Molecular Cloning, Expression, Functional Characterization, Chromosomal Localization, and Gene Structure of Junctate, a Novel Integral Calcium Binding Protein of Sarco(endo)plasmic Reticulum Membrane*

Screening a cDNA library from human skeletal muscle and cardiac muscle with a cDNA probe derived from junctin led to the isolation of two groups of cDNA clones. The first group displayed a deduced amino acid sequence that is 84% identical to that of dog heart junctin, whereas the second group had a single open reading frame that encoded a polypeptide with a predicted mass of 33 kDa, whose first 78 NH2-terminal residues are identical to junctin whereas its COOH terminus domain is identical to aspartyl β-hydroxylase, a member of the α-ketoglutarate-dependent dioxygenase family. We named the latter amino acid sequence junctate. Northern blot analysis indicates that junctate is expressed in a variety of human tissues including heart, pancreas, brain, lung, liver, kidney, and skeletal muscle. Fluorescence in situ hybridization analysis revealed that the genetic loci of junctin and junctate map to the same cytogenetic band on human chromosome 8. Analysis of intron/exon boundaries of the genomic BAC clones demonstrate that junctin, junctate, and aspartyl β-hydroxylase result from alternative splicing of the same gene. The predicted lumenal portion of junctate is enriched in negatively charged residues and is able to bind calcium. Scatchard analysis of equilibrium 45Ca2+ binding in the presence of a physiological concentration of KCl demonstrate that junctate binds 21.0 mol of Ca2+/mol protein with a kD of 217 ± 20 μm (n = 5). Tagging recombinant junctate with green fluorescent protein and expressing the chimeric polypeptide in COS-7-transfected cells indicates that junctate is located in endoplasmic reticulum membranes and that its presence increases the peak amplitude and transient calcium released by activation of surface membrane receptors coupled to InsP3 receptor activation. Our study shows that alternative splicing of the same gene generates the following functionally distinct proteins: an enzyme (aspartyl β-hydroxylase), a structural protein of SR (junctin), and a membrane-bound calcium binding protein (junctate).

The sarcoplasmic reticulum (SR) 1 is an intracellular membrane compartment that controls the intracellular Ca 2ϩ concentration thereby playing an important role in the excitationcontraction coupling mechanism (for review see Refs. [1][2][3]. The anatomical site of excitation-contraction coupling is the triad, a unique intracellular synapsis that is formed by the association of the following membrane compartments: transverse tubules, which are an invagination of the sarcolemma, and the SR terminal cisternae (3). The portion of terminal cisternae facing the transverse tubules is referred to as junctional face membrane SR (4). Ordered arrays of junctional feet (3,4), referable to as ryanodine-sensitive Ca 2ϩ release channels (RYR) (5)(6)(7)(8)(9), bridge the gap of 90 -120 Å that separates the membrane of the transverse tubules from the junctional face membrane. The dihydropyridine-sensitive calcium channel of the transverse tubules acts as the voltage sensor for excitation-contraction coupling and plays a crucial role in the regulation of the RYR calcium channel (10 -14). In addition to the RYR, the junctional face membrane contains several proteins including the histidine-rich calcium binding protein, triadin, calsequestrin, and junctin (15)(16)(17)(18). During the past decades numerous studies have appeared concerning the biochemical characterization of the protein constituents of the junctional face membrane (19 -29). The most abundant polypeptide appears to be calsequestrin, the SR calcium storage protein (30), which might also be involved in the regulation of the RYR (31,32). Whether this effect is mediated by a direct interaction between the two proteins or via bridging calsequestrin binding proteins such as triadin or junctin is still controversial (17,21,29,33). The junctional face membrane is endowed with numerous other less abundant proteins having a molecular mass ranging from 20 kDa up to 120 kDa, which have yet to be identified and characterized at the molecular level (4). Because of their localiza-* This work was supported in part by grants from Telethon Italy number 908, Ministero Università e Ricerca Scientifica e Tecnologica 60 and 40%, ERBFMRXCT 960032, and Agenzia Spaziale Italiana. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AF306765.
‡ ‡ To whom correspondence should be addressed: Dept. of Experimental and Diagnostic Medicine, Section of General Pathology, University of Ferrara, Via Borsari 46, 44100 Ferrara, Italy. Tel.: 41-61-265-2371; Fax: 41-61-263-3702; E-mail: zor@dns.unife.it. tion, the proteins that are present on the junctional face membrane are deemed to be involved in the excitation-contraction coupling mechanism, and a defect in the function of these proteins could potentially lead to alterations of the Ca 2ϩ release process and/or intracellular Ca 2ϩ homeostasis. Thus the characterization of all the molecular components of the junctional face membrane is important not only for our understanding of the basic mechanism of Ca 2ϩ storage and release in muscle and non-muscle cells but also in view of the enormous effort aimed at identifying novel genes which, upon mutation, may be linked to neuro/muscular diseases (34,35).
In the present report, we demonstrate the existence of junctate, a novel integral Ca 2ϩ binding protein of sarco(endo)plasmic reticulum membranes. The Ca 2ϩ binding properties of junctate indicate that the protein might have an active role in the Ca 2ϩ storage/release process in ER membranes in a variety of tissues including heart, brain, pancreas, lung, liver, kidney, and skeletal muscle. In addition, we report that an enzyme (aspartyl ␤-hydroxylase), a structural protein of sarco(endo)plasmic reticulum membranes (junctin), and a calcium binding protein (junctate) belong to a family of single membrane-spanning proteins that result from alternative splicing events of the same gene located in human chromosome 8.

Materials
Nitrocellulose was from Amersham Pharmacia Biotech; isopropyl-␤-D-thiogalactoside, restriction enzymes, the chemiluminescence kit, synthetic primers, DNA-modifying enzymes, the DNA-digoxigenin labeling kit, and anti-digoxigenin peroxidase-conjugated antibodies were from Roche Molecular Biochemicals and New England Biolabs; the triplex, gt10 human skeletal muscle, and gt10 human cardiac cDNA libraries, pEGFP plasmids, and multiple tissue Northern blots were from CLONTECH; the Bluescript cloning vector was from Stratagene; the pGex plasmid was from Amersham Pharmacia Biotech; heparin-agarose and glutathione were from Sigma; protein molecular weight markers were from Bio-Rad; [ 45 Ca] and [ 32 P]dATP were from PerkinElmer Life Sciences; Lipofectin was from Life Technologies, Inc.; indo-1 was from Molecular Probes; and all other chemicals were reagent grade.

