Endoplasmic reticulum form of calreticulin modulates glucocorticoid-sensitive gene expression.

Calreticulin is a ubiquitously expressed Ca2+-binding protein of the endoplasmic reticulum (ER), which inhibits DNA binding in vitro and transcriptional activation in vivo by steroid hormone receptors. Transient transfection assays were carried out to investigate the effects of different intracellular targeting of calreticulin on transactivation mediated by glucocorticoid receptor. BSC40 cells were transfected with either calreticulin expression vector (ER form of calreticulin) or calreticulin expression vector encoding calreticulin minus leader peptide, resulting in cytoplasmic localization of the recombinant protein. Transfection of BSC40 cells with calreticulin expression vector encoding the ER form of the protein led to 40-50% inhibition of the dexamethasone-sensitive stimulation of luciferase expression. However, in a similar experiment, but using the calreticulin expression vector encoding cytoplasmic calreticulin, dexamethasone-stimulated activation of the luciferase reporter gene was inhibited by only 10%. We conclude that the ER, but not cytosolic, form of calreticulin is responsible for inhibition of glucocorticoid receptor-mediated gene expression. These effects are specific to calreticulin, since overexpression of the ER lumenal proteins (BiP, ERp72, or calsequestrin) has no effect on glucocorticoid-sensitive gene expression. The N domain of calreticulin binds to the DNA binding domain of the glucocorticoid receptor in vitro; however, we show that the N+P domain of calreticulin, when synthesized without the ER signal sequence, does not inhibit glucocorticoid receptor function in vivo. Furthermore, expression of the N domain of calreticulin and the DNA binding domain of glucocorticoid receptor as fusion proteins with GAL4 in the yeast two-hybrid system revealed that calreticulin does not interact with glucocorticoid receptor under these conditions. We conclude that calreticulin and glucocorticoid receptor may not interact in vivo and that the calreticulin-dependent modulation of the glucocorticoid receptor function may therefore be due to a calreticulin-dependent signaling from the ER.

ticoid-sensitive gene expression. The N domain of calreticulin binds to the DNA binding domain of the glucocorticoid receptor in vitro; however, we show that the N؉P domain of calreticulin, when synthesized without the ER signal sequence, does not inhibit glucocorticoid receptor function in vivo. Furthermore, expression of the N domain of calreticulin and the DNA binding domain of glucocorticoid receptor as fusion proteins with GAL4 in the yeast two-hybrid system revealed that calreticulin does not interact with glucocorticoid receptor under these conditions. We conclude that calreticulin and glucocorticoid receptor may not interact in vivo and that the calreticulin-dependent modulation of the glucocorticoid receptor function may therefore be due to a calreticulin-dependent signaling from the ER.
Calreticulin is a ubiquitous and highly conserved Ca 2ϩ -binding protein of the endoplasmic reticulum (ER) 1 (for review see ). The protein is multifunctional and may play an important role in modulation of a variety of different cellular processes. These include chaperone activity, control of intracellular Ca 2ϩ homeostasis, modulation of cellular adhesion, antithrombotic activity, long term "memory" in Aplysia, cytotoxic T-cell function/activation, and a role in neutrophil phagocytosis, in viral RNA replication, in sperm cell function, and in autoimmunity (Kennedy et al., 1992;Nakamura et al., 1993;Sontheimer et al., 1993;Liu et al., 1994;Stendahl et al., 1994;Singh et al., 1994;Bastianutto et al., 1995;Camacho and Lechleiter, 1995;Gray et al., 1995;Kubawara et al., 1995;Nauseef et al., 1995;White et al., 1995;Peterson et al., 1995;Otteken and Moss, 1996;Mery et al., 1996). Surprisingly, calreticulin was also found to have a role in the control of the steroid-sensitive gene expression (Burns et al., 1994a(Burns et al., , 1994bDedhar et al., 1994aDedhar et al., , 1994b. The protein in vitro binds to the conserved amino acid sequence KXFF(K/R)R found between the two Zn 2ϩ fingers in the DNA binding domain of steroid receptors and prevents them from interacting with their DNAresponsive elements (Burns et al., 1994b;Dedhar et al., 1994b). Overexpression of calreticulin in mouse L cells and Vero fibroblasts inhibited transcriptional activation by glucocorticoid and androgen receptors in vivo, respectively (Burns et al., 1994b;Dedhar et al., 1994b). Calreticulin also binds, in vitro, to other transcription factors that contain the KXFF(K/R)R consensus amino acid sequence: the peroxisome proliferator-activated receptor-retinoid X receptor heterodimers, the hepatocyte nuclear factor-4, the chicken ovalbumin upstream promoter-transcription factor (Winrow et al., 1995), the vitamin D 3 receptor (Wheeler et al., 1995;St-Arnaud et al., 1995) and retinoic acid receptor (Dedhar et al., 1994;Desai et al., 1996). Surprisingly, in vivo calreticulin inhibits transcriptional activation by the vitamin D 3 receptor, retinoic acid receptor, androgen receptor, and glucocorticoid receptor (Dedhar et al., 1994;Burns et al., 1994;Wheeler et al., 1995;St-Arnaud et al., 1995;Desai et al., 1996) but not by the peroxisome proliferator-activated receptor-retinoid X receptor heterodimers (Winrow et al., 1995). It was also proposed that calreticulin binds to the KXFF(K/R)R motif conserved in the cytoplasmic tails of ␣ integrins and contributes to the regulation of cell adhesiveness (Rojiani et al., 1991;Leung-Hagesteijn et al., 1994;Coppolino et al., 1995). To directly affect integrins at the cytoplasmic site of the plasma membrane and to modulate steroid receptor function in the cytoplasm or in the nucleus, calreticulin would have to be present in the cytoplasm, but as yet there is no direct evidence for this.
