Bi-directional Regulation of Brown Fat Adipogenesis by the Insulin Receptor*

Insulin is a potent inducer of adipogenesis, and differentiation of adipocytes requires many components of the insulin signaling pathway, including the insulin receptor substrate IRS-1 and phosphatidylinositol 3-kinase (PI3K). Brown pre-adipocytes in culture exhibit low levels of insulin receptor (IR), and during differentiation there is both an increase in total IR levels and a shift in the alternatively spliced forms of IR from the A isoform (–exon 11) to the B isoform (+exon 11). Brown pre-adipocyte cell lines from insulin receptor-deficient mice exhibit dramatically impaired differentiation and an inability to regulate alternative splicing of the insulin receptor. Surprisingly, re-expression of either splice isoform of IR in the IR-deficient cells fails to rescue differentiation in these cells. Likewise, overexpression of IR in control IRlox cells also results in inhibition of differentiation and a failure to accumulate expression of the adipogenic markers peroxisome proliferator-activated receptor gamma, Glut4, and fatty acid synthase, although cells overexpressing IR retain the ability to activate PI3K and down-regulate mitogen-activated protein kinase (MAPK) phosphorylation. Thus, differentiation of brown adipocytes requires a timed and regulated expression of IR, and either the absence or overabundance of insulin receptors in these cells dramatically inhibits differentiation.

Insulin is a potent inducer of adipogenesis, and differentiation of adipocytes requires many components of the insulin signaling pathway, including the insulin receptor substrate IRS-1 and phosphatidylinositol 3-kinase (PI3K). Brown pre-adipocytes in culture exhibit low levels of insulin receptor (IR), and during differentiation there is both an increase in total IR levels and a shift in the alternatively spliced forms of IR from the A isoform (؊exon 11) to the B isoform (؉exon 11). Brown pre-adipocyte cell lines from insulin receptor-deficient mice exhibit dramatically impaired differentiation and an inability to regulate alternative splicing of the insulin receptor. Surprisingly, re-expression of either splice isoform of IR in the IR-deficient cells fails to rescue differentiation in these cells. Likewise, overexpression of IR in control IR lox cells also results in inhibition of differentiation and a failure to accumulate expression of the adipogenic markers peroxisome proliferator-activated receptor gamma, Glut4, and fatty acid synthase, although cells overexpressing IR retain the ability to activate PI3K and down-regulate mitogen-activated protein kinase (MAPK) phosphorylation. Thus, differentiation of brown adipocytes requires a timed and regulated expression of IR, and either the absence or overabundance of insulin receptors in these cells dramatically inhibits differentiation.
Adipocytes play a central role in energy metabolism, both as an energy reserve for the storage of triglycerides and as a source of secreted hormones and cytokines that affect metabolic regulation. Although white adipose tissue is the primary site for lipid storage and fatty acid release, brown adipocytes are uniquely responsible for basal thermogenic energy expenditure through expression of the mitochondrial protein uncoupling protein-1 (UCP-1). 1 Recently, significant progress has been made in clarifying the molecular mechanisms of adipocyte differentiation, which includes the expression and activation of a series of transcription factors, including the C/AAT enhancer binding protein (C/EBP) family members ␤, ␦, and ␣ and the peroxisome proliferator-activated receptor gamma (PPAR␥). Expression and activation of these transcription factors are required for the expression of adipogenic markers such as fatty acid synthase (FAS) and the insulin-sensitive glucose transporter Glut4.
Upstream signals regulating the expression and activation of these transcription factors during adipocyte differentiation are not fully understood. However, the derivation of white and brown adipocyte cell lines that are deficient in insulin or insulin-like growth factor I (IGF-I) signaling genes have provided useful tools for determining the contribution of the insulin/ IGF-I signaling pathways to both adipogenesis and differentiation-dependent signaling. Pre-adipocyte cell lines derived from IRS-1-deficient cells are impaired in their ability to differentiate into mature adipocytes and show defects in lipid synthesis (1,2). Although IRS-2-deficient adipocytes show only modest defects in adipogenesis, insulin-stimulated Glut4 translocation and glucose uptake in mature adipocytes is dramatically reduced (3). In addition, pharmaceutical inhibition of phosphatidylinositol 3-kinase (PI3K) or expression of dominant negative mutant subunits abrogates differentiation of both white and brown adipocytes in vitro (1,4,5). Although these signaling components are downstream of both insulin and IGF-I, it is currently hypothesized that IGF-I is the more essential inducer of adipocyte differentiation, because the IGF-I receptor (IGF-IR) is the predominant receptor in preadipocytes, and high levels of insulin are required for differentiation, presumably by activity through IGF-IR. However, mice with either a total fat-specific (FIRKO) or a brown fat-specific (BATIRKO) inactivation of the insulin receptor have reduced fat mass compared with their wild type littermates (6,7). These data suggest that insulin action through the insulin receptor is a critical factor in the regulation of adipogenesis.
