Use of Fluorescence Resonance Energy Transfer-based Biosensors for the Quantitative Analysis of Inositol 1,4,5-Trisphosphate Dynamics in Calcium Oscillations*

Inositol 1,4,5-trisphosphate (IP3) is an intracellular messenger that elicits a wide range of spatial and temporal Ca2+ signals, and this signaling versatility is exploited to regulate diverse cellular responses. In this study, we have developed a series of IP3 biosensors that exhibit strong pH stability and varying affinities for IP3, as well as a method for the quantitative measurement of cytosolic concentrations of IP3 ([IP3]i) in single living cells. We applied this method to elucidate IP3 dynamics during agonist-induced Ca2+ oscillations, and we demonstrated cell type-dependent differences in IP3 dynamics, a nonfluctuating rise in [IP3]i and repetitive IP3 spikes during Ca2+ oscillations in COS-7 cells and HSY-EA1 cells, respectively. The size of the IP3 spikes in HSY-EA1 cells varied from 10 to 100 nm, and the [IP3]i spike peak was preceded by a Ca2+ spike peak. These results suggest that repetitive IP3 spikes in HSY-EA1 cells are passive reflections of Ca2+ oscillations, and are unlikely to be essential for driving Ca2+ oscillations. In addition, the interspike periods of Ca2+ oscillations that occurred during the slow rise in [IP3]i were not shortened by the rise in [IP3]i, indicating that IP3-dependent and -independent mechanisms may regulate the frequency of Ca2+ oscillations. The novel method described herein as well as the quantitative information obtained by using this method should provide a valuable and sound basis for future studies on the spatial and temporal regulations of IP3 and Ca2+.

Inositol 1,4,5-trisphosphate (IP 3 ) 2 is an important intracellular messenger produced by phospholipase C (PLC)-dependent hydrolysis of phosphatidylinositol 4,5-bisphosphate (PIP 2 ). IP 3 releases Ca 2ϩ from intracellular stores via IP 3 receptors (IP 3 Rs), and the resulting Ca 2ϩ signals often exhibit complex spatial and temporal organizations, such as Ca 2ϩ oscillations (1). The mechanism responsible for Ca 2ϩ oscillations has been a longstanding question, and a number of experimental approaches and mathematical models have been reported to account for these Ca 2ϩ signals, yet the mechanism responsible remains controversial (2). There are two general classes of Ca 2ϩ oscillation models (3). In one class, Ca 2ϩ oscillations are generated in the presence of constant cytosolic IP 3 concentrations ([IP 3 ] i ) (4); in the other class, oscillating [IP 3 ] i are required to drive Ca 2ϩ oscillations (5). The physiological relevance of the former class has been supported experimentally by using nonmetabolizable IP 3 analogs (6), and by the observation of repetitive Ca 2ϩ release in permeabilized cells with clamped IP 3 concentrations (7,8). On the other hand, oscillatory changes in [IP 3 ] i have been suggested by the observed cyclical translocation of a GFPtagged pleckstrin homology domain of PLC-␦ (GFP-PHD) (9,10). However, in other experiments using more specific IP 3 biosensors, IP 3 was shown to accumulate gradually with little or no fluctuation during Ca 2ϩ oscillations (11). These discrepant observations may be attributable to differences between various IP 3 biosensors and a lack of quantitation.
There are two types of IP 3 biosensors, GFP-PHD and IP 3 Rbased FRET sensors. GFP-PHD binds to both PIP 2 and IP 3 ; thus it has been thought that changes in [IP 3 ] i could be monitored indirectly by the release of membrane-bound GFP-PHD (9). IP 3 R-based FRET biosensors consist of the ligand-binding domain of IP 3 R and a pair of fluorescent proteins, cyan fluorescent protein and yellow fluorescent protein. Since the successful monitoring of IP 3 with LIBRA (12), the first IP 3 R-based FRET biosensor, several different groups have used similar biosensors for IP 3 monitoring (11,13,14). In principle, quantitative measurements of [IP 3 ] i are not possible with GFP-PHD. In addition, it is recognized that GFP-PHD may be released from the plasma membrane by decreases in available PIP 2 (15), which could be attributed to PIP 2 hydrolysis or the occupation by other molecules. IP 3 R-based FRET biosensors offer significant benefits for monitoring IP 3 based on their high selectivity for IP 3 and ratiometric measurement.
