Hormone stimulation of type III adenylyl cyclase induces Ca2+ oscillations in HEK-293 cells.

Various forms of cross-talk between the Ca2+ and cAMP signal transduction systems can occur in animal cells depending upon the types of adenylyl cyclases present. Here, we report that Ca2+ oscillations can be generated by hormone stimulation of type III adenylyl cyclase expressed in HEK-293 cells. These Ca2+ oscillations are apparently due to the unique regulatory features of type III adenylyl cyclase, which is stimulated by hormones and inhibited by elevated Ca2+in vivo. Ca2+ oscillations were generated by glucagon, isoproterenol, or forskolin stimulation of type III adenylyl cyclase and were dependent upon the activity of cAMP- and calmodulin-dependent protein kinases. Ca2+ oscillations were not solely dependent upon cAMP increases since dibutyryl cAMP or (Sp)-cAMP did not stimulate Ca2+ oscillations. We hypothesize that stimulation of type III adenylyl cyclase leads to increased cAMP, activation of inositol 1,4,5-trisphosphate receptors, and elevation of intracellular Ca2+. As free Ca2+ increases, type III adenylyl cyclase activity is attenuated by CaM kinase(s) and intracellular cAMP levels decrease. When cAMP levels drop below a threshold level, the inositol 1,4,5-trisphosphate receptor is dephosphorylated and Ca2+ is resequestered. This cycle is repeated if type III adenylyl cyclase is chronically exposed to an activator. This unique mechanism for generation of Ca2+ oscillations in cells is distinct from others documented in the literature.

Various forms of cross-talk between the Ca 2؉ and cAMP signal transduction systems can occur in animal cells depending upon the types of adenylyl cyclases present. Here, we report that Ca 2؉ oscillations can be generated by hormone stimulation of type III adenylyl cyclase expressed in HEK-293 cells. These Ca 2؉ oscillations are apparently due to the unique regulatory features of type III adenylyl cyclase, which is stimulated by hormones and inhibited by elevated Ca 2؉ in vivo. Ca 2؉ oscillations were generated by glucagon, isoproterenol, or forskolin stimulation of type III adenylyl cyclase and were dependent upon the activity of cAMP-and calmodulin-dependent protein kinases. Ca 2؉ oscillations were not solely dependent upon cAMP increases since dibutyryl cAMP or (S p )-cAMP did not stimulate Ca 2؉ oscillations. We hypothesize that stimulation of type III adenylyl cyclase leads to increased cAMP, activation of inositol 1,4,5-trisphosphate receptors, and elevation of intracellular Ca 2؉ . As free Ca 2؉ increases, type III adenylyl cyclase activity is attenuated by CaM kinase(s) and intracellular cAMP levels decrease. When cAMP levels drop below a threshold level, the inositol 1,4,5-trisphosphate receptor is dephosphorylated and Ca 2؉ is resequestered. This cycle is repeated if type III adenylyl cyclase is chronically exposed to an activator. This unique mechanism for generation of Ca 2؉ oscillations in cells is distinct from others documented in the literature.
In most mammalian tissues the Ca 2ϩ and cAMP signal transduction systems are tightly coupled, and cross-talk between these two regulatory systems may play an important role for various physiological phenomena including synaptic plasticity (Xia et al., 1991;Choi et al., 1993b). Intracellular free Ca 2ϩ (Ca 2ϩ i ) 1 can affect cAMP levels by modulation of adenylyl cyclase or phosphodiesterase activities (reviewed by Choi et al. (1993a) and Beavo and Reifsnyder (1990)). On the other hand, cAMP-dependent protein kinase (PKA) or cAMP can affect Ca 2ϩ i by regulating Ca 2ϩ ion channel activity (reviewed by Hell et al. (1994)). Because of the regulatory diversity of adenylyl cyclases, phosphodiesterases, and protein kinases, different patterns of cross-talk between the Ca 2ϩ and cAMP regulatory systems may be established in specific cell types.
