Interaction of Calreticulin with Protein Disulfide Isomerase*

We report here that calreticulin interacts with protein disulfide isomerase (PDI). The PDI-calreticulin complex can be dissociated by Zn 2 (cid:49) -iminodiacetate-substituted Sepharose-agarose chromatography, suggesting that these interactions may be Zn 2 (cid:49) dependent. Direct interaction between calreticulin and PDI is also documented by calreticulin affinity chromatography. PDI was the only pancreatic microsomal protein retained on the cal- reticulin affinity column. Calreticulin and PDI were identified by their NH 2 -terminal amino acid sequence analysis, mobilities in SDS-polyacrylamide gel electrophoresis, binding of 45 Ca 2 (cid:49) , and their reactivity with specific antibodies. Using glutathione S -transferase-cal- reticulin fusion proteins, we show that PDI interacts strongly with the P-domain and only weakly with the N-domain of calreticulin. Expression of calreticulin domains and PDI as fusion proteins with GAL4 in the yeast two-hybrid system revealed that calreticulin interacted with PDI also under normal cellular conditions. Inter- action with PDI required only the NH 2 -terminal region of the N-domain (amino acid residues 1–83) and the P- domain (amino acid residues 150–240) of calreticulin. affinity.

The endoplasmic reticulum (ER) 1 plays an important role in control of a variety of different intracellular processes, including synthesis of membrane and secretory proteins, protein folding and modification, synthesis of membrane lipids, and regulation of Ca 2ϩ storage and release. The ER lumen contains a characteristic set of resident proteins that is involved in virtually every aspect of the ER function including Ca 2ϩ binding and storage. One of the major Ca 2ϩ binding proteins of the ER is calreticulin . The protein has two classes of Ca 2ϩ binding sites; a high capacity site (Ͼ25 mol/mol of protein) and a high affinity Ca 2ϩ binding site (K d ϭ ϳ10 M) (Ostwald and MacLennan, 1974;Baksh and Michalak, 1991). Calreticulin also binds other ions including Zn 2ϩ (K d ϭ 300 M, B max ϭ 14 mol of Zn 2ϩ /mol of protein) (Khanna et al., 1986). Khanna et al. (1986) showed by fluorescence spectroscopy that calreticulin undergoes Zn 2ϩ -dependent conformational changes and suggested that the structure of the protein becomes more compact upon binding of Zn 2ϩ . This property of calreticulin was utilized for a Zn 2ϩ -dependent purification of the protein by phenyl-Sepharose chromatography (Heilmann et al., 1993). Zn 2ϩ binding to calreticulin may, therefore, play an important role inducing the folding of the protein or its domains.
Although calreticulin was first identified as a Ca 2ϩ binding protein (Ostwald and MacLennan, 1974), recent reports indicate that the protein is multifunctional . For example, calreticulin modulates steroid-sensitive gene expression (Burns et al., 1994a(Burns et al., , 1994bDedhar et al., 1994) and affects adhesion properties of some cells (Rojiani et al., 1991;Leung-Hagesteijn et al., 1994). Furthermore, calreticulin affects intracellular Ca 2ϩ homeostasis (Liu et al., 1994;Bastianutto et al., 1995;Camacho and Lechleiter, 1995), replication of rubella virus RNA (Singh et al., 1994), and has a chaperone activity (Nigam et al., 1994;Nauseef et al., 1995;Wada et al., 1995). In cytolytic T lymphocytes, calreticulin is found in the lytic granules (Dupuis et al., 1993) where it may play a role in killing of target cells. In human neutrophils, calreticulin is concentrated in the area around the phagocytosed particle (Stendahl et al., 1994), suggesting that the protein may contribute to the process of phagocytosis.
Based on the amino acid sequence of the protein, calreticulin can be divided into distinct structural/functional domains . The NH 2 -terminal half of the protein (designated N-domain) is proposed to be a globular domain containing eight anti-parallel ␤-strands. This region leads into a proline-rich sequence (P-domain) followed by the carboxylterminal quarter of the protein (C-domain). The C-domain of calreticulin is acidic and binds Ca 2ϩ with high capacity and low affinity, whereas the P-domain binds 1 mol of Ca 2ϩ with high affinity . The N-domain of the protein does not bind Ca 2ϩ , but it may interact with the DNA binding domain of steroid receptors (Burns et al., 1994b).
In our earlier studies, we have identified calreticulin as one of the major Ca 2ϩ binding proteins of the ER membrane and localized the Ca 2ϩ binding sites to different domains in the protein . In this paper, we report that calreticulin interacts with PDI. Interactions between calreticulin and PDI are Zn 2ϩ dependent and involve the P-domain and the NH 2 -terminal region of the N-domain of calreticulin. Importantly, interaction between PDI and calreticulin reduced Ca 2ϩ binding to the high affinity site (in the P-domain) of calreticulin and inhibited the PDI activity as determined by its ability to refold scrambled RNase A.

