Regulation of Inositol 1,4,5-Trisphosphate 3-Kinases by Calcium and Localization in Cells*

Inositol 1,4,5-trisphosphate (Ins(1,4,5)P3) 3-kinases (IP3Ks) are a group of calmodulin-regulated inositol polyphosphate kinases (IPKs) that convert the second messenger Ins(1,4,5)P3 into inositol 1,3,4,5-tetrakisphosphate. However, what they contribute to the complexities of Ca2+ signaling, and how, is still not fully understood. In this study, we have used a simple Ca2+ imaging assay to compare the abilities of various Ins (1,4,5)P3-metabolizing enzymes to regulate a maximal histamine-stimulated Ca2+ signal in HeLa cells. Using transient transfection, we overexpressed green fluorescent protein-tagged versions of all three mammalian IP3K isoforms, including mutants with disrupted cellular localization or calmodulin regulation, and then imaged the Ca2+ release stimulated by 100 μm histamine. Both localization to the F-actin cytoskeleton and calmodulin regulation enhance the efficiency of mammalian IP3Ks to dampen the Ins (1,4,5)P3-mediated Ca2+ signals. We also compared the effects of the these IP3Ks with other enzymes that metabolize Ins(1,4,5)P3, including the Type I Ins(1,4,5)P3 5-phosphatase, in both membrane-targeted and soluble forms, the human inositol polyphosphate multikinase, and the two isoforms of IP3K found in Drosophila. All reduce the Ca2+ signal but to varying degrees. We demonstrate that the activity of only one of two IP3K isoforms from Drosophila is positively regulated by calmodulin and that neither isoform associates with the cytoskeleton. Together the data suggest that IP3Ks evolved to regulate kinetic and spatial aspects of Ins (1,4,5)P3 signals in increasingly complex ways in vertebrates, consistent with their probable roles in the regulation of higher brain and immune function.

Inositol 1,4,5-trisphosphate (Ins(1,4,5)P 3 ) 3-kinases (IP 3 Ks) are a group of calmodulin-regulated inositol polyphosphate kinases (IPKs) that convert the second messenger Ins(1,4,5)P 3 into inositol 1,3,4,5-tetrakisphosphate. However, what they contribute to the complexities of Ca 2؉ signaling, and how, is still not fully understood. In this study, we have used a simple Ca 2؉ imaging assay to compare the abilities of various Ins (1,4,5)P 3metabolizing enzymes to regulate a maximal histamine-stimulated Ca 2؉ signal in HeLa cells. Using transient transfection, we overexpressed green fluorescent protein-tagged versions of all three mammalian IP 3 K isoforms, including mutants with disrupted cellular localization or calmodulin regulation, and then imaged the Ca 2؉ release stimulated by 100 M histamine. Both localization to the F-actin cytoskeleton and calmodulin regulation enhance the efficiency of mammalian IP 3 Ks to dampen the Ins (1,4,5)P 3 -mediated Ca 2؉ signals. We also compared the effects of the these IP 3 Ks with other enzymes that metabolize Ins(1,4,5)P 3 , including the Type I Ins(1,4,5)P 3 5-phosphatase, in both membrane-targeted and soluble forms, the human inositol polyphosphate multikinase, and the two isoforms of IP 3 K found in Drosophila. All reduce the Ca 2؉ signal but to varying degrees. We demonstrate that the activity of only one of two IP 3 K isoforms from Drosophila is positively regulated by calmodulin and that neither isoform associates with the cytoskeleton. Together the data suggest that IP 3 Ks evolved to regulate kinetic and spatial aspects of Ins (1,4,5)P 3 signals in increasingly complex ways in vertebrates, consistent with their probable roles in the regulation of higher brain and immune function.
From the first studies of the IP 3 Ks, it has been assumed that because they are enzymes with a higher affinity for Ins (1,4,5)P 3 than the Type I Ins(1,4,5)P 3 5-phosphatase (26) and differ from the 5-phosphatase in that they are positively regulated by CaM and CaM kinase (see Refs. 13,14,27, and 28 for reviews), they must play a preferential role in regulating the pools of Ins (1,4,5)P 3 involved in signaling. However, the exact nature of this contribution is still far from clear, since type I Ins (1,4,5)P 3 5-phosphatases are also very effective attenuators of Ins(1,4,5)P 3 signals (29 -31). The relative contributions of the kinase and phosphatase pathways of Ins(1,4,5)P 3 removal have been explored genetically (31) and in cells (29,32), and these studies show that the phenotypes of Ca 2ϩ signal attenuation are not identical. Mathematical modeling suggests that the two enzyme types can differentially regulate Ca 2ϩ oscillations (22). Because type I Ins(1,4,5)P 3 5-phosphatases have a higher affinity and lower V max for Ins(1,3,4,5)P 4 than for Ins(1,4,5)P 3 , in some cases the product of the IP 3 Ks may saturate the 5-phosphatase and protect Ins(1,4,5)P 3 from dephosphorylation (33). IP 3 Ks are composed of at least two main functional domains. The C-terminal domain is highly conserved among species and isoforms and contains the catalytic activity, whereas the aminoterminal regions dictate further regulation and targeting of the enzyme. The catalytic structures of the A (34,35) and B (36) isoforms have been solved, revealing how nature has elaborated on the core IPK structure in a way that restricts and optimizes its specificity, catalyzing only the Ins (1,4,5)P 3 3-kinase reaction. In contrast, highly divergent N-terminal domains enable the IP 3 Ks to possess significant diversity, both via signal feedback from Ca 2ϩ /CaM (37,38) and by intracellular localization. In neurons, for example, IP 3 KA is concentrated in dendritic spines by way of an interaction with the filamentous actin (Factin) cytoskeleton (39), a localization that is subject to dynamic regulation by Ca 2ϩ (40). Similarly, IP 3 KB can associate with F-actin (41), but it can also be targeted to the endoplasmic reticulum (24), a targeting that may be dynamically regulated by proteases (25). IP 3 KC has been reported to be cytosolic in humans (24), yet it is also reported to actively shuttle between the cytoplasm and the nucleus in the rat (42,43).