How can calreticulin, an ER lumenal protein, modulate the function of nuclear hormone receptors? We postulated that for calreticulin to inhibit steroid-sensitive gene expression in vivo the protein must first gain access to the cytoplasm and then to the nucleus in order to interact with the receptors (Burns et al., 1994a). In this study we have tested this hypothesis by targeting calreticulin and calreticulin domains to different intracellular compartments followed by the analysis of calreticulindependent glucocorticoid-sensitive gene expression. We show that the ER, but not a cytosolic, form of calreticulin is responsible for inhibition of glucocorticoid receptor-mediated gene expression. Our results suggest that in vivo calreticulin and glucocorticoid receptor may not interact and that the observed modulation of glucocorticoid receptor function may therefore be due to signal transduction from the ER mediated by calreticulin.

EXPERIMENTAL PROCEDURES
Materials-Pipes, luciferin, glycine, coenzyme A, dithiothreitol, ATP, Nonidet P-40, Triton X-100, dexamethasone, FITC-ConA, and o-nitrophenyl-␤-D-galactopyranoside were from Sigma. Dimethyl sulfoxide and protease inhibitors were from Boehringer Mannheim. FITC-and Texas Red-conjugated secondary antibodies were purchased from Bio/ Can Scientific (Mississauga, Ontario). Restriction endonuclease and DNA-modifying enzymes were obtained from Boehringer Mannheim, Life Technologies, Inc., and Bio/Can Scientific. Peroxidase-conjugated secondary anti-rabbit, anti-goat, and anti-human antibodies were from Bio/Can Scientific. Vinol 205S was from St. Lawrence Chemical (Toronto, Ontario, Canada). Matchmaker two-hybrid system was purchased from Clontech Laboratories. The pBluescript and pRc/CMV plasmids were from Clontech and Invitrogen, respectively. pSVL plasmid was from Pharmacia. Plasmid purification kits were purchased from QIAGEN Inc. (Chatsworth, CA). Rabbit anti-calreticulin antibodies were from Affinity Probes. Human anti-lamin B receptor autoantibodies were a gift from Dr. H. J. Worman (College of Physicians and Surgeons of Columbia University). NIH/3T3 cells were obtained from the ACCT, and the BSC40 cells (ATCC CCL-26) were a gift from Dr. R. Rachubinski (Department of Anatomy, University of Alberta). All chemicals were of the highest grade available.
Plasmids and Epitope Tagging of Calreticulin-For expression experiments cDNAs encoding different versions of calreticulin were subcloned into either pSVL (Pharmacia) or pRc/CMV (Invitrogen) plasmids. pSVL-CRT (pSCR) calreticulin expression vector was constructed by inserting the DraI/SmaI restriction DNA fragment (nucleotides 20 -1653) of pcDx-CRT (GenBank TM accession number J05138) (Fliegel et al., 1989) into SmaI-digested pSVL vector (Pharmacia). This vector was used for epitope-tagging of calreticulin. First the XhoI/SstI DNA restriction fragment, containing cDNA encoding full-length calreticulin, was excised from the pSCR vector and inserted into pBluescript to form pBCR. Recombinant calreticulin was tagged with a specific 12-amino acid epitope (NH 2 -PSSRGRNTPGKP-COOH) referred to as the dystrophin tag (DT). DNA encoding this epitope was engineered into the carboxyl terminus of calreticulin. This peptide is unique to the carboxyl terminus of the muscular dystrophy gene product, dystrophin (Milner et al., 1992). The following complementary oligodeoxynucleotides, which encoded for the DT, were synthesized with the 5Ј NotI and 5Ј StyI restriction sites 5Ј-GGC CGC CCC TAG TTC AAG AGG AAG AAA ATA CCC CTG GA AAG CC-3Ј and 5Ј-CTT GGG CTT TCC AGG GGT ATT TCT TCC TCT TGA ACT AGG GGC-3Ј, respectively. The synthetic oligodeoxynucleotides were phosphorylated with DNA kinase, annealed, and inserted into the NotI and StyI sites of the pBCR vector to generate the pBCR-DT plasmid. This ligation reaction deleted nucleotides 1290 -1299 of the cDNA encoding calreticulin resulting in the loss of three amino acid residues, 394 -396 (Fliegel et al., 1989), and the addition of 12 amino acids encoding the DT. For the transfection experiments, the XhoI/SacI fragment from pBCR-DT was cloned into XhoI/ SacI restriction sites of pSVL plasmid to generate pSCR-DT calreticulin expression vector. This vector was used for transient transfection experiments of the full-length ER form of calreticulin.
In order to express the cytoplasmic form of calreticulin, cells were transfected with pSCR-L-DT vector, which encoded full-length, dystrophin-tagged calreticulin minus leader peptide (referred to throughout the paper as cytoplasmic calreticulin). To generate pSCR-L-DT vector, calreticulin cDNA was synthesized by polymerase chain reaction-driven amplification of pRCR-DT using primers 5Ј-ATACTCGAG ATG GAG CCC GTC GTC ACT TCA-3Ј (encoding six NH 2 -terminal amino acids of the mature calreticulin and 5Ј-ATG initiation codon) and 5Ј-GGGAAT-TCAGAGACATTATTGGCTCTGCG-3Ј (nucleotide sequence 1242-1264) (36 base pairs behind the stop codon of the calreticulin clone; Fliegel et al. (1989)) with flanking XhoI and EcoRI restriction sites, respectively. The DNA fragment was first inserted into XhoI/EcoRI restriction sites of the pBluescript to obtain a vector designated pBCR-L-DT. The XhoI/SacI fragment of pBCR-L-DT was inserted into XhoI/ SacI restriction sites in pSVL plasmid to obtain pSCR-L-DT. To generate the pSCR-NP vector encoding the NϩP domain of calreticulin  the BamHI DNA fragment (encoding the C domain of calreticulin) was excised and removed from the pSCR-L-DT vector. Dystrophin tagging of the N domain was carried out using synthetic oligodeoxynucleotides as described for the pSCR-DT calreticulin expression vector.
