G-protein-coupled Receptor Kinase-interacting Proteins Inhibit Apoptosis by Inositol 1,4,5-Triphosphate Receptor-mediated Ca2+ Signal Regulation*

The inositol 1,4,5-trisphosphate (IP3) receptor (IP3R) is an intracellular IP3-gated calcium (Ca2+) release channel and plays important roles in regulation of numerous Ca2+-dependent cellular responses. Many intracellular modulators and IP3R-binding proteins regulate the IP3R channel function. Here we identified G-protein-coupled receptor kinase-interacting proteins (GIT), GIT1 and GIT2, as novel IP3R-binding proteins. We found that both GIT1 and GIT2 directly bind to all three subtypes of IP3R. The interaction was favored by the cytosolic Ca2+ concentration and it functionally inhibited IP3R activity. Knockdown of GIT induced and accelerated caspase-dependent apoptosis in both unstimulated and staurosporine-treated cells, which was attenuated by wild-type GIT1 overexpression or pharmacological inhibitors of IP3R, but not by a mutant form of GIT1 that abrogates the interaction. Thus, we conclude that GIT inhibits apoptosis by modulating the IP3R-mediated Ca2+ signal through a direct interaction with IP3R in a cytosolic Ca2+-dependent manner.

Small Interfering RNAs-siRNA duplexes were purchased from B-Bridge as duplexes or a mixture of three duplexes and were originally designed for human GIT1 (accession number NM_014030) and GIT2 (accession numbers NM_139201, NM_057169, NM_014776, and NM_057170). The target sequences were: GIT1-1, 5Ј-CGAGGUGGAUCGAAGAGAA-3Ј; GIT1-2, 5Ј-GCACUGAGCUAGAGGACGA-3Ј; GIT1-3, 5Ј-CCAAGAACAUUCAGGAACU-3Ј; and GIT2, 5Ј-GGAA-AUACAGUAUGAGCUA-3Ј. No difference in the efficiency of knockdown of GIT1 between each of the three siRNAs and the mixture of the three siRNAs were found (data not shown). For knockdown of endogenous GIT1 in HeLa cells, a mixture of the three GIT1 siRNAs or GIT1-2 siRNA was used. For the rescue experiments, HeLa cells transfected with both GIT1-2 siRNA and mouse GIT1 cDNA, which has three substitutions for the GIT1-2 siRNA-targeted sequence (mouse GIT1 sequence, with substitutions underlined, 5Ј-GCACGGAGCUCGAAGACGA-3Ј), were used. No knockdown effect of GIT1-2 siRNA on the expressed mouse GIT1 was detected (data not shown). For FIGURE 1. GIT1 and GIT2 bind to all three subtypes of IP 3 R. A, schematic of ER residential IP 3 R. The CTT of IP 3 R1 is used as bait in a yeast two-hybrid screen. B, schematic representation of GIT1, GIT2, and two GIT1 fragments identified from the yeast two-hybrid screen. Functional domains are indicated. ARF-GAP, ARF-specific GTPase-activating protein domain; ANK-REP, ankyrin repeats; CC, coiled-coil domains; SHD, the Spa2-homology domain; EF, EF-hand; IQ, IQ-like motifs; aa, amino acid. C, GIT1 binds to IP 3 R1 in vitro. GST and GST-IP 3 R1/CTT were incubated with mouse brain lysate for a pull-down assay. The input and pulled-down samples were probed with ␣-GIT1. D and E, GIT1 binds to IP 3 R1 in vivo. Mouse brain lysates were processed to control IgG and ␣-IP 3 R1 (D) or ␣-GIT1 (E) for IP. The input and IP samples were probed with ␣-GIT1 and ␣-IP 3 R1. F and G, both GIT1 and GIT2 bind to all three IP 3 R subtypes. HeLa cells coexpressing GFP-fused IP 3 R1, IP 3 R2, or IP 3 R3 and mRFP-fused GIT1 (F) or GIT2 (G) were processed for IP using ␣-RFP. The input and IP samples were blotted with ␣-GFP (top) and ␣-RFP (bottom). COS-7 cells, GIT1-2 siRNA and GIT2 siRNA, which recognize monkey GIT1 (accession number XM_001108023) and GIT2 (accession number XM_001106038, predicted sequence), respectively, were used. The control siRNA was a mixture of three random siRNAs corresponding to the following sequences: control 1, 5Ј-ATCCGCGCGATAGTACGTA-3Ј; control 2, 5Ј-TTACGCGTAGCGTAATACG-3Ј; control 3, 5Ј-TATTCGCGCGTATAGCGGT-3Ј.
Cell Culture and Transfection-HeLa and COS-7 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. Transfection of plasmids (16) and siRNA (19) were performed as described previously.
Yeast Two-hybrid Assay-A yeast two-hybrid assay was performed as described previously (20). The liquid yeast two-hybrid quantification assay was performed using 5-bromo-4chloro-3-indolyl-␤-D-galactopyranoside (X-gal) as substrate according to the methods described in the Clontech manual.
Pull-down and Immunoprecipitation Assay-Pull-down and immunoprecipitation were performed as described previously (16). For Ca 2ϩ -dependent in vitro and in vivo binding assay, appropriate concentrations of EGTA and Ca 2ϩ were added to lysis buffer and wash buffer.
Measuring Ca 2ϩ Release from Cerebellar Microsomes-Mice cerebellar microsome fractions were incubated with purified His-GIT1 or His-L712A for 10 min on ice, and then IP 3 -induced Ca 2ϩ release from microsomes was measured with Fura-2 and a fluorospectrometer, CAF110 (Jasco, Tokyo, Japan) as described previously (19). Data are presented as mean Ϯ S.D. of at least three independent experiments.
Ca 2ϩ Imaging-Ca 2ϩ imaging was performed as described previously (19) by placing the cells in HEPES-buffered saline with or without 2 mM CaCl 2 . Acquisition was performed with the custom software TI Workbench. Off-line analysis was performed with TI Workbench combined with Igor Pro software (WaveMetrics). The Ca 2ϩ response was quantitated as the area under the response curve (AUC). Data are presented as mean Ϯ S.D. of at least five independent experiments.
TUNEL Staining and Microscopy-TUNEL staining was performed according to the protocol of Chemicon International. Coverslips were mounted using Vectashield mounting medium with 4Ј,6-diamidino-2-phenylindole (Vector Laboratories). Fluorescent images were obtained using a digital microscope A, liquid yeast two-hybrid quantification assay of the interactions between various lengths of IP 3 R/CTT and the short GIT1 fragment identified from the yeast two-hybrid screen. Values of interaction strength were normalized against the interaction between IP 3 R1/CTT and GIT1 fragment. B, sequence alignment of IP 3 R/GITCOREs (top) and schematic of IP 3 R/GITCORE (bottom). The boxes indicate essential residues for the binding. C and D, IP 3 R/GITCORE binds to GIT in vitro. The purified GIT1-(417-716)-His fragment was incubated with purified GST-fused IP 3 R1/GITCORE, IP 3 R2/GITCORE, IP 3 R3/GITCORE, or GST for pull-down assay. The applied and pulled-down samples were detected by Western blotting using ␣-GIT1 (C) and staining with Coomassie Brilliant Blue (CBB; D). E, IP 3 R/GITCORE overlaps the coiled-coil domain in IP 3 R/CTTs. The probability of coiled-coil domains for the whole length of IP 3 R/CTTs was predicted using the method of Lupas et al. (38). Outputs at window size of 14 residues are shown. F, schematic diagram of mRFP-fused wild-type and mutant GIT1 used in this study. G, lysate of HeLa cells expressing mRFP-fused full or partial lengths of GIT1 was incubated with GST-IP 3 R2/GITCORE for the pull-down assay. Applied and pulled-down samples were probed with ␣-RFP. Arrows indicate the peptides that failed to bind. WB, Western blot. H, lysate of HeLa cells expressing mRFP-fused GIT1 wild-type or point mutants were incubated with GST-IP 3 R2/GITCORE for the pull-down assay as described in G. I, sequence comparison of the COOH-terminal region of GIT proteins. Dashed box and arrows indicate necessary and critical residues responsible for binding to IP 3 R, respectively.  OCTOBER 16, 2009 • VOLUME 284 • NUMBER 42

JOURNAL OF BIOLOGICAL CHEMISTRY 29161
(BZ-8000, Osaka, Japan). Data are presented as mean Ϯ S.D. of at least three independent experiments.