Dewaste et al. (24), showed that it is possible to gain some insight into the in vivo efficacy of different IP 3 K constructs by transiently transfecting them into cells and then assessing their impact on Ca 2ϩ mobilization driven by Ins(1,4,5)P 3 . We recently used IP 3 KB overexpression to demonstrate the differing efficacy of IP 3 KB when it is bound to the cytoskeleton compared with when is cytosolic (25), thus demonstrating that localization of IP 3 KB alters its ability to modulate Ins(1,4,5)P 3 levels. Here we report a more systematic exploration of IP 3 K function using the same simple assay system (25), and we have extended the analysis to take in the individual patterns of Ca 2ϩ evoked in single cells. Moreover, we have directly addressed the impact of endogenous CaM on overexpressed IP 3 K in cells by using mutant enzymes that do not bind CaM. Third, we have explored the effect of localization of IP 3 KA and Type I Ins(1,4,5)P 3 5-phosphatase on the Ca 2ϩ signal. By comparing directly all the IP 3 Ks with the two Drosophila isoforms, the mammalian IPMK and the Type I Ins(1,4,5)P 3 5-phosphatase in the same cellular model, we reveal new insights into how IP 3 Ks sculpt Ca 2ϩ signals and how their regulation has evolved.

cDNA Constructs
cDNA clones encoding each of the enzymes of interest were used as PCR templates, and the amplified inserts were subcloned into the pEGFP vectors (Clontech) for expression as GFP fusion proteins or the pQE30 vector for expression in bacteria. All PCRs were initiated by the addition of the Pfu Turbo polymerase (Stratagene) and performed in the presence of Me 2 SO (10%, v/v). Unless otherwise stated, PCRs were allowed to proceed for 25 cycles with cycling segments of 1 min at 96°C, 1 min at 60°C, and 3 min at 72°C. The identity and cloning junctions in all constructs were confirmed by restriction endonuclease mapping and DNA sequencing. Details are as follows.
InsP 3 3-Kinase A-Portions of cDNA encoding full-length and truncated (amino acids 66 -489) rat InsP 3 3-kinase A were amplified and subcloned into the HindIII and BamHI sites of the pEGFPC3 vector, as described in Ref. 39.
InsP 3 3-Kinase B-The full-length cDNA of rat InsP 3 3-kinase B (44), a gift from Dr. G Banting (Department of Biochemistry, University of Bristol, Bristol, UK) was amplified by PCR, and the purified PCR product was subcloned into the HindIII and BamHI sites of the pEGFPC1 vector. Sequencing of this clone revealed a single point mutation at T220C in the open reading frame, which was corrected using QuikChange (Stratagene) mutagenesis, as described in Ref. 25.
InsP 3 3-Kinase C-The full open reading frame of mouse IP 3 KC cDNA was amplified by PCR from IMAGE Clone 6508198, obtained from the United Kingdom Human Genome Mapping Project Resource Centre (UK HGMPRC, Babraham, Cambridge, UK). The purified PCR product was subcloned into the HindIII and BamHI sites of the pEGFPC1 vector.
IPMK-The full open reading frame of human IPMK was amplified by PCR from IMAGE clone 4510867 (UK HGMPRC). The purified PCR product was subcloned into KpnI and BamHI sites of the pEGFPC1 vector. A superfluous nucleotide at position 451 was deleted using QuikChange site-directed mutagenesis, as previously described in Ref. 45.
The full open reading frame of dmIP 3 K␣ (also called IP3K1; GenBank TM AY061607, CG4026) was amplified by PCR from Resgen clone SD07279 (Invitrogen) and subcloned into the HindIII and BamHI sites of the pEGFPC1 vector. For expression in bacteria, the protein was truncated in an attempt to improve solubility. Hence, the cDNA encoding amino acids 103-441 of dmIP 3 K␣ was subcloned into the pQE30 vector for expression as a hexahistidine fusion protein in Escherichia coli. dmIP 3 K␤-The open reading frame of dmIP 3 K␤ (also called IP 3 K2; GenBank TM AY084158, CG1630) was amplified by PCR from Resgen clone RE35745 (Invitrogen) and subcloned into the KpnI and BamHI sites of the pEGFPC1 vector. Again, for expression in bacteria, the protein was truncated in an attempt to improve solubility. Hence, the cDNA encoding amino acids 319 -669 of dmIP 3 K␤ was subcloned into the pQE30 vector for expression as a hexahistadine fusion protein in E. coli.