In order to generate pSVL-CSQ vector encoding canine cardiac calsequestrin, the SmaI restriction DNA fragment from IC3 S/S 1378 pBluescript vector (a generous gift of Dr. L. R. Jones, Krannert Institute of Cardiology, Indianapolis) was subcloned into the SmaI restriction site of pSVL plasmid. MMTV-GRE-luciferase (pJA358) and MMTV-luciferase plasmids with deleted GRE were a gift from Dr. The nucleotide sequence of all cDNAs encoding calreticulin was confirmed by DNA sequencing. Sequencing was performed in the DNA Sequencing Laboratory in the Department of Biochemistry, University of Alberta, using an Applied Biosystems model 373A DNA sequencer. Synthetic oligodeoxynucleotides and sequencing primers were made in the DNA Sequencing Laboratory in the Department of Biochemistry, using an Applied Biosystems model 392 DNA/RNA synthesizer.
DNA Mobility Shift Assay-The DNA binding domain of glucocorticoid receptor was expressed in Escherichia coli as a fusion protein with glutathione S-transferase as described by Burns et al. (1994b). The DNA shift assay was carried out in 1 mM EDTA, 20 mM NaCl, 0.05% bovine serum albumin, 4 mM dithiothreitol, 10% glycerol, 2 g of poly-(dI:dC), and 20 mM Tris, pH 7.5. Complementary oligodeoxynucleotide containing the GRE (5Ј-TCC TTG TTT TAA GAA CAG TTA TCG ATT ATA AAC-3Ј and 5Ј-GTT TAT AAT CGA TAA CGT TTC TTA TTA AAA CAA CGA-3Ј) were used in this assay. Calreticulin was isolated from canine pancreas by the ammonium sulfate precipitation procedure (Milner et al., 1991).
Isolation and Extraction of Nuclei-Nuclei were isolated from rat liver by the procedure described by Blobel and Potter (1966). Tissue was homogenized in 2 volumes of a solution containing 250 mM sucrose, 25 mM KCl, 5 mM MgCl 2 , and 50 mM Tris, pH 7.5. The homogenate was overlaid on a step sucrose gradient consisting of 1.62 M and 2.3 M sucrose followed by centrifugation at 124,000 ϫ g for 30 min using a Beckman SW 50.1 rotor. The white pellet containing pure nuclei was collected, suspended in a solution containing 25 mM KCl, 5 mM MgCl 2 , and 50 mM Tris, pH 7.5, and centrifuged for 12 min at 13,000 ϫ g. The pellet contained pure, intact nuclei. To remove the outer nuclear membranes, the purified nuclei were solubilized with 5% Triton X-100 in 25 mM KCl, 5 mM MgCl 2 , and 50 mM Tris, pH 7.5, followed by centrifugation at 800 ϫ g for 5 min. The supernatant (Triton X-100-soluble fraction) contained solubilized outer nuclear membrane. The final pellet, containing the outer membrane-stripped nuclei, was suspended in a solution containing 25 mM KCl, 5 mM MgCl 2 , and 50 mM Tris, pH 7.5, and centrifuged at 13,000 ϫ g for 10 min. The quality of the isolated nuclei was monitored by electron microscopy (Blobel and Potter, 1966).
Yeast Two-hybrid System-To study the interaction between the glucocorticoid receptor and calreticulin's domains in vivo, we employed the yeast two-hybrid system (Fields and Song, 1989;Chien et al., 1991). For the purpose of this study, we have generated the GAL4 DNA binding domain fusion protein with the full-length calreticulin (designated pGB-CRT) or with the N domain of the protein (amino acid residues 1-174) (designated pGB-N) as described by Baksh et al. (1995). To generate the GAL4 activating domain fusion protein with the DNA binding domain of the glucocorticoid receptor, (designated pGAD-GR) cDNA was generated by polymerase chain reaction-driven amplification as described by Burns et al. (1994). Sequences of every construct were confirmed by nucleotide sequencing. Yeast strain SFY526 was co-transformed with either pGAD-GR and pGB-N or pGB-CRT using 200 ng of each construct following the protocol recommended by the manufacturer. As negative controls, the pGB-N and pGB-CRT were each cotransformed with the pGAD424 vector, and pGAD-GR was co-transformed with the pGBT9 vector. As positive control, yeast was transformed with pCL1 vector encoding GAL4. The colony lift filter assay for ␤-galactosidase activity was performed according to the manufacturer's protocol. Color development was monitored for 8 h.
Immunoblotting and Immunocytochemistry-Cellular extracts were prepared for immunoblotting as described by Mery et al. (1996). Proteins (equal to approximately 100,000 cells/lane) were separated by SDS-polyacrylamide gel electrophoresis on 10% polyacrylamide gels as described by Laemmli (1970) and then transferred to nitrocellulose membranes (Towbin et al., 1979). Standards were Bio-Rad prestained markers, phosphorylase b (135,000), bovine serum albumin (85,000), ovalbumin (50,000), carbonic anhydrase (39,000), soybean trypsin inhibitor (27,000), and lysozyme (17,000). Immunoblotting was carried out as described by Milner et al. (1991). Goat anti-calreticulin and rabbit anti-DT synthetic peptide antibodies were described earlier by Milner et al. (1991) and Milner et al. (1992), respectively. Rabbit anti-DT was raised against a synthetic peptide (NH 2 -PSSRGRN-TPGKP-COOH) encoding part of the carboxyl-terminal region of dystrophin (Milner et al., 1992). The peptide was chemically synthesized on a peptide synthesizer (model 430A; Applied Biosystems, Inc., Foster City, CA) and then coupled to keyhole limpet hemocyanin by the Alberta Peptide Institute (University of Alberta). For affinity purification, the bovine serum albumin-conjugated dystrophin synthetic peptide was coupled to the CNBr-activated Sepharose as recommended by the manufacturer. Antibody purification was carried out as described by Harlow and Lane (1988). Human anti-lamin B receptor autoantibodies (Ye and Worman, 1994;Lin et al., 1996) were used at a 1:200 dilution.
