Calreticulin inhibits commitment to adipocyte differentiation

Obesity is rapidly becoming a worldwide epidemic, leading to numerous medical conditions such as type 2 diabetes and cardiovascular disease. The development of obesity is marked by both an increase in adipocyte size and number as a consequence of differentiation of the stem cells or preadipocytes into mature adipocytes (for review see Gesta et al., 2007). Spontaneous differentiation of embryonic stem (ES) cells into adipocytes in culture is a rare event; thus, for ES cells to commit toward the adipocyte lineage, it is necessary to pretreat cells with retinoic acid (RA), a ligand of the nuclear receptor retinoid X receptor (RXR), during the first phase of differentiation (Dani et al., 1997). RXR in response to RA heterodimerizes with peroxisome proliferator–activated receptor (PPAR), resulting in activation of genes involved in terminal differentiation and lipid metabolism (Gregoire et al., 1998). PPARγ2, CCAAT enhancer–binding protein α (C/EBPα), and aP2 are crucial for the development of adipose cells (Rosen et al., 2000). The process of adipogenesis may also be influenced by different extrinsic factor and intracellular signaling pathways (Rosen and MacDougald, 2006), yet very little is known about the role of Ca²⁺ homeostasis in adipogenesis. Increased Ca²⁺ levels lead to inhibition of adipocyte differentiation, which is accompanied by the decrease or, in some cases, the loss of PPARγ2, C/EBPα, and aP2 expression (Jensen et al., 2004; Serlachius and Andersson, 2004; Zhang et al., 2007). Thapsigargin, an inhibitor of the sarco/ER Ca²⁺-ATPase, inhibits adipocyte differentiation in 3T3-L1 preadipocytes (Ntambi and Takova, 1996; Shi et al., 2000), suggesting a role for ER stores and ER Ca²⁺-binding proteins in the modulation of adipogenesis.

Calreticulin is a major Ca²⁺-buffering protein in the lumen of the ER, which also acts as a molecular chaperone and modulator of gene expression (Bedard et al., 2005). The two major functions of calreticulin, chaperoning and Ca²⁺ buffering, are confined to specific protein domains. The N+P domain of calreticulin forms a folding module, and the acidic C-terminal C domain binds and buffers Ca²⁺ with high capacity (Baksh and Michalak, 1991; Nakamura et al., 2001). Calreticulin deficiency is embryonic lethal (Mesaeli et al., 1999), and calreticulin-deficient cells have a reduced capacity for Ca²⁺ storage within the ER and impaired agonist-stimulated Ca²⁺ release from ER stores, whereas cells overexpressing calreticulin have higher levels of lumenal ER Ca²⁺ and a larger pool of releasable Ca²⁺ (Nakamura et al., 2001). Calreticulin's impact on protein folding and Ca²⁺ homeostasis have been implicated in several signaling pathways (Bedard et al., 2005), thus affecting gene expression and, consequently, the behavior of individual cells and cell communities. In this study, we show that calreticulin may act as a Ca²⁺-dependent molecular switch that negatively regulates commitment to adipocyte differentiation by preventing the expression and transcriptional activation of critical proadipogenic transcription factors.

Fig. 1 A shows that the adipogenic potential of ES cells, as measured by oil red O staining at differentiation day 20 (D20), was ∼30-fold higher in the absence of calreticulin. The number of adipocyte colonies in calreticulin-deficient embryoid bodies (EBs) was approximately ninefold higher than that in the calreticulin-containing EBs at D20 (Fig. 1 B). This was irrespective of whether the calreticulin gene was removed by homologous recombination (G45crt^−/− cells) or by Cre recombinase–mediated excision (L7crt^−/− cells; Fig. S1 A). Adipogenesis in ES cells was RA and insulin dependent and required an initial 3-d exposure to RA followed by insulin treatment (unpublished data). Microscopically, by Nile red staining, lipid droplets in the wild-type (WT) adipocytes were indistinguishable from those in calreticulin-deficient cells (Fig. S1 B), indicating that the gross morphology of lipid stores was not affected in the absence of calreticulin. Nile red is a red-emitting fluorescent lysochrome (Greenspan et al., 1985) that can be used as a fluorescent alternative to oil red O lipid visualization (Fowler and Greenspan, 1985). Fig. S2 shows the nucleotide sequence of the calreticulin promoter.

Next, we examined the expression of both early (C/EBPα and PPARγ2) and late (aP2) adipogenic markers. The abundance of PPARγ2 and C/EBPα adipogenesis markers was markedly increased at D20 in G45crt^−/− and L7crt^−/− cells compared with WT (Fig. 1, C and D). The aP2 mRNA levels were undetectable in extracts from the WT and L7 cells, whereas they were high at D20 in extracts from calreticulin-deficient cells (Fig. 1, C and D). It should be stressed that in Fig. 1 C as well as in subsequent figures, adipogenic markers in calreticulin-containing cells are barely discernible. This is caused by a large disparity in the markers' abundance between calreticulin-deficient and calreticulin-containing cells, which makes it difficult to visualize all of them when equally loaded. Modulation of the expression of calreticulin also impacted adipogenesis of 3T3-L1 preadipocytes, a commonly used model for adipogenesis (Otto and Lane, 2005). Increased expression of calreticulin in 3T3-L1 preadipocytes inhibited their adipogenesis, as indicated by oil red O staining (Fig. 1 E). In agreement with ES cell results, molecular markers of adipogenesis (lipoprotein lipase, aP2, PPARγ2, C/EBPα, and C/EBPβ) were all down-regulated in 3T3-L1 cells overexpressing calreticulin (Fig. 1 F).

