and Lodish HF. Profiling gene transcription in vivo reveals adipose tissue as an immediate target of tumor necrosis factor-alpha: implications for insulin resistance. Diabetes 51: , Sartipy P and Loskutoff DJ. Expression profiling identifies genes that continue to respond to insulin in adipocytes made insulin-resistant by treatment with tumor necrosis factor-alpha. J Biol Chem 278: !

108 50. Sethi JK and Hotamisligil GS. The role of TNF alpha in adipocyte metabolism. Semin Cell Dev Biol 10: , Smas CM and Sul HS. Control of adipocyte differentiation. Biochem J 309 (Pt 3): , Sorisky A, Magun R, and Gagnon AM. Adipose cell apoptosis: death in the energy depot. Int J Obes Relat Metab Disord 24 Suppl 4: S3-7, Souza SC, Palmer HJ, Kang YH, Yamamoto MT, Muliro KV, Paulson KE, and Greenberg AS. TNF-alpha induction of lipolysis is mediated through activation of the extracellular signal related kinase pathway in 3T3-L1 adipocytes. J Cell Biochem 89: , Torti FM, Torti SV, Larrick JW, and Ringold GM. Modulation of adipocyte differentiation by tumor necrosis factor and transforming growth factor beta. J Cell Biol 108: , Umek RM, Friedman AD, and McKnight SL. CCAAT-enhancer binding protein: a component of a differentiation switch. Science 251: , Unger RH. Minireview: weapons of lean body mass destruction: the role of ectopic lipids in the metabolic syndrome. Endocrinology 144: Unger RH and Orci L. Lipotoxic diseases of nonadipose tissues in obesity. Int J Obes Relat Metab Disord 24 Suppl 4: S28-32., Viswakarma N, Yu S, Naik S, Kashireddy P, Matsumoto K, Sarkar J, Surapureddi S, Jia Y, Rao MS, and Reddy JK. Transcriptional regulation of Cidea, 101!

109 mitochondrial cell death-inducing DNA fragmentation factor alpha-like effector A, in mouse liver by peroxisome proliferator-activated receptor alpha and gamma. J Biol Chem 282: , Williams PM, Chang DJ, Danesch U, Ringold GM, and Heller RA. CCAAT/enhancer binding protein expression is rapidly extinguished in TA1 adipocyte cells treated with tumor necrosis factor. Mol Endocrinol 6: , Wu Z, Rosen ED, Brun R, Hauser S, Adelmant G, Troy AE, McKeon C, Darlington GJ, and Spiegelman BM. Cross-regulation of C/EBP alpha and PPAR gamma controls the transcriptional pathway of adipogenesis and insulin sensitivity. Mol Cell 3: , Wu Z, Xie Y, Bucher NL, and Farmer SR. Conditional ectopic expression of C/EBP beta in NIH-3T3 cells induces PPAR gamma and stimulates adipogenesis. Genes Dev 9: , Xing H, Northrop JP, Grove JR, Kilpatrick KE, Su JL, and Ringold GM. TNF alpha-mediated inhibition and reversal of adipocyte differentiation is accompanied by suppressed expression of PPARgamma without effects on Pref-1 expression. Endocrinology 138: , Yu S, Matsusue K, Kashireddy P, Cao WQ, Yeldandi V, Yeldandi AV, Rao MS, Gonzalez FJ, and Reddy JK. Adipocyte-specific gene expression and adipogenic steatosis in the mouse liver due to peroxisome proliferator-activated receptor gamma1 (PPARgamma1) overexpression. J Biol Chem 278: , !

110 64. Zhou Z, Yon Toh S, Chen Z, Guo K, Ng CP, Ponniah S, Lin SC, Hong W, and Li P. Cidea-deficient mice have lean phenotype and are resistant to obesity. Nat Genet 35: !

111 Acknowledgements We thank Dr. Han Fei Ding (University of Toledo Health Science Campus) for advice on apoptosis assays, and David Sowa for excellent technical assistance. We also thank Dr. W.A. Maltese (University of Toledo Health Science Campus) for the epitope tagged CVM expression vectors and Dr. David Giovannucci (University of Toledo Health Science Campus) and members of his laboratory for contributing their expertise in confocal microscopy. Disclosures None Grants Supported by an institutional monies from University of Toledo Health Science Campus. 104!

112 Figure Legends Figure 1. Northern Blot Assessment of FSP27 Transcript Upregulation in Multiple Adipogenesis Models. A. 3T3-L1 cells. RNA was harvested from 3T3-L1 preadipocytes (day 0) and at daily intervals through day 5 post-induction of adipogenesis. B. Rat primary cultures. Primary cultures of rat preadipocytes were harvested for RNA preparation at day 0 and at 2, 5, and 7 days post-adipogenic induction. C. Murine brown preadipocytes. RNA was harvested from cultures as preadipocytes at day 0 and at 3 and 8 days post-induction of adipogenesis. D. In vivo adipose tissue fractions. RNA was harvested from stromal-vascular (SV) fraction or adipocyte (Ad) fraction cell populations prepared from C57BL/6 murine WAT and BAT. E. Analysis of primary human preadipocytes (P) and from in vitro differentiated human adipocytes (A). For each of the above, Northern blot analysis was performed on 5 mg total RNA with the indicated random-primed 32 P-labeled probes and EtBr staining of rrna is shown below the autoradiogram. For A, B, and C, the numbers above the lanes indicate days. In the case of A-E, all data shown in each row of the boxed panel was obtained from the same original northern blot, however some data lanes were either removed or rearranged for economy and/or clarity of presentation. Figure 2. Concentration and Temporal Effects of TNF! on FSP27 and Select Adipocyte-Enriched Transcripts in 3T3-L1 Adipocytes. A. 3T3-L1 adipocytes were treated with the indicated concentration of TNF! for 24 h and FSP27, ATGL, resistin (Retn) and SCD1 transcript levels were analyzed by Northern blot using random-primed 32 P-labeled probes. B. 3T3-L1 adipocytes were treated with 10 ng/ml of TNF! for the 105!

113 indicated time points. 5 mg of total RNA was analyzed by Northern blot using randomprimed 32 P-labeled probes for FSP27, PPARg, ATGL, Retn, and SCD1 probes. EtBr staining of rrna, shown below the autoradiogram, was used to assess gel loading. Dose response and time course studies were conducted in duplicate and representative data shown. In the case of A and B, all data shown in each row of the boxed panel was obtained from the same original northern blot, however some data lanes were either removed or rearranged for economy and/or clarity of presentation. Figure 3. Intracellular Signaling Pathways Involved in TNF!-Mediated Downregulation of FSP27 Transcript in 3T3-L1 Adipocytes. A. 3T3-L1 adipocytes were pretreated for 1 h with DMSO vehicle, PD98059 (PD, 50 µm), LY (LY, 50 µm), and rapamycin (Rap, 1 µm) before addition of 10 ng/ml TNF! for 16 h or without TNF! addition. RNA was analyzed for FSP27 and 36B4 transcripts by Northern blot using random-primed 32 P-labeled probes. EtBr staining of rrna is shown below the autoradiogram. Data shown in each row of the boxed panel was obtained from the same original northern blot, however some data lanes were either removed or rearranged for economy and/or clarity of presentation. The experiment was carried out two times, and representative data is shown. B. Graphical representation of the response of FSP27 transcript to pharmacological inhibitors of intracellular signaling. Northern blot analyses were conducted as in A and the FSP27 transcript level for each sample normalized against its 36B4 control by phosphorimager analysis to correct for variations in sample loading. The numbers shown below the graph are the average change in FSP27 transcript level per treatment group, expressed as percent of the value for the DMSO vehicle 106!

114 control untreated with TNF!. Data shown is from triplicate studies conducted studies (represented by dark, medium and light gray shaded bars), however one of the +TNF!/+LY was lost in processing. The vertical dashed line demarcates data from two wholly separate studies with the indicated inhibitors, either for the effects of PD treatment (to left of line) or for the effects of LY and Rap treatment (to right of line); as such the left and right regions of the graph each have their own respective control values. The data in A and B derive from wholly distinct sets of studies. Figure 4. Concentration and Temporal Effects of Insulin on FSP27 Transcript Level in 3T3-L1 Adipocytes. A. 3T3-L1 adipocytes were incubated in serum-free medium for 16 h (time 0) or in regular growth media (R), at which time serum-free cultures were further incubated with the indicated concentration of insulin for 36 h. EtBr staining of rrna, shown below the autoradiogram, was used to assess gel loading. B. 3T3-L1 adipocytes were incubated in serum-free medium for 16 h (designated as 0) or in regular serum-containing growth media (R), at which time serum-free cultures were further incubated for the indicated times in the presence of 100 nm insulin ((+) insulin) or in its absence (SF). Northern blot hybridization was conducted using random-primed 32 P- labeled probe for FSP27. EtBr staining of rrna, shown below the autoradiogram, was used to assess gel loading. Dose response and time course studies were carried out two times and representative data shown. All data shown in each row of the single boxed panel was obtained from the same original image of a Northern blot, however in the case of (B) data lanes were either removed or rearranged for economy and/or clarity of presentation. 107!

115 Figure 5. Intracellular Signaling Pathways Involved in Insulin-Mediated Increase of FSP27 Transcript Level in 3T3-L1 Adipocytes. A. 3T3-L1 adipocytes were serumstarved for 6 h and pretreated for 1 h with DMSO vehicle, PD98059 (PD, 50 µm), LY (LY, 50 µm), and rapamycin (Rap, 1 µm) at which time they were subject to continued incubation under these conditions, in either the presence or absence of 100 nm insulin for an additional 16 h. 5 &g of total RNA was analyzed by Northern blot for FSP27 and 36B4 transcript using random-primed 32 P-labeled probes. EtBr staining of rrna is shown below the autoradiogram. All data shown in each row of the single boxed panel was obtained from the same original image of a Northern blot, however some data lanes were either removed or rearranged for economy and/or clarity of presentation. The experiment was carried out two times, and representative data is shown. B. The FSP27 transcript expression level was quantitated as described for Figure 3. The numbers shown below the graph are the average change in FSP27 transcript level per treatment group, expressed as percent of the value for the DMSO vehicle control untreated with insulin. Data shown is from three independently conducted studies (represented by dark, medium and light gray shadings), however one of the (+) insulin and one of the (-) insulin samples was lost in processing. The vertical dashed line demarcates data from two wholly separate studies with the indicated inhibitors, either for the effects of PD treatment (to left of line) or for the effects of LY and Rap treatment (to right of line); as such the left and right regions of the graph each have their own respective control values. The data in A and B derive from wholly distinct sets of studies. 108!

116 Figure 6. Expression of FSP27 Transcript in Wild Type Murine Tissues. A. Northern blot analysis was conducted on indicated tissues using random-primed 32 P-labeled probes for FSP27, Cidea, and Cideb. All data shown in each row of the boxed panel was obtained from the same original image of a Northern blot, however, some data lanes were either removed or rearranged for economy and/or clarity of presentation. B. Real-time PCR analysis of expression of FSP27 in murine tissues. Sem. Tubule, seminiferous tubule; Sal. Gland, salivary gland. Figure 7. Expression of FSP27 Transcript in ob/ob Murine Tissues. A. Wild type C57BL/6 (wt) and obese (ob/ob) WAT. Level of FSP27 transcript in subcutaneous WAT (A), or liver and kidney (B), or of Cidea in liver (C), was assessed in individual wild type and ob/ob mice by Northern blot analysis. EtBr staining of rrna is shown below the respective autoradiogram. For B and C, blots were rehybridized to 36B4 as an internal control. Separate lanes represent tissue from different individual mice, with a minimum of three mice per genotype. All data shown in each row of the boxed panels were obtained from the same original image of a Northern blot, however, some data lanes were either removed or rearranged for economy and/or clarity of presentation. Figure 8. Apoptosis Induced by Murine FSP27 Results in Cleavage of PARP and!- Fodrin. A. DNA fragmentation induced by FSP T cells transfected with 2 µg FSP27 expression construct or empty vector (EV) were harvested at indicated time points, genomic DNA was prepared and analyzed for fragmentation. A control apoptotic 109!

117 ladder (L) is shown to the right side of the panel and DNA marker (M) is shown in the left-most lane. B. 293T cells were transfected with 2 µg FSP27 expression construct or empty vector (EV) and total protein harvested at the indicated time points posttransfection. Protein was subjected to SDS/PAGE and analyzed by Western blot for full length and cleaved PARP, full length and cleaved!-fodrin and for "-tubulin. C. Time course of FSP27-induced cell death. 293T cells were co-transfected with 0.1 µg of a LacZ expression construct and either 1.90 µg of empty vector (EV) or an FSP27 expression construct. At the indicated hours post-transfection cells were fixed and stained for "- galactosidase activity. Representative stained culture dishes are shown, where blue color indicates LacZ positive cells. D. Dose response of FSP27-mediated cell death. 293T cells were transfected with an expression construct for murine FSP27 or empty vector (EV) and a LacZ expression construct. A total of 2 µg of DNA was used per transfection comprised of 0.01 µg of a LacZ expression construct and either empty vector (0) or an FSP27 expression construct, where the amount of FSP27 plasmid (second column) is 2 µg, and the 5 subsequent columns to the right represent transfections using decreasing mass of FSP27 expression that are 1/2, 1/4, 1/8, 1/16 and 1/32 of 2 µg. At 48 h posttransfection, blue cells were enumerated as described in Materials and Methods, and expressed as percent of the empty vector transfectants (set at 100%). Data shown is the average value of duplicate transfections. The images of culture dishes below the graph are for the 2 µg transfection (left) and the empty vector transfection (right). E. Level of FSP27 transcript in transfected 293T cells compared to adipocytes. Real-time PCR was carried out on total RNA of 293T cells transfected with empty vector (EV) or with the mouse FSP27 expression construct, and harvested at the indicated times (in hours) post- 110!

118 transfection. A longer-term study is shown in the upper panel, and a shorter-term study in the lower panel. For the upper panel the level in 9 h EV sample was set to a value of 1. For the lower panel the level in the 4 h EV sample was set to a value of 1. The plotted data represents FSP27 transcript level after correction for the transfection efficiency of the 293T cell population and the percentage of adipocytes in either the whole WAT (WAT) or differentiated 3T3-L1 (3T3-L1) cell populations, as described in the text. In the case of A and B, all data shown in each single inclusive panel was obtained from the same original image of an agarose gel (A), or western blot exposure (C), however, some data lanes were either removed or rearranged for economy and/or clarity of presentation. Figure 9. Assessment of FSP27 Apoptotic Activity and Protein Expression in the Adipocyte Lineage. A. DNA fragmentation assay in 3T3L1 preadipocytes and adipocytes. The left panel shows genomic DNA prepared from duplicate cultures of preadipocytes (P) or adipocytes (A) and assessed by electrophoresis through 1.2% agarose gels. A control apoptotic ladder (L) is shown to the right side of the panel and DNA marker (M) is shown in the left-most lane. Data shown in this panel was obtained from the same original image of an agarose gel, however some data lanes were either removed or rearranged for economy and/or clarity of presentation. B. Western blot analysis of 30 µg cell lysate of 3T3-L1 preadipocytes (Pre) or adipocytes (Ad) with FSP27 antibody. C. Western blot analysis of cell lysates of 30 µg of cell lysates from transfection of COS cells with a double epitope tag FSP27 expression construct, bearing an N-terminal HA tag and a C-terminal FLAG tag. Both panels shown are from a single 111!