III-AC is stimulated by Ca 2ϩ and calmodulin (CaM) in isolated membranes when the enzyme is also activated by G s (Choi et al., 1992). However, we recently discovered that Ca 2ϩ inhibits hormone-stimulated III-AC activity in vivo (Wayman et al., 1995). Ca 2ϩ inhibition of III-AC is not due to activation of G i or protein kinase C and is apparently mediated by one of the CaM kinases. For example, Ca 2ϩ inhibition of III-AC is blocked by KN-62, which is an inhibitor of CaM kinases. Furthermore, III-AC is inhibited by coexpression of III-AC and constitutively activated CaM kinase-II in HEK-293 cells. The CaM kinase construct used was under the control of the metallothionein promoter, which allowed the induction CaM kinase-II expression with Zn 2ϩ . Ca 2ϩ inhibition of III-AC in vivo provides a feedback mechanism for attenuation of hormone-stimulated adenylyl cyclase activity. Since activation of PKA can increase Ca 2ϩ i and hormone stimulation of III-AC is inhibited by Ca 2ϩ , one might expect Ca 2ϩ oscillations to be generated by hormone stimulation of III-AC. Here, we report glucagon and isoproterenol stimulation of Ca 2ϩ oscillations in HEK-293 cells expressing III-AC and the glucagon receptor.

EXPERIMENTAL PROCEDURES
Cell Culture-Human embryonic kidney 293 (HEK-293) cells were grown at 37°C in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum in a humidified 95% air, 5% CO 2 incubator. Cell culture materials were from Life Technologies, Inc. unless otherwise noted.
Expression of I-AC, III-AC, and the Glucagon Receptor in HEK-293 Cells-The I-AC cDNA clone was isolated from a bovine brain cDNA library as described by Xia et al. (1991). The III-AC cDNA clone in pBluescript SK- (Bakalyar and Reed, 1990) was obtained from R. R. Reed (John Hopkins University, Baltimore, MD). The coding sequence of III-AC was ligated to CDM-8 (CDM-8(III-AC)) for expression in HEK-293 cells. Neomycin-resistant HEK-293 cells stably transfected with an expression vector CDM8 that contained cDNA for I-AC (CDM8(I-AC)), III-AC (CDM8(III-AC)), or no exogenous DNA were used for this study. These clones have been characterized previously (Choi et al., 1992a(Choi et al., , 1992b(Choi et al., , 1993a(Choi et al., , 1993bWu et al., 1993) and were used for subsequent cotransfection with the rat glucagon receptor (Jelinek et al., 1993). Each of these cell lines was stably transfected with either the pZCEP expression vector encoding the rat glucagon receptor (pLJ4) or vector alone. For DNA transfections, cells were plated on 100-mm dishes at a density of 2 ϫ 10 6 cells/plate, grown overnight, and transfected with the pZCEP control vector (1 g of DNA/plate) and a hygro-mycin resistance vector (1 g of DNA/plate) by the Ca 2ϩ phosphate method (Chen and Okayama, 1987). Hygromycin-resistant cells were selected in culture medium containing hygromycin B (Sigma, 460 units/ ml) and 300 g/ml G418. Hygromycin/neomycin-resistant cells were assayed for glucagon-stimulated adenylyl cyclase activity by use of a cAMP accumulation assay. After selection, cells were maintained in medium containing 230 units/ml hygromycin B and 300 g/ml G418. Multiple hygromycin/neomycin-resistant clones of each type, expressing the rat glucagon receptor (GluR) and III-AC were isolated. Cells were grown for imaging as follows. Day 1 cells were plated on poly-Llysine-or poly-D-lysine-coated Lab-Tek four-chambered coverglass slides (60,000 cells/well) and were Ca 2ϩ imaged on day 6.
