Localization and Characterization of the Calsequestrin-binding Domain of Triadin 1

Triadin is an integral membrane protein of the junctional sarcoplasmic reticulum that binds to the high capacity Ca2+-binding protein calsequestrin and anchors it to the ryanodine receptor. The lumenal domain of triadin contains multiple repeats of alternating lysine and glutamic acid residues, which have been defined as KEKE motifs and have been proposed to promote protein associations. Here we identified the specific residues of triadin responsible for binding to calsequestrin by mutational analysis of triadin 1, the major cardiac isoform. A series of deletional fusion proteins of triadin 1 was generated, and by using metabolically labeled calsequestrin in filter-overlay assays, the calsequestrin-binding domain of triadin 1 was localized to a single KEKE motif comprised of 25 amino acids. Alanine mutagenesis within this motif demonstrated that the critical amino acids of triadin binding to calsequestrin are the even-numbered residues Lys210, Lys212, Glu214, Lys216, Gly218, Gln220, Lys222, and Lys224. Replacement of the odd-numbered residues within this motif by alanine had no effect on calsequestrin binding to triadin. The results suggest a model in which residues 210–224 of triadin form a β-strand, with the even-numbered residues in the strand interacting with charged residues of calsequestrin, stabilizing a “polar zipper” that links the two proteins together. This small, highly charged β-strand of triadin may tether calsequestrin to the junctional face membrane, allowing calsequestrin to sequester Ca2+ in the vicinity of the ryanodine receptor during Ca2+ uptake and Ca2+ release.

polar zippers (14 -16). Guo and co-workers (9, 10) first localized the calsequestrin-binding domain of triadin to the common lumenal region (residues 69 -264 for the rabbit triadins), where short runs of alternating lysine and glutamic acid residues motifs are present in very high content. The lumenal region of junctin also contains a high density of KEKE motifs and also binds calsequestrin avidly (6,10). Thus, it was proposed that KEKE motifs may be responsible for triadin (and junctin) binding to calsequestrin (10). This idea is reinforced by the fact that calsequestrin is a highly charged, acidic protein (17), suggesting that KEKE association motifs of triadin (and junctin) could bind to KEKE-binding sites (13) of calsequestrin. To date, however, the region of triadin (or junctin) binding to calsequestrin has not been definitively localized, and the role of the KEKE motifs, if any, in the relatively specific interaction between these proteins has not been investigated.
Here we have utilized deletional and site-specific mutagenesis of triadin 1 in combination with the radiolabeled calsequestrin overlay technique (6,18) to localize the calsequestrinbinding domain of cardiac triadin. Our results demonstrate that a charged cluster of KEKE residues is indeed required for triadin binding to calsequestrin but that only one of several KEKE motifs of triadin is essential for binding to occur. Alanine mutagenesis suggests that the charged residues of triadin interacting with calsequestrin form a ␤-strand, in which only one face of the putative ␤-strand contributes amino acid side chains that bind to calsequestrin.

EXPERIMENTAL PROCEDURES
Materials-Restriction enzymes were purchased from New England Biolabs. Growth media constituents for plasmid propagation in JM109 Escherichia coli and protein induction in BL21 (DE3) E. coli were purchased from Fisher and Sigma. Pefabloc® was obtained from Roche Molecular Biochemicals. All other reagents were from Sigma unless otherwise stated. All PCR products were generated using AmpliTaq® DNA polymerase (Perkin-Elmer), and the fidelity of all constructs was confirmed by nucleic acid sequencing using Sequenase TM (Amersham Pharmacia Biotech). All oligonucleotides were purchased from Life Technologies, Inc. Radioactive reagents were obtained from NEN Life Science Products.
