Functional importance of polymerization and localization of calsequestrin in C. elegans.

Dual roles of calsequestrin (CSQ-1) being the Ca2+ donor and Ca2+ acceptor make it an excellent Ca2+-buffering protein within the sarcoplasmic reticulum (SR). We have isolated and characterized a calsequestrin (csq-1)-null mutant in Caenorhabditis elegans. To our surprise, this mutant csq-1(jh109) showed no gross defects in muscle development or function but, however, is highly sensitive to perturbation of Ca2+ homeostasis. By taking advantage of the viable null mutant, we investigated the domains of CSQ-1 that are important for polymerization and cellular localization, and required for its correct buffering functions. In transgenic animals rescued with various CSQ-1 constructs, the in vivo patterns of polymerization and localization of several mutated calsequestrins were observed to correlate with the structure-function relationship. Our results suggest that polymerization of CSQ-1 is essential but not sufficient for correct cellular localization and function of CSQ-1. In addition, direct interaction between CSQ-1 and the ryanodine receptor (RyR) was found for the first time, suggesting that the cellular localization of CSQ-1 in C. elegans is indeed modulated by RyR through a physical interaction.


Introduction
Calsequestrin is a high-capacity (40-50 mol Ca 2+ mol -1 calsequestrin) and low-affinity Ca 2+ -binding protein (K d =40 M) localized in the lumen of the junctional sarcoplasmic reticulum (SR) of skeletal and cardiac muscles (Cozens and Reithmeier, 1984;Ikemoto et al., 1972;MacLennan and Wong, 1971). It stores Ca 2+ in sufficient quantities to allow repetitive contractions. Albeit its ability to sequester a large amount of Ca 2+ , being a low-affinity Ca 2+ -binding protein ensures the rapid delivery of Ca 2+ when the need arises. Thus calsequestrin efficiently functions as a Ca 2+ -buffering protein.
Overexpression of calsequestrin results in severe hypertrophy and heart failure in mice which is accompanied with impaired Ca 2+ loading (Jones et al., 1998;Sato et al., 1998;Wang et al., 2000). A missense mutation in calsequestrin resulting in a single amino acid substitution (D307H) results in a genetic disease called catecholaminergic polymorphic ventricular tachycardia (Lahat et al., 2001).
In mammals, calsequestrin protein is in physical proximity with the ryanodine receptor (RyR) through its interaction with triadin (Guo and Campbell, 1995) and junctin (Zhang et al., 1997). In our previous paper, we reported that the RyR protein UNC-68 appeared to affect the correct localization of CSQ-1 in vivo (Cho et al., 2000).
The crystal structure of rabbit skeletal calsequestrin shows that calsequestrin consists of three domains (I, II, and III), each with a topology similar to that of E. coli thioredoxin . The individual domains fold into an ␣-␤ structure with five ␤-strands in the core, flanked by four-␣-helices, two on each side of the sheet . The folding and compaction of calsequestrin is dependent on Ca 2+ and the ionic strength, and occurs optimally with Ca 2+ and K + concentrations close to those in the lumen of the SR in vivo (~1 mM Ca 2+ and 100 mM K + ). Crystal structures show that calsequestrin forms two different dimers: front-to-front and back-to-back (Gatti et al., 2001;Wang et al., 1998). In the back-to-back interaction, the C-terminal segments form a negatively charged pocket within the dimer that favors Ca 2+ binding. The front-to-front interaction also generates a negatively charged pocket that supports Ca 2+ binding and, like the back-to-back dimer, is likely to be favored by Ca 2+ binding. With these two types of interactions, calsequestrin is predicted to form a continuous, linear polymer when Ca 2+ concentration is increased. In addition, it has been confirmed that both back-to-back (Cterminal) and front-to-front (N-terminal) interactions are necessary for calsequestrin polymerization (Park et al., 2003).