Type I Ins(1,4,5)P 3 5-Phosphatase-A plasmid containing the open reading frame of the Type 1 Ins(1,4,5)P 3 5-phosphatase in vector pCDNA3 (29) was the kind gift of Dr. C. Erneux (Institut de Récherche Interdisciplinaire, Université libra de Bruxelles, Brussels). This was used as a PCR template, and the PCR products were cloned into the KpnI/BamHI sites of pEGFPC1 and pEGFPN1 to make GFP fusions.

Site-directed Mutagenesis
The enhanced GFP constructs of IP 3 KA, -B, and -C and dmIP 3 K␤ were used as PCR templates for QuikChange mutagenesis. Primers were designed to introduce a single point mutation at the Trp residue homologous to Trp 165 in rat InsP 3 3-kinase A: for InsP 3 3-kinase A W165R, 5Ј-GCAGAAAAGC-CATCGGCAGAAGATCCG-3Ј; for InsP 3 3-kinase B W636R, 5Ј-GTGAGCAAGTCAAGGAGGAAGATAAAG-3Ј; for InsP 3 3-kinase C W383R, 5Ј-GCGCTGGAAGCAAGCCCAGGAA-GAAGCTGAAGACAG-3Ј; for dmIP 3 K␤ W337R, 5Ј-GAAGA-GCTCCGGCAGGCGAAAGATTCGC-3Ј. Mutagenesis PCRs were initiated by the addition of the Pfu Turbo polymerase (Stratagene) and performed in the presence of Me 2 SO (10%, v/v) and consisted of 19 cycles of 30 s at 96°C, 1 min at 55°C, and 14 min at 68°C. Parental template digestion was performed by incubation of the PCR product with 10 units of DpnI (Stratagene) for 2 h at 37°C. The kinase-dead mutant of IP 3 KA was the double mutant D260A/K262A.

Cell Culture and Transfection
Both COS-7 and HeLa cells were cultured routinely in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% heat-inactivated fetal bovine serum (Invitrogen), 100 units/ml penicillin, and 100 mg/ml streptomycin (Invitrogen). HeLa cells were grown on poly-D-lysine (Sigma)-coated glass coverslips. COS-7 cells were transfected with various plasmids using Fugene 6 (Roche Applied Science) according to the manufacturer's instructions. HeLa cell transfection was performed according to a calcium phosphate precipitation technique.

CaM Binding Assay
All steps were performed at 4°C. 12 mM ␤-mercaptoethanol and protease inhibitors (Calbiochem Protease Inhibitor Mixture Set III; 1:100 dilution) were present throughout. XL10-Gold E. coli were transformed with pQE30 constructs encoding the proteins of interest. Following induction of protein synthesis, cells were harvested by centrifugation (2500 ϫ g; 15 min) and resuspended in Bugbuster protein extraction reagent (Novagen) supplemented with Benzonase nuclease (Novagen; 1:1000) and protease inhibitors (Calbiochem protease inhibitor mixture set III; 1:100 dilution). The suspension was incubated on a rotor for 20 min before removal of cell debris by centrifugation. Following a single freeze-thaw cycle, crude lysates were clarified by centrifugation (4°C; 13,000 ϫ g; 1 h).
To test CaM binding from bacterial lysates, we adopted a published procedure (46). 10 l of crude lysate was diluted in 800 l of binding buffer (20 mM Tris, pH 7.5, 0.2 mM CaCl 2 , 0.1 M NaCl, 1% Triton X-100, 1% fish skin gelatin) and applied to a 50-l 1:1 slurry of pre-equilibrated CaM-agarose resin (Calbiochem). Samples were incubated for 1 h at 4°C with gentle rotation. The resin was consequently rapidly washed three times in wash buffer (20 mM Tris, pH 7.5, 0.1 mM CaCl 2 , 0.4 M NaCl, 0.5% Triton X-100) at 4°C. The washed resin was then incubated for 15 min at room temperature with 30 l of elution buffer (20 mM Tris, pH 7.5, 10 mM EGTA, 0.4 M NaCl, 0.1% Triton X-100). Eluates were subjected to SDS-PAGE and analyzed by Western blotting. To test CaM binding from COS-7 cell lysates, cells were grown to 50% confluence on 100-mm culture dishes before transfection and then allowed to grow for a further 48 h. Immediately before harvesting of cells, the culture dishes were washed with ice-cold PBS without calcium or magnesium. 500 l of COS lysis buffer (40 mM Tris/HCl, 0.2 mM CaCl 2 , 0.2 M NaCl, 2% Triton X-100, 2% gelatin, 0.02% NaN 3 ), protease inhibitors (Calbiochem protease inhibitor mixture set III; 1:100 dilution), and phenylmethylsulfonyl fluoride (Sigma; 1 mM final concentration) were then added to each dish on dry ice. A disposable scraper was used to collect lysates, which were then frozen to Ϫ20°C. The COS-7 cell scrapings were thawed and sonicated to aid cell lysis, before centrifugation (4°C; 13,000 ϫ g; 30 min). 400 l of crude lysate was diluted 1:1 with distilled water and incubated for 1 h with 50 l of a 1:1 slurry of preequilibrated calmodulin affinity resin (Stratagene). All subsequent steps were performed as outlined above.