RESULTS
Both GIT1 and GIT2 Bind to All Three IP 3 R Subtypes-To understand the function of the CTT of IP 3 R (Fig. 1A), we performed a yeast two-hybrid screen of a mixture of embryonic and adult human brain cDNA libraries with the CTT of IP 3 R1 as bait to search for novel IP 3 R-binding molecules. Nineteen prey clones were obtained. Seven of these clones encoded protein 4.1N fragments (20), and 2 clones encoded the different lengths of the GIT1 fragment (Fig. 1B). The interaction between GIT1 and IP 3 R1/CTT was confirmed by a glutathione S-transferase (GST) pull-down assay (Fig. 1C) and also by an immunoprecipitation (IP) assay from mouse whole brain lysates using anti-IP 3 R1 (Fig. 1D) and anti-GIT1 antibodies (Fig. 1E). To determine whether both GIT1 and GIT2 bind to all three subtypes of IP 3 R, mRFP-fused GIT1 or GIT2 was coexpressed with green fluorescent protein (GFP)-fused IP 3 R1, IP 3 R2, or IP 3 R3, in HeLa cells, and IP was performed using anti-RFP antibody from the cell lysates. All three subtypes of IP 3 R were coprecipitated with GIT1 ( Fig. 1F) and GIT2 ( Fig.  1G), implying that both GIT1 and GIT2 bind to all three subtypes of IP 3 R.
The Coiled-coil Domain in the CTT of IP 3 R Is Involved in the Interaction with GIT-The interaction between the shorter GIT1 fragment identified from the yeast two-hybrid screen and full-length or serial deletions of the CTT of IP 3 R1, IP 3 R2, and IP 3 R3 were examined by a liquid yeast two-hybrid quantification assay to determine the minimal region responsible for IP 3 R binding to GIT. As shown in Fig. 2, A and B, we found that the last 50 (IP 3 R1 and IP 3 R2) or 44 (IP 3 R3) residues within CTT showed 5-8-fold stronger binding to the GIT1 fragment as compared with the full-length CTT of the corresponding IP 3 R subtype. Thereafter, we called these peptides "IP 3 R/GIT-COREs," which means "GIT-binding cores in IP 3 Rs." Within IP 3 R/GITCOREs, the first 10 residues in all three IP 3 R/GIT-COREs and the last 14 residues in both IP 3 R1 and IP 3 R2 or the last 8 residues in IP 3 R3 are essential for the interaction (Fig. 2, A and B). All three purified GST-fused IP 3 R/GITCOREs pulled down the purified GIT1 protein fragment in vitro (Fig. 2, C and  D), indicating that the IP 3 R/GITCORE is sufficient for GIT binding and that the interaction is direct. By searching functional protein motifs within the CTT of IP 3 Rs, we found that a potential coiled-coil domain, which is usually involved in protein-protein interaction, was localized in the CTT of all three IP 3 Rs, and that IP 3 R/GITCORE partially overlapped with the coiled-coil domain (Fig. 2E). These findings suggested that the coiled-coil domain within the CTT of IP 3 R is involved in the GIT-IP 3 R interaction.
The Second IQ-like Motif of GIT Is Critical for Binding to IP 3 R-To determine the precise IP 3 R binding region in GIT, mRFPfused full-length fragment or subfragments of GIT1 (Fig. 2F) were expressed in HeLa cells and subjected to a pull-down assay using GST-IP 3 R2/GITCORE. Peptides containing residues 691-721 of GIT1 bound to IP 3 R, whereas peptides not containing these residues failed to bind (Fig. 2G). Because these residues contain the second IQ-like motif, consisting of LQXXXR (where X represents any amino acid; Fig. 2, F and I), we further examined whether this IQ-like motif of GIT is involved in the binding. Several single, double, and triple point mutants within this IQ-like motif of mRFP-GIT1 (Fig. 2F) were expressed in HeLa cells and subjected to pull-down assay. We found that mRFP-L712A, L712A/Q713A, and L712A/Q713A/R717A abolished the interaction with GST-IP 3 R2/GITCORE (Fig. 2H). Thus, the Leu 712 residue of GIT1 is critical for binding to IP 3 R. By comparing of the homology in the COOH-terminal region of GIT proteins, we found that the Leu 712 of GIT1 is conserved in all GIT proteins of all species (Fig. 2I), which suggests that the GIT-IP 3 R interaction is evolutionarily conserved.