Insulin receptor (IR) mRNA is alternatively spliced, creating two isoforms that either contain (IR B ) or lack (IR A ) exon 11, which encodes 12 amino acids at the carboxyl terminus of the IR ␣-subunit. Functional studies have consistently shown that IR A has a 2-fold higher affinity for insulin. Additionally, IGF-II preferentially binds IR A , and IR A may have a higher internalization rate than IR B (8 -11). Expression of the IR splice isoforms is regulated both developmentally and in a tissue-specific manner in mammals (12)(13)(14). Despite the variable expression of IR isoforms in insulin-sensitive tissues, such as liver, muscle, and white fat, the regulation of splicing and function of these isoforms in brown fat is not understood.
In the current study we have investigated the direct role of the insulin receptor and its two isoforms in brown fat adipogenesis by generating immortalized brown pre-adipocytes from mice lacking the insulin receptor (FIRKO) and their control (IR lox ) littermates, as well as IR-deficient and normal preadipocytes that overexpress IR. We find that brown adipocytes undergo IR isoform switching during differentiation and that there is a critical need for the proper level of insulin receptors to initiate adipocyte differentiation. used for immunoprecipitation and immunoblotting included anti-IRS-1, anti-IRS-2, and anti-phosphotyrosine  4G10 (kindly provided by Morris White, Joslin Diabetes Center, Boston,  MA); anti-insulin receptor (kindly provided by Bentley Cheatham, Joslin Diabetes Center); anti-PPAR␥, anti-C/EBP␣, and anti-UCP1 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA); anti-Glut4 (Chemicon International, Inc., Temecula, CA); and anti-phospho-AKT and anti-phospho-MAPK (New England BioLabs, Beverly, MA). Protein A-Sepharose was purchased from Amersham Biosciences (Piscataway, NJ), and [␥-32 P]ATP was from PerkinElmer Life Sciences (Boston, MA). Phosphoinositol was obtained from Avanti Polar-Lipids, Inc. (Alabaster, AL), polyvinylidene difluoride membrane was from Fisher Scientific (Pittsburgh, PA), thin-layer chromatography plates were from VWR (Bridgeport, NJ), and electrophoresis supplies were from Bio-Rad Laboratories (Hercules, CA). Troglitazone was a gift from Warner-Lambert Co. (Ann Arbor, MI). All other supplies, unless indicated, were purchased from Sigma Chemical Co. (St. Louis, MO).

Materials-Antibodies
Cell Isolation and Culture-Cells that were homozygous for a floxed allele of exon 4 of the insulin receptor (IR lox ) were used as controls for all studies. Brown pre-adipocytes were isolated from newborn control IR lox and FIRKO mice by collagenase digestion as described previously (15). Pre-adipocytes were immortalized by infection with a pBABE retrovirus encoding SV40 T-antigen and selected with 2 g/ml puromycin. Cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% v/v fetal calf serum in a 5% CO 2 environment. For in vitro recombination of the insulin receptor, pre-adipocytes harboring a floxed allele of the insulin receptor (IR lox ) were first plated at a subconfluent density. After 24 h, cells were infected with an adenovirus encoding cre recombinase at a titer of 10 9 plaque-forming units. After 1 h the viral supernatant was replaced with culture medium. Individual colonies were selected for IR recombination using PCR with genomic DNA (16). To differentiate cells, pre-adipocytes were first grown to confluence in culture medium containing 20 nM insulin and 1 nM triiodothyronine (differentiation medium). Adipogenesis was induced by treating confluent cultures (day 0) with differentiation medium supplemented with 0.5 mM isobutylmethylxanthine, 0.5 M dexamethasone, and 0.125 mM indomethacin (induction medium) for 48 h. After this induction phase (day 2), cells were returned to differentiation medium, which was replenished every 2 days until day 6 when cells were considered mature adipocytes. For activation of PPAR␥, 25 M troglitazone was included in the induction medium for 48 h.