119 transfer where the membrane was cut vertically such that one portion was subject to Western blot with anti-ha antibody and the other with anti-flag antibody. Figure 10. Cell Death Effects of FSP27 in 3T3-L1 Preadipocytes. A. Cell death assay of 3T3-L1 preadipocytes (pre) and 293T cells (293T) of cells transfected with empty vector (EV) or a murine FSP27 expression construct (1.5 µg) in combination with a LacZ expression construct (0.5 µg). Cells were enumerated as described in Materials and Methods, and cell viability in empty vector transfectants was set to 100% (*, p< 0.05). B. Image of microscopic field of "-galactosidase stained cells of preadipocytes (left panel) or CHO cells (right panel) transfected with 0.5 µg of a lacz expression construct in combination with 3.5 µµg of empty vector (EV) or an FSP27 expression construct. Cells were stained at 48 post-transfection. C. 3T3-L1 preadipocytes were transfected with an egfp-fsp27 fusion construct and observed by fluorescence (dark panels) for egfp signal and Hoechst-stained nuclei, or by and phase microscopy as indicated. Black arrow indicates typical morphology of transfected cells at 48 h post transfection; white arrow indicates bright Hoechst staining indicative of condensed chromatin. Supplemental Figure Legends Supplemental Figure 1. Confocal microscopy analysis of egfp-fsp27 transfected COS cells stained with MitoTracker Red. A. Localization of egfp-fsp27 and MitoTracker Red signals in COS cells transfected with egfp-fsp27 expression construct. Images were collected as described in Methods, with signal for egfp, 112!

120 MitoTracker Red and the merged signals shown in the left, middle and right panels, respectively. The upper panel indicates the projected image of 18 Z-sections of 0.5 microns each. The lower panel shows three individual Z-sections. Live cells were analyzed at ~20 h post-transfection. B. Similar study conducted in transfected 3T3-L1 preadipocytes. Supplemental Figure 2. Confocal microscopy analysis of egfp-fsp27 and pdsred2-mito transfected COS cells. Localization of egfp-fsp27 and pdsred2- Mito signals in transfected COS cells. Images were collected as described in Methods, with signal for egfp, DSRed2-Mito, and the merged signals shown in the left, middle and right panels, respectively. The upper panel indicates the projected image of 12 Z- sections of 0.5 microns each. The lower panel shows three individual Z-sections. Methanol-fixed cells were analyzed at ~20 h post-transfection. Supplemental Figure 3. Confocal microscopy analysis of mitochondrial morphology in egfp-fsp27 transfected COS cells. Upper panel shows signal for egfp-fsp27 in two of the cells in the microscopic field. Lower panel shows result of MitoTracker Red signal in these cells (left portion of image), and in two adjacent non-transfected cells (right portion of image). Live cells were analyzed at ~20 h post-transfection. 113!

121 114!

122 115!

123 116!

124 117!

125 118!

126 119!

127 120!

128 121!

129 122!

130 123!

131 124!

132 125!

133 126!

134 MANUSCRIPT #2 Functional Analysis of FSP27 Protein Regions for Lipid Droplet Localization, Caspase-Dependent Apoptosis and Dimerization with CIDEA Kun Liu #, Shengli Zhou #, Ji-Young Kim #, Kristin Tillison #, David Majors #, David Rearick #, Jun Ho Lee #, Ruby F. Fernandez-Boyanapalli #, Katherine Barricklow^, M. Sue Houston^ and Cynthia M. Smas #* ( Denotes equal co-authorship) Am J Physiol Endocrinol Metab 297: E , These authors contributed equally to the work. # Department of Biochemistry and Cancer Biology and Center for Diabetes and Endocrine Research, The University of Toledo College of Medicine, Toledo, OH USA ^ Food and Nutrition Program, School of Family and Consumer Sciences, Bowling Green State University, Bowling Green OH USA Running Head: Functional Analysis of FSP27 Protein Regions Please address reprint requests to: Cynthia M. Smas, D.Sc. (* Corresponding author) Department of Biochemistry and Cancer Biology and Center for Diabetes and Endocrine Research, The University of Toledo College of Medicine, Toledo, OH USA 127!

135 Phone: FAX: !

136 Abstract The adipocyte-specific protein FSP27 is one of three cell death-inducing DFF45- like effector (CIDE) proteins. The first known function for CIDEs was promotion of apoptosis upon ectopic expression in mammalian cells. Recent studies in endogenous settings demonstrated key roles for CIDEs in energy metabolism. FSP27 is a lipid droplet associated protein whose heterologous expression enhances formation of enlarged lipid droplets and is required for unilocular lipid droplets typical of white adipocytes in vivo. Here we delineate relationships between apoptotic function and lipid droplet localization of FSP27. We demonstrate ectopic expression of FSP27 induces enlarged lipid droplets in multiple human cell lines, indicative that its mechanism involves ubiquitously present, rather than adipocyte-specific, cellular machinery. Furthermore, promotion of lipid droplet formation in HeLa cells via culture in exogenous oleic acid offsets FSP27- mediated apoptosis. Using transient co-transfections and analysis of lipid droplets in HeLa cells stably expressing FSP27, we show FSP27 does not protect lipid droplets from action of ATGL lipase. Domain mapping with egfp-fsp27 deletion constructs indicates lipid droplet localization of FSP27 requires amino acids of its CIDE-C domain. The apoptotic mechanism of FSP27, which we show involves caspase-9 and mitochondrial cytochrome c, also requires this 19 AA region. Interaction assays determine the FSP27 CIDE-C domain complexes with CIDEA and Western blot reveals FSP27 protein levels are reduced by co-expression of CIDEA. Overall, our findings demonstrate the function of the FSP27 CIDE-C domain and/or regions thereof for apoptosis, lipid droplet localization and CIDEA interaction. 129!

137 Introduction Obesity and its related co-morbidities are approaching epidemic levels (1, 9). The development of new therapeutic inroads to treat these conditions would be greatly facilitated by a full understanding of lipid homeostasis (35, 36). Lipotoxicity is a highly detrimental outcome of the obese state, leading to derangement of cell function and/or cell death in various tissues (34, 35, 37). White adipocytes present in white adipose tissue are the major sites storage of excess energy in the form of triacylglycerol, contained within intracellular lipid droplets (7, 38). Efficient storage of excess fatty acids within adipocyte lipid droplets also serves to protect other cells and tissues from their lipotoxic effects (7, 38). Lipid droplets are highly dynamic organelles consisting of a neutral lipid core, a phospholipid monolayer, and a large number of lipid-droplet associated proteins (7, 38). The role for the vast majority of these proteins in lipid droplet function is undetermined. While most cells of the body are thought to contain small lipid droplets that serve to sequester fatty acids and to meet ongoing energy needs, the white adipocyte is unique in that nearly all its cell volume is filled by a large unilocular lipid droplet. The Cell death-inducing DFF45-like Effector (CIDE) protein family consists of three ~22-27 kda proteins, FSP27, CIDEA, and CIDEB; each are newly recognized lipid-droplet associated proteins with key roles in lipid homeostasis and energy balance (14-16, 21, 26, 41, 42). Curiously, the first described function of CIDEs was promotion of apoptosis and FSP27, CIDEA and CIDEB each exerts robust apoptotic activity upon ectopic expression in mammalian cells (5, 8, 11, 13, 17). CIDE proteins have a region of amino acid sequence homology in their N-terminal halves, termed the CIDE-N domain, that is also present in the major proapoptotic nuclease DFF40 and its inhibitory partner 130!

138 protein DFF45 (11). A CIDE-C domain, present in their C-terminal halves, is found only in FSP27, CIDEA and CIDEB (11). While the physiological role of CIDE-induced apoptosis remains undetermined, recent studies for each of FSP27, CIDEA, CIDEB have greatly illuminated the role of endogenous CIDE proteins and indicate they have crucial roles in lipid metabolism (14-16, 24, 33, 42, 44). Gene knockout has demonstrated that FSP27 is requisite for formation of the unilocular lipid droplet that typifies white adipocytes in vivo. In vitro knockdown of FSP27 in adipocytes results in an apparent fragmentation and/or a failed fusion of lipid droplets to result in cells with many markedly smaller lipid droplets and that evidence enhanced lipolysis (21, 24). FSP27 and CIDEA are expressed only in adipocytes, with distinctions noted for transcript expression in human vs. mouse. FSP27 transcript is expressed in both human and murine white adipocytes (13, 25), and is present at lower levels in murine brown adipocytes (13). In mice, CIDEA is found only in brown adipocytes (13, 44) while in humans a high level of CIDEA is noted in white adipocytes (26). CIDEB transcript is markedly enriched in human and murine liver with expression also reported for murine kidney and intestine (11, 13, 15). Earlier reports described a mitochondrial localization for FSP27, CIDEA and CIDEB (5, 17, 44). However recent studies strongly suggest that this localization was incorrect and that FSP27, CIDEA, and CIDEB are lipid droplet associated proteins (12, 21, 25, 33, 42). Signal for an FSP27-eGFP fusion protein has been shown colocalized to lipid droplets in 3T3-L1 adipocytes and lipid-loaded 293T cells (12). Immunostaining for endogenous or ectopically expressed FSP27 localized it to lipid droplets of 3T3-L1 adipocytes and cultured human white adipocytes (21, 25). Ectopically expressed CIDEA- 131!

139 egfp localized to lipid droplets in 3T3-L1 adipocytes and COS cells and endogenous CIDEA protein is found at lipid droplets of cultured human white adipocytes and cultured murine brown adipocytes (26). Most recently, ectopically expressed CIDEB was shown to be localized to lipid droplets of lipid-loaded hepatocytes and the regions of the CIDEB protein that governed lipid droplet localization examined (42). Ectopic expression of an HA-tagged or egfp fusion construct containing amino acids of CIDEB was sufficient for lipid droplet targeting in lipid-loaded COS cells and HepG2 hepatocytes, respectively (42). CIDEA and CIDEB are also reportedly localized to the endoplasmic reticulum (ER), an organelle from which biogenesis of intracellular lipid droplets initiates (27, 42). Moreover, in addition to being lipid droplet localized proteins, per se, ectopic expression of FSP27 and CIDEA has been demonstrated to promote the formation and/or enlargement of lipid droplets in several non-adipocyte cell types, a phenomenon that is particularly evident with addition of exogenous oleic acid to culture media. This has been shown for FLAG-tagged FSP27 in 3T3-L1 preadipocytes (12) and for egfp-fsp27 in 3T3-L1 preadipocytes, 293T and COS cells (12, 25). For CIDEA, FLAG-tagged CIDEA has been demonstrated to promote lipid droplet formation in 3T3-L1 preadipocytes (12). egfp-cidea has been reported to enhance lipid droplet size in lipid-loaded 3T3-L1 preadipocytes and lipid-loaded COS cells (26). On the other hand, FSP27 knockdown reduced the size and increased the number of smaller lipid droplets in 3T3-L1 adipocytes (12, 25) and human adipocytes (21). Analysis of FSP27 null mice indicates that FSP27 facilitates efficient energy storage in WAT by promoting formation of unilocular lipid droplets to restrict lipolysis (21). FSP27 null mice are refractory to diet-induced obesity and insulin resistance (21, 132!

140 24, 33). They have markedly reduced WAT mass that evidences some features of BAT including increased mitochondrial biogenesis and enhanced "-oxidation, with FSP27 null white adipocytes containing smaller, multilocular lipid droplets (21, 33). On the other hand, transgenic expression of FSP27 in murine liver results in hepatosteatosis (20). CIDEA null mice exhibit elevated lipolysis in BAT and resist diet induced obesity (44). A recent report demonstrated that one mechanism of CIDEA action is via its interaction with the major metabolic regulator AMPK that reduces AMPK protein level by enhancing proteosome-mediated AMPK degradation (27). The phenotype of CIDEA null mice was initially attributed to the ability of CIDEA to regulate thermogenesis via interaction with UCP1 to inhibit UCP1 activity (44). Given that the initial mitochondrial localization reported for CIDEA appears to not be the case (27), CIDEA interaction with UCP1 remains to be fully clarified. CIDEB interacts with apob and promotes the formation of triacylglycerol-enriched VLDL particles (42). CIDEB null mice display decreased plasma triglycerides and free fatty acids and are refractory to diet induced obesity (15). Livers of CIDEB null mice have higher levels of triacylglycerols and lower VLDL secretion, with VLDL containing less triacylglycerol (42). Studies in humans have reported that the level of FSP27 and CIDEA transcript are higher in the WAT of obese insulin sensitive persons compared with WAT of obese insulin resistant persons. This indicates a possible positive protective effect of elevated WAT FSP27 and CIDEA expression in the relationship between fat mass and the detrimental impact of obesity on systemic metabolism in humans (25, 26). Additionally, partial lipodystrophy, alteration in WAT morphology to multilocular lipid droplets, and insulin resistant diabetes has recently been reported for a patient with a homozygous nonsense mutation in 133!

141 FSP27/CIDEC that generates a truncated protein form largely lacking the CIDE-C domain (23). Taken together, studies to date indicate that endogenous FSP27 protein has a primary role in lipid droplet formation and energy balance. On the other hand, when FSP27 is expressed outside of the lipid droplet context it manifests proapoptotic activity. In this report we provide a detailed examination of the apoptotic mechanisms of FSP27 action and conduct structure-function analysis of FSP27 in regard to regions of the protein involved in lipid droplet localization, apoptotic response, and interaction with CIDEA. Materials and Methods Cell Culture and Co-transfection-Based Assessment of Cell Death All cell lines were maintained in DMEM supplemented with 10% FBS. The pancaspase inhibitor Z-VAD-FMK (R&D Systems, Minneapolis MN) and the negative control peptide VA-FMK (BD Biosciences, San Jose CA) were used at 20 µm and added to cultures at the time of transfection. All transfections were carried out using Lipofectamine 2000 (Invitrogen, Carlsbad CA). For studies using combinations of expression constructs for FSP27 and dominant-negative caspase-9 (CS9DN) DNAs were co-transfected at the indicated mass ratios of plasmids and assessed at 48 h posttransfection. We utilized a "-galactosidase co-transfection assay as an indirect visual measurement of cell death; this protocol has been described in previous studies of apoptosis (6, 13). The "-galactosidase construct serves as a reporter to mark transfected 134!

142 cells, which are also co-transfected with a test "effector" plasmid(s), for example FSP27 or empty vector pcdna3.1. Cells which die are lost from cultures and therefore not counted among the LacZ+ blue cells. Comparison of the numbers of blue cells in cultures transfected with empty vector vs. those transfected with the effector plasmid(s) allow for detection of the degree of cell death. For experiments involving enumeration of "- galactosidase (LacZ) positive cells, co-transfections included 10 ng of a "-galactosidase expression construct. Transfections were done in triplicate and unless otherwise stated blue cells were counted at 48 h. For "-galactosidase staining cells were fixed for 5 minutes at room temperature in 0.5% glutaraldehyde in PBS. Following two PBS washes, cells were incubated in staining solution (2 mm MgCl 2, 5 mm K 3 Fe (CN) 6, 5 mm K 4 Fe(CN) 6, 1 mg/ml 5-bromo-4-chloro-3-indolyl-"-D-galactopyranoside (X-gal) in PBS) and incubated at 37 C for 4 h. After incubation, blue cells per microscopic field were enumerated with 10 independent randomly chosen fields analyzed per dish or well. Single factor ANOVA was used for statistical assessments. Assessment of DNA Fragmentation Genomic DNA was prepared using either an Apoptotic DNA-Ladder Kit (Roche Diagnostics, Nutley NJ) exactly per manufacturer directions or by manual preparation using standard methods. For the latter, cells were collected from the media and culture plates and subjected to low speed centrifugation. The pellet was resuspended in lysis buffer (20 mm EDTA, 5 mm Tris HCl ph 8.0, 0.5% SDS) and incubated on ice for 20 min. Insoluble material was removed by centrifugation and the supernatant extracted with phenol/chloroform. DNA was ethanol precipitated and the pellet resuspended in 135!

