The Role of the S4-S5 Linker and C-terminal Tail in Inositol 1,4,5-Trisphosphate Receptor Function*

In previous studies we have suggested that spatial proximity of the C- and N-terminal domains of inositol 1,4,5-trisphosphate receptors (IP3Rs) may be critical for the channel gating mechanism. In the present study we have examined the sites of C-N interaction in more detail. We report that deletion mutations within the S4-S5 linker (amino acids 2418–2437) prevent co-immunoprecipitation of the C- and N-terminal domains, inhibit channel activity and enhance IP3 binding. We also show that a region of the C-terminal tail (amino acids 2694–2721), predicted to be a coiled-coil, is also required for channel activity. Circular dichroism spectroscopy and gel filtration studies confirm that this region has a helical structure with the ability to form tetramers. We propose a model in which IP3-induced conformational changes in the N-terminal domain are mechanically transmitted to the opening of the pore through an attachment to the S4-S5 linker. The coiled-coil domain in the C-terminal tail may play a critical role in maintaining the structural integrity of the channel.

In previous studies we have suggested that spatial proximity of the C-and N-terminal domains of inositol 1,4,5-trisphosphate receptors (IP 3 Rs) may be critical for the channel gating mechanism. In the present study we have examined the sites of C-N interaction in more detail. We report that deletion mutations within the S4-S5 linker (amino acids 2418 -2437) prevent co-immunoprecipitation of the C-and N-terminal domains, inhibit channel activity and enhance IP 3 binding. We also show that a region of the C-terminal tail (amino acids 2694 -2721), predicted to be a coiled-coil, is also required for channel activity. Circular dichroism spectroscopy and gel filtration studies confirm that this region has a helical structure with the ability to form tetramers. We propose a model in which IP 3 -induced conformational changes in the N-terminal domain are mechanically transmitted to the opening of the pore through an attachment to the S4-S5 linker. The coiled-coil domain in the C-terminal tail may play a critical role in maintaining the structural integrity of the channel.
Inositol 1,4,5-trisphosphate receptors (IP 3 R) 2 are tetrameric ligand-gated cation channels located in the membrane of the endoplasmic reticulum that serve to mobilize Ca 2ϩ into the cytosol in response to cell stimulation (1,2). There are three mammalian IP 3 R isoforms (types I, II, and III) encoded by three distinct genes (3). Each subunit of the IP 3 R can be functionally divided into a ligand binding domain (LBD), a regulatory domain and a channel domain (4). Based on crystal structure and mutagenesis studies the LBD can be subdivided into three regions: the suppressor domain (amino acids 1-223), the ␤-domain (amino acids 224 -436), and the ␣-domain (amino acids 437-604) (5,6). Two models have been put forward to explain how the binding of IP 3 at the N terminus gates the opening of the channel pore some 2000 amino acids away at the C terminus. The first envisions that conformational changes in the LBD may be transmitted through the regulatory domain and suppressor domain to the channel domain (7). The second model is based on the observation that the LBD is directly coupled to the channel domain (8,9). This latter model predicts that conformational changes in the LBD can gate the opening of the channel through a direct mechanical interaction between the domains. Overall, very little is known about the exact gating mechanism of IP 3 R channels.
The coupling of the LBD to the channel domain was proposed to be intermolecular and to involve interaction sites located between amino acids 1-340 in the N terminus of one subunit and amino acids 2418 -2749 in the C terminus of the adjacent subunit (9). In the present study we have utilized mutations in the C-terminal channel domain to further define the sites of C-N interaction. We show that one C-terminal site of interaction is the cytosol-exposed loop between transmembrane segments 4 and 5 (S4-S5 linker). The 159 amino acids of the cytosol-exposed C-terminal tail were not involved in the C-N interaction, but deletion of 60 amino acids from the tail, which completely removes a predicted coiled-coil, was sufficient to impair channel function. A model of IP 3 R channel gating is proposed that involves the S4-S5 linker and a coiled-coil in the C-terminal tail as key components of the gating mechanism.

EXPERIMENTAL PROCEDURES
Expression Cloning-The cDNA encoding the IP 3 R type I in pCMV3 was the kind gift of Dr. Thomas Südhof (University of Texas Southwestern Medical Center). All amino acid numbering is with reference to the rat type I IP 3 R (10). The splice variant used in this study was SI (Ϫ), SII (ϩ), SIII (Ϫ). All point mutants were made using the QuikChange or the QuikChange Multi Site-directed mutagenesis kit (Stratagene) utilizing a cassette encompassing the BstBI/XbaI fragment of the type I IP 3 R in pBluescript (Invitrogen). Mutants were confirmed by sequencing and the BstBI/XbaI-digested inserts were subcloned into the full-length IP 3 R or HA-tagged full-length IP 3 R cDNA in pCMV3. For the GST fusion protein an insert of nucleotides 8050 -8205 was amplified using Pfu DNA polymerase (Stratagene, CA) and primers encoding BamHI and EcoRI sites. The BamHI/EcoRI PCR product was ligated into similarly digested pGEX-2T (Amersham Biosciences). This construct expresses amino acids 2684 -2735 of type I IP 3 R with GST fused to its N terminus. All primer sequences are available upon request.
Cell Culture and Transfection-COS-7 cells were grown on 100-mm or 150-mm plates (Sarstedt) in Dulbecco's modified Eagle's medium (Sigma) supplemented with 10% fetal bovine serum (Invitrogen), 0.1 mg/ml streptomycin (Invitrogen), 100 international units/ml penicillin (Invitrogen), and 0.04 mg/ml Gentamicin (Invitrogen) until 70 -80% confluent. Transfections typically were done overnight in DMEM without serum. LT-1 (Mirus) and NovaFECTOR (VennNova, Inc.) were used together during transfections and each was added at a cationic lipid to DNA ratio of 1:1. Transfections typically involved 5 g and 20 g of DNA for 100-mm and 150-mm plates, respectively. Co-transfections were done on 100-mm plates with 5 g of IP 3 R and 5 g of SERCA2b. After 24 h, serum containing DMEM was added, and cells were allowed to grow for 48 -72 h.
Expression, Purification, and Concentration of GST Fusion Proteins-GST-(2684 -2735) was propagated in BL21-Gold (DE3) Escherichia coli (Stratagene) and induced with 1 mM isopropyl ␤-D-thiogalactopyranoside at 37°C for 3 h. After induction, bacteria were pelleted and resuspended in a lysis buffer (PBS, pH 7.8, 10 mM EDTA, 0.25% lysozyme, 5 mM DTT, 1% Triton X-100, and a broad spectrum protease inhibitor mixture (Roche Applied Science). Bacteria were rotated at room temperature for 30 min then sonicated. Lysed bacteria were spun down at 20,000 ϫ g for 30 min at 4°C. The supernatant was rotated overnight at 4°C with 5 ml of a 50% slurry of glutathione-Sepharose 4B (Amersham Biosciences). Beads were spun down, washed three times with PBS and resuspended in elution buffer (50 mM Tris-HCl pH 8.0, 10 mM glutathione, 150 mM KCl). After overnight rotation at 4°C the supernatant was removed and concentrated 10ϫ using an Amicon-10 Ultra centrifugal device (Millipore, MA). Cleaved GST fusion protein was concentrated using Vivaspin 2 concentrators with a 3-kDa molecular mass cutoff (ISC BioExpress).
Sephacryl S-200 Column-2-5 mg of cleaved and purified GST fusion protein was loaded onto the column. Fractions were eluted with a column buffer (10 mM sodium phosphate pH 7.4, 150 mM NaCl) and 130 fractions were collected (1 ml/tube). 50 -200 l were removed from each fraction and assayed for protein. Fractions containing the eluted coiled-coil were pooled and concentrated for use in circular dichroism spectroscopy.
Circular Dichroism-Far UV CD spectra from 190 -250 nm (in steps of 1 nm) were obtained by averaging 2-5 scans on a CD spectrometer (Jasco J-810). The recordings were done with 0.45 mg/ml peptide in column elution buffer using a cuvette with a 1-mm path length. The helical content of the peptide was predicted by deconvolution of the spectra using Selcon3 (11).
Trypsin Digestion and Co-immunoprecipitation-Microsomal vesicles containing recombinant IP 3 Rs were incubated (0.25 g/l) in 200 l of trypsin digestion buffer (20 mM Tris-HCl pH 7.8, 120 mM KCl, 1 mM EDTA). Trypsin was added to the vesicles at 4 g/ml and incubated at 37°C for 10 min. The reaction was stopped by the addition of 40 g/ml soybean trypsin inhibitor (Sigma), 1ϫ Complete protease inhibitor mixture (Roche Applied Science), and 1 mM phenylmethylsulfonyl fluoride (Sigma). Microsomes were solubilized in 600 l of solubilization buffer (50 mM Tris-HCl, pH 7.8, 1 mM EDTA, 150 mM NaCl, 1% Triton X-100) and 10 l of Staphylococcus aureus cell wall (Pansorbin cells; Calbiochem). The supernatants were collected after centrifugation (12,000 ϫ g; 10 min) and 70 l of PrA-Sepharose (20% slurry) and immunoprecipitating antibody were added for 4 -16 h at 4°C on a rocker. PrA-Sepharose beads were spun down and washed twice in solubilization buffer. Samples were quenched, boiled for 5 min, and loaded on 10% SDS-PAGE gels (unless stated otherwise).
Trypsin Digestion in the Presence of a Cross-linking Reagent-Microsomal vesicles containing recombinant IP 3 Rs were incubated (0.5 g/l) in 100 l of PBS with 0.7 mM dithiobis(sulfosuccinimidylpropionate) (DTSSP) for 2 h. After the incubation, 1.5 M Tris-HCl, pH 8.8 was added to a concentration of 35 mM to stop the cross-linking reaction. An equal volume of trypsin digestion buffer was added, and microsomes were processed for trypsin digestion as described above. After immunoprecipitation, the protein A-Sepharose beads (Sigma) were incubated an additional 15 min at 4°C in 1% Zwittergent 3-14 to disrupt any non-covalent interactions. Samples were then washed in PBS and quenched in sample buffer supplemented with 10 mM DTT to ensure cleavage of the cross-linker.
3 H-IP 3 Binding Assays-COS-7 cells from a 150-mm plate were harvested by trypsinization, washed once in 10 ml of HRB-HEDTA (20 mM Tris-HEPES pH 7.5, 120 mM KCl, 2 mM HEDTA, 1ϫ protease inhibitor mixture, and 1 mM phenylmethylsulfonyl fluoride) and resuspended in 800 l of HRB-HEDTA. The suspension was permeabilized by the addition of 40 g saponin/mg protein. Saponin-treated cells (0.8 ml) were incubated with a medium containing 120 mM KCl, 20 mM Tris-HEPES (pH 7.2) and 10 nM [ 3 H]IP 3 . Nonspecific binding was estimated by inclusion of 10 M cold IP 3 . After 5 min on ice, the incubations were vacuum-filtered on type A/E glass fiber filters (Pall Corporation) and washed with 15 ml of buffer (50 mM Tris-HCl, pH 8.3, 1 mM EDTA, 1 mg/ml bovine serum albumin). Filters were counted in Budget-Solve complete counting mixture (Research Products International, Corp.). 45 Ca 2ϩ Flux Assays-Assays were performed as previously described (12). Briefly, microsomal vesicles prepared from COS-7 expressing SERCA2b and IP 3 R were incubated for 25 min at 30°C in a 200 nM Ca 2ϩ buffer supplemented with ATP and an ATP regenerating system, 45 Ca 2ϩ , and 20 M ruthenium red. 45 Ca 2ϩ uptake was estimated in the absence of any addition, 10 M IP 3 , or 1 M A23187. After incubation, microsomes were vacuum-filtered over a 0.3-m filter (Millipore), washed (150 mM KCl), and filters were counted in Budget-Solve complete counting mixture (Research Products International, Corp.). Depending on the exact level of SERCA2b expression the 45 Ca 2ϩ counts accumulated in the vesicles ranged from 16,000 to 112,000 cpm in the absence of IP 3 for all the constructs used in this study with the exception of TL-6. For the latter construct the accumulated counts were significantly lower (2342 Ϯ 30.5 cpm, n ϭ 4), which may indicate that TL-6 encodes a constitutively open channel.
Membrane Attachment Assays-Microsomal vesicles were prepared from COS-7 cells transiently transfected with various IP 3 R constructs. 40 mg of vesicles were incubated in a buffer (120 mM KCl, 20 mM Tris, 1 mM EDTA, 1 mM DTT) supplemented with 10 g/ml of trypsin for 0 -60 min. The reactions were quenched by the addition of soybean trypsin inhibitor (100 g/ml) and 1 mM phenylmethylsulfonyl fluoride. After digestion the vesicles were spun down at 62,000 ϫ g, and then pellets and supernatant fractions were quenched in SDS sample buffer and processed by 10% SDS-PAGE. Alternatively, a single time point of 7 min of trypsin digestion was used for the quantitative analysis of point mutations. All mutants were run in parallel with LoopA so that the loss of fragment I could be compared for each mutant. After SDS-PAGE both the pellets and the supernatants were immunoblotted with NT-1, an N-terminal specific IP 3 R antibody. Densitometric analysis of bands was carried out using the program ImageJ (NIH, Bethesda, MD). The loss of fragment I into the supernatant for each loop mutant was expressed as a percentage of the loss of fragment I for the LoopA mutant. For analysis of the tail-less mutants the TL-2 mutant was used as a normalization control. Fig. 1A depicts a linear cartoon of the IP 3 R showing the three main functional domains. IP 3 Rs are cleaved by limited trypsin digestion into five distinct fragments (I-V) whose boundaries and molecular weights are labeled in Fig. 1A. We and Yoshikawa et al. (13) have shown that these five fragments remain bound to one another through non-covalent interactions after tryptic cleavage, such that immunoprecipitation of the 95-kDa tryptic fragment V with an antibody against the C terminus will co-immunoprecipitate the 40-kDa fragment I of an adjacent subunit in the tetramer (8,9,13). Further studies showed that fragment I and V could be directly cross-linked indicating that they were within the cross-linking distance of ϳ12 Å. GST pull-down assays using a GST-tagged LBD and various in vitro translated transmembrane segments from the channel domain, showed that the interacting sites within the channel domain resided downstream of S4 (i.e. between amino acids 2418 and 2749, see also Fig. 1B) (9). Within these residues the only exposed cytosolic regions, which could be involved in the C-N interaction are the S4-S5 linker and C-terminal tail (Fig. 1B). To further narrow down the interaction site(s) we deleted 10 amino acids from the proximal or distal portion of the S4-S5 linker and also made progressive deletions of the C-terminal tail. All mutant constructs were tagged with an HA epitope for detection and immunoprecipitation. We refer to the S4-S5 linker mutants as ⌬LoopA (⌬2418 -2427) and ⌬LoopB (⌬2428 -2437), and the C-terminal tail deletions as "tail-less" mutants (TL-1 through TL-6) (Fig. 1B).