EXPERIMENTAL PROCEDURES
Materials-Mops, sodium deoxycholate, and Triton X-100 were purchased from Sigma. 45 CaCl 2 (2.52 Ci/mmol) was obtained from Amersham. Peroxidase-conjugated rabbit anti-goat IgG and goat anti-rabbit IgG were from Bio-Rad and Bio/Can, respectively. The ECL detection kit for Western blotting was obtained from Amersham. The Mono Q fast protein liquid chromatography column, metal chelating Sepharose, and CNBr-activated Sepharose 4B were from Pharmacia Biotech Inc. Isopropyl ␤-D-thiogalactopyranoside and 5-bromo-4-chloro-3-indolyl-␤-Dgalactopyranoside were obtained from Boehringer Mannheim. Vent polymerase was from New England Biolabs. Glutathione-agarose was obtained from Molecular Probes, Inc. p5A5 plasmid containing cDNA encoding human liver PDI was obtained from ATCC (batch 89-06). Nitrocellulose and Immobilon (polyvinylidene difluoride) membrane filters were from Scheicher and Schuell and Millipore, respectively. Matchmaker two-hybrid system was purchased from Clontech. SDS-PAGE reagents and molecular weight markers were from Bio-Rad. Fresh canine pancreas and hearts were obtained from the Surgical Medical Research Institute at the University of Alberta. All chemicals were of the highest grade available.
Purification of Proteins and Membrane Vesicles-Calreticulin and canine cardiac calsequestrin were purified by the ammonium sulfate precipitation procedures as described earlier (Milner et al., 1991;Baksh et al., 1992). Recombinant full-length calreticulin, the domains of calreticulin, and recombinant GST were expressed in Escherichia coli and purified Baksh et al., 1992). For details on construction of the cDNA clones, E. coli expression, and purification of the recombinant proteins, see Baksh and Michalak (1991).
PDI was isolated in the presence of protease inhibitors by the procedure as outlined by Lambert and Freedman (1983). PDI was precipitated with 85% ammonium sulfate followed by anion exchange chromatography on carboxymethylcellulose, precipitation with 100% ammonium sulfate, and cation exchange chromatography on fast protein liquid chromatography Mono Q column.
Zn 2ϩ Chelating Sepharose Chromatography-Zn 2ϩ -iminodiacetatesubstituted Sepharose (IDA) chromatography of purified proteins was carried out essentially as described by Porath and Olin (1983) and Picello et al. (1992). The Zn 2ϩ -IDA column (8 ϫ 1 cm; flow rate ϭ 0.5 ml/min) was first washed with a solution containing 50 mM NaH 2 PO 4 , pH 7.0, 5 mM EDTA, and 100 mM NaCl followed by saturation of the column with 10 mM ZnCl 2 . The excess of Zn 2ϩ was removed by washing the column with a solution containing 50 mM NaH 2 PO 4 , pH 7.0, and 100 mM NaCl. Purified calreticulin (2-8 mg of protein) was applied to a column and eluted with a linear gradient of 0 -50 mM imidazole in 50 mM NaH 2 PO 4 , pH 7.0. 3-ml fractions were collected, and the proteins were analyzed on a 12.5% SDS-PAGE gel. The peak containing calreticulin was pooled, dialyzed, and analyzed for immunoreactivity with specific antibodies.
Calreticulin Affinity Chromatography-A calreticulin affinity column was synthesized by coupling recombinant calreticulin to CNBractivated Sepharose 4B (Pharmacia) as recommended by the manufacturer. 2 mg of calreticulin were coupled to 1 ml of activated CNBr-Sepharose 4B. Efficiency of coupling was greater than 95%. A precolumn was prepared by incubation of CNBr-activated Sepharose 4B with 1 M ethanolamine and 100 mM NaHCO 3 , pH 9.0. The calreticulin affinity column was subsequently used to purify calreticulin binding proteins from pancreatic microsomes.
Pancreatic microsomes (a rich source of the ER lumenal proteins) were isolated by the procedure of Walter and Blobel (1983). Prior to calreticulin affinity column chromatography, pancreatic microsomes containing ϳ20 mg of protein were thawed at 37°C and then cooled on ice, followed by centrifugation at 100,000 ϫ g for 30 min using a Ti70.1 rotor. The membrane pellet was resuspended by Dounce homogenization in 800 l of a solution containing 10 mM Mops, pH 7.0, 100 mM KCl, 2 mM MgCl 2 , 0.5 mM EGTA. The microsomes were then solubilized with 1% Triton X-100 and 0.5% sodium deoxycholate by incubation on ice for 30 min. Solubilized microsomes were centrifuged at 100,000 ϫ g for 30 min, and the soluble extract was diluted 10 fold with a solution containing 10 mM Mops, pH 7.0, 100 mM KCl, 2 mM MgCl 2 , 0.5 mM EGTA, 0.1% Triton X-100, and 0.05% sodium deoxycholate. Ca 2ϩ , Zn 2ϩ , or EGTA were added to give a final concentration of 2, 2, or 5 mM, respectively. The extracts were first applied onto a precolumn to remove any proteins that nonspecifically associate with the matrix. The flowthrough (ϳ2 mg/ml protein) was then directly applied to the calreticulin affinity column. The column was washed with 15 ml of a solution containing 10 mM Mops, pH 7.0, 100 mM KCl, 2 mM MgCl 2 , 0.5 mM EGTA, 0.1% Triton X-100, and 0.05% sodium deoxycholate. Bound proteins were eluted with a 25-ml linear gradient of 100 -750 mM KCl at a flow rate of 0.2 ml/min. Bound proteins were concentrated using an Amicon concentrator, separated by SDS-PAGE, and electrophoretically transferred to Immobilon membrane for NH 2 -terminal sequencing or to nitrocellulose membrane for immunoblotting.