Methods
Preparation of Sarcoplasmic Reticulum Fractions-Terminal cisternae, longitudinal sarcoplasmic reticulum, transverse tubules, and sarcolemma were obtained from the white skeletal muscle of New Zealand rabbits as described (27). Protein concentration was according to Lowry et al. (36) using bovine serum albumin as standard.
Heparin-Agarose Chromatography-Vesicles obtained from the various fractions of rabbit skeletal muscle were resuspended at a final concentration of 1 mg/ml in 1% Triton X-100, 200 mM NaCl, 50 mM Tris-HCl, pH 8.5, 1 mM dithiothreitol, 1 mM EDTA containing 1 mg/ml leupeptin, 100 M phenylmethylsulfonyl fluoride, and benzamidine, and 1 mM pepstatin as anti-proteolytic agents (mixture of anti-proteolytic agents). The vesicles were solubilized for 30 min at room temperature under gentle agitation and centrifuged at 100,000 ϫ g max for 30 min at 4°C. The supernatant was incubated with a heparin-agarose resin (1 ml of resin/mg of protein) previously equilibrated in 50 mM Tris-HCl, pH 8.5, 200 mM NaCl, 1 mM dithiothreitol, and 1 mM EDTA (Buffer A) for 60 min at room temperature under gentle agitation. The column was washed with 10 bed volumes of Buffer A containing 0.1% Triton X-100 and proteins eluted with a step gradient of NaCl (300 mM, 500 mM, 750 mM, and 1 M) in Buffer A. Identification of the proteins contained in the 500 mM NaCl fraction was carried by Western blot analysis and microsequencing.
Electrophoresis and Immunological Staining-SDS-PAGE was carried out as described by Laemmli (37); proteins were visualized by Coomassie Brilliant Blue staining or by Stains All staining as described previously (38). Western blots were prepared as described previously (38).
Protein Microsequencing-The fraction eluting from the heparinagarose column at 500 mM NaCl was concentrated by addition of 3 volumes of cold acetone. The insoluble protein fraction was collected as a pellet after centrifugation at 4000 rpm (3000 ϫ g max ) in an ALC 5400 centrifuge. The pellet was resuspended in 50 mM Tris-HCl, pH 6.8, 2% SDS, and 0.1% bromphenol blue, and the proteins were separated by SDS gel electrophoresis in a 10% slab gel. The proteins were revealed by imidazole/ZnSO 4 reverse staining. To concentrate the protein, gel slices containing the 27-kDa protein bands (ϫ8) were transferred to a rod gel in a Pasteur pipette. The glass of the Pasteur pipette tube was treated with 5% Cl 2 (CH 3 ) 2 Si/CHCl 3 (dichloromethylsylane), dried, and the glass tubes were filled with a solution containing 5% acrylamide, 0.13% bisacrylamide, 125 mM Tris-HCl, pH 6.7. Polymerization was initiated by the addition of ammoniumpersulfate and TEMED. The gel slices were loaded on the rod gel in the presence of 5 ml of 50 mM Tris-HCl, pH 6.8, 2% SDS, 0.1% bromphenol blue. Electrophoresis was carried out at 150 -200 mV until the bromphenol blue reached the narrow part of the Pasteur pipette. The protein band was visualized by Coomassie Brilliant Blue staining. The slice of gel containing the protein was washed 3 times with 1 ml of MilliQ H 2 O and then incubated in 140 l of CNBr/HCO 2 H (58.1 mg of CNBr/1.162 ml of HCO 2 H) plus 60 ml of H 2 O overnight at room temperature under vortex stirring. The next day the supernatant was transferred to another microvial and dried in a speed vac. The peptides derived from the 27-kDa protein were separated by SDS-PAGE and blotted onto polyvinylidene difluoride membrane. The bands were visualized by Coomassie Brilliant Blue staining. The NH 2terminal sequence and the polypeptides derived from CNBr cleavage were subjected to amino acid sequencing several times (carried out by Prof. V. Hoppe, Physiologische Chemie II, Institute fü r Biowissenschaften der Univesität Wü rzburg, Wü rzburg, Germany).
cDNA and BAC Cloning, FISH Analysis, and Sequencing-A human triplex oligo(dT)-primed skeletal muscle library was screened using a 32 P-labeled cDNA probe obtained by RT-PCR of human skeletal muscle using forward and reverse primers obtained from the protein sequence of peptides 1 and 3; the sequences were 5Ј-AGA ATT CAC AGA GGA CAA AGA G-3Ј and 5Ј-AGA ATT CTA ACG CGA TGA CCA T-3Ј, respectively. Amplification conditions were as follows: 45Љ annealing at 45°C, 30Љ extension at 72°C, and 30Љ denaturation at 95°C for a total of 35 cycles. The addition of an EcoRI restriction enzyme sequence was used to facilitate subcloning and sequencing of the amplified cDNA. After overnight hybridization of the filters at 42°C with the labeled probe, the filters were washed at high stringency at 65°C in 0.2ϫ SSCP, 0.1% SDS; positive plaques were identified and purified, and the DNA was prepared according to the manufacturer's recommendations. Templates for sequencing were prepared in the Bluescript cloning vector. To obtain the full-length sequence we carried out nested exonuclease III/mung bean nuclease deletions according to a previously described procedure (39). Random/oligo(dT)-primed gt10 human skeletal and cardiac muscle libraries using a cDNA 32 P-labeled probe were obtained from triplex library clone. The primary sequence of the clones was obtained with an automated DNA sequencer using the dideoxy method. Human BAC genomic libraries were screened as described previously (40). To identify BAC clones, colonies were lifted onto nylon membrane for hybridization probe analysis with radioactively labeled cDNA probes encompassing the 3Ј end of both human junctin and junctate. FISH analysis was performed by Genomics Inc. (St. Louis, MO).