Upon induction of adipogenesis with RA, the abundance of calreticulin increased dramatically in the WT ES cells (Fig. 2 A), whereas the abundance of PPARγ2 and C/EBPα remained persistently low (Fig. 2, B and C). In contrast, in calreticulin-deficient (G45crt^−/−) cells after induction of adipogenesis with RA, PPARγ2 and C/EBPα levels steadily increased over 20 d of differentiation (Fig. 2, B and C). Similar to the WT ES cells, the 3T3-L1 preadipocytes and 3T3-L1 cells overexpressing calreticulin showed an increase in calreticulin abundance and reduced adipogenic potential upon RA treatment (Fig. 2, D and E). We concluded that the increased expression of calreticulin plays a negative regulatory role during adipogenesis.

One important function of calreticulin is modulation of Ca²⁺ homeostasis (Nakamura et al., 2001; Bedard et al., 2005). Thus, we next examined whether calreticulin-dependent changes in Ca²⁺ homeostasis might be responsible for modulation of adipogenesis. We used BAPTA-AM, a membrane-permeable Ca²⁺ chelator (Tsien, 1980), to lower cytosolic Ca²⁺ concentration ([Ca²⁺]_Cyto) as well as ionomycin, a Ca²⁺ ionophore (Bergling et al., 1998), to increase [Ca²⁺]_Cyto. Treatment of cells with BAPTA from 20 min to 2 h produced exactly the same results (unpublished data), which implies that BAPTA was not secreted from the cells. BAPTA-AM–treated WT and L7 cells showed a dramatic increase in oil red O staining, indicating that chelation of cytoplasmic Ca²⁺ promoted adipogenesis in these cells (Fig. 3, A and B). There was no significant increase in oil red O staining in BAPTA-AM–treated calreticulin-deficient ES cells (G45crt^−/− or L7crt^−/−; Fig. 3, A and B). In contrast, after treatment with ionomycin, there was no detectable oil red O–positive adipocytes derived from WT and L7 cells, whereas differentiation of both calreticulin-deficient cell lines (G45crt^−/− and L7crt^−/−) produced an extremely low number of oil red O–positive cells (Fig. 3, A and B).

In agreement with oil red O staining results, BAPTA-AM treatment significantly increased the abundance of all three adipogenic markers in adipocytes derived from WT and L7 cells containing calreticulin but did not have a significant effect on adipocytes derived from calreticulin-deficient (G45crt^−/− and L7crt^−/−) ES cells (Fig. 3, C and D). In the presence of ionomycin, the levels of PPARγ2, C/EBPα, and aP2 remained nearly undetectable in extracts from WT and L7 cells (Fig. 3, C and E). However, ionomycin significantly decreased the abundance of PPARγ2, C/EBPα, and aP2 levels in extracts from calreticulin-deficient cell lines (G45crt^−/− and L7crt^−/−) at D20 (Fig. 3, C and E).

Calreticulin has two structural and functional domains (Nakamura et al., 2001); one responsible for chaperoning, and another for Ca²⁺ buffering (Fig. 4 A). To determine which of calreticulin's functions (domains) may be involved in the modulation of adipogenesis, two ES cell lines expressing single functional modules of calreticulin in calreticulin-deficient ES cells (G45crt^−/−) were created and tested for adipogenic potential. Because the C domain of calreticulin cannot be stably maintained, it was fused to the P domain of calreticulin (Nakamura et al., 2001). Expression of the domains was tracked using anti-HA antibodies (Fig. 4 A).

Fig. 4 B shows that reexpression of the P+C domain in calreticulin-deficient cells inhibited adipogenesis (Fig. 4, B and D). Overexpression of the P+C domain in 3T3-L1 preadipocytes also resulted in reduced adipogenesis (Fig. 4 C). Treatment of the P+C domain–expressing crt^−/− ES cells with BAPTA-AM restored their adipogenic potential (Fig. 4, B and D). In contrast, expression of the N+P domain in crt^−/− ES cells had no significant effect on their ability to differentiate into adipocytes, and they maintained a phenotype identical to calreticulin-deficient cells (Fig. 4 B). This was further supported by BATPA-AM and ionomycin experiments (Fig. 4), indicating that the chaperone function of calreticulin was not involved in the modulation of adipogenesis. Therefore, crt^−/− cells expressing the N+P domain of calreticulin displayed features resembling those of crt^−/− ES cells, whereas crt^−/− cells expressing the Ca²⁺-buffering region of the protein (P+C domain) had properties characteristic of those of WT cells, indicating that calreticulin modulates adipogenesis via its Ca²⁺ homeostatic function.