Interactions of PDI with Domains of Calreticulin-The domains of calreticulin and recombinant GST were expressed in E. coli and purified . For these experiments, we have expressed, as GST-fusion proteins, three domains of calreticulin: N-domain (amino acid residues 1-182), the P-domain (amino acid residues 139 -273), and the C-domain (amino acid residues 270 -401) . Prior to incubation with PDI, the purified GST or GST fusion proteins were bound to the Glutathione-agarose beads (200 l of 10% solution). The beads were extensively washed with phosphatebuffered saline followed by incubation for 60 min at 4°C with 70 g of purified PDI. The mixture was spun down for 3 min at 2,000 ϫ g in an Eppendorf centrifuge. The supernatant containing the unbound material was collected and analyzed by SDS-PAGE. Beads were washed with phosphate-buffered saline, and bound proteins were released by incubation with 5 mM glutathione for 60 min at 4°C. The beads were spun down for 3 min at 2,000 ϫ g, and the released proteins were subjected to SDS-PAGE.
Yeast Two-hybrid System-To study interaction between PDI and calreticulin domains, the yeast two-hybrid system was employed (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 N1-domain (amino acid residues 1-86), the N2-domain (amino acid residues 83-174), and the P1-domain (amino acid residues 150 -240) of calreticulin and the GAL4 activation domain fusion protein with PDI.
The nucleotide sequence of primer A contained a 5Ј-EcoRI restriction site and corresponded to the nucleotides 112-131 (underlined) of the cDNA encoding the first 6 amino acids of the mature calreticulin (Fliegel et al., 1989). Primer A was used as a 5Ј (forward) primer for the synthesis of the N1-domain. Primer B contained a 5Ј-SalI restriction site and corresponded to the nucleotide sequence 352-372 (underlined) encoding residues 80 -86 of the mature rabbit calreticulin (Fliegel et al., 1989). Primer B was used as a 3Ј (reverse) primer for the synthesis of the N1-domain. Primer C with 5Ј-EcoRI restriction site corresponded to the nucleotide sequence 616 -636 (underlined) encoding residues 150 -157 of the mature rabbit calreticulin. This primer was used as the 5Ј (forward) primer for the PCR-driven synthesis of the P1-domain. Primer D with a 5Ј-SalI restriction site was used as a 3Ј (reverse) primer for the synthesis of the P1-domain. Primer D corresponded to the nucleotide sequence 811-831 (underlined) encoding residues 233-240 of the mature rabbit calreticulin (Fliegel et al., 1989). The nucleotide sequence of primer E contained a 5Ј-EcoRI restriction site and corresponded to the nucleotides 346 -366 (underlined) of the cDNA encoding residues 77-83 of the mature calreticulin (Fliegel et al., 1989). Primer E was used as a 5Ј (forward) primer for the synthesis of the N2-domain. Primer F contained a 5Ј-SalI restriction site and corresponded the nucleotide sequence 633-653 (underlined) encoding residues 168 -174 of the mature rabbit calreticulin (Fliegel et al., 1989). Primer F was used as a 3Ј (reverse) primer for the synthesis of the N2-domain. Linearized pcDX-CRT plasmid encoding rabbit skeletal muscle calreticulin cDNA (GenBank accession no. J05138) (Fliegel et al., 1989) was used as the template. PCR reactions were carried out in a 100-l reaction using 20 ng of template, 0.8 M of the appropriate primers, 200 M of each dNTP in 10 mM KCl, 10 mM (NH 4 ) 2 SO 4 , 20 mM Tris-HCl, pH 8.8, 2 mM MgCl 2 , and 0.1% Triton X-100. Each reaction was overlaid with mineral oil and heated for 3 min at 94°C. The reaction was initiated by the addition of 1 unit of Vent polymerase. Amplification was carried out for 25 cycles with each cycle consisting of 1 min at 94°C, 1 min at 57°C, and 2 min at 72°C. This was followed by one cycle for 10 min at 72°C. The PCR products were purified using Millipore 30,000 MW cut off spin filters, digested with EcoRI and SalI restriction endonuleases, phenol:chloroform extracted, and precipitated with ethanol. The PCR products were then ligated into EcoRI and SalI sites of the phosphatase-treated pGBT9 to generate pGB-N1, pGB-N2, and pGB-P1 vectors encoding GAL4 DNA binding domain and the N1-domain fusion protein, the N2-domain fusion protein or GAL4 DNA-binding domain and the P1-domain fusion protein, respectively.