143 water and subject to RNase digestion. DNA was assessed by fractionation on 1.2% agarose gels, stained with ethidium bromide or SYBR green, visualized under UV illumination and photographed. All images shown accurately represent the original data, however minor adjustments to brightness and contrast were made to allow for better visualization. Images shown as a single panel in DNA fragmentation assays were run on the same agarose gel. However, in some instances lanes have been removed and/or rearranged for economy and clarity of presentation. Immunocytochemical Staining and Western Blot Analysis for Cytochrome c Release For immunocytochemistry, COS cells were grown overnight on laminin-coated coverslips in 6-well plates and transfected with 2 µg of DNA of the indicated egfp- FSP27 construct or empty egfp vector. At 20 h post-transfection, cells were fixed with cold methanol for 10 min. Coverslips were blocked by incubation in 0.1% BSA in PBS-T for 1 h and incubated at room temperature with monoclonal antibody for cytochrome c. Coverslips were washed 3 times for 3 min each with PBS. Secondary antibody was Alexafluor 568-conjugated goat anti-mouse. After three 3 min washes with PBS, nuclei were stained with 10 nm DAPI for 10 min, coverslips were mounted on glass slides, and images obtained. Negative controls showed no signals and consisted of egfp empty vector transfectants and immunocytochemical staining of egfp-fsp27 transfectants with secondary antibody only. Signals were documented using a Nikon Eclipse E800 fluorescence microscope equipped with a digital camera and image acquisition and merging was performed with Image-Pro Plus software (Media Cybernetics, Carlsbad, CA) or with an Olympus IX70 microscope using Spot Advanced software (Diagnostic 136!

144 Instruments, Inc., Sterling Heights MI). All images shown accurately represent the original data, however minor adjustments to brightness and contrast were made to allow for better visualization. Similar observations were observed in multiple microscopic fields and in duplicate studies, with representative data presented. For Western blot studies of cytochrome c release, COS cells were collected at 18 h post-transfection. Cytosolic fraction was prepared using a Mitosciences Cell Fractionation Kit, exactly per manufacturer instructions. Transfections and protein preparations were conducted in triplicate. Western blots were carried out using a 1:5000 dilution of anti-cytochrome c antibody (Mitosciences, MSA06) and a 1:2000 dilution of secondary antibody (Santa Cruz, SC-2005). GAPDH monoclonal primary antibody at 1:10,000 (Santa Cruz, SC-47724) and monoclonal ATP synthase! primary antibody was used at 1:1000 (Mitosciences, MS507). Antibody incubations were carried out for 1 h at room temperature. Blocking, washing and ECL are described in section entitled Other Western Blot Analysis and Co-immnoprecipitation. Digital images were obtained and data were quantified using FluorChem HD2 software and an Alpha Innotech Digital imaging system. Statistical analysis was by single factor ANOVA. Other Western Blot Analysis and Co-immunoprecipitation For all other western blot studies, exclusive of cytochrome c studies, cell lysates were harvested at 48 h post-transfection by lysis in TNN(+) buffer (10 mm Tris ph 8.0, 120 mm NaCl, 0.5% NP-40, 1 mm EDTA supplemented with a protease inhibitor cocktail). Lysates were incubated on ice for 30 min with intermittent vortexing, 137!

145 supernatant collected via centrifugation, and protein content determined (Bio-Rad, Hercules, CA). For co-immunoprecipitation experiments, 500 µg of protein extract was incubated with 20 µl of anti-flag M2-agarose affinity gel (Sigma-Aldrich, Minneapolis MN) and co-immunoprecipitation performed per manufacturer directions. For analysis of protein half-life, cells were treated with 100 µg/ml cycloheximide at 40 h posttransfection. For Western blot analyses, typically 50 µg of protein extract was fractionated on SDS-PAGE, followed by electroblotting onto PVDF membrane with M Tris/0.192 M glycine transfer buffer supplemented with 20% methanol. Membranes were blocked for 1 h in 5% non-fat milk in PBS containing 0.5% Tween 20 (PBS-T) followed by either 1 h incubation at room temperature, or overnight at 4 o C, with a 1:2,000 dilution of antibody to full-length p116 PARP, cleaved p89 PARP, cleaved p150!-fodrin, cleaved p37 caspase-9, cleaved p20 caspase-7, cleaved p19 caspase-3 (Cell Signaling Technology, Beverley MA) or egfp antibody (Covance Research Products, Berkeley CA). Secondary antibody was HRP-conjugated goat anti-rabbit (Bio- Rad), used at a 1:2000 dilution. All washes were conducted in PBS-T. Signal was detected by ECL Plus enhanced chemiluminescence (GE Healthcare, Waukesha WI) and exposure to X-ray film or using a FluorChem HD2 digital imaging system (Alpha Innotech, San Leandro CA). Signal intensity was quantified using AlphaEaseFC software (Alpha Innotech, San Leandro CA). For protein half-life studies, relative signal intensity was normalized to the time 0 time point, which was set to 100%. Excel software was used to generate a line fitting equation and half-life calculated. All images shown accurately represent the original data, however minor adjustments to brightness and contrast were made to allow for better visualization. Images shown in the same horizontal Western blot 138!

146 panel were run on the same gel and processed on the same membrane. However, in some instances lanes have been removed and/or rearranged for economy and clarity of presentation. Lipid Droplet Subcellular Localization and Lipid Droplet Formation Studies Expression constructs for native full-length FSP27, CIDEA, CIDEB and ATGL contained the complete open reading frame. Expression constructs for egfp full-length FSP27, CIDEA, and CIDEB, DSRed-CIDEA, or egfp CIDE-C or CIDE-N of FSP27 lacked the respective initiator methionine and were generated by PCR based cloning using sequence-verified I.M.A.G.E. cdna clones as template. The FSP27-N constructs contained AA and the FSP27-C constructs AA For egfp constructs of '173-FSP27 and '192-FSP27 an Erase-A-Base kit (Promega Corp., Madison WI), was employed according to manufacturer instructions. For the FSP27-pBABE-puro construct, the open reading frame of FSP27 was cloned by PCR into the pbabe-puro vector. All constructs were confirmed by full sequencing of inserts. For lipid droplet localization studies, HeLa cells were grown for 3 d in DMEM with 10% FCS supplemented with 400 µm BSA-complexed oleic acid to induce lipid droplet formation. Cells were then transfected with the indicated expression construct or corresponding empty vector. During transfection, DNA complexes were incubated with cells for 6 h in the absence of exogenous oleic acid, after which media was changed to DMEM with 10% FCS and 400 µm BSA-complexed oleic acid. At 16 h posttransfection, lipid droplets were stained with Nile Red by incubating cells for 15 minutes with 0.5 &g/ml Nile Red (Invitrogen, Carlsbad CA). Confocal documentation of egfp 139!

147 and Nile Red signals in live cells transfected with egfp-fsp27 or regions thereof used the resources of the Advanced Microscopy and Imaging Center at the University of Toledo Health Science Campus. Images were captured with a Leica TCS SP5 broadband confocal microscope (Leica, Mannheim, Germany) equipped with Argon-488 and diodepumped solid-state-561 laser sources and 63.0 x 1.40 N.A. oil immersion objective. A series of optical Z sections, 0.5 &M in thickness and totaling 5 6 &M, were collected and visualized as projection images using Leica LAS software. Laser intensities and microscope settings between samples were maintained constant. Other imaging studies of lipid droplet localization used an Olympus IX70 microscope using Spot Advanced software (Diagnostic Instruments, Inc., Sterling Heights MI). For transient transfection studies of the effects of FSP27, CIDEA or CIDEB on lipid droplet formation across multiple cell lines, cells were used without preculture in exogenous oleic acid. HeLa, HT1080, ZR75, LNCaP, MG63 or U2OS cells were transfected with expression constructs for non-tagged native versions of FSP27, CIDEA or CIDEB in pcdna3.1, or empty pcdna3.1 vector, as indicated. Media was supplemented with 400 µm BSA-complexed oleic acid at 4 h post-transfection. Cells were documented at 18 h post-transfection by fixation in 4% formalin and staining with Oil Red O. For studies of effects of ATGL expression by transient co-transfection with FSP27, HeLa cells were transfected with a 1:5 mass ratio of expression construct for FSP27 and ATGL in pcdna3.1, or FSP27 and empty vector (pcdna3.1). At 4 h posttransfection, media was changed to include 400 µm of BSA-complexed oleic acid. Studies using ATGL expressed from the vector pires2-egfp (pires2-egfp-atgl) wherein cells transfected with ATGL could be tracked based on green signal were 140!

148 conducted in a similar manner except in this case a 5:1 ratio of FSP27 and ATGL constructs was used. Cells were analyzed at post-transfection for lipid content by photography, cell counting and flow cytometry (described below). Cell counting and flow cytometry studies were carried out in triplicate. Statistical analysis was by single factor ANOVA. For cell counting studies wherein ATGL was expressed from pcdna3.1, the number of cells with large lipid droplets per microscopic field were enumerated. For cell counting studies wherein ATGL was expressed as pires2-egfp-atgl, green cells per field were scored for the presence of large lipid droplets. For both of these counting studies, unstained live cells were examined with 10 individual fields analyzed per each of triplicate transfections, with a minimum of 100 cells analyzed per replicate. Data was expressed as a percent of cells, or green cells in the case of pires2-egfp-atgl, with lipid droplets. For studies of the effects of ATGL on preformed FSP27-induced lipid droplets, we prepared stable cell populations of HeLa cells expressing FSP27 via retroviral infection of a FSP27-pBABE-puro construct, with cells infected with empty vector pbabe-puro as a control population. Cells surviving following selection in 2 µg/ml puromycin were pooled for analysis. For assessment of ATGL effects we conducted studies using ATGL expressed in pcdna3.1 and studies using pires2-egfp-atgl. In either case, HeLa- FSP27 cells were pretreated with BSA-complexed 400 µm oleic acid for 1 d or 3 d, to induce lipid droplet formation, and then transiently transfected with either an ATGL expression construct (in pcdna3.1 or in pires2-egfp) or corresponding empty vector. Cells were analyzed at post-transfection for lipid content by photography, cell counting and flow cytometry (described below). Cell counting and flow cytometry 141!

149 studies were carried out on quadruplicate transfected samples. For cell counting studies wherein ATGL was expressed from pcdna3.1, the number of cells with large lipid droplets per microscopic field was enumerated, with at least 5 individual fields analyzed per each of quadruplicate transfections, and at least 100 cells scored for each replicate. For cell counting studies employing pires2-egfp-atgl, green cells per field were scored for the presence of large lipid droplets, with at least 10 fields and 100 total green cells analyzed per replicate. For both of these counting studies, unstained live cells were examined. Data was expressed as a percent of cells, or green cells in the case of pires2- egfp-atgl, with evident lipid droplets. For photographic documentation of lipid droplet formation, live cells were stained for neutral lipid with either Bodipy 493/503 (Molecular Probes, D-3922) or LipidTox Deep Red (H34477, Invitrogen Corp.). For this, cells on tissue culture plates were rinsed with PBS and stained for 10 minutes with 5 µm Bodipy 493/503 or a 1:2000 dilution of LipidTox Deep Red in PBS. Staining and incubations were at 37 o C. After staining, cells were washed once with PBS, followed by a 10 min incubation in PBS; this was then replaced with fresh PBS. Digital images were obtained with an Olympus IX70 microscope using Spot Advanced software (Diagnostic Instruments, Inc., Sterling Heights MI). All images shown accurately represent the original data, however minor adjustments to brightness and contrast were made to allow for better visualization. Similar observations were observed in multiple microscopic fields and in duplicate studies, with representative data presented. For flow cytometry analysis of neutral lipid content per cell, cells were stained with LipidTox Deep Red or Bodipy 493/503, as described above for photographic 142!

150 imaging. Cells were then trypsinized, washed with media in suspension, pelleted with low-speed centrifugation and resuspended in PBS. For each study 10,000 cells of independent quadruplicate transfections were analyzed using a BD Biosciences FACSCalibur flow cytometer using the Flow Cytometry Core Facility of the University of Toledo College of Medicine. Yeast Two-Hybrid Assessment of Protein-Protein Interactions Coding sequences for full-length FSP27, the CIDE-N or CIDE-C domains of FSP27, or full-length CIDEA were generated by PCR-based cloning and inserted into the EcoRI and BamHI sites of pgbkt7 to produce Gal4 DNA binding fusion constructs and of pgadt7 to produce the Gal4 activation domain fusion constructs. Reading frame across the vector-insert junction and the insert was fully sequence verified. The indicated pair-wise combinations of activation domain fusion and binding domain fusion constructs were co-transformed into S. cerevisiae yeast strain AH109. Co-transformants were selected following incubation for four d at 30 o C on -Leu, -Trp double dropout (DDO) media. Colonies from each pair-wise co-transformation were patched onto DDO dropout media plates, -His, -Leu, -Trp triple dropout (TDO) media plates, and TDO media plates containing the chromogenic substrate X-!-galactosidase. After 4 d of growth at 30 o C, growth patterns were assessed and digital images generated. All images shown accurately represent the original data, however minor adjustments to brightness and contrast were made to allow for better visualization. Images shown in a boxed area arose from the same 143!

151 agar plate. However in some instances lanes have been removed and/or rearranged for economy and clarity of presentation. Results and Discussion FSP27, CIDEA and CIDEB Promote Lipid Droplet Formation in Multiple Cell Types To date ectopic expression of FSP27 protein has been tested for promotion of formation of enlarged lipid droplets in just several non-adipocyte cell types. To further address the range of cells types wherein FSP27 and CIDEA can exert a lipid droplet enlargement phenotype we tested effects of transient expression of FSP27, CIDEA, or empty vector on six human cancer cell lines of various cellular origins including cervical (HeLa), fibrosarcoma (HT1080), breast cancer (ZR75), prostate cancer (LNCaP), and osteosarcoma (MG63, U2OS), Figure 1A. At 4 h post-transfection culture media was supplemented with 400 mm oleic acid and intracellular lipid stained with Oil Red O 14 h later. Transfection of pcdna3.1 empty vector resulted in accumulation of multiple tiny lipid deposits/droplets in oleic acid treated cells. In contrast, expression of FSP27 or CIDEA resulted in the robust accumulation of visibly enlarged lipid droplets that were first observed at ~ 10 h post-transfection. The lipid droplet enlargement phenotype was observed in each of the six cell lines tested. We also noted that the appearance of enlarged lipid droplets was concomitant with the disappearance of the tiny lipid deposits/droplets from within the same cell, suggestive that FSP27 and CIDEA proteins may act by mediating fusion of tiny lipid deposits/droplets to generate much lipid larger 144!

152 droplets. CIDEB was recently described as a lipid droplet associated protein in hepatocytes (42), one of several tissues that express high levels of endogenous CIDEB transcript. Although CIDEB has been shown to function in VLDL synthesis (42), the ability of ectopic CIDEB to promote lipid droplet formation in heterologous cell types (i.e. non-hepatocytes) has not been addressed. As the focus of our study was on the two CIDE proteins present in adipocytes, FSP27 and CIDEA, we assessed the lipid droplet promoting effects of CIDEB in only two cell lines, HeLa and HT1080. However, our data suggests that like CIDEA and FSP27, CIDEB also possesses a robust lipid droplet formation/enlargement activity. Overall, these observations point to the likely ability of FSP27 and CIDEA to promote formation of enlarged lipid droplets in most, and possibly all, cell types. This underscores the notion that their mechanisms of action in lipid droplet formation utilizes cellular machinery generally present in many cell types, rather than pathways and factors specific to adipocytes or other sites of endogenous CIDE protein expression. On the other hand, in the course of our studies we never observed lipid droplets of the strikingly large and unilocular morphology which characterize white fat cells in vivo in our studies with FSP27, suggesting that additional and possibly adipocytespecific proteins are needed for such. To date, two apparently disparate roles have been demonstrated for CIDE proteins, lipid droplet function and proapoptotic function. Recent in vitro and in vivo studies demonstrate a key role for FSP27 in lipid metabolism and support the idea that the primary function for endogenous FSP27 in vivo is the formation of large unilocular lipid droplets in adipocytes (21, 33). In the presence of oleic acid supplementation of media, we show herein that FSP27 promotes efficient packaging of triacylglycerol into 145!