To assess whether the Ca²⁺ content of intracellular Ca²⁺ stores is modified in crt^−/− cells and calreticulin-deficient cells expressing Ca²⁺ handling the P+C domain of calreticulin, we performed equilibrium loading experiments with ⁴⁵Ca²⁺. Cells were cultured for 54 h in the regular culture medium containing 10 mCi/ml ⁴⁵Ca²⁺. The total cellular Ca²⁺ content was then calculated based on the cell-associated radioactivity and on the specific activity of Ca²⁺ in the culture medium. We used thapsigargin, an inhibitor of sarco/ER Ca²⁺-ATPase, to measure the amount of Ca²⁺ associated with ER-exchangeable intracellular Ca²⁺ stores in either ES (Fig. 4 F) or 3T3-L1 (Fig. 4 H) cells. To assess the residual amount of Ca²⁺ contained within thapsigargin-insensitive lumenal Ca²⁺ stores, we used the Ca²⁺ ionophore ionomycin (Fig. 4, F and H). Measurement of ionomycin- and thapsigargin-induced Ca²⁺ discharge from the ER indicated that the WT ES cells and calreticulin-deficient ES cells expressing the P+C domain had higher [Ca²⁺]_ER and [Ca²⁺]_Tot, respectively, compared with the crt^−/− ES cells (Fig. 4 F). To provide yet another measure of ER-releasable Ca²⁺, we measured thapsigargin-induced Ca²⁺ discharge from the ER versus cytosolic [Ca²⁺] using the Ca²⁺-sensitive fluorescent dye fura-2-AM under conditions preventing dye sequestration into the ER (Mery et al., 1996). Fig. 4 G shows that calreticulin-containing WT ES cells and calreticulin-deficient ES cells expressing the P+C domain had higher [Ca²⁺]_ER and [Ca²⁺]_Cyto, respectively, in comparison to the crt^−/− ES cells. Similar to the ES cells, overexpression of calreticulin or the P+C domains in the 3T3-L1 preadipocytes also resulted in increased [Ca²⁺]_ER and [Ca²⁺]_Tot compared with control 3T3-L1 cells (Fig. 4 H). Finally, as ionomycin may be inactive in releasing Ca²⁺ from acidic intracellular compartments, we added the sodium proton ionophore monensin. Monensin-induced Ca²⁺ release was very small (unpublished data), suggesting that WT, crt^−/− cells, and cells expressing the N+P domain did not contain substantial quantities of Ca²⁺ stored within acidic compartments. Thus, we conclude that the effects of calreticulin on adipogenesis may be mediated by calreticulin-dependent changes in intracellular Ca²⁺.

The results so far suggested that calreticulin plays a modulatory role during adipogenesis. We next wanted to determine whether there was a functional relationship between calreticulin and the PPARγ transcriptional complex. Upon RA-dependent induction of adipogenesis, RXR and PPARγ form a transcriptionally active complex. The calreticulin promoter contains two PPARγ-binding sites termed peroxisome proliferator responsive elements (PPREs); one is found at −1,944 bp (designated PPRE1), and the other is found at −590 bp (designated PPRE2; Fig. 5 A), suggesting that the PPARγ transcription factor may regulate the calreticulin gene. To test this, NIH3T3 fibroblasts were cotransfected with PPARγ expression vector and RXRα expression vector (pLC0, pLC1, pLC2, pLC0mt1, and pLC0mt2) or luciferase reporter gene vectors (pCL2 and pCL3; Fig. 5 A) under control of the calreticulin promoter. pSVβ-galactosidase was used as an internal control. In the pLC0 vector, luciferase was controlled by 2.1 kb of the calreticulin promoter containing both PPRE sites (Fig. 5 A and Fig. S2). Fig. 5 B shows that PPARγ induced luciferase activity in NIH3T3 cells. Cells transfected with promoterless control plasmids showed no detectable luciferase activity (unpublished data). This finding indicates that PPARγ activates transcription of the calreticulin gene.

To identify the PPARγ-binding site on the calreticulin promoter, we performed reporter gene assay with calreticulin promoter deletions and mutations (Fig. 5, B and C). Deletion of PPRE1 and PPRE2 sites (pLC2 vector) completely abolished PPARγ-dependent activation of the calreticulin promoter (Fig. 5 B). Deletion of PPRE site 2 at −1,944 bp (Fig. 5 A) or mutation of PPRE site 1 or 2 (Fig. 5 A) had no effect on PPARγ activation of the promoter (Fig. 5, B and C). Thus, upon induction of adipogenesis with RA, RXR and PPARγ form a complex, which binds PPRE sites 1 and 2 on the calreticulin promoter and transcriptionally activates the calreticulin gene.

To further elucidate a physical interaction between the calreticulin promoter and PPARγ, an electrophoretic mobility shift assay (EMSA) was performed (Fig. 5 C). Synthetic oligodeoxynucleotides corresponding to PPRE sites 1 and 2 were used along with a positive control probe, an ideal PPAR site. Fig. 5 C shows that the PPARγ–RXRα complex bound to PPRE1 and PPRE2. This DNA–protein interaction was only observed in the presence of RXRα, indicating that PPARγ is only functional once complexed with RXRα (Fig. 5 C). The specificity of the PPRE and PPARγ–RXRα binding was confirmed by reduced signal intensity after the addition of 30-fold excess of cold probe (Fig. 5 C). The lesser intensity in PPRE1 and PPRE2 complexes (Fig. 5 C, lanes 4 and 6) when compared with complexes containing the ideal PPRE site (Fig. 5 C, lanes 1, 3, and 10) is likely the result of the different nucleotide sequences of these probes. Specificity of the binding was further confirmed by supershift EMSA and mutation of both PPRE sites (Fig. 5 D). Chromatin immunoprecipitation (ChIP) was also performed to determine whether there was a direct interaction between the calreticulin promoter and PPARγ. ChIP analysis indicated that PPARγ bound to both PPRE sites 1 and 2 on the calreticulin promoter (Fig. 5 E). We concluded that the PPARγ–RXRα complex binds to calreticulin PPRE1 and PPRE2 sites and activates the calreticulin gene.

PPARγ2, along with C/EBPα, directly affect gene transcription in the nucleus (Rosen, 2005; Rosen and MacDougald, 2006). Given that PPARγ2 is both necessary and sufficient for adipogenesis (Rosen et al., 2000), we determined its spatial expression during adipogenesis in our ES cell system (Fig. 6). On D20 of differentiation in the crt^−/− ES cells (which exhibit increased adipogenesis), PPARγ2 was distinctly nuclear (Fig. 6 A). However, in the WT ES cells, PPARγ2 was barely detectable and appeared cytosolic (Fig. 6 B). Given that the WT ES cells show reduced adipogenesis and, thus, only sparse adipocyte colonies, Fig. 6 B represents a region of low adipogenic potential that predominates in WT outgrowths. Localization of C/EBPα was also investigated in the ES cells, and it showed an essentially identical pattern to that observed for PPARγ2 (unpublished data).