Generation and Purification of Radiolabeled Calsequestrin-cDNA encoding canine cardiac calsequestrin without its signal sequence was PCR-amplified from clone IC3A (17), using a 5Ј primer with an NdeI restriction enzyme site, and a 3Ј primer with a BamHI restriction enzyme site, then subcloned into the pET5a vector (Novagen). The calsequestrin/pET5a construct produced encodes mature cardiac calsequestrin (amino acid residues 1-391) (17) with an amino-terminal methionine contributed by the NdeI restriction enzyme site. (Amino acid residues 1-42 of the purified, recombinant protein were verified by Edman sequencing.) The calsequestrin/pET5a construct was transformed into BL21 (DE3) E. coli cells (Novagen), and calsequestrin was metabolically labeled in E. coli and purified as follows.
A fresh colony of BL21 (DE3) cells carrying the calsequestrin/pET5a construct was inoculated into 10 ml of Luria broth supplemented with 125 g/ml ampicillin and grown at 37°C to an A 600 ϭ 0.6. Isopropyl-␤-D-thiogalactopyranoside was added at 0.5 mM to induce expression along with 10 mCi of 35 S-EasyTag TM (specific activity 1175.0 Ci/mmol, 11 mCi/ml). Protein induction proceeded for 3 h at 37°C with constant shaking. Cells were harvested and washed in cold phosphate-buffered saline and then resuspended in 5.0 ml of lysis buffer containing 50 mM Tris-Cl (pH 7.5), 5 mM DTT, 1 mM EDTA, and 0.1 mg/ml lysozyme, along with the protease inhibitors aprotinin (1 g/ml), leupeptin (2 g/ml), Pefabloc TM (100 M), and benzamidine (1 mM). Cells were lysed on ice for 30 min and tip-sonicated with three 10-s bursts. The soluble fraction was separated from insoluble material by centrifugation at 50,000 ϫ g for 30 min at 4°C. 20 mM CaCl 2 was added to the cleared bacterial lysate to precipitate the expressed calsequestrin (19), and the sample was incubated on ice for 20 min followed by centrifugation at 10,000 ϫ g for 10 min at 4°C. The Ca 2ϩ -precipitated pellet of calsequestrin was clarified by resuspension in 5 ml of buffer containing 20 mM MOPS (pH 7.2), 5 mM DTT, 10 mM EGTA, and 0.5 M NaCl and allowed to sit on ice for 10 min. The solubilized recombinant calsequestrin was then loaded over a 1-ml phenyl-Sepharose column pre-equilibrated with 20 mM MOPS (pH 7.2), 5 mM DTT, 1 mM EGTA, and 0.5 M NaCl, and the column was washed with 8 column volumes of the same buffer. Bound, metabolically labeled calsequestrin was subsequently eluted with 5 column volumes of elution buffer containing 20 mM MOPS (pH 7.2), 1 mM DTT, 10 mM CaCl 2 , and 0.5 M NaCl (20). Elution fractions containing purified, recombinant calsequestrin were pooled and concentrated. The concentrated, 35 S-labeled calsequestrin was dialyzed against 20 mM MOPS (pH 7.2), 150 mM NaCl and stored frozen in small aliquots.
Generation of Triadin Deletion Mutants-To localize the calsequestrin-binding site of canine triadin 1, different triadin domains and subdomains with progressive amino-and carboxyl-terminal deletions were produced as GST fusion proteins. GST fusion protein constructs were amplified by PCR using a series of oligonucleotides complementary to targeted regions of canine triadin 1 cDNA (11). Sense and antisense oligonucleotides were engineered with 5Ј EcoRI and XhoI sites, respectively, to facilitate cloning into pGex4T-1. Constructs were transformed into BL21 (DE3) E. coli cells and grown overnight in Luria broth supplemented with 125 g/ml ampicillin. Overnight cultures were washed two times, diluted 10-fold, and grown to A 600 ϭ 0.6 at 37°C. Optimal induction conditions to maximize expression with the least amount of protein degradation were determined empirically for each fusion protein. Protein expression proceeded for either 1-3 h at 37°C or overnight at 22°C with 0.1-0.5 mM isopropyl-␤-D-thiogalactopyranoside for induction of expression. Cultures were harvested by centrifugation at 5,000 ϫ g for 10 min, and cells were resuspended for 30 min at room temperature in 20 mM Tris-Cl (pH 7.5), 1 mM EDTA, 1 mM DTT, 1% Triton X-100, 0.2 mg/ml lysozyme, and the protease inhibitors described above. Lysates were sonicated with three 10-s bursts on ice and then centrifuged at 10,000 ϫ g for 10 min. Solubilized fusion proteins were purified using glutathione-Sepharose 4B chromatography (Amersham Pharmacia Biotech) (10). Protein concentrations were determined by the method of Schaffner and Weissmann (21).