To generate the GAL4-activating domain fusion protein with PDI, cDNA encoding the full-length mature human PDI was generated by PCR amplification of p5A5 vector containing cDNA encoding full-length PDI (GenBank accession no. J02783) (Cheng et al., 1987). The following primers were used: primer G, 5Ј-ATATGAATTCGACGCCCCCGAG-GAG-3Ј and primer H, 5Ј-ATATGGATCCGGGTCTGGCTTTGCGTA-3Ј.
Primer G contained an EcoRI restriction site and was used as a 5Ј (forward) primer. Primer G corresponded to the nucleotide sequence 115-129 (underlined) and encoded the first 5 amino acids of the fulllength mature human PDI (Cheng et al., 1987). Primer H contained a BamHI restriction site and corresponded to the nucleotide sequence 1591-1607 (underlined) immediately behind the stop codon of the PDI clone. This primer was used as a 3Ј (reverse) primer. PCR reactions were carried out as described above for calreticulin constructs. The PCR products were then ligated into EcoRI and BamHI sites of the phosphatase-treated pGAD424 to generate pGAD-PDI vector. Sequences of every construct were confirmed by nucleotide sequencing.
To detect interactions between the calreticulin domains and PDI, yeast strain SFY526 was transformed with pGB-N1, pGB-N2, pGB-P1, or pGAD-PDI or was cotransformed with either pGB-N1, pGB-N2, or pGB-P1 with pGAD-PDI using 200 ng of each construct following the protocol recommended by the manufacturer. As negative controls the pGB-N1, pGB-N2, and pGB-P1 were each cotransformed with the pGAD424 vector, and pGAD-PDI was cotransformed with the pGBT9 vector. The colony lift filter assay for ␤-galactosidase activity was performed according to the manufacturer's protocol. Color development was monitored for 24 h.
PDI Assay-Activity of PDI was followed by PDI ability to refold scrambled RNase A, which in turn will regain its ability to cleave highly polymerized yeast RNA (Lambert and Freedman, 1983). PDI (2.0 g/ ml) was dissolved in 100 mM sodium phosphate, pH 7.5, 2 mM glutathione, and 0.2 mM reduced glutathione. The reaction was started by the addition of 200 g of scrambled RNase A (Hillson et al., 1984). PDI activity was determined in the presence of recombinant calreticulin, purified N-, P-, and C-domains of calreticulin, and in the presence of cardiac calsequestrin. For the assay, the proteins were used in 1:1 molar ratio. The A 260 /A 280 ratio over a period of 60 min was measured as indicative of the extent of cleavage of yeast RNA by refolded RNase A.
Antibodies-Goat anti-rabbit calreticulin antibody was described by Milner et al. (1991). Anti-bovine PDI was made in rabbit by injection of 1 mg of purified bovine liver PDI in 1 ml of either complete or incomplete Freund's adjuvant. The antigen emulsified in complete Freund's adjuvant was injected subcutaneously in four sites at the back of New Zealand White male rabbits, followed by two Booster injections at 2-3-week intervals. 2 weeks after the third injection, the sera were immunoreactive with PDI at a dilution of 1:300. Preimmune sera at a dilution of 1:200 did not result in any detectable immunoreaction with the antigen.
Miscellaneous-All recombinant techniques were conducted according to standard protocols (Ausubel et al., 1989). Quantitative Ca 2ϩ binding to calreticulin in the presence of PDI was conducted by equilibrium dialysis according to Baksh and Michalak (1991). 45 Ca 2ϩ binding to proteins on nitrocellulose membranes was carried out by the method of Maruyama et al. (1984). Protein was determined by the method of Lowry et al. (1951) or Bradford (1976). NH 2 -terminal sequence analysis of all proteins was carried out on protein electroblotted to Immobilon (polyvinylidene difluoride) membrane (Matsudaira, 1987).