In this study, we developed a series of improved IP 3 biosensors that exhibit high pH stability and varying IP 3 affinities. They also possess higher selectively and afford a larger dynamic range than that of original LIBRA. In combination with these * This work was supported by Grant-in-aid for Scientific Research 16390532 (to A. T.), by HAITEKU (2007) from the Ministry of Education, Culture, Sports, Science and Technology of Japan, and the Japan Science and Technology Agency. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental Figs. S1-S4 and Tables S1 and S2. 1  Plasmid Construction-Venus, EYFP with F46L, F64L, M153T, V163A, and S175G, was constructed using site-directed mutagenesis of pEYFP-N1 (Clontech) and a QuikChange II XL site-directed mutagenesis kit (Stratagene, La Jolla, CA), according to the manufacturer's instructions. Primers 1-6 were used for mutagenesis (all primers used in this study are listed in supplemental Table S1). The sequence of the Venus construct (pVenus-N1) was verified by DNA sequencing.
To construct pH-stable LIBRA (LIBRAvIII), the portion of LIBRA containing the membrane-targeting signal, enhanced cyan fluorescent protein, the IP 3 -binding domain of the rat type 3 IP 3 R (amino acids 1-604), and NheI and EcoRI sites at either end was amplified by PCR using primers 7 and 8 and LIBRA plasmid as the template. This PCR product was then ligated into the NheI and EcoRI sites of pVenus-N1.
To construct LIBRA variants with type 1 IP 3 R (LIBRAvI), the IP 3 -binding domain of the rat type 1 IP 3 R (amino acids 1-604) with XhoI sites incorporated at both ends was amplified by PCR using primers 9 and 10 and a rat brain cDNA library as the template. This PCR product was then ligated into the linker region XhoI site in the LIBRAvIII plasmid. In a similar way, we constructed LIBRAvII with the IP 3 -binding domain of the rat type 2 IP 3 R (amino acids 1-604) using primers 11 and 12 and a rat parotid cDNA library as the template. The high affinity variant of LIBRAvIII (LIBRAvIII with mutation R440Q; LIBRA-vIIIS) was prepared by site-directed mutagenesis of the ligandbinding domain of LIBRAvIII using primers 13 and 14. In a similar way, the high affinity variant of LIBRAvII (LIBRAvIIS) was prepared by site-directed mutagenesis of LIBRAvII using primers 15 and 16. Some PCR errors were identified within the IP 3 -binding domain of LIBRAvI, LIBRAvII, and LIBRAvIIS. Nucleotide sequences of these constructs were corrected by site-directed mutagenesis, and the corrected sequences were confirmed by DNA sequencing.
We previously constructed an IP 3 -insensitive mutant (K507A) of LIBRA, which we referred to as LIBRA-N. The IP 3 -insensitive variant of LIBRAv (LIBRAvN) was constructed by cutting the mutated IP 3 -binding domain of LIBRA-N-plasmid with XhoI and ligating the fragment into LIBRAvIII-plasmid cut with the same enzyme.
Cell Culture and Transfection-COS-7 cells, obtained from RIKEN Cell Bank (Tokyo, Japan), were cultured in Dulbecco's modified Eagle's medium with low glucose (1000 mg/liter), supplemented with 10% fetal calf serum, 584 mg/liter L-glutamine, 110 mg/liter sodium pyruvate, 100 units/ml penicillin, and 100 g/ml streptomycin (all from Invitrogen). HSY-EA1 cell was subcloned from HSY human parotid cell line, a generous gift from Dr. Mitsunobu Sato (Tokushima University, Japan), by a dilution plating technique. HSY-EA1 cells were cultured in Dulbecco's Eagle's medium nutrient mixture Ham's F-12 (Sigma) supplemented with 10% newborn calf serum, 2 mM glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin, as described previously. These cells were grown in fibronectin-coated experimental chambers consisting of plastic cylinders (7 mm in diameter) glued to round glass coverslips.