cAMP Accumulation-Changes in intracellular cAMP levels were measured by determining the ratio of [ 3 H] cAMP to total ATP, ADP, and AMP pool in [ 3 H]adenine-loaded cells as described by Wong et al. (1991). This assay system allows rapid and extremely sensitive measurements of relative changes in intracellular cAMP levels in response to various effects. Absolute ratios for cAMP accumulation generally show some variation between experiments using different sets of cells (Federman et al., 1992;Dittman et al., 1993). It is important to emphasize, however, that relative changes in cAMP were highly consistent between experiments. Confluent cells in six-well plates were initially incubated in Dulbecco's modified Eagle's medium containing [ 3 H]adenine (2.0 Ci/ml, ICN) for 16 -20 h, washed once with 150 mM NaCl, and incubated at 37°C for 30 min in incubation buffer (118 mM NaCl, 4.7 mM KCl, 2.0 mM CaCl 2 , 1.2 mM MgSO 4 , 1.2 mM KH 2 PO 4 , 0.5 mM EDTA, 10.0 mM glucose, 20.0 mM HEPES, pH 7.4) containing 1.0 mM IBMX, and various effectors as indicated. Reactions were terminated by aspiration, washing cells once with 150 mM NaCl, and the addition of 1.0 ml of ice-cold 5% trichloroacetic acid containing 1.0 M cAMP. Culture dishes were maintained at 4°C for 1-4 h, and acid-soluble nucleotides were separated by ion-exchange chromatography as described (Salomon et al., 1979). Reported data are the average of triplicate determinations.
Determination of Phosphoinositide Production-Cells were plated in six-well (35 mm diameter) plates and allowed to grow until nearly confluent. Cells were then assayed for phosphoinositide turnover in response to forskolin, isoproterenol, glucagon, and carbachol by the method of Masters et al. (1985) and Subers et al. (1988). Reported data are the average of triplicate determinations.
Calcium Imaging-A 75-watt xenon lamp and a Metaltek (NC) filter wheel and shutter were separated from the Nikon microscope to prevent vibrations from affecting the optical recordings. A G. W. Ellis fiberoptic light scrambler (Technical Video Ltd.) was used to transmit the light to the microscope. Excitation, emission, and neutral density filters were from Omega Optical. The objectives used were a Nikon Fluor 20/1.3 and a Nikon Fluor 40/0.85 NA. The images were intensified with a GenIIsys image intensifier (Dage-MTI) and acquired with a Dage-MTI CCD-72 series camera. All image acquisition was computercontrolled with the Universal Imaging Corporation's Image-1/FL program. Images were viewed on a Sony Trinitron color video monitor (PVM-1343MD) and printer (UP-5000) and a Javelin Electronics video monitor (model BWM9x) and stored via a Panasonic optical disk drive (LF-7010). Images were acquired at 2-3-s intervals to reduce photobleaching. Preliminary experiments verified the linearity of the Fura-2 response at the camera settings utilized in these experiments. In vivo calibrations were performed by incubating Fura-2-loaded HEK-293 cells with 20 M A23187 for 30 min in either (a) Ca 2ϩ -free imaging buffer containing 1 mM EGTA to determine R min or (b) imaging buffer containing 30 mM CaCl 2 to determine R max (S F/B 380 is the ratio of the signal at 380 in Ca 2ϩ -free buffer divided by the 380 signal in the presence of Ca 2ϩ ). Cells were loaded with 100 nM Fura-2-AM in Dulbecco's modified Eagle's medium at 26 C for 30 min and allowed to rest in imaging buffer at 26 C for 1 h before imaging. The Ca 2ϩ imaging buffer contained 140 mM NaCl, 10 mM HEPES, 5 mM KCl, 0.5 mM MgCl 2 , 1.5 mM CaCl 2 , 10 mM glucose in distilled H 2 O at pH 7.4. The buffers were filter sterilized through a nalge 0.2 filter. Ca 2ϩ -free buffer was identical to imaging buffer except that Ca 2ϩ was omitted and 1 mM EGTA was included. All images were corrected for background fluorescence and shading across the field of view. Conversion of the ratio of the fluorescent intensities at each excitation wavelength (340 nm and 380 nm) to intracellular free Ca 2ϩ was determined through standard equations (Grynkiewicz et al., 1985). The Fura-2 K d of 224 nM was utilized in this equation.