Zn 2ϩ Chelating Chromatography of Calreticulin-Calreticu-
lin binds approximately 14 mol of Zn 2ϩ /mol of protein with a K d of ϳ300 M (Khanna et al., 1986). We have investigated interactions of calreticulin with Zn 2ϩ using Zn 2ϩ chelating chromatography in IDA-agarose (Porath and Olin, 1983;Hemdan et al., 1989;Picello et al., 1992). Fig. 1 shows that recombinant calreticulin bound to the Zn 2ϩ -IDA-agarose and was eluted with an increasing gradient of imidazole concentration. Recombinant calreticulin eluted at approximately 22 mM imidazole (Fig. 1), confirming that the protein interacts with Zn 2ϩ under native conditions. We were surprised, however, to find that the native calreticulin, isolated from canine pancreas (Baksh et al., 1992) failed to bind to the Zn 2ϩ -IDA-agarose and was found in the column "flow-through" (Fig. 2). Fig. 2 also shows that canine pancreatic calreticulin migrated in Laemmli SDS-PAGE as a 56-kDa protein. Variability in electrophoretic mobility of different calreticulins has been reported by others . The identity of calreticulin in the column flow-through was confirmed using specific goat anti-calreticulin antibodies (Fig. 3A,  lanes 1-3) and by the NH 2 -terminal amino acid sequence analysis. The goat anti-calreticulin antibody recognized the P-and C-domain of the protein (lanes 7 and 8). The NH 2 -terminal amino acid sequence of the protein found in the flow-through fraction was EPAIYK. This sequence is identical to the NH 2 -FIG. 1. Zn 2ϩ -dependent chromatography of calreticulin. Recombinant calreticulin was loaded onto Zn 2ϩ -IDA-agarose column, equilibrated with 50 mM NaH 2 PO 4 , pH 7.0, and 100 mM NaCl (fractions 1-11) followed by elution with a linear (0 -50 mM) imidazole gradient (arrow). 1-ml fractions were collected and assayed for protein content (Bradford, 1976).
terminal amino acid sequence of calreticulin . These results suggested that another protein may have masked the site(s) in calreticulin involved in interaction with Zn 2ϩ -IDA-agarose. When an imidazole gradient was applied to the column, a protein with a molecular weight of approximately 58,000 eluted at 33 mM imidazole (Fig. 2, lanes 5 and 6). The NH 2 -terminal amino acid sequence of this protein band was APEED. This sequence corresponds exactly to the NH 2 -terminal amino acid sequence of PDI (Edman et al., 1985). Identity of this protein band to PDI was further confirmed using specific anti-PDI antibodies (Fig. 3B, lane 4). Fig. 3C shows that calreticulin found in the Zn 2ϩ -IDA-agarose flow-through bound large quantities of 45 Ca 2ϩ (lane 1). PDI also bound 45 Ca 2ϩ but at much lower quantities compared to calreticulin (Fig. 3C,  compare lanes 1 and 2). We concluded that calreticulin and PDI may interact, and this interaction may prevent calreticulin binding to the Zn 2ϩ -IDA-agarose.
PDI Binds to a Calreticulin Affinity Column-To test for the direct interaction between calreticulin and PDI, we have synthesized a calreticulin affinity column using the recombinant protein. Pancreatic ER vesicles (rich source of PDI) (Naiva and Lennarz, 1992) were solubilized with detergent, and the soluble extract, containing PDI, was applied onto the calreticulin affinity column (Fig. 4, lane 1). The column was extensively washed, and bound proteins were eluted with a gradient of increasing concentration of KCl. Eluted fraction contained a single protein band of M r approximately 58,000 (Fig. 4, lane 2, silver-stained gel). The NH 2 -terminal amino acid analysis of this protein band was APEED, which corresponds to the amino acid sequence of PDI (Edman et al., 1985). The identity of the eluted protein as PDI was further confirmed by its reactivity with anti-PDI antibodies (data not shown). Interestingly, the association between calreticulin and PDI was observed only when the affinity chromatography was carried out in the presence of EGTA but not in the presence of Ca 2ϩ and Zn 2ϩ . These experiments further confirmed that calreticulin and PDI, two ER resident proteins, interact and that this interaction may be ion dependent.
PDI Interacts in Vitro with the N-domain and P-domain of Calreticulin-We were interested in identification of the specific region of calreticulin involved in the interactions with PDI. Three domains of calreticulin were used in this study: the GST-N-domain (encompassing the NH 2 -terminal 182 amino acids of the protein), the GST-P-domain (residues 139 -273), and the GST-C-domain (residues 270 -401) (Fig. 5). GST was FIG. 2. Zn 2ϩ chelating Sepharose chromatography of native calreticulin. Tissue-purified pancreatic calreticulin was separated on a Zn 2ϩ -IDA-agarose column, equilibrated with 50 mM NaH 2 PO 4 , pH 7.0, and 100 mM NaCl followed by washing with the same buffer (flowthrough) and elution with a linear imidazole gradient (lanes 5 and 6) as described under "Experimental Procedures." 1-ml fractions were collected and analyzed by SDS-PAGE followed by staining with Coomassie Blue. Lane 1, purified calreticulin; lane 2, purified PDI; lanes 3 and 4, tissue-purified calreticulin (additional protein bands in lane 4 are degradation products of the tissue-purified calreticulin); lane 5 and 6, fraction bound to the column and eluted with imidazole gradient (these fractions contain PDI); S, low molecular weight Bio-Rad standards; Flow-through, fractions 1-6 of the column wash (unbound material containing calreticulin).