For intracellular localization of calreticulin, polyclonal goat or rabbit anti-calreticulin antibodies were used at a 1:50 dilution in PBS Nauseef et al., 1995). For localization of the dystrophin-tagged protein, the affinity-purified rabbit anti-dystrophin synthetic peptide (anti-DT antibody) was used at a 1:10 dilution (Milner et al., 1992). The ER was visualized with FITC-ConA. For fluorescence microscopy, cells on coverslips were fixed for 10 min in a solution containing 3.8% formaldehyde in PBS, extracted with 0.1% Triton X-100 in a buffer containing 100 mM Pipes, pH 6.9, 1 mM EGTA, and 4% (w/v) polyeth-ylene glycol 8000 for 3 min, washed in PBS for 10 min, and then processed for labeling with primary antibodies followed by appropriate secondary antibodies. The secondary antibodies were FITC-conjugated donkey anti-goat (diluted 1:30 in PBS) and Texas Red-conjugated donkey anti-rabbit (used at 1:30 dilution). For double labeling, all incubations were done sequentially. After the final wash (3 times for 5 min), the slides were mounted in Vinol 205S containing 0.25% 1,4-diazabicyclo-(2,2,2)-octane and 0.002% p-phenylenediamine to prevent photobleaching. A Bio-Rad model MRC-600 confocal fluorescence microscope equipped with a krypton/argon laser was used. To maintain pixel registration necessary for subtractive imaging, fluorescence images were collected using a simultaneous double labeling K1/K2 filter set. However, the images were recorded using single excitation wavelengths in order to avoid any bleed-through of emissions. Furthermore, no appreciable specific signal could be detected if any of the primary antibodies were omitted from the staining procedure. No signal could be detected when both the primary and secondary reagents were omitted. Finally, staining of mock-transfected cells with the anti-DT antibody did not produce any appreciable specific labeling. Subtractive imaging was performed by arithmetic subtraction of pixel-registered images of two labels in the same cells using the Image-1 program (West Chester, PA).

ER Form of Calreticulin Inhibits Transactivation by Glucocorticoid Receptor in Vivo-Transient transfection assays
were carried out to investigate whether targeting of calreticulin to different intracellular compartments had an effect on transactivation mediated by glucocorticoid receptors. Two cell lines were used in these experiments, the BSC40 African green monkey kidney monolayer cells derived from BS-C-1 cells (Hopps et al., 1963) and the mouse fibroblast NIH/3T3 cells. BSC40 and NIH/3T3 cells were co-transfected with different calreticulin expression vectors, MMTV-GRE-luciferase and p␤GAL control vector. Since the BSC40 cells do not contain an endogenous glucocorticoid receptor, they were also co-transfected with VERO plasmid encoding the glucocorticoid receptor. All results were normalized for transfection efficiency with p␤GAL, ␤-galactosidase expression vector. In the presence of 10 Ϫ6 M Dex, BSC40 or NIH/3T3 cells co-transfected with MMTV-GRE-luciferase and the glucocorticoid receptor expression vector (VERO) exhibited Ͼ20-fold increase in the reporter gene activity over cells grown in the absence of Dex (Fig. 1). In agreement with earlier observations (Burns et al., 1994b;Winrow et al., 1995), co-transfection of these cells with MMTV-GRE-luciferase, VERO plasmid (BSC40 cells only), and calreticulin expression vector (pSCR-DT) led to significant inhibition of the Dex-sensitive stimulation of luciferase expression (Fig. 1). The effect of calreticulin on transactivation of dexamethasone-sensitive reporter gene was dose-dependent ( Fig. 2). Maximal effects were observed when BSC40 cells were transfected with 10 g of calreticulin expression vector (Fig. 2). In order to determine the level of calreticulin expression in BSC40 cells transiently transfected with calreticulin expression vector (10 g), we performed quantitative immunological analysis of calreticulin-transfected and control cells using antibodies raised against calreticulin (Fig. 3A). The immunoblot shown in Fig. 3A was scanned, and the level of calreticulin in cells transfected with calreticulin expression vector pSCR-DT was determined to be increased approximately 4-fold over that of the endogenous protein.
In order to test if calreticulin's effects on glucocorticoid receptor function are due to the cytoplasmic protein, we generated a calreticulin expression vector encoding calreticulin minus leader peptide (pSCR-L-DT). Fig. 1 shows that cotransfection of either BSC40 or NIH/3T3 cells with MMTV-GRE-luciferase, VERO plasmid encoding glucocorticoid receptor (BSC40 cells only), and pSCR-L-DT plasmid resulted in less than 10% inhibition of Dex-stimulated activation of the luciferase reporter gene. This inhibition was not statistically significant (p Ͼ 0.42). Co-transfection of BSC40 cells with pSCR-L-DT cytoplasmic calreticulin expression vector and an increasing amount of glucocorticoid receptor expression vector did not have any additional effect on the glucocorticoid receptor-sensitive expression of the luciferase reporter gene (Fig. 4), ruling out the possibility that the observed lack of inhibition was not due to a limited number of glucocorticoid receptor molecules in BSC40 cells. Immunoblot analysis with antibodies raised against calreticulin showed that BSC40 cells transfected with pSCR-L-DT had a ϳ2.5-fold increase in the level of calreticulin (Fig. 3B).
We have previously shown that the in vitro interaction between the DNA binding domain of glucocorticoid receptor and calreticulin is confined to the N domain and NϩP domain of the protein (Burns et al., 1994b). To test the in vivo role of the N domain of calreticulin in modulation of glucocorticoid receptorsensitive gene expression BSC40 cells were co-transfected with plasmids encoding the NϩP domain (pSCR-NP). The pSCR-NP vector was chosen for these experiments, since it encodes calreticulin minus the C domain, a high capacity Ca 2ϩ binding region of the protein . This was to avoid any nonspecific effects of cytoplasmic Ca 2ϩ sequestration by calreticulin on the function of the glucocorticoid receptor. In addition, this expression vector encoded calreticulin domains minus leader peptide, resulting in the targeting of the recombinant protein to the cytoplasm. Furthermore, the DT was introduced into the C terminus of the NϩP domain to enable monitoring expression and intracellular localization of the recombinant protein. Fig. 1 shows that expression of NϩP domain in either BSC40 or NIH/3T3 cells transfected with MMTV-GRE-luciferase had no effect on Dex-sensitive luciferase expression. The presence of the NϩP domain of calreticulin in BSC40 cells transfected with the NϩP domain expression vector was established using the anti-DT antibodies (Fig. 3C).