Oil Red O Staining-Dishes were washed once with phosphatebuffered saline and fixed with 10% buffered formalin for at least 16 h at 4°C. Cells were then stained for 4 h at room temperature with oil red O solution (5 g/liter in isopropyl alcohol), washed three times with water, and visualized.
Plasmids and Retroviral Infection-The PPAR␥ retroviral expression vector has been described previously (1). Coding sequences for the individual splice isoforms of the human IR either containing or lacking exon 11 (kindly provided by Ingo Leibiger, Karolinska University, Stockholm, Sweden) were cloned into pBABE-bleo or -hygro vectors. Subconfluent NX packaging cells were transfected by Lipofectamine TM reagent (Invitrogen, Carlsbad, CA) with 10 g of retroviral vector, and viral supernatants were collected 48 h after transfection. Control or IRKO pre-adipocytes were infected with Polybrene (4 g/ml)supplemented virus for 48 h and then placed in selection medium containing either the bleomycin analog zeocin (250 g/ml) or hygromycin (200 g/ml, Invitrogen).
Cell Lysis, Immunoblotting, and Immunoprecipitation-Cells were lysed in extraction buffer containing 50 mM HEPES, 137 mM NaCl, 1 mM MgCl 2 , 1 mM CaCl 2 , 10 mM NaP 2 O 7 , 10 mM NaF, 2 mM EDTA, 10% glycerol, 1% Igepal CA-630, 1 mM sodium orthovanadate, 10 g/ml trypsin inhibitor, 0.5 g/ml leupeptin, and 0.2 mM phenylmethylsulfonyl fluoride. Extracts were sonicated and clarified by centrifugation for 20 min, and protein concentrations were determined by the Bradford method (18). For immunoblotting of whole cell lysates, equal amounts of protein were directly solubilized in Laemmli sample buffer and resolved on SDS-containing polyacrylamide gels. Protein was transferred to polyvinylidene difluoride membranes and immunoblotted using the indicated primary antibodies. Blots were incubated with the appropriate peroxidase-conjugated secondary antibody and developed with enhanced chemiluminescence. For immunoprecipitations, 500 g of protein was incubated with the indicated antibodies and protein A-Sepharose for 16 h at 4°C. Immune complexes were collected by centrifugation, washed in extraction buffer, solubilized in Laemmli sample buffer, and analyzed as above.
PI3K Assays-Assays were completed as previously described (1) with the following modifications. Reactions were terminated by the addition of 150 l of chloroform:methanol:12 N HCl (100:200:2, v/v). The organic phase was washed once with methanol:1 N HCl (1:1) and concentrated by centrifugation under vacuum. The lipids were resolved with chloroform and spotted onto a silica gel thin-layer chromatography plate. The plate was developed in CHCl 3 :MeOH:H 2 O:NH 4 OH (120:94: 23:2.4, v/v), dried, exposed to a PhosphorImager screen, and quantitated with a Molecular Dynamics densitometer.

RESULTS
Alternative Splicing of the Insulin Receptor Is Regulated during Brown Fat Adipogenesis-Insulin receptor-deficient brown pre-adipocytes for these studies were derived in two independent fashions. In one, brown fat precursor cells from the interscapular fat pad of the fat-specific insulin receptor knockout (FIRKO) mice were directly immortalized from neonatal mice using a retrovirus expressing the SV40 T-antigen. In these FIRKO cells, cre recombinase is driven by the adiposespecific promoter aP2. While this promoter is mainly expressed in mature adipocytes, low levels of aP2 mRNA expression have been observed in pre-adipocytes before the onset of adipogenesis (19). A low level of cre recombinase expression driven by this promoter was sufficient to recombine the insulin receptor alleles in these pre-adipocytes. Alternatively normal differentiating cells in which both alleles of exon 4 of the insulin receptor were flanked by loxP sites (IR lox (20)) were immortalized using SV40 T-antigen then infected in vitro with an adenovirus encoding cre recombinase. Individual colonies containing the recombined allele (IRKO) were selected and propagated. Cells derived by both methods behaved similarly, and we have focused on cells derived by in vitro recombination for simplicity. Fig. 1 shows a representative IRKO pre-adipocyte cell line in which exon 4 of IR has been deleted. A 480-bp PCR product was present in IR lox cells, whereas a smaller 220-bp product was observed in IRKO cells, confirming that recombination had occurred. To restore IR expression, individual splice isoforms of the human IR were introduced into IRKO cells by retroviral infection. While IR mRNA was still present in IRKO cells, no mature protein was detected in these cells, and reconstituted levels of IR are higher than that seen in control cells (Fig. 1,  bottom panel).
To ascertain the role of the insulin receptor in differentia-tion, control IR lox and IRKO pre-adipocytes were differentiated into brown adipocytes using insulin, triiodothyronine, isobutylmethylxanthine, dexamethasone, and indomethacin as described under "Experimental Procedures." At the end of the differentiation protocol (day 6), cells from both genotypes were stained using the fat-specific dye oil red O ( Fig. 2A). In control IR lox cells, 90 -100% of cells accumulated multiocular fat droplets. This was coordinated with an induction of various adipogenic mRNAs such as PPAR␥ and Glut4. In control adipocytes, splicing of exon 11 of the insulin receptor mRNA was also regulated in a differentiation-dependent manner (Fig. 2B). In pre-adipocytes (day 0) the level of insulin receptor mRNA was low and predominantly the A isoform. During differentiation there was an increase in the total levels of IR mRNA with a coordinated increase in the B isoform of IR such that the ratio of IR B :IR A isoform was ϳ60%:40% by day 6. As has been shown previously (1), during differentiation there was an increase in mRNA levels of the adipogenic markers PPAR␥ and Glut4 in control cells.
Insulin Receptor-deficient Pre-adipocytes Fail to Differentiate-In comparison to controls, no fat droplets were observed in either type of IRKO pre-adipocytes after the differentiation protocol ( Fig. 2A). This failure to accumulate lipid was observed in ten independent IR-deficient cell lines (data not shown) and was accompanied by an inability to induce the differentiation markers PPAR␥ and Glut4 (Fig. 2B). Although IR mRNA was still present in IRKO cells, because deletion of exon 4 by cre recombination creates a frameshift mutation predicting a premature stop codon, no mature IR protein was present in these cells (see above). In parallel with the lack of differentiation, IRKO cells also showed no change in the splicing of IR mRNA during time in culture. Similarly, the increased expression of PPAR␥ and Glut4 was not observed in IRKO cells.
Reconstitution of IR in IRKO Cells Fails to Restore Differentiation-To determine whether either of the IR splice isoforms has a unique role in adipogenesis, IRKO cells were infected with a retrovirus encoding for either human IR A or IR B . Surprisingly, IRKO cells reconstituted with either IR isoform failed to differentiate as measured by oil red O staining (Fig.  2C) or accumulation of the adipogenic markers PPAR␥ and Glut4 (Fig. 2D). These results were verified in three independent IRKO cell lines. Coexpression of both splice isoforms also failed to restore differentiation in IRKO cells (data not shown).
To ensure that the re-expressed IR isoforms were functional proteins, control, IRKO, and IR-reconstituted KO pre-adipocytes were stimulated with insulin, and phosphorylation of downstream targets was analyzed (Fig. 3). IR protein expression was very low in control IR lox pre-adipocytes, but was undetectable in IRKO cells. IR-reconstituted KO cells showed dramatically increased IR protein levels and a markedly increased level of IR tyrosine phosphorylation following insulin stimulation. Insulin was also able to stimulate phosphorylation of AKT in control and IR-reconstituted KO cells, but not in IRKO cells. Therefore, IR is required for insulin-stimulated AKT phosphorylation. Surprisingly, despite the absence of insulin receptors, insulin was able to induce phosphorylation of MAPK in IRKO cells, although phosphorylation was slightly enhanced in cells re-expressing IR. Because IRKO cells lack insulin receptors, this increase in MAPK phosphorylation in IRKO pre-adipocytes most likely occurs via insulin action through the IGF-I receptor. This result was not due to receptor compensation, because IGF-I receptor levels were not different between the cell lines (data not shown). Both splice isoforms of IR (hIR A and hIR B ) were equivalent in their ability to activate downstream signaling molecules.
Phosphorylation of IR, IRS-1, and IRS-2 during Adipogenesis Is Impaired in IRKO Cells-A number of proteins implicated in insulin signaling are modified during differentiation.
To determine the contribution of IR to these events, tyrosine phosphorylations of IR, IRS-1, and IRS-2 were examined in control IR lox -, IRKO-, and IR-reconstituted KO cells during the differentiation protocol. Because hIR A and hIR B signaled similarly in these cells (see Fig. 3), only results for cells expressing hIR A are shown for simplicity. Similar to mRNA levels, protein expression of IR was low in control pre-adipocytes but was increased in mature adipocytes (Fig. 4). IRKO cells failed to express mature IR protein. IR-reconstituted KO cells showed an abundant increase in expression of IR in pre-adipocytes (day 0). Although these levels were maintained throughout differentiation, the absolute level of IR was only ϳ3-fold higher at day 6 compared with control adipocytes.
Following induction of differentiation, tyrosine phosphorylation of IR was elevated at day 2 in control cells. At days 4 -6 of differentiation, levels of IR tyrosine phosphorylation declined even though total IR protein levels were elevated. As expected, IRKO cells showed no increase in IR tyrosine phosphorylation. In IR-reconstituted KO cells tyrosine phosphorylation of IR was faintly detected at day 0 but was dramatically increased at day 2 after induction of differentiation. Following the pattern of IR tyrosine phosphorylation, both IRS-1 and IRS-2 were tyrosine-phosphorylated at day 2 of differentiation in control cells. This phosphorylation was absent in IRKO cells and dramatically reduced in IR-reconstituted KO cells despite the high levels of IR tyrosine kinase activity.
Reconstitution of IR Expression in IRKO Cells Restores PI3K Activation and MAPK Dephosphorylation-Inhibition of PI3K activity by pharmaceutical inhibitors or expression of dominant-negative subunits disrupts adipocyte differentiation (1,4), indicating PI3K activity is essential for the adipogenic program. To determine whether PI3K activation was directly coupled to differentiation, PI3K activity in phosphotyrosine immunoprecipitates was analyzed in the three cell lines. Similar to the pattern of IRS-1 and IRS-2 tyrosine phosphorylation, PI3K activity was induced ϳ4-fold at day 2 and then returned to basal levels at days 4 and 6 in control cells (Fig. 5A). There was no induction of PI3K activity in IRKO cells. In IR-reconstituted FIG. 1. Insulin receptor expression is absent in IRKO pre-adipocytes. IR lox , IRKO, and IRKO pre-adipocytes reconstituted with IR splice variants (hIRA or hIRB) were grown to confluence and mRNA or protein was harvested. Semi-quantitative RT-PCR was performed for mouse IR exon 4 and human IR exon 11 as described under "Experimental Procedures." Immunoblots were analyzed for total IR expression.
KO cells basal PI3K activity was higher, and there was a further induction at day 2 of differentiation. Thus, although the insulin receptor appears to be required for the PI3K activation observed during differentiation, activation of PI3K, in the presence of an inappropriate level of IR, is not sufficient to drive the adipogenic program.
To investigate signaling further downstream of IR, IRS, and PI3K, phosphorylation of AKT and MAPK was examined. In control cells, there was a transient increase in AKT phospho-rylation at day 2 of differentiation that paralleled PI3K activation (Fig. 5B, upper panel). Phosphorylation of AKT was absent in IRKO cells. IR-reconstituted KO cells showed an increase in AKT phosphorylation at day 0, which declined during differentiation. MAPK activity has been associated with inhibition of differentiation (21,22), and in accordance with this, MAPK phosphorylation was elevated in control pre-adipocytes (day 0) but then decreased during differentiation (Fig. 5B, lower   FIG. 3. Reconstitution of IR restores insulin signaling in IRKO pre-adipocytes. IR lox and IRKO pre-adipocytes, as well as IRKO cells reconstituted with either hIRA or hIRB, were serum-starved for 16 h then stimulated with 10 nM insulin for 10 min. Protein was extracted and analyzed for IR expression or phosphorylation of IR (PY-IR), AKT, and MAPK.

FIG. 4. Differentiation-dependent phosphorylation of IRS-1 and IRS-2 is impaired in IRKO cells.