Calmodulin–Ca²⁺/calmodulin-dependent protein kinase II (CaMKII) and cAMP response element binding (CREB pathways affect adipogenesis (Wang et al., 1997; Reusch et al., 2000). Western blot analysis with anti-Thr286–phosphorylated CaMKII, anticalmodulin, and anti-Ser133–phosphorylated CREB antibodies showed that their expression was higher in calreticulin-deficient G45crt^−/− and L7crt^−/− cells than in WT and L7 cells at the end of the 20 d of differentiation (Fig. 7, A and B). These results indicate that CaMKII and CREB play a positive role during adipogenesis of crt^−/− cells and that they are up-regulated in the absence of calreticulin.

To elucidate the role of the CaMKII pathway, we used specific inhibitors of CaMKII, KN-62 and KN-93 (Hidaka and Kobayashi, 1994). Inhibition of CaMKII in G45crt^−/− and L7crt^−/− cells with either KN-62 or KN-93 decreased adipogenesis (Fig. 7, C and D). Inhibition of CaMKII also resulted in the decreased expression of PPARγ2, C/EBPα, and aP2 in all cell lines, with the most dramatic effects in calreticulin-deficient cells (Fig. 7, E and F). KN-92, an inactive analogue of KN-93, did not alter oil red O staining (Fig. 7, C and D) or adipogenic marker expression in either crt^−/− or calreticulin-containing ES cells (Fig. 7, E and F). These data suggest that the calmodulin–CaMKII pathway plays an important role during adipogenesis from ES cells.

Calcineurin is a negative regulator of adipogenesis (Neal and Clipstone, 2002; Kennell and MacDougald, 2005) and is known to affect calcineurin activity (Guo et al., 2002; Lynch et al., 2005). On D20, calcineurin activity was significantly higher in the WT ES cells than in the calreticulin-deficient cells (Fig. 7 G). In the 3T3-L1 preadipocytes, inhibition of calcineurin with cyclosporin-A promoted adipogenesis (Fig. 7, H and I), whereas constitutive expression of activated calcineurin decreased adipogenesis (Fig. 7, H and I). These data give support for the role of calcineurin in adipogenesis of both ES cells and 3T3-L1 preadipocytes. The present findings are consistent with earlier studies relating activation of calcineurin to the presence of calreticulin (Lynch and Michalak, 2003; Lynch et al., 2005).

In this study, we show that calreticulin affects adipocyte differentiation from either ES cells or 3T3-L1 preadipocytes. The absence of calreticulin sways ES cell differentiation toward the adipocyte lineage. This choice of fate is reversed to the WT phenotype by overexpression of calreticulin or by the Ca²⁺-buffering domain of calreticulin but not by expression of the chaperoning domain of the protein. These effects of calreticulin are likely mediated through its Ca²⁺ regulatory function. In comparison to WT cells, calreticulin-deficient ES cells have reduced ER Ca²⁺ storage capacity, resulting in reduced ER Ca²⁺ release upon stimulation. Expression of full-length calreticulin or the calreticulin Ca²⁺-buffering domain in ES cells or 3T3-L1 cells increases ER Ca²⁺ content and reduces the commitment of these cells to adipogenesis. In WT cells, chelating cytoplasmic [Ca²⁺] enhances adipogenesis, and, conversely, increasing cytoplasmic [Ca²⁺] inhibits adipogenesis. These Ca²⁺ effects may negatively affect Ca²⁺-dependent transcriptional pathways that are required for adipogenesis. Calreticulin exerts its Ca²⁺ effects within the first 3 d of differentiation, as manipulation of intracellular [Ca²⁺] at later stages of differentiation (i.e., D7 or D9) did not alter the respective adipogenic potential of ES cells (unpublished data). This is consistent with what has been done to elicit calcium-dependent effects during ES cell differentiation by Li et al. (2002), who used 100 nm ionomycin at D4, and Grey et al. (2005), who used 100 nm ionomycin at D6. Interestingly, Li et al. (2002) showed that ionomycin could be used at concentrations of 10–1,000 nM at days 4–7 with the same effects. Moreover, as calreticulin levels were increasing during the first 3 d of differentiation, PPARγ2 and C/EBPα levels were decreasing, implying that the effects of calreticulin fall within the commitment and/or initial stages of adipogenesis. We conclude that calreticulin may act as a Ca²⁺-dependent molecular switch that negatively regulates commitment to adipocyte differentiation by down-regulating the expression and transcriptional activation of critical proadipogenic transcription factors.

An important finding of this study is that calreticulin, at least in part, exerts its effects on adipogenesis via its role as a modulator of Ca²⁺ homeostasis. Ca²⁺ has previously been shown to affect adipogenesis of 3T3-L1 preadipocytes (Shi et al., 2000; Jensen et al., 2004). For example, 3T3-L1 cells exposed to elevated external Ca²⁺ levels accumulated little or no cytoplasmic lipids and showed the diminished expression of PPARγ2, C/EBPα, and aP2 (Jensen et al., 2004). Increasing cytoplasmic Ca²⁺ levels by thapsigargin also inhibited early stages of adipogenesis (Shi et al., 2000), all pointing out the importance of Ca²⁺ on adipogenesis. Indeed, we show here that an increase in intracellular Ca²⁺ concentration leads to a decrease in adipocyte differentiation. Conversely, decreasing intracellular Ca²⁺ concentrations promotes adipogenesis of calreticulin-deficient ES cells. The relationship between calreticulin expression, intracellular Ca²⁺ concentration, and adipogenesis implies that calreticulin exerts its effects on adipogenesis via its Ca²⁺ homeostatic function. However, the formal proof comes from experiments in which functional modules of calreticulin were expressed in calreticulin-deficient ES cells. Upon expression of calreticulin's Ca²⁺-buffering P+C domain, adipogenesis was halted, whereas expression of the N+P domain had no effect on the progression of adipogenesis.