153 enlarged lipid droplets in multiple cell types. On the other hand, in the same cell types, cell death ensues upon heterologous CIDE protein expression under standard culture conditions (i.e. lacking exogenous oleic acid supplementation). This is illustrated by the fact that for each of the cell lines examined in Figure 1A for the ability of FSP27 to promote lipid droplet formation, expression of FSP27 in the absence of exogenous oleic acid supplementation resulted in cell death (Figure 1B). Partitioning FSP27 Protein into Lipid Droplets Attenuates its Apoptotic Activity We have previously shown in studies with 293T cells that the level of ectopically expressed FSP27 transcript is similar to that found in mature fat cells (13). However, despite this high expression level of FSP27 in white adipocytes, we have failed to find evidence of basal apoptosis in 3T3-L1 adipocytes under normal culture conditions (13). This indicates that the induction of FSP27 gene expression that occurs during the normal adipogenic program, and which is concomitant with lipid droplet accumulation, does not result in increased cellular apoptosis. Given that FSP27 is a lipid droplet associated protein, we postulated that in the lipid droplet milieu of the adipocyte, FSP27 is unable to exert its apoptotic action. We therefore reasoned that by partitioning FSP27 into lipid droplets, we might be able to attenuate its proapoptotic action. To test this we assayed the ability of FSP27 to promote HeLa cell death in the absence and presence of exogenous oleic acid. We first determined the ability of FSP27 to localize to lipid droplets in HeLa cells, as has been reported for a few other nonadipocyte cell types. Confocal analysis in Figure 2A reveals that empty vector egfp 146!

154 shows uniformly cytoplasmic signal. On the other hand, there is clear localization of egfp-fsp27 fusion protein in a discrete ring-like signal at the surface of lipid droplets. The vast majority of visible egfp-fsp27 signal localizes with staining for intracellular lipid, in some cases to extremely small lipid droplets. Likewise, nearly all the Nile Red signal is coincident with that for egfp-fsp27, even in areas of the cell where clearly spherical lipid droplets are not yet apparent. We also used confocal analysis to determine localization of CIDEA and CIDEB to HeLa cell lipid droplets. CIDEA had been previously described to localize to lipid droplets in 3T3-L1 adipocytes and expression of egfp-cidea enhanced lipid droplet formation in two non-adipocyte cell types, 3T3-L1 preadipocytes and COS cells (12, 26). However, it was noted that in these non-adipose cells that most of the egfp- CIDEA signal was not colocalized with that for lipid droplets (26), but rather appeared as punctuate signals in the cytoplasm. As such, it was suggested that other, likely adipocyte differentiation-dependent proteins, are needed to enable a lipid droplet localization for CIDEA (26). The assessments by these investigators were conducted at 24 h post-transfection, with oleic acid added at 8 h posttransfection. Based on our prior experience we have found that, in media that is not supplemented with oleic acid, CIDEs induce robust apoptosis with morphological alterations visible beginning at ~20 h post-transfection (13). Close inspection of the various cell morphology data presented by these authors (26), suggests that the cells examined for CIDEA expression may have been already undergoing apoptosis. Our data for lipid-loaded HeLa cells in Figure 2A, obtained at ~16 h post-transfection clearly show that for both CIDEA and CIDEB nearly all the respective egfp-cide signal localizes to lipid droplets in this non-adipocyte cell type. During the preparation of this manuscript 147!

155 the lipid droplet localization for CIDEB, which had been previously unknown, was reported (42). This was assessed in CIDEB null hepatocytes, cells that normally express CIDEB, and in lipid-loaded COS cells (42). Our date demonstrates that the generalized cellular machinery present in HeLa cells is sufficient for lipid droplet localization of CIDEs, without the need for additional cell type specific proteins present in adipocytes in the case of CIDEA and FSP27 or in hepatocytes in the case of CIDEB. We next assessed the degree of FSP27-mediated apoptosis of HeLa cells in the presence and absence of exogenous oleic acid supplementation. Figure 2B shows that 40% cell death was observed in FSP27 induced-apoptosis in HeLa cells when cultured in regular growth media (i.e. no exogenous oleic acid). However, induction of lipid droplets in these cells via oleic acid supplementation of media, results in partially rescuing FSP27- mediated apoptosis such that only 20% cell death is observed. This suggests that physical localization of FSP27 at lipid droplets can inhibit its pro-apoptotic action. These data also indicate that the same region of FSP27 protein may be responsible for both apoptotic effect and lipid droplet localization. It is not currently known if FSP27 might undergo regulated dissociation from lipid droplets, however lipid droplets are highly dynamic organelles that possess multiple proteins associated with intracellular trafficking (10, 18, 38). ATGL Expression Inhibits FSP27-Induced Lipid Droplets Adipose triglyceride lipase (ATGL) has recently been demonstrated to be the first and rate limiting step in triacylglycerol hydrolysis. The sequential action of ATGL, 148!

156 hormone sensitive lipase and monoglyceride lipase result in release of energy from stored lipid droplets as fatty acid and glycerol (43). In addition to its role in adipocyte hormonesensitive triacylglycerol hydrolysis, ATGL has also been demonstrated to function in basal lipolysis (31). Ectopic expression of ATGL in oleic acid cultured HeLa cells significantly diminished triacylglycerol stores and the size of lipid droplets, whereas knockdown of ATGL under these culture conditions results in enhanced triacylglycerol accumulation and the formation of markedly larger lipid droplets (31). Although the mechanism(s) used by FSP27 to enhance lipid droplet size and triacylglycerol storage are not yet identified, observations to date support the general hypothesis of shielding the triacylglycerol in lipid droplets from hydrolysis by lipases rather that by stimulating lipogenesis. We examined the effect of ATGL expression on FSP27-mediated lipid droplet accumulation in HeLa cells cultured in the presence of oleic acid, wherein cells were subjected to co-transfection for concomitant expression of ATGL and FSP27. Cells were co-transfected with a 1:5 ratio of expression constructs for FSP27 and ATGL, or expression construct for FSP27 with empty pcdna3.1 vector and were supplemented with 400 µm BSA-complexed oleic acid upon media change at 4 h post-transfection. The next day cells were stained for neutral lipid content. The photomicrographs in Figure 3A show representative microscopic fields of these cultures stained with Bodipy 493/503 or LipidTox Red. Many of the cells co-transfected with FSP27 and pcdna EV evidence enlarged lipid droplets (right panels), however those co-transfected with ATGL have a dramatic reduction in numbers of cells with enlarged lipid droplets (left panels). The 1:5 ratio was chosen to ensure that if cells were transfected with FSP27, they likely also 149!

157 harbored the expression construct for ATGL. We have found that by using the maximum mass of FSP27 DNA for transfection, the vast majority of cells form enlarged lipid droplets. However the mass of FSP27 expression construct we could utilize for these transfections was constrained by the 1:5 ratio. As such some cells in the population escaped FSP27 transfection, and therefore did not form enlarged lipid droplets, as evidenced in the right panels of Figure 3A. The data for the effect of co-transfection of ATGL on FSP27-induced lipid droplet enlargement is quantified in Figure 3B, wherein the numbers of cells with large lipid droplets per microscopic field were enumerated. The percentage of cells with one or more large lipid droplets is reduced by 70% in the presence of ATGL co-transfection. To more specifically address the effect of ATGL on formation of FSP27-induced lipid droplets, we generated an ATGL expression construct wherein transfected cells were trackable by egfp signal. To eliminate concerns over alterations in ATGL action due to expression as a fusion protein, we chose not to use an ATGL-eGFP fusion for these studies. Rather we expressed ATGL using the internal ribosome entry site vector pires2-egfp, from a construct we designate as pires2-egfp-atgl. Co-transfection studies were conducted using pires2-egfp-atgl and FSP27. Cells were cotransfected, media was supplemented with 400 µm oleic acid at 4 h post-transfection, and cultures stained the next day for neutral lipid with LipidTox Deep Red. As ATGLexpressing cells could be tracked by egfp signal, this allowed us to use a 5:1 ratio of FSP27 vs. pires2-egfp-atgl. This enabled a higher degree of transfection efficiency in regard to FSP27, with a large majority of cells in the cultures co-transfected with FSP27 and pcdna empty vector demonstrating lipid droplet formation. In Figure 3C 150!

158 FSP27 transfectants that harbor empty vector pires2-egfp, enlarged lipid droplets are evident in nearly all the green cells (lower panels). On the other hand none of the green cells in the upper panel of Figure 3C evidence enlarged lipid droplets, whereas their neighboring non-green cells do. These results are quantified by cell counting in Figure 3D and by flow cytometry analysis of transfected egfp+ green cell populations for LipidTox Deep Red signal in Figure 3E. Figure 3D indicates a 78% reduction in the number of green cells with enlarged lipid droplets. Flow cytometry analysis in Figure 3E indicates a 60% reduction in the mean LipidTox Deep Red signal intensity for green cells harboring pires2-egfp-atgl vs. green cells harboring pires2-egfp empty vector. Together this data indicates that under conditions where FSP27 and ATGL are simultaneously co-expressed by co-transfection, ATGL is highly effective at inhibiting FSP27-mediated lipid droplet content. The above study utilized co-transfection of ATGL and FSP27. As such, it is possible that the inhibitory action observed for ATGL is due to its lipolysis-promoting effects on the tiny nascent lipid deposits/droplets that form in response to incubation with oleic acid, and whose formation is not dependent on FSP27 action. As these nascent droplets may serve as a substrate for FSP27-mediated lipid droplet enlargement, it is possible that ATGL inhibits the ability of FSP27 to exert lipid droplet enlargement by diminishing the cellular content of these nascent droplets. To more fully explore the relationship between ATGL and FSP27 in regard to lipid droplet content, we next addressed the effect of ATGL on lipid droplets that were preformed via FSP27 action. To do so we generated a stable cell population of FSP27-expressing HeLa cells using retroviral expression. Figure 4A compares lipid droplet content of these HeLa-FSP27 151!

159 cells with that of empty vector control cells, in the presence and absence of exogenous oleic acid. HeLa-FSP27 cells readily form small clearly visible and demarcated lipid droplets in the absence of exogenous oleic acid, and evidence enlarged lipid droplets when cultured in oleic acid. Although the control empty vector cells show multiple punctuate areas of Bodipy lipid staining, as is usually observed for naive HeLa cells under oleic acid culture conditions, they do not form clearly evident lipid droplets. To test the effects of ATGL on preformed FSP27-mediated lipid droplets, HeLa-FSP27 cells were induced to form lipid droplets by culturing for 24 h in 400 µm BSA-complexed oleic acid followed by transfection with an expression construct for ATGL. Bodipy 493/503 staining in Figure 4B shows the result of transient transfection with ATGLpcDNA compared to cells transfected with pcdna empty vector. While nearly all cells in the empty vector show enlarged lipid droplet content, many cells in the ATGLtransfected population lack evident lipid droplets. This is enumerated in Figure 4C, which shows a 25% reduction in cells with enlarged lipid droplets upon ATGL expression and flow cytometry analysis for Bodipy signal in Figure 4D shows a 28% reduction in mean signal intensity. As was done in studies in Figure 3, we also utilized pires2-egfp-atgl for these analyses. Figure 4E reveals that while all of the green cells in the pires2-egfp- EV population possess enlarged lipid droplets, as shown by LipidTox Deep Red staining, lipid droplet staining is dramatically reduced in the green cells harboring the pires2- egfp-atgl construct. Enumeration of green cells for lipid droplets in Figure 4F and the corresponding flow cytometry data for LipidTox Deep Red signal in Figure 4G, reveals a respective 80% and 63% decrease upon expression of ATGL. As shown in Figures 4H, 4I 152!

160 and 4J, we also conducted these studies using HeLa-FSP27 cells wherein FSP27- mediated lipid droplets had been preformed by a 3 d incubation in 400 µm BSAcomplexed oleic acid and found effects largely similar to those of the 1 d oleic acid incubation. Thus in our experimental cell culture model, whether tested by concomitant expression via cotransfection or in the context of lipid droplets pre-formed via FSP27- mediated action, FSP27 is not effective at protecting lipid droplets from the effects of ATGL. Caspase-Dependent Apoptosis by FSP27 Given that lipid droplet accumulation within cells diminished the cell death effect of FSP27, we next set out to further define its apoptotic mechanism. Our previous report on the detection of PARP and!-fodrin cleavage upon ectopic expression of FSP27 in mammalian cells suggested the involvement of caspase activation in the proapoptotic effects of FSP27 (13); however the caspase-dependence of FSP27-mediated apoptosis and other details of its apoptotic mechanism has not yet been examined. We first investigated effects of the pan-caspase inhibitor Z-VAD-FMK on FSP27-mediated apoptosis using transient expression in 293T cells, with inhibitor added at time of transfection. Extent of apoptosis was determined by cell death assay, DNA fragmentation and cleavage of PARP and!-fodrin. As shown in Figures 5A, 5B and 5C, Z-VAD-FMK effectively blocked the proapoptotic effects of FSP27. The level of cell death in the absence or presence of Z-VAD-FMK is 92% and 39%, respectively (p<0.01). As shown in Figure 5B, DNA fragmentation upon FSP27 transfection was completely inhibited by 153!

161 Z-VAD-FMK. In the Western blot in Figure 5C, the disappearance of full-length p116 PARP, the appearance of the p89 caspase cleavage product of PARP, and the p150 caspase cleavage product of!-fodrin are blocked by Z-VAD-FMK treatment. To address involvement of specific caspases in FSP27-mediated apoptosis, the levels of cleaved caspase-9, 7, and 3 were examined by Western blot. Figure 5D reveals that, compared to the empty vector controls, in cells transfected with an FSP27 expression construct the protein levels for these cleaved caspases are increased at each of the four posttransfection time points examined. To further validate the involvement of the caspase-9 mediated cell death pathways in FSP27-mediated apoptosis we utilized a wellcharacterized dominant negative form of caspase-9, CS9DN. CS9DN contains a single point mutation of C287A at the site of processing of procaspase-9 to the active cleaved form and has previously been demonstrated to be highly effective in inhibiting activation of endogenous caspase-9 (32). The left panel of Figure 5E shows that transfection of FSP27 alone resulted in 53% cell death. Co-transfection of the CS9DN expression construct effectively rescued this effect; these cultures evidenced only 13% cell death. Co-transfection of CS9DN also diminished the appearance of p89 cleaved PARP (Figure 5E, right panel). To determine if the caspase pathways activated by FSP27 during apoptosis involved mitochondria-mediated actions, we conducted immunocytochemistry analysis for localization of cytochrome c in FSP27-transfected cells. In order to correlate cytochrome c localization with FSP27 transfection on a cell-by-cell basis, we utilized an egfp-fsp27 expression construct allowing for visualization of FSP27 transfection with fluorescence microscopy. As a negative control, cells were transfected with empty egfp 154!

162 vector. egfp expression and cytochrome c immunostaining was assessed at 20 h posttransfection. Figure 5F reveals that cells expressing egfp-fsp27, and therefore green in appearance, show a diffuse cytoplasmic distribution of red signal for immunostained cytochrome c, indicative of release of cytochrome c from mitochondria. These cells are designated by arrows in the left middle panel. A non-transfected cell present in this same field is depicted by an asterisk and shows cytochrome c staining only in the distinctive spaghetti-like pattern that is typical of intact mitochondria. The right three panels of Figure 5F show an empty vector egfp transfected and a non-transfected cell, indicated by asterisks, in which the typical mitochondrial localization pattern for cytochrome c is observed. We quantified the release of cytochrome c to the cytosol by Western blot analysis, Figure 5G. COS cells were transfected with egfp-fsp27 or egfp empty vector and cytosolic fractions prepared and analyzed by Western blot for cytochrome c, as well as for the cytosolic marker GAPDH and the mitochondrial marker ATP synthase!. The Western blot data in the upper panels of Figure 5G reveal a dramatic increase of cytosolic cytochrome c level in cells transfected with egfp-fsp27 vs. empty vector. The GAPDH and ATP synthase! data panels reveal this is due to the effects of egfp-fsp27 rather than protein loading differences or any differential mitochondrial contamination of samples. The cytochrome c signal is quantified in the graph in the lower portion of Figure 5G. Overall, our studies indicate that a caspase-dependent mitochondrial-mediated mechanism is involved in FSP27-induced cell death. The role of caspases in the apoptotic mechanism of CIDEA and CIDEB remains to be fully explored. In the initial cloning report for CIDEA and CIDEB, it was indicated that apoptosis induced by transient 155!

163 expression of CIDEA in 293T cells was caspase independent due to the failure of caspase inhibitor, when added at 8 h post-transfection, to block CIDEA-mediated apoptosis (11). On the other hand, a study of CIDEB supported a caspase-dependent mechanism of action in that activity of caspase-3 and release of mitochondrial cytochrome c occurred upon transient expression of CIDEB in COS cells (8). It remains to be determined whether the apoptotic mechanism of CIDEA differs from that for FSP27 and CIDEB, or whether the early report of caspase-independence for CIDEA might be due to the rather late timing of addition of caspase inhibitor in that study (11). The Same Subregion of the FSP27 CIDE-C Domain Governs Apoptotic Activity and Lipid Droplet Localization We next addressed the region(s) of FSP27 responsible for apoptotic function and for lipid droplet localization, initially using egfp fusions for full-length FSP27, or containing the CIDE-N domain (FSP27-N), or the CIDE-C domain (FSP27-C) of FSP27, fused C-terminal to egfp. Figure 6A reveals that 70% cell death is observed for fulllength FSP27 and 65% for FSP27-C. In contrast, FSP27-N evidences 18% cell death. Thus the vast majority of the apoptotic effect of FSP27 is attributable to actions of its CIDE-C domain. In contrast to FSP27-N, expression of the FSP27-C leads to the appearance of marked DNA fragmentation (Figure 6B) and generation of the cleaved proteins for PARP,!-fodrin and for active cleaved forms of caspase-9, -7, and -3 (Figure 6C). Figures 6D and 6E confirms that FSP27-C mediated apoptosis, as we have shown herein for full-length FSP27, is caspase-dependent in that it is effectively inhibited by Z- 156!

164 VAD-FMK. This is both in regard to cleaved PARP and!-fodrin levels (Figure 6D) and DNA fragmentation (Figure 6E). While a slight increase in cell death effect was also found for FSP27-N in the assay used in Figure 6A, no evidence of FSP27-N mediated apoptosis was noted in regard to DNA fragmentation or by Western blot analysis of apoptotic markers (Figures 6B and 6C). We also observed, as we found for full-length FSP27, expression FSP27-C promoted release of mitochondrial cytochrome c (Figure 6F), but expression of FSP27-N did not. As was done for Figure 5G, we compared and quantified cytochrome c release for FSP27-N and FSP27-C by Western blot, Figure 6G. While a significant increase in cytosolic cytochrome c was found for FSP27-C vs. empty egfp vector, this was not observed for FSP27-N. We next used confocal microscopy to examine the ability of the FSP27-N or FSP27-C to localize to lipid droplets. egfp-fsp27, egfp-fsp27-n or egfp-fsp27-c were transfected into HeLa cells that had been pre-cultured in the presence of exogenous oleic acid to stimulate lipid droplet accumulation. Figure 6H clearly shows that signal for egfp-fsp27-n is distributed throughout the cytoplasm with no particular signal enrichment at lipid droplets. In contrast, egfp-fsp27-c shows distinct and clear localization with green signal that is specifically localized in a ring around the Nile Red stained lipid droplets. In order to better map the region of FSP27 governing its apoptotic activity and lipid droplet localization, two N-terminal deletion constructs of egfp-fsp27-c were generated, egfp-fsp27-'173 and egfp-fsp27-'192 (Figure 7A, upper panel). These were tested for apoptotic activity and lipid droplet localization. The lower panel of Figure 7A indicates that FSP27-'173 has robust cell death activity. In contrast, egfp-fsp27-157!

165 '192 had no discernable effects on cell viability. egfp-fsp27-'173 is capable of inducing DNA fragmentation (Figure 7B) and the generation of cleaved p89 PARP and active caspase-9,-7 and -3, as shown by the Western blot in Figure 7C. This is in marked contrast to egfp-fsp27-'192 which has only minor effects on chromosomal DNA integrity and no evident generation of apoptotic markers by Western blot. Apoptosis mediated by egfp-fsp27-'173 is caspase-dependent as the addition of Z-VAD-FMK abrogates the appearance of p89 PARP and activated caspases, as shown by Western blot in Figure 7D. These results suggest the 19 AA region from 173 to 191 is critical for the major apoptotic activity of FSP27 and that the mechanism of apoptosis that maps to this region appears to be the same as that for full-length FSP27 in that both are caspasedependent. We next assessed if the ability of the FSP27 deletions to induce caspasedependent apoptosis tracked with their lipid droplet localization, as we had demonstrated in Figure 6 in respect to FSP27-C. Figure 7E shows confocal localization of egfp- FSP27-'173 and egfp-fsp27-'192 assessed in transfected HeLa cells cultured in the presence of exogenous oleic acid, with lipid droplets stained with Nile Red. egfp- FSP27-'173 shows clear localization to lipid droplets. In contrast, egfp-fsp27-'192 shows a uniform distribution of egfp signal throughout the cytoplasm, as we previously noted for egfp empty vector and for egfp-fsp27-n. To obtain additional information on the amino acid sequences that may function as a discrete signal for FSP27 localization to lipid droplets, we generated a small peptide fused C-terminal to egfp which contained the 19 AA spanning from amino acid 173 to 191 of the FSP27 CIDE-C domain, termed egfp-fsp27-19aa. Analysis of HeLa cells cultured with exogenous oleic acid and 158!

166 transfected with egfp-fsp27-19aa shows signal for this fusion protein fails to localize to lipid droplets, appearing instead evenly distributed in throughout the cytoplasm, similar to that found for egfp-fsp27-'192. egfp-fsp27-19aa also failed to show evidence of cell death inducing activity (data not shown). Thus our data indicates that amino acids 173 to 191 of the FSP27 CIDE-C domain are necessary for both robust apoptosis and for localization to lipid droplets, but that this region alone is not sufficient for either effect. Our findings on the role of the CIDE-C domain of FSP27 are in line with that reported by MacDougald who observed that full-length FSP27 and an expression construct containing its CIDE-C domain promotes apoptotic morphology of 293T cells (12). These workers also showed that 3T3-L1 adipocytes expressing full-length FSP27 or the CIDE-C expression construct showed enhanced sensitivity to TNFa-mediated apoptosis as assessed by TUNEL staining (12). However no further indices or markers of apoptotic activity were assessed, nor was further mapping of these functions or those for regions of FSP27 governing lipid droplet targeting further examined. A recent report by Savage and coworkers have described partial lipodystrophy and insulin resistant diabetes in a patient with a homozygous nonsense mutation in FSP27/CIDEC at position 186 (23). This mutation truncates and disrupts the CIDE-C domain, further underscoring the role of this region in FSP27 function in WAT. It is of interest to note that this mutation occurs within that 19 amino acids stretch from 173 to 191 of the FSP27 CIDE-C domain that we show herein are required for FSP27 localization to lipid droplets. During the preparation of this manuscript, a study on the localization of CIDEB to lipid droplets was reported (42). This revealed that the region of CIDEB from 166 to 195 directed localization of an egfp-cideb fusion to lipid droplets in hepatocytes (42). In 159!

167 regard to protein sequence homologies between the three CIDE proteins, while they share homology within their respective CIDE-C domains, the regions C-terminal to their CIDE- C domains is unique to each respective CIDE protein. We postulate that the region of the FSP27 CIDE-C domain from amino acids 173 to 191 is involved in its lipid droplet localization; it is also likely that the homologous region governs lipid droplet localization for CIDEA. Assessment of Protein Interaction for FSP27 and CIDEA Studies to date indicate protein-protein interactions are important to CIDE protein function and both CIDEA and CIDEB have been studied somewhat in this regard. Homo and heterodimeric interaction of CIDEA and/or CIDEB (i.e., CIDEA:CIDEA, CIDEB:CIDEB and CIDEA:CIDEB) has been reported (5, 8, 11, 19, 27). CIDEB interacts with viral protein NS2 (8) and apob (42), and CIDEA with AMPK (27); such interactions appear to impact the physiological function of these CIDE partner proteins. In cases where protein regions for CIDEA and CIDEB mediated-interactions have been mapped, they involve the respective CIDE-C region (5, 8, 11, 27, 42). CIDEA and CIDEB were initially cloned based on sequence homology to the CIDE-N domain of DFF45, the inhibitory regulatory subunit of the major apoptotic nuclease DFF40. To our knowledge, direct interactions mediated by CIDE-N domains have only been detected in respect to interaction of CIDEB with the CIDE-N domains of DFF40 and DFF45, and between the respective CIDE-N domains for DFF40 and DFF45 (19). The physiological endogenous role of CIDEB-mediated apoptosis and its CIDE-N domain mediated interactions remain undetermined. No information exists regarding protein-protein interactions for FSP27. Since FSP27 and CIDEA are both present at lipid droplets of 160!

168 human white adipocytes (25, 26), this raised the possibility of interaction of CIDEA and FSP27. To investigate this we carried out co-transfection studies in 293T cells and assessed for FSP27-CIDEA interaction using co-immunoprecipitation. The upper panel of Figure 8A reveals interaction of full-length FSP27 with full-length CIDEA. Yeast two hybrid had previously been successfully employed and validated for assessment of CIDEB protein-protein interactions in regard to homo-dimerization, and CIDEB hetero-dimerization with CIDEA, AMPK and NS2 (8, 42). Technical issues with Western blots showing heavy signal arising from the immunoglobulin light chain precluded our further assessments by co-immunoprecipitation. We therefore utilized the yeast two hybrid protein-protein interaction method. For our analyses full-length FSP27, FSP27-N, FSP27-C and full-length CIDEA were expressed as Gal4 DNA binding domain and Gal4 DNA activation domain fusion proteins in yeast. Various pair-wise combinations of the pgbkt7 binding domain and the pgad activation domain-based expression constructs were transformed into AH109 strain S. cerevisiae. Effective cotransformation is illustrated by robust growth of yeast on DDO media lacking Trp and Leu (Figure 8B, top panel). The interaction of the indicated pairs of expressed proteins is assessed in the middle and lower panel of Figure 8B. Protein-protein interaction was scored by growth on TDO media lacking Trp, Leu and His and the ability to cleave the chromogenic substrate X-!-gal to produce blue-colored growth. Here we find the anticipated homodimeric interaction for CIDEA, which has previously been reported (27, 44). Figure 8B also shows that while full-length FSP27 does not evidence homodimerization, homo-dimerization is observed with the individual FSP27-N and FSP27-C constructs. We also find, as we had using co-immunoprecipitation, heterodimerization 161!

169 for CIDEA and full-length FSP27 and further demonstrate that FSP27-C functions in this interaction. FSP27-C interacts with full length CIDEA and full length FSP27, either when FSP27-C is expressed as a binding domain or activation domain fusion. In the intracellular milieu of the lipid droplet, it is possible that the proapoptotic CIDE-C region of FSP27 is kept in check via interaction with other lipid droplet proteins, perhaps including CIDEA. While the FSP27-N domain interacts with full length FSP27 when it is expressed as an activation domain fusion in pgad, this is not evidenced when FSP27-N is expressed as a binding domain fusion in pgbtk7. We have determined that full length FSP27 localizes to lipid droplets, as does FSP27-C domain, whereas FSP27-N domain does not. These observations and our yeast two hybrid data suggested that by interaction with full length FSP27, the FSP27-N domain might show lipid droplet localization. To test this we transiently co-transfected lipid-loaded HeLa cells with egfp-fsp27-n and non-tagged full length FSP27. As shown in Figure 8C, we observed multiple cells that showed a degree of egfp signal at a subset of lipid droplets, suggesting lipid droplet localization of egfp-cide-n via interaction with full length FSP27. We note, however, that this was observed in only 1-3% of green cells. The vast majority of green cells demonstrated diffuse cytoplasmic signal, as shown in Figure 6H. On the hand, no evidence of lipid droplet localization of egfp-fsp27-n was observed when cells were co-transfected with CIDEA in place of FSP27 co-transfection (data not shown). This is in line with our yeast two hybrid data showing no evidence of interaction for CIDEA with FSP27-N domain. Interestingly, we had previously reported that both ectopically expressed FSP27 162!

170 and FSP27 protein present in fat cells exists as multiple distinct species of size(s) consistent with that predicted to result from N-terminal truncations (13). A major shorter FSP27 protein species of ~14 kda that is consistent with the predicted mass of FSP27-C is present at readily detectable levels in 3T3-L1 adipocytes and FSP27 transfected COS cells (13). It is possible that in vivo not only does FPS27 exist as a full-length form but also as a processed/cleaved form that may generate FSP27 protein species containing CIDE-N or CIDE-C regions. Based on our data herein, these species of FSP27 protein would be predicted to demonstrate distinct activities in regard to apoptosis, lipid droplet localization and protein-protein interaction. Our demonstration that the same region of FSP27 required for specific subcellular localization at the lipid droplet coincides with that required for its apoptotic activity suggests that the localization of FSP27 in the unique molecular niche of the lipid droplet may concomitantly mask the subregion of FSP27 that is necessary for its apoptotic activity. By extension, disrupting these interactions, some of which may be mediated by amino acids of FSP27, may derail its lipid droplet association possibly allowing FSP27 to possibly exert other actions. Studies in several mammalian cell lines of the stability of ectopically expressed CIDEA protein have shown that it undergoes rapid proteasomal degradation with a halflife of less than 30 minutes (4). Moreover, the recently identified CIDEA interaction partner AMPK undergoes greatly enhanced proteasomal degradation as a result of its interaction with CIDEA (27). The rapid proteasomal degradation of CIDEA is largely governed by ubiquitination of a lysine at amino acid position 23 of the CIDEA protein (4). This is N-terminal to the CIDE-N domain and as such outside of the region of shared CIDE-N domain homology for CIDEA, CIDEB, and FSP27. A lysine does not appear 163!

171 conserved in this position or region of either CIDEB or FSP27, although other lysines are present. As we demonstrated interaction of CIDEA with FSP27, we investigated whether CIDEA might affect FSP27 protein stability. We first transfected 293T cells with CIDEA only, FSP27 only, or both constructs and assessed levels of respective protein expression at 48 h post-transfection. As is shown in the Western blot in Figure 9A, co-expression of FSP27 with CIDEA resulted in diminished levels of FSP27 protein compared to that seen when FSP27 alone is expressed. As CIDEA has been reported to be a short-lived protein, it is possible that since FSP27 can complex with CIDEA this may contribute to the reduced steady state level of FSP27 of protein we observed, possibly through effects on FSP27 protein half-life. To address this we utilized the protein synthesis inhibitor cycloheximide and used Western blot analysis to assess levels of FSP27 protein in the absence and presence of CIDEA protein expression in transfected 293T cells, a cell type previously employed in studies of CIDEA protein stability (4). As shown in the left panel of Figure 9B, and in the accompanying graph of FSP27 protein half-life measurements in Figure 9C (left panel), we find that FSP27 protein has a half-life of ~ 1.3 h. The middle panel in Figure 9B and the accompanying graph in Figure 9C (right panel), reveals that protein half-life of FSP27 is moderately reduced to ~1 h in the presence of CIDEA co-expression. The Western blot in the rightmost panel of Figure 9B compares the steady state level of FSP27 in the absence and presence of CIDEA, used as the time 0 point in the cycloheximide treatment study and is consistent with our observations in Figure 9A. Quantification of the steady state signal for FSP27 protein in the absence and presence of CIDEA indicates a 36% 164!