Assembly Status of IP 3 R Mutants-
Several criteria were used to ascertain if the mutations caused gross structural alterations of the protein. We employed size exclusion FPLC to determine if the mutants were able to form tetramers. Type I IP 3 R from rat cerebellum microsomes in the presence and absence of Zwittergent 3-14 were used as controls to distinguish the monomeric from the tetrameric fraction, respectively (9, 10). All mutant constructs were able to form stable tetramers that peaked at fraction 24 (supplemental data, Fig. S1). These results were replicated using 5-20% sucrose density gradients (data not shown). Previous work had suggested that amino acids 2629 -2654 in the C-terminal tail may be critical to the assembly of IP 3 Rs (14). Surprisingly, both TL-5 and TL-6 retained the ability to form tetramers despite the absence of amino acids 2629 -2654 (supplemental data, Fig.  S1). Secondly, we tested the ability of each of the mutant constructs to be cleaved by trypsin using microsomal vesicles prepared from transiently transfected COS-7 cells. Previous studies have shown that mutant IP 3 R constructs with gross structural alterations produce aberrant trypsin cleavage patterns (7). All the mutant constructs used in this study showed the expected pattern of cleavage into five fragments (data not shown).
Limited Tryptic Digestion and Co-immunoprecipitation-We analyzed the C-N interaction of ⌬LoopA, ⌬LoopB, and TL-6 mutants employing a co-immunoprecipitation assay we have used previously (8,9). All the mutants formed a 40-kDa N-terminal fragment I ( Fig. 2A, lanes 4, 7, and 10) and a ϳ90 -95-kDa HA-tagged C-terminal fragment V ( Fig. 2A, lanes [13][14][15][16][17][18][19][20]. As observed previously, the immunoprecipitation of wild-type (FL-HA) fragment V with HA antibody resulted in the co-immunoprecipitation of fragment I (Fig. 2A, lane 2). This was also observed for TL-6, a mutant IP 3 R lacking 159 amino acids of the C-terminal tail ( Fig. 2A, lane 11). From this result we conclude that the C-terminal tail region is not necessary for C-N interaction. The same procedure, however, showed that fragment I of the ⌬LoopA and ⌬LoopB mutants was not coprecipitated by HA-tagged fragment V ( Fig. 2A, lanes  5 and 8). Dithiobis(sulfosuccinimidylpropionate) (DTSSP) is a cross-linking reagent reactive toward lysines that has previously been shown to cross-link fragments V & I in a manner that survives immunoprecipitation in the presence of Zwittergent 3-14, a detergent that disrupts all intra-and intermolecular non-covalent interactions (9,10,15). If the loop mutants caused gross structural changes in the receptor, then DTSSP crosslinking of fragments V & I would be expected to be altered. Fig.  2A (lanes 6 and 9) shows this was not the case, indicating that the cross-linked lysine side chains in the two tryptic fragments remain within ϳ12 Å in both loop mutants. These data suggest that the S4-S5 linker may be one major site of interaction between the ligand binding and channel domains. In the absence of cross-linking reagents, this interaction may be destabilized in the loop mutants during the prolonged period required for immunoprecipitation.
Fragment I (amino acids 1-340) encompasses the entire suppressor domain (amino acids 1-223) as well as some of the IP 3 binding core (amino acids 224 -576) (5,13). We considered the possibility that a loss of C-N interaction could affect the IP 3 binding characteristics of the IP 3 R. We performed [ 3 H]IP 3 binding assays on each of our constructs as previously described (16). To correct for different expression levels the data were normalized to the binding and expression levels of wild-type type I IP 3 R measured in parallel. Both loop deletions displayed statistically significant increases in the binding of IP 3 relative to the wildtype IP 3 R (Fig. 2B). The finding that deletions made to the S4-S5 linker have effects on the binding of IP 3 to the receptor is consistent with an interaction between these two domains and suggests that the C-N interaction may indirectly affect the structure of the N-terminal LBD. None of the tail-less mutants tested showed any significant changes in [ 3 H]IP 3 binding (data not shown; for TL-5 see also Refs. 7 and 17).
Structural and Functional Characterization of Loop Mutants-Previous studies have shown that a major proportion of the cleaved fragment I remains associated with the membrane fraction after trypsin digestion (8,13). If the interaction between fragment I and fragment V is weakened in the loop mutants we would predict that a greater proportion of fragment I would be released into the supernatant. To test this, the supernatants of trypsin treated microsomal vesicles expressing FL-HA, ⌬LoopA or ⌬LoopB were assayed for fragment I. FL-HA had almost no loss of fragment I from the membrane after 60 min of trypsin digestion (Fig. 3A). However, fragment I dissociated from the membrane in a time-dependent manner in both ⌬LoopA and ⌬LoopB mutants (Fig. 3A). The results observed with the loop mutants in membrane attachment assays are consistent with the findings obtained in coimmunoprecipitation assays and suggest that C-N interactions are weakened by deletions made within the S4-S5 linker.
To further narrow down which amino acids might be directly involved in the C-N interaction we created a series of alanine point mutations within the ⌬LoopA and ⌬LoopB regions. The mutations were designed to remove charged or hydrophobic residues and also to determine the relative importance of residues within the ⌬LoopA and ⌬LoopB regions (Fig. 3B). All the point mutants were expressed at similar levels in microsomal vesicles (upper panel, Fig. 3C). The amount of fragment I lost from the membrane by each point mutant was expressed as a percentage of the amount lost from ⌬LoopA membranes, which were run in parallel and assigned as 100% (Fig. 3C,  mutant 1). ⌬LoopB behaved identically to ⌬LoopA in membrane attachment assays (Fig. 3C, mutant 2). Mutations made to the charged residues within the ⌬LoopA region led to a sub- Microsomal vesicles were prepared and digested with trypsin as described under "Experimental Procedures" in a final volume of 400 l (0.125 g of protein/l). An aliquot (5 g of protein) of each digestion was saved for immunoblot analysis (lanes 1, 4, 7, and 10). The remaining digested IP 3 R was subjected to immunoprecipitation with HA mAb and protein A-Sepharose for 4 -16 h at 4°C (lanes 2, 5, 8, and 11). Alternatively, vesicles were cross-linked with 0.7 mM DTSSP for 2 h on ice prior to trypsin digestion and immunoprecipitation (lanes 3, 6, 9, and 12). Lanes 13-20 show analysis of lysates by immunoblotting with HA mAb and indicate that fragment V is formed normally in the loop and tail-less mutants. B, WT and loop mutant cDNA were transiently transfected into COS-7 cells. The cells were permeabilized with saponin, and binding was measured with a subsaturating concentration of [ 3 H]IP 3 (10 nM) as described under "Experimental Procedures." The expression of the loop mutants was quantified densitometrically from immunoblots and expressed as a ratio (R) relative to the WT receptor. The binding values of the loop mutants were divided by R to correct for small variations in expression (16). WT binding values were between 0.8 -1 pmol/mg protein. Data are the means Ϯ S.E. of 6 -8 independent measurements. stantial loss of fragment I from the membrane (Fig. 3B, mutants  3, 4, and 5). Mutation of charged residues in ⌬LoopB (mutant 6) or two hydrophobic residues in ⌬LoopA (mutant 7) had minimal effects. The partial effect on stability of removing a positive charge in ⌬LoopA (mutant 4) was not augmented by the additional removal of positive charges in ⌬LoopB (mutant 10). The data suggest that C-N interactions may involve electrostatic associations primarily with charged residues in the ⌬LoopA portion of the S4-S5 linker.
To test if mutations in the S4-S5 linker affected function, we employed a 45 Ca 2ϩ flux assay using microsomal vesicles prepared from COS-7 cells co-transfected with various mutant IP 3 R constructs along with the Ca 2ϩ pump, SERCA2b. The assay measures the activity of transfected, rather than endoge-nous IP 3 Rs, and has been described in detail elsewhere (9,12). The last lane in Fig. 3D (labeled SERCA2b) shows the IP 3 -mediated inhibition of 45 Ca 2ϩ flux in microsomes prepared from COS-7 cells transfected with SERCA2b alone (no recombinant IP 3 Rs) and represents the background contribution from endogenous IP 3 Rs. Both ⌬LoopA and ⌬LoopB were inactive in IP 3 -mediated 45 Ca 2ϩ flux assays and behaved like the control cells transfected with SERCA2b alone (mutants 1 and 2, respectively). Functional characterization of selected loop point mutants by 45 Ca 2ϩ flux assays showed qualitatively similar results to the membrane attachment assays. Mutation of either acidic residues (Asp 2418 , Glu 2423 , Glu 2424 ) or a single basic residue (Arg 2422 ) in the ⌬LoopA region had partial inhibitory effects on channel function (Fig. 3D, mutants 3 and 4), which . C, membrane attachment assay of S4-S5 linker mutants. Upper panel, Western blot of microsomal lysate (pellet ϩ supernatant) after trypsin digestion using NT-1. The molecular mass marker is shown on the left, and lanes are aligned with the respective bar for each mutant construct. Lower panel, supernatants and pellets were collected after 7 min of trypsin digestion, and the levels of fragment I were determined by immunoblotting with NT-1 antibody. The amount of fragment I released for each mutant was expressed relative to ⌬LoopA (run in parallel) and defined as the 100% value. *, significance of p Յ 0.05 versus WT and is representative of at least three independent measurements. The absolute level of release for ⌬LoopA expressed as a supernatant/pellet ratio was 35.2 Ϯ 2.6% (n ϭ 10). D, effect of IP 3 on 45 Ca 2ϩ flux was measured in microsomes prepared from COS-7 cells co-transfected with SERCA2b and IP 3 R cDNA as described under "Experimental Procedures." The cells transfected with SERCA2b alone were taken as the control (12). *, significance of p Յ 0.05 versus WT and is representative of at least four independent measurements. was further increased when the mutations were combined (Fig.  3D, mutant 5). However, mutation of the charged residues in ⌬LoopB region did not affect IP 3 R function (Fig. 3B, mutant 6).