Glucagon Stimulates Ca 2ϩ Oscillations in HEK-293 Cells
Expressing III-AC-In these experiments we used several types of transformed HEK-293 cell lines to analyze the contribution of III-AC to Ca 2ϩ transients. Glucagon does not increase intracellular Ca 2ϩ or cAMP in the control HEK-293 cells because they do not express glucagon receptors . I-AC-expressing cells were also used as a control. I-AC is stimulated by concentrations of intracellular Ca 2ϩ that inhibit the activity of glucagon-stimulated III-AC activity, and it is not stimulated by glucagon or ␤-adrenergic agonists in vivo .
HEK-293 cells stably expressing the glucagon receptor (293-G), the glucagon receptor with I-AC (I-AC-G), or III-AC (III-AC-G) were treated with 100 nM glucagon and individual cells were Ca 2ϩ imaged using Fura-2 ( Fig. 1). In agreement with the observations of Jelinek et al. (1993), treatment of 293-G cells with glucagon caused a single spike of intracellular Ca 2ϩ (Fig.  1A). Additional exposures to glucagon resulted in no further increase in Ca 2ϩ . Cells expressing I-AC and the glucagon receptor gave a similar response; a single peak of Ca 2ϩ with no additional increase with subsequent exposures to glucagon (Fig. 1B). In contrast, Ca 2ϩ oscillations were generated when III-AC-G cells were treated with glucagon (Fig. 1C). These oscillations were dependent upon the continued presence of glucagon and were not generated by transient exposure to the hormone.
Glucagon elicited three general types of Ca 2ϩ response in these cells (Fig. 2, Table I). Only 7% of III-AC-G cells gave a single Ca 2ϩ spike ( Fig. 2A), 9% showed an intermediate response best described as spike-plateau (Fig. 2B), and 84% exhibited Ca 2ϩ oscillations (Fig. 1C). In contrast, only 4% of the control 293-G cells responded with Ca 2ϩ oscillations ( Table I).
The initial Ca 2ϩ spike in III-AC-G cells averaged 500 Ϯ 136 nM and was followed by multiple Ca 2ϩ transients (336 Ϯ 128 nM), which continued for at least 60 min. These oscillations were at an average frequency of 4.3 peaks/15 min. In 293-G cells, the single Ca 2ϩ spike averaged 332 Ϯ 75 nM (Table I). Several different stable cell lines expressing III-AC and the glucagon receptor were examined with analogous results.
Isoproterenol Stimulates Ca 2ϩ Oscillations in HEK-293 Cells Expressing III-AC-To determine if initiation of Ca 2ϩ oscillations was strictly dependent on glucagon, III-AC-G cells were treated with the ␤-adrenergic agonist isoproterenol. HEK-293 cells express endogenous ␤-adrenergic receptors that are coupled to the stimulation of III-AC in vivo . Incubation of HEK-293, I-AC-G (data not shown), or 293-G cells with 10 M isoproterenol caused a single transient Ca 2ϩ peak (Fig. 3A). This response was similar in amplitude and duration to that elicited by glucagon. Exposure of III-AC-G cells to isoproterenol caused Ca 2ϩ oscillations that strongly resembled those induced by glucagon (Fig. 3B). Therefore, the phenomenon under consideration is not specific to glucagon stimulation of III-AC.
Forskolin Stimulates Ca 2ϩ Oscillations in HEK-293 Cells Expressing III-AC-To determine if other adenylyl cyclase activators stimulate Ca 2ϩ oