A novel recombinant hyperaffinity inositol 1,4,5-trisphosphate (IP(3)) absorbent traps IP(3), resulting in specific inhibition of IP(3)-mediated calcium signaling.

We have developed a novel recombinant hyperaffinity inositol 1,4,5-trisphosphate (IP(3)) absorbent, called the "IP(3) sponge," which we constructed on the basis of the ligand-binding site of the mouse type 1 IP(3) receptor (IP(3)R1). The IP(3) sponge exhibited approximately 1000-fold higher affinity for IP(3) than the parental IP(3)R1 and specifically competed with the endogenous IP(3)R for binding to IP(3). Trapping IP(3) with the IP(3) sponge inhibited IP(3)-induced Ca(2+) release (IICR) from cerebellar microsomes in a dose-dependent manner. The IP(3) sponge expressed in HEK293 cells also inhibited IICR in response to stimulation with carbachol or ATP. Its inhibitory effects were dependent upon the level of its expression over the increased IP(3) contents. Moreover, the IP(3) sponge significantly reduced the carbachol-induced phosphorylation of cAMP-response element-binding protein in HEK293 cells, indicating that the activation of cAMP-response element-binding protein by Ca(2+)-dependent phosphorylation may be partly attributable to IICR.

Many cellular responses to diverse biological stimuli, such as neurotransmitters, hormones, and growth factors, are mediated by the intracellular second messenger inositol 1,4,5trisphosphate (IP 3 ). 1 IP 3 subsequently induces Ca 2ϩ mobilization from intracellular stores by activating its receptor (IP 3 R) (1,2). IP 3 R channels form homo-or heterotetramers via the co-assembly of distinct types of IP 3 R subunits (types 1-3) (3) and bind IP 3 in a stoichiometric manner (4). The IP 3 -induced Ca 2ϩ release (IICR) not only results in a transient increase in intracellular Ca 2ϩ concentration ([Ca 2ϩ ] i ) but also evokes complex spatiotemporal dynamics of [Ca 2ϩ ] i called Ca 2ϩ waves and Ca 2ϩ oscillations (5). Thus, IP 3 R functions as a signal converter from IP 3 to Ca 2ϩ , which exhibits more complex dynamics in time and space inside cells and acts on thousands of downstream targets that play key roles in many aspects of cell physiology (1,2). Type 1 IP 3 R (IP 3 R1) is composed of 2749 amino acids (molecular mass about 313 kDa) and is structurally divided into three parts as follows: a large N-terminal cytoplasmic arm region (83% of the receptor molecule); a putative six membrane-spanning region clustered near the C terminus, which is thought to constitute an ion channel by forming a tetramer; and a short C-terminal cytoplasmic tail region (6). In our previous studies (7,8) on the ligand-binding site of mouse type 1 IP 3 R (mIP 3 R1), we found that the core region, essential for the high affinity of ligand binding, was localized in amino acid residues 226 -576. We also found that the affinity of residues 1-604 (T604) of mIP 3 R1 for IP 3 (K d ϭ 45 nM) was comparable with that of the native cerebellar IP 3 R (83 nM) (4), indicating that T604 could form a well folded or compact conformation for the functional IP 3 -binding pocket (7,9). The N-terminal residues 1-223, which we assigned as a suppresser region, instead of being directly responsible for IP 3 binding, suppressed the binding activity of the core region (7,9). Without the putative suppresser region, residues 224 -604 showed markedly higher affinity for IP 3 than did the parental T604 (7). We therefore inferred that such a high affinity IP 3 -binding protein could function as an IP 3 absorbent in living cells and compete with the native IP 3 R for binding to IP 3 .