Properties of the IP 3 R C-terminal Tail Region-The C-terminal tail region encompasses amino acids 2590 -2749 and has been shown to be involved in the assembly and function of the IP 3 R (7,14,18). Previous studies have shown that deletion of 13 amino acids from the C terminus does not affect channel func-tion (7,19) but removal of 139 amino acids inactivates the channel (7). Two cysteine residues just downstream of S6 were also shown to be crucial for activation gating (white circles, Fig. 4A) (7). Other studies located a "minimal interaction domain" (MID) between amino acids 2629 -2654, which plays a role in the dimerization of C-terminal tails as well as in the assembly and stability of tetramers (Fig. 4A) (14). Membrane attachment assays were carried out on the series of tail-less mutant constructs with progressive deletions of the C-terminal tail (shown in Fig. 1B). All the tail-less deletions were expressed at similar levels in microsomal vesicles (upper panel, Fig. 4B). Wild type, TL-1, TL-1A, and TL-1B did not show a substantial loss of fragment I from the membrane. However, the deletion of more than 43 amino acids (i.e. TL-2 through TL-6) caused a loss of fragment I from the membrane compared with wild type (Fig.  4B). 45 Ca 2ϩ flux studies on the tailless deletions are shown in Fig. 4C. Interestingly, function was completely lost when more than 43 amino acids were deleted from the C terminus (TL-2 through TL-6). Because the membrane attachment and the functional assays both show major effects beginning at TL-2 we conclude that the amino acids between 43 and 60 residues from the C terminus are also crucial to maintaining a functional channel.
Structural Characterization of the C-terminal Tail Region-Analysis of the C-terminal tail using the program MultiCoil (20) indicates that the residues between TL-1 and TL-2 (amino acids 2690 -2725) have a high probability of being part of a coiled-coil (Fig. 5A). The analysis indicates that there are four heptad repeats, which make up the coiledcoil beginning at Leu 2694 and continuing through Gln 2721 . The inset in Fig. 5A is an end-on view of C-terminal tail residues 2694 -2721 shown as a single helix of the coiled-coil. The core hydrophobic residues at positions a and d of the heptad are indicated (black circles). Removal of the last two heptads displayed a statistically significant inhibition of function (TL-1B, Fig. 4C) but the removal of all 4 heptads from the C-terminal tail correlated with a complete loss of channel function (TL-2, Fig. 4C). , and a predicted coiled-coil domain. The C-terminal end of each tail-less deletion (italics) is marked by an arrow. B, membrane attachment assay of tail-less deletion mutants. Upper panel, Western blot of microsomal lysate (pellet ϩ supernatant) after trypsin digestion using NT-1. The molecular mass marker is shown on the left, and lanes are aligned with the respective bar for each mutant construct. Lower panel, supernatants and pellets were collected after 7 min of trypsin digestion and the levels of fragment I were determined by immunoblotting with NT-1 antibody. The amount of fragment I released for each mutant was expressed relative to TL-2 (run in parallel) and defined as the 100% value. *, significance of p Յ 0.05 versus WT and is representative of at least three independent measurements. The absolute release for TL-2 expressed as a supernatant/pellet ratio was 30.3 Ϯ 4.7% (n ϭ 4) C, channel function of the tail-less deletion mutants was assessed using a 45 Ca 2ϩ flux assay as given in Fig. 3D. *, significance of p Յ 0.05 versus WT and is representative of at least four independent measurements.
To examine the structural and biophysical properties of this region of the C-terminal tail, a GST fusion protein of amino acids 2684 -2734 was expressed and purified (Fig. 5B, lane 1). The fusion protein was cut with thrombin to separate the GST (24 kDa) from the predicted coiled-coil region (6 kDa) (Fig. 5B, lane 2) and then further purified by incubation with glutathione-Sepharose 4B beads (lane 3) followed by size exclusion FPLC (lane 4). A portion of the linker region remained attached to the coiled-coil after thrombin digestion and created a peptide with the predicted molecular mass of 10.5 kDa. Analysis of the molecular mass of the purified protein on a Sephacryl S-200 column indicated a size of ϳ45-kDa (fraction 48) with no evi-dence for the significant formation of a monomer (fraction 82) (Fig. 5C). We conclude that amino acids 2684 -2734 from the C-terminal tail has the ability to assemble into tetramers and that this higher order structure may be necessary for channel function. The secondary structure of amino acids 2684 -2734 was assessed by circular dichroism (CD) spectroscopy. CD wavelength scans showed intense minima near 208 nm and 222 nm, which are characteristic of highly helical proteins (Fig. 5D). When the spectra was analyzed by a secondary structure prediction program Selcon3, (11) this region of the C-terminal tail was predicted to be 42% helical (data not shown).