Extra Long PCR and Sequence Analysis of BAC Clones and Genomic DNA-Extra long PCR products were obtained using the GeneAmp XL PCR kit (Applied Biosystems, Foster City, CA) from 1 g of BAC or genomic DNA with either 28 or 37 amplification cycles using the PCR primers reported in Table II. PCRs were carried out according to the manufacturer's instructions; the products were purified with Microcon-100 (Millipore) and sequenced with the ABI PRISM Big Dye terminator cycle sequencing ready reaction kit using the ABI PRISM 377 DNA sequencer (PE Applied Biosystems, Foster City, CA).
Northern Blot Analysis and Southern Blot Analysis-2 g of poly(A) ϩ RNA from eight different human tissues (CLONTECH) and 20 g of total human genomic DNA digested with EcoRI and PvuII and blotted onto nylon membranes were hybridized with the appropriate cDNA probes as described previously (38). Hybridization was performed twice with two distinct membranes. Blots were washed under high stringency (0.2ϫ SSCP, 0.1%SDS at 65°C for 60 min), and autoradiography was performed for 1 week (Northern blot) or 3 weeks (Southern blots) at Ϫ80°C with an intensifying screen. Northern blots were also probed with digoxigenin-labeled ␤-actin probe and washed under high stringency, and the tissue distribution of ␤-actin was revealed using an anti-digoxigenin peroxidase-conjugated (1:10,000) antibody followed by chemiluminescence.
Expression of the Recombinant Protein-A GST fusion protein containing the COOH-terminal domain of junctate was constructed by fusing the 610-bp PCR-amplified cDNA (amino acid residues 98 -298) in frame into the multiple cloning site of pGex5 ϫ 3. The forward and reverse primers were, respectively, 5Ј-TCT CGA GGGGCA GTC TTT TTG AA-3Ј and 5Ј-AGA ATT CTA CTT CAG AGC CAG CA-3Ј. Amplification conditions were as described for cDNA cloning except that the annealing temperature was 60°C; the addition of the XhoI/EcoRI restriction enzyme sequence was used to facilitate cloning and sequencing of the amplified sequence. The GST fusion protein was purified from the bacteria using glutathione-Sepharose 4B as described by the manufacturer. 45 Ca 2ϩ Ligand Overlay and 45 Ca 2ϩ Equilibrium Binding-Calcium overlays on Western blots were carried out as described (41). Calcium equilibrium binding was carried out by the use of a continuos flow microdialysis chamber as described previously (38).
Expression of Recombinant junctate in Eukaryotic Cells and Intracellular Distribution-To monitor the intracellular distribution of junctate, we cloned it into the pEGFPC1 mammalian expression vector and monitored expression of the recombinant protein after transfection into COS-7 cells. Two NH 2 -terminal green fluorescent protein-tagged constructs were made; one encompassed the whole coding sequence of junctate (nucleotides 1-979; EGFP-junctate), whereas the other encompassed the putative hydrophobic transmembrane domain between nucleotides 137 and 254 (EGFP-TM-junctate). PCR amplification conditions were as described for cDNA cloning except that the annealing temperature was 58°C. The EcoRI/BamHI restriction enzyme sequence was used to facilitate cloning and sequencing of the amplified sequence into the MCS of the pEGFPC1 plasmid. Forward and reverse primers for the EGFP-junctate construct were 5Ј-AGA ATT CAC AAA TGG CTG AAG-3Ј and 5Ј-GAA GCT TTT AGG ATC CTG GTG-3Ј, respectively; forward and reverse primers for the EGFP-TM-junctate construct were 5Ј-GGA ATT CCA CCA TGA GGA AAG GCG GAC TCT CA-3Ј, and 5Ј-GGG ATC CCT TTG CTT TGG CTA GA, respectively. COS-7 cells grown on glass coverslips were transfected using Lipofectin as described previously (42). 24 or 48 h after transfection cells were washed 3 times with phosphate-buffered saline, fixed with 3.7% formaldehyde in phosphate-buffered saline for 20 min, and examined under fluorescent light (excitation, 480 nm; emission, 510 nm) using a Nikon Diaphot 300 inverted microscope equipped with a PlanApo ϫ 100/1.40 objective.
[Ca 2ϩ ] i Ratio Measurements-The free cytoplasmic Ca 2ϩ concentration of COS-7 cells was determined using the fluorescent Ca 2ϩ indicator indo-1 using a Nikon Diaphot 300 inverted fluorescent microscope equipped with a ϫ 60 Plan-Apo oil immersion objective attached to two photomultipliers (P100; Nikon Inc.). The cell to be measured was identified by its green fluorescence and then ratio fluorescence measurements (410/480 nm) were obtained as described previously (42). Excitation of indo-1 was achieved using a 100-watt mercury lamp attenuated by neutral density filters to avoid dramatic photobleaching. Stimulation of individual cells was obtained by an 8-way 100-m diameter quartz micromanifold computer-controlled microperfusor (ALA Scientific). The total amount of calcium released and statistical analysis were performed using the Origin computer program (Microcal Software, Inc., Northampton, MA).
Preparation of Kidney Microsomes-Kidneys from 3.0 -3.5-kg male New Zealand rabbit were homogenized (10% w/v) in a buffer containing 10 mM HEPES, pH 7.2, 150 mM KCl plus a mixture of anti-proteolytic agents (see "Heparin-Agarose Chromatography"). The homogenate was centrifuged at 3,000 ϫ g max for 10 min, and the resulting supernatant was centrifuged at 10,000 ϫ g max for 15 min. The 10,000 ϫ g max supernatant was then filtered through 10 layers of cheesecloth and centrifuged at 150,000 ϫ g max for 60 min. The pellet was resuspended in a solution containing 10 mM HEPES, pH 7.2, 0.6 M KCl plus antiproteolytic agents and was centrifuged at 150,000 ϫ g max for 60 min. The KCl-washed membranes were resuspended in a solution containing 10 mM HEPES, pH 7.2, 150 mM KCl at a final concentration of 1-2 mg/ml. The microsomal suspension was then incubated for 30 min at 4°C in the presence of 100 mM Na 2 CO 3 , pH 11. The membrane fraction pellet was obtained by centrifugation at 150,000 ϫ g max for 60 min and washed with 0.6 M KCl. Biochemical analyses of kidney microsomes were carried out with three different preparations with microsomes isolated from two different animals.
Preparation of Rabbit Heart Microsomes-Preparation of rabbit heart microsomes was essentially as described by Pessah et al. (43), except that the total microsomal fraction was washed with 0.6 M KCl before being layered onto the sucrose gradient to remove cytoskeletal proteins.