Our findings that calreticulin affects the commitment to adipocyte differentiation are in line with reports that calreticulin-dependent Ca²⁺ signaling also affects several aspects of the differentiation of cardiomyocytes (Li et al., 2002; Grey et al., 2005;Puceat and Jaconi, 2005) and human myeloid cells (Clark et al., 2002). Interestingly, diminished intracellular Ca²⁺ stores attenuate cardiomyogenesis but promote adipogenesis and differentiation of myeloid cells. Combined, these findings imply that calreticulin and ER Ca²⁺ must play important roles in a variety of differentiation pathways. Moreover, our data regarding timing of the [Ca²⁺] manipulations indicate that in adipogenesis, as in cardiomyogenesis (Li et al., 2002), there is a calreticulin-regulated calcium-sensitive step referred to as a checkpoint by Li et al. (2002).

Both CaMKII and calcineurin are Ca²⁺ dependent and have been implicated in affecting adipogenesis (Wang et al., 1997; Neal and Clipstone, 2002; Kennell and MacDougald, 2005). The CaMKII pathway is involved in C/EBPα regulation and, thus, indirectly in PPAR regulation. CaMKII activates CREB, which, in turn, activates C/EBPα (Fajas et al., 2002; Wang et al., 2003; Zhang et al., 2004). CREB plays a role in adipogenesis, and, indeed, expression of dominant-negative CREB blocks adipogenesis (Reusch et al., 2000; Klemm et al., 2001; Reusch and Klemm, 2002; Fox et al., 2006; Vankoningsloo et al., 2006). CaMKII is also involved in the activation of PPARγ (Paez-Pereda et al., 2005; Park et al., 2007), indicating that the CaMKII pathway may be critical during adipogenesis. Once C/EBPα is expressed, PPARγ and C/EBPα regulate each other, thereby promoting adipogenesis (Rosen, 2005). Here, we show that increased levels of threonine-phosphorylated CaMKII in calreticulin-deficient cells correlate with their increased adipogenesis, and inhibition of CaMKII attenuates adipogenesis in these cells. In calreticulin-deficient cells, in addition to increased CaMKII activity, calmodulin and CREB are also up-regulated, thus explaining the elevated adipogenesis in these cells.

Given that [Ca²⁺]_ER is lower in calreticulin-deficient cells, the question arises as to how CaMKII activity could be increased in these cells. In calreticulin-deficient cells, the level of tyrosine phosphorylation, including that of c-Src (Papp et al., 2007), is higher than in WT cells (Fadel et al., 1999, 2001; Szabo et al., 2007). Activated c-Src phosphorylates calmodulin on tyrosine 99, thereby increasing the affinity of calmodulin for CaMKII (Abdel-Ghany et al., 1990; Benaim and Villalobo, 2002). This phosphorylation event is inhibited by high Ca²⁺ concentration (Fukami et al., 1986). Tyrosine-phosphorylated calmodulin effectively activates CaMKII (Corti et al., 1999). In addition, after initial activation, CaMKII becomes autophosphorylated, remaining active even though its activators are removed (Meyer et al., 1992). Therefore, CaMKII may remain active and promote adipogenesis in calreticulin-deficient cells even under reduced intracellular Ca²⁺ concentration.

We have found that endogenous calcineurin activity was significantly higher in calreticulin-expressing cells that have reduced adipogenic potential compared with the calreticulin-deficient cells that are highly adipogenic. This is in agreement with previous studies on calcineurin-dependent inhibition of adipogenesis (Neal and Clipstone, 2002; Kennell and MacDougald, 2005). Thus, we propose that although the calcineurin pathway inhibits adipogenesis, the Ca²⁺-independent CaMKII pathway might be responsible for the promotion of adipogenesis in the absence of calreticulin observed here.

PPARγ2 and C/EBPα are crucial transcription factors during adipogenesis (Rosen, 2005); however, although PPARγ2 is necessary for adipogenesis to take place (Rosen et al., 2000), C/EBP has a more accessory role during this process (Rosen et al., 2002). We showed here that PPARγ2 and C/EBPα are down-regulated in cells overexpressing calreticulin. Therefore, calreticulin may act as a transcriptional regulator of PPAR, whereby it inhibits the binding of PPAR/RXR to PPRE in the promoter region of the target genes, thus inhibiting transcription activation by peroxisome proliferators and by fatty acids (Burns et al., 1994; Dedhar et al., 1994; Winrow et al., 1995). Here, we show that the initial process that precedes the effects of calreticulin on the activity of the PPARγ–RXR complex involves PPARγ transcriptional activation of the calreticulin gene. PPARγ is a potent transcriptional activator of the calreticulin gene as a result of its direct interaction with the calreticulin promoter. Thus, PPARγ can up-regulate calreticulin, and, conversely, calreticulin can inhibit its activity. We also found an inverse relationship between calreticulin and PPARγ2 expression upon induction of adipogenesis with RA. RA induces the expression of calreticulin but reduces the expression of PPARγ2. This supports the notion of a hierarchical process in which transcriptional activation of the calreticulin gene by PPARγ is the early event followed by calreticulin modulation of PPARγ transcriptional activity. These data provide evidence for a previously unrecognized negative feedback loop whereby PPARγ regulates the expression of calreticulin, and calreticulin modulates PPARγ2 activity and expression. This is likely one important mechanism whereby calreticulin and ER Ca²⁺ homeostasis regulate adipogenesis.