In this paper, we report the development of a novel recombinant IP 3 absorbent, the "IP 3 sponge," which has hyperaffinity for IP 3 (K d ϭ 0.092 nM). The IP 3 sponge was constructed on the basis of the ligand-binding site of mIP 3 R1, and we assumed that this hyperaffinity IP 3 sponge could compete with IP 3 R for the IP 3 signal. We showed that its application actually caused specific inhibition of in vitro IICR in a dose-dependent manner. Moreover, the IP 3 sponge exogenously expressed in HEK293 cells could trap IP 3 produced in response to stimulation with either carbachol (CCh) or ATP, resulting in inhibition of IICR. The IP 3 sponge appeared to be effective in competing for IP 3 with multiple types of tetrameric IP 3 R channels, because HEK293 cells expressed all three IP 3 R subunits (10). The effectiveness of the IP 3 sponge on IP 3 -Ca 2ϩ signaling-dependent cell physiology was verified by providing an invaluable insight into the functional role of IICR in the phosphorylation of cAMPresponse element-binding protein (CREB), which was previously thought to be activated mainly by Ca 2ϩ influx (11,12). cDNA was used for all constructs used in this study (13). pET-T604 consisted of the IP 3 -binding sequence encoding residues 1-604 cloned in pET-3a (7,8). pET-(224 -604) containing the residue 224 -604-coding sequence was produced by PCR-based mutagenesis, as described elsewhere (8,9). To express an N-terminal fused protein of residues 224 -604 with the glutathione S-transferase gene (G224), pGEX-G224 was obtained by cloning the BamHI-EcoRI fragment containing the residue 224 -604 coding sequence isolated from pET-(224 -604) into pGEX-2T (Amersham Biosciences). Site-directed mutagenesis of K508A and R441Q was introduced into G224 by two-step PCR, as described elsewhere (8). The sequences and junctions of all constructs were verified by DNA sequencing. Expression of recombinant proteins in Escherichia coli cells was carried out by the low temperature method (7,8) with minor modifications, followed by glutathione-Sepharose 4B column chromatography (Amersham Biosciences) performed according to the manufacturer's protocol. The glutathione S-transferase (GST)-fused proteins were then applied to a PD-10 column containing Sephadex G-25M (Amersham Biosciences) pre-equilibrated with elution buffer (10 mM HEPES-KOH, pH 7.2, 88 mM NaCl, 1 mM KCl). Protein concentration assay and Western blot analysis were carried out as described elsewhere (7).
Measuring Ca 2ϩ Release from Microsomes-IICR from cerebellar microsomes was measured with fura2 and a fluorospectrometer, CAF110 (Nihon Bunko), as described elsewhere (14). Ca 2ϩ was loaded into microsomes by activating Ca 2ϩ -ATPase with 1 mM ATP (Sigma), and Ca 2ϩ release was triggered by activating IP 3 R with 100 nM, 500 nM, and 1 M IP 3 . Various amounts of the GST-fused proteins were added 1 min before the addition of IP 3 . Before completing each set of experiments, Ca 2ϩ release via the ryanodine receptor was determined by adding 40 mM caffeine (Sigma). Maximum and minimum values were obtained in the presence of an excess amount of CaCl 2 and EGTA, respectively, to calculate Ca 2ϩ concentration as described elsewhere (15).
Phospholipase C (PLC) Assay-PLC activity was assayed by the methods described previously (18). In brief, a reaction mixture containing 50 mM Mes buffer, pH 6.0, 400 M CaCl 2 , 1 mg/ml bovine serum albumin, 1 pmol of PIP 2 , 22,000 dpm of [ 3 H]PIP 2 , and the homogenate of established stable HEK293 cells or native HEK293 cells (130 g) in the presence of the bacterially expressed GST-fused IP 3 -binding proteins (1, 10, and 100 pmol) was incubated at 37°C for 10 min. The reaction was terminated by the addition of 2 ml of chloroform/methanol (2:1), and radioactive inositol trisphosphate was extracted with 1 N HCl. We measured the radioactivity with a scintillation counter.