Measurement of Fluorescence-Cells were washed with HBSS-H and rested for at least 5 min prior to experiments. In some experiments, cells were incubated at room temperature for 1-2 min in HBSS-H containing 2.5 M fura-2 acetoxymethyl ester (Dojin Chemicals, Kumamoto, Japan). Permeabilization was performed by applying ICM containing 100 g/ml (w/v) saponin (ICN, Cleveland, OH) for 1-1.5 min. In some experiments, cells were permeabilized with ICM containing 200 M ␤-escin (Sigma). Fluorescence images were captured using a dual wavelength ratio imaging system (Hamamatsu photonics, Shizuoka, Japan) consisting of a C9100-13 EM-CCD camera and W-View optics coupled to a Nikon TE2000 inverted fluorescence microscope equipped with a Nikon S Fluor 60 oil immersion objective (NA 1.3). Fluorescence of IP 3 biosensors was monitored with excitation at 425 nm and dual emission at 480 and 535 nm. Fluorescence of fura-2 was monitored with sequential excitation at 345 and 380 nm and emission at 535 nm. AQUACOSMOS 2.6 software (Hamamatsu Photonics) and ImageJ program were used for image analyses. To minimize the effect of photobleaching on the estimation of [IP 3 ] i , time-dependent decreases in the fluorescence intensities at two emission wavelengths (480 and 535 nm) were monitored for 5-10 min prior to the stimulations, and the photobleaching rates were calculated by fitting the data to linear or exponential equations.
Determinations of K d and Hill Coefficient-Data were analyzed by simulated annealing procedure with GOSA-fit software (BIO-LOG, Toulouse, France) to fit the data to Equation 1, where C is the IP 3 concentration; E max is the maximal effect (100%), and n is the Hill coefficient. 3 ] i , emission ratios of LIBRAvIIS in intact cells before (R rest ) and after (R min ) treatment with U73122, and emission ratios in the absence (R min Ј) and presence (R max Ј) of saturating concentrations of IP 3 were monitored after permeabilization. [IP 3 ] i can be calculated by Equation 2,   MARCH 27, 2009 • VOLUME 284 • NUMBER 13

JOURNAL OF BIOLOGICAL CHEMISTRY 8911
Determinations of Response Rate-LIBRAvIIS-expressing COS-7 cells were permeabilized with saponin, and changes in emission ratio were monitored with the first acquisition protocol (20 -50 ms/frame). The response rate for the IP 3 -dependent increase in ratio was examined by application of 10 M IP 3 to the experimental chamber containing saponin-treated LIBRA-vIIS-expressing cells. To examine the rate of decrease in ratio due to the removal of IP 3 , permeabilized LIBRAvIIS-expressing cells were initially exposed to 1 M IP 3 , and subsequently washed out using IP 3 -free ICM. In this experimental condition, ϳ70% of medium was replaced by a single wash. Using a nonlinear least squares method, these data were fitted by the equation derived from a solution of the chemical reaction model, Equation 4, where R 0 is the emission ratio at time 0 (the time when the medium in the chamber was replaced); t is the time after the replacement of the medium; R t is the emission ratio at time t; and A, C, and ␥ are arbitrary constants. This equation yields the time for the half-maximal increase or decrease in ratio as t1 ⁄ 2 ϭ ␥ Ϫ1 ln(2 Ϫ C Ϫ1 ). Note that this model is based on the assumption that medium (ICM with or without IP 3 ) within the permeabilized cell is replaced completely at t ϭ 0, whereas the actual time for the replacement is later than at t ϭ 0. It is therefore expected that actual t1 ⁄2 values are at least less than those calculated with this method. See details in the supplemental material.