SDS-PAGE and Western Blotting
Samples for electrophoresis were subjected to 10% SDS-PAGE followed by transfer to nitrocellulose membranes using a standard protocol. Membranes were blocked for 30 min at 22°C with 5% nonfat dry milk (Marvel) in Tris-buffered saline (50 mM Tris-HCl, pH 7.5, 140 mM NaCl) containing 0.05% Tween 20. Membranes were immunolabeled overnight at 4°C with primary antibody. For visualization of hexahistidinetagged proteins, a mouse anti-His 6 monoclonal antibody (catalog number 631212; BD Biosciences) was used at a 1:4000 dilution. For visualization of GFP fusion proteins, a rabbit anti-GFP polyclonal antiserum (catalog number Ab-290-50; Abcam) was used at a 1:1000 dilution. Following five washes over 30 min, the membranes were incubated with species-specific horseradish peroxidase-conjugated secondary antibody (Amersham Biosciences) for 45 min at room temperature. This was followed by a further series of washes and detection of immunolabeling using the Pierce SuperSignal West Dura extended duration substrate kit, according to the manufacturer's instructions.

Confocal Microscopy
24 h after transfection HeLa cells were fixed for 30 min in 4% paraformaldehyde, 0.1 M sodium phosphate, pH 7.4. Cells were permeabilized and labeled with Alexa 568 phalloidin to co-stain the F-actin cytoskeleton. Fluorescent images of the two channels were collected a on Zeiss LSM 510 confocal microscope using the multitracking function (laser excitation at 488 and 543 nm); exported images were collated for presentation in Adobe Photoshop and Illustrator.

Calcium Imaging
Imaging experiments were performed 36 h after transfection. HeLa cells were incubated with the indicator FuraRed-AM (2.5 M; Molecular Probes, Inc., Eugene, OR) for 20 min at 37°C in Dulbecco's modified Eagle's medium. Cells were then placed on an open air 37°C heated microscope stage for imaging of the intracellular calcium response. The extracellular solution for these experiments was imaging buffer (121 mM NaCl, 5.4 mM KCl, 0.8 mM MgCl 2 , 1.8 mM CaCl 2 , 6 mM NaHCO 3 , 5.5 mM glucose, 25 mM HEPES, pH 7.3, 37°C). Florescence images were obtained using a Zeiss 510 confocal microscope using a ϫ20 Plan-Apochromatic objective. GFP was excited by a 488-nm argon laser line, and emission was collected through a bandpass filter of 505-550 nm, with pinhole set to achieve an optical slice of 1 m. FuraRed was excited similarly, but emission was collected above 585 nm, with pinhole set at maximum (12.5 M). 512 ϫ 512 resolution scans were collected at a rate of 2 Hz before, during, and after the bath application of 100 M histamine. Parallel experiments, comparing directly different constructs and their ability to alter Ca 2ϩ signals, were performed on the same day to ensure that different levels of expression did not account for the differences seen.

Data Analysis
Using the Zeiss LSM 510 software, regions of interest were defined around individual cells to produce plots of fluorescence change over time. FuraRed fluorescence values were then normalized to prestimulatory levels (F 0 ) by calculating (F/F 0 ) ϫ 100 for each data point. These data were used to plot fluorescence profiles for individual cells, representing changes in [Ca 2ϩ ] i over time. A one-way analysis of variance test, assuming Gaussian distribution, was used to test whether the mean and variances of cells on different coverslips, but with the same transfection and experimental conditions, were significantly different. No significant difference was found, allowing us to pool the normalized fluorescence profiles in order to calculate the mean, S.E., and n values for each transfection condition. These data were used to plot the mean response for the period immediately after stimulation. A two-tailed, unpaired Student's t test, with Welch's correction, was used to test for significant differences between the peak mean responses under different transfection conditions.

RESULTS
Individual Cell Responses Are Heterogeneous-In order to validate our assay for Ca 2ϩ signals in transiently transfected cells, it was crucial to establish that any differences in the Ca 2ϩ responses observed were not due merely to differences in transfection levels among the many different constructs used. All of the constructs were GFP-tagged, so in all of our experiments where we compared directly two or more constructs, we used the same laser setting and measured GFP fluorescence to ensure that we were comparing coverslips with similar expression levels. As an example, Table 1 shows quantitation of GFP readings from a typical experiment (one of many that contributed to the data shown in Fig. 2) using four different construct transfections, where, although expression efficiency varied modestly among constructs and coverslips, the level of GFPtagged protein expressed was very similar with respect to both the mean and the variability among cells (S.D.).
To measure and control for intercellular variability in Ca 2ϩ signals, we classified the fingerprint of each cell's Ca 2ϩ response and quantified the distribution of these responses. These individual cell responses (Fig. 1) range from no response (A), to a single Ca 2ϩ pulse (B), to a short series of oscillations (C), to a prolonged series of oscillations (D), and finally to a sustained Ca 2ϩ rise (E).