Cellular Ca 2ϩ Imaging-HEK293 cells were grown for 48 h on 35-mm glass-bottomed dishes (coated with 50 g/ml poly-L-lysine; Sigma) in the presence or absence of 20 ng/ml Dox. The cells were then incubated in Dulbecco's modified Eagle's medium containing 5 M fura2-AM (Dojindo) and 10% fetal bovine serum at 37°C for 30 min, washed once with Hanks' balanced salt solution, and then kept in Hanks' balanced salt solution in the dark at 30°C for 30 min. Before measurements, cells were washed twice in Ca 2ϩ -free Hanks' balanced salt solution supplemented with 1 mM MgCl 2 and 15 M EGTA. Fura2 fluorescence images were analyzed using a video image analysis system (Argus-50/CA, Hamamatsu Photonics), as described elsewhere (19). Small volumes of stock solutions containing carbachol (CCh) (Sigma), ATP, or thapsigargin (Wako) were added to the bath medium to achieve appropriate concentrations. At the end of each recording, intracellular Ca 2ϩ concentration ([Ca 2ϩ ] i ) was estimated from the maximal and minimal ratio (15) determined by adding 20 mM CaCl 2 and 1 M ionophore 4-Bromo A23187, followed by the addition of 30 mM EGTA. Viable At the top is a schematic representation of the N-terminal region of the IP 3 R1, with the IP 3 -binding core (residues 226 -576) indicated by a solid line. T604 is residues 1-604 cloned in pET-3a. All GST (hatched boxes) fusion constructs were cloned in pGEX-2T; G224, GST ϩ residues 224 -604; K508A, the substitution of Ala for Lys-508 on G224; R441Q, the substitution of Gln for Arg-441 on G224; GST alone (pGEX-2T). B, Western blot analysis (7) of GST-fused IP 3 -binding proteins expressed in E. coli. Soluble proteins (0.05 g) were probed with anti-GST antibody. The expression of G224, K508A, or R441Q was also confirmed by using polyclonal antibodies N2, N3, and N5 against the IP 3 binding region (14). C, Scatchard analysis of the inhibition of specific [ 3 H]IP 3 (9.6 nM) binding to K508A (2 g) by cold 1,4,5-IP 3 (7,8). To analyze G224 and R441Q (0.02 g), specific binding was determined by subtracting nonspecific binding measured in the presence of 1 M unlabeled IP 3 from the total binding obtained with 0.15-9.6 nM [ 3 H]IP 3 . The results of a typical experiment are shown. D, inhibition of specific [ 3 H]IP 3 (9.6 nM) binding to G224 and R441Q (0.1 g) by various inositol phosphates. G224 is represented by the open symbols, and R441Q is represented by the filled symbols: 1,4,5-IP 3 (circles); 2,4,5-IP 3 (triangles); 1,3,4,5-IP 4 (rhombuses); 4,5-IP 2 (squares); IP 6 (inverse triangles). Values are normalized to the 100% control measured without cold inositol phosphates as a competitor. Nonspecific binding was measured in the presence of 10 M cold 1,4,5-IP 3 . Each point represents the mean value from two separate experiments. E, the IP 3 absorbancy of G224 and R441Q (0.781-50 g/ml) was assayed by directly adding them to binding mixtures containing 0.96 or 9. cells, judged by their impermeability to trypan blue, were used throughout all the experiments. After every experiment, the cells were fixed for 20 min by adding formaldehyde solution (38% formalin; Wako) to a final concentration of 4%, and cells were then permeabilized with 0.1% Triton X-100 in PBS (PBS-T) for 5 min. After being washed and blocked with 2% normal rabbit serum in PBS-T, the cells were incubated with 1:500 goat anti-GST antibody (Amersham Biosciences) and then with 1:200 FITC-conjugated anti-goat antibody (Vector). Immunofluorescence was analyzed with a Zeiss LSM 410 inverted laser scan microscope and a ϫ40 objective lens.
CREB Phosphorylation-HEK293 cells (4 ϫ 10 5 cells per 6-cm dish) were cultured for 48 h in the presence or absence of 20 ng/ml Dox. Cells were stimulated with CCh or forskolin for 15 min under Ca 2ϩ -free conditions, as described above. Stimulated cells were fixed for 30 min by adding 10% trichloroacetic acid solution and lysed in 100 l of SDS-PAGE sampling buffer supplemented with 1 mM Na 3 VO 4 , 1 mM NaF, and protease inhibitors. After sonication, the lysates were neutralized with 2 l of 1 M Tris and boiled for 10 min. A 25-l volume of cell lysate was then resolved by 10% SDS-PAGE, transferred to nitrocellulose, and probed with anti-phospho-CREB or anti-CREB antibody (Upstate Biotechnology, Inc.). The anti-phospho-CREB blots were quantitatively analyzed with a Molecular Dynamics FluorImager after staining with the alkaline phosphatase substrate AttoPhos (JBL Scientific).
Immunocytochemical Staining of CREB Phosphorylation-HEK293 cells were grown for 48 h on 35-mm glass-bottomed dishes in the presence or absence of 20 ng/ml Dox. After stimulation with CCh for 15 min under Ca 2ϩ -free conditions, dual labeling of CREB phosphorylation and GST-fused IP 3 -binding proteins was carried out by the methods described previously (20). In brief, cells were stained with 1:200 antiphospho-CREB rabbit polyclonal antibody (Upstate Biotechnology, Inc.) and 1:150 anti-GST (B-14) mouse monoclonal antibody (Santa Cruz Biotechnology), followed by 1:200 goat anti-rabbit Alexa Fluor 488 and 1:200 goat anti-mouse Alexa Fluor 594 secondary antibodies (Molecular Probes). Images were analyzed with an Olympus laser scan confocal microscope with an Olympus Fluoview system and a ϫ60 objective lens.