FIG. 3. SDS-PAGE of Zn 2ϩ -IDA-agarose purified calreticulin and PDI. Native pancreatic calreticulin was fractionated on Zn 2ϩ -IDA chromatography as described under "Experimental Procedures." Unbound material (see Fig. 2, flow-through fractions) and bound and eluted fractions (see Fig. 2, lanes 5 and 6)  Pancreatic microsomes were isolated and solubilized with 1% Triton X-100 and 0.5% sodium deoxycholate. Solubilized microsomes were centrifuged at 100,000 ϫ g for 30 min, and the soluble extract was applied to the calreticulin affinity column. The column was washed with 15 ml of a solution containing 10 mM Mops, pH 7.0, 100 mM KCl, 2 mM MgCl 2 , 0.5 mM EGTA, 0.1% Triton X-100, and 0.05% sodium deoxycholate, and bound proteins were eluted with 300 mM KCl. Proteins were separated by SDS-PAGE and stained with silver. Lane 1, detergent-solubilized pancreatic microsomes (6 g of protein/lane); lane 2, protein bound and eluted from the calreticulin affinity column (1 g of protein/lane).
used as a control. GST or the GST-fusion proteins were first bound to the Glutathione-agarose beads followed by the addition of purified PDI and centrifugation to collect the unbound material. The pellet was incubated with 5 mM glutathione to elute proteins bound to the GST affinity beads. We predicted that if PDI interacts with calreticulin domain(s), the two proteins should elute from the beads together as a protein complex. First, we established that PDI did not interact with GST. Fig. 5 shows that PDI did not bind to the GST-glutathioneagarose and was found in the unbound fraction (GST, lane U), whereas the recombinant GST was eluted with 5 mM glutathione (GST, lane B). We conclude that PDI did not interact with the recombinant GST. When incubated with the GST-N-domain affinity beads, the majority of PDI eluted in the unbound fraction ( Fig. 5; GST-N, lane U). The remaining PDI and GST-N-domain were eluted with 5 mM glutathione (GST-N, lane B), suggesting that there were limited interactions between GST-N-domain of calreticulin and PDI. Based on densitometry scanning of protein bands, PDI and GST-N-domain of calreticulin interacted with a 6:1 ratio (Fig. 5, GST-N, lane B). When GST-P-domain affinity beads were incubated with PDI, almost all of the PDI and the GST-P-domain eluted together from the column with 5 mM glutathione ( Fig. 5; GST-P, lane U), indicating that the two proteins interacted. Based on densitometry of the Coomassie Blue-stained intensities of these two protein bands, we conclude that they interacted with 1:1 ratio. Fig. 5 shows that the GST-C-domain of calreticulin did not interact with PDI. PDI eluted alone in the unbound fraction (GST-C, lane U), followed by elution of the GST-C-domain with 5 mM glutathione (GST-C, lane B). We conclude that PDI interacts tightly with the P-domain of calreticulin and to a lesser extent with the N-domain of the protein.
In Vivo Interactions between PDI and Domains of Calreticulin-We employed the yeast two-hybrid system (Fields and Song, 1989;Chien et al., 1991) to test if PDI and calreticulin domains interact under normal cellular conditions. For these experiments, instead of the full-length N-domain, two regions of the N-domain of calreticulin were fused to the GAL4-DNA binding domain: the N1-domain (amino acid residues 1-86) and the N2-domain (amino acid residues 83-174). The aminoterminal part of the P-domain (designated the P1-domain; amino acid residues 150 -240) was also fused to the GAL4-DNA binding domain. This region of the P-domain contains the three PXXIXDPDAXKPEDWDE amino acid repeats conserved in all calreticulins and calnexins . Fig. 6 (filters 2  and 3) shows that the N1-and P1-domain fusion proteins did not activate reporter gene expression on its own when coexpressed with the GAL4 transcriptional activation domain. Next, PDI was fused to the GAL4 transcriptional activation domain. This fusion protein also did not activate the reporter gene when coexpressed with the GAL4-DNA binding domain (Fig. 6, filter 1). When jointly expressed as GAL4 fusion proteins, the N1-domain with PDI (Fig. 6, filter 4) induced expression of ␤-galactosides, indicating that the two proteins interacted under these in vivo conditions. In contrast, the N2domain when coexpressed with PDI did not induce expression of ␤-galactosidase (data not shown), suggesting that the NH 2terminal part of the N-domain interacted with PDI. Next, we investigated interactions between the P1-domain of calreticulin and PDI GAL4 fusion proteins. Fig. 6 ( filter 5) shows that in the presence of the P1-domain and PDI, expression of ␤-galactoside was induced, indicating that this region of calreticulin interacted with PDI. These results confirmed the in vitro studies and showed that calreticulin interacts with PDI and that these interactions are mediated by the N-and P-domain of calreticulin.