However, in vivo roles of either of the two interactions of calsequestrin have yet to be investigated in any of the conventional model organisms. In our previous study, we reported RNAi mediated knockdown phenotypes in C. elegans (Cho et al., 2000). Here, we report the isolation and characterization of the csq-1 deletion mutant csq-1(jh109) that Dual roles of calsequestrin (CSQ-1) being the Ca 2+ donor and Ca 2+ acceptor make it an excellent Ca 2+ -buffering protein within the sarcoplasmic reticulum (SR). We have isolated and characterized a calsequestrin (csq-1)-null mutant in Caenorhabditis elegans. To our surprise, this mutant csq-1(jh109) showed no gross defects in muscle development or function but, however, is highly sensitive to perturbation of Ca 2+ homeostasis. By taking advantage of the viable null mutant, we investigated the domains of CSQ-1 that are important for polymerization and cellular localization, and required for its correct buffering functions. In transgenic animals rescued with various CSQ-1 constructs, the in vivo patterns of polymerization and localization of several mutated calsequestrins were observed to correlate with the structure-function relationship. Our results suggest that polymerization of CSQ-1 is essential but not sufficient for correct cellular localization and function of CSQ-1. In addition, direct interaction between CSQ-1 and the ryanodine receptor (RyR) was found for the first time, suggesting that the cellular localization of CSQ-1 in C. elegans is indeed modulated by RyR through a physical interaction.
provides a valuable tool to study the structure-function relationship of CSQ-1 in vivo. Our results demonstrate that CSQ-1 polymerization in C. elegans is essential, but not sufficient for its correct localization and function in SR.

Results
Characterization of the calsequestrin knockout mutant csq-1(jh109) in C. elegans We have isolated the csq-1 deletion mutant csq-1(jh109) by removing several exons in the middle of the gene (Fig. 1A). When genomic DNA from a single worm was used as template, the PCR amplification product confirmed the isolation of a homozygote deletion mutant (Fig. 1B). To further evaluate the nature of this allele, we performed western blot analysis of protein extracts from wild-type (N2) worms and csq-1(jh109). As evidenced from Fig. 1C, no CSQ-1 protein was detected in C. elegans carrying the csq-1(jh109) mutation, verifying the isolated allele is functionally null.
To our surprise, csq-1(jh109) mutants were viable and showed no obvious muscle defects. This is consistent with our previous characterization of RNAi (RNA interference) knockdown phenotypes. We have shown that animals deficient of calsequestrin exhibit normal movement in body-wall muscle (Cho et al., 2000). Since CSQ-1 is abundantly expressed in body-wall muscles of C. elegans, we further investigated potential muscle-specific function of CSQ-1 by characterizing the csq-1(jh109) mutant. We found no obvious alteration in muscle functioning, as evidenced by the fact that both the wildtype and csq-1(jh109) animals moved with equal efficiency ( Fig. 2A). As expected, body-wall muscle structures examined by electron microscope showed normal sarcomere structures compared with wild-type animals (Fig. 2F,G). However, Ca 2+ signaling plays a crucial role in fertilization, embryonic development, post-embryonic development and contraction of vulval muscles, a prerequisite for the expulsion of embryos.
Based on the strong expression of CSQ-1 in vulval muscle (Cho et al., 2000), which was resistant to RNAi treatment, we sought to know whether there is a defect associated with the reproductive function in csq-1(jh109) mutant animals. However, as shown in Fig. 2B, the brood size of csq-1(jh109) animals was indistinguishable from that of wild-type animals, precluding the possible role of CSQ-1 in reproduction and/or embryonic development in C. elegans.
Since calsequestrin has been attributed to be a major Ca 2+ buffering protein in the SR (Fliegel et al., 1987;Scott et al., 1988), we reasoned that csq-1-null mutants might not acclimatize that well in an environment of either high or low Ca 2+ concentration [Ca 2+ ]. As expected, the survival rate of csq-1(jh109) was reduced dramatically to 50% compared with wild-type worms, when grown on plates with nematode growth medium (NGM) containing high [Ca 2+ ] (Fig. 2C). Furthermore, the survival rate of csq-1(jh109) was reduced to 40% when grown on plates containing the Ca 2+ chelator EGTA (Fig. 2D). Taken together, our results suggest that CSQ-1 functions as a major Ca 2+ -buffering protein in C. elegans.