172 reduction of FSP27 protein level in the presence of CIDEA. This is consistent with the degree of effect we note for reduction of FSP27 half-life by CIDEA expression. The biological consequence of interaction of FSP27 and CIDEA remains to be investigated. As these proteins interact with each other, the possibility exists that they may each also interact with similar or the same subset of partner proteins at lipid droplets. Studies with the PAT lipid droplet family, the first and to date the best studied group of lipid droplet associated proteins which includes perilipin, adipophilin and TIP47 (2, 3, 7, 22, 39, 40), have revealed that specific PAT proteins are preferentially and/or exclusively associated with differing sizes of nascent through large lipid droplets within a single cell (2, 7). It has also been reported that expression of certain PAT proteins can cause the loss of lipid droplet association of other PAT proteins (22, 39). These observations support a working model for exchangeable PAT proteins in the structure, function and dynamic nature of lipid droplets. To begin to address the association of FSP27 and CIDEA among lipid droplets within a single cell, we examined distribution of these proteins using transient transfection of lipid-loaded HeLa cells utilizing a 1:1 ratio of egfp-fsp27 and DSRed-CIDEA to test for either co-localization or mutual exclusivity. As shown by the confocal analysis in Figure 10, we find that within single cells, FSP27 and CIDEA are both co-localized on the full size range of cellular lipid droplets. While the projected image in Figure 10A appears to show some enrichment of FSP27 at smaller lipid droplets, and CIDEA at larger droplets, close inspection of the individual Z-sections for either efgp-fsp27 or DSRed-CIDEA signal, shown in Figure 10B, indicates that all evident lipid droplets evidence association of both FSP27 and CIDEA. In addition FSP27 and CIDEA are co-localized in all other regions within the cells; these are presumably 165!

173 very tiny lipid droplets/deposits that do not yet have the evident clearly circular morphology of discernable lipid droplets. Thus unlike observations for particular PAT proteins, the two CIDE family proteins we tested, FSP27 and CIDEA, fail to demonstrate any degree of mutual exclusivity in respect their association with varying size lipid droplets. While the interplay between FSP27, CIDEA and lipid droplets remains to be more fully explored, our observations suggests that FSP27 cannot effectively displace CIDEA, nor vice versa. In conclusion FSP27 has dual functions; as a lipid droplet protein in cellular lipid metabolism and as robust proapoptotic factor (12, 13, 25). Our studies are the first to demonstrate interaction of CIDEA and FSP27, and that this interaction likely involves the FSP27 CIDE-C domain. To our knowledge, only a single study to date has addressed potential functional interaction for FSP27 and CIDEA, albeit in an indirect manner. It was reported that in cultured brown adipocytes sirna-mediated depletion of FSP27 had no apparent effects on CIDEA localization to the lipid droplet surface (26). However our findings raise the possibility that CIDEA and FSP27 may work synergistically via heterodimerization in regard to certain aspects of adipocyte metabolism and/or lipid droplet function. Our data also suggest that in studies addressing the role of FSP27 or CIDEA in cells where both proteins exist, for example human white adipocytes, that the impact of dimerization of FSP27 and CIDEA should be considered when assessing the respective functional roles of these proteins. It is currently not clear if the proapoptotic effect of FSP27 is merely a functional remnant of its evolutionary relationship with the major apoptotic nuclease DFF40/DFF45 (41). However it is intriguing that we demonstrate that the proapoptotic function and lipid 166!

174 droplet localization function of FSP27 requires a subregion of its CIDE-C domain, a protein motif that is unique to the three CIDE family members (FSP27, CIDEA, and CIDEB) and not present in DFF40/DFF45 (11, 41). In adipocytes wherein a high degree of lipolysis occurs, lipid droplet content is diminished. The pro-lipolytic agent TNF! results in loss of lipid content, diminution of transcript expression for a large number of adipocyte-expressed genes and can induce adipocyte apoptosis (28-30). It is not currently known if a portion of the lipid droplet proteome becomes released from the lipid droplet milieu as a result of TNF!-induced or other lipolysis. One can speculate that these freed lipid droplet proteins might be involved various events to impact or initiate intracellular signaling pathways. These may function to communicate status of cellular triacylglycerol energy stores, and by extension overall health of the adipocyte, to other organelles within the cell. 167!

175 References 1. Badman MK, and Flier JS. The adipocyte as an active participant in energy balance and metabolism. Gastroenterology 132: , Bickel PE, Tansey JT, and Welte MA. PAT proteins, an ancient family of lipid droplet proteins that regulate cellular lipid stores. Biochim Biophys Acta 1791: , Brasaemle DL. Thematic review series: adipocyte biology. The perilipin family of structural lipid droplet proteins: stabilization of lipid droplets and control of lipolysis. J Lipid Res 48: , Chan SC, Lin SC, and Li P. Regulation of Cidea protein stability by the ubiquitin-mediated proteasomal degradation pathway. Biochem J Chen Z, Guo K, Toh SY, Zhou Z, and Li P. Mitochondria localization and dimerization are required for CIDE-B to induce apoptosis. J Biol Chem 275: , Ding HF, Lin YL, McGill G, Juo P, Zhu H, Blenis J, Yuan J, and Fisher DE. Essential role for caspase-8 in transcription-independent apoptosis triggered by p53. J Biol Chem 275: , Ducharme NA, and Bickel PE. Lipid droplets in lipogenesis and lipolysis. Endocrinology 149: , !

176 8. Erdtmann L, Franck N, Lerat H, Le Seyec J, Gilot D, Cannie I, Gripon P, Hibner U, and Guguen-Guillouzo C. The hepatitis C virus NS2 protein is an inhibitor of CIDE-B-induced apoptosis. J Biol Chem 278: Flier JS. Obesity wars: molecular progress confronts an expanding epidemic. Cell 116: , Granneman JG, and Moore HP. Location, location: protein trafficking and lipolysis in adipocytes. Trends Endocrinol Metab 19: 3-9, Inohara N, Koseki T, Chen S, Wu X, and Nunez G. CIDE, a novel family of cell death activators with homology to the 45 kda subunit of the DNA fragmentation factor. Embo J 17: , Keller P, Petrie JT, De Rose P, Gerin I, Wright WS, Chiang SH, Nielsen AR, Fischer CP, Pedersen BK, and MacDougald OA. Fat-specific protein 27 regulates storage of triacylglycerol. J Biol Chem 283: , Kim JY, Liu K, Zhou S, Tillison K, Wu Y, and Smas CM. Assessment of fatspecific protein 27 in the adipocyte lineage suggests a dual role for FSP27 in adipocyte metabolism and cell death. Am J Physiol Endocrinol Metab 294: E , Li JZ, and Li P. Cide proteins and the development of obesity. Novartis Found Symp 286: ; discussion , , Li JZ, Ye J, Xue B, Qi J, Zhang J, Zhou Z, Li Q, Wen Z, and Li P. Cideb regulates diet-induced obesity, liver steatosis, and insulin sensitivity by controlling lipogenesis and fatty acid oxidation. Diabetes 56: , !

177 16. Li P. Cidea, brown fat and obesity. Mech Ageing Dev 125: , Liang L, Zhao M, Xu Z, Yokoyama KK, and Li T. Molecular cloning and characterization of CIDE-3, a novel member of the cell-death-inducing DNAfragmentation-factor (DFF45)-like effector family. Biochem J 370: , Liu P, Ying Y, Zhao Y, Mundy DI, Zhu M, and Anderson RG. Chinese hamster ovary K2 cell lipid droplets appear to be metabolic organelles involved in membrane traffic. J Biol Chem 279: Lugovskoy AA, Zhou P, Chou JJ, McCarty JS, Li P, and Wagner G. Solution structure of the CIDE-N domain of CIDE-B and a model for CIDE-N/CIDE-N interactions in the DNA fragmentation pathway of apoptosis. Cell 99: , Matsusue K, Kusakabe T, Noguchi T, Takiguchi S, Suzuki T, Yamano S, and Gonzalez FJ. Hepatic steatosis in leptin-deficient mice is promoted by the PPARgamma target gene Fsp27. Cell Metab 7: , Nishino N, Tamori Y, Tateya S, Kawaguchi T, Shibakusa T, Mizunoya W, Inoue K, Kitazawa R, Kitazawa S, Matsuki Y, Hiramatsu R, Masubuchi S, Omachi A, Kimura K, Saito M, Amo T, Ohta S, Yamaguchi T, Osumi T, Cheng J, Fujimoto T, Nakao H, Nakao K, Aiba A, Okamura H, Fushiki T, and Kasuga M. FSP27 contributes to efficient energy storage in murine white adipocytes by promoting the formation of unilocular lipid droplets. J Clin Invest 118: , !

178 22. Orlicky DJ, Degala G, Greenwood C, Bales ES, Russell TD, and McManaman JL. Multiple functions encoded by the N-terminal PAT domain of adipophilin. J Cell Sci 121: , Oscar R-C, Vishwajeet P, Incoronata M, Vladimir S, Robert KS, Satya D, Caroline SSH, William B, Corinne V, Jocelyne M, Philippa R-B, Peter RM, Anil C, Jeremy NS, Chatterjee VK, Sara S, Consortium LDS, Ann-Marie P, Anil KA, Abhimanyu G, Inês B, Saverio C, Michael PC, Jesús A, Stephen OR, and Savage DB. Partial lipodystrophy and insulin resistant diabetes in a patient with a homozygous nonsense mutation in CIDEC. EMBO Molecular Medicine 1: , Puri V, and Czech MP. Lipid droplets: FSP27 knockout enhances their sizzle. J Clin Invest 118: , Puri V, Konda S, Ranjit S, Aouadi M, Chawla A, Chouinard M, Chakladar A, and Czech MP. Fat specific protein 27: A novel lipid droplet protein that enhances triglyceride storage. J Biol Chem Puri V, Ranjit S, Konda S, Nicoloro SM, Straubhaar J, Chawla A, Chouinard M, Lin C, Burkart A, Corvera S, Perugini RA, and Czech MP. Cidea is associated with lipid droplets and insulin sensitivity in humans. Proc Natl Acad Sci U S A 105: , Qi J, Gong J, Zhao T, Zhao J, Lam P, Ye J, Li JZ, Wu J, Zhou HM, and Li P. Downregulation of AMP-activated protein kinase by Cidea-mediated ubiquitination and degradation in brown adipose tissue. EMBO J 27: , !

179 28. Ruan H, Hacohen N, Golub TR, Van Parijs L, and Lodish HF. Tumor necrosis factor-alpha suppresses adipocyte-specific genes and activates expression of preadipocyte genes in 3T3-L1 adipocytes: nuclear factor-kappab activation by TNFalpha is obligatory. Diabetes 51: , Ruan H, and Lodish HF. Insulin resistance in adipose tissue: direct and indirect effects of tumor necrosis factor-alpha. Cytokine Growth Factor Rev 14: , Ruan H, Miles PD, Ladd CM, Ross K, Golub TR, Olefsky JM, and Lodish HF. Profiling gene transcription in vivo reveals adipose tissue as an immediate target of tumor necrosis factor-alpha: implications for insulin resistance. Diabetes 51: , Smirnova E, Goldberg EB, Makarova KS, Lin L, Brown WJ, and Jackson CL. ATGL has a key role in lipid droplet/adiposome degradation in mammalian cells. EMBO Rep 7: , Srinivasula SM, Ahmad M, Fernandes-Alnemri T, and Alnemri ES. Autoactivation of procaspase-9 by Apaf-1-mediated oligomerization. Mol Cell 1: , Toh SY, Gong J, Du G, Li JZ, Yang S, Ye J, Yao H, Zhang Y, Xue B, Li Q, Yang H, Wen Z, and Li P. Up-regulation of mitochondrial activity and acquirement of brown adipose tissue-like property in the white adipose tissue of fsp27 deficient mice. PLoS ONE 3: e2890, Unger RH. Lipotoxic diseases. Annu Rev Med 53: , !

180 35. Unger RH. Minireview: weapons of lean body mass destruction: the role of ectopic lipids in the metabolic syndrome. Endocrinology 144: Unger RH. The physiology of cellular liporegulation. Annu Rev Physiol 65: , Unger RH, and Orci L. Lipotoxic diseases of nonadipose tissues in obesity. Int J Obes Relat Metab Disord 24 Suppl 4: S28-32., Walther TC, and Farese RV, Jr. The life of lipid droplets. Biochim Biophys Acta Wolins NE, Brasaemle DL, and Bickel PE. A proposed model of fat packaging by exchangeable lipid droplet proteins. FEBS Lett 580: , Wolins NE, Quaynor BK, Skinner JR, Schoenfish MJ, Tzekov A, and Bickel PE. S3-12, Adipophilin, and TIP47 package lipid in adipocytes. J Biol Chem 280: , Wu C, Zhang Y, Sun Z, and Li P. Molecular evolution of Cide family proteins: novel domain formation in early vertebrates and the subsequent divergence. BMC Evol Biol 8: 159, Ye J, Li JZ, Liu Y, Li X, Yang T, Ma X, Li Q, Yao Z, and Li P. Cideb, an ER- and lipid droplet-associated protein, mediates VLDL lipidation and maturation by interacting with apolipoprotein B. Cell Metab 9: , !

181 43. Zechner R, Kienesberger PC, Haemmerle G, Zimmermann R, and Lass A. Adipose triglyceride lipase and the lipolytic catabolism of cellular fat stores. J Lipid Res 50: 3-21, Zhou Z, Yon Toh S, Chen Z, Guo K, Ng CP, Ponniah S, Lin SC, Hong W, and Li P. Cidea-deficient mice have lean phenotype and are resistant to obesity. Nat Genet 35: !

182 Acknowledgements We thank Dr. H. F. Ding (Medical College of Georgia) for the dominant negative caspase-9 expression construct. Disclosures None Grants Medicine. Supported by an institutional grant from the University of Toledo College of 175!

183 Figure Legends Figure 1. FSP27 Promotes Lipid Droplet Formation and Apoptosis Across Multiple Cell Lines. A. Promotion of lipid droplet formation by CIDE proteins. The indicated cell lines were transfected with either empty vector, CIDEA, FSP27, or CIDEB. Cells were switched to culture media supplemented with 400 µm BSA-complexed oleic acid at 4 h post-transfection and the next day stained with Oil Red O. B. FSP27 induces apoptosis in multiple cell types. Cells were transfected with an empty vector or an FSP27 expression construct along with a marker "-galactosidase expression construct, as described in Materials and Methods. At 24 h post-transfection the numbers of "-galactosidase (LacZ+) cells were enumerated, with that for empty vector transfectants set to 100% for each respective cell line. Data are shown as mean ± SD with * = p<0.01. Figure 2. Localization of FSP27 at Lipid Droplets Attenuates its Apoptotic Effect. A. Localization of FSP27 to lipid droplets in HeLa cells. Lipid-loaded HeLa cells were transfected with an egfp expression construct for full-length FSP27, CIDEA or CIDEB. At 16 post-transfection, live cells were stained with Nile Red to visualize lipid droplets and assessed by confocal microscopy. The top panel set shows egfp signal, the second panel set shows lipid droplet signal stained with Nile Red, the third panel set shows the merged image for egfp and Nile Red signal. The bottom single panel shows merged imaged of empty vector (EV) egfp signal and Nile Red signal. B. Presence of intracellular lipid droplets attenuates FSP27-mediated apoptosis. HeLa cells were cultured for 1 d in the presence or absence of 400 µm BSA-complexed oleic acid (OA) 176!