Construction of a Novel Recombinant IP 3 Absorbent Protein-
The IP 3 -binding protein of amino acid residues 224 -604 of mouse type 1 IP 3 R (mIP 3 R1), with markedly high affinity for IP 3 , was not as efficiently expressed as a soluble active form in E. coli as residues 1-604 (T604) (7,9). To solve this problem, we fused residues 224 -604 of mIP 3 R1 with the GST gene instead of the putative suppresser region (Fig. 1A). Expression of the GSTfused protein, named G224, improved to ϳ30 mg/liter of E. coli culture using a low temperature method (versus 19 mg/liter for T604). Fig. 1B shows a specific immunoreactive band of G224 (66 kDa) after partial purification with column chromatography for glutathione affinity and gel filtration. G224 showed strikingly greater binding affinity, 500 -1000-fold higher (K d ϭ 92 Ϯ 2.1 pM; n ϭ 3), than T604 (45 nM) (7) and cerebellar IP 3 R (83 nM) (4) (Fig.  1C). Site-directed mutational analyses revealed that three basic amino acid residues (Arg-265, Lys-508, and Arg-511) are critical for specific IP 3 binding and that substitution of Gln for Arg-441 enhances binding activity (7,8). To determine the effect of these mutations on G224, we produced two mutants, K508A and R441Q, each having a single amino acid substitution on G224. In mutant K508A Ala was substituted for Lys-508, and in R441Q Gln was substituted for Arg-441 (Fig. 1A). As expected, K508A showed an enormous loss of binding affinity (K d ϭ 340 Ϯ 3.9 nM) (Fig. 1C), to 3700-fold less than that of parental G224. Mutant R441Q had double the IP 3 binding activity (K d ϭ 45 Ϯ 2.1 pM) (Fig. 1C) of G224 and 1900-fold the binding activity of cerebellar IP 3 R.
G224 and R441Q Compete with IP 3 R for Binding to IP 3 -To determine whether G224 and R441Q behave as IP 3 absorbents, we initially assessed the inhibitory effect of these high affinity IP 3 -binding proteins on [ 3 H]IP 3 binding to cerebellar microsomes, an enriched source of IP 3 R protein (4) (Fig. 1E). In assays with 0.96 nM [ 3 H]IP 3 , IP 3 binding to microsomes de- creased to about 20% of the control in the presence of either G224 or R441Q at 6.25 g/ml. By contrast, even at an 8-fold higher concentration (50 g/ml), K508A reduced binding to only 70%, and GST had no effect. Moreover, in assays with 9.6 nM [ 3 H]IP 3 , G224 and R441Q at 12.5 g/ml reduced IP 3 binding to 30%. Therefore, both G224 and R441Q appeared to function as effective IP 3 absorbents, and they dose-dependently competed with microsomal IP 3 R for binding to IP 3 .
G224 and R441Q Trap IP 3 , Resulting in Inhibition of in Vitro IICR-We next examined the inhibitory effect of G224 and R441Q on IICR from mouse cerebellar microsomes. In the absence of IP 3 -absorbent proteins, half-maximal Ca 2ϩ release occurred in response to stimulation with 100 nM IP 3 (data not shown). In the presence of either G224 or R441Q at 12.5 g/ml, the peak height of IICR with 100 nM IP 3 decreased to about 50% of the control, and almost maximal inhibition was recorded at 100 g/ml (Figs. 2A, a and d, and B). IICR stimulated with 500 nM and 1 M IP 3 was inhibited in the presence of 150 g/ml G224 and R441Q to as low as 15 and 30% of the control, respectively (Fig. 2B). By contrast, at 200 g/ml K508A and GST had no effect on IICR (Figs. 2A, b and c, and B) (even at 500 g/ml; data not shown). Ca 2ϩ release induced with 40 mM caffeine, which activates the ryanodine receptor (21), was unaffected by the addition of G224 ( Fig. 2A, insets). Therefore, both G224 and R441Q specifically and dose-dependently inhibited in vitro IICR from microsomes in a competitive manner. Thus, we named G224 an "IP 3 sponge." Carbachol-and ATP-evoked Ca 2ϩ Increases Are Inhibited by IP 3 Sponge Expressed in HEK293 Cells-To analyze the inhibitory effect of the IP 3 sponge on IICR in HEK293 cells, we obtained tetracycline (Tet)-controllable stable transformants (16,17), in which expressions of the IP 3 sponge, K508A, or GST could be induced only by removing doxycycline (Dox) from the culture medium (Fig. 3A). More specifically, we established two IP 3 sponge-expressing cell lines (clone 2 (C2) and clone 17 (C17)). Immunoblotting analysis using total cell lysates demonstrated that the level of IP 3 sponge expression in C2 cells was higher than that in C17 cells. This difference of the level of induced expression was in agreement with immunostaining analysis with anti-GST antibody (Fig. 3B). Exogenous protein expression in this Tet-controllable expression system reached its maximal level within 48 h (data not shown). The IP 3 sponge, K508A, and GST expression-induced HEK293 cells were viable 48 h after induction, as judged by their impermeability to trypan blue (data not shown), and they showed no remarkable changes of cell morphology at a light microscopic level when compared with non-induced cells. The levels of expression of the endogenous IP 3 Rs were unaltered in cells that expressed IP 3 sponge, K508A, or GST (Fig. 3A). The IP 3 sponge and K508A expressed were homogeneously distributed throughout the cytoplasm (Fig. 3B). After removing Dox, significant IP 3 binding was present in the homogenate of the IP 3 sponge expression-induced C2 cells but not in the K508A, GST expression-induced cells, or the native HEK293 cells (Fig. 3C).