Untransfected HeLa cells stimulated with low doses of histamine show just such heterogeneous responses, believed to be governed in part by intercellular variation in the concentration of Ins(1,4,5)P 3 (47,48), whereas at the high concentration of histamine that we used here, the untransfected cells consistently gave a uniform plateau response (see Fig. 2, and compare Refs. 47 and 48). All enzymatically active constructs caused a shift toward a blunted Ca 2ϩ signal and exposed intercellular heterogeneity, which we quantitated as   a response profile (displayed as histograms in the relevant figures) to produce a sensitive assay for the impact of the constructs on Ins(1,4,5)P 3 metabolism and Ca 2ϩ release. It is also noteworthy that many of the mutations used here are single amino acid substitutions, which would in any case be highly unlikely to engender systematic changes in transfection efficiency.
To control for nonspecific artifacts caused by cell transfection or any possible influence of GFP fluorescence on the Fura-Red Ca 2ϩ signals, we transfected cells with a kinase-dead construct of IP 3 KA. The cell population mean response ( Fig. 2A, graph) and the population profile ( Fig. 2A, histogram) show that transfection in general has no detectable effect on the signal. Thus, overall we believe that we can attribute the Ca 2ϩ responses observed to the ability of the various transfected enzymes to remove Ins (1,4,5)P 3 .
IP 3 Ks, IPMK, and Ins(1,4,5)P 3 5-Phosphatase All Reduce Ca 2ϩ Responses-The remainder of Fig. 2 shows pooled data from experiments comparing the three mammalian IP 3 Ks, the mammalian IPMK, mammalian Type I Ins (1,4,5)P 3 5-phosphatase, and the two Drosophila IP 3 K isoforms. We first compared the three mammalian IP 3 Ks. The summed data from the populations (graphs; left panels) suggest that isoform C is most effective, followed by A and then B. This conclusion is supported by differences in the distribution of response profiles (histograms; right panels). For example, 25% of IP 3 KA-transfected cells gave no response, and 75% gave a single pulse of Ca 2ϩ , whereas with IP 3 KC transfection, the corresponding numbers are 46 and 54%, respectively. We next compared the effect of Ins(1,4,5)P 3 metabolism via the 5-phosphatase route versus the 3-kinase route. Overexpression of Ins(1,4,5)P 3 5-phosphatase, an enzyme with a higher K m and V max for Ins(1,4,5)P 3 than IP 3 Ks (26), was as effective as IP 3 KC at blunting the Ca 2ϩ response and more efficacious than IP 3 KA and IP 3 KB (Fig. 2E). These experiments emphasize that the phosphatase is an efficient regulator of Ins(1,4,5)P 3 signals (see also Refs. 29 and 32).
The two Drosophila IP 3 Ks were both very effective removers of Ins(1,4,5)P 3 but to different degrees. The ␣ isoform was the single most efficacious enzyme tested in these experiments; 80% of transfected cells gave no response at all, and the remaining 20% showed a single transient Ca 2ϩ signal (Fig. 2G). By contrast, the dmIP 3 K␤ was only slightly more effective than mammalian IP 3 KB (Fig. 2H). Thus, perhaps surprisingly, both isoforms of IP 3 Ks in Drosophila function as well as or better than their mammalian counterparts at reducing the signaling pool of Ins(1,4,5)P 3 . The significance of these data is discussed below, where the CaM regulation of these Drosophila isoforms is addressed.
Regulation by Localization near Sites of Ins(1,4,5)P 3 Production-We have previously highlighted the significant effect that the plasma membrane localization (by virtue of its F-actin binding) of IP 3 KB has on its efficacy in HeLa cells (25). Fig. 3 depicts the various cellular localizations of some of the constructs used in this study. Previous work suggests that the mammalian isoforms A and B are localized to the filamentous actin cytoskeleton, whereas isoform C is soluble and/or nuclear. We confirm these studies and also demonstrate that neither isoform from Drosophila exhibits cytoskeletal targeting, suggesting that it is their catalytic activity that accounts for their high efficacy at removing Ins(1,4,5)P 3 (Fig. 2) rather than their targeting near sources of its generation. Fig. 4 illustrates the effects of intracellular localization on mammalian IP 3 K efficacy. As we previously reported, removal of the N-terminal 66 amino acids in isoform A makes it cytosolic (39), and this has a marked effect on its efficacy, as assessed either in a way similar to that used for IP 3 KB by Yu et al. (25) (Fig. 4A, graph) or the population profile (Fig. 4A, histograms). For example, if the IP 3 KA was cytosolic, no cells showed a zero response, whereas 25% did if transfected with the full-length (actin-binding) construct. A very similar influence on efficacy was revealed for the Ins(1,4,5)P 3 5-phosphatase if its plasma membrane targeting was compromised by tagging it with GFP at the C terminus in order to destroy the consensus site for prenylation (Fig. 4B), confirming a previous report (29). Note that in our experiments (Fig. 3), the C-terminally tagged Ins(1,4,5)P 3 5-phosphatase GFP construct localized signifi- cantly to the nucleus. Thus, overall, most mammalian enzymes that metabolize Ins(1,4,5)P 3 are targeted near sites of Ins(1,4,5)P 3 generation, giving them better spatial and temporal control over the Ca 2ϩ signal.
Effect of Regulation by Calmodulin-In vitro, the A and B isoforms are both significantly activated by Ca 2ϩ /CaM, with IP 3 KB demonstrating the greatest degree of activation (27,28,54). The effect on IP 3 KC, however, appears to be more complex and may be species-specific, since varying degrees of activation and even inhibition by Ca 2ϩ /CaM have been reported (42,55). These data are all derived from experiments in vitro, and the significance of calmodulin regulation of IP 3 Ks in the cell has not been explored. Moreover, although Seeds et al. (11) reported no CaM-binding consensus sequence in either of the dmIP 3 K genes they identified, another study has suggested that dmIP 3 K␤ but not dmIP 3 K␣ contains a putative CaM binding sequence (56). Here we have expressed both enzymes and addressed this issue experimentally.
An alignment of putative CaM-binding regions of the IP 3 Ks under consideration, plus some other homologues, revealed the conserved CaM-binding motif (Fig. 5A), highlighting in particular the essential tryptophan residue (arrow) that is highly conserved and crucial to CaM binding (46). The C. elegans enzyme is known not to bind CaM (18), and this is consistent with the absence of this tryptophan in the C. elegans sequence. An analogous tryptophan is also missing from the Drosophila ␣ isoform, whereas it is present in the ␤ sequence (Fig. 5A). We tested this issue with an in vitro CaM-binding assay with rat IP3KA as a positive control and found that dmIP 3 K␤ bound CaM in a Ca 2ϩ -dependent manner, whereas dmIP 3 K␣ did not (Fig. 5B). Thus, Drosophila possess both a CaM-independent form IP3K similar to the one found in C. elegans and a CaMsensitive form more closely resembling the mammalian versions, and we suggest that the CaM regulation of IP 3 Ks evolved some time after the divergence from the last common ancestor  between nematodes and flies. Moreover, these data emphasize that CaM regulation has not improved on the basic enzymatic efficacy of IP3Ks, since the more "primitive" wormlike CaMinsensitive, ␣ isoform from Drosophila is so efficacious in metabolizing Ins (1,4,5)P 3 (Fig. 2).
To explore the relevance of CaM-binding to the efficacy of the three mammalian IP3K isoforms in cells, we engineered site-directed mutations in the key tryptophan of the three mammalian isoforms and dmIP 3 K␤ and confirmed that the mutations did indeed ablate CaM-IP 3 K interactions (Fig. 5C), as previously demonstrated for IP3KA (46). It is interesting to note in passing that the expressed IP 3 KB in particular shows evidence of being proteolyzed, much more so than the other isoforms, consistent with it being particularly sensitive to proteolysis in vitro (44) and in vivo (25); we have shown that its subcellular localization is altered by that proteolysis (25).
We then assessed the effect of removal of CaM binding in the Ca 2ϩ mobilization assay (Fig. 6). For IP 3 KA and IP 3 KB (Fig. 6, A  and B), there was a significant decrease in efficacy, showing for the first time that CaM binding does stimulate the activity of these enzymes in vivo. Looking at the cellular profiles (histograms), it appears that the effect on the B isoform is slightly greater, as might be expected from the greater stimulation of this isoform in vitro (57), although this is not statistically significant in the pooled data. The effect of removing CaM binding from the C isoform was particularly pertinent, since it is controversial as to whether in vitro it is stimulated or inhibited by CaM (42,55). In our experimental protocol, the difference between the wild-type and W383R mutant was not statistically significant over the whole population (Fig. 6C), although both the graph and the bar diagram of Fig.  6C suggest that if there was any small effect, it is a decrease in efficacy, implying a stimulation by CaM of this isoform. Thus, in our hands, all three isoforms of mammalian IP 3 K are positively regulated by CaM in cells, albeit with a much lower effect on IP 3 KC.
The effect of preventing dmIP 3 K␤ from binding CaM was especially notable. The decrease in overall efficacy was the most pronounced of all the isoforms (Fig. 6D, graph and histogram), suggesting that CaM activation is probably more profound than in the mammalian isoforms. More interestingly, in cells transfected with the W337R mutant, nearly all of the cells oscillated; this includes cells that showed sustained oscillations after a delay (15% of cells), a pattern that was only reproducibly seen with this particular transfection. Our simple interpretation of this is that with CaM activation removed, the K m of this "deregulated" enzyme happens to be at the correct level to hold Ins(1,4,5)P 3 at a steady-state concentration ideal for Ca 2ϩ oscillation generation in HeLa cells. Thus, whereas in untransfected cells the response to this high dose of histamine is an almost universally sustained Ca 2ϩ increase, the dmIP 3 K␤ W337R mutant "clamps" the Ins(1,4,5)P 3 to the right concentration to generate Ca 2ϩ oscillations (47,48). This is reminiscent of the observations of Dupont et al. (32), who reported that raising Ins(1,4,5)P 3 5-phosphatase (by microinjection) and then increasing agonist doses to compensate, and thus generating a consistent steady-state level of Ins(1,4,5)P 3 , led to Ca 2ϩ oscillations.
These catalytic properties of dmIP 3 K␤ are an interesting contrast with the ␣ isoform, which is more efficacious than the (wild-type) ␤ at removing Ins(1,4,5)P 3 (Fig. 2), yet it is not CaMregulated. This is inconsistent with a view that CaM activation of IP 3 Ks evolved simply to increase their effectiveness at removing Ins(1,4,5)P 3 . Rather, it argues that CaM regulation enables the IP 3 Ks to participate in increasingly complex Ca 2ϩ signaling through feedback regulation by signals coming from Ins(1,4,5)P 3 receptors.

DISCUSSION
The principal finding of this study is that all major enzymes thought to metabolize the second messenger Ins(1,4,5)P 3 are capable of blunting Ca 2ϩ signals when overexpressed in cells, but their ability to do so depends on their variable catalytic activities, feedback regulation by Ca 2ϩ via CaM, and their cellular localization. In our experimental approach, we transfected cells with high levels of GFP-tagged enzymes and monitored the effect on Ins(1,4,5)P 3 Ca 2ϩ signals produced by a maximal dose of histamine. Our comprehensive and quantitative comparison of heterogeneous enzymes in a single assay system reveals significant new insights into IP 3 Ks function. For example, we have shown positive CaM regulation of the three mammalian IP 3 K isoforms for the first time in intact cells, and we have demonstrated that Drosophila possess one isoform of IP 3 K that is regulated by CaM and one that is not. In the case of mammalian isoforms, we show that targeting near sites of receptor-triggered IP 3 production increases the efficacy of IP 3 Ks and also the Type 1 Ins(1,4,5)P 3 5-phosphatase. Surprisingly, we found that dmIP 3 K␣, despite having neither CaM regulation nor localization near the plasma membrane, was the single most efficacious enzyme tested in this study. These data suggest that the more recent evolutionary elaborations of IP 3 K involve fine tuning of the localization and regulation in order to better control Ca 2ϩ signals in time and space, as would be required for the increasingly complex signaling demands of metazoan cells and tissues.
Comparison of Core Catalytic Efficacies-Among the three mammalian IP 3 K varieties, isoform C was the most effective at attenuating Ca 2ϩ signals. This isoform is the least regulated by feedback from Ca 2ϩ /CaM and lacks a consensus phosphorylation site for CaM kinase. Thus, it has been suggested that IP 3 KC is adapted for regulating the resting Ins(1,4,5)P 3 levels in cells (42,43). The reported K m values for mammalian IP 3 Ks range from about 0.2 to 10 M (28, 38), whereas the K m reported for dmIP 3 K␤ is 25 nM (11). The K m for dmIP 3 K␣ has not been reported. In terms of its primary structure, dmIP 3 K␣ resembles the IP3K from C. elegans, and it may represent a phylogenetically older version of IP3K (56). Taken together, the data support the notion that, in terms of their ability to bind substrate, IP3Ks have probably not improved much since the occurrence of the defining event in the evolution of these enzymes: the appearance of a greatly elaborated inositol phosphate (IP) lobe, which dictates the exquisite specificity of the Ins(1,4,5)P 3 3-kinase reaction (34,58). Indeed, our experiments indicate that, if the mammalian IP 3 Ks are compared with their Drosophila counterparts in cells, it appears that a modest decrease in catalytic efficiency has ensued (i.e. the ability of the Drosophila IP 3 Ks to ablate Ins(1,4,5)P 3 signals is greater, despite the fact that they lack membrane targeting and are cytosolic) (Figs. 2 and 3).
We also demonstrated that the type I Ins(1,4,5)P 3 5-phosphatase is quantitatively comparable with the IP 3 Ks in terms of its ability to remove Ins (1,4,5)P 3 . A previous study reported that overexpressed 5-phosphatase is actually more efficacious than IP 3 Ks in terms of blunting Ca 2ϩ signals stimulated by ATP (29). Expression of 5-phosphatases, in multiple isoforms, is both widespread and ancient (59). These data emphasize that IP 3 Ks arose not simply as an ATP-dependent alternative to the widely expressed 5-phosphatase Ins(1,4,5)P 3 removal pathway (22,58). Rather, IP 3 Ks arose to impart the cell greater control over its Ca 2ϩ signals.
Our study also compared IP 3 Ks with the human IPMK, which exhibits both 3-kinase and 6-kinase activity for Ins(1,4,5)P 3 among its various catalytic activities (Fig. 2).
Although human IPMK can definitely remove Ins(1,4,5)P 3 in vivo (and thereby blunt Ca 2ϩ signals), its greater promiscuity for other substrates (15,16,52,53,60) makes it less suited for regulating Ca 2ϩ signals and better suited for the synthesis of higher inositol phosphates, as suggested by genetic studies in yeast (51), Drosophila (11), and mammals (12,61). One or more IPMK-like genes can be found in most or all eukaryotic genomes, suggesting that IPMK represent a more ancient form of IPK than IP 3 Ks. Indeed, the recent crystal structure of the yeast IPMK Arg82 reveals a more rudimentary form of IPK, containing a much smaller IP lobe capable of accommodating multiple IP substrates (49).