In conclusion, for the first time, we show an essential role for organellar Ca²⁺ in adipogenic regulation. The presence of calreticulin inhibits adipogenesis through its effects on Ca²⁺ homeostasis, causing the indirect regulation of crucial pathways affecting adipogenesis, such as the calcineurin and CaMKII pathways. Although the CaMKII pathway is important for adipogenic differentiation, the calcineurin pathway may be involved in other choices of fate of ES cells. Finally, we show that the expression of calreticulin is tightly regulated during adipogenesis, and any modification of this expression influences adipogenesis such that aberrant calreticulin expression may lead to a variety of cellular pathologies, including obesity and diabetes.

G45 crt^−/− ES cells and the WT ES cell line (CGR8) were derived from J1 129/Sv mice. To generate the L7 ES cell line, J1 129/Sv ES cells were electroporated with a targeting vector containing the calreticulin gene with the loxP site inserted in introns 1 and 2 (Fig. S1 A) and maintained on mitomycin C–treated G-418–resistant mouse embryonic fibroblast feeder cells (Mesaeli et al., 1999). Recombinant clones were selected with 0.2 mg/ml G418 and 2 mM gancyclovir. Several hundred colonies were picked after 10 d in selection medium and expanded. The L7 clone was selected based on Southern blot analysis and Western blot analysis with anticalreticulin antibodies (Mesaeli et al., 1999). Cre recombinase with additional mutant estrogen receptor ligand-binding domains (Cre-ER and MerCreMer) was used to excise exons 2–4 from the calreticulin gene (Fig. S1 A). In brief, L7 ES cells were stably transfected with Cre recombinase mutant expression vector pANMerCreMER (a gift from J.D. Molkentin, University of Cincinnati, Cincinnati, OH) to generate the L7-Cre cell line. To generate L7crt^−/−, 10 μM 4-hydroxytamoxifen was added to promote nuclear translocation and consequently activation of Cre recombinase. L7crt^−/− contained a truncated and silenced calreticulin gene with interrupted reading frame and no expression of the protein (Fig. 1 C). Calreticulin-deficient ES (G45crt^−/−) cell lines expressing N+P and P+C domains of the protein were also created and designated G45crt^−/− N+P and G45crt^−/− P+C.

Adipocyte differentiation was performed by hanging drop method. In brief, 1,000 ES cells in 20-μl drops were placed on the lids of tissue culture dishes for 2 d. The differentiation medium contained high glucose DME (Multicell; Wisent) supplemented with 15% FBS (Multicell; Wisent), sodium pyruvate, l-glutamine, minimal essential medium nonessential amino acids, and β-mercaptoethanol. After 2 d, the aggregated ES cells formed EBs, which were floated in differentiation medium supplemented with 10⁻⁷ M RA for an additional 3 d. The EBs were then plated onto tissue culture dishes coated with 0.1% gelatin (in PBS) in differentiation medium supplemented with 1 μg/ml insulin and 2 nM triiodothyronine. The differentiation medium was changed every 2 d for the duration of differentiation (15 d). The EBs were then collected for Western blot analysis, RNA isolation, or staining with oil red O.

Mouse embryonic fibroblasts were isolated from calreticulin-deficient (K42) and WT embryos (K41) as previously described (Nakamura et al., 2000). The 3T3-L1 preadipocyte cells were cultured in DME supplemented with 10% bovine growth serum at 37°C. For adipocyte differentiation, 2-d postconfluent cells were cultured for 2 d in differentiation medium containing 1 μM dexamethasone, 0.5 mM methylisobutylxanthine, 5 μg/ml insulin, and 0.5 μM rosiglitazone (Cayman Chemical). Afterward, the 2-d insulin (5 μg/ml) and 0.5 μM rosiglitazone were added. At day 9 of differentiation, the cells were harvested for Western blot analysis, RNA isolation, or staining with oil red O.

pSG5-rRxRα and pSG5-mPPARγ2 were gifts from R. Rachubinski (University of Alberta, Edmonton, Alberta, Canada). Plasmid DNA was purified using Mega plasmid preparation and columns (QIAGEN) as recommended by the manufacturer. Expression vectors encoding the HA-tagged N+P domain (pcDNA3.1_Zeo-CRTNPd-1) and P+C domain (pcDNA3.1_Zeo-CRTPCd-3) of calreticulin were previously described (Nakamura et al., 2001). ES cells were electroporated with 30 μg/cuvette of pcDNA3.1_Zeo-CRTNPd-1 or pcDNA3.1_Zeo-CRTPCd-3 expression vector by electroporation (1,500 V/cm; 25 IxF), and Zeocin (30 μg/ml)-resistant clones were isolated. Expression of the recombinant N+P and P+C domains was monitored with anti-HA antibodies (Nakamura et al., 2001).

3T3-L1 cells were stably transfected with pcDNA3.1-CRT (encoding HA-tagged full-length calreticulin), pcDNA3.1_Zeo-CRTNPd-1, or pcDNA3.1_Zeo-CRTPCd-3 using Lipofectamine reagent (Invitrogen) according to the manufacturer's instructions. 3T3-L1 clones expressing full-length calreticulin or calreticulin domains were selected with 50 μg/ml Zeocin. Expression of recombinant proteins was monitored with anti-HA antibodies (Nakamura et al., 2001).