To exclude the possibility that the IP 3 sponge affected the production of IP 3 , we examined the activity of PLC in the IP 3 sponge expression-induced C2 cells. As shown in Fig. 3D, the expression of neither the IP 3 sponge nor K508A inhibited the PLC activity in the homogenate of HEK293 cells. Additionally, the presence of the bacterially expressed IP 3 sponge as well as K508A and GST had no effect on the PLC activity of native HEK293 cells (Fig. 3E). These results indicated that the IP 3 sponge expressed in this system was a functional IP 3 -binding protein that did not affect the production of IP 3 and that K508A could be used as a negative control.
We found that there were variations in the level of expres-sion of exogenous proteins from cell to cell in each stable cell line generated by selection with G418 and hygromycin (Figs. 3B, 4A, and 5A), and a similar variety of levels of expression in the Tet-controllable expression system has been described for other proteins (22, 23), although the reason was unknown. We therefore identified the expressing cells immunocytochemically were verified by Western blot analysis with the anti-IP 3 binding region antibody N2 (14) using total cell lysates (100 g) of HEK293 transformants. Protein expression was suppressed and induced in the presence (ϩ) and absence (Ϫ) of Dox, respectively. As a control, lane 1 was for cerebellar microsomes (2 g) prepared from ddY mice; lane 2 was for soluble the IP 3 sponge (G224) (0.05 g) prepared from E. coli, and lane 3 was for a cell lysate (100 g) of non-transformants. The same results were obtained by using polyclonal antibodies N3 and N5 (14). GST expression (31 kDa) was immunoblotted with the anti-GST antibody (bottom). B, photomicrographs showing K508A or IP 3 sponge expression induced by removing Dox for 48 h; C2 or C17 HEK293 cells, which expressed high or low concentrations of the IP 3 sponge, respectively. Ph, phase contrast images; FITC, the same HEK293 cells stained with anti-GST antibody, including induced K508A cells that did not show any immunoreactivity (open arrowheads). Scale bar represents 10 m. C, IP 3 binding to the IP 3 sponge expressed in HEK293 cells. Specific [ 3 H]IP 3 (9.6 nM) binding to sonicated nuclei-free S1 fractions (100 g) prepared from IP 3 sponge expression-induced C2 cells, K508A or GST expression-induced cells, non-induced C2 cells, and parental HEK293 cells. The result is the mean of two individual experiments. D, PLC activities of sonicated nuclei-free S1 fractions (130 g) prepared from IP 3 sponge expression-induced C2 cells, K508A expression-induced cells, and non-induced C2 cells. The productions of radioactive IP 3 by hydrolyzing [ 3 H]PIP 2 with endogenous PLC activity were measured as described under "Experimental Procedures." The result is the mean of two individual experiments. E, effect of the bacterially expressed GSTfused IP 3 -binding proteins on PLC activity of HEK293 cells. The total activity of the control cell homogenates used was 1900 dpm. PLC activities using 1 pmol PIP 2 as a substrate were measured in the presence of IP 3 sponge, K508A, or GST, in which IP 3 -binding sites (1, 10, and 100 pmol) were calculated by the B max value (Fig. 1C).
with anti-GST antibody (indicated as FITC) after each experiment and used induced cells that did not show any immunoreactivity (open arrowheads in the top column in Fig. 3B, and in the middle column in Figs. 4A and 5A) as an internal control in the same culture dish.
We used these stable transformants to analyze IICR in response to stimulation with the muscarinic agonist CCh (24) in the absence of extracellular Ca 2ϩ (Fig. 4). There was no significant difference in resting levels of [Ca 2ϩ ] i regardless of expression of the exogenous proteins (Fig. 4A, Rest). In the noninduced C2 cells in which expression of the IP 3 sponge was suppressed in the presence of Dox (Fig. 4, B, a, and C, a) and K508A-expressing cells (Fig. 4, A, top column, B, b, and C, b), stimulation with 20 M CCh evoked a rapid rise in [Ca 2ϩ ] i followed by oscillations with decreasing amplitude, and stimu-lation with 200 M CCh triggered a larger increase in [Ca 2ϩ ] i with a slowly decreasing phase. The same results were obtained in GST-expressing cells and non-induced K508A or GST cells in which expression of K508A or GST was suppressed, respectively (data not shown).