Cytosolic Ca 2ϩ Dynamics Modulates the GIT-IP 3 R Interaction-Cytosolic Ca 2ϩ elevation triggers the conformational change of IP 3 R (21). Alternatively, GIT contains putative EFhand and IQ-like motifs for Ca 2ϩ and calmodulin binding, respectively (Fig. 1B). We therefore investigated whether the GIT-IP 3 R interaction is regulated by a cytosolic Ca 2ϩ concentration by a pull-down assay with GST-IP 3 R2/GITCORE against mRFP-GIT1 expressed in HeLa cells in the presence of various concentrations of Ca 2ϩ . We found that the amount of mRFP-GIT1 pulled down by GST-IP 3 R2/GITCORE increased with serial elevation of the Ca 2ϩ concentration (Fig. 3A). To corroborate this observation in vivo, HeLa cells were stimulated with 10 M ATP, an activator of purinergic receptors (Fig. 3B), or 1.0 M thapsigargin (TG) (Fig. 3C), an inhibitor of the sarcoendoplasmic reticulum Ca 2ϩ -ATPase, to increase cytosolic Ca 2ϩ concentration, and the cell lysates were subjected to the IP assay using anti-GIT1 antibody. We found that both stimulants increased the amount of all three IP 3 R subtypes coprecipitated with GIT1. Thus, the above data suggested that an elevated cytosolic Ca 2ϩ concentration favors GIT-IP 3 R association.
GIT Suppresses IP 3 -gated Ca 2ϩ Release in Vitro-We investigated whether GIT regulates the channel activity of IP 3 R by an in vitro Ca 2ϩ release assay using mouse cerebellar microsomes to explore the functional consequences of the GIT-IP 3 R inter-  OCTOBER 16, 2009 • VOLUME 284 • NUMBER 42 JOURNAL OF BIOLOGICAL CHEMISTRY 29163 action. The microsomes were preincubated with or without recombinant wild-type or mutant GIT1, and the Ca 2ϩ release activity was examined by addition of various concentrations of IP 3 . We found that wild-type GIT1 reduced Ca 2ϩ release in response to low concentrations of IP 3 (2-200 nM), which were in the range of physiological concentrations (22), whereas the L712A point mutant GIT1 markedly attenuated the ability to suppress Ca 2ϩ release (Figs. 4, A and B). These results suggest that the GIT-IP 3 R interaction suppresses the channel activity of IP 3 R by reducing the apparent sensitivity of the channel to IP 3 .

GIT Proteins Inhibit Apoptosis
GIT Inhibits IP 3 -gated Ca 2ϩ Release in Intact Cells-To investigate whether GIT also modulates IP 3 -gated Ca 2ϩ release in intact cells, we imaged Ca 2ϩ signals of HeLa cells expressing mRFP-fused wild-type or mutant GIT1, or mRFP alone in response to 3.0 M ATP stimulation in an extracellular Ca 2ϩfree medium. Whereas the Ca 2ϩ response of cells expressing mRFP or mRFP-L712A was indistinguishable from that of nonexpressing cells, cells expressing mRFP-GIT1 typically exhibited significantly smaller Ca 2ϩ transients (Fig. 4, C and D). The difference in Ca 2ϩ transients in cells expressing mRFP-GIT1 was not due to a decrease in Ca 2ϩ store size within the ER because the amount of passive Ca 2ϩ leakage elicited by TG was unchanged (Fig. 4, E and F). We also observed a similar effect of GIT1 overexpression on IP 3 -induced Ca 2ϩ release in COS-7 cells (Fig. 4, G-J). Hence, we conclude that GIT generally inhibits the IP 3 -gated Ca 2ϩ release in intact cells.
Specific Knockdown of GIT Results in Amplified Intracellular Ca 2ϩ Oscillations and IP 3 R Channel Activity-Next, we used the RNA interference technique to examine the effect of endogenous GIT on IP 3 R channel activity. Because there are no suitable antibodies that specifically recognize GIT2 because of the high homology between GIT1 and GIT2 (13), we used a pan-GIT antibody that almost equally recognizes GIT1 and GIT2 to detect the effect of knockdown of GIT2 in comparison with that of GIT1 (Fig. 5A). We found that GIT1 was robustly expressed in HeLa cells and was efficiently knocked down by GIT1 siRNA. The pan-GIT antibody did not detect any bands in the HeLa cell lysates treated with GIT1 siRNA (Fig. 5B), indicating that GIT2 expression was very low and under detection levels. Alternatively, in COS-7 cells, both GIT1 and GIT2 were robustly expressed and efficiently knocked down by GIT1 and GIT2 siRNA, respectively (Fig. 5C). The expression of all three IP 3 R subtypes was not affected by any of the siRNAs (Fig. 5, B and C). Therefore, we investigated the effects of GIT1 knockdown in HeLa cells and GIT1 and/or GIT2 knockdown in COS-7 cells on the IP 3 R Ca 2ϩ release activity. The siRNA-transfected HeLa cells were consecutively stimulated with 0.3, 1.0, and 10 M ATP in the presence of 2 mM extracellular Ca 2ϩ (Fig. 5D). We found that the knockdown of GIT1 in HeLa cells resulted in more cells showing discernible Ca 2ϩ responses to the low ATP concentration (0.3 and 1.0 M) (Fig. 5E). The average Ca 2ϩ response (AUC) was greater in the GIT1 siRNA-transfected cells than in control cells at all ATP concentrations (Fig. 5H). Furthermore, when GIT1 was knocked down, more cells responded with Ca 2ϩ oscillations and fewer cells responded with Ca 2ϩ spikes to all ATP concentrations (Fig. 5, F and G). The knockdown of GIT1 in HeLa cells had no effect on the level of IP 3 production and the Ca 2ϩ leakage induced by TG (data not shown). These data imply that the knockdown of GIT1, the predominantly expressed form of GIT, in HeLa cells led to more intracellular Ca 2ϩ oscillations and amplified IP 3 R channel Ca 2ϩ release activity. In COS-7 cells, on the other hand, we found that both single knockdown of either GIT1 or GIT2 and double knockdown of GIT1 and GIT2 significantly amplify intracellular Ca 2ϩ oscillations and IP 3 R Ca 2ϩ release activity (Fig. 5,  J-M). These findings demonstrate that knockdown of GIT proteins generally amplifies intracellular Ca 2ϩ oscillations and IP 3 R Ca 2ϩ release activity.
GIT Knockdown Induces Caspase-dependent Apoptosis-We noticed that many cells transfected with GIT siRNAs died, and therefore we examined whether apoptosis occurred in these cells. The GIT1 siRNA-transfected HeLa cells showed a significantly increased number of TUNEL-positive cells as compared with non-siRNA and control siRNA-transfected cells (Fig. 6, A  and C). The increased number of TUNEL-positive cells was abolished by preincubation with Z-VAD-fmk, a pan-inhibitor for caspases (Fig. 6, B and C), confirming the involvement of caspase activation in GIT1 siRNA-transfected cells. Cleaved caspase-9 and caspase-3 were consistently detected only in cells transfected with GIT1 siRNA (Fig. 6D, lane 3); Z-VAD-fmk efficiently blocked the activation of these two caspases (Fig. 6D,  lane 6).
GIT Knockdown Accelerates Staurosporine (STS)-stimulated Apoptosis-It is known that apoptosis-inducible stimuli, such as STS (23)(24)(25) and ceramide (26), can induce IP 3 R Ca 2ϩ release. We therefore examined the effect of GIT knockdown on STS-stimulated apoptosis to further investigate the role of GIT in apoptosis. Forty-eight hours after transfection with GIT1 siRNA, we treated the HeLa cells with 1 M STS for 4 h and examined the number of apoptotic cells. We observed a significant increase in TUNEL-positive cells among the GIT1 siRNA-transfected cells (Fig. 6, E and G). STS-stimulated apoptosis in the GIT1 siRNA-transfected cells was abolished by Z-VAD-fmk treatment (Fig. 6, F and G). In addition, Western blot analysis revealed the activation of caspase-9 and caspase-3 only in GIT1 siRNA-transfected cells that were treated with STS for 4 h (Fig. 6H, lane 9). Although the activation of caspase-9 and caspase-3 was detected even in control cells after STS treatment for 8 h (Fig. 6H, lanes 13-15), it was much stronger in cells transfected with GIT1 siRNA (Fig. 6H, lane 15).  Activated caspase-9 and caspase-3 were not detected when the cells were pretreated with Z-VAD-fmk (Fig. 6H, lanes 12 and  16 -18). Altogether, these data demonstrate that GIT has a critical role in protection against caspase-dependent apoptosis.