184 and then transfected with either empty vector (EV) or an FSP27 expression construct, along with a marker "-galactosidase (LacZ+) expression construct, as described in Materials and Methods. LacZ+ blue cells were enumerated 24 h later. The number of cells in the respective EV transfectants was set to 100%. Data are shown as mean ± SD with # = p<0.05 for FSP27 in absence vs. presence of OA and *, p <0.05 vs. respective EV control. Figure 3. Effect of Co-Transient Expression of ATGL on Ability of FSP27 to Promote Lipid Droplet Formation. A. HeLa cells were transfected with an expression construct for FSP27 in the presence or absence of a 5-fold excess of an ATGL expression construct or an empty vector (EV, pcdna3.1). Cells were switched to culture media supplemented with 400 µm BSA-complexed oleic acid at 4 h post-transfection and at h post-transfection were stained for lipid droplet content with Bodipy 493/503 (upper panel set) or LipidTox Deep Red (lower panel set) and photographed. Typical fields are depicted. B. The numbers of cells with one or more prominent enlarged lipid droplets (LDs) were enumerated per microscopic field as described in Materials and Methods. Data are shown as mean ± SD with *= p<0.01 vs. empty vector (EV). C. HeLa cells were transfected with a 5:1 ratio of FSP27 and either pires2-egfp-atgl or pires2-egfp empty vector (EV). Cells were switched to culture media supplemented with 400 µm BSA-complexed oleic acid at 4 h post-transfection and at h post-transfection stained for lipid droplet content with LipidTox Deep Red, imaged for egfp and LipidTox Red signals and photographed. Typical fields are depicted. D. The numbers of green cells with one or more prominent enlarged lipid droplets were enumerated per microscopic 177!

185 field, for quadruplicate transfections as described in Materials and Methods. E. Quadruplicate cultures of cells as depicted in C and D were subjected to flow cytometry analysis for egfp and LipidTox Deep Red signals. Mean LipidTox Deep Red signal intensity per green cell is shown on Y-axis. A.U. indicates arbitrary units. For B, D, and E, data are shown as mean ± SD with *= p<0.01. Figure 4. Effect of ATGL on Pre-Formed FSP27-Induced Lipid Droplets. A. Lipid droplet accumulation in HeLa cell populations stably expressing retrovirally-driven FSP27 or pbabe-puro EV. The first four panels show phase contrast pictures of EV and FSP27-expressing cells cultured for 1 d with (+OA) and without (-OA) 400 µm BSAcomplexed oleic acid. The second four panels show Bodipy 493/503 staining for lipid droplets. Phase contrast and Bodipy panels represent different cell fields with typical fields depicted. B. Transient transfection of HeLa FSP27 cells with ATGL or EV. Cells were incubated with 400 µm BSA-complexed oleic acid for 24 h then transiently transfected with ATGL-pcDNA3.1 or EV. Cells were stained for lipid with Bodipy 493/503 at h post-transfection with typical fields depicted. C. Quantification of cells with lipid droplets that were transfected and cultured as in B. The numbers of cells with one or more prominent enlarged lipid droplets were enumerated per microscopic field, in quadruplicate transfections as described in Materials and Methods. D. Quantification of lipid droplet content per cell. Flow cytometry analysis for Bodipy 493/503 signal. Cells were transfected and cultured as for B and C in quadruplicate with mean signal intensity per cell shown on Y-axis. E. Transient transfection of HeLa FSP27 cells with pires2-egfp-atgl or pires2-egfp EV. Cells were incubated with 178!

186 400 µm BSA-complexed oleic acid for 24 h then transiently transfected with pires2- egfp-atgl or pires2-egfp EV. Cells were stained for lipid with LipidTox Deep Red at h post-transfection. Typical fields are depicted. F. Quantification quadruplicate samples of cells with lipid droplets that were transfected and cultured as in E. The percentage of green cells with one or more prominent enlarged lipid droplets were enumerated per microscopic field, in quadruplicate as described in Materials and Methods. G. Quantification of lipid droplet content per green cell. Cells were transfected and cultured as in E and F. Flow cytometry analysis was done on quadruplicate samples for egfp and LipidTox Deep Red signal, with mean LipidTox Deep Red signal intensity per green cell shown on Y-axis. H, I and J. Experiments were carried out in quadruplicate as for E, F and G except that cells were cultured with 400 µm BSAcomplexed oleic acid for 3 d prior to transfection. For D, G and J, A.U. indicates arbitrary unit. For C, D, F, G, I and J data are shown as mean ± SD with *= p<0.01. Figure 5. The FSP27 Apoptotic Mechanism Involves Caspase-9 and Release of Mitochondrial Cytochrome c. A. The pan caspase inhibitor Z-VAD-FMK diminishes FSP27-mediated cell death. 293T cells were co-transfected with empty vector (EV, black bars) or a FSP27 expression construct (white bars) along with a marker "-galactosidase (LacZ+) expression construct, as described in Materials and Methods, with 20 µm Z- VAD-FMK or DMSO vehicle added at time of transfection. Cells were stained for LacZ activity at 24 h post-transfection and LacZ+ blue cells counted. Data are shown as mean ± SD. #, p<0.001 for untreated or DMSO treated FSP27-transfected cells vs. their EV transfected counterparts. The value for the leftmost column was set to 100%. *, p< !

187 for Z-VAD-FMK treated FSP27-transfected cells vs. untreated or DMSO treated FSP27- transfected cells. B. Inhibition of FSP27-mediated fragmentation by Z-VAD-FMK. 293T cells were transfected with FSP27 expression construct or empty vector and incubated with 20 µm Z-VAD-FMK, DMSO vehicle, 20 µm of the negative control peptide VA- FMK or no additions. Genomic DNA was prepared and analyzed for fragmentation using ethidium bromide staining. M, DNA marker, with numbers to right indicating base pairs. C. Inhibition of FSP27-mediated PARP and!-fodrin cleavage by Z-VAD-FMK. 293T cells were transfected with FSP27 expression construct and incubated with 20 µm Z- VAD-FMK, DMSO vehicle, 20 µm of the negative control peptide VA-FMK, or no additions. Cell lysates were harvested and Western blot analysis was performed for fulllength p116 and p89 cleaved PARP, and p150 cleaved!-fodrin. NT, non-transfected. D. Western blot assessment of cleaved caspases. 293T cells were transfected with FSP27 expression construct or empty vector (EV) and total protein prepared at the indicated time points post-transfection. Western blot analysis was performed for p37 cleaved caspase-9 (CS9), p20 cleaved caspase-7 (CS7), and p19 cleaved caspase-3 (CS3). E. Effects of expression of dominant negative caspase-9 (CS9DN) on FSP27-mediated apoptosis. Upper: 293T cells were co-transfected with a "-galactosidase (LacZ) expression construct as and either empty vector (EV) or expression constructs for FSP27 and/or CS9DN in the indicated combinations. Values of 1 and 7 represent 0.5 µg and 3.5 µg of the FSP27 and the CS9DN expression constructs, respectively. Cells were stained for "-galactosidase activity at 48 h post-transfection and blue (LacZ+) cells counted. Data are shown as mean ± SD. *, p<0.001 for FSP27-transfected cells compared to EV transfected cells; #, p<0.001 for FSP27 and CS9DN co-transfected cells compared to FSP27-transfected cells. 180!

188 The value of the leftmost samples was set to 100%. Lower: 293T cells were cotransfected, as described directly above for cell death assay and analyzed by Western blot using anti-caspase-9 and anti p89 cleaved PARP antibodies. The mass of the FSP27 and CS9DN expression construct used in micrograms are indicated at top of the respective figure. For B-E images shown in the same agarose gel or in the boxed horizontal Western blot panels were run on the same gel and in case of Western blots, processed on the same membrane. However, in some instances lanes have been removed and/or rearranged for economy and clarity of presentation. F. Assessment of cytochrome c release. COS cells were transiently transfected with an expression construct for egfp-fsp27 or empty egfp vector (EV). Shown are egfp signal under FITC fluorescence (green signal, top panels), cytochrome c (Cyto c) immunostaining using a monoclonal cytochrome c primary antibody with Alexafluor 568-conjugated secondary antibody (red signals, middle panels). Nuclei stained with DAPI is shown in the left and right bottom panels. The meaning of the arrowhead and asterisk is described in the text. Representative images are shown. G. Western blot analysis for quantification of cytochrome c release to cytoplasm. COS cells were transiently transfected with egfp-fsp27 (Full) or empty egfp vector (EV) and analyzed at 18 h post-transfection. Upper panel shows 6 µg of cytosolic fraction protein analyzed by Western blot for cytochrome c, the cytosolic marker protein GAPDH and the mitochondrial marker ATP synthase!; this blot also contained 10 µg of purified mitochondria fraction protein with short (S) and long (L) exposures shown. Samples were independently prepared from triplicate transfections. Data shown within each boxed area arose from the same protein gel. Lower panel shows 181!

189 digital quantification of Western blot signals; A.U. indicates arbitrary unit. Data are shown as mean ± SD with *= p<0.01 vs. EV. Figure 6. FSP27-C is Necessary and Sufficient for Caspase-Mediated Apoptosis and Lipid Droplet Localization. A. Cell death assay. 293T cells were co-transfected with empty vector (EV), full-length FSP27 (Full), FSP27-N or FSP27-C as egfp fusion constructs, together with a LacZ expression construct. Cells were stained for "- galactosidase (LacZ+) and blue cells counted at 48 h post-transfection. Data are shown as mean ± SD, where *, p<0.001 compared to EV transfected cells. Levels in the EV sample was set to 100%. B. DNA fragmentation assay. 293T cells were transfected with empty vector (EV) or the indicated egfp FSP27 fusion constructs (Full, full-length FSP27; N, FSP27-N; C, FSP27-C). Genomic DNA was prepared at 48 h post-transfection and analyzed by SYBR Green staining. M, DNA marker, with numbers to right indicating base pairs. C. Western blot assessment of apoptotic indices. Total protein was harvested from 293T cells at 48 h post-transfection and Western blot analysis performed for indicated proteins. D. Inhibition of cleavage of PARP and!-fodrin by Z-VAD-FMK. 293T cells were transfected with EV or egfp-cide-c expression construct and cultured in the absence (-) or presence (+) of Z-VAD-FMK for 24 h. Total protein was harvested and analyzed by Western blot for PARP or!-fodrin. E. Inhibition of DNA fragmentation by Z-VAD-FMK. 293T cells were transfected with empty vector egfp (EV) or egfp- CIDE-C expression construct and cultured in the absence (-) or presence (+) of Z-VAD- FMK for 24 h. Genomic DNA was harvested and assessed by SYBR Green staining. M, DNA marker; numbers to right indicate base pairs. Due to the degree of cell death 182!

190 mediated FSP27 and FSP-C, the respective egfp fusion proteins are not visible in the exposures shown in Western blot of these cell lysates. For B-E images shown in the same agarose gel or in the boxed horizontal Western blot panels were run on the same gel and in case of Western blots, processed on the same membrane. However, in some instances lanes have been removed and/or rearranged for economy and clarity of presentation. F. FSP27-C mediates release of mitochondrial cytochrome c to cytoplasm. COS cells were transiently transfected with egfp-fsp27-cide-n (N) or egfp-fsp27-cide-c (C). Shown are egfp signal under FITC fluorescence (green signal, top panel), cytochrome c (Cyto c) immunostaining using a monoclonal cytochrome c primary antibody with Alexafluor 568 conjugated secondary antibody (red signals, middle panel) and nuclei stained with DAPI (blue signal, lower panel). Representative images are shown. G. Western blot analysis for quantification of cytochrome c release by egfp-fsp27-cide- C. COS cells were transiently transfected with egfp-fsp27 (Full), egfp-fsp27-cide- N (N), egfp-fsp27-cide-c (C) or empty egfp vector (EV) and analyzed at 18 h posttransfection. Upper panel shows 6 µg of cytoplasmic fraction protein analyzed by Western blot for cytochrome c, the cytoplasmic marker protein GAPDH and the mitochondrial marker ATP synthase!; this lower blot also contained 10 µg of purified mitochondria fraction protein with short (S) and long (L) exposures shown. Samples were independently prepared from triplicate transfections. Data shown within each boxed area arose from the same protein gel. Lower panel shows digital quantification of Western blot signals with A.U. indicating arbitrary units. Data are shown as mean ± SD with *= p<0.01 vs. EV. H. Localization of the FSP27-N and FSP27-C in lipid-loaded 183!

191 HeLa cells. Cell culture, transfection and analysis were carried out as described for Figure 2A, with merged images shown. Figure 7. Deletion Analysis of the Apoptotic and Lipid Droplet Localization Function of FSP27-C. A. Effects of regions of the FSP27-C on apoptosis. Upper: Schematic representation of the egfp-fsp27-c and deletion constructs. The CIDE-C domain or regions thereof is represented as black rectangle and numbers indicate amino acid positions in the FSP27 protein sequence. Constructs contain egfp fusion 5 to the indicated FSP27 coding regions. Lower: 293T cells were co-transfected with empty vector (EV), egfp-fsp27-c (C), '173 and '192 expression constructs, together with a LacZ expression construct as described in Materials and Methods. Cells were stained for "-galactosidase activity and blue cells (LacZ+) counted at 48 h post-transfection. Value for the EV was set to 100%. Data are shown as mean± SD, where *= p<0.001 compared to EV transfected cells. B. DNA fragmentation assay. 293T cells were transfected empty vector (EV) or the indicated egfp FSP27 fusion constructs. Genomic DNA was prepared at 24 h post-transfection and analyzed by SYBR Green staining. M, DNA marker, with numbers to right indicating base pairs. C. Western blot assessment of apoptotic indices for FSP27 deletions. Total protein was harvested at 48 h post-transfection and Western blot analysis performed for p89 cleaved PARP, p37 cleaved caspase-9 (CS9), p19 cleaved caspase-7 (CS7), and p20 cleaved caspase 3 (CS3). D. Effect of the pan-caspase inhibitor Z-VAD-FMK on '173-mediated apoptosis. 293T cells were transfected with empty vector (EV) or the '173 expression construct and cultured in the presence (+) or absence (-) of Z-VAD-FMK for 24 h. Total protein was harvested and analyzed by 184!

192 Western blot for p89 cleaved PARP, p37 cleaved caspase-9 (CS9) and p19 cleaved caspase-7 (CS7). For B-D images shown in the same agarose gel or in the boxed horizontal Western blot panels were run on the same gel and in case of Western blots, processed on the same membrane. However, in some instances lanes have been removed and/or rearranged for economy and clarity of presentation. E. Lipid droplet localization of FSP27 deletion constructs. Cell culture, transfection and analysis were carried out as described for Figure 2A. Figure 8. Protein-Protein Interaction Analysis of FSP27. A. Co-immunoprecipitation analysis. 293T cells were co-transfected with HA-tagged FSP27 (HA-FSP27) (+), FLAGtagged CIDEA (+) or empty vector (-). Total protein was harvested 48 h post-transfection and cell lysates immunoprecipitated with anti-flag M2-agarose followed by Western blotting with anti-ha antibody. A 10 sec exposure is shown. Arrowhead indicates the HA signal and asterisk indicates non-specific signal from antibody light chain. Images shown in the same boxed horizontal Western blot panel were run on the same gel and processed on the same membrane. However, in some instances lanes have been removed and/or rearranged for economy and clarity of presentation. B. Yeast two hybrid interaction assay. Yeast harboring the indicated combinations of the pgbtk7-based or pgad-based yeast two-hybrid expression constructs, listed at top and left were inoculated onto on selective media. DDO indicates -Leu, -Trp dropout media; TDO indicates -His, -Leu, -Trp dropout media; X-!-gal indicates TDO media containing X-!- gal. Macroscopic view of yeast media plates is shown with the middle and bottom panels demonstrating growth and color indicative of protein-protein interaction. A thin dotted 185!