3T3-NIH cells were cotransfected with reporter plasmid containing the calreticulin promoter, deletion of the calreticulin promoter, or the calreticulin promoter with mutations of PPRE1 or PPRE2 (pLC0, pLC1, pLC2, pLC0mt1, and pLC0mt2), PPARγ, and RXRα expression vectors. pLC1 and pLC2 plasmids encode the luciferase reporter gene under the control of 1.7-kb and 0.4-kb calreticulin promoters, respectively (Waser et al., 1997). pLC0 encoded the luciferase reporter gene under control of the 2.1-kb calreticulin promoter (Waser et al., 1997). To generate pLC0mt1 and pLC0mt2 plasmids, site-directed mutagenesis of PPRE1 and PPRE2 of the calreticulin promoter, respectively, was performed using the QuikChange Site-Directed Mutagenesis kit (Stratagene). Specifically, PPRE1 (AGGTCAGAGGACA) was mutated to AGGctcGAGGAtc, whereas PPRE2 (TGGCCCTTGACCT) was mutated to ccGggCTgctCCc (lowercase letters are mutated nucleotides). After 48 h, cells were harvested in a lysis buffer containing 100 mM Tris, pH 7.8, 0.5% NP-40, and 0.5 mM DTT. Luciferase and β-galactosidase activity were measured as described previously (Waser et al., 1997).

At the floating stage (days 3–5 of differentiation), EBs were incubated with KN-62 CaMK inhibitor or its inactive analogue (KN-93; Hidaka and Kobayashi, 1994) at concentrations of 10 μM for 2 h during the 3 d of EB flotation stage. The optimal inhibitory concentrations of KN-62 and KN-93 were determined to be in the range of 10 to 15 μM. Under these conditions, the drugs had no effect on cell survival/proliferation. For inhibition of calcineurin, 3T3-L1 preadipocytes were incubated with 1 μg/ml cyclosporin-A. The cells were then allowed to differentiate for an additional 15 d and were harvested for Western blot analysis, RNA isolation, or staining with oil red O.

The ⁴⁵Ca²⁺ measurements as well as [Ca²⁺] measurements using the Ca²⁺-sensitive fluorescent dye fura-2-AM were performed as previously described in detail by Mery et al. (1996). The ionomycin and BAPTA-AM treatment regimen was performed as previously described by Li et al. (2002) and Grey et al. (2005) with the following modifications: to obtain the optimal functional output, three consecutive pulse treatments were used for either ionomycin or BAPTA-AM. To chelate cytoplasmic Ca²⁺, floating EBs were incubated with 50 nM BAPTA-AM for 30 min at days 3–5 of differentiation. To increase cytoplasmic Ca²⁺, the EBs were incubated for 2 h with 500 nM ionomycin in the same manner. Exposure to ionomycin or BAPTA-AM exerts a significant change even after 1 d of treatment, but optimal results were obtained with three pulses of treatment on three consecutive days of differentiation. The cells were then allowed to differentiate on gelatinized (0.01%) tissue culture dishes for 15 d, after which they were collected for Western blot analysis, RNA isolation, or staining with oil red O.

Cells were harvested in a lysis buffer containing 50 mM Tris-HCl, pH 8.0, 120 mM NaCl, and 0.5% NP-40. Protein concentration was determined by the method of Bradford (Bradford, 1976). Protein samples (10 μg per lane) were separated by SDS-PAGE and transferred to nitrocellulose membrane (Nakamura et al., 2001). Western blot analysis was performed with the following antibodies: anti-PPARγ2 (Sigma-Aldrich and Santa Cruz Biotechnology, Inc.) at 1:1,000 dilution; anticalreticulin (Opas et al., 1991) at 1:300 dilution; anti-C/EBPα (Santa Cruz Biotechnology, Inc.) at 1:1,000 dilution; anti–glyceraldehyde 3-phosphate dehydrogenase (GAPDH; Labfrontiers) at 1:5,000 dilution; anti-CaMKII (Affinity BioReagents) at 1:1,000 dilution; antiphospho-CaMKII Thr286 (Invitrogen) at 1:500 dilution; anticalmodulin (Affinity BioReagents) at 1:1000 dilution; and anti-CREB pSer133 (Sigma-Aldrich) at 1:1,000 dilution. All secondary antibodies were horseradish peroxidase conjugated (Jackson Immunochemicals) and were used at 1:100,000 dilution. Immunoreactive bands were detected with a chemiluminescence ECL Western blotting system (GE Healthcare). Western blots were normalized by using anti-GAPDH antibodies. Relative mRNA levels were normalized to the housekeeping gene L32. Bands were quantified using ImageJ software (National Institutes of Health).

Total RNA was isolated from cells grown in 10-cm tissue culture dishes using TRIzol reagent (Invitrogen) according to the manufacturer's instructions. 1 μg RNA was used for synthesis of cDNA followed by RT-PCR using Superscript II (Invitrogen) according to the manufacturer's protocol. The following primers were used for PCR analysis: for PPARγ, forward primer 5′-ATCAAGTTCAAACATATCACC-3′ and reverse primer 5′-TTGTCTTGGATGTCCTCGATG-3′; for C/EBPα, reverse primer 5′-CGCAAGAGCCGAGATAAAGC-3′ and forward primer 5′-GCGGTCATTGTCACTGGTCA-3′; for aP2, reverse primer 5′-CATCAGCGTAAATGGGGATT-3′ and forward primer 5′-TCGACTTTCCATCCCACTTC-3′; and for L32, reverse primer 5′-CATGGCTGCCCTTCGGCCTC-3′ and forward primer 5′-CATTCTCTTCGCTGCGTAGCC-3′. PCR products were separated in 1.5% agarose gel. The mRNA levels were normalized using L32 as the housekeeping gene, and relative mRNA levels were quantified using ImageJ software.