Regulation via CaM-All known IP 3 Ks exhibit substantial activity in the absence of Ca 2ϩ , but most are also activated by Ca 2ϩ . For example, in vitro studies of IP 3 KB suggest about a 10-fold activation through the combination of direct CaM binding and phosphorylation by CaMKII (62). This contrasts to type I Ins(1,4,5)P 3 5-phosphatase, which has no known positive regulation by Ca 2ϩ and is 80% inhibited when phosphorylated by CaMKII (63). Activation by Ca 2ϩ is a typical physiological feedback mechanism for terminating intracellular signals, consistent with the role of Ins(1,4,5)P 3 as a major participant in Ca 2ϩ signaling in metazoan cells. By comparing mutant enzymes unable to bind CaM to their wild type counterparts, we have studied the Ca 2ϩ feedback regulation of IP 3 Ks for the first time in intact cells and found all three mammalian isoforms to be activated by Ca 2ϩ , although the C isoform is the least affected. It is likely that we have underestimated the true amount of this stimulation, because we are expressing the IP 3 Ks at supraphysiological levels such that endogenous CaM may be limiting. If this is a factor, then a difference in its magnitude between different cell lines may in part account for a previous study's inability to detect any significant kinetic differences in Ca 2ϩ signal attenuation between the three mammalian isoforms (24). However, since the CaM-binding mutants were also different from one another (Fig. 6), this cannot be the whole explanation for the difference in the results obtained by Dewaste et al. (24) and our data. Other factors might be the difference between cell types in such parameters as the different agonists used, the absolute magnitude of the changes in Ins(1,4,5)P 3 , the level of Ins(1,4,5)P 3 receptors (thus altering the relationship between these changes and the Ca 2ϩ signals), or the relative contribution of the 5-phosphatase pathway.
We have shown, both by CaM-binding experiments in vitro and by Ca 2ϩ imaging of mutants in cells, that one of the Drosophila isoforms (dmIP 3 K␤) is CaM-regulated (Figs. 5 and 6). Our direct comparison of the two Drosophila isoforms reveals, however, that the CaM-stimulated isoform removes Ins (1,4,5)P 3 less effectively than does the CaM-insensitive enzyme dmIP 3 K␣ (Fig. 2). These observations suggest a possible autoinhibitory mechanism for the CaM binding domains of IP 3 Ks in the absence of CaM, and they also imply that regulation by CaM did not arise simply to increase the efficiency of bulk Ins (1,4,5)P 3 removal.
Regulation via Localization-An evolutionarily more recent elaboration on IP 3 K structure, which as far as we know is confined to vertebrates (and is here exemplified by our demonstration that the Drosophila isoforms are cytoplasmic) (Fig. 3), is their specific targeting near sources of IP 3 generation via their divergent N termini. We show here for the A isoform, as we did earlier for the B (25), that such targeting enhances the ability of the enzyme to phosphorylate receptor-generated Ins(1,4,5)P 3 . However, such structural modifications cannot be viewed as absolute improvements in the enzyme's ability to lower the bulk Ins(1,4,5)P 3 signals in the cells, but rather as a means for the cell to better restrict Ins(1,4,5)P 3 within microdomains.
Implications for the Evolution of Signal Transduction Pathways-Our data provide new evidence to support the idea that IP 3 Ks evolved specifically to regulate Ins(1,4,5)P 3 signals by removal of receptor-generated Ins(1,4,5)P 3 in metazoans. Based on the fact that some protozoans have Ins (1,4,5)P 3 receptors but none have yet been shown to have IP 3 Ks, these enzymes probably arose after the emergence of Ins(1,4,5)P 3 as a second messenger (10,56). Recent studies of the relative contributions of IPMK and IP 3 K to the synthesis of the higher inositol phosphates indicate that IP 3 Ks need not have evolved for the purpose of making inositol 1,2,3,4,5,6-hexakisphosphate via the sequential phosphorylation of Ins(1,4,5)P 3 , because this is achieved via IPMK (11,12). Indeed, the genetic and biochemical evidence overwhelmingly supports the idea that that IP 3 Ks are enzymatic specialists, dedicated to removing the second messenger Ins(1,4,5)P 3 with increasing spatiotemporal control. As a consequence, of course, they also produce Ins(1,3,4,5)P 4 , whose most likely function is to regulate Ras and Ras-like small G-proteins (64,65), which are themselves major effectors of intracellular Ca 2ϩ signals (66).
In this context, any role in cell signaling attributed to Ins(1,3,4,5)P 4 must be a recent evolutionary innovation, restricted to complex metazoan tissues, and current evidence supports this view. For example IP 3 KB (and possibly Ins(1,3,4,5)P 4 ) is essential in the process by which thymocytes mature to T-cells in a p42 mitogen-activated protein kinase-dependent way (5,6). IP 3 KA, by contrast, has a physiologically significant role in the brain (67) and participates in learning and memory processes (68). Indeed, the example of the dynamic, Ca 2ϩ -regulated localization of IP 3 KA in dendritic spines of hippocampal neurons (39,40) may represent the most extreme case of specialization of an IP 3 K in cellular function.