Intracellular localization of the recombinant calreticulin and FIG. 1. Calreticulin represses the in vivo transactivation of the glucocorticoid-sensitive gene expression in BSC40 and NIH/3T3 cells. BSC40 or NIH/3T3 cells were co-transfected with MMTV-GREluciferase vector, p␤GAL control vector, VERO plasmid expressing glucocorticoid receptor (BSC40 cells only), and different calreticulin expression vectors as indicated. Cellular extracts were prepared and assayed for luciferase and ␤-galactosidase expression as described under "Experimental Procedures." Cells were incubated in DMEM containing 10% charcoal-treated calf serum for 12 h followed by incubation for 24 h with DMEM alone (Dex Ϫ) or in DMEM containing 10 Ϫ6 dexamethasone (Dex ϩ). The values shown are relative activities from four independent transfections done in triplicate and normalized for ␤-galactosidase activity. The means Ϯ S.D. are given.

FIG. 2. Dose-dependent effects of calreticulin on the in vivo transactivation of the glucocorticoid receptor in BSC40 cells.
BSC40 cells were co-transfected with MMTV-GRE-luciferase vector, p␤GAL control vector, VERO plasmid expressing glucocorticoid receptor, and different amounts of the calreticulin expression vector (pSVL-CRT) as indicated. Cellular extracts were prepared and assayed for luciferase and ␤-galactosidase expression as described under "Experimental Procedures." Cells were incubated in charcoal-treated calf serum for followed by incubation for 24 h with DMEM alone (Dex Ϫ) or in DMEM containing 10 Ϫ6 dexamethasone (Dex ϩ). The values shown are relative activities from three independent transfections done in triplicate and normalized for ␤-galactosidase activity. The means Ϯ S.D. are given. The recombinant proteins were co-localized with an ER marker, ConA (Virtanen et al., 1980;Tartakoff and Vassalli, 1983) (Fig. 5A), or endogenous calreticulin (Fig. 5B). As expected, in cells transfected with the pSCR-DT calreticulin expression vector the full-length protein was found in the ER as demonstrated by both the ConA fluorescence (Fig. 5A) and labeling of the endogenous calreticulin (Fig. 5B). In cells transfected with the pSCR-L-DT cytoplasmic calreticulin expression vector, the recombinant protein was localized to the cytoplasm (Fig. 5A). The recombinant NϩP domain of calreticulin lacking leader peptide also localized to the cytoplasm (Fig. 5, NϩP domain). Subtractive imaging (Fig. 5A, difference) of both the cytoplasmic calreticulin and the NϩP domain against the ER clearly demonstrated cytoplasmic targeting of these recombi-nant molecules. This was further confirmed by subtractive imaging of DT on the recombinant proteins against the anticalreticulin label (Fig. 5B). In the pSCR-L-DT-transfected cells (Cytoplasmic CRT) the goat antibody detected calreticulin in both the ER and the cytoplasm; thus, the differential image exposed the ER-localized protein (Fig. 5B, Cytoplasmic CRT, difference). In contrast, since this antibody recognizes an epitope close to the C terminus of calreticulin, it did not label the NϩP domain in the pSCR-NP-transfected cells, and the subtractive imaging yielded familiar "ER exclusion" pattern in the differential image (Fig. 5B, NϩP domain, difference). Based on the transfection experiments (Fig. 1) and the immunological localization of the recombinant proteins (Fig. 5), we concluded that the ER, but not the cytosolic, form of calreticulin inhibited glucocorticoid receptor-dependent activation of the luciferase reporter gene in both BSC40 and NIH/3T3 cells.
Inhibition of Glucocorticoid Receptor Function Is Specific to Calreticulin-In order to test if the observed inhibition of glucocorticoid receptor-sensitive gene expression is calreticulinspecific or if it could be mediated by other lumenal ER proteins, we performed a series of co-transfection experiments. BSC40 cells were transfected with MMTV-GRE-luciferase and expression plasmid encoding BiP, ERp72, or with calsequestrin the sarcoplasmic reticulum lumenal Ca 2ϩ -binding protein. Fig. 6 shows that overexpression of BiP or ERp72 did not have any effect on glucocorticoid receptor-dependent expression of luciferase reporter gene. However, transient co-transfection of BSC40 cells with both calreticulin (pSCR-DT) and ERp72 expression vectors led to 50% inhibition of the Dex-dependent  (NϩP domain row). The full-length recombinant calreticulin and the ER co-localize. Subtractive imaging (difference) exposes the ER only in a cell that does not express a detectable level of DT. Both the recombinant calreticulin without leader peptide (Cytoplasmic CRT) and the recombinant NϩP domain of calreticulin lacking leader peptide (NϩP domain) protein are found in the cytoplasm throughout the cell (dystrophin tag). The cytoplasmic localization of both recombinant molecules is shown in their respective differential images as the DT fluorescence outside the ER, which appears dark after the image subtraction. In mock-transfected cells (Empty pSVL row), no labeling could be detected with anti-DT antibodies. B, co-localization of anticalreticulin labeling (calreticulin column) and the dystrophin-tagged (dystrophin tag column) recombinant calreticulin (Full length CRT row), recombinant calreticulin lacking leader peptide (Cytoplasmic CRT row), and recombinant NϩP domain of calreticulin lacking leader peptide (NϩP domain row). Because calreticulin and the DT on the full-length protein co-localize, the differential image is black (difference). The recombinant calreticulin without leader peptide (Cytoplasmic CRT) is detected in the cytoplasm by both the anti-calreticulin (calreticulin) and anti-dystrophin (dystrophin tag). Since the recombinant calreticulin is absent from the ER, the differential image exposes only the endogenous, ER-resident calreticulin (difference). The NϩP domain of calreticulin (NϩP domain) is also found in the cytoplasm by the DT detection (dystrophin tag). However, in contrast to the cytoplasmic calreticulin, the NϩP domain is not recognized by the anti-calreticulin antibody, which hence detects only the ER form of calreticulin in the pSCR-NP-transfected cells (calreticulin). Consequently, the subtractive imaging produces an ER exclusion pattern in the differential image (NϩP domain, difference). In cells transfected with the empty pSVL vector (Empty pSVL row), no dystrophin labeling could be detected. luciferase activity (Fig. 6). Overexpression of cardiac calsequestrin, a Ca 2ϩ -binding protein of similar physicochemical properties to calreticulin (Scott et al., 1988), also did not have any effect on the Dex-sensitive expression of luciferase (Fig. 6). To confirm that BiP, ERp72, and calsequestrin were expressed in transfected cells, we carried out Western blot analysis of cellular extracts (Fig. 7). Since BiP expression vector encoded BiP tagged with specific c-myc epitope, we detected expression of this protein using the anti-c-myc antibodies (Fig. 7A). Fig. 7B shows that cells transfected with the ERp72 expression vector had a 3-fold increased level of the immunoreactive ERp72. Finally, using anti-calsequestrin antibodies, we showed that the protein was expressed in BSC40 cells transfected with the calsequestrin expression vector (Fig. 7C). These results suggest that calreticulin-dependent modulation of the Dex-sensitive luciferase reporter gene was specific to calreticulin and that it was not a result of protein overexpression in the lumen of the ER or increased sequestration of ER lumenal Ca 2ϩ as documented by overexpression of calsequestrin.
Calreticulin Is Not a Resident Nuclear Protein-As shown in Fig. 5, calreticulin was not detected in the nucleus of transfected cells, yet we observed significant inhibition of the glucocorticoid receptor-dependent expression of the reporter gene (Fig. 1). However, we and others have reported that a calreticulin-like antigen is detected in the nuclei of L6 cells  and in the isolated nuclei from P19 cells (Dedhar et al., 1994a). In order to determine if calreticulin is a resident nuclear protein, we have immmunolabeled isolated nuclei with two different anti-calreticulin antibodies. Furthermore, we have fractionated rat liver nuclei to determine if there is any association of calreticulin with the nucleus and/or the nuclear membranes. Fig. 8 shows staining of the isolated nuclei either with goat anti-calreticulin (B) or rabbit anti-calreticulin (D) antibody. The goat anti-calreticulin antibody gave no nuclear staining of the isolated nuclei, whereas the rabbit anti-calreticulin showed some patches of nuclear staining. The reason for the different staining by these two antibodies is not clear at present, but it might be due to a different degree of contamination of the isolated nuclei with a nuclear envelope mem-brane. To further test for the presence of calreticulin in the nucleus, we have fractionated rat nuclei followed by immunological analysis of different fractions with the goat and the rabbit anti-calreticulin antibodies. As a control, we employed the anti-lamin B receptor antibodies (Ye and Worman, 1994;Lin et al., 1996). Fig. 9 shows that a 60-kDa calreticulin was present in the isolated, intact nuclei (lane 3). Extraction of purified nuclei with 2% Triton X-100 leads to separation of the detergent-soluble outer nuclear membrane from the nuclear content surrounded by inner nuclear membrane (Blobel and Potter, 1966). Resident nuclear proteins are associated with the Triton-insoluble nuclear fractions, whereas the Triton-soluble fractions contain proteins of the ER origin. Fig. 9, A and B, lane 4 shows that calreticulin was associated with the Tritonsoluble outer nuclear membrane and not with the nuclei (lane 6). An identical pattern of immunological reactivity of different nuclear fractions was obtained using either goat or rabbit anticalreticulin antibodies (Fig. 9, A and B). The same protein blot was also tested for the reactivity with the antibodies against the lamin B receptor, a specific integral inner nuclear membrane protein (Ye and Worman, 1994;Lin et al., 1996). As expected, a 60-kDa lamin B receptor was present in the intact nuclei and in the Triton-insoluble fractions containing nuclei content surrounded by inner nuclear membrane (Fig. 9C). There was no immunoreactive lamin B receptor in the Tritonsoluble fraction (Fig. 9C, lane 4), which contained calreticulin (Fig. 9, A and B, lane 4). Taken together, these results suggested that calreticulin is not a resident nuclear protein and, therefore, that calreticulin may not inhibit function of the glucocorticoid receptor by direct interaction with its DNA binding domain in the cytoplasm or the nucleus. This conclusion is further supported by the in vitro DNA mobility shift experiments. Fig. 10 shows that, as reported earlier (Burns et al., 1994b), calreticulin inhibits glucocorticoid receptor binding to the GRE in vitro. However, when glucocorticoid receptor was first bound to the GRE prior to incubation with calreticulin, the protein no longer inhibited glucocorticoid receptor interaction with the GRE (Fig. 10). This suggests that in vitro the DNA bound (nuclear form) of the glucocorticoid receptor does not interact with calreticulin.
Calreticulin Does Not Interact with the DNA Binding Domain of the Glucocorticoid Receptor in the Yeast Two-hybrid System-Finally, we employed the yeast two-hybrid system (Fields and Song, 1989;Chien et al., 1991) to test if the DNA binding domain of the glucocorticoid receptor and calreticulin interact under the in vivo cellular conditions. For these experiments, calreticulin and the N domain of calreticulin (amino acid residues 1-174) were fused to the GAL4-DNA binding domain. Fig. 11 (filters 3 and 6) shows that calreticulin and the N domain fusion proteins did not activate reporter gene expression when co-expressed with the GAL4 transcriptional activation domain. However, in a control experiment (Fig. 11, filter 1) expression of GAL4 transcription factor induced expression of ␤-galactosidase. Next the DNA binding domain of the GR was fused to the GAL4 transcriptional activation domain. This fusion protein did not activate the reporter gene when co-expressed with the GAL4-DNA binding domain (Fig. 11, filters 2 and 5). When jointly expressed as GAL4 fusion proteins, fulllength calreticulin, or the N domain with the DNA binding domain of GR (Fig. 11, filters 4 and 7), they also did not activate expression of ␤-galactosidase, indicating that the two proteins did not interact under these conditions.