Preparation of Antibodies-Polyclonal Abs were raised by immunizing mice with the proteins present in the fraction eluting at 500 mM NaCl from the heparin-agarose column. To affinity purifiy anti-junctin Ab we blotted the fraction eluting at 500 mM NaCl from the heparinagarose column onto nitrocellulose; to visualize the proteins, mem-branes were stained with a solution containing 3% (w/v) trichloroacetic acid, 0.1% (w/v) Ponceau red. The band that corresponds to junctin as revealed by NH 2 -terminal amino acid sequencing data was cut out of the membrane and used to affinity purify the Abs as described previously (38). To raise anti-junctate antibodies, the glutathione-Sepharosepurified GST fusion protein was used to immunize rabbits as described previously (38); the serum was tested for the presence of antibodies, and the IgG fraction was purified through a protein A-Sepharose column. Antibodies reacting with the bacterial apo-protein were removed by batch extraction on the purified GST protein.
Peptide Synthesis-A peptide corresponding to the first 10 amino acids of rabbit skeletal muscle junctin was synthesized as described previously (44). Quality control was carried out by mass spectroscopy analysis.

RESULTS
The biochemical and functional characterization of the molecular components present in SR membranes is important for understanding the fine mechanisms underlying Ca 2ϩ homeostasis not only in skeletal muscle but also in other tissues. To analyze novel components involved in Ca 2ϩ homeostasis, we identified the proteins present in terminal cisternae that copurify with the ryanodine receptor. We obtained a fraction eluting at 500 mM NaCl from a heparin-agarose column that was particularly enriched in a 27-kDa protein (Fig. 1A), as well as other proteins, including the RYR and triadin as determined by immunostaining (not shown). The NH 2 -terminal amino acid sequence of the 27-kDa protein did not match that of any protein present in the Swiss-Prot and NCBI BLAST data bases. Internal sequencing subsequently revealed the protein as junctin (Fig. 1B), a protein previously identified as a calsequestrin binding protein expressed in both skeletal and cardiac muscles (15,16,29).
cDNA Sequence Determination-The divergence of the NH 2terminal sequence between rabbit skeletal muscle and dog heart junctin could be due to tissue-specific or species-specific differences. We addressed this issue by determining and comparing the primary sequences of skeletal muscle and cardiac junctin deduced from cDNA clones pulled out from human skeletal and cardiac muscle libraries. A 98-bp cDNA probe was obtained by RT-PCR on human skeletal muscle RNA using the forward and reverse primers indicated under "Experimental Procedures." The PCR product was sequenced to confirm its identity and then used to screen a triplex human skeletal muscle library. After screening approximately 1 ϫ 10 5 plaqueforming units we obtained a strong positive signal; single plaque purification was carried out, and a cDNA insert of approximately 750 bp was obtained and subjected to nucleotide sequencing. The insert contained a 150-bp 5Ј untranslated region; the initial methionine residue was followed by an open reading frame of approximately 500 nucleotides, after which the sequence was terminated by a series of (A) 8 . Because the triplex human skeletal muscle library was oligo(dT) primed, the stretch of poly(A) introduced a second priming site that prevented the isolation of overlapping clones containing the 3Ј end of the cDNA. Thus, we screened random/oligo(dT)-primed gt10 human skeletal and cardiac muscle libraries using (i) a cDNA 32 P-labeled probe obtained from the whole 750-bp segment (probe A) of the triplex library or (ii) a 200-bp cDNA segment obtained from the most 3Ј end (probe B) of the triplex insert. Several clones were pulled out, and those larger than 800 bp were characterized. Partial nucleotide sequence, restriction map, and hybridization probe analyses revealed that the cDNA clones fell into two groups. The prototype of the first group, clone 16, pulled out from both cardiac and skeletal muscle libraries with both probes (A and B), was similar to the published sequence of dog heart junctin (not shown). Comparison of the amino acid sequence deduced from the only open reading frame within the cDNA of human skeletal muscle junctin clone 16 with that of dog heart junctin revealed an overall identity of 84% (not shown). Of interest (i) the first few amino acids that are usually the fingerprint of a protein are not identical in rabbit, human skeletal muscles, and dog heart junctin (Table I); (ii) the NH 2 -terminal methionine is not conserved in the mature rabbit muscle protein; (iii) the putative transmembrane sequence starting at amino acid 30 -60 is identical in the three species examined; (iv) the molecule has an overall highly positive charge; and (v) residues 6 to 78 of human skeletal muscle junctin are 100% identical to those between residues 35 and 107 of human aspartyl ␤-hydroxylase (see Fig. 3, panel A for comparison and alignment of the sequences).
The prototype of the second class of clones is represented by clone 58, which was pulled out from the human cardiac gt10 library with probe A and is depicted in Fig. 2. The amino acid sequence similarity of clone 58 to dog heart and human skeletal muscle junctin was limited to the first 78 residues. Immediately after residue 78 the sequences diverged and showed very little homology thereafter. The differences are located in the COOH-terminal domain of the molecule, in a region that has been proposed to be localized in the lumen of ER/SR. A BLAST search revealed that the amino acid sequence deduced from the entire length of the COOH-terminal region of clone 58 matched that of aspartyl ␤-hydroxylase, a member of the ␣-ketoglutaratedependent dioxygenase family. However, it should be stressed that the region of aspartyl ␤-hydroxylase that is identical to the sequence of clone 58 is restricted to the portion that is not endowed with enzymatic activity. Henceforth the chimeric mol-ecule having the first 0 -78 residues identical to junctin and the remaining COOH-terminal sequence identical to aspartyl ␤-hydroxylase (residues 35-312) will be referred to as junctate.