Site-directed Mutagenesis of Amino Acids within the Calsequestrinbinding Domain-The GST fusion protein construct encoding residues 178 -224 of canine triadin 1 was produced as described above. Using this fusion protein construct as template, mutations encoding alanine substitutions between residues 200 and 224 of triadin were generated by PCR following the methodology described in Kobayashi and Jones (11). Residues 178 -224 of triadin with the targeted alanine mutations were then expressed and purified as a series of GST fusion proteins. Amino acid residues of triadin 1 that would appear on either side of the calsequestrin-binding domain when plotted on a ␤-strand were mutated to alanine.
Generation of the ⌬CBD/Triadin 1 Construct-Production of cDNA encoding canine cardiac triadin 1 with amino acid residues 200 -224 deleted was done by a three-step PCR procedure. The first PCR reaction produced a product that encoded amino acid residues 1-230 of triadin 1 but with the coding sequence for residues 200 -224 deleted. This PCR reaction utilized sense-Primer 1 (11), which encoded the first 9 amino acids of triadin 1 and had an engineered 5Ј EcoRI restriction enzyme site just upstream of the Kozak initiation sequence and ATG start codon, and antisense-Primer 3Ј ⌬CBD (5Ј GACTTTTGCAGCTGTTG-GTGCCATTGTCTTTGTTTC 3Ј), which comprised the antisense codons for amino acids 194 -199 and 225-230 of triadin 1 but had the antisense codons for residues 200 -224 deleted. The second PCR reaction produced a product that encoded amino acids 194 -278 of triadin 1 with the coding sequence for residues 200 -224 deleted. This PCR reaction utilized sense-Primer 5Ј ⌬CBD (5Ј GAAACAAAGACAATGGCACCAA-CAGCTGCAAAAGTC 3Ј), which encoded amino acids 194 -199 and 225-230 of triadin 1 but had the coding sequence for residues 200 -224 deleted, and antisense-Primer 3Ј T1, which comprised the antisense codons for the last 8 amino acids of triadin 1, followed by the antisense stop codon, and then an engineered 5Ј XhoI restriction enzyme site just downstream of the TGA stop codon. The third PCR reaction used the products from the first and second PCR reactions and Primer 1 and Primer 3Ј T1 to amplify the cDNA encoding the complete amino acid sequence of triadin 1 (amino acid residues 1-278) but with the calsequestrin-binding domain (residues 200 -224) deleted (⌬CBD/triadin 1). The final PCR product was digested with EcoRI and XhoI and then ligated into the baculovirus transfer vector pBlueBac4.5 to produce the ⌬CBD/pBB4.5 construct, which encoded ⌬CBD/triadin 1.
Expression and Purification of Recombinant Triadin from Insect Cells-The ⌬CBD/pBB4.5 construct was co-transfected with Bac-N-Blue linearized baculovirus DNA (Invitrogen) into Sf21 insect cells (Invitrogen) according to manufacturer's specifications. Isolation and amplification of the recombinant baculovirus encoding ⌬CBD/triadin 1 was performed as described in baculovirus protocols (22). Recombinant wild-type triadin 1 and ⌬CBD/triadin 1 were expressed in Sf21 cells and purified from Triton X-100-solubilized microsomes by phosphocellulose chromatography as recently described (11).