193 line has been added to aid in distinguishing each row. Images shown in each boxed area arose from the same agar plate. However in some instances lanes have been removed and/or rearranged for economy and clarity of presentation. C. FSP27 can mediate lipid droplet localization of the FSP27 CIDE-N domain. HeLa cells were incubated for 2 d with 400 µm BSA-complexed oleic acid to induce lipid droplet formation and then transiently co-transfected with expression constructs for non-tagged FSP27 and egfp- FSP27-CIDE-N. At ~ 16 h post-transfection cells were stained for lipid with LipidTox Deep Red and observed by fluorescence microscopy and photographed. The left and right panel sets depict two examples typical for cells showing egfp-fsp27-cide-n signal at lipid droplets. Within each panel set, the left image shows egfp signal for egsp-fsp27- N and the right image lipid droplets stained with LipidTox Deep Red. Figure 9. Effect of CIDEA on FSP27 Protein Levels. A. Western blot analysis. 293T cells were transfected with expression constructs for HA-tagged FSP27 (FSP27), FLAGtagged CIDEA (CIDEA), both or empty vector (EV). Total protein was harvested 48 h post-transfection and cell lysates assessed by Western blot using anti-flag and anti-ha antibodies. B. Assessment of FSP27 protein stability. 293T cells were transfected with expression constructs for HA-tagged FSP27 (FSP27) in the absence and presence of cotransfection of FLAG-tagged CIDEA. Cells were treated with 100 µg/ml cycloheximide at 40 h post-transfection, with the 0 time point harvested just prior to cycloheximide treatment. Total protein was harvested at the indicated time points post cycloheximide treatment, shown in minutes. Western blots were probed with indicated antibodies. For A and B, images shown in the same boxed horizontal Western blot panel 186!

194 were run on the same gel and processed on the same membrane. However, in some instances lanes have been removed and/or rearranged for economy and clarity of presentation. C. Half-life determination. Graphical representation of quantitated signals from Western blot analyses is shown. Figure 10. FSP27 and CIDEA Co-localize on Small and Large Lipid Droplets. HeLa cells were incubated for 2 d with 400 µm BSA-complexed oleic acid to induce lipid droplet formation. Cells were then co-transfected with full length fluorescent protein fusion expression constructs for FSP27 (egfp-fsp27) and CIDEA (DSRed-CIDEA) and observed at 14 h post-transfection with confocal microscopy. A. Projected image of all Z- sections is shown, with upper and lower panels depicting signal for egfp-fsp27 and DSRed-CIDEA, respectively. Merged image for egfp-fsp27 and DSRed-CIDEA is shown in the middle panel. B. Individual Z-sections. Three representative Z-sections (upper, middle and lower panels) are shown. Signal for egfp-fsp27 and DSRed-CIDEA are shown in the right and left panel set, respectively. 187!

195 188!

196 189!

197 190!

198 191!

199 192!

200 193!

201 194!

202 195!

203 196!

204 197!

205 198!

206 DISCUSSION AND SUMMARY This dissertation is focused on defining the roles of FSP27 in lipid droplet formation and apoptosis. I was also looking for potential interplay mediated through FSP27 to bridge these two distinct cellular processes. Despite the fact that FSP27 gene knockout studies have been recently carried out (Nishino et al., 2008; Toh et al., 2008), and that these and other studies showed FSP27 to be a key metabolic player (Puri et al., 2007; Keller et al., 2008; Kim et al., 2008; Nishino et al., 2008; Toh et al., 2008; Liu et al., 2009), the story of this protein with respect to lipid droplet biology and apoptosis is far from being over. Rather, the profound phenotype of FSP27 null mice (Nishino et al., 2008; Toh et al., 2008) provides highly compelling evidence of the need to carefully dissect FSP27 function(s) and mechanism(s). Interesting questions remain regarding this protein that should be addressed in the future studies. I believe its further study will provide key insights for the understanding and treatment of obesity and lipotoxicity. From what I have observed and based on the literature, it appears that there may be three potential means whereby FSP27 and its manipulation can control energy storage. Namely, the regulation of lipid droplet formation and lipolysis, the establishment of a white versus brown adipocyte identity, and the prospect of harnessing of its apoptotic potential to control fat mass. The first two are clearly supported by biochemical and genetic analysis as described here. The third one is speculative at this time. With respect to function in lipid droplet formation and lipolysis, the mechanisms of how FSP27 maintains TG lipid droplet content and regulates lipid enlargement remain 199!

207 uncertain (Puri et al., 2007; Keller et al., 2008; Nishino et al., 2008; Toh et al., 2008; Liu et al., 2009). The majority of data indicate that FSP27 serves to protect TG from basal lipolysis (Nishino et al., 2008; Toh et al., 2008), as also is the case for perilipin (Martinez-Botas et al., 2000; Tansey et al., 2001). However, knowledge to date indicates that the mechanisms used by FSP27 and perilipin in respect to lipid droplet metabolism are different. This conclusion is based on the observation that multilocular lipid droplets are seen in WAT after knockout of FSP27, while WAT of perilipin null mice has unilocular lipid droplets (Nishino et al., 2008; Toh et al., 2008). Additionally, perilipin null mice have clear defects of hormone-stimulated lipolysis (Nishino et al., 2008; Toh et al., 2008), whereas this is not the case for FSP27 null mice (Nishino et al., 2008; Toh et al., 2008). To further investigate the possible functional interrelationships between FSP27 and perilipin, it would be intriguing to determine whether FSP27 can functionally compensate for perilipin by introducing FSP27 into perilipin knockout cells, or vice versa. In regard to FSP27 in lipid droplet enlargement, and whether it controls lipid droplet enlargement by regulating lipid droplet fusion, it would also be interesting to test whether FSP27 can associate with those SNARE proteins that have been demonstrated to be involved in the fusion of lipid droplets (Boström et al., 2007). Another method to address the role of FSP27 in lipid droplet fusion would be to undertake real-time imaging to assess lipid droplet fusion rates in the presence or absence of FSP27. In regard to the nature of adipocyte identity, accumulation of the unilocular lipid droplet is morphological evidence of successful differentiation from white preadipocytes to white adipocytes. However, even though the lipid droplet is the defining 200!

208 morphological and functional feature of white adipocytes, the ectopic expression of the FSP27 gene in preadipocytes does not affect adipogenesis (Keller et al., 2008). On the other hand, the absence of FSP27 has striking effect on white adipocyte identity (Nishino et al., 2008; Toh et al., 2008). This was observed to some extent during in vitro knockdown based on the appearance of smaller lipid droplets and large numbers of mitochondria (Keller et al., 2008). The ability of FSP27 to impact adipocyte identity is most striking for in vivo models, where a functional and transcriptional transition of WAT to BAT was observed (Nishino et al., 2008; Toh et al., 2008). Unlike WAT, which mainly serves as a form of energy storage, BAT is an energy consumption tissue (Gesta et al., 2007). Hence, this interesting observation provides another potential means that can be employed by FSP27 to control fat mass. It is apparent that FSP27 is a pro-apoptotic protein when ectopically expressed in non-adipocyte cell lines. However, at this point, I still do not know if there is any physiological significance of this cell-death inducing function of FSP27. Since FSP27 has a restricted and high expression in adipose tissues, it is! temping to postulate that under specific circumstances the pro-apoptotic activity of FSP27 might be triggered, which could consequently lead to the reduction of fat cell numbers. Under controlled conditions, this may be able to affect a degree of reduction of white adipose mass. Pharmacological effectors that reduce FSP27 activity provide an alternative option to diminish adipose mass other than through FSP27 triggering apoptosis in these cells, as mentioned above. Study of FSP27 -/- mice clearly demonstrated that the enhanced "-oxidative capacity of their white adipocytes acts to spare other tissues from lipotoxic effects of high-fat diet, where toxic lipid overflow would normally be anticipated. This 201!

209 makes FSP27 an ideal target for pharmacological manipulation with the aim of controlling fat mass while protecting from lipotoxicity. Drugs might be identified that can produce a functional FSP27 knockout, perhaps by dissociating FSP27 from the lipid droplet. This may result in a phenotype similar to FSP27 null mice, with the concomitant benefits in respect to enhanced energy expenditure via increased!-oxidation within WAT, protection from lipotoxicity, and increased insulin sensitivity. Additionally, future investigations and identification of FSP27 pathway players and interacting proteins may result in the identification of additional targets for therapeutic intervention in relation to FSP27 function in fat tissues.! In regard to the implications of FSP27 action in obesity and metabolism, it is important to keep in mind that distinctions have already been noted for murine versus human settings. For example, FSP27 null mice show increased insulin sensitivity (Nishino et al., 2008; Toh et al., 2008), whereas the single patient identified to date with lipodystrophy and mutation in FSP27 is insulin resistance (Rubio et al., 2009). Also, the tissue specific expression patterns of the FSP27 interacting protein CIDEA is distinct for mouse versus human. Murine white adipocytes express abundant FSP27 but are negative for CIDEA expression. On the other hand, human white adipocytes express high levels of both FSP27 and CIDEA. In humans, there is no direct evidence as to whether a relationship between mitochondrial number or activity in WAT and FSP27 level exists. Additionally, in human, elevated levels of FSP27 correlated with enhanced insulin sensitivity in obese patients (Puri et al., 2008a), while enhanced insulin sensitivity was noted in the complete absence of FSP27 in knockout mice (Nishino et al., 2008; Toh et al., 2008). Due to these differences obtained from mice and humans, it is essential to 202!

210 assess the outcomes of the genetic studies in mice with the careful application of descriptive and correlative studies in humans. It will have a huge impact on controlling fat mass if manipulation of the level of FSP27 in humans can transform WAT from an energy storage tissue into an energy consuming tissue like BAT. At this time, while it is clear that FSP27 is a lipid droplet-associated protein, how FSP27 is targeted to lipid droplets remains a mystery. Researchers in the lipid droplet field have been searching for discrete lipid droplet targeting sequences, such as is known for protein targeting to other organelles. Still, no sequence meeting this criterion has been identified. This may simply reflect the fact that to date very few lipid droplet proteins have undergone detailed structure-function mapping in regard to lipid droplet localization. Past studies have demonstrated certain dispersed region of lipid droplet proteins can mediate their lipid droplet targeting. Even for the PAT proteins, a longstanding prototype of lipid droplet research, to date researchers can only narrow the lipid droplet targeting signal to several rather widely dispersed protein regions (McManaman et al, 2003; Garcia et al., 2003). For instance, three dispersed hydrophobic regions within the approximately 100 amino acid central portion of perilipin are involved in its lipid droplet localization (Garcia et al., 2003). Nonetheless, to date, no consensus sequence or potential mechanism has been illustrated from these types of studies (Fujimoto et al., 2001; Garcia et al., 2003; Brasaemle, 2007). Here I show that amino acids 173 to 191 in the CIDE-C domain of FSP27 is required for its lipid droplet localization (Liu et al., 2009). However, this peptide per se failed to target egfp to lipid droplets. Nonetheless, it may be a bona fide lipid droplet targeting sequence in the context of full, intact FSP27 protein. It is possible that this short 203!

211 sequence requires a unique topology that is not recapitulated when fused to egfp. It may be premature to discount the potential of this sequence to function in lipid droplet targeting based on a single type of fusion. Indeed, no lipid droplet localization was observed by direct fusing any of the three LDs targeting regions of caveolin-2 to EGFP (Fujimoto et al., 2001). It would be of interest to further test the lipid droplet targeting potential of this 19 amino acid sequence by fusion with other proteins, perhaps those smaller in size than egfp. I also do not exclude the possibility that other regions of FSP27 may harbor the lipid droplet targeting sequence or, alternatively, act to significantly stabilize and/or maintain the structure of the 19 amino acid region in order for it to be exposed in an appropriate conformation. In general, there are three ways a protein may localize at the lipid droplet. A region of the protein may be embedded into the hydrophobic TG core. Second, the protein may associate with or insert into the phospholipid monolayer or, third, it may associate indirectly by interaction with other lipid droplet proteins. FSP27 association with lipid droplets may be via interacting with other lipid droplet-associated proteins. As I have shown in the second manuscript, FSP27 interacts with another lipid droplet-associated protein, CIDEA (Liu et al., 2009). This observation increases the plausibility of this hypothesis. However, it is not yet known how CIDEA is targeted to or how it associates with lipid droplets. To further investigate protein interactions of FSP27 at lipid droplets, one could test candidate lipid droplet-associated proteins such as perilipin for interaction with FSP27. In addition, a more comprehensive proteomic experiment might be carried out. I could analyze the lipid droplet pool following overexpression of FSP27, or use 204!

212 endogenous protein of adipocyte cell lines. FSP27 immunoprecipitates would be separated in SDS-polyacrymide gel. By determining the identity of the protein band(s) unique to samples that harbored FSP27, the FSP27 lipid droplet interactome can be identified by proteomic methods. Following this, knockdown of potential candidates would be carried out and effects of such on FSP27 lipid droplet localization determined. There is also the possibility that FSP27 may interact with the surface of lipid droplet via electrostatic interaction. To distinguish this possibility, I could treat purified lipid droplets from FSP27 transfected cells or mature adipocytes with alkaline carbonate. Western blots would then be used to determine if FSP27 is still remained in the lipid droplet pool. To date, based on my study and observations by others, it is clear that FSP27 can induce mitochondrial-mediated apoptosis (Liu et al., 2009). However, the identity of the most proximal steps in FSP27-mediated apoptosis, prior to cytochrome c release, remains to be uncovered. I showed that the same 19 amino acid in CIDE-C domain was necessary for both its lipid droplet localization and apoptotic function (Liu et al., 2009). Evidence from our finding suggests that localization of FSP27 in lipid droplets can attenuate its pro-apoptotic activity (Liu et al., 2009). FSP27 can induce intrinsic apoptosis pathway, which is tightly controlled by activities of BCL-2 family members. It would be intriguing to analyze whether FSP27 can associate with any of the pro- or anti-apoptotic BCL-2 family members, thereby providing some insights into the details of FSP27-induced apoptosis. Based on their sequence similarities, studies on CIDEA and CIDEB may possibly provide clues towards a better understanding the role of FSP27 in lipid metabolism as 205!

213 well as in apoptosis. For instance, I show that the 173 to 191 amino acid portion of the CIDE-C domain of FSP27 is important for its lipid droplet targeting and apoptosis (Liu et al., 2009). Similarly, a report by others showed that amino acids in the CIDE-C domain of CIDEB was sufficient for its lipid droplet targeting in lipid-loaded COS and HepG2 cells (Ye et al., 2009) and also for apoptosis (Chen et al., 2000). Comparison across these two regions of the respective CIDE proteins may provide additional clues as to which amino acids within these short regions are key to lipid droplet localization and apoptosis. In my study, I noted that FSP27 could interact with another lipid droplet-associated protein, CIDEA (Liu et al., 2009). Furthermore, I observed that the CIDE-C domain of FSP27 was critical for this interaction. In the same study, the CIDE-C domain of FSP27 was also shown to have the ability to form dimers with FSP27, CIDEA and itself, whereas the FSP27 CIDE-N domain only formed dimers with itself (Liu et al., 2009). A previous study by others reported that the CIDE-C domain of CIDEB could form homodimers during apoptosis (Chen et al., 2000). Additionally, in structural studies, the CIDE- N domain of CIDEB was shown to harbor the ability to associate with the CIDE-N domain of DFF40 and DFF45. This relies on bipolar surface features of the respective CIDE-N domains (Lugovskoy et al., 1999). Since the CIDE-N domain is highly conserved among CIDE family members, I speculate that the CIDE-N domain might be an interaction interface in FSP27 as well. It is possible that the presence of the CIDE-C domain masks the critical part of the CIDE-N domain for used for interaction and that the CIDE-N domain interaction surface is only available under certain conditions. 206!

214 In summary, my discoveries and additional findings by the Smas laboratory (FIG. 8), combined with information from the literature, makes it obvious that FSP27 is a key factor controlling lipid droplet metabolism and also a protein with robust pro-apoptotic activity, at least in the setting outside of adipocytes. Further investigation into the molecular mechanisms that underlie these two processes in adipocytes will without doubt shed more light on whether FSP27 can serve as a therapeutic target against obesity and its associated metabolic diseases. Figure 8. Summary of Proposed Roles of FSP27 in Lipid Droplet Metabolism and Apoptosis.! 207!