RESULTS AND DISCUSSION
Characterization of Improved IP 3 Biosensors- Fig. 1A shows schematic diagrams of IP 3 biosensor domain structures. To improve the pH stability of the LIBRA emission ratio (480/535 nm), EYFP was replaced with Venus, a pH-stable yellow fluorescent protein mutant (16). The resulting construct (LIBRAvIII) showed strong pH stability in the range of pH 6 -8 (Fig.  1B). Next the LIBRAvIII ligandbinding domain was replaced with that of type 1 or type 2 IP 3 R, to create high affinity IP 3 biosensors. These constructs were designated LIBRAvI and LIBRAvII, respectively. It is known that changing Arg to Gln at position 441 (R441Q) increases the affinity of type 1 IP 3 R (17). To further increase the affinity, comparable amino acid substitutions in type 2 and type 3 IP 3 Rs (R440Q) (18) were made in LIBRAvII and LIBRAvIII. We also generated IP 3 -insensitive LIBRAv (LIBRAvN) by substituting Ala for Lys at position 508 (K508A) in the IP 3 -binding domain of LIBRAvIII. These biosensors were distributed on the plasma membrane and in vesicular structures (Fig. 1C), and most of the associated fluorescence (Ͼ90%) was retained after permeabilization.
To quickly characterize these new candidate biosensors, we examined changes in emission ratios in permeabilized cells in response to stepwise increases in IP 3 concentration (supplemental Fig. S1). The results of this preliminary characterization indicated that the apparent affinities were LIBRA Ϸ LIBRA-vIII Ͻ LIBRAvIIIS Ϸ LIBRAvI Ϸ LIBRAvII Ͻ LIBRAvIIS, and the extent of maximal changes in the emission ratios of LIBRA-vIII, LIBRAvIIIS, LIBRAvII, and LIBRAvIIS were ϳ2-fold larger than that of LIBRA. We also noticed that LIBRAvI worked well in HSY-EA1 cells (supplemental Fig. S1G) but not in COS-7 cells (supplemental Fig. S1D).
Concentration-response curves of the new IP 3 biosensors for IP 3 , inositol 4,5-bisphosphate (IP 2 ), and inositol 1,3,4,5-tetrakisphosphate (IP 4 ) are shown in Fig. 1D. The apparent K d values of LIBRAvIII, LIBRAvI, LIBRAvII, and LIBRAvIIS for IP 3 were 491.5, 268.7, 233.6, and 117.2 nM, respectively. A high concentration (10 M) of IP 2 or IP 4 induced 20 -40% of the maximal response of these biosensors, and the estimated K d values of LIBRAvIII, LIBRAvII, and LIBRAvIIS for IP 2 and IP 4 were estimated (supplemental Table S2). LIBRAvII and LIBRA-vIIS exhibited greater selectivity than LIBRA, and the selectivity of LIBRAvIIS for IP 3 was 300-fold higher than for IP 2 and 96-fold higher than for IP 4 .
The characteristics of our IP 3 biosensors include the use of Venus, ligand-binding domains of different IP 3 R isoforms, and a membrane-targeting signal. The superior selectivity of IP 3 Rbased FRET sensors provides a clear advantage over GFP-PHD sensors. In addition, utilizing Venus for a FRET-based biosensor is particularly important to avoid pH-related artifacts (16,19).
The ligand-binding domain of IP 3 Rs is composed of two functional domains, the amino-terminal suppressor domain and the carboxylterminal IP 3 -binding core domain (17). Unlike other IP 3 R-based FRET sensors (11,13), LIBRA and LIBRAv variants contain both IP 3 suppressor and core domains. Because the molecular properties of IP 3 R isoforms have been studied extensively, it is relatively easy to construct biosensors with appropriate levels of affinity. In addition, it has been reported that structural differences in the suppressor domains contribute to functional diversity in ligand sensitivity among IP 3 R isoforms (20), assuming that the ligand-binding properties of the LIBRAv series reflect this feature of IP 3 R isoforms. We previously reported a method using LIBRA for determining IP 3 R ligands (21), and we identified novel ligands of IP 3 Rs from newly synthesized cyclopentane derivatives (22). This method could be extended easily for identifying subtype-specific ligands of IP 3 Rs by using a series of new LIBRAv variants.
