SPARC Inhibits Adipogenesis by Its Enhancement of β-Catenin Signaling*

SPARC (secreted protein acidic and rich in cysteine) modulates interactions between cells and extracellular matrix and is enriched in white adipose tissue. We have reported that SPARC-null mice accumulate significantly more fat than wild-type mice and maintain relatively high levels of serum leptin. We now show that SPARC inhibits adipogenesis in vitro. Specifically, recombinant SPARC inhibited (a) adipocyte differentiation of stromal-vascular cells isolated from murine white adipose tissue and (b) the expression of adipogenic transcription factors and adipocyte-specific genes. SPARC induced the accumulation and nuclear translocation of β-catenin and subsequently enhanced the interaction of β-catenin and T cell/lymphoid enhancer factor 1. The activity of integrin-linked kinase was required for the effect of SPARC on β-catenin accumulation as well as extracellular matrix remodeling. During adipogenesis, fusiform preadipocytes change into sphere-shaped adipocytes and convert the extracellular matrix from a fibronectin-rich stroma to a laminin-rich basal lamina. SPARC retarded the morphological changes exhibited by preadipocytes during differentiation. In the presence of SPARC, the deposition of fibronectin was enhanced, and that of laminin was inhibited; in parallel, the expression of α5 integrin was enhanced, and that of α6 integrin was inhibited. Lithium chloride, which enhances the accumulation of β-catenin, also inhibited the expression of α6 integrin. These findings demonstrate a role for SPARC in adipocyte morphogenesis and in signaling processes leading to terminal differentiation.

The cellular composition of WAT includes primarily adipocytes and preadipocytes as well as endothelial cells and macrophages. Obesity is the result of both over-proliferation (number) and overgrowth (size) of adipocytes. Adipocytes are not only the storage depots of energy but also the source of various cytokines and hormones. These so-called adipokines, e.g. tumor necrosis factor-␣, leptin, adiponectin, and resistin, target the central nervous system and peripheral tissues (fat, liver, and muscle) to modulate energy metabolism (1,2). SPARC (secreted protein acidic and rich in cysteine) belongs to the family of matricellular proteins, which generally do not contribute to the structure of extracellular matrix (ECM) but regulate its interaction with cells. SPARC is typically anti-adhesive in vitro and regulates angiogenesis and collagen production/fibrillogenesis in vivo. It is also a major participant in wound healing, tumor progression, and inflammation (3). Recent findings have attracted new interest in SPARC and its proposed role(s) in adipose tissue formation. SPARC-null mice exhibit significantly more fat accumulation than wild-type (WT) mice (4); consistent with this observation, SPARC-null bone marrow cells showed an increased tendency to differentiate into adipocytes rather than osteoblasts (5). Expression of SPARC in fat is enhanced in various murine obesity models that include diet-induced obesity, gold thioglucose treatment, and the ob/ob strain (6). In a clinical study, the plasma concentration of SPARC was correlated positively with body mass index (7). These data imply that SPARC is involved in the regulation of adipocyte differentiation and adipose tissue turnover.
Adipocytes are derived from mesenchymal stem cells, which first differentiate into preadipocytes and, subsequently, adipocytes, a process termed adipogenesis. Extensive studies have probed into mechanisms by which transcription factors and exogenous hormones regulate adipogenesis in cultured 3T3-L1/F442 cells. CAAT/enhancer-binding protein ␣ (C/EBP␣) and peroxisome proliferator-activated receptor ␥ (PPAR␥) are the key factors required for adipogenesis in addition to signaling mediated by insulin/insulin-like growth factor-1 and nuclear receptors (1,2). The Wnt/␤-catenin pathway has been shown to inhibit adipogenesis and enhance osteoblastogenesis (1,8,9). Activation of this pathway is sufficient to inhibit the differentiation and apoptosis of preadipocytes through an inhibition of C/EBP␣ and PPAR␥ (9,10). Wnt proteins bind to frizzled (Fz) receptors and low density lipoprotein receptorrelated protein coreceptors to activate several signaling pathways. Importantly, the inhibition of glycogen synthase kinase 3␤ (GSK3␤) via Wnt results in the stabilization of ␤-catenin in the cytoplasm as opposed to its proteasomal degradation. After translocation to the nucleus, ␤-catenin binds to and coactivates transcription factors that include members of the T-cell factor/ lymphoid-enhancing factor (TCF/LEF) family. Moreover, constitutively activated Fz1 increases the stability of ␤-catenin, inhibits apoptosis, inhibits adipogenesis, and induces osteoblastogenesis (11).
We have recently reported that SPARC regulates the activity of integrin-linked kinase (ILK) in lung fibroblasts (12). Another group also demonstrated that ILK activity mediates oncogenic effects of SPARC in glioma cells (13). ILK also regulates the ␤-catenin pathway through its phosphorylation of GSK3␤, an inhibition resulting in the stabilization of ␤-catenin (14,15). Further accumulation of free ␤-catenin in the cytoplasm is a consequence of the inhibition of E-cadherin production by ILK (14,16). SPARC represses expression of E-cadherin and promotes tumorigenesis in melanoma cells (17). Therefore, we hypothesized that SPARC could inhibit adipogenesis through ILK-␤-catenin-mediated signaling.
Herein we have established that SPARC inhibits adipogenesis and enhances osteoblastogenesis. SPARC not only retarded morphological changes in preadipocytes but also inhibited the expression of most adipocyte transcription factors and other adipocyte-specific genes. Significantly, SPARC inhibited the degradation of ␤-catenin and enhanced its nuclear translocation. ILK was required for the effect of SPARC on ␤-catenin accumulation. Consistent with its effects in other tissues (18 -20), SPARC also regulated the production of ECM proteins and integrins, in an ILK-dependent manner. These data identify a novel pathway by which SPARC inhibits adipogenesis.

EXPERIMENTAL PROCEDURES
Mice-A colony of WT and SPARC-null mice has been described (21). C57BL/6 WT and SPARC-null mice are maintained in a specific pathogen-free facility. All animal experiments conformed to NIH guidelines and were approved by the Institutional Animal Care and Use Committee. These WT and SPARCnull mice have a mixed genetic background (129SV ϫ C57BL/6) with four to ten backcrosses into the C57BL/6 background.
Isolation and Differentiation of Preadipocytes from Adipose Tissue-Stromal-vascular cells (SVCs) from mouse adipose tissue were prepared as described (25). The subcutaneous inguinal fat deposits from 3ϳ4-week-old male mice were dissected under sterile conditions, and the lymph nodes were carefully removed. The SVCs were obtained from the minced fat tissues by collagenase (Invitrogen) digestion (2 mg/ml at 37°C for 60 min in DMEM, with orbital agitation). The resulting cell suspension was homogenized by pipetting with a 10-ml serological plastic pipette, passed through a 100-mm nylon filter, and centrifuged at 400 ϫ g for 10 min. The floating (top) layer of mature adipocytes was removed, and the pellets were resuspended. Erythrocytes were removed by exposure to hypotonic buffer for 1 min. The cells were plated at a density of 3 ϫ 10 4 cells/cm 2 in DMEM/F12 supplemented with 10% fetal bovine serum, penicillin G, and streptomycin sulfate. When the cells reached confluence, differentiation was initiated by the addition of 0.1 mM dexamethasone, 0.25 mM 3-isobutyl-1-methylxanthine (IBMX), and 17 nM insulin in DMEM/F-12 containing 10% fetal bovine serum. After 48 h, the differentiation medium was replaced by DMEM/F-12 containing 10% fetal bovine serum and insulin only.
Oil-Red-O Staining-Cells were fixed in 10% formalin for 10 min. After two washes in water, cells were stained with 0.5% Oil-Red-O in isopropanol:H 2 O (3:2, v:v) for 1 h. After washes with water to remove unbound dye, cells were photographed under a Leica inverted microscope equipped with a digital camera.
Glycerol-3-phosphate Dehydrogenase (GPDH) Activity Assay-Cells were lysed in 50 mM Tris-HCl (pH 8.0) containing 1 mM 2-mercaptoethanol and 1 mM EDTA. Next, cell lysates were sonicated and centrifuged to clear debris. The supernatant was incubated with 0.12 mM ␤-NADH and 0.2 mM dihydroxyacetone phosphate in 100 mM triethanolamine/HCl buffer (pH 7.5), and the reaction was monitored by absorbance at 340 nm.
Immunoblotting-Cells cultured in media were lysed in radiolabeled immunoprecipitation assay buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.25% sodium deoxycholate, 0.1% Nonidet P-40, and a mixture of protease inhibitors), cleared by centrifugation, and analyzed by SDS-PAGE followed by detection with a specific antibody as detailed in the figure legends.
Immunostaining-For immunohistochemistry, WAT was fixed with methyl Carnoy's solution for 24 h and subsequently embedded in paraffin. Five-m-thick sections were deparaffinized and incubated with control goat IgG or goat anti-mouse SPARC IgG. Bound primary antibodies were detected with horseradish peroxidase-conjugated donkey anti-goat IgG, and histochemical reactions were performed with 3, 3Ј-diaminobenzidine (Vector Laboratories, Burlingame, CA) as the substrate. Sections were counterstained with hematoxylin.
For immunocytochemistry, preadipocytes were plated on coverslips. At certain stages of differentiation, cells were fixed with 10% formalin for 10 min followed by permeabilization with 20% Aquablock and 0.1% Triton X-100. Specific primary antibodies were used as indicated, with fluorescein isothiocyanate-or rhodamine-conjugated secondary antibodies. Sections and cells were examined with a Leica DMR microscope, and images were captured digitally with an RT-Spot camera (Diagnostic Instruments, Sterling Heights, MI).
Alkaline Phosphatase Staining-After 2 washes with phosphate-buffered saline, cells were fixed in methanol/acetone (1:1) for 10 min at Ϫ20°C. The cells were subsequently washed twice with water and stained with 5-bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium alkaline phosphatase substrate buffer (Sigma) for 45 min at 37°C. Cells were photographed under a Leica inverted microscope equipped with a digital camera.
ILK Activity Assay-200 -250 g of cell lysate was incubated with 5 g of rabbit polyclonal anti-ILK antibody (Cell Signaling Technology) or IgG isotype control, and protein A-Sepharose beads (Amersham Biosciences). Complexes were washed with high salt lysis buffer and kinase reaction buffer, and kinase assays were performed with 5 g of myelin basic protein as a substrate. Kinase activity was measured by immunoblotting with a specific antibody against phosphorylated myelin basic protein (Upstate Biotechnology).