This second class of clones is not only restricted to the human cardiac gt10 library, because a nearly identical clone was also pulled out of the skeletal muscle gt10 library (not shown), a result that would also exclude possible cloning artifacts. Restriction enzyme mapping, Southern blot analysis, and partial sequencing of the latter clone obtained from the skeletal muscle library revealed it to be identical to cardiac clone 58, except that it was missing a sequence, AKAKDFRYNLSEVLQ, defined by residues 56 -70 of cardiac clone 58. Comparison of the deduced amino acid sequences of junctate and aspartyl ␤-hydroxylase reveals that the two molecules were identical between residues 6 -55 and 35-85 and between residues 71-298 and 86 -312 of junctate and aspartyl ␤-hydroxylase, respectively (Fig. 3, panel A). Unlike junctin, junctate has an overall negative charge in its lumenal COOH terminus. In fact in the region of divergence there are 64 negative and 6 positive residues and 26 negative and 48 positive residues in junctate and junctin, respectively. Other important differences are as follows: (i) the presence of a single Cys residue in the lumenal portion of junctate that may be involved in interchain disulfide bond formation; and (ii) the presence of a consensus sequence for glycosylation (NLS) in the region immediately after the putative transmembrane hydrophobic region (Fig. 3, panel A).
Expression of junctate and junctin in Human Tissues-The identification of junctate cDNA was rather intriguing and prompted us to investigate the expression pattern of the pro-FIG. 1. Heparin-agarose fraction eluting at 500 mM NaCl. Panel A, fractions eluting at 500 mM NaCl were pooled, concentrated, and resuspended in 100 l of Laemmli buffer. 10 l of resuspended proteins were loaded on a 10% SDS polyacrylamide gel. The arrow indicates the protein band on which NH 2 -terminal sequencing was carried out. The amino acid sequence is given in single letter code; X indicates an unknown residue. Panel B, the band of 27 kDa was cut from the 10% SDS polyacrylamide gel and digested with CNBr. The proteolytic fragments were separated on a 15% SDS polyacrylamide gel and blotted onto polyvinylidene difluoride membrane. The proteins were stained with Coomassie Brilliant Blue, cut off the membrane, and used for microsequencing. The amino acid sequence is given in single letter code; X indicates an unknown residue. Sequences obtained from peptides 2 and 3 are similar to dog heart junctin.

TABLE I
Comparison of amino acid sequences of junctin from dog heart, rabbit, and human skeletal muscle Amino acid sequence is given as single letter codes. X represents unknown residues.  3. Comparison of the amino acid sequence deduced from the cDNAs of human junctate, human aspartyl ␤-hydroxylase, human junctin, and human cardiac junctin isoform 1. Panel A, multiple sequence alignment was carried out with ClustalW computer program, which is available at the Swiss node of the European Molecular Biology network. Black boxes indicate identical residues; gray boxes indicate conserved residues. Underlined residues indicate the amino acid sequence obtained from protein microsequencing (Table 1). Human junctate and junctin sequences were from the present work. Human cardiac junctin isoform 1 and human aspartyl ␤-hydroxylase sequences were downloaded from the NCBI data bank. Note that junctin isoform 1 contains a 15-amino acid residue insert at position 56, which was not found in tein in different human tissues. We used two different 32 Plabeled cDNA probes to determine the pattern of expression of human junctin and junctate mRNAs by Northern blot analysis. 2 g of poly(A) RNA from eight different human tissues were hybridized with a radiolabeled XbaI-BamHI probe obtained from junctate cDNA defined by nucleotides 850 -1237, which is different from the cDNA of aspartyl ␤-hydroxylase beginning from nucleotide 981. Although the probe we used for junctate exhibits a 130-bp overlap with the cDNA sequence of aspartyl ␤-hydroxylase, we observed an expression pattern clearly different from that reported for aspartyl ␤-hydroxylase by Korioth et al. (50). This result is accounted for by the small overlapping sequence and by the high stringency washing conditions we used. This junctate probe detected a single major transcript of approximately 2.6 kb in heart, brain, placenta, lung, liver, skeletal muscle, kidney, and pancreas; the weakest level of transcription was found in skeletal muscle whereas the highest was in heart, brain, and pancreas (Fig. 4, panel A). The size of the mRNA transcript recognized by the junctate probe is slightly higher than that of the cDNA clone we characterized, a result consistent with a small portion of the 3Ј end including the poly(A) tail, which was not sequenced. On the other hand, when the same blot was hybridized with a 0.2-kb SacI fragment obtained from the clone encoding the COOH-terminal portion of junctin (amino acid residues 100 to 165), hybridizing transcripts were only detected in skeletal and cardiac muscles (Fig.  4, panel B). The latter result confirms the observation of Jones et al. (15). In fact cardiac muscle displays a transcript of approximately 3.5 kb, whereas in skeletal muscle two bands were positive; the major one has the same size of that present in heart, and the minor transcript is of approximately 4.5 kb.
Structural Analysis of the Gene Encoding junctate, Aspartyl ␤-Hydroxylase, and junctin-We next performed Southern blot analysis of human genomic DNA with the same cDNA probes that were used for the Northern blots. Human genomic DNA was digested with EcoRI or PvuII, and the fragments were separated on a 0.8% agarose gel and electrically transferred onto a nylon membrane. The blots were then probed with either a radiolabeled XbaI-BamHI (nucleotides 850 -1237) fragment obtained from junctate cDNA (Fig. 5, panel A) or a 0.2-kb SacI fragment from junctin cDNA (amino acid residues 100 to 165) (Fig. 5, panel B). Digestion with EcoRI (Fig. 5A, lane 1) yielded either a single 4-kb hybridizing band with a junctate-specific cDNA probe or two bands of about 2.4 and 5.0 kb with a junctin-specific cDNA probe (Fig. 5B, lane 1). Similarly, different hybridization patterns were obtained when the genomic DNA was digested with PvuII (Fig. 5B, lane 2). These results are compatible with the existence of either two independent single copy genes or a single copy of one gene encoding both proteins. In the latter case one would expect a large gene, and the hybridization with different cDNA probes would pick up different regions of the gene. We investigated this possibility by analyzing the gene(s) encoding junctin and junctate. We first isolated two BAC genomic clones by using the unique 3Ј end cDNA sequences of junctin and junctate as probes. The identity of the BAC clones was confirmed by nucleotide sequencing. The two BAC clones cover up to 200 kb of human genomic sequence and contain overlapping sequences as revealed by Southern blot analysis (not shown). FISH analysis revealed that the loci of both junctin and junctate map to the cytogenetic band q12.1 on human chromosome 8 (Fig. 6). Interestingly, the aspartyl ␤-hydroxylase locus was also mapped to the same region of chromosome 8 (45). These data indicate that either the genes for aspartyl ␤-hydroxylase, junctin, and junctate are very close to each other or that the three proteins result from alternative splicing events of the same gene. We addressed this issue by determining the intron/exon boundaries in correspondence with the 5Ј end of the gene (Fig. 7). The two BAC clones were cut with EcoRI, BglII, or PstI, and some of the fragments were cloned and partially sequenced. To amplify both BAC clones and human genomic DNA by extra long PCR, we designed a set of primers on the basis of sequences we obtained from BAC clones and cDNA (see Table II). The nucleotide sequences we obtained were then compared with those of the cDNAs of junctate, junctin, and aspartyl ␤-hydroxylase. The combination of data obtained by sequencing, PCR amplification, and Southern blotting allowed us to define the structure of the 5Ј region of the locus of aspartyl ␤-hydroxylase, junctin, and junctate (see Fig.  7 and Table III). The data obtained indicate that the first exon of aspartyl ␤-hydroxylase is located more than 20 kb upstream from the first exon of junctin/junctate. The second and the third exons, which are common to the three proteins and encompass the amino acid sequence of the predicted transmembrane segment, as well as the cytoplasmic loop of junctin and junctate, are located 5 to 6 kb downstream the first exon of junctin. Interestingly, the sequence AKAKDFRYNLSEVLQ, defined by residues 56 -70 of cardiac clone 58, is encoded by a small exon (number 4) located between the third and fifth exons, which are common to the three proteins. We found a rare GCAAG splice donor site at the 5Ј end of the fifth intron. Such a splice site has been also described for other genes (46 -48). The use of differ-ential splice donors has been shown to be involved in the generation of protein diversity by alternative splicing (47). Taken together these data clearly show that three distinct proteins, i.e. junctin, junctate, and aspartyl ␤-hydroxylase result from alternative splicing events of the same gene localized in the human chromosome 8 (40).
Targeting of junctate to the ER Membranes of COS-7 Cells and Its Functional Characterization-Junctate contains a hydrophobic sequence that has been predicted to form a membrane-spanning segment, i.e. the structure necessary to anchor the protein to sarco(endo)plasmic reticulum membrane. To verify this we expressed the full-length junctate cDNA in COS-7 cells. To visualize the cells expressing the junctate cDNA, we fused EGFP to the NH 2 -terminal portion of junctate. As expected cells expressing the full-length EGFP-junctate fusion protein display granular perinuclear fluorescence (Fig. 8, panel  B), which is quite different from that of cells expressing EGFP alone (Fig. 8, panel A). This type of fluorescence is consistent with localization of the recombinant fusion protein into the ER membranes. To define in greater detail the minimal sequence necessary for targeting of junctate we synthesized by PCR the cDNA sequence encoding the membrane-spanning segment of junctate and fused it in frame to the COOH terminus of EGFP. As can be seen the fluorescence pattern of the EGFP-TMjunctate is similar if not identical to its full-length counterpart (Fig. 8, panel C).
The COOH terminus of junctate differs from that of junctin because of the prevalence of negatively charged residues (Fig.  3). This observation led us to investigate whether junctate displays different functional properties from junctin. Fusion proteins consisting of the COOH-terminal region of junctate were constructed. The expression of the recombinant proteins was monitored by SDS-PAGE, and a protein of the expected size is present only in the soluble fraction of the induced Escherichia coli culture (not shown). Because of the prevalence of negatively charged residues in the COOH terminus domain of junctate, we investigated whether such a domain is able to bind Ca 2ϩ . Fig. 9A shows an autoradiogram of a 45 Ca 2ϩ ligand overlay with bacterially expressed, recombinant COOH-terminal junctate (GST-C-junctate), which is able to bind calcium (Fig. 9A, lane 4). The binding activity of GST-C-junctate is specific, because GST alone is not able to bind calcium (Fig. 9A,   lane 2). This result was confirmed by staining the proteins with Stains All, a carbocyanine cationic dye that metachromatically stains calcium binding proteins blue (Fig. 9B, lane 1). Finally Scatchard analysis (Fig. 9C) shows that the COOH-terminal portion of junctate binds 650 mol of Ca 2ϩ /mg of protein (21.0 mol of Ca 2ϩ /mol of protein) with a k D of 217 Ϯ 20 M (n ϭ 5).
We next examined the effect of transiently overexpressing junctate on the [Ca 2ϩ ] i elevations in response to ATP, an agonist that has been shown to release Ca 2ϩ from intracellular stores via InsP 3 production (49). Cells were transfected either with the full-length pEGFP-junctate construct or with the truncated version, pEGFP-TM-junctate, as control; 48 h post transfection cells were loaded with indo-1, and the [Ca 2ϩ ] i was monitored. Both the peak amplitude and the total amount of calcium released by ATP were significantly higher in cells transfected with the cDNA encoding full-length junctate, compared with those transfected with the cDNA encoding TMjunctate (Fig. 10, panel A). In fact COS-7 cells transfected with the cDNA encoding pEGPF-junctate showed a 55% increase (student's t test for paired samples, p Ͻ 0.000001) in peak [Ca 2ϩ ] i compared with those transfected the pEFGP-TM alone (Fig. 10, panel B). When the total amount of Ca 2ϩ released by 10 M ATP was examined (by calculating the integral), the difference between the two constructs was even greater (increase of 116%; student's t test for paired samples, p Ͻ 0.000001) (Fig. 10, panel C). No difference was observed either in peak amplitude or total calcium released, between mock-transfected cells and cells transfected with the pEGFP-TM cDNA (not shown).