DISCUSSION
A hypothetical model of IP 3 R gating based on our studies is shown in Fig. 6. The main feature of the model is that the gating mechanism couples the conformational change in the LBD, resulting from IP 3 binding, to the mechanical movement of the S4-S5 linker through a direct interaction between these regions of the IP 3 R. In our mutagenesis experiments, shortening the S4-S5 linker (⌬LoopA or ⌬LoopB) prevented interactions between the C-and N-terminal domains and yielded a non-functional channel. Point mutations suggested that charged residues in the proximal region of the linker (⌬LoopA) may have a dominant role. The importance of the S4-S5 linker has also been emphasized in studies of the gating mechanism of several voltage-gated channels. In the Kv1.2 channel the S4-S5 linker is an amphipathic helix that runs parallel to the plane of the membrane and is arranged in a manner that constricts the S6 pore-lining helix bundle at the cytosolic aspect of the membrane, thereby maintaining the channel in a closed conformation (21,22). The voltage-induced movement of the charged S4 helix is proposed to displace the S4-S5 linker and thereby permit the S6 helix bundle to separate which allows ion conduction (22). Similarly, in KirBac1.1 channels a homologous S4-S5 linker structure (referred to as a "sliding helix") has been suggested to play a critical role in its gating mechanism (23). In both these K ϩ channels, interactions between the S4-S5 linker and the distal segments of the S6 helix have been documented (24,25).
Secondary structure programs predict that the S4-S5 linker would also form an amphipathic helix in IP 3 Rs (data not shown). Whereas the voltage-dependent shift of the S4 helix drives the movement of the S4-S5 linker in K ϩ channels, we suggest that the direct coupling of the LBD to the S4-S5 linker causes the opening of the channel upon IP 3 binding. Previous studies have shown that the presence of IP 3 does not interfere with the co-immunoprecipitation of C-and N-terminal domains in the absence (8) or presence of a cross-linking reagent (data not shown). This suggests that any changes in the interactions between these domains involved in the gating mechanism may cause a subtle mechanical movement of the S4-S5 linker rather than a gross formation/ disruption of molecular interactions. The movement of the linker would allow the S6 helix bundle to separate and open an ion conduction pathway for Ca 2ϩ (Fig. 6B). Ca 2ϩ itself has profound effects on IP 3 R channel gating (26). Although not specifically considered in Fig. 6, it is clear that the multiple binding sites for Ca 2ϩ and calmodulin present in the receptor could potentially influence the binding of IP 3 and the conformational movements involved in the gating mechanism at many steps.
What is the evidence that C-N interactions are direct and involve the suppressor domain (amino acids 1-223) as indicated in Fig. 6? GST-LBD (amino acids 1-605) has been shown to specifically pull down an in vitro translated segment of the IP 3 R containing the S4-S5 linker (amino acids 2418 -2749), which together with crosslinking data (9) suggest that the interaction is direct and lies within amino acids 1-340. The suppressor domain makes up the majority of this region and we have therefore speculated that it contains the site of interaction. Interestingly, deletion of the suppressor domain does inactivate IP 3 Rs despite increasing the affinity of the receptor for IP 3 by over 10-fold (7). Others have proposed that segments from the regulatory domain maintain the channel in the closed state and that binding of IP 3 relieves this inhibition (27). Varnai et al. (28) have shown that expression of an ER-tethered form of the ␣-domain of the LBD (amino acids 427-605) can gate endogenous IP 3 Rs. Further work is required to identify the exact N-terminal segments interacting with the S4-S5 linker.