Subcellular Fractionation-Cells were separated into cytosolic and nuclear fractions (NUCLEI EZ Prep Nuclei Isolation kit, Sigma). The cells were scraped into Nuclei EZ lysis buffer and were pelleted by centrifugation after two washes with ice-cold serum-free DMEM. Cell pellets were washed again with Nuclei EZ lysis buffer. The nuclei were subsequently pelleted by cen-FIGURE 1. Expression of SPARC in WAT. A, SPARC protein in WAT as shown by immunohistochemistry of epididymal WAT from WT (ϩ/ϩ) and SPARCnull adult mice (Ϫ/Ϫ). Tissue sections were stained with normal goat IgG or goat anti-mouse SPARC (SPC) antibody. Scale bar, 50 m. B, WT preadipocytes were cultured in GM or adipogenic medium (D2) for 2 days. Cells were subsequently fixed, rendered permeable, and exposed to anti-SPARC IgG and fluorescein isothiocyanate-conjugated secondary antibody. Scale bar, 20 m. C, expression of SPARC during adipogenesis evaluated by immunoblot. Preadipocytes from WT mice were induced to differentiate into adipocytes. Cells were lysed at 0 min, 10 min, 6 h, and at 1, 2, 3, 4, and 6 days after the induction of differentiation. Total protein concentration was determined by a BCA assay, and equal amounts of cell lysate were resolved by SDS-PAGE and probed for SPARC, ␤-actin, phosphorylated ERK1/2 (pERK1/2), and total ERK1/2. trifugation at 4°C. Nuclear fractions were solubilized with cold Nuclei EZ storage buffer. Proteins comprising the different fractions were analyzed by SDS-PAGE under reducing conditions and, subsequently, by immunoblotting.
Transfection of Small Interfering RNAs (siRNAs)-Equimolar amounts of nonspecific siRNAs or siRNAs targeting murine ILK or SPARC (Qiagen) were incubated with Hiperfect Transfection Reagent (Qiagen) according to the manufacturer's instructions. The oligo mixtures were added to preadipocyte cultures 18 h after plating the cells. Thirty hours later the media were replaced with standard differentiation media as described above.
Luciferase Reporter Gene Assay-5 ϫ 10 5 cells were transfected with 3 g of Super 8ϫ TOPflash (From Dr. Randall T. Moon, University of Washington, Seattle, WA) and 20 ng of pRenilla-TK (Promega, Madison, WI) by electroporation. Super 8ϫ TOPflash contains eight TCF/LEF transcription factor binding sites upstream of the luciferase gene. Forty-eight hours later, cells were lysed in passive lysis buffer (Promega).
Luciferase/Renilla assays were performed with the Dual-luciferase reporter assay kit (Promega). The average ratio of luciferase activity (relative light units) to Renilla activity was calculated. Experiments were repeated more than three times in triplicates.
ECM Protein Deposition-For quantification of ECM proteins, cells were plated into 6-well tissue culture plates and were induced to differentiate as described above. The cells in each well were removed by repeated washing with cold Mg 2ϩ /Ca 2ϩ -free phosphate-buffered saline. Detached cells were counted and subsequently lysed with a lysis buffer (1% Nonidet P-40, 0.1% SDS, 10 mM Tris-HCl, pH 7.5, and 5 mM EDTA) containing a complete protease inhibitor mixture (Roche Applied Science). Proteins deposited on the plates were collected by scraping with a policeman into radiolabeled immunoprecipitation assay buffer. Protein concentrations were determined by a bicinchoninic acid (BCA) assay (Pierce). Proteins were resolved by SDS-PAGE and subjected to immunoblotting with antibodies against LN-1 or cellular FN.

SPARC Inhibits Adipogenesis and Enhances Osteoblastogenesis-SPARC
is expressed extensively in WAT and in most adipocytes ( Fig. 1A) as previously reported (6,7). Given the high levels of SPARC in WAT and the data linking SPARC with obesity (4, 6, 7), we asked whether SPARC controlled one (or more) of the processes contributing to adipogenesis. We separated SVCs from murine WAT and established an in vitro adipogenesis model as described (25). SVCs are composed mainly of preadipocytes, as verified by immunostaining (data not shown); contamination by either macrophages or endothelial cells is less than 5%, according to staining for CD31 and F4/80, respectively. At least 50% of cells differentiate into adipocytes by day 7 after induction of adipogenesis, although the actual differentiation capacity of cells varied between preparations by ϳ10%. More than 90% of WT SVCs expressed SPARC. SPARC protein was distributed in a diffuse pattern in cells maintained in growth medium (GM), whereas the location of SPARC appeared more concentrated in the endoplasmic reticulum/Golgi-like compartment surrounding nuclei at day 2 of adipogenesis ( Error bars represent one S.D. from the mean for a sample population of n ϭ 3. **, p Ͻ 0.005, Student's t test, for SPARC-treated cells compared with BSA-treated cells. #, p Ͻ 0.005, Student's t test, for cells treated with partially denatured SPARC compared with active SPARC-treated cells (20 g/ml). C, concentration-dependent inhibitory effect of SPARC. SPARC was added to cultures at different concentrations (0, 10 ng/ml, 100 ng/ml, 500 ng/ml, 1 g/ml, or 2 g/ml). GPDH activity was measured at day 7. The level of endogenous SPARC in WT preadipocytes is indicated by a dashed line. D, anti-SPARC IgG partially blocks the inhibitory effect of SPARC. Rabbit anti-SPARC IgG (Ab) was added to adipogenic medium together with BSA (D/BSA) or SPARC (D/SPC) at a 5:1 molar ratio (Ab:SPARC). GPDH activity was measured at day 7. Error bars represent one S.D. from the mean. Assays were performed twice in triplicate. *, p Ͻ 0.05, Student's t test.
ent stages: the first, between 6 and 24 h, and the second, at days 3-4 of adipogenesis (Fig. 1C). The concentration of SPARC in conditioned media on day 1 of adipogenesis is ϳ200 ng/ml. Levels of ␤-actin are diminished during differentiation due to the major cytoskeletal remodeling required for this process (26). ERK1/2 is phosphorylated and activated only at an early stage of differentiation when mitotic clonal expansion occurs ( Fig. 1C and supplemental Fig. 2) (27). On day 7 of differentiation, cultured SPARC-null cells were fixed and stained with Oil-Red-O or were extracted for GPDH activity assays. Purified SPARC protein was added simultaneously to the culture with the differentiation medium and was replenished every 2 days. SPARC protein did not change the cellular apoptotic index, according to an acridine orange assay (data not shown). As shown in Fig.  2, A and B, 2 g/ml SPARC significantly inhibited adipogenesis in these primary cultures. Both hevin (a SPARC homolog) and SPARC proteins are purified from insect cells. Hevin, as well as bovine serum albumin (BSA), were therefore used as controls, neither of which showed effects on adipogenesis, and SPARC protein after partial denaturation exhibited significantly less activity (Fig. 2B). The inhibitory effect of SPARC was concentrationdependent (Fig. 2C). The endogenous level of SPARC (200 ng/ml) should exert certain inhibitory effects on adipogenesis; consequently, anti-SPARC IgG rescued adipogenesis in the presence of SPARC and blocked SPARC activity by 80% (Fig. 2D). Preadipocytes and mesenchymal stem cells differentiate into osteoblasts under long term culture. SPARC significantly enhanced osteoblast formation 21 days after the induction of differentiation (supplemental Fig. 3). These data support previous reports that SPARC inhibits adipogenesis and enhances osteoblastogenesis (4,5,28).
To confirm the effect of SPARC on adipocyte differentiation, we compared the differentiation in vitro of preadipocytes from WT and SPARC-null mice. Preadipocytes were isolated from epididymal WAT of each genotype and were induced to differentiate into adipocytes. After 7 days, more adipocytes and more fat accumulation were observed in cultures of SPARC-null preadipocytes compared with those of WT preadipocytes (Fig.  3, A-C). We also tested the capacity of these preadipocytes to differentiate into osteoblasts. On day 28, cells were stained for the osteoblast marker, alkaline phosphatase. There were fewer stained cells in SPARC-null preadipocytes (Fig. 3, D and E). These data further demonstrate that SPARC regulates mesenchymal stem cell lineage commitment.
SPARC Inhibits Adipocyte Gene Expression and Enhances Osteoblast Gene Expression-Adipogenesis requires the sequential expression of adipogenesis transcription factors and adipocyte genes. We next examined the levels of specific mRNAs in the presence or absence of SPARC. By RT-PCR (Fig.  4A), the transcription factors C/EBP␣, C/EBP␤, and PPAR␥2 were all significantly decreased by SPARC, whereas C/EBP␦ was unchanged. Adipocyte genes such as leptin and lipoprotein lipase were also substantially decreased by SPARC. Conversely, genes required for osteoblast differentiation were enhanced by SPARC (Fig. 4B). RUNX2 is one of the key transcription factors for osteoblast formation (29). SPARC enhanced the expression of RUNX2, and the osteoblast-specific gene osteocalcin was also stimulated by SPARC. These data support the claim that SPARC blocks the expression of certain genes associated with adipocyte differentiation and enhances that of two osteoblastspecific genes.

FIGURE 4. SPARC inhibits adipocyte gene expression (A) and enhances osteoblast gene expression (B).
Semiquantitative RT-PCR on RNA from mouse preadipocytes cultured in GM or adipogenic medium in the presence of 2 g/ml BSA or 2 g/ml SPARC (SPC). A, C/EBP␤ and C/EBP␦ mRNAs were derived from preadipocytes at day 2 of differentiation; PPAR␥2, C/EBP␣, leptin, lipoprotein lipase (LPL), and 36B4 mRNAs were derived from preadipocytes at day 7 of differentiation. Below each band is the comparative band density, which was normalized to that of 36B4 to allow quantification. B, levels of mRNAs were determined by RT-PCR on days 2, 4, and 6 of preadipocyte differentiation. Band density was normalized to that of 36B4 to allow quantification. Numbers represent the comparative ratio of band densities for the two conditions (ϩ/ϪSPARC).
The differentiation mixture is composed of insulin, dexamethasone, and IBMX, each of which activates complex downstream signaling. To identify signaling pathways affected by SPARC, we first probed the effect of SPARC in differentiation medium lacking one or more of these factors. As shown in supplemental Fig. 4, SPARC appeared to inhibit primarily IBMXinduced differentiation. This compound is required for the maximal differentiation of preadipocytes. In contrast, SPARC did not significantly affect the differentiation induced by insulin alone or insulin together with dexamethasone. Therefore, SPARC exhibits significant inhibitory effects only in the presence of IBMX.
SPARC Retards Changes in the Morphology of (Pre)adipocytes during Adipogenesis-Fibroblast-like preadipocytes undergo major changes in morphology to differentiate into sphere-shaped adipocytes. During this process, cell adhesion is weakened, and actin stress fibers are decreased to allow fat accumulation in the cytosol. Given the function of SPARC in the modulation of cell-ECM interactions, we next examined whether SPARC affected cytoskeletal remodeling and cell adhesion during adipogenesis. We stained cells for the focal adhesion protein vinculin and for F-actin (Fig.  5). On day 2 of adipogenesis, the differentiation mix (D/BSA) triggered preadipocytes to lose most of their focal adhesions; actin stress fibers were diminished, and ruffles were formed at the cell periphery. However, in the presence of SPARC (D/SPC, 2 g/ml), the configuration of vinculin-containing focal adhesions and actin stress fibers differed from those observed in both growth medium (GM; top panel) and differentiation medium (D/BSA; middle panel). Cells retained some vinculin-containing focal adhesions, although their number, size, and density were significantly decreased (Fig. 5B). Moreover, the rearrangement of F-actin from stress fibers (top panel), largely to a cortical configuration (middle panel), was partially inhibited by SPARC (bottom panel) (Fig. 5A) as previously described (12). Therefore, SPARC interferes with the changes in cell morphology necessary for adipogenesis. SPARC Promotes ILK Activity at an Early Stage of Adipogenesis-SPARC regulates ILK activity in both lung fibroblasts and glioma cells (12,13). Does SPARC regulate ILK activity during adipogenesis? We first tested the expression levels of ILK. There were no obvious changes in the level of ILK protein during adipocyte differentiation or in the presence of SPARC protein (Fig. 6A). However, at 10 min after the induction of differentiation, ILK activity was enhanced by SPARC (Fig. 6B).  We also tested two downstream targets of ILK, namely GSK-3␤ and AKT. The phosphorylation of AKT was not changed by SPARC during early stages of adipogenesis (data not shown). ILK can phosphorylate GSK3␤ at Ser-9 to inactivate the kinase and inhibit its effects on the degradation of ␤-catenin (14). Consistently, SPARC enhanced the Ser phosphorylation of GSK3␤ 6 h after the induction of differentiation (Fig. 6C). SPARC Induces ␤-Catenin Accumulation and Nuclear Translocation-Enhancement of the Wnt/␤-catenin pathway inhibits adipogenesis and increases osteoblastogenesis (8). The levels of ␤-catenin increase during the first 8 h and subsequently decrease during adipogenesis (14,30); ␤-catenin is also one of the identified downstream targets of ILK (14). We therefore asked whether SPARC regulates adipogenesis through a pathway that includes ILK and ␤-catenin. As shown in Fig. 7A, SPARC enhanced the accumulation of ␤-catenin in total cell lysates during the first 8 -24 h of differentiation. Moreover, in the presence of SPARC, levels of ␤-catenin in the nuclear fraction were higher at 8 h of differentiation (Fig. 7B). At 24 h of differentiation, the ␤-catenin signal was decreased in the cytosol, relative to that of cells in growth medium (Fig. 7D, D versus GM). In contrast, in the presence of SPARC, ␤-catenin was observed in the nucleus and cytosol immediately surrounding the nucleus (Fig. 7D, D/SPC). Consistently, levels of ␤-catenin were lower in SPARC-null preadipocytes compared with those in WT preadipocytes at 8 h after the induction of differentiation (Fig. 7C).
␤-Catenin translocates into the nucleus to bind to a member of the TCF/LEF transcription factor family and activates its transcriptional activity. In the presence of SPARC, significantly more TCF4 protein was coimmunoprecipitated with anti-␤catenin antibody (Fig. 8A). Moreover, SPARC enhanced the transactivation of the TCF binding element in a Super 8ϫ TOPflash construct. Although the effect of 2 g/ml SPARC on TOPflash activity was less robust than that of 20 mM LiCl, SPARC elicited a consistent increase in TOPflash activity that was statistically significant (Fig. 8B). Taken together, these data indicate that SPARC enhances ␤-catenin-mediated signaling.
ILK Activity Is Required for the Inhibitory Effect of SPARC during Adipogenesis-Rho and Rho kinase regulate adipogenesis through the regulation of cell shape (31). However, inhibitors of Rac (NSC23766), Rho kinase (Y27632), and phosphoinositide 3-kinases (LY294002) did not block the inhibitory effects of SPARC (data not shown). KP747, a specific inhibitor of ILK (32), alone did not enhance adipogenesis; however, it partially blocked the inhibitory effect of SPARC upon adipocyte differentiation. As shown in Fig. 9A, 0.5 g/ml (0.012 M) SPARC inhibited adipogenesis by 60%. In the presence of 10 M KP747, SPARC inhibited adipogenesis by 20%. Increasing concentrations of KP747 further blocked the inhibitory effects of SPARC, although the ILK inhibitor did not effect a complete inhibition (Fig. 9A).
To test whether SPARC induces the accumulation of ␤-catenin through ILK, we used specific siRNA targeting ILK. Nonspecific siRNA or siRNA-targeting ILK were transfected into cells 30 h before the induction of differentiation. Levels of ILK expression were decreased by more than 60% 8 h after the induction (Fig. 9B). siRNA targeting ILK, but not the nonspecific siRNA, blocked the effect of SPARC on ␤-catenin accumulation (Fig. 9, B and C). These data indicate that ILK is required for SPARC to induce the accumulation of ␤-catenin at an early stage of adipocyte differentiation.
SPARC Regulates the Expression, Deposition, and Organization of ECM Proteins-ECM proteins interact with cell surface receptors, especially integrins, and transduce outside-in signaling to regulate cell function. During adipogenesis, extensive ECM remodeling modulates morphological changes in preadipocytes. Flat fibroblast-like preadipocytes round up into spherical adipocytes, focal adhesions disappear, and stress fibers are lost. ECM is converted from a FN-rich stromal matrix into a LN-rich basement membrane. Given the reports that FN inhibits adipogenesis and LN/Matrigel enhances adipogenesis (26,33), we asked whether SPARC modulates FN and LN production during this process.