Identification of the Protein Product of the junctate Transcript in Kidney and Cardiac Microsomes-Based on the deduced amino acid sequence and analysis of the expressed recombinant protein, it is clear that junctate is (i) an integral membrane protein; (ii) has a molecular mass of 33 kDa; and is (iii) structurally related to junctin. To confirm the existence of a protein product encoded by the junctate mRNA transcript, we looked for a protein having the above biochemical properties in the membrane fraction of kidneys, an organ expressing junctate mRNA. The total microsomal fraction isolated from rabbit kidney was washed with 0.6 M KCl (Fig. 11, lanes 1 and 5) and then extrinsic proteins were removed from membrane vesicles by treatment with 100 mM Na 2 CO 3 at alkaline pH (Fig. 11,  lanes 2 and 6). The microsomal fraction resulting from the treatment with sodium carbonate was also washed with 0.6 M KCl (Fig. 11, lanes 3 and 7). Integral proteins were separated on a 10% SDS-PAGE and blotted onto nitrocellulose. To establish whether kidney membranes are endowed with an integral membrane protein structurally related to junctin we carried out Western blot analysis with affinity-purified anti-rabbit skeletal muscle junctin Ab. The affinity-purified Ab immunodecorated a protein of approximately 27 kDa in rabbit terminal cisternae (Fig. 11, lane 4; **), as well as a protein of slightly slower mobility (Fig. 11, lanes 5-7; *), as expected from the cDNA and Northern blot analysis, which is present in kidney microsomal membranes. Analysis of the second exon indicates that junctin and junctate share the first 6 NH 2 -terminal amino acid residues. To confirm the identity of the immunopositive band we carried out a Western blot in the presence of a competing peptide encompassing amino acids encoded by the second exon plus the first 4 amino acids of exon 3. The immunological reactivity of the anti-junctin Ab with the 32-kDa protein was abolished by the addition of the competing peptide. Thus the 32-kDa protein present in kidney microsomal membranes contains an amino acid sequence that is identical to the NH 2terminal sequence of rabbit skeletal muscle junctin and junctate. This 32-kDa protein is enriched 2-3-fold in the microso- mal fraction resulting from high salt wash and sodium carbonate treatment, indicating that it is an integral membrane protein. Altogether, these data are consistent with the conclusion that the 32-kDa intrinsic protein from kidney microsomes represents the product encoded by junctate mRNA.
We also analyzed the presence and the distribution of junctate in cardiac SR membranes. The total microsomal fraction was isolated from rabbit hearts and fractionated according to the procedure described under "Experimental Procedures" (43). The very same blot affinity-purified anti-junctin Abs that were used to immunodecorate the proteins present in kidney microsomes were used to immunostain the proteins of cardiac SR fractions. Western blotting revealed the presence of two immunopositive proteins in cardiac microsomes; one has the expected molecular mass of 27 kDa and is referable to as junctin ( Fig. 12; **), and the other band has a higher apparent molec-ular mass of approximately 32 kDa ( Fig. 12; *). The distribution of the two proteins roughly overlaps; they are mainly distributed in the fractions collected from the 32-34 and 34 -38% sucrose interfaces. The 32-kDa protein also appears to be present, though to a lower extent, in the fraction collected from the 27-32% sucrose interface. We could not discern gross differences in their apparent abundance (Fig. 12, panel C). The proteins present in the cardiac SR membranes were also stained with a polyclonal Ab raised against the C-terminal domain of junctate, i.e. the portion of protein that is in common with the central non-catalytic domain of aspartyl ␤-hydroxylase. As expected, when we used the latter Ab, two bands having a molecular mass of 32 and 90 kDa, respectively, were immunopositive (Fig. 12, panel B). The component having a higher molecular mass displays a distribution clearly distinct from the 32-kDa band. The 90-kDa protein is referable to as the  active form of aspartyl ␤-hydroxylase and is partitioned in the light membranes and collected from the 27-32% sucrose interface (Fig. 12, panel B, lane 2). The 32-kDa immunopositive band, on the other hand, exhibits a distribution similar, if not identical, to the 32-kDa protein stained with anti-junctin Ab, a result consistent with the notion that such a protein represents junctate.

DISCUSSION
In the present report we describe for the first time the existence of junctate, a protein that, because of its biochemical properties and tissue distribution, is deemed to play an important role in calcium homeostasis. Junctate is made up of 298 amino acids and contains one membrane-spanning segment; the first 23 residues of the protein are predicted to be in the cytoplasm whereas the bulk of the molecule is located in the lumen of sarco(endo)plasmic reticulum membranes. The NH 2terminal portion of junctate displays an amino acid sequence that is identical to that of junctin, a component of the calcium release site of cardiac and skeletal muscle SR. During the revision of this manuscript the NCBI data bank released the amino acid sequences of human cardiac junctin and human cardiac junctin isoform 1. The human cardiac junctin isoform 1 is identical to our junctin sequence (Fig. 3, panel A) but contains the 15-amino acid insert (residues 56 -70) that was found in the junctate sequence encoded by cardiac cDNA clone 58. Thus, junctate is identical to human cardiac junctin isoform 1 up to residue 93 (Fig. 3, panel A). On the other hand, the COOH-terminal domain of junctate is identical to the central region of aspartyl ␤-hydroxylase, a widely distributed protein responsible for the post-translational hydroxylation of aspartic acid/asparagine residues. However, because junctate lacks the domain endowed with enzymatic activity it must play a different role from amino acid hydroxylation within the cell.
This result was rather intriguing and prompted us to investigate the matter in greater detail; FISH, gene structure, and cDNA analysis clearly revealed that via an alternative splicing event of the same gene, which is located at position q12.1 of human chromosome 8, three functionally distinct proteins could be generated. As to the regulation of the expression of this gene, we think that it is rather complex most likely under the control of different promoters and transcriptional factors. In fact, Northern blot analysis of human tissues demonstrated that two transcripts of 2.6 and 4.3 kb encoding aspartyl ␤-hydroxylase are approximately equally distributed in the heart, placenta, skeletal muscle, and kidney, whereas the most enriched tissue appeared to be the lung (50). On the contrary, junctate mRNA is particularly abundant in pancreas and heart followed by brain and kidney, whereas skeletal muscle exhibits the least amount of transcription. The tissue distribution of junctin does not overlap either with that of junctate or with that of aspartyl ␤-hydroxylase (50).
The presence of mRNA strongly suggests, but does not necessarily prove, the existence of a protein product. We approached this issue by performing biochemical analyses on the microsomal fraction of kidney, a tissue enriched in the junctate transcript. Western blot analysis using an affinity-purified anti-junctin antibody revealed the presence of a 32-kDa integral membrane protein. Competition experiments unambiguously show that the binding of the anti-junctin Ab to the rabbit kidney 32-kDa integral membrane protein occurs through the first 10 NH 2 -terminal residues. Because (i) such a sequence is unique for junctin and for junctate, (ii) Northern blot analysis shows that there is no detectable amount of junctin mRNA expressed in kidney, and (iii) the molecular mass (32 kDa) of the immunopositive integral protein is compatible with that of junctate, we provide unequivocal evidence for the existence of the protein junctate.