In addition to its primary effect on chelating and thereby depleting Ca 2+ , EGTA is also known to elicit ER stress. Hence, our observation of decreased survivability of csq-1(jh109) on EGTA-containing plates could be due to either defective Ca 2+ buffering or ER stress. To rigorously discern between these two possibilities, we grew the worms on plates containing 1 mM DTT, a substance known to cause ER stress. Survival rates of csq-1(jh109) animals were slightly reduced compared with those of wild-type animals (Fig. 2E). However, crt-1(jh101) animals deficient in calreticulin CRT-1, an ER-resident chaperon protein, showed dramatically decreased survival (Fig.  2E), but survived relatively well on plates with excess Ca 2+ (Fig. 2C). Furthermore, the csq-1(jh109);crt-1(jh101) double mutant showed only additive defects under growing conditions with EGTA, but not when DTT was present. Taken together, our results suggest that the primary cause for decreased survivability of csq-1 mutants on EGTA plates is not ER stress but ineffective Ca 2+ buffering ability.
In vivo analysis of polymerization and cellular localization of CSQ-1 within the SR The quaternary structure of calsequestrin changes from dimers to long, linear polymers as [Ca 2+ ] is increased. Two types of interaction between calsequestrin (front-to-front and back-toback interactions) could be related to this polymerization . Recently, this idea was exploited further by Park et al. using several mutant versions of canine calsequestrin analyzed with respect to their property of polymerization in vitro (Park et al., 2003). Since a viable C. elegans knockout mutant of calsequestrin provides an excellent in vivo model, we were interested in how the subtle changes in the predicted intermolecular contact regions of CSQ-1 are translated into differential CSQ-1 polymerization and cellular localization of CSQ-1 in the SR compartment.
In rabbit skeletal-muscle calsequestrin, stable formation of the calsequestrin dimer is necessary for the subsequent polymerization process . From a structural point of view , an interaction between serine acidic hydrophobic (SAH) residues of one calsequestrin with the dibasic hydrophobic (DBH) of the other appears to be important for back-to-back interaction. To have a schematic Journal of Cell Science 120 (9) and csq-1(jh109) animals (lane 2). Structure-function relationship of C. elegans CSQ-1 model of C. elegans CSQ-1, we performed homology modeling of C. elegans CSQ-1 by using rabbit calsequestrin as a template (Fig. 3A). In this model, the back-to-back interface has a basic residue, Lys111, that is highly conserved among all three calsequestrins (Fig. 3B,C). The side chain of Lys111 was found to be protruding out towards the interfacial region. Interestingly, three acidic residues (Asp196, Glu247 and Glu251) of the interacting CSQ-1 were found to be in the vicinity of Lys111 (Fig. 3B). Our modeling suggests that the interaction of Lys111 of one CSQ-1 with some or all of the three acidic residues of the other CSQ-1 at the interface may be essential for the formation of a stable dimer. To test our hypothesis, we changed Lys111 to Ala by site-directed mutagenesis. This mutation completely changed the cellular localization of CSQ-1 and appears to be non-functional (Fig.  4G,I). To rule out general misfolding of the CSQ-1(K111A) mutant protein, we also expressed CSQ-1(K111Q) in the csq-1(jh109) mutant. It showed a diffused localization throughout the muscle similar to the localization of CSQ-1(K111A) (see Fig. S2C,D in supplementary material). Our result indicates that back-to-back interaction mediated by a possible salt bridge of Lys111 is essential for the dimer formation. Hence, the K111A mutation in CSQ-1 allowed neither polymerization nor proper localization (adjacent to RyR), as evidenced by its diffuse distribution. We therefore conclude that polymerization is essential for the correct localization of CSQ-1.
The C-terminal residues are not essential for polymerization but important for correct localization The in vitro light-scattering experiment with canine mutant calsequestrin (⌬C27, deletion of C-terminal 27 residues) showed Ca 2+ -dependent dimerization of ⌬C27. However, they failed to form polymers like wild-type protein (Park et al., 2003). The exact mode of interaction (front-to-front vs back-to-back) has not been unequivocally determined . By contrast, we found no evidence for any defect in the oligomerization of the corresponding mutant in C. elegans. Instead, our results suggests that the C-terminal 26 amino acids are necessary for correct localization of CSQ-1 to the proximity of RyR. The staining pattern of csq-1-null mutants expressing either the ⌬C26mutant form or a ⌬C17 mutant were indistinguishable from that of the RyRnull mutant [unc-68(e540)] stained with anti-CSQ-1 antibody (Fig. 4D,H). This result indicates that C-terminal residues, at least up to the last 26 amino acids, are not essential for polymerization per se; rather, they are important for correct localization of CSQ-1 to the proximity of RyR.