DISCUSSION
In order to examine a role of calreticulin in regulation of the steroid-sensitive gene expression, the protein was targeted either to the lumen of the ER or to the cytoplasm followed by FIG. 8. Immunostaining of the isolated rat liver nuclei with anti-calreticulin antibodies. Rat liver nuclei were isolated by the procedure described under "Experimental Procedures" and immunostained with either goat anti-calreticulin (B) or rabbit anti-calreticulin (D) antibodies. A and C are phase contrast picture of the isolated nuclei.
analysis of the effect of calreticulin on the glucocorticoid receptordependent reporter gene expression. We show that the ER, but not cytosolic, form of calreticulin is the most effective in inhibition of the glucocorticoid receptor-mediated gene expression. This effect is specific to calreticulin, since transfection of BSC40 cells with expression vectors for ER lumenal proteins such as BiP, ERp72, or calsequestrin did not have a significant effect on the glucocorticoid-sensitive gene expression. Furthermore, we show that the NϩP domain of calreticulin, when synthesized without the ER signal sequence, also does not inhibit glucocorticoid receptor function. The N domain of the protein interacts with the DNA binding domain of the glucocorticoid receptor in vitro (Burns et al., 1994b). Furthermore, expression of either calreticulin or the N domain of calreticulin and the DNA binding domain of glucocorticoid receptor as fusion proteins with GAL4 in the yeast two-hybrid system revealed that they do not interact with glucocorticoid receptor under these in vivo conditions. We conclude that calreticulin may not interact with steroid receptors in vivo but that it may function as a "signaling" molecule from the lumen of the ER.
Calreticulin-like immunoreactivity was detected in the nucleus of some cells Dedhar et al., 1994aDedhar et al., , 1994b, suggesting that the nuclear antigen, if it was calreti-culin, could have been responsible for affecting gene expression. Despite the apparent presence of calreticulin in intact cells, a calreticulin-like antigen is not consistently found by immunofluorescence in the isolated nuclei, and the immunostaining for calreticulin is not reproducible when different antibodies raised against the protein are used. Calreticulin also fractionates away with the outer nuclear membrane, but not with the nuclear matrix or with the inner nuclear membrane (containing lamin B receptor, a nuclear membrane marker), indicating that calreticulin is associated with the ER but not with the nucleus. We were unable to detect any nuclear and cytoplasmic calreticulin by immunofluorescence analysis of the BSC40 and NIH/3T3 cells overexpressing calreticulin, using two different anti-calreticulin antibodies. This is similar to other cells overexpressing calreticulin including HeLa, mouse L fibroblasts, A10, and COS-7 cells (Sönnichsen et al., 1994;Bastianutto et al., 1995;Mery et al., 1996). 2 It is possible that levels of calreticulin in the cytoplasm or the nucleus may be too low for the detection methods used in this study. This is unlikely, however, since overexpression of cytoplasmic calreticulin did not significantly affect glucocorticoid receptor function. FIG. 9. Identification of calreticulin in rat liver nuclear fractions. Nuclei were isolated from rat liver by sucrose gradient centrifugation and fractionated with Triton X-100 as described under "Experimental procedures." Proteins were separated by SDS-polyacrylamide gel electrophoresis, transferred to nitrocellulose filters, and incubated with goat anti-calreticulin (A), rabbit anti-calreticulin (B), or human anti-lamin B receptor (C) antibodies. Sucrose gradient-purified nuclei (lane 3) were washed with 25 mM KCl, 5 mM MgCl 2 , and 50 mM Tris, pH 7.5. The purified nuclei were solubilized with 5% Triton X-100 followed by centrifugation at 800 ϫ g for 5 min. The Triton X-100 supernatant contains solubilized nuclear envelope, whereas the outer membranestripped nuclei were collected in the pellet (Blobel and Potter, 1966). Removal of the outer nuclear membrane was confirmed by electron microscopy of the intact nuclei and the Triton X-100-extracted nuclei. Lane 1, purified recombinant calreticulin; lane 2, wash of the intact nuclei; lane 3, purified, intact nuclei; lane 4, Triton X-100-soluble, outer nuclear membrane fraction; lane 5, wash of the Triton X-100-fractionated nuclei; lane 6, outer nuclear membrane-stripped nuclei. The arrows indicate calreticulin. The two low molecular weight immunoreactive protein bands in C may be degradation products of the lamin B receptor. The positions of Bio-Rad prestained molecular mass marker proteins (85, 50, 39, and 27 kDa) are indicated by arrowheads or dots. FIG. 11. Calreticulin does not interact with the DNA binding domain of the glucocorticoid receptor in the yeast two-hybrid system. Appropriate expression vectors were transformed into yeast strain SFY526 as described under "Experimental Procedures." The colony lift filter assay for ␤-galactosidase activity was performed according to the manufacturer's protocol using 5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside as a substrate. Filter 1, yeast was co-transformed with the pCL1 control vector encoding GAL4; filter 2, yeast was co-transformed with pGAD-GR (encoding the DNA binding domain of glucocorticoid receptor) and the pGBT9 vectors (encoding the GAL4 DNA binding domain); filter 3, yeast was co-transformed with the pGB-CRT (full-length calreticulin) and the pGAD (encoding GAL4 activation domain) vectors; filter 4, yeast was co-transformed with the pGB-CRT and pGAD-GR vectors; filter 5, yeast was co-transformed with pGAD-GR and the pGBT9 vectors; filter 6, yeast was co-transformed with the pGB-N (encoding the N domain) and the pGAD vectors; filter 7, yeast was co-transformed with the pGB-N and pGAD-GR vectors. ␤-galactosidase-positive colonies were detected only on filter 1.
We concluded that calreticulin may not be a nuclear resident protein and that, in vivo, calreticulin may not interact directly with the DNA binding domain of steroid receptors, but more likely the protein inhibits their activation of transcription indirectly, from the lumen of the ER. This is further supported by experiments using the yeast two-hybrid system. Under these conditions we also did not observe any interaction between the DNA binding domain of the glucocorticoid receptor (fused to the GAL4 transcriptional activation domain) and full-length calreticulin or its N domain (fused to the GAL4-DNA binding domain).
Calreticulin modulates steroid-sensitive gene expression (Burns et al., 1994a(Burns et al., , 1994bDedhar et al., 1994aDedhar et al., , 1994b. The protein in vitro binds to the conserved amino acid sequence KXFF(K/R)R found between the two Zn 2ϩ fingers in the DNA binding domain of steroid receptors and prevents their interaction with the DNA-responsive element (Burns et al., 1994b;Dedhar et al., 1994b). Overexpression of calreticulin in mouse L cells and Vero fibroblasts inhibited transcriptional activation by glucocorticoid receptor and androgen receptor in vivo, respectively (Burns et al., 1994b;Dedhar et al., 1994b). Furthermore, overexpression of calreticulin inhibits retinoic acid-induced differentiation of P19 cells (Dedhar et al., 1994) and decreases the induction of protein kinase C by retinoic acid (Desai et al., 1996). Calreticulin also binds, in vitro, to other transcription factors that contain the KXFF(K/R)R consensus amino acid sequence: the peroxisome proliferator-activated receptor-retinoid X receptor heterodimers, the hepatocyte nuclear factor-4, the chicken ovalbumin upstream promoter-transcription factor (Winrow et al., 1995), and the vitamin D 3 receptor (Wheeler et al., 1995;St-Arnaud et al., 1995). Surprisingly, in vivo calreticulin inhibits transcriptional activation by the vitamin D 3 receptor, retinoic acid receptor, androgen receptor, and glucocorticoid receptor (Dedhar et al., 1994;Burns et al., 1994;Wheeler et al., 1995;St-Arnaud et al., 1995;Desai et al., 1996) but not by the peroxisome proliferator-activated receptor-retinoid X receptor heterodimers (Winrow et al., 1995). All of these experiments were carried out using the ER form of calreticulin, further supporting the conclusion of the present work.
How can cells respond to signals from the lumen of the ER? For example, cells respond to ER stress by inducing novel gene expression (McMillan et al., 1994). Consequently, a signal must be transduced from the lumen of ER to the nucleus, to the plasma membrane, and to the cytoplasm to activate a specific response (adhesion, gene expression, or ion fluxes). Very little is known about "ER signaling" in mammalian cells. This pathway has been studied extensively in yeast (McMillan et al., 1994). An ER-resident protein kinase IRE-1/ERN-1 was identified in yeast whose kinase activity is essential for ER-nuclear signal transduction (Cox et al., 1993;Mori et al., 1993). However, there is no evidence that these kinases function in the mammalian cells (Cao et al., 1995). Cellular cholesterol homeostasis is controlled by ER-nuclear signaling via sterol-regulated proteolysis of ER membrane-bound transcription factors called sterol regulatory element-binding proteins . In sterol-deprived cells, a protease cleaves sterol regulatory element-binding protein, which is an integral ER membrane protein, and releases an N-terminal fragment of the protein that enters the nucleus and acts as a transcription factor that activates the genes for the low density lipoprotein receptor Wang et al., 1994). Sterol regulatory element-binding proteins are also cleaved by CPP32 protease during programmed cell death (apoptosis) (Wang et al., 1996). Recently, a new ER-nucleus pathway has been described involving NF-B (Pahl and Baeuerle, 1995;Pahl et al., 1996).
Accumulation of proteins in the lumen of ER induces NF-B DNA binding and B-dependent gene expression (Pahl and Baeuerle, 1995;Pahl et al., 1996). If calreticulin activates NF-B, the protein may affect several aspects of cell function via NF-B-dependent activation of expression of a large number of genes including interferons, cytokines, cell adhesion molecules, and growth factors (Baeuerle and Henkel, 1994). Similar to calreticulin, gene expression is also regulated by ERp61, an ER protein that belongs to a family of protein disulfide isomerase proteins (Johnson et al., 1992). In leukemia cells from patients with chronic myelogenous leukemia, ERp61 has been demonstrated to alter complex formation between nuclear proteins and regulatory regions of interferon-inducible genes (Johnson et al., 1992). It is not clear at present if calreticulin is involved in the activation of any of these pathways.
Ca 2ϩ may play an important role in many calreticulindependent cellular functions. Calreticulin is an ER Ca 2ϩ -binding protein, and overexpression of the protein results in increased Ca 2ϩ concentration in the thapsigargin-sensitive intracellular Ca 2ϩ stores (ER) (Bastianutto et al., 1995;Mery et al., 1996). The protein might also alter Ca 2ϩ concentrations not only in the lumen of ER but also in the cytoplasm (Camacho and Lechleiter, 1995). Depletion of Ca 2ϩ from the lumen of the ER (Greber and Gerace, 1995) or nuclear envelope (Stehno-Bittel et al., 1995), as well as modulation of the cytoplasmic Ca 2ϩ concentrations (Macaulay and Farbes, 1996), regulates transport of molecules into the nucleus. Therefore, calreticulindependent Ca 2ϩ fluctuations either in the lumenal ER/nuclear envelope or in the cytoplasm may also affect steroid receptor function and/or its nuclear compartmentalization.
Calreticulin, from the lumen of the ER, may modulate other cellular processes. Bastianutto et al. (1995) and Mery et al. (1996) showed that overexpression of calreticulin in the lumen of the ER results in inhibition of the store-operated Ca 2ϩ influx via plasma membrane Ca 2ϩ release-activated channel (CRAC) (Hoth and Penner (1992); reviewed in Bennett et al. (1995) and Berridge (1995)). Entry of Ca 2ϩ through the CRAC is increased when the internal ER stores are drained of their Ca 2ϩ content (Putney, 1986;Hoth and Penner, 1992). Calreticulin may modulate CRAC function from the lumen of the ER by a similar mechanism to the inhibition of steroid-sensitive gene expression. Calreticulin also contributes to the regulation of cell adhesiveness (Leung-Hagesteijn et al., 1994;Coppolino et al., 1995). It will be important to test the effects of the cytoplasmic versus the ER form of calreticulin on cell adhesiveness and Ca 2ϩ influx to establish if calreticulin may also affect these processes indirectly from the lumen of the ER.