Before staining with oil red O, cells were washed twice with PBS, fixed with 10% formaldehyde for 15 min at room temperature, and washed twice with distilled water and once with 70% isopropanol. Next, cells were stained for 1 h at room temperature with filtered oil red O at a ratio of 60% oil red O stock solution (0.5% wt/vol in isopropanol) to 40% distilled water. The cells were washed twice with distilled water, twice with PBS, and examined under a light microscope. An invertoscope (Diaphot; Nikon) equipped with a 10/0.25 DL dry plan Apochromatic objective (Nikon) was used for imaging at room temperature. A camera (Coolpix 4500; Nikon) was used for image acquisition. For quantitative analysis, oil red O was extracted with 5 ml isopropanol for 2 min, and optical density of each sample was determined at 540 nm.

Full-length RXRα and PPARγ and luciferase control proteins were synthesized using a coupled transcription and translation reticulocyte system (Promega; Guo et al., 2001). Synthetic oligodeoxynucleotides corresponding to PPAR site 1 (PPRE1; 5′-AGTGTGGAGGTCAGAGGACACCGGCTC-3′) and PPAR site 2 (PPRE2; 5′-TCCTGGCTGGCCCTTGACCTTATCCTG-3′) in the calreticulin promoter were used. An ideal PPRE oligodeoxynucleotide sequence, 5′-CAAAACTAGGTCAAAGGTCAAGGCATC-3′, was used as a positive control. Underlined residues correspond to the PPAR-binding site. Deoxyoligonucleotides were labeled with γ-[³²P]ATP (GE Healthcare) using T4 polynucleotide kinase. EMSA was performed as described previously (Guo et al., 2001). For supershift analysis, anti-HA tag antibodies were used.

ChIP assay was performed as described previously (Lynch et al., 2005). In brief, NIH-3T3 cells were transiently transfected with HA-PPARγ and cross-linked in 1% formaldehyde at room temperature for 20 min. Cells were then lysed with the Extract-N-Amp kit (Sigma-Aldrich) according to the manufacturer's instructions. Chromatin was sheared by sonication followed by centrifugation for 10 min. Supernatants were precleared with protein A–Sepharose beads for 1 h at 4°C. Immunoprecipitation was performed with mouse anti-HA antibodies at 4°C overnight. DNA was purified and analyzed by PCR using the following primers: PPRE site 1, forward primer 5′-TGTGTCTGAAGACAGCTACAGTG-3′ and reverse primer 5′-GCAGCAGGAGAAAAGAAGAGAG-3′; PPRE site 2, forward primer 5′-CTCTATGGCCTGAACAACTGTG-3′ and reverse primer 5′-TGGTCAGAGGGAAGAAGTAAGG-3′

Calcineurin activity assay was performed as previously described (Fruman et al., 1992). In brief, a peptide corresponding to the regulatory domain of protein kinase A (Sigma-Aldrich) was used as the substrate in an in vitro dephosphorylation assay (RII peptide). 1.0 × 10⁶ cells were lysed in 50 ml hypotonic lysis buffer containing 50 mM Tris, pH 7.5, 0.1 mM EGTA, 1 mM EDTA, 250 mM DTT, and protease inhibitors. 20 ml of lysate was added to 5 mM γ-[³²P]–labeled RII peptide. The reaction was performed for 20 min at 30°C in the presence of 0.5 mM okadaic acid with a total reaction volume of 60 ml. The released phosphate reported the activity of calcineurin and was expressed in picomoles of phosphate released per milligrams of total protein.

Cells on coverslips were fixed in 3.7% formaldehyde in PBS for 10 min. After washing (three times for 5 min) in PBS, the cells were permeabilized with 0.1% Triton X-100 in buffer containing 100 mM Pipes, pH 6.9, 1 mM EGTA, and 4% (wt/vol) polyethylene glycol 8000 for 2 min, washed three times for 5 min in PBS, and incubated with goat polyclonal anti-PPARγ2 antibody (diluted 1:50 in PBS; Santa Cruz Biotechnology, Inc.) for 30 min at room temperature. After washing three times for 5 min in PBS, the cells were incubated with the secondary antibody for 30 min at room temperature. The secondary antibody was FITC-conjugated donkey anti–goat IgG(H+L) diluted 1:50 in PBS. The cells were then incubated with 0.2 mM ribonuclease A (Sigma-Aldrich) for 2 h at room temperature to digest RNA and washed in PBS (three times for 5 min) followed by incubation with 1 μl/ml propidium iodide in PBS for 30 min to identify nuclei. For Nile red staining, the dye stock solution of 0.5 mg/ml in acetone was diluted in a glycerol/PBS mixture (0.05 ml of Nile red stock solution in 1 ml of 3:1 glycerol/PBS). Cells were fixed and permeabilized as for immunofluorescence and incubated with Nile red working solution for 10 min. After the final wash (three times for 5 min in PBS), the slides were mounted in fluorescent mounting medium (to prevent photobleaching; Dako). A confocal fluorescence microscope (MRC 600; Bio-Rad Laboratories) equipped with a 60/1.40 plan Apochromatic oil immersion objective (Nikon) and a krypton/argon laser was used for fluorescence imaging at room temperature. COMOS software (Bio-Rad Laboratories) was used for image acquisition.

Differences between mean values for different treatments were calculated using analysis of variance or two-tailed unpaired t test (where applicable) and were considered to be significant at P < 0.05 and P < 0.01.

Fig. S1 shows genomic configuration of the calreticulin gene. Fig. S2 shows nucleotide sequence of the calreticulin promoter.

The superb technical help of Ewa Dziak is greatly appreciated. We thank Dr. L. Guo for generating the targeting vector with the calreticulin gene flanked by loxP and thank R. Rachubinski and J. Molkentin for plasmid DNA.

M. Opas is a member of the Heart and Stroke/Richard Lewar Centre of Excellence. This work was supported by grants from the Canadian Institutes of Health Research (CIHR; 36384) and the Heart and Stroke Foundation of Ontario (T6181) to M. Opas and CIHR grants 15415 and 53050 to M. Michalak.