For the correct front-to-front interaction, 13 N-terminal residues were found to be essential in canine calsequestrin (Park et al., 2003). C. elegans CSQ-1 has a similar sequence (residues 32 to 43). Proximal to this region, we also found seven additional amino acids that seem to be unique to C. elegans CSQ-1 (see gap in sequence alignment of Fig. 3C). To resolve which of these two regions are essential for the frontto-front interaction, we made two mutant constructs: ⌬N23-31 and ⌬N32-43, lacking aa residues at positions 23 to 31 and 32 to 43, respectively. Immunostaining of csq-1 mutants expressing construct ⌬N23-31 was indistinguishable (data not shown) from that of wild-type animals (Fig. 4E): typical sievelike patterns were observed on each case. This result indicates that the N-terminal residues 23-31 are not essential for polymerization or localization of CSQ-1. Interestingly, staining of the ⌬N32-43 mutant CSQ-1 proteins appeared as random aggregates (Fig. 4F) that were not localized properly, suggesting CSQ-1 precipitation in the SR. Note that the observed precipitation is probably not due to the retention of CSQ-1 to the cytoplasm, because the mutant harbors an intact signal sequence. We conclude that correct linear polymerization is essential for correct localization of CSQ-1 in vivo.

Functional relevance of CSQ-1 polymerization and localization
To better correlate the differences in the polymerization and localization of CSQ-1 with the efficiency of protein-activity, we injected different mutant constructs into csq-1(jh109) animals that were instrumental in studying the structurefunction relationship of calsequestrin. Sensitivity of the csq-1(jh109) mutant worm towards slight perturbations in [Ca 2+ ] was fully reversed by injection of the wild-type csq-1 gene (Fig. 4I). Similarly, constructs ⌬N23-31 and the Bedouin mutations (missense mutations of human CASQ2) also fully rescued CSQ-1 function in the csq-1-null mutant phenotype, which was not surprising because these mutated proteins still polymerized and localized correctly. Interestingly, mutant constructs ⌬C17 as well as ⌬C26 failed to rescue CSQ-1 function in the csq-1-null mutant phenotype (Fig. 4I). These two mutant proteins, although polymerized correctly, failed to display wild-type localization. As mentioned above, polymerization is essential but not sufficient for correct localization. We, therefore, suggest that polymerization, although necessary, is not sufficient for correct functioning of CSQ-1. The K111A mutant is futile in terms of its rescuing ability, underscoring the importance of back-to-back interaction in polymerization, localization and function.
Journal of Cell Science 120 (9) Fig. 3. Homology model and schematic diagram of mutated C. elegans calsequestrin. (A) Schematic prediction of polymerized CSQ-1. The monomeric model of CSQ-1 was built using the rabbit calsequestrin structure. The tetramer is shown as C ␣ -traces to display putative polymerizing interactions between monomers. Monomers are either green or yellow. N and C-termini of each molecule are shown as dashed lines because they are disordered in the template structure. Note the positively charged residues in the C-terminus specific to C. elegans calsequestrins. The location of the conserved back-to-back interface formed between SAH (Serine Acidic Hydrophbic) and DBH (Dibasic Hydrophobic) is boxed in red. Lys111 residues are shown as spheres. (B) Close-up view of the back-to-back interface of CSQ-1. The two interacting molecules are green and yellow. Three negatively charged residues (Glu247, Glu251 and Asp196) are placed in the vicinity of Lys111 and might participate in a salt bridge. (C) Amino acid sequence of C. elegans calsequestrin aligned with the rabbit skeletal calsequestrin and human cardiac calsequestrin CSQ2 amino acid sequences. Green box, signal sequence; first to fourth red and first and second yellow boxes indicate ⌬C26 and ⌬C17, ⌬N23-31, ⌬N32-43, DBH, SAH, respectively. (D) Each DNA construct represents wild-type (WT), N-terminal 9-amino-acid truncation, N-terminal 12-amino-acid truncation, C-terminal 17-amino-acid truncation and C-terminal 26amino-acid truncation. Mutant CSQ-1 constructs K111A and Bedouin contain a Lys111 to Ala and Asp318 to His mutation, respectively. These constructs were introduced into the csq-1(jh109) mutant and their expression was detected by western blotting using anti-CSQ-1 and anti-PYP-1 (internal control) antibodies. Structure-function relationship of C. elegans CSQ-1 Conservation of Lys111 across the calsequestrins from different species indicates the existence of high selectionpressure over this residue.