A second significant observation of this study is the identification of a coiled-coil domain in the distal portion of the C-terminal tail (Figs. 4A and 5). Removal of this region eliminated IP 3 R activation and destabilized C-N interactions as measured in membrane attachment assays (Fig. 4, B and C), but did not increase IP 3 binding or affect the co-immunoprecipitation of tryptic fragments as observed for the loop mutants (Fig. 2). Therefore, we conclude that deletion of the C-terminal tail must destabilize the IP 3 R in a manner distinct from the loop deletions. Co-immunoprecipitation assays ( Fig. 2A) or pulldown assays using a GST fusion protein of the C-terminal tail (amino acids 2590 -2749) and recombinant LBD (amino acids 1-605) from transiently transfected COS-7 cells provide no evidence that the tail is directly involved in the interaction with the N-terminal domain (supplemental data, Fig. S2). This conclusion may appear to be inconsistent with the finding that an enhanced release of fragment I was noted for both the loop and tail-less mutants in the membrane attachment assays. However, it should be pointed out that the co-immunoprecipitation and membrane attachment measurements are very different FIGURE 6. Model for the gating mechanism of the IP 3 R. For simplicity only the S4, S5, and S6 helices of two subunits of the IP 3 R tetramer are shown. The S4-S5 linker is shown as a helix running parallel to the membrane and arranged in a manner that constricts the pore-lining S6 helix into a closed configuration (panel A). Based on the present study we propose that the S4-S5 linker is connected via multiple interactions with residues in the N-terminal suppressor domain (SD) and that binding of IP 3 to the LBD causes a conformational change that induces a displacement of the S4-S5 linker that allows the constriction of the S6 helix at the bundle crossing to be relieved and thereby provide a conduction pathway for Ca 2ϩ (panel B). The maintenance of the tertiary structure of the gating/conduction pathway is also dependent on the integrity of two key cysteine residues (filled circles) and a coiled-coil domain (CC) in the C-terminal tail. For additional details see text.
assays and that many additional factors (other than interaction between the S4-S5 linker and fragment I) may be involved in maintaining the adherence of trypsin-digested receptor segments to membranes. For example, the void created by deletion of the C-terminal tail and the absence of the stabilization contributed by the coiled-coil domain could potentially interfere with the packing of the tetrameric subunits and could indirectly impair interactions between the five tryptic fragments.
Coiled-coils have been identified in the C-terminal tails of a diverse set of ion channels such as voltage-gated potassium channels (EagI, EagII, Elk, Erg) (29,30), cyclic nucleotide channels (CNQ and KCNQ) (31)(32)(33)(34), calcium-dependent potassium channels (SK) (35), non-selective cation channels (PKD) (36), and calcium channels (TRP) (37), where they typically aid in the assembly and stability of tetramers (38). However, the primary determinants for tetramer assembly in IP 3 Rs are contained within the S5-S6 transmembrane domains (18). Therefore the coiled-coil may contribute to the stability of the IP 3 R tetramer but is unlikely to have a primary role in tetramer formation. The C-terminal tails of both ryanodine receptors (RyR) and IP 3 Rs have been shown to self associate (14) and the recombinant IP 3 R coiled-coil domain (amino acids 2684 -2734) forms tetramers as determined by FPLC chromatography (Fig. 5C). At present we can only speculate as to why the deletion of 60 amino acids from the C-terminal tail inactivates the IP 3 R channel. Clearly the loss of the coiled-coil domain correlates with the loss of structural stability of the receptor as indicated from the membrane attachment assays. The lack of the proper associations between the C-terminal tails in the tail-less mutants could affect a number of channel characteristics such as the packing of subunits around the channel, the structural integrity of the pore-lining S6 helix or the occlusion of the normal exit pathway for the Ca 2ϩ ion. Interestingly, a coiled-coil domain is absent from the shorter C-terminal tail of RyRs, although removal of 15 amino acids from the C-terminal tail of RyRs is sufficient to disrupt RyR activation (39,40).
Ramos-Franco et al. (41) have shown that deletion of the first four transmembrane domains (⌬2211-2416) produced a constitutively open channel (41). Presumably, the absence of a membrane anchor on the N-terminal end of the S4-S5 linker allows the linker to adopt a position where it can no longer compress the S6 pore-lining helix bundle. This mutant was also found not to be gated by IP 3 as would be expected if the channel lacked the C-N interaction necessary to relay the signal to the channel domain. In our studies, none of the loop mutants or TL-1 through TL-5 resulted in a constitutive activation of the channel, as indicated by a robust 45 Ca 2ϩ uptake into microsomal vesicles in the absence of IP 3 (data not shown). However, preliminary studies suggest that the TL-6 mutant, which includes deletion of the cytoplasmic end of the helix bundle, may be constitutively active (data not shown).
A number of predictions and future directions arise from the model in Fig. 6. The S6 helix is shown as extending beyond the membrane but its exact boundaries remain to be established. Many ion channels contain a glycine which is proposed to act as a hinge to allow the S6 helix to bend during channel opening (24,25,42). Mutagenesis of a putative hinge residue Gly 2586 in the S6 helix of IP 3 Rs should test this proposed mechanism.
Interestingly, mutation of the analogous residue in RyR2 (G4864A) did not affect caffeine-mediated Ca 2ϩ release (43). The section of the S6 helix, proposed to extend into the cytoplasm and interact with the S4-S5 linker, is presumably critical for channel gating. Cys 2610 and Cys 2613 in this region have been identified as being crucial for activation gating of IP 3 Rs (7). Salt bridges between residues in the S4-S5 linker and the region adjacent to S6 have been suggested to stabilize the closed state of HCN channels (44). Further studies are needed to identify interaction sites between the S4-S5 linker and S6 helix in IP 3 Rs and also to document the movement of these regions predicted from the proposed gating mechanism.