In marked contrast, such an increase in [Ca 2ϩ ] i evoked by 200 M CCh was completely inhibited in sponge-expressing C2 cells (Fig. 4, A, middle column, and B, c), whereas induced C2 cells that did not show any immunoreactivity (open arrowheads in Fig. 4A, middle column) displayed the same Ca 2ϩ responses as K508A-expressing cells. Neither the Ca 2ϩ increase evoked by 20 or 200 M CCh was detected in 60 -70% of spongeexpressing C2 cells (Fig. 4, B, c, and C, c), and the other cells also displayed a smaller and often delayed (29 Ϯ 18, mean Ϯ S.D., seconds) rise in [Ca 2ϩ ] i (filled arrowheads in Fig. 4A,  , and B, d). Thus, the exogenously expressed IP 3 sponge was able to inhibit significantly CCh-evoked IICR (Fig.  4, D and C2; p Ͻ 0.001).
We next examined whether the inhibitory effect of IP 3 sponge would be influenced by the expression levels. As in the result with C2 cells, we observed a significant small and often delayed Ca 2ϩ response in sponge-expressing C17 cells after CCh stimulation (Fig. 4, A, bottom column, B, e, C, d, and D, C17; p Ͻ 0.001). However, no Ca 2ϩ increase in response to stimulation with 20 M CCh was seen in only 26% of the sponge-expressing C17 cells (Fig. 4C, e), indicating that the inhibitory effect in a high level sponge-expressing cell line, C2 cells, could be significantly more potent than that in a low level sponge-expressing cell line, C17 cells ( Fig. 4D; p Ͻ 0.001, Student's t test). These results indicate that the exogenously expressed IP 3 sponge reached a concentration sufficient to trap IP 3 , resulting in the complete inhibition of IICR.
The subsequent Ca 2ϩ oscillation induced by stimulation with 20 M CCh was not detected in most of the sponge-expressing C2 and C17 cells (Fig. 4C, d). Interestingly, some IP 3 spongeexpressing cells displayed a small rise in [Ca 2ϩ ] i in response to 200 M CCh followed by oscillation (Fig. 4B, e) that resembled the response elicited by 20 M CCh in the non-induced cells.
We also investigated whether IP 3 sponge would inhibit the Ca 2ϩ response to the purinergic agonist ATP (25) under the same conditions we used to measure the response to CCh. Application of ATP to K508A-expressing cells (Fig. 5, A, top  column, B, b, and C, b) and GST-expressing cells (data not shown) as well as all non-induced cells (Fig. 5, B, a, and C, a) caused a steep rise in [Ca 2ϩ ] i that depended on the ATP concentration. In contrast, a smaller and delayed Ca 2ϩ increase was typically seen in sponge-expressing C2 cells (filled arrowheads in Fig. 5, A, middle column, B, c, C, c, and D, C2; p Ͻ 0.001), whereas induced C2 cells that did not show any immunoreactivity (open arrowheads in Fig. 5A, middle column) displayed marked Ca 2ϩ release. Moreover, in the same manner as CCh-evoked IICR, complete inhibition of the increase in [Ca 2ϩ ] i was observed in 20 -30% of sponge-expressing C2 cells stimulated with both 10 and 100 M ATP (Fig. 5, A, middle column,  B, d, and C, d). Most Ca 2ϩ increases evoked by ATP in spongeexpressing C17 cells were significantly reduced (Fig. 5, B, e, C, e, and D, C17; p Ͻ 0.001) and often delayed (filled arrowheads in Fig. 5A, bottom column). However, no rise in Ca 2ϩ was detected in only 13% of cells stimulated with 10 M ATP (data not shown). These findings show that IP 3 sponge could also efficiently trap IP 3 produced in response to ATP and significantly inhibited ATP-evoked IICR. All IP 3 sponge-expressing cells examined after CCh or ATP stimulation showed obvious 0.2 M thapsigargin (TG)-induced Ca 2ϩ release as well as K508A-and GST-expressing cells and all non-induced cells (Fig. 4, B and C, and Fig. 5, B and C), indicating that the inhibition of IICR by IP 3 sponge did not result from a lack of Ca 2ϩ storage.