35
S-Calsequestrin Overlay Assay-A filter overlay assay (6,10,18) was used to localize the calsequestrin-binding domain of triadin 1. 2 g of different purified GST-triadin fusion proteins or of full-length recombinant triadin 1 were separated by SDS-PAGE and transferred to nitrocellulose sheets. Blots were blocked for 1 h with 1% horse hemoglobin dissolved in 20 mM MOPS, 150 mM KCl (pH 7.2), followed by three 5-min washes with the same buffer except in the absence of hemoglobin. 35 S-Labeled, recombinant cardiac calsequestrin was added to each blot to a final concentration of 0.0125 M in 20 mM MOPS (pH 7.2), 150 mM KCl, and 1 mM EGTA. Blots were allowed to air dry and were directly exposed to autoradiographic film at room temperature.
SDS-PAGE, Immunoblotting, and Antibodies-SDS-PAGE was conducted according to Laemmli (23) or Porzio and Pearson (24) using 10 or 8% polyacrylamide, respectively. Immunoblotting with the T 1 -specific antibody was performed as recently described using 125 I-protein A for antibody detection (11).

Expression and Purification of Metabolically Labeled
Calsequestrin-Previously, iodinated calsequestrin was used to identify calsequestrin-binding proteins in junctional SR vesicles isolated from heart (6,18). However, iodination may oxidize amino acids and produce a heterogeneous pool of labeled product. Since the goal here was to localize and characterize the calsequestrin-binding domain of triadin, it was important to ensure that a high quality radiolabeled calsequestrin preparation was utilized for the studies. To this end, canine cardiac calsequestrin was expressed and metabolically labeled at cysteine and methionine residues in E. coli and then purified to homogeneity in two easy steps. Five criteria were used to verify that the metabolically labeled, recombinant calsequestrin was structurally and functionally equivalent to its native counterpart as follows: 1) the inherent ability of calsequestrin to aggregate in its Ca 2ϩ -bound state (18,19,25); 2) the ability of calsequestrin to undergo a conformational change, internalizing a hydrophobic site and exposing hydrophilic sites in the presence of Ca 2ϩ (17,18,20); 3) the solubility of recombinant calsequestrin; 4) the anomalous electrophoretic mobility of cardiac calsequestrin on SDS-PAGE (17); and 5) the characteristic "blue staining" of calsequestrin by the dye Stains-all (26).
Recombinant calsequestrin was readily expressed and purified. The protein was extracted from E. coli without use of detergents (Fig. 1A, Sup), indicating that it was properly folded and soluble. Ca 2ϩ in millimolar concentrations selectively precipitated the recombinant protein, purifying it to near homogeneity in one step (19) (Fig. 1A, Ca 2ϩ pel). The precipitated 35 S-calsequestrin was resolubilized in Ca 2ϩ -free buffer and then purified to homogeneity in one additional step, employing Ca 2ϩ -dependent elution from phenyl-Sepharose (20) (Fig. 1A, lanes [5][6][7][8][9]. Successful purification with phenyl-Sepharose shows that the hydrophobic site of the protein is exposed at low Ca 2ϩ concentration and internalized at high Ca 2ϩ concentration, one of the hallmarks of native calsequestrin (17,18,20). Recombinant calsequestrin migrated with identical mobility on SDS-PAGE as native calsequestrin in cardiac SR vesicles, and the intense blue staining by Stains-all was preserved (Fig. 1B). The apparent molecular mass of the protein on SDS-PAGE, 55 kDa, was substantially greater than the calculated molecular mass of 45 kDa (17), demonstrating that the anomalous electrophoretic mobility of calsequestrin was also retained by the recombinant protein. 10 l (48.6 pmol) of the purified 35 Scalsequestrin was run on an SDS-PAGE gel and exposed to film (Fig. 1C), demonstrating the purity of the probe used in the overlay assays described below. 