Direct interaction between the Cterminus of CSQ-1 and intraluminal loop regions of RyR in C. elegans Previously, we have reported that RyR affects the localization of CSQ-1 in C. elegans (Cho et al., 2000). Calsequestrin in mammals has been shown to regulate the RyR through triadin and junctin (Guo and Campbell, 1995;Zhang et al., 1997). But no homolog of triadin and junctin has (so far) been found in C. elegans. According to our homology modeling, the C-terminal region of CSQ-1 that contains positive charges might directly interact electrostatically with the intraluminal loops of RyR. Therefore, to investigate the nature of interaction between CSQ-1 and RyR, an in vitro binding assay was performed at different concentrations (0-2 mM) of Ca 2+ . As shown in Fig. 5A, we found that CSQ-1 directly interacted with loop I and loop II of the RyR.
To identify the regions of CSQ that interact with the loop regions of the RyR, in vitro binding assay were performed using mutated CSQ-1 protein in which the Cterminal 17 amino acids were deleted. As shown Fig. 5B, the CSQ-1 deletion mutant (⌬C17-CSQ-1) failed to bind RyR loop regions, suggesting that the C-terminal (17 amino acids) region of CSQ-1 is essential for binding these loop regions. Since the RyR loop regions have contain negatively charged amino acids, positive charges in the Cterminal region of CSQ-1 might interact electrostatically with RyR loop regions and, therefore, stabilize any binding between CSQ-1 and RyR (as schematically presented in Fig. 5C).

Discussion
Consistent with RNAi-mediated knockdown phenotypes previously studied by us (Cho et al., 2000), csq-1-null mutants did not show an obvious phenotype. In our previous study, body-wall muscle expression of csq-1 was completely abolished by RNAi treatment. However, ineffective silencing of csq-1 transcripts in the vulval and pharyngeal muscle left the open question of potential vulva-specific or pharynx-specific functions we might have missed. We observed that the brood size of csq-1-null mutants (Fig. 2B), as well as pharyngeal pumping (data not shown), were comparable with those of wild-type animals. Furthermore, we did not observe any other defects associated with egg-laying or pharyngeal pumping. Our current data based on csq-1-knockout animals suggests that CSQ-1 is not required for vulval or pharyngeal muscle function. This is not surprising and, actually, similar to the recent finding that cardiac calsequestrin is not essential for providing sufficient Ca 2+ storage for normal function of cardiac muscle (Knollmann et al., 2006). Despite absent cardiac calsequestrin, calsequestrin-knockout mice maintain relatively normal Ca 2+ ] conditions. Young F1 transgenic worms were grown on NGM plates with 20 mM EGTA and 20 mM CaAc. Survival rates of worms with rescued CSQ-1 function was compared with that of wild-type animals at 20°C. Over 50 animals were tested for each data point of a single set of experiments, and each experiment was repeated three times. release and contractile function. Calreticulin is another endoplsmic reticulum (ER)-resident Ca 2+ -binding protein, and our study indicates that C. elegans calreticulin (crt-1) is also expressed both in body-wall muscle as well as in pharyngeal muscle (Park et al., 2001). Therefore, we investigated the expression levels of other Ca 2+ -binding proteins, such as calreticulin (CRT-1) and calnexin (CNX-1), in csq-1(jh109) mutants to determine whether a deficiency in CSQ-1 protein induces a compensatory overexpression of the other Ca 2+binding proteins. Protein levels of CRT-1 and CNX-1 were unchanged (data not shown). More interestingly, the levels CSQ-1 itself was unchanged in wild-type animals stressed by low [Ca 2+ ]. Taken together, CRT-1 and CNX-1 are probably not overexpressd to compensate for CSQ-1 function in the csq-1-null mutant. This is consistent with the study of calsequestrin knockout mice, where no apparent upregulation of other Ca 2+binding protein was detected (Knollmann et al., 2006).