GIT Inhibits STS-stimulated Apoptosis by Interacting with IP 3 R-We then examined the effect of overexpression of the wild-type or L712A-mutant GIT on apoptosis in GIT1 siRNAtransfected HeLa cells to verify whether GIT inhibits apoptosis by modulating IP 3 R channel activity. We found that the overexpression of mRFP-GIT1 attenuated the acceleration of apoptosis in GIT1 siRNA-transfected cells induced by STS treatment for both 4 (Fig. 7A, lane 11) and 8 h (Fig. 7A, lane 17). In contrast, overexpression of mRFP-L712A abolished the attenuating effect: apoptosis observed in cells expressing mRFP-L712A (Fig. 7A, lanes 12 and 18) was similar in cells expressing mRFP alone (Fig. 7A, lanes 10 and 16). We also found that apoptosis in the control HeLa cells, triggered by 8 h STS treatment, was attenuated by the wild-type and not the mutant form of GIT1 (Fig. 7A, lanes 13-15). Furthermore, application of 2-aminoethoxydiphenyl borate (Fig. 7B, lane 4) and xestospongin C (Fig. 7B, lane 5), two pharmacological inhibitors of IP 3 R, inhibited the acceleration of apoptosis in GIT1 siRNA-trans-fected cells. Altogether, we conclude that the inhibitory effect of GIT on caspase-dependent apoptosis is caused by the modulation of cytosolic Ca 2ϩ dynamics achieved by the control of IP 3 R channel Ca 2ϩ release activity through direct GIT-IP 3 R interaction.

Inhibition of Apoptosis by GIT Regulation of IP 3 R-mediated
Ca 2ϩ Signal-Here, we found that GIT functions as an antiapoptotic modulator by inhibiting IP 3 R activity. Based on the following two observations, we conclude that GIT presumably inhibits channel activity by elevating the threshold of the IP 3 concentration necessary for gating of the channel. First, purified GIT1 reduced IP 3 R channel activity in vitro in response to a low concentration of IP 3 (Fig. 4, A and B). Second, knockdown of both GIT proteins resulted in more cells responding to a lower but not to a higher concentration of the IP 3 -generating agonist without a change in the level of IP 3 production (Fig. 5, D-M; data not shown). Based on these data, we propose a model by which GIT regulates apoptosis through IP 3 R (Fig. 7, C-E). In resting cells, in which the cytosolic Ca 2ϩ concentration is low (ϳ10 Ϫ7 M), relatively small amounts of GIT bind to IP 3 R with low affinity and desensitize IP 3 R to IP 3 that may exist in unstimulated cells, thereby inhibiting the opening of IP 3 Rs (Fig. 7C). When the cells are stimulated and the IP 3 concentration rises, IP 3 R binds to IP 3 and releases Ca 2ϩ from the ER (Fig.  7D). Subsequently, the elevated cytosolic Ca 2ϩ concentration enhances GIT-IP 3 R association, which in turn inhibits IP 3 R channel activity as a negative feedback (Fig. 7E), thereby inhibiting the elevation of excessive cytosolic Ca 2ϩ and preventing the activation of the caspase cascade. Interestingly, several key regulators of apoptotic signaling, such as cytochrome c (9), Bcl-2 (10), and Bcl-x L (12), have been reported to interact with the CTT of IP 3 R (Rong et al. (11) argued that Bcl-2 interacts with the internal coupling domain, but not the CTT of IP 3 R) and modulate IP 3 R activity by changing the apparent affinity for IP 3 . Together with previous findings that cysteine mutation within the CTT (3) or deletion of the coiled-coil domain from the CTT (27) impairs channel activity, the CTT of IP 3 R, by itself and with accessory proteins, plays a critical role in regulating IP 3 R channel opening. Because the CTT is reported to contribute to IP 3 R tetramer stability (27), it would be interesting to study the relationship between the effect of accessory proteins on structural changes of IP 3 R and the channel activity.
Why does GIT knockdown increase the number of cells showing Ca 2ϩ oscillations in cells (Fig. 5, D, F, I, and K)? One possible explanation is the affinity of GIT for each IP 3 R subtype. Previous studies have found that each IP 3 R subtype has different biophysical properties; for example, IP 3 R1 and IP 3 R2 tend to generate Ca 2ϩ oscillations, whereas IP 3 R3 acts as an anti-Ca 2ϩ oscillatory unit in HeLa and COS-7 cells (28). IP 3 R2 is required for long-lasting regular Ca 2ϩ oscillations, whereas IP 3 R3 generates only monophasic Ca 2ϩ responses in DT40 cells (29). Kuroda et al. (30) recently demonstrated that IP 3 R2 contributes to Ca 2ϩ oscillation in osteoclasts, using IP 3 R knock-out mice. Although we should regard these results with caution because of the presence of so many regulatory factors of IP 3 R activity in vivo (for example, Ca 2ϩ , IP 3 , ATP, accessory proteins, and IP 3 R localization), these papers suggest that both IP 3 R2 and IP 3 R1, particularly IP 3 R2, generally contribute to the generation of Ca 2ϩ oscillations in vivo. Judging from our data showing that GIT seems to bind to IP 3 R2 and IP 3 R1 with higher affinity compared with IP 3 R3 in vivo (Figs. 1, F and G, and 2A), GIT may predominantly inhibit IP 3 R1 and IP 3 R2 that contribute to the generation of Ca 2ϩ oscillations in HeLa and COS-7 cells. Thus, GIT knockdown would increase the number of cells showing Ca 2ϩ oscillations.
Regulation of the GIT-IP 3 R Interaction by Cytosolic Ca 2ϩ Dynamics-The precise mechanism by which Ca 2ϩ enhances the GIT-IP 3 R interaction (Fig. 3) is still unknown, although conformational changes in IP 3 R and/or GIT may underlie this Ca 2ϩ -dependent interaction. It was demonstrated that the NH 2 -terminal IP 3 binding region, which was shown to bind to the S4 -S5 linker near the CTT (27), relocates from the corners of the square form to the corners of the radial wings of the windmill form in the presence of Ca 2ϩ (21). This Ca 2ϩ -dependent spatial rearrangement of the NH 2 -and COOH-terminal regions of IP 3 R may at least in part facilitate the accessibility of GIT to the CTT of IP 3 R. Alternatively, Ca 2ϩ may change the conformation of GIT. Because GIT has an IQ-like motif critical for IP 3 R binding (Fig. 2, F-I) and some interactions between proteins containing IQ-like motifs and their partners can be regulated by Ca 2ϩ (31), the putative Ca 2ϩ -binding EF-hand of GIT (Fig. 1B) might be responsible for the Ca 2ϩ -dependent GIT-IP 3 R interaction. Further studies on the structural and biochemical regulation of GIT and IP 3 R by Ca 2ϩ will be necessary to elucidate the precise mode of their Ca 2ϩ -dependent interaction.
Significance of the Identification of GIT as an Anti-apoptotic Molecule-Our data will shed light on the diverse physiological and pathological significance of GIT-dependent apoptotic processes in the future because apoptosis is not only essential for normal tissue development and homeostasis but also contributes to many forms of pathological cell loss. For example, the COOH-terminal fragment of GIT1, which was shown to bind to IP 3 R in this study, has been found in Huntington disease but not in the healthy brain (32) and decreases the number of synapses and neurite formation (33,34). Nef, a protein that is encoded by the human immunodeficiency virus (HIV)-1 and HIV-2, which can induce apoptosis in the cells of the immune system (35), has been reported to associate with GIT in HIV-1infected primary T lymphocytes and macrophages (36), and with IP 3 R1 in Nef-transfected Jurkat T cells and HIV1-infected primary human peripheral mononuclear cells (37). It would be interesting to investigate the possible association of the fulllength and/or COOH-terminal fragment of GIT with IP 3 R and its effect on neurogenesis, neurodegeneration, and HIV infection, an endeavor that might yield new diagnostic or therapeutic tools.