The primary sequences predicted from junctate and aspartyl ␤-hydroxylase cDNAs reveal that these two proteins share the membrane-spanning sequence with junctin, a protein selectively localized in the junctional face membrane of skeletal and cardiac muscles, i.e. the calcium-releasing site of sarcoplasmic reticulum. The membrane-spanning segment of junctin also shows a high degree of similarity with that of triadin, another protein selectively localized to the junctional sarcoplasmic reticulum (20,29). Although aspartyl ␤-hydroxylase and junctin share their transmembrane segment, the distribution of these two proteins within the cardiac SR membranes is distinct. Western blot analysis of cardiac SR fractions clearly shows that the 90-kDa component, which represents the enzymatically active form of aspartyl ␤-hydroxylase, is mainly present in the light fraction collected from 27-32% sucrose interface, whereas junctin and junctate are enriched in the 32-34% and 34 -38% sucrose interface fractions, the same fractions that have been shown to be enriched by [ 3 H]ryanodine binding activity (43). To our knowledge this is the first time that the protein aspartyl ␤-hydroxylase has been shown to exist in cardiac SR membranes, and for the time being we have no clear cut answers as to its exact role in cardiac physiology. On the other hand, junctin and junctate have a similar distribution that is distinct from that of aspartyl ␤-hydroxylase; thus the transmembrane segment is not sufficient to target proteins to their ultimate FIG. 9. Ca 2؉ binding of GST-C-junctate. Panel A, 45 Ca 2ϩ ligand overlay. Proteins contained in 15 l of a total bacterial extract from an isopropyl-␤-D-thiogalactoside-induced culture of E. coli was transformed with pGEX (lanes 1 and 2), and the glutathione-Sepharose 4B-purified GST-C-junctate (lanes 3 and 4) were electrophoretically separated in a 10% SDS gel and blotted onto nitrocellulose.  10. [Ca 2؉ ] i increases in response to Ca 2؉ release from intracellular stores. Cells were either transfected with the cDNA-encoding fulllength junctate or that encoding the truncated version encoding the transmembrane domain portion (pEGFP-TM) fused in frame with pEGFP as described for Fig. 7. Cells were loaded with the fluorescent Ca 2ϩ indicator indo-1 and stimulated with 10 M ATP to elicit calcium release from intracellular stores; the fluorescence ratio (410/480 nm) of individual cells was measured as described under "Experimental Procedures." Panel A, trace representing typical response of cells transfected for 48 h with the indicated construct. As can be seen both the peak amplitude and integral calcium increases are elevated when cells overexpressing full-length junctate were examined. The calcium increase elicited by ATP in mock-transfected cells or cells transfected with the truncated junctate were not significantly different. The peak amplitude (panel B) and integral calcium (panel C) released by ATP in the presence or absence of extracellular calcium in cells transfected with the cDNA encoding the full-length junctate or the transmembrane portion (pEGFP-TM) fused inframe with pEGFP was calculated. Results are expressed as the mean peak increase in fluorescence ratio or integral of the total calcium released (calculated with Origin software) in at least four separate experiments and the indicated number of cells. subcellular compartment.
A peculiar feature of junctate is the net negative charge in the domain that is predicted to be in the lumen of the ER. These negative charges account for its ability to bind calcium as demonstrated by 45 Ca 2ϩ ligand overlay data. Scatchard analyses of equilibrium binding indicate that junctate is a high capacity moderate affinity calcium binding protein. This is quite an intriguing result in view of the fact that junctate is an integral membrane protein, whereas the majority of calcium binding proteins are soluble proteins preferentially localized in the cytoplasm or in the lumen of endocellular membrane compartments (51). The junctate transcript is expressed to different extents in a variety of tissues, including pancreas, brain, kidney, liver, and lung. Our expression studies in COS-7 cells clearly demonstrate that the fluorescence of recombinant EFGP-junctate is consistent with its localization to the ER, suggesting that junctate is associated with this membrane compartment. Cells overexpressing junctate clearly exhibit an increased intracellular Ca 2ϩ content, and thus this protein may be part of the agonist-sensitive rapidly exchangeable intracellular calcium stores, as has been shown with calreticulin, a soluble protein found in the lumen of the endoplasmic reticulum (52). Nevertheless it should be kept in mind that even in the heart, a tissue particularly enriched in the junctate transcript, the content of this protein is roughly comparable with that of junctin, a minor component of junctional SR. The accumulation of calcium tightly associated with the inner leaflet of the SR/ER membrane may be important to increase the local calcium concentration next to the lumenal mouth of the calcium release channel or in regulating protein-protein interaction with key elements of the molecular machinery involved in calcium homeostasis. As depicted in Fig. 3, junctin and junctate have opposite net charges in the predicted lumenal loop. If these proteins are adjacent to each other in the junctional SR membrane one could envisage a potential electrostatic interaction between the two. Under resting conditions this seems unlikely because of the high calcium concentration in the lumen of the SR. Thus, under these conditions the calcium binding sites of junctate would be saturated by their ligand, and the protein would not be available for electrostatic interactions with junctin and/or other proteins. Immediately after calcium release and before refilling of the calcium stores, however, the intralumenal calcium concentration is lower than the k D of junctate for calcium. In the latter condition, junctate would be in its calcium-free form, whereby it would be available to interact directly with other components of junctional SR, including junctin and triadin.
Thus, by analogy with its muscle membrane counterpart, it is possible that the calcium bound to junctate could increase the local calcium concentration in microdomains that are adjacent to the lumenal mouth of the calcium release channel of ER membranes. The prevalent Ca 2ϩ release channel of the ER is the InsP 3 receptor (53), and further studies are needed to clarify whether the COOH terminus of junctate interacts and/or co-localizes with the InsP 3 receptor. Alternatively, the capacity of junctate to bind calcium could be interpreted as an adaptive response of the protein to a high [Ca 2ϩ ] environment, such as that found in the lumen of SR/ER (54). If this is the case, whatever function junctate might have would be protected by neutralizing the high [Ca 2ϩ ] by binding this ion, as has been recently suggested to occur for other ER proteins including calnexin, protein disulfide isomerase, and calreticulin (54).