However, the expression of SERCA1 protein has been reported to be reduced by up to 50% in CSQ-2 and CSQ-1-CSQ-2 knockdown myotubes (Wang et al., 2006). So we investigated the expression levels of the sarcoplasmic reticulum Ca 2+ ATPase homolog sca-1 (SCA-1; mammalian homolog: SERCA1) in csq-1-null mutants. To measure mRNA expression levels of sca-1 in csq-1-null mutants, we performed reverse transcriptase (RT)-PCR using two different C. elegans strains (N2 and csq-1(jh109)) grown on plates with NGM and NGM supplemented with EGTA. mRNA levels of sca-1 were significantly reduced in csq-1(jh109) mutants (supplementary material Fig. S1), which is consistent with a recent report (Wang et al., 2006). But, interestingly, the sca-1 mRNA levels in csq-1(jh109) grown on EGTA-supplemented plates were not significantly different (supplementary material Fig. S1). However, the mechanism which leads to the reduction of SERCA1 expression remains unexplained, and genetic dissection of C. elegans mutants might provide some clues in the future.
In vitro, calsequestrin forms a long, linear polymer as dimers (back-to-back and front-to-front dimer) (Park et al., 2003). However, whether such polymerization does indeed happen in vivo has so far not been investigated. We exploited the csq-1null mutant to directly visualize the status of CSQ-1 polymerization in C. elegans. We took two main parameters into consideration: (1) status of polymerization and, (2) RyRdependent localization of CSQ-1. We then correlated these with their functional property. In wild-type worms, CSQ-1 polymerized correctly (punctate and mesh-like pattern). However, in unc-68(e540) mutants, the typical sieve-like pattern was lost and CSQ-1 distribution appeared as random spots, suggesting a role for RyRs in CSQ-1 localization.
Mouse, rat and rabbit calsequestrins are known to bind to junctin and triadin. The C-terminal of triadin (especially the KEKE motif) binds with the second luminal loop of RyRs (Lee et al., 2004), thus bringing calsequestrin into the proximity of RyRs. Thus, RyRs define the localization of CSQ-1. The similar immunostaining pattern of CSQ-1 corroborates with our result. Since wild-type CSQ-1 failed to form the typical sieve-like pattern in RyR mutants, the appearance of random spots indicates successful polymerization but failed correct localization. It is worthwhile to note that polymerization does not depend upon any other protein; purified calsequestrin successfully forms polymers at high [Ca 2+ ] in vitro (Park et al., 2003) and intraluminal [Ca 2+ ] is around 1 mM. Hence, wildtype CSQ-1 must have succeeded in polymerization and our observation of CSQ-1 appearing as random spots in RyR mutants is solely indicative of defective localization of CSQ-1. Furthermore, polymerization does not assure correct localization. In other words, polymerization and localization appear to be independent processes.
Immunostaining patterns of ⌬C26 and ⌬C17 CSQ-1 are similar to that of wild-type CSQ-1 in unc-68(e540) mutants, indicating that both ⌬C27 and ⌬C17 mutant proteins have succeeded to polymerize; however, they have failed to localize adjacent to RyRs. In vitro studies, though, carried out by Park et al. (Park et al., 2003) indicate that the canine calsequestrin mutants with C-terminal deletions failed to form polymers. The apparent discrepancy could be the manifestation of differences in the structure of the two calsequestrins. Alternatively, because the C-terminal region of the C. elegans CSQ-1 is unique, the function of this region may not be generalized to the mechanism of mammalian calsequestrin. Therefore, the importance of the C-terminal region for the polymerization of calsequestrin remains to be studied further. In addition, the Cterminus of C. elegans CSQ-1 does not contain as many Journal of Cell Science 120 (9) Positively charged residues at the C-terminal end of C. elegans calsequestrin may directly interact with the negatively charged residues of the loops of RyRs to localize calsequestrin into the vicinity of the SR membrane. This could allow fast and localized release of Ca 2+ through RyRs once the receptor is activated. negatively charged amino acids as mammalian calsequestrins; indeed, the C-terminus of CSQ-1 contains many positively charged amino acids. Hence, an overall negative charge, hypothesized to be essential for binding Ca 2+ and for Ca 2+mediated polymerization might not be relevant to C. elegans CSQ-1 owing to the existence of a relatively large number of positively charged residues. Based on our homology modeling, we have found that the C-terminal region of CSQ-1, which contains positive charges, interacts with intraluminal loops of RyR as seen in our in vitro binding assay. Therefore, Ca 2+ release from RyRs could be modulated by calsequestrin through a physical interaction in C. elegans.