35 S-Calsequestrin Binding to Native and Recombinant Tria-din 1-Binding of 35 S-calsequestrin to native triadin 1 in SR vesicles and to purified recombinant triadin 1 was demonstrated by use of the calsequestrin overlay method (6,18). Canine cardiac junctional SR proteins and purified triadin 1 were separated by SDS-PAGE, transferred to nitrocellulose, and blots were overlaid with 35 S-calsequestrin. Three major 35 S-calsequestrin-binding proteins were detected in cardiac SR vesicles (Fig. 2, 1st lane) as follows: a 26-kDa protein or junctin (6) (asterisk), and a doublet of 35-and 40-kDa molecular mass proteins (arrows), corresponding to the deglycosylated and glycosylated () forms of triadin 1 (T1), respectively (11). These same major calsequestrin-binding proteins in junctional SR vesicles were previously identified by probing overlays with native cardiac calsequestrin that had been iodinated (6,18). The identities of the 35-and 40-kDa molecular mass proteins as triadin 1 were confirmed with use of the T 1 -specific antibody (lane 2), which was raised to a peptide with the same sequence as the carboxyl-terminal 17 residues of triadin 1 (11). Recombinant canine triadin 1 purified from insect cells also bound 35 S-calsequestrin strongly (3rd lane, arrows) and was recognized by the T 1 -specific antibody (4th lane). The two glycosylated forms of triadin 1 () expressed by Sf21 cells were noted previously (11) and also bound 35 S-calsequestrin. Localization of the Calsequestrin-binding Region of Triadin 1-Three major regions of canine triadin 1, excluding the transmembrane domain, were expressed and purified as GST fusion proteins to determine which domain binds 35 S-calsequestrin (Fig. 3A). Fig. 3B shows a Coomassie Blue-stained gel of the expressed and purified fusion proteins along with purified recombinant triadin 1, and Fig. 3C shows the results of the 35 S-calsequestrin overlay, following SDS-PAGE and transfer of identical samples to nitrocellulose. Neither the amino-terminal cytoplasmic domain common to all triadins (Fig. 3A, residues 2-47) nor the unique carboxyl terminus of triadin 1 (Fig. 3A,  residues 258 -278) bound 35 S-calsequestrin (Fig. 3C, lanes 3  and 5, respectively). The GST peptide expressed by itself also did not bind 35 S-calsequestrin (lane 2). The lumenal domain common to all triadins (Fig. 3A, residues 69 -257), in contrast, bound 35 S-calsequestrin strongly (Fig. 3C, lane 4), as did the purified recombinant protein (Fig. 3C, lane 1). These results concur with previous findings using a fusion protein pull-down assay, which showed that the calsequestrin-binding domain of rabbit triadin is located between residues 69 and 264 (7,9).
The common lumenal domain of canine triadin 1 (residues 69 -257) has several KEKE motifs (Fig. 4), defined as regions of sequence greater than 12 residues in length, with more than 60% Lys and Glu/Asp residues, and lacking five positively or negatively charged residues in a row (13). However, the proximal half of this region of triadin is more negatively charged than the distal half and contains only one KEKE motif (Fig. 4). To see which subdomain of this region binds 35 S-calsequestrin,

FIG. 2. Binding of 35 S-calsequestrin to proteins in canine cardiac junctional SR vesicles (JSR) and to purified recombinant triadin 1 (rec. T-1).
JSR proteins and rec. T1 were separated on SDS-PAGE and transferred to nitrocellulose. The 1st and 3rd lanes show blots overlaid with 35 S-calsequestrin ( 35 S-CSQ). The 2nd and 4th lanes show blots overlaid with the T 1 -specific antibody (Ab), which recognizes the unique carboxyl terminus of triadin 1 (T1). Arrows designate triadin 1 mobility forms. designates glycosylated forms of triadin 1. The asterisk indicates the location of junctin.
residues 69 -164 and 165-257 of triadin 1 were expressed separately (Fig. 5A) and purified as GST fusion proteins (Fig. 5B). 35 S-Calsequestrin overlay assay (Fig. 5C)  To resolve the calsequestrin-binding domain with greater precision, residues 188 -257 of triadin were progressively deleted from the carboxyl-terminal end, holding residues 165-187 constant and anchored to GST (Fig. 6A). This approach should delimit the amino terminus of the calsequestrin-binding domain. The fusion proteins were expressed and purified (Fig.  6B) and then used for overlay assay (Fig. 6C). Analysis of the results showed that calsequestrin binding remained approximately constant when triadin residues 165-224 were retained by the fusion proteins (Fig. 6C, lanes 2-5) but dropped off precipitously with the fusion protein containing only residues 165-211 (lane 6), with no binding at all observed with the fusion protein containing residues 165-199 (lane 7). These results eliminated residues 165-199 of triadin from binding to calsequestrin and also excluded the two KEKE motifs in this region (Fig. 4) from interacting with calsequestrin.
A second series of fusion protein deletions was made to confirm the carboxyl-terminal end of the calsequestrin-binding domain of triadin. This time residues 165-238 of triadin were progressively deleted from the amino-terminal side, holding the carboxyl terminus constant (Fig. 7A). Overlay assay (Fig.  7C) of the purified fusion proteins (Fig. 7B) showed that only the two shortest fusion peptides tested (lanes 7 and 8) did not bind 35 S-calsequestrin. This excluded residues 225-257 of triadin from binding to calsequestrin. Combining the data in Fig. 3 and Figs. 5-7, the calsequestrin-binding region of canine triadin is localized to amino acid residues 200 -224, a select run of only 25 amino acids. This region also contains a KEKE motif. The high concentration of lysine and glutamic acid here is conserved across several mammalian species, especially upon aligning residues 208 -224 of canine triadin with the homologous triadin regions from the other species (Fig. 8).
Lack of Binding of 35 S-Calsequestrin to ⌬CBD/T1-If the fusion protein analyses accurately localized the calsequestrinbinding domain of triadin to a single site at residues 200 -224, then triadin 1 with these amino acids deleted (⌬CBD/T1) should not bind calsequestrin. This was tested by expressing and purifying ⌬CBD/T1 from Sf21 cells and then checking for the ability of the mutated protein to bind 35 S-calsequestrin. Immunoblotting with the T 1 -specific antibody revealed that ⌬CBD/T1 was purified by phosphocellulose chromatography like native triadin 1 and also appeared to be partially glycosylated as evidenced by the endoglycosidase H-induced mobility shift (Fig. 9A). However, overlay assay demonstrated that ⌬CBD/T1 did not bind 35 Scalsequestrin (Fig. 9B), confirming that triadin 1 contains a single calsequestrin-binding site at amino acids 200 -224.
Evidence That a ␤-Strand of Triadin Mediates Protein Binding to Calsequestrin-When residues 200 -224 of triadin are plotted on a ␤-strand, most of the KEKE residues line up on one side of the strand (Fig. 10A). This asymmetry could allow the KEKE residues on the more polar face of the strand to form hydrogen bonds with charged residues of calsequestrin, thus linking the two proteins together (15). To test this hypothesis, 4-residue blocks of amino acids on either side of the putative ␤-strand of triadin were mutated to alanine. Fusion proteins carrying the mutations were isolated (Fig. 10B) and tested for the ability to bind 35 S-calsequestrin (Fig. 10C). Considering the even-numbered residue mutations, the first fusion protein tested (carrying the E202A, K204A, K206A, and E208A mutations) (lane 2) exhibited an increased level of 35 S-calsequestrin binding, relative to the binding obtained with the fusion protein with no mutations in this region (lane 1). However, the next two fusion proteins tested, which had downstream alanine substitutions on this side of the putative ␤-strand (K210A, K212A, E214A, and K216A) (lane 3) and (G218A, Q220A, K222A, and K224A) (lane 4), were incapable of binding 35 Scalsequestrin. The next series of fusion proteins investigated contained the odd-numbered residue mutations that would appear on the opposite side of the putative ␤-strand (Fig. 10A). All of the fusion proteins encompassing this region ((E201A, R203A, and T207A; lane 5) (E209A, I211A, K213A, and V215A; lane 6); and (G217A, K219A, E221A, and V223A; lane 7)) bound 35 S-calsequestrin as well as the fusion protein with no mutations (lane 1) or as well as intact triadin 1 (Rec. T1) (Fig. 10C).
These results provide strong evidence that the calsequestrinbinding domain of triadin is a ␤-strand. The amino acids of FIG. 4. Amino acid sequence of the common lumenal domain of canine triadin. Amino acids 69 -257 of canine triadin (11) are listed in single letter abbreviations. Residue numbers are indicated in the right margin. The even-numbered amino acids essential for triadin binding to calsequestrin are gray-shaded.

FIG. 5. 35 S-Calsequestrin binding to triadin 1 lumenal subdomains.
A, schematic, formatted as described in the legend to Fig. 3, of the two subdomains analyzed. GST fusion protein 2 contained residues 69 -164 of canine triadin 1, and GST fusion protein 3 contained residues 165-257 of canine triadin 1. B, Coomassie Blue-stained gel of purified recombinant triadin 1 (lane 1) and the two fusion proteins (lanes 2 and  3). C, 35 S-calsequestrin overlay of the same proteins. triadin 1 most critical to formation of the calsequestrin-binding site are the even-numbered, mostly charged residues located between residues 210 and 224 (Fig. 4). DISCUSSION In this study we have used deletional and site-specific mutagenesis to localize the calsequestrin-binding domain of triadin 1. Only a single binding site was found, located at residues 210 -224 of triadin. Here, replacement of Lys 210 , Lys 212 , Glu 214 , and Lys 216 by alanine or replacement of Gly 218 , Gln 220 , Lys 222 , and Lys 224 by alanine abolished the ability of triadin to bind calsequestrin. The results are consistent with a model in which residues 210 -224 of triadin form a ␤-strand. The even-numbered, mostly charged residues above may line up along one side of the ␤-strand to interact directly with charged residues of calsequestrin (17), forming a "polar zipper" (15) that links the two proteins together. Remarkably, replacement of charged residues on the opposite side of the proposed ␤-strand by alanine did not affect calsequestrin binding to triadin at all, demonstrating that only one face of the strand contributes amino acid side chains that are essential for binding to calsequestrin. Notably, all of the even-numbered, charged residues above are completely conserved across several mammalian species (Fig.  8), suggesting that they comprise a critical interaction site for calsequestrin binding to all triadins.
KEKE motifs have been defined as short runs of amino acid sequence greater than 12 residues in length, which contain at least 60% alternating lysine and glutamic/aspartic acid residues, lack five positively or negatively charged residues in a row, and are devoid of tryptophan, tyrosine, phenylalanine, and proline (13). Such KEKE motifs have been implicated in a variety of different types of protein associations (13)(14)(15)(16). Since triadin (and junctin) are replete with KEKE motifs throughout their lumenal domains (Fig. 4), we strongly suspected in initiating these studies that several charged sites of triadin, conforming to KEKE motifs, would be required for tight binding to calsequestrin. We were very surprised to find, however, that out of the numerous charged residues located in the lumenal domain of triadin, only six (Lys 210 , Lys 212 , Glu 214 , Lys 216 , Lys 222 , and Lys 224 ) were critical for binding to calsequestrin. This demonstrates that the triadin-calsequestrin binding interaction is very specific and is not simply the result of nonspecific electrostatic interactions between two oppositely charged, highly polar protein molecules (6,17). Residues 200 -224 of triadin satisfy all of the criteria for a KEKE motif (13). Although it has been suggested the KEKE motifs are predominantly ␣-helical (13, 16), our results clearly suggest that the calsequestrin-binding region of triadin is a ␤-strand.