PDI Modulates Ca 2ϩ Binding to Calreticulin, and Calreticulin Effects PDI Activity-We then investigated if interaction between calreticulin and PDI has any effect in the function of FIG. 5. PDI binding to calreticulin domains. GST, GST-N-domain, GST-P-domain, and GST-C-domain of calreticulin were expressed in E. coli and purified as described by Baksh and Michalak (1991). The GST and GST fusion proteins were bound to the Glutathione-agarose beads, washed with phosphate-buffered saline, and then incubated with purified PDI as described under "Experimental Procedures." The beads were spun down to obtain the unbound material (lanes U, ϳ5 g of protein/lane) followed by incubation with 5 mM glutathione (3-fold dilution) to obtain the bound fractions (lanes B, ϳ2 g of protein/lane). Lanes C, purified control fusion proteins ( FIG. 6. The N1-and P1-domains of calreticulin interact with PDI 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 cotransformed with the pGAD-PDI with the pGBT9 vectors; filter 2, yeast was cotransformed with the pGB-N1 (encoding the N1-domain) and the pGAD424 vectors; filter 3, yeast was cotransformed with the pGB-P1 (encoding the P1-domain) and the pGAD424 vectors; filter 4, yeast was cotransformed with the pGB-N1 and the pGAD-PDI vectors; filter 5, yeast was cotransformed with the pGB-P1 and the pGAD-PDI vectors. these proteins. Recombinant or native calreticulin has two Ca 2ϩ binding sites: a high capacity site (Ͼ25 mol/mol of protein) and a high affinity Ca 2ϩ binding site (K d ϳ ϭ 10 M) (Ostwald and MacLennan, 1974;Baksh and Michalak, 1991). In the presence of PDI, Ca 2ϩ binding to the high affinity site in calreticulin was dramatically reduced. This is not surprising, as the high affinity Ca 2ϩ binding site is localized to the Pdomain in calreticulin , a site of interaction with PDI (see Fig. 5). PDI had no effect on the low affinity and high capacity Ca 2ϩ binding (located in the Cdomain) to the protein (data not shown).
To test the ability of calreticulin to affect PDI function, PDI activity was measured as the ability of the protein to assist in refolding of scrambled RNase A (Naiva and Lennarz, 1992;Freedman et al., 1994). Scrambled RNase A was incubated in the presence and absence of PDI and PDI complexed with calreticulin, calreticulin domains, or calsequestrin. In the presence of PDI, the recovery of RNase activity was significantly enhanced (Fig. 7). This effect of PDI was not observed in the presence of calreticulin (Fig. 7). Cardiac calsequestrin, a Ca 2ϩ binding protein of similar physicochemical properties to calreticulin, had no affect on PDI activity (Fig. 7). Calreticulin alone had no effect on the recovery of the scrambled RNase activity (data not shown). We have tested calreticulin domains for their ability to inhibit PDI activity as measured by the ability of PDI to assist in refolding of scrambled RNase A. The N-domain inhibited PDI activity similar to that observed for the fulllength recombinant calreticulin, suggesting that the N-domain may be binding near the active site of PDI (Fig. 7). Surprisingly, the P-domain of calreticulin had no effect on PDI activity (Fig. 7). DISCUSSION In this work, we have established that calreticulin interacts with PDI. PDI interacts with the P-domain and with the Ndomain of calreticulin and reduces the high affinity Ca 2ϩ binding to the protein. Calreticulin binding to PDI inhibits PDI activity as measured by refolding of the scrambled RNase A.
The PDI-calreticulin complex can be broken up by the Zn 2ϩ -IDA chromatography, suggesting that these interactions may be Zn 2ϩ dependent. In the presence of PDI, calreticulin does not bind to the Zn 2ϩ -IDA-agarose. PDI interacts with calreticulin at or near its Zn 2ϩ binding site(s) or induces conformational changes in the protein, preventing interaction of calreticulin with Zn 2ϩ -IDA-agarose.
Calreticulin and PDI belong to a family of the resident ER proteins. Both proteins terminate with the KDEL ER retention signal (Edman et al., 1985;Michalak et al., 1992) and bind Ca 2ϩ with a large capacity Lebeche et al., 1994). Several other ER lumenal proteins, including ERp72, ER calcistorin/PDI (calstorin), reticulocalbin, Grp94 (endoplasmin, ERp99), and BiP (Grp78), may also play an important role in Ca 2ϩ binding and storage within the ER membrane (Koch, 1987(Koch, , 1990Mazzarella et al., 1990;Mazzarella and Green, 1987;Milner et al., 1992;Ozawa and Muramatsu, 1993;Lucero et al., 1994;Racchetti et al., 1994). Many of these proteins are involved in the posttranslation modification/folding of newly synthesized secretory proteins. Their retention in the lumen of ER is dependent on the KDEL carboxyl-terminal amino acid sequence (Pelham, 1989). Koch (1987) proposed the name reticuloplasmin for the ER lumenal proteins to emphasize their specific location and to consider their possible role in ER structure and function including storage of Ca 2ϩ . Interactions between calreticulin and PDI reported here support this hypothesis and show that these two reticuloplasmins interact. Importantly, these interactions may modulate function of these ER lumenal proteins.
Using GST-fusion protein affinity chromatography we have determined that PDI interacts with the P-domain and to a lesser extent with the N-domain of calreticulin. Furthermore, expression of calreticulin domains and PDI as fusion proteins with GAL4 in the yeast two-hybrid system revealed that the Nand P-domain of calreticulin interact with PDI also under normal cellular conditions. The interaction with PDI requires only the amino-terminal region of the N-domain (amino acid residues 1-83) of calreticulin. This is one of the most conserved regions of the amino acid sequence of calreticulins . The N-domain is also the region of calreticulin responsible for the protein's interaction with the DNA binding domain of steroid receptors (Burns et al., 1994b), Zn 2ϩ binding properties (Baksh et al., 1995), rubella virus RNA binding (Singh et al., 1994), and interaction with the ␣-subunit of integrin (Leung-Hagesteijn et al., 1994). Importantly, the N-domain of calreticulin inhibits PDI activity as measured by its ability to refold scrambled RNase A, suggesting that it may interact at or near the active site(s) of PDI.
Using the yeast two-hybrid system, we have also confirmed interaction between PDI and the P1-domain of calreticulin (amino acid residues 150 -240). This is one of the most interesting regions of the P-domain of calreticulin. The high affinity, low capacity Ca 2ϩ binding site is localized in the P-domain of calreticulin . This part of the Pdomain is highly charged and contains a set of repeats of the amino acid sequence PXXIXDPDAXKPEDWDE . A similar amino acid sequence is found in calnexin, an integral ER membrane chaperone (Bergeron et al., 1994) and in calmegin, a Ca 2ϩ binding protein specifically expressed during male meiotic germ cell development (Watanabe et al., 1994). Similar to calreticulin, a recombinant peptide, encompassing amino acids 254 -334 of calnexin (P-domain) and containing the amino acid sequence PXXIXDPDAXKPEDWDE repeats was sufficient to bind Ca 2ϩ (Tjoelker et al., 1994). It is not surprising, therefore, that PDI inhibits binding of Ca 2ϩ to the high affinity site in calreticulin. No significant effects of PDI on FIG. 7. Calreticulin effects PDI activity. Recovery of ribonuclease activity of the scrambled RNase A was measured in the absence (RNase A alone) and presence of PDI, PDI and calreticulin (PDI ϩ CRT), calreticulin domains (PDI ϩ N, PDI ϩ P, PDI ϩ C), and canine cardiac calsequestrin (CCS) as described under "Experimental Procedures." The A 260 /A 280 ratio was measured to determine the extent of cleavage of yeast RNA by refolded RNase A. Each value represents the mean Ϯ S.D. (n ϭ 3). the high capacity Ca 2ϩ binding site to calreticulin was observed. Hence, binding of PDI to calreticulin influences a function of the central portion of the molecule. It is conceivable that both calnexin and calmegin may also interact with PDI, and their Ca 2ϩ binding properties may be modified by these interactions.
Interactions between calreticulin and PDI are supported by recent studies by Nigam et al. (1994). While studying a set of ER chaperones, these authors isolated a complex of ER resident proteins that bind to denatured protein columns in an ATP and Ca 2ϩ -dependent manner. These proteins were identified as BiP (Grp78), Grp94, calreticulin, PDI, ERp72, p50, and 46-kDa protein. Direct interaction between calreticulin and denatured proteins was not reported. It is possible that calreticulin binds to the denatured proteins via interaction with PDI as described here. This hypothesis is supported by the finding that PDI interacts with calreticulin as determined by calreticulin affinity chromatography under the conditions used in this study. The interaction between PDI and calreticulin is ion dependent.
PDI is an abundant protein approaching millimolar levels in the lumen of ER (Naiva and Lennarz, 1992;Freedman et al., 1994). The protein catalyzes the in vitro folding of a variety of proteins by assisting in isomerization of intramolecular disulfide bridges (Naiva and Lennarz, 1992;Freedman et al., 1994). Calreticulin and PDI have both been proposed to be multifunctional proteins capable of interaction with a variety of proteins Naiva and Lennarz, 1992;Freedman, 1994;Burns et al., 1994a). In addition to their ion binding and protein binding activities, both proteins (or their homologues) have been implicated to modulate gene expression (Johnson et al., 1992;Burns et al., 1994b;Dedhar et al., 1994), were identified as autoimmune antigens (McCauliffe et al., 1990;Yokoi et al., 1993), and were found complexed with the intracellular lipid droplets (Ghosal et al., 1994). Importantly, interaction between calreticulin and PDI leads to modulation of their functions. In the presence of PDI, calreticulin does not bind Ca 2ϩ with high affinity, whereas PDI shows significantly reduced ability to refold scrambled RNase. It is tempting to speculate that calreticulin/PDI while complexed may perform additional functions likely involved in posttranslational modification of the newly synthesized proteins.