Interaction between the SAH motif and the DBH motif is found in many protein-protein interactions (He et al., 1993;Mitchell et al., 1988). For instance, binding of calmodulin to its target is mediated by SAH-DBH interaction (Ikura et al., 1992;Meador et al., 1992;Meador et al., 1993). Structural analysis of rabbit calsequestrin shows the interaction of the SAH domain of one calsequestrin and the DBH domain of another calsequestrin molecule. Likewise, our homology modeling hinted to the possibility that the interaction between the SAH motif and the DBH motif is essential to stabilize backto-back interactions. The highly conserved Lys111 in the DBH motif was found to be the point of interaction with some or all three of the negatively charged residues in the back-to-back interaction. When mutant CSQ-1 (K111A point mutation) was expressed, a diffuse localization throughout the muscle was seen instead of a punctate appearance like in the wild-type polymerized CSQ-1 in unc-68 mutants. Thus, we believe that csq-1(K111A) mutants neither polymerized nor localized properly. However, to rule out a general misfolding of the CSQ-1(K111A) mutant protein, we also expressed the CSQ-1(K111Q) mutant protein in csq-1(jh109) worms. As shown in supplementary material Fig. S2C,D, the mutant protein CSQ-1(K111Q) showed a diffused localization throughout the muscle, similar to the localization of CSQ-1(K111A). Additionally, we predicted that the highly conserved Lys111 in the DBH motif interacts with the three negatively charged residues (Asp197, Glu247 and Glu251) in our homology model. To determine whether these negatively charged amino acids are -like Lys-111 -important for polymerization and localization of calsequestrin, we expressed CSQ-1(E247Q), CSQ-1(E251Q) and doubly mutated protein CSQ-1(E247Q/E251Q) in csq-1-null mutants. As shown in supplementary material Fig. S2E-J, CSQ-1(E247Q) expression resulted in a diffused protein pattern, mostly similar to that of the CSQ-1(K111A) mutant protein (supplementary material Fig. S2E), sometimes showing a weak dispersed pattern (supplementary material Fig. S2D). Sometimes, the staining for CSQ-1(E247Q) was brighter than that of CSQ-1(K111A), but the pattern was generally consistent with the one shown in supplementary material Fig. S2A. Like CSQ-1(E247Q), mutant protein CSQ-1(E251Q) mostly showed a diffused pattern (supplementary material Fig. S2G) and much less frequently a weak dispersed pattern (supplementary material Fig. S2H). Interestingly, the doubly mutated protein CSQ-1(E247Q/E251Q) showed completely diffused localization (supplementary material Fig. S2I,J). It appears that the single mutations (E247Q or E251Q) affect the interaction between Glu247 or Glu251 and Lys111. Furthermore, the doubly mutated protein CSQ-1(E247Q/E251Q) disrupted this interaction completely and, therefore, failed to both polymerize and localize. Therefore, Glu247 and Glu251 are, like Lys111, also important for polymerization of calsequestrin.
Similar to canine calsequestrin, the N-terminal region of CSQ-1 (residues 32 to 43) is essential for the formation of a physiologically relevant linear polymer. Deletion of this region led to the formation of aggregates that were not localized appropriately. We infer from this that only long, linear polymers -and not random aggregates -are localized adjacent to RyR. Interestingly, the structural defects were directly reflected at the functional level. Only those mutant constructs that polymerized properly (⌬N23-31, Bedouin) rescued CSQ-1 function in csq-1(jh109) animals. By contrast, mutant proteins that polymerized correctly but failed to localize (constructs ⌬C17 and ⌬C26), formed random aggregates (construct ⌬N32-43) or both failed to polymerize and localize [mutant protein CSQ-1(K111A)] did not rescue CSQ-1 function in the csq-1(jh109) animals. In addition, mammalian calsequestrin (rabbit skeletal) could rescue CSQ-1 function in the csq-1(jh109) animals (supplementary material Fig. S3). Even though the C-terminus of C. elegans CSQ-1 is different from that in mammals, these data suggest that calsequestrin functions are conserved from C. elegans to mammals. The possible reason why the C-terminus of C. elegans CSQ-1 has the unique feature of its many positively charged residues is to facilitate the direct interaction between CSQ-1 and RyR.
In summary, we found the following: (1) polymerization and localization of CSQ-1 are independent processes in C. elegans. Additionally, we observed a clear correlation between polymerization-status and correct localization of CSQ-1 in the proximity of RyRs. (2) Polymerization is essential but not sufficient for the correct localization of CSQ-1. (3) Back-toback inter-CSQ-1 interaction involving Lys111 is important for the polymerization. (4) Formation of long, linear polymersnot aggregation -accounts for correct localization. (5) Correct polymerization followed by RyR-dependent localization of CSQ-1 is essential for its function as an efficient Ca 2+buffering protein.