The ⌬CBD/triadin 1 mutant, which is full-length recombinant triadin 1 with residues 200 -224 deleted, failed to bind to calsequestrin in the overlay assay (Fig. 9). Moreover, several small fusion peptides, which shared only residues 200 -224 of triadin 1, bound calsequestrin as well as full-length triadin 1 (Figs. 6 and 7). This demonstrates that only a few key amino acids, located between residues 200 and 224 of triadin, are necessary and sufficient for binding to calsequestrin. The specific loss of calsequestrin binding to this region, when charged residues along one side of the putative ␤-strand were changed to alanine, suggests that the charged side chains of these residues form hydrogen bonds with charged side chains of calsequestrin, stabilizing a ␤-sheet, which may zip the two proteins together (15). It should be noted, however, that the results of Fig. 10 call into question the requirement for a rigorously defined KEKE motif, at least as originally proposed (13), as necessary for triadin binding to calsequestrin. For example, fusion peptides 6 and 7, which had charged residues on the opposite side of the putative ␤-strand changed to alanine, bound calsequestrin as well as intact triadin 1. However, these two fusion peptides no longer contained a KEKE motif between residues 200 and 224, because they retained only 56% charged residues here. Similarly, fusion peptide 2 in Fig. 10, which had four closely spaced KEKE-charged residues replaced by alanine (residues 202, 204, 206, and 208), actually bound 35 S-calsequestrin more strongly than did full-length triadin 1. Only fusion peptides 4 and 5, which had blocks of even-numbered charged residues removed between residues 210 and 224, failed to bind 35 S-calsequestrin. Although the atomic structure of triadin 1 is not known, the results of Fig. 10 strongly suggest that the calsequestrin-binding domain of triadin is a ␤-strand.
Junctin is another calsequestrin-binding protein that is homologous to triadin and has similar KEKE association motifs scattered throughout its lumenal domain (6,10). However, when similar deletional mutagenesis-analysis was done with junctin, we observed that deletion of any one of several widely separated segments of its lumenal domain was sufficient to abolish calsequestrin binding (data not shown). It thus appears that junctin does not have a single discrete binding site for calsequestrin, which distinguishes it from triadin. The region of calsequestrin binding to triadin (and junctin) has not been identified. Cardiac calsequestrin does not have any consensus KEKE motifs but does have numerous small clusters of lysine and glutamic acid residues dispersed throughout its sequence (17). Thus, the concept of triadin binding to calsequestrin, stabilized by a polar zipper (15), seems a viable one. Recent crystallization of skeletal muscle calsequestrin showed that negatively charged sequences of calsequestrin are arranged into platforms of negative charges (28). Since amino acids 200 -224 of triadin are sufficient for binding to calsequestrin, it should be feasible to co-crystallize a triadin synthetic peptide of these residues with purified recombinant calsequestrin, in order to determine the three-dimensional structure of the binding complex. These experiments are currently in progress.
In other experiments we observed that the interaction between the calsequestrin-binding site of triadin and calsequestrin was enhanced at low Ca 2ϩ concentration, 3 as has been reported previously for intact triadin (6,10,18). Similar Ca 2ϩ concentrations may occur intralumenally in junctional SR following Ca 2ϩ release (29). One of the functions of triadin may be to anchor calsequestrin to the junctional membrane in proximity to the RyR, such that during muscle relaxation, when Ca 2ϩ levels in the SR are beginning to rise, Ca 2ϩ will be concentrated near the RyR where calsequestrin is tethered. When calsequestrin becomes saturated with Ca 2ϩ , its propensity to aggregate may help to keep it localized to the junctional SR lumenal space. Triadin 1 with the calsequestrin-binding domain deleted (⌬CBD/triadin 1) would make an interesting construct for overexpression in transgenic mouse hearts (11) to test for a possible dominant-negative effect on excitation-contraction coupling. Conversely, triadin 1 with enhanced calsequestrin binding, as achieved by alanine mutagenesis of the proximal end of the calsequestrin-binding domain, would be an interesting construct to overexpress to test for a possible dominant-positive effect. It remains to be investigated whether these mutations of triadin alter its binding to the RyR (6,7,9).