Imaging of IP 3 -dependent Response of LIBRAvIIS-Emission ratios within an individual cell, whether intact or permeabilized, were uneven, whereas the magnitudes of the maximal change in ratio in response to 10 M IP 3 were reasonably consistent (supplemental Fig. S2). Therefore, we visualized IP 3 -dependent changes in the fluorescence emission ratio with normalized ratio images ( Fig. 2A), where each ratio image was divided by an image before the application of IP 3 . The normalized LIBRAvIIS emission ratio increased homogeneously except for the area including intracellular vesicles in cell 2 (Fig. 2B). It was noticed that abundant accumulation of IP 3 sensors on intracellular vesicles tended to decrease the extent of IP 3 -dependent changes in the emission ratio. We then applied this procedure to visualize the ATP-induced increase in LIBRAvIIS emission ratio and subsequent decrease because of the removal of ATP in intact COS-7 cells (Fig. 2, C  and D).
Response rates of IP 3 biosensors were examined by high speed monitoring of IP 3 -dependent changes in the emission ratio of LIBRAvIIS in permeabilized cells and a least squares curve-fitting technique. The emission ratio increased to the half-maximal level in ϳ100 ms following the application of 10 M IP 3 . The estimated time for the half-maximal increase (t1 ⁄ 2 on) and decrease (t1 ⁄ 2 off) in the emission ratio was 162 and 169 ms, respectively (supplemental Fig. S3). These analyses indicate that the response rate of LIBRAvIIS is sufficiently rapid to reflect responses that occur ϳ100 ms or longer. However, the spatial resolution and signal to noise ratio necessary to detect subcellular difference in the high speed monitoring of IP 3 were not achieved in our experimental system.
Inclusion of a membrane-targeting signal provided a rapid means for examining biosensors expressed in permeabilized cells (supplemental Fig. S1) and enabled calibration of FRET signals in single cells (Fig. 3). However, nonuniform distributions of biosensors caused a subcellular variability in the ratio. This variability was overcome by using normalized emission ratio images, although this method is not suitable for measurements in motile cells. Further improvement of IP 3 biosensor dynamic ranges is required for the high speed monitoring of subcellular IP 3 responses.

Methods for Quantitative Measurement of [IP 3 ] i -Emission ratios of LIBRAvIII and LIBRAvIIS in intact cells increased
upon the application of 10 M ATP, and returned to basal levels after the addition of 5 M U73122 (Fig. 3). These reagents did not change the emission ratio of LIBRAvN (data not shown). To obtain the maximal changes in emission ratio (⌬R max ) in each individual cell, cells were permeabilized and exposed to IP 3 following the completion of measurements in intact cells (Fig. 3, A  and B). Biosensor-expressing cells and LIBRAvN-expressing cells showed decreases in their fluorescence ratios upon permeabilization. These changes in fluorescence are thought to be due to effects on the fluorescent proteins rather than to effects on the IP 3 -binding domain, and thus they are unlikely to interfere with IP 3 -dependent changes in emission ratios.
In Fig. 3, C and D, changes in emission ratios are quantitatively shown as % of maximal change in ratio (% max). The difference in % max values of ATP-induced changes in emission ratios between LIBRAvIII and LIBRAvIIS is thought to reflect the difference in affinities of these IP 3 sensors. We further examined resting [IP 3 ] i by treating unstimulated cells with U73122. The LIBRAvIII emission ratio did not change with U73122 treatment (Fig.  3E), whereas the LIBRAvIIS ratio decreased slowly to a new steady state (Fig. 3F). These results indicate that the emission ratio of LIBRAvIIS reflects the resting [IP 3 ] i , and U73122 decreases [IP 3 ] i to levels below the detectable range of LIBRAvIIS. Consistent with this interpretation, Sato et al. (13) reported that U73122-induced decreases in the emission ratio of intact cells were observed using a high affinity IP 3 biosensor. Based on this idea, we used the U73122-induced decrease in emission ratios, K d , and Hill coefficient of LIBRA-vIIS (shown in Fig. 1D) (Fig. 4D), whereas the effects of adjusting calculations for resting [IP 3 ] i was limited for LIBRAvIII (Fig. 4C). In addition, the impact of basal noise of the LIBRAvIIS fluorescence on the basal level of calculated [IP 3 ] i was negligible (Fig. 4D, arrowhead) (Fig. 4C, arrowheads). These results indicate that LIBRAvIII is convenient for monitoring broad ranges of [IP 3 ] i .