C. elegans strains and maintenance
The C. elegans Bristol N2 strain was obtained from the Caenorhabditis Genetics Center (CGC) at the University of Minnesota, MN. The knockout calsequestrin (csq-1) mutant csq-1(jh109) was isolated using a reverse genetics method (this study) and crt-1(jh101) was previously isolated in our laboratory (Park et al., 2001). Worm breeding and handling were conducted as previously described (Brenner, 1974).

Construction of double mutant strain and phenotype analysis
Double mutantion animals csq-1(jh109);unc-68(r1161) and csq-1(jh109);crt-1(jh101) were constructed by standard genetic methods. PCR was used to detect the unc-68(r1161), csq-1(jh109) and crt-1(jh101) mutants. The csq-1(jh109) deletion mutant was characterized microscopically by phenotypic defects. The brood size, survival rate and motility of both the wild-type and mutant worms were estimated at 20°C. To study sensitivity to EGTA and Ca 2+ acetate (CaAc), gravid adults were allowed to lay eggs for 4 hours on plates containing 20 mM EGTA or 20 mM CaAc and were then removed from the plates. Survival rates of worms were determined 3 days later.

Generation of CSQ mutant constructs
To generate the site-specific mutations K111A, K111Q, E247Q, E251Q and the double mutation E247Q/E257Q, we used overlap extension method. Overlapping PCR fragments were mixed, denatured and annealed to generate hetero-duplexes that were extended and amplified into full-length DNA using two primers. This PCR product was cloned into the pPD49.83 vector containing the upstream region of csq-1. A similar strategy was applied to make deletions constructs (⌬N23-32, ⌬N32-43, ⌬C26, and ⌬C17).

Microinjection of mutant constructs and immunofluorescence microscopy
The mutated CSQ-1 constructs were microinjected into csq-1-null mutant csq-1(jh109) animals to obtain germline transformants as described by Mello and Fire (Mello and Fire, 1995). As a transformation marker, plasmid pRF4 containing a dominant gene roller [rol-6(su1006)] was co-injected. Mutant constructs (150 g/l) were injected in the gonads of the adult hermaphrodites as described (Mello and Fire, 1995). After recovery in bacteria-seeded worm agar plates, the injected worms were transferred to fresh worm plates at 16-hour intervals. The progeny of injected animals was selected for roller phenotype and were collected for immunostaining or EGTA sensitivity test.

In vitro binding assay
In vitro binding assays were performed as described previously (Lee et al., 2004). An affinity column was prepared by immobilizing GST-RyR loop I and GST-RyR loop II fusion proteins on glutathione-Sepharose beads for 2 hours at 4°C (Amersham Biosciences). Solubilized proteins CSQ-1 and CSQ-1(⌬C17) (CSQ-1 lacking the 17 C-terminal amino acids) were obtained by centrifugation, mixed with buffer (20 mM Tris-Cl pH 7.4, 1 mM DTT and protein inhibitor mixture) and incubated with the purified GST-RyR loop I and GST-RyR loop II protein Sepharose complex for 4 hours at 4°C. After incubation, the fusion protein-Sepharose complexes were washed three times with 20 mM Tris-Cl pH 7.4, 0.15 M NaCl, 0.2% Triton X-100. The bound proteins were eluted by boiling in SDS-PAGE sample buffer, separated by SDS-PAGE and immunoblotted using anti-CSQ antibody.
Comparative modeling of CSQ-1 A homology model of C. elegans CSQ-1 was built by using the X-ray structure of rabbit calsequestrin (PDB id 1A8Y) . The alignment and modeling building was performed using the program MODELER version 6 (Sali and Blundell, 1993). The model covers residues Gly32 to Ser371. The best model with the lowest MODELLER restraint energy was kept and the side chains of the model were further refined by energy minimization using the GROMOS96 implementation of Swiss-PdbViewer (Guex and Peitsch, 1997). Construction of the CSQ-1 polymer was done by superposition of each monomer to the separate crystallographic symmetry position of the rabbit calsequestrin structure. Figures were prepared using PyMOL (DeLano, 2002).