The assembly of basement membrane starts with the selfpolymerization of LN (34,35). The major type of LN expressed in adipose tissue is LN-8 (␣4␤1␥1) (36), which has been detected in basement membrane but does not self-polymerize (34). In most tissues, LN-1 (␣1␤1␥1) is the component that initiates the assembly of the basement membrane. Thus, we examined the expression of LN ␣1 and ␣4 chains during adipocyte differentiation. In (pre)adipocytes, the level of the LN ␣1 chain was lower than that of the LN ␣4 chain (Fig. 10A). In primary culture, LN ␣1 chains were detected on the surface of mature adipocytes as a reticular network (data not shown), and . SPARC regulates ILK activity during early stages of adipocyte differentiation. A, ILK protein and GAPDH were measured by immunoblotting at 0, 6 h, and 24 h after the induction of adipogenesis in the presence or absence of SPARC (2 g/ml). B, 10 min and 6 h after the induction of differentiation ILK activity was measured in preadipocytes by a kinase assay on an immunoblot with an anti-phospho-myelin basic protein (pMBP) antibody. Below each band is the comparative band density. C, phosphorylated GSK3␤ (pGSK3␤, Ser-9), total GSK3␤, and GAPDH were measured by immunoblotting at 0, 30 min, and 6 h after the induction of adipogenesis in the presence or absence of SPARC (2 g/ml). Band density was normalized to that of GAPDH to allow quantification. Numbers represent the comparative ratio of band densities for the two conditions (ϩ/ϪSPARC).
the expression of the ␣1 chain was increased during adipogenesis (Fig. 10A), data indicating that the ␣1 chain might initiate the assembly of basement membrane. SPARC significantly inhibited the expression of the LN ␣1 chain (Fig. 10A), consistent with our data that SPARC inhibits the formation of intact, functional basement membrane in the lens (37,38). Conversely, expression of the ␣4 chain was slightly enhanced by SPARC, and that of the ␥1 chain was enhanced on day 6 of adipogenesis (Fig. 10A).
The total LN protein deposited on the culture dish was also diminished by SPARC (Fig. 10B). In contrast, the deposition of FN was enhanced by SPARC (Fig. 10B), although the mRNA levels of FN were not affected (data not shown). SPARC also influenced the formation of LN networks around cells. As shown in Fig. 10C, the formation of a LN reticular network on the adipocyte surface and of a more extensive intercellular LN matrix among undifferentiated cells was diminished in the presence of SPARC. These results support the claim that SPARC inhibits adipogenesis in part by its modulation of ECM protein expression and deposition.
SPARC Regulates the Expression of Integrins-␣5-and ␣6-containing integrins are the major receptors for FN and LN, respectively. We next examined the effect of SPARC on the expression and activity of ␣5 and ␣6 integrin subunits during adipocyte differentiation. ␣5 was decreased during adipocyte differentiation (Fig. 11A), consistent with the report that overexpression of the ␣5 integrin subunit significantly inhibits adipogenesis (39). SPARC enhanced the expression of ␣5 integrin on days 1-2 and 6 of adipogenesis (Fig. 11, A and B). In contrast to ␣5 integrin, levels of ␣6 integrin are generally enhanced during adipocyte differentiation (39) (Fig. 11A), and the expression of ␣6 integrin was significantly diminished in the presence of SPARC at later stages of adipogenesis (Fig. 11, A and B). Therefore, the capacity of SPARC to regulate the selective expression of ␣ integrin subunits is consistent with its effects on the corresponding ECM ligands and contributes to its inhibitory activity on adipogenesis. SPARC Regulates the Expression of Integrins through ␤-Catenin-mediated Signaling during Adipogenesis-Wnt/ ␤-catenin inhibits adipogenesis; however, it is not clear whether Wnt/␤-catenin regulates the expression of integrins and ECM remodeling during adipogenesis. We next measured the expression of ␣6 integrin in LiCl-treated cells during adipocyte differentiation. LiCl is an ATP noncompet-  A and B, 0, 8 h, 24 h, and 48 h after the induction of adipogenesis in the absence (Ϫ) or presence (ϩ) of SPARC, SPARC-null cells were lysed or were separated into nuclear and cytosolic fractions. A, levels of ␤-catenin and GAPDH were determined by immunoblotting of total cell lysates. B, levels of ␤-catenin, GAPDH, and histone 1 were determined by immunoblotting of cytosolic and nuclear protein extracts. C, preadipocytes from WT (ϩ/ϩ) and SPARC-null (Ϫ/Ϫ) mice were induced to differentiate under the same conditions described above. Levels of ␤-catenin, endogenous SPARC (mSPARC), and GAPDH were measured by immunoblotting at 8 h and 24 h after the induction of differentiation. D, SPARC-null preadipocytes were cultured in GM or adipogenic medium in the absence (D) or presence of 2 g/ml SPARC (D/SPC) for 24 h. Cells were subsequently fixed, rendered permeable, and exposed to anti-␤-catenin IgG and fluorescein isothiocyanate-conjugated secondary antibody and Hoechst 33258 dye (blue, right panel). Scale bar, 20 m.
itive inhibitor of GSK-3␤ and leads to the accumulation of ␤-catenin (40,41). On day 2 of adipogenesis, 20 mM LiCl significantly inhibited the expression of ␣6 integrin (Fig.  12A). In addition, LiCl enhanced the SPARC-induced decrease in the expression of ␣6 integrin on day 4 of adipogenesis (Fig. 12A). Activation of ILK increases the accumulation of ␤-catenin through GSK-3␤. To test whether ILK regulates the expression of relevant integrins, we used specific siRNA targeting ILK. RNA interference of ILK blocked the inhibitory effect of SPARC on the expression of ␣6 integrin (Fig. 12B). These data indicate that ␤-catenin-mediated signaling influences the expression of certain integrins during adipogenesis. SPARC, therefore, regulates ECM remodeling through an ILK/␤-catenin-mediated signaling pathway.

DISCUSSION
SPARC is a matricellular protein that regulates cell-ECM interaction and, thereby, modulates changes in cell morphology. SPARC-null mice show increased fat accumulation. The precise biological and molecular mechanisms responsible for this phenotype remain largely unknown. In this study we have uncovered a novel function of SPARC in adipocyte differentiation. Exogenous SPARC inhibits the differentiation of mouse primary preadipocytes. Primary preadipocytes from SPARCnull mice exhibit an enhanced capacity for adipocyte differentiation and a diminished propensity toward osteoblast differentiation. SPARC inhibits the expression of key adipogenic transcription factors, such as C/EBP␤, C/EBP␣, and PPAR␥ and adipocyte-specific genes. SPARC also induces the accumu-lation and nuclear translocation of ␤-catenin, a pathway controlled in part by ILK, a known target of SPARC (12). Subsequently, SPARC enhances the deposition of FN and the expression of ␣5 integrin and inhibits the expression of LN ␣1 chain and ␣6 integrin, the latter couple necessary for the formation of a basal lamina by mature adipocytes. Previous studies have shown that stimulation of FN assembly by SPARC is ILKdependent (12).
SPARC inhibits adipogenesis by its enhancement of Wnt/␤catenin signaling. Accumulation of ␤-catenin was increased in the presence of exogenous or endogenous SPARC. At an early stage of adipogenesis, SPARC inhibited IBMX-induced differentiation (supplemental Fig. 4). One of the effects of IBMX is the induction of C/EBP␤ during the first few hours of differentiation (42). Consistently, SPARC inhibited the expression of C/EBP␤ at an early stage of this process. C/EBP␤ expression in the presence of dexamethasone contributes to the induction of PPAR␥ and decreased levels of ␤-catenin (30). During early adipogenesis, the decrease in the abundance of ␤-catenin coincides with the accumulation of PPAR␥ (30). PPAR␥ can suppress Wnt signaling by its targeting of phosphorylated ␤-catenin to the proteasome (30,43,44). Conversely, ␤-catenin not only inhibits expression of PPAR␥ but also inhibits its transcriptional activity through a direct interaction involving its TCF/LEF binding domain (44). In addition, Wnt/␤-catenin enhances osteoblastogenesis via repression of the expression of   C/EBP␣ and PPAR␥ (45). Consistently, SPARC substantially inhibited the expression of PPAR␥ and C/EBP␣ at later stages of differentiation. These data provide further evidence that SPARC inhibits the key transcriptional cascades of adipogenesis through the Wnt/␤-catenin pathway. The Wnt/␤-catenin pathway also plays an important role during skeletogenesis. Inactivation of ␤-catenin in mesenchymal stem cells prevents osteoblast differentiation (46). During postnatal development, ␤-catenin also promotes the capacity of differentiated osteoblasts to inhibit osteoclast differentiation (47,48). Consistently, conditional inactivation of ␤-catenin in skeletal progenitors or inactivation of Lrp5 (a Wnt receptor) results in low bone mass (osteopenia) (46). The ␤-catenin-TCF/ LEF complex regulates the expression of the osteogenesis genes RUNX2, RANK ligand, osteoprotegerin, and osteocalcin (47,49,50). Similarly, SPARC-null mice develop osteopenia that becomes more severe as the animals age (28). Consistent with our data, mesenchymal stem cells from SPARC-null bone marrow show an increased tendency to form adipocytes and a decreased tendency to form osteoblasts (5). Therefore, SPARC could regulate osteoblast formation, maturation, and survival through a Wnt/␤-catenin pathway.
SPARC interacts not only with ECM components but also with integrins and growth factors. SPARC interacts with ␤1 integrin and subsequently enhances the activation of ILK (51). In preadipocytes, SPARC-mimicking peptides bind to ␣5␤1 integrin (52). SPARC is also involved in growth factor signaling; e.g., it binds to platelet-derived growth factor and vascular endothelial growth factor and inhibits the binding of these growth factors to their cognate receptors (53)(54)(55)(56).
Our data now establish that SPARC enhances the activity of ILK at an early stage of adipocyte differentiation, and an ILK-specific inhibitor partially blocks the inhibitory effects of SPARC. ILK can be activated through an interaction with the cytoplasmic tail of ␤1/␤3 integrins; alternatively, insulin, plateletderived growth factor, and other growth factors can stimulate ILK activity through phosphoinositide 3-kinases (57,58). Thus, SPARC might enhance ILK activity through ␤1/␤3 integrins or certain growth factors and/or their receptors. SPARC also induces the phosphorylation and activation of focal adhesion kinase in glioma cells (13). At an early stage of adipocyte differentiation, preadipocytes partially detach, focal adhesion complexes are lost, and focal adhesion kinase activity (phosphorylation) is decreased. SPARC could indeed function to maintain the activity of ILK and focal adhesion kinase in preadipocytes. Consequently, ILK activity would preserve the linkage among ECM components, integrins, and the actin cytoskeleton, with subsequent preservation of focal adhesion complexes. That the Rho kinase inhibitor did not block the inhibitory effects of SPARC on differentiation indicates that Rho kinase does not participate in the downstream signaling conferred by SPARC and ILK.
Adipogenesis is characterized by the conversion from a FNrich stromal matrix to a LN-rich basal lamina; consistently, integrin expression switches from ␣5 (FN) to ␣6 (LN) (39 . SPARC regulates ECM production during adipogenesis. A, mRNA levels of LN␣1, LN␣4, LN␥1, and 36B4 were determined by RT-PCR from total RNA isolated from preadipocytes cultured in adipogenic medium in the presence (ϩ) or absence (Ϫ) of SPARC at days 2, 4, and 6. Band density was normalized to that of 36B4 to allow quantification. The number represents the comparative ratio of band densities under two conditions (ϩ/ϪSPARC). B, 24 h after induction of differentiation preadipocytes were detached in 5 M EDTA in phosphate-buffered saline. ECM proteins on the culture plates (ECM) and in the cell lysates (cell) were solubilized in SDS-PAGE buffer containing 0.5% ␤-mercaptoethanol and were analyzed by SDS-PAGE. LN (LN-␤/␥) and cellular FN were detected by immunoblotting. Numbers represent the comparative ratio of band densities under two conditions (ϩ/ϪSPARC). C, 7 days after induction of adipogenesis in the presence of 2 g/ml BSA (BSA, a and c) or 2 g/ml SPARC (SPARC, b and d), cells were exposed to anti-murine LN␣1␤1␥1 followed by Cy3-conjugated secondary antibody (red) and Hoechst 33258 dye (blue). a and b, mature adipocytes; c and d, undifferentiated cells. Scale bar, 20 m. and poly-L-lysine exert inhibitory effects on adipogenesis (1,33,39). The inhibitory activity of FN requires cell spreading, and cytochalasin D, via its disruption of actin filaments, overcame the inhibitory effects of FN (26). Remodeling of ECM proteins induces changes in the cytoskeleton. FN is linked to actin stress fibers, whereas LN polymerization on the cell surface promotes the formation of a cortical actin network (59). SPARC modulates ECM protein expression and deposition. For example, SPARC regulates the expression and fibrillogenesis of type I collagen. The dermis of SPARC-null mice exhibits decreased collagen fibril diameters and a reduction in tensile strength (20). During adipogenesis, we found that SPARC enhanced the deposition of FN and the expression of its receptor, ␣5 integrin. In lung fibroblasts, SPARC enhanced cell-mediated partial unfolding of FN molecules and FN-induced stress fiber formation through ILK-dependent contractile signaling (12). The unfolding of FN molecules is an important step in FN fibril formation. Here we suggest that SPARC might also enhance this process through ILK signaling and contribute to the increased deposition of FN. PPAR␥ inhibits transcription of the ␣5 integrin gene in lung carcinoma cells (60). Therefore, the increase in ␣5 integrin expression that we observed would be a downstream effect of the diminished expression of PPAR␥ elicited by SPARC.
We also found that SPARC inhibits the expression of the LN ␣1 chain and one of its major receptors, ␣6 integrin, as well as the deposition of LN itself. SPARC regulates the secretion and deposition of LN in lens cells (37). The expression of ␣6 integrin correlates with the growth-arrest of preadipocytes during adipogenesis, and it favors differentiation over proliferation of preadipocytes (39). ␣6 integrin and LN enhance adipogenesis and lipid accumulation. The interaction of LN and ␣7␤1 integrin is required for basal lamina formation and actin reorganization in muscle cells (59). Similarly, ␣6 integrin might be required for receptor-facilitated LN self-assembly, basal lamina formation, actin reorganization, and other adipogenic signaling events. Therefore, SPARC inhibits adipogenesis by its interference with the formation of basal lamina.
Does Wnt/␤-catenin signaling regulate ECM remodeling and the expression of integrin and ECM components during adipogenesis? Although it is not clear whether Wnt/␤-catenin directly regulates the expression of integrins or any ECM components, we have presented evidence that ␤-cateninmediated signaling is involved in the regulation of integrin expression. LiCl, a strong inducer of ␤-catenin signaling, significantly inhibited the expression of ␣6 integrin. ILK activity, which also enhances ␤-catenin accumulation, is required for SPARC to inhibit the expression of ␣6 integrin. SPARC could regulate ECM remodeling through ␤-cateninmediated signaling, an effect that might be exerted through PPAR␥ (44).
In conclusion, we have shown that 1) SPARC inhibits adipogenesis and enhances osteoblastogenesis, 2) SPARC activates ILK, resulting in the accumulation and nuclear translocation of ␤-catenin, 3) SPARC inhibits the subsequent ECM remodeling required for adipocyte differentiation, and 4) SPARC regulates ECM remodeling through ␤-catenin-mediated signaling. Taken together, our findings indicate that SPARC exerts inhibitory effects on adipogenesis and might, therefore, be a potential therapeutic target for obesity and related metabolic disorders. . SPARC modulates the expression of integrins during adipogenesis. A, levels of ␣5 and ␣6 integrin mRNA were determined by RT-PCR from total RNA isolated from preadipocytes cultured in adipogenic medium in the presence (ϩ) or absence (Ϫ) of SPARC (2 g/ml) at days (D) 2, 4, and 6. The band density was normalized to that of 36B4 to allow quantification. Numbers represent the comparative ratio of band densities for two conditions (ϩ/ϪSPARC). B, preadipocytes were lysed at days 1, 2, 4, and 6 after the induction of differentiation. Cell lysates (10 g) were resolved by SDS-PAGE under non-reducing conditions and were immunoblotted for ␣5 and ␣6 integrins. Equal protein loading was shown by staining with Ponceau S (Ponceau S).
Band density was normalized to that of the corresponding Ponceau S staining to allow quantification. Numbers represent the comparative ratio of band densities for two conditions (ϩ/ϪSPARC). FIGURE 12. ␤-Catenin-mediated signaling regulates the expression of ␣6 integrin during adipogenesis. Expression of ␣6 integrin was analyzed by semiquantitative RT-PCR. A, preadipocytes were cultured in adipogenic medium in the presence (ϩ) or absence (Ϫ) of SPARC (2 g/ml) or LiCl (20 mM) at days (D) 2 and 4. B, RNA interference of ILK blocks the inhibitory effects of SPARC on the expression of ␣6 integrin. Preadipocytes were transfected with equimolar quantities of siRNA of a non-targeting control (CK-siRNA) or ILK (ILK-siRNA) 30 h before the induction of differentiation. Cells were processed on day 2 for the isolation of total RNA. The band density was normalized to that of 36B4 to allow quantification. Numbers represent the comparative band densities.