IP 3 Trapping by the IP 3 Sponge Reduces CCh-induced CREB Phosphorylation in HEK293 Cells-We next analyzed whether trapping of IP 3 by the IP 3 sponge would influence cell physiology downstream. It has been well established that Ca 2ϩ signaling patterns in the form of single transients and sustained plateaux, including repetitive oscillations generated by cell permeant caged IP 3 ester (26), play a critical role in gene transcription (27). One of the responses to an increase in [Ca 2ϩ ] i is activation of the transcription factor CREB (28, 29). Although the results of several studies have implied that influx of extracellular Ca 2ϩ via L-type Ca 2ϩ channels or N-methyl-Daspartate (NMDA) receptors acts on the phosphorylation of Ser-133 of CREB (11,12), it was not fully understood whether IICR could also induce CREB phosphorylation. Sutton et al. (30) reported that application of CCh to HEK293 cells activated the transcription of the target gene through a Ca 2ϩ -dependent pathway. Therefore, to determine whether IICR would be involved in the activation of CREB, we examined the influence of overexpression of the IP 3 sponge in HEK293 cells on CREB phosphorylation by using an antibody that specifically recognizes the phosphorylated serine 133 of CREB (anti-phospho-CREB). As shown in Fig. 6A, application of 20 M CCh increased immunoreactivity to anti-phospho-CREB about 2-fold in all non-induced cells, even in the absence of extracellular Ca 2ϩ . Neither exogenously expressed K508A nor GST influ-enced the level of CREB phosphorylation in cells treated with 20 M CCh (Fig. 6). In contrast, the level of CREB phosphorylation in IP 3 sponge expression-induced C2 cells was less than half that in non-induced C2 cells (Fig. 6, A and B, n ϭ 8; p Ͻ 0.001, Student's t test), indicating that the IP 3 sponge had a significant inhibitory effect on the CCh-induced CREB phosphorylation. Immunocytochemical analysis of CREB phosphorylation using anti-phospho-CREB antibody (Fig. 6C) showed that the nuclear immunoreactivity of K508A-expressing cells treated with 20 M CCh (Fig. 6C, d) was up-regulated in comparison with that of unstimulated cells (Fig. 6C, a and b). In IP 3 sponge-expressing C2 cells (Fig. 6C, c), CREB phosphorylation triggered by 20 M CCh was almost completely inhibited, whereas the increase of nuclear immunoreactivity to anti-phospho-CREB antibody was clearly observed in both non-induced C2 cells (Fig. 6C, e) and induced C2 cells that did not show the expression of IP 3 sponge (arrowhead in Fig. 6C, c). On the other hand, trapping IP 3 did not influence 20 M forskolin-induced CREB phosphorylation (Fig. 6, A and B, n ϭ 3-4), suggesting that the IP 3 sponge specifically inhibited IICR-dependent CREB phosphorylation. DISCUSSION We have developed a novel recombinant IP 3 -absorbent protein, called the IP 3 sponge, that has Ͼ1000-fold higher affinity for IP 3 than the endogenous IP 3 R, and we found that it specifically and dose-dependently inhibited in vitro IICR in a competitive manner. Exogenously expressed IP 3 sponge was also found to trap efficiently IP 3 and inhibit IICR in living cells, depending on the level of expression. Trapping IP 3 with the IP 3 sponge significantly reduced CCh-induced CREB phosphorylation at Ser-133.
All of the inhibitors of IICR reported previously have targeted IP 3 R, and there have been problems with specificity, isoform selectivity, and/or permeability. Application of heparin, a competitive inhibitor for IP 3 binding (31), is limited, because it performs actions on multiple targets in cells and is not membrane-permeable. Microinjection of the monoclonal antibody 18A10 specifically inhibits IICR via type 1 IP 3 R by binding to its C terminus (32), but this approach cannot be applied to phenomena likely to involve type 2 or 3 IP 3 R, and it is difficult to deliver efficiently this antibody inside cells (33). There are two potent membrane-permeable IICR inhibitors that do not block IP 3 binding, the xestospongins, a group of macrocyclic bis-1-oxaquinolizidines isolated from the Australian sponge (34), and 2-aminoethoxy diphenylborate, which we developed previously (35). However, Missiaen and colleagues recently reported that xestospongin C (36) and 2-aminoethoxy diphenylborate (37) are not specific for IICR, because both equally block the endoplasmic reticulum Ca 2ϩ pump. In the present study, we have demonstrated that the IP 3 sponge could overcome these problems.