In this study, [IP 3 ] i was estimated under the following assumptions: 1) that the properties of IP 3 biosensors in permeabilized cells are comparable with those in intact cells, and 2) that U73122 reduces [IP 3 ] i below the detectable level of the biosensors. To examine possible effects of cytosolic proteins on biosensor response, we compared IP 3 -induced changes in the ratios of LIBRAvIIS in ␤-escin-permeabilized cells and in saponin-permeabilized cells and found comparable IP 3 -induced responses (supplemental Fig. S4). ␤-Escin-permeabilized cells have been used to examine the functions of cytosolic proteins, 3 3 A. Tanimura, unpublished observations. whereas cytosolic proteins are lost in saponin-permeabilized cells (7,23). Thus, the effects of endogenous cytosolic proteins on the IP 3 -induced response of biosensors could be ruled out. In addition, our previous study indicates that IP 3 -dependent changes in the ratio of LIBRA are not altered by Ca 2ϩ or ATP, and that the effect of pH on fluorescent proteins does not alter the IP 3 -dependent changes in the fluorescence of LIBRA (12). Although these experiments do not exclude possible effects of other small endogenous molecules, it is reasonable to assume at this stage that IP 3 -dependent changes in the ratios of these types of biosensors in permeabilized cells are comparable with those in intact cells.
We also examined the effects of U73122 on ATP-induced Ca 2ϩ responses, and we found that the pretreatment of COS-7 cells with 5 M U73122 completely blocked the rise in [Ca 2ϩ ] i obtained by 3 M ATP, and strongly decreased responses obtained with 10 M ATP (data not shown). It is therefore thought that U73122 pretreatment is sufficient to block the low level of PLC activity in unstimulated cells. Indeed, we have used this method to estimate the [IP 3 ] i required to elicit Ca 2ϩ responses (see below), and these values are reasonably close to the threshold concentration of photoreleased [IP 3 ] i (60 nM) that triggers Ca 2ϩ spikes in Xenopus oocytes (24).

Changes in [IP 3 ] i during Agonist-induced Ca 2ϩ
Oscillations-The mechanism responsible for Ca 2ϩ oscillations and the associated dynamics of IP 3 have been of long-standing interest (2). Experiments based on the translocation of GFP-PHD have suggested that [IP 3 ] i oscillate (9, 10), whereas an IP 3 R-based FRET biosensor has shown that IP 3 accumulates gradually in the cytosol with little or no fluctuation during Ca 2ϩ oscillations (11). Because of conflicting data and a lack of quantitation of IP 3 dynamics, it remains controversial whether [IP 3 ] i truly fluctuates and whether proposed [IP 3 ] i fluctuations drive Ca 2ϩ oscillations. To clarify these important questions, we quantitatively monitored IP 3 dynamics during Ca 2ϩ oscillations using LIBRAvIIS.
The upper panels of Fig. 5, A-D, show changes in emission ratios of LIBRAvIIS and fura-2, and the lower panels indicate calculated [IP 3 ] i during ATP-induced Ca 2ϩ oscillations. In COS-7 cells, stimulation with either 1 or 3 M ATP increased [IP 3 ] i slowly to a sustained level without detectable fluctuations, and Ca 2ϩ oscillations were observed during this increase and sustained elevation of [IP 3 ] i (Fig. 5, A and B). Unlike ATP-induced increases in [Ca 2ϩ ] i , the large increase in [Ca 2ϩ ] i caused by ionomycin (2 M) treatment had no effect on the emission ratio of LIBRAvIIS, indicating that Ca 2ϩ itself does not activate IP 3 generation in COS-7 cells.