IR lox -, IRKO-, and IR-reconstituted IRKO cells were induced to differentiate as described under "Experimental Procedures." At the indicated days, protein was harvested and analyzed for IR expression. Additionally, lysates were subjected to immunoprecipitation with IR, IRS-1, or IRS-2 antibodies followed by immunoblotting with a phosphotyrosine antibody (PY) .   FIG. 2. IRKO pre-adipocytes fail to differentiate. Brown adipose precursor cells isolated from control (IR lox ), and IRKO mice were grown to confluence and induced to differentiate as described under "Experimental Procedures." A, at day 6 of differentiation, cells were fixed and stained with oil red O. B, at the indicated days, mRNA was harvested and RT-PCR was performed for the indicated markers. IRKO cells overexpressing IR splice variants (hIRA or hIRB) were induced to differentiate and cells were stained with oil red O (C) or mRNA was collected for RT-PCR (D) as indicated above. panel). By contrast, MAPK phosphorylation in IRKO cells was high and remained elevated throughout the culture period, correlating with the inability of these cells to differentiate. Phosphorylation of MAPK in IR-reconstituted KO cells resembled the pattern observed in control cells, i.e. phosphorylation was elevated in pre-adipocytes and declined at the later days of the differentiation program. However, as noted above, these cells failed to differentiate.
PPAR␥ Expression and Activation Partially Restores Differentiation in IRKO Cells-The finding that IR-reconstituted KO cells were similar to control cells in their ability to alter differentiation-dependent signaling such as activation of PI3K and down-regulation of MAPK phosphorylation, but failed to differentiate as measured by oil red O staining or accumulation of adipogenic markers such as PPAR␥ and Glut4, suggested that the block in adipogenesis in these cells may occur at a later stage in differentiation.
To test this hypothesis, the adipogenic transcription factor PPAR␥ was introduced into IRKO cells via retroviral-mediated gene transfer. To enhance PPAR␥ activation, some experiments were also performed with the addition of the PPAR␥ agonist troglitazone. Differentiation was measured by the induction of the adipogenic proteins PPAR␥, Glut4, FAS, and C/EBP␣ (p42 and p30 products). As described above, fully differentiated control cells showed high expression of all the measured adipogenic markers (Fig. 6, lane 1). Addition of PPAR␥ or troglitazone in control cells exhibited no significant changes in protein expression, presumably due to the natural differentiation potential of these cells (Fig. 6, lanes 2-4). In contrast with the results in IRS-1 knockout cells (1), expression of PPAR␥ or its activation by troglitazone only slightly improved the ability of IRKO cells to differentiate as measured by a modest increase in the levels of endogenous PPAR␥, Glut4, and FAS (Fig. 6, compare lanes 6 and 7 to lane 5). The combination of PPAR␥ expression with activation by troglitazone further enhanced differentiation in IRKO cells (Fig. 6, compare  lane 8 with lanes 6 and 7). However, under these conditions FIG. 6. PPAR␥ expression improves differentiation in IRKO cells. IR lox (lanes 1-4) and IRKO pre-adipocytes overexpressing PPAR␥ (lanes 5-8) were prepared as described under "Experimental Procedures." Additionally, hIRA or hIRB was expressed in IRKO cells overexpressing PPAR␥ (lanes 9 -12). All cell lines were induced to differentiate in the absence (odd lanes) or presence (even lanes) of 25 M troglitazone. At day 6 of differentiation, protein lysates were prepared and immunoblots were performed using antibodies specific for the indicated adipogenic markers. Arrows indicate the 42-and 30-kDa forms of C/EBP␣.  A and B) and IR lox cells overexpressing IR (C and D) were differentiated and lysates were harvested at the indicated days. PI3K activity was measured in phosphotyrosine immunoprecipitates as described under "Experimental Procedures" (A and C). These panels show the mean and standard error for at least three separate experiments. Protein lysates were directly immunoblotted with antibodies specific for the phosphorylated (P) forms of AKT or MAPK (B and D). differentiation of IRKO cells was still markedly reduced compared with control cells. To determine whether PPAR␥ expression and IR signaling could coordinately enhance differentiation, human insulin receptor (hIR A , lanes 9 and 10; or hIR B , lanes 11 and 12) was reintroduced into IRKO cells expressing PPAR␥. Expression of either IR splice isoform was able to prevent adipogenesis induced by moderate overexpression of PPAR␥ in IRKO cells in the presence or absence of troglitazone (Fig. 6, compare lanes 10 and 12 to lane 8). These data support the hypothesis that high levels of IR may be inhibitory to adipogenesis.
Overexpression of IR in Control Pre-adipocytes Inhibits Differentiation-To determine whether high IR expression disrupts adipogenesis, the human insulin receptor or an empty vector (ϩpBABE) was introduced into control IR lox cells (Fig.  7). Similar to uninfected cells, vector-infected cells accumulated multiocular fat droplets as measured by oil red O staining (Fig. 7A). However, fat accumulation was not detected in cells overexpressing either isoform of IR. This defect in differentiation was mirrored by the failure of these cells to induce expression of the adipogenic markers PPAR␥, Glut4, FAS, C/EBP ␣, and UCP-1, whereas control and vector-infected cells showed a robust induction of these proteins by day 6 of differentiation (Fig. 7B).
Because IRKO cells reconstituted with IR failed to differentiate but exhibited some differentiation-dependent signaling (see above), PI3K activation was examined in control cells that overexpress IR. As shown above, PI3K activity associated with phosphotyrosine increased ϳ4-fold at day 2 in control cells (Fig.  5C). Although basal PI3K activity was higher in cells overexpressing IR, a similar maximal level of activity was observed at day 2 of differentiation. The pattern of PI3K activation was paralleled by phosphorylation of AKT, a downstream target (Fig. 5D, upper panel). In cells overexpressing IR, AKT phosphorylation was initially high at the beginning of the differentiation program (days 0 and 2) but decreased thereafter. MAPK phosphorylation was also elevated in pre-adipocytes (Fig. 5D,  lower panel). This phosphorylation was absent during the middle phases of the differentiation protocol but was increased again by day 6. Control cells that overexpress IR fail to differentiate but maintain insulin signaling through PI3K, AKT, and MAPK. DISCUSSION In the current study we have utilized SV40T-antigen-immortalized brown pre-adipocytes isolated from control and IRKO mice to examine the role of the insulin receptor in adipocyte differentiation. Similar to the pattern in models of white adipogenesis, such as 3T3-L1 cells, the expression level of the insulin receptor is very low in brown pre-adipocytes and is dramatically induced during adipocyte maturation. In addition to an increase in total levels, insulin receptor mRNA is also alternatively spliced during brown fat differentiation such that there is an isotype switch from the A isoform (Ϫexon 11) in pre-adipocytes to the B isoform (ϩexon 11) in mature adipocytes. A similar isotype switch from IR A to IR B has been observed when 3T3-L1 cells are differentiated in the presence of dexamethasone. This pattern of alternative splicing is also observed in differentiating hepatocytes. Thus, HepG2 cells cultured at low density exhibit a fetal phenotype and express predominantly the A isoform of the insulin receptor, whereas HepG2 cells cultured to confluence in the presence of dexamethasone display an adult phenotype and express the B isoform of the insulin receptor (23). Likewise, in rat liver in vivo, IR B expression increases from 60% in fetal liver to 95% in adult liver (12). In addition to dexamethasone, insulin itself may regulate splicing of the insulin receptor gene. For example, FAO hepatoma cells treated with insulin show a time-dependent decrease in IR A with a relative increase in IR B (24). Therefore, both differentiation itself and the hormonal milieu used during differentiation of adipocytes may regulate the alternative splicing of the insulin receptor.
Expression of either IR A or IR B restores insulin signaling in IR-deficient pre-adipocytes. In contrast to observations in pancreatic ␤ cells (25), both receptors are equally capable of initiating phosphorylation of AKT, MAPK, or p70S6K (data not shown) in response to insulin. However, we were unable to examine the individual contribution of each isoform in mature adipocytes where expression of both receptors occurs naturally.
Using standard differentiation protocols, we observed that insulin receptor-deficient brown pre-adipocytes failed to differentiate as measured by the lack of lipid accumulation or the expression of the adipogenic markers PPAR␥, FAS, and Glut4. The requirement for the insulin receptor in adipocyte differentiation in vitro is supported by gene dosage studies performed in 3T3-L1 cells. 3T3-L1 fibroblasts with a somatic inactivation of one allele of the insulin receptor showed a 50 -70% reduction in IR levels and exhibited impaired differentiation with only 30 -50% of cells displaying morphological evidence of adipogenesis (26). Although insulin receptor levels are low in pre-adipocytes, this minimal expression is required to initiate the FIG. 7. Overexpression of IR inhibits differentiation of control cells. IR lox cells expressing pBABE vector, hIRA, or hIRB were induced to differentiate as indicated under "Experimental Procedures." At day 6 of differentiation, cells ere stained with oil red O (A). Protein was extracted at 0 or day 6, and immunoblots were performed using antibodies directed against the indicated adipogenic markers (B). Arrows indicate the 42-and 30-kDa forms of C/EBP␣. adipogenic program. The failure of IRKO cells to differentiate also resembles defects seen in IRS-1-deficient cells, reiterating the importance of insulin signaling in brown fat adipogenesis.