One of the unique properties of the IP 3 sponge is that it directly traps IP 3 signals. The specificity of the IP 3 sponge for various inositol phosphates was similar to that of native cerebellar IP 3 R (Fig. 1D), indicating that the IP 3 sponge retained specific IP 3 binding properties that were the same as those of endogenous IP 3 R. The inhibitory effect of IP 3 sponge on in vitro IICR from microsomes (Fig. 2) resulted from trapping extravesicular IP 3 in a dose-dependent manner rather than affecting the de novo production of IP 3 . In addition, the caffeine-induced Ca 2ϩ response was unaffected by adding the IP 3 sponge ( Fig.  2A, insets), indicating that the IP 3 sponge had no effect on Ca 2ϩ release from ryanodine-sensitive stores. We also identified that the exogenously expressed IP 3 sponge in HEK293 cells was distributed homogeneously in the cytoplasm (Figs. 3B, 4A, and 5A) and had no effect on the activity of PLC (Fig. 3, D and E). , and mean increases in immunoreactivity to anti-phospho-CREB (Ϯ S.D.) measured by quantitative analysis were normalized to those seen in non-induced cells (B); *, p Ͻ 0.001. No marked differences in basal immunoreactivity were seen between unstimulated C2 and K508A cells. The extracts were immunoblotted for CREB to demonstrate equal protein loading (A, bottom, CREB). C, confocal immunofluorescence microscopy analysis of 20 M CCh-triggered CREB phosphorylation in HEK293 cells. The increases in nuclear immunoreactivity following CCh treatment (c, d and e) were detected by anti-phospho-CREB antibody (green). IP 3 sponge (a and c)-or K508A (b and d)expressing cells were visualized using anti-GST antibody (red). Arrowheads indicated induced C2 cells (c) and induced K508A cells (b) that showed no or less immunoreactivity with anti-GST antibody than the non-induced C2 cells (e) did. Similar staining patterns were observed in four separate experiments. Scale bar represents 10 m. Thus, IP 3 sponge could trap the increased intracellular IP 3 without affecting the production of IP 3 . Because the inhibition of IICR by the IP 3 sponge was due to the superiority in the affinity and specificity for IP 3 over the parental IP 3 R, the inhibitory effect could be enhanced by using mutant R441Q, the substitution of Gln for Arg-441 on the IP 3 sponge (G224), which had double the IP 3 binding activity of the IP 3 sponge.
Another unique feature of the IP 3 sponge is the product encoded in its cDNAs, which allows trapping of intracellular IP 3 by exogenously expressing the IP 3 sponge. Moreover, HEK293 cells express all three subunits of the IP 3 R family (10), and thus the IP 3 sponge appears to be effective in competing for IP 3 with multiple types of tetrameric IP 3 R channels, consisting of various subunit compositions. Therefore, spatiotemporally well controlled expression of the IP 3 sponge will enable it to trap intracellularly produced IP 3 and to approach IP 3 signaling-dependent Ca 2ϩ physiology, irrespective of cell or tissue type. Because the levels of IP 3 trapping seem to influence Ca 2ϩ oscillation, the IP 3 sponge might be useful for the analysis of temporal Ca 2ϩ dynamics. Moreover, targeting the expression of this recombinant IP 3 sponge protein would permit analysis of cell type-and cell stagespecific IICR events or intracellular compartment-specific miniature IICR events.
We also used the IP 3 sponge to measure whether inhibition of IICR would affect activation of the transcription factor CREB, which was crucially dependent on phosphorylation of Ser-133 (28,29). Although it was well known that an increase in [Ca 2ϩ ] i via L-type Ca 2ϩ channels or NMDA receptors acts on CREB phosphorylation (11,12), the involvement of IP 3 -Ca 2ϩ signaling in CREB phosphorylation had never been directly investigated. Finkbeiner et al. (38) reported that the brainderived neurotrophic factor-induced activation of PLC␥ stimulated CREB phosphorylation via IICR, and Hardingham et al. (39) recently showed that IICR was critical for propagation of Ca 2ϩ waves from synaptic NMDA receptors to the nucleus, which consequently induced Ca 2ϩ -dependent phosphorylation of CREB. The present study showed that the up-regulation of CREB phosphorylation in response to 20 M CCh was inhibited in the IP 3 sponge-expressing C2 cells, in which the rise in Ca 2ϩ evoked by 20 M CCh was almost completely blunted. Thus, the application of the IP 3 sponge provided the first direct evidence that IICR enhanced phosphorylation at Ser 133 of CREB. These data suggest a link between IICR and gene transcription via CREB phosphorylation, involved in many neuronal activities, such as development, plasticity, and survival (40 -42). Trapping of IP 3 by the IP 3 sponge actually influenced cell physiology including a downstream cascade and would be beneficial for exploring cellular events mediated by IP 3 -Ca 2ϩ signaling.
In conclusion, we have characterized the recombinant hyperaffinity IP 3 sponge protein, which is useful for exploring IP 3 -Ca 2ϩ signal conversion. This novel inhibitory system should provide a powerful means of analyzing not only the signaling effects of different intracellular IP 3 levels and dynamics but also a variety of aspects of IP 3 -Ca 2ϩ physiology in vivo by expressing cell-, tissue-, and organelle-specific IP 3 sponge.