We also examined IP 3 dynamics in HSY-EA1 cells. This cell line is characterized by long lasting Ca 2ϩ oscillations in response to wide ranges of agonist concentrations (23,25). In fact, 65% of HSY-EA1 cells (55 of 84 cells) exhibited Ca 2ϩ oscillations following treatment with 3-100 M ATP or 10 -100 M carbachol. Interestingly, 56% of Ca 2ϩ -oscillating HSY-EA1 cells (31 cells) showed associated fluctuations in LIBRAvIIS emission ratios (Fig. 5, C and D). Similar fluctuations in the LIBRAvIIS emission ratio were observed when IP 3 dynamics were examined in the absence of fura-2 loading (data not shown). In contrast, no change in emission ratio was observed during Ca 2ϩ oscillations in LIBRAvN-expressing HSY-EA1 cells (Fig. 5E). These experiments exclude the possibility of resulting artifacts derived from IP 3 -independent changes in LIBRAvIIS fluorescence and possible interference by fura-2 fluorescence. These results demonstrate cell type-specific differences in IP 3 dynamics, nonfluctuating rises in [IP 3 ] i and repetitive IP 3 spikes in COS-7 cells and HSY-EA1 cells, respectively. Quantitative examinations revealed that repetitive IP 3 spikes in HSY-EA1 cells occurred concomitantly with a slow basal accumulation of [IP 3 ] i . The size of IP 3 spikes varied from 10 to 100 nM, and the second and later spikes were initiated before the decline of [IP 3 ] i to resting levels, resulting in a slow increase in the [IP 3 ] i interspike.
It is generally thought that IP 3 diffuses rapidly within the cell, and thus we think that the fluorescence of membrane-bound biosensors reflects overall cytosolic [IP 3 ] i . In agreement with this idea, we observed nonfluctuating IP 3 responses in COS-7 cells and repetitive IP 3 spikes in HSY-EA1 cells in both cytosolic and nuclear areas using a biosensor lacking the membrane targeting sequence. 3 Although repetitive IP 3 spikes were observed in HSY-EA1 cells, our observations do not support the requirement of IP 3 fluctuations in Ca 2ϩ oscillations. Although [IP 3 ] i showed clear fluctuations at the beginning of Ca 2ϩ oscillations, repetitive spikes of [IP 3 ] i were gradually obscured during Ca 2ϩ oscillations in 64% of HSY-EA1 cells. In addition, the [IP 3 ] i spike peak was preceded by a Ca 2ϩ spike peak (Fig. 5, C and D). These results suggest that repetitive IP 3 spikes in HSY-EA1 cells are passive reflections of the Ca 2ϩ oscillations, and are unlikely to be essential for driving Ca 2ϩ oscillations.
[IP 3 ] i fluctuations could be induced by the effects of Ca 2ϩ on IP 3 synthesis and/or IP 3 degradation. Applications of 2 M ionomycin had little or no effect on the emission ratio of LIBRAvIIS in HSY-EA1 cells (Fig. 5C), suggesting that the direct effect of Ca 2ϩ on IP 3 production is very small or absent in this cell line. Thus, [IP 3 ] i fluctuations are likely to be caused by Ca 2ϩ -induced potentiation of agonist-dependent IP 3 generation. The IP 3 spikes described here resemble the pattern pre-dicted by the oscillator model, in which positive feedback via Ca 2ϩ -dependent activation of PLC is added to the dual positive and negative feedback effect of Ca 2ϩ on IP 3 R, rather than the model including negative feedback via Ca 2ϩ -dependent IP 3 degradations by IP 3 3-kinases (26).
Ca 2ϩ oscillations were observed primarily when [IP 3 ] i was less than 300 nM (Fig. 6, A and B). More than 50% of COS-7 cells exhibited Ca 2ϩ oscillations when [IP 3 ] i was 50 -100 nM, and the percentage of oscillating COS-7 cells decreased as [IP 3 ] i increased (Fig. 6C). In HSY-EA1 cells, Ca 2ϩ oscillations were observed in 50 -70% of cells, when interspike [IP 3 ] i was less than 250 nM, and the percentage of oscillating cells decreased abruptly (11.1%) when interspike [IP 3 ] i was greater than 250 nM (Fig. 6D). These results indicate that low concentrations of IP 3 (Ͻ100 nM) induce Ca 2ϩ oscillations in both cell types, whereas HSY-EA1 cells are more likely to exhibit Ca 2ϩ oscillations at higher concentrations of [IP 3 ] i (100 -250 nM) than are COS-7 cells. Large increases in [IP 3 ] i (Ͼ250 nM) induced the peak plateau-type Ca 2ϩ response in both cell types.