Even though control or IRKO cells overexpressing IR fail to terminally differentiate, some insulin receptor-and differentiation-dependent signaling is preserved in these cells. For example, PI3K activity peaks at day 2 in control cells. Although PI3K activity is absent in IRKO cells, IR reconstitution restores a similar pattern of PI3K activity after the hormonal induction of adipogenesis, suggesting that the insulin receptor is required for this event. PI3K activity is also maintained in control pre-adipocytes that overexpress IR. Previous data has suggested that PI3K activity during the induction phase is essential for adipogenesis in both white and brown fat cell models. Our data would suggest that, although required, PI3K activity is not sufficient to drive adipocyte differentiation.
Previous studies have shown that high MAPK phosphorylation correlates with the inability to differentiate, and overexpression of MAPK or its upstream activator MEK1 inhibits 3T3-L1 differentiation, possibly by reducing PPAR␥ activity (22,27). In control pre-adipocytes, MAPK phosphorylation is high then declines with differentiation. IRKO cells, which fail to differentiate, also fail to attenuate MAPK phosphorylation. However, IR-reconstituted cells and control cells overexpressing IR exhibit a decrease in MAPK phosphorylation during the differentiation program similar to control cells but show similar defects in adipogenesis. These data suggest that the classic insulin-dependent signaling induced during the early phases of differentiation can be uncoupled from the production of the mature adipocyte phenotype.
Unlike IRS-1 reconstitution in IRS-1 KO cells, reconstitution of the insulin receptor in IRKO cells fails to restore differentiation (1). IR-reconstituted KO pre-adipocytes do not accumulate lipid or induce adipogenic proteins when induced to differentiate. This appears to be due to the high level of insulin receptor in the pre-adipocyte stage. This hypothesis is validated by the fact that control pre-adipocytes overexpressing IR also fail to differentiate. Thus it appears that the elevated levels of IR re-introduced in IRKO or control cells upset the balance between IR and IGF-IR that may be required for normal differentiation.
Insulin and IGF-I receptors are highly homologous receptors, each composed of two ␣␤-heterodimers linked by disulfide bonds (28,29). Individual ␣␤-heterodimer subunits from IR and IGF-IR can also combine to form hybrid receptors (30). Previous studies have shown that both IGF-I and insulin are able to induce adipogenic and thermogenic genes in brown adipocytes (31). Fetal brown adipocytes possess a number of high affinity IGF-I binding sites and express both IGF-I and IGF-IR mRNA during development (32). As noted above, while insulin receptor expression increases during differentiation, overexpression of IR inhibits differentiation. In brown preadipocytes and 3T3-L1 fibroblasts, IR levels are low and IGF-IR is the primary signaling receptor (33). 2 With the induction of differentiation, IR expression increases dramatically such that it is the predominant receptor in mature adipocytes. Hybrid receptor formation rises in parallel with the increase in insulin receptor levels. High IR expression would increase the number of binding sites for insulin through classic insulin receptors but place more of the IGF-I receptors in hybrid complexes. We hypothesize that these hybrids are like the insulin receptor in that they are unable to support differentiation. Thus, the increase in the insulin signaling pathway indirectly alters the response to the IGF-I receptor, which adversely affects cell growth or differentiation. Indeed, Chinese hamster ovary cells overexpressing the insulin receptor are more sensitive to insulin but become resistant to IGF-I, although the levels of IGF-IR remain unchanged, suggesting that high levels of IR impair IGF-I signaling (34). Taken together, these data suggest that abnormal expression of IR in pre-adipocytes might disrupt crucial IGF-I signaling required for adipogenesis. Therefore, a correct balance of IR and IGF-IR must be tightly controlled at all stages of differentiation. Depletion or inappropriate overexpression of the insulin receptor early in the process impairs brown adipocyte differentiation. Thus, the increase in insulin receptor later in adipogenesis may be important for both insulin regulation of metabolism and the normal differentiation of this cell type.