This study showed the time delay of IP 3 spike peaks from Ca 2ϩ spike peaks in HSY-EA1 cells, and Ca 2ϩ oscillations with nonfluctuations of [IP 3 ] i in COS-7 cells. Together these results suggest it is likely that IP 3 spikes are not essential to drive Ca 2ϩ oscillations. Regarding the mechanism of Ca 2ϩ oscillations, the importance of dual feedback effects of Ca 2ϩ on IP 3 Rs has been demonstrated experimentally (6 -8). However, it has been pointed out that this dual feedback effect explains relatively short period Ca 2ϩ oscillations, but it cannot reproduce long interspike intervals. Thus, additional mechanisms responsible for establishing these oscillations remain unclear (3,26,27). One model study reported that frequency properties of oscillation are modulated by the incorporation of Ca 2ϩ activation of PLC into the Ca 2ϩ oscillation model based on dual feedback regulations of IP 3 R properties, which enhances the range of frequency encodings of agonist stimulations (26). Interestingly, HSY-EA1 cells exhibited Ca 2ϩ oscillations for a wider range of agonist and [IP 3 ] i concentrations than observed in COS-7 cells. Thus, it may be possible that repetitive IP 3 spikes or fluctuations play some role in supporting and/or regulating Ca 2ϩ oscillations. The quantitative information provided here would be useful for future studies of the mechanisms and roles of IP 3 oscillations.

Effect of [IP 3 ] i on the Refractory Period of Ca 2ϩ Oscillations-
We analyzed the effects of [IP 3 ] i on the interspike period of Ca 2ϩ oscillations. In COS-7 cells, increases in agonist concentrations resulted in either shortening of interspike periods of Ca 2ϩ oscillations (Fig. 5A) or shifting of oscillations to the peak plateau-type Ca 2ϩ response (Fig. 5B) in association with increases in [IP 3 ] i . Similarly, in HSY-EA1 cells, the interspike period of Ca 2ϩ oscillations appeared to be correlated inversely with the increase in interspike levels of [IP 3 ] i rather than the spike peak level (Fig. 5, C and D). These observations agree with observations that the refractory period of Ca 2ϩ oscillations is controlled by [IP 3 ] i (7,8).
However, interspike periods of Ca 2ϩ oscillations occurring during a slow rise in [IP 3 ] i were not shortened by the rise in [IP 3 ] i (Fig. 5, A and C). Small increases or no change in interspike periods during an increase in [IP 3 ] i was observed in 18 of 55 HSY-EA1 cells and in 12 of 25 COS-7 cells. These data suggest that IP 3 is an important, but not exclusive, regulator of Ca 2ϩ oscillation frequency.
Although the mechanism underlying the dissociation of the interspike period from the rise in [IP 3 ] i is unknown, it might be associated with decreases in IP 3 R sensitivity. Consistent with this idea, Matsu-ura et al. (11) described progressive decreases in apparent IP 3 sensitivity for generation of Ca 2ϩ spikes during Ca 2ϩ oscillations. It is known that Ca 2ϩ within the lumen of the endoplasmic reticulum regulates IP 3 R sensitivity. Thus, a slow rise in [IP 3 ] i may be balanced by a decrease in IP 3 R sensitivity via a slow decrease in stored Ca 2ϩ (28). In addition, the involvement of stored Ca 2ϩ and sarco/endoplasmic reticulum Ca 2ϩ -ATPase activity in determining the frequency of Ca 2ϩ oscillations has been proposed recently (27). Additional work is required to explore the possible involvement of stored Ca 2ϩ and other cytosolic factors in controlling the frequency of Ca 2ϩ oscillations.
In summary, we developed improved IP 3 biosensors and a relatively simple method for the quantitative measurement of [IP 3 ] i , and we applied this method for monitoring IP 3 dynamics during Ca 2ϩ oscillations. This method revealed cell type-specific differences in IP 3 dynamics as follows: nonfluctuating rises in [IP 3 ] i and repetitive IP 3 spikes. Our results provide the first quantitative information for repetitive IP 3 spikes and present new aspects concerning the regulation of Ca 2ϩ oscillation frequency. In addition, the method demonstrated here offers a powerful means for studying the IP 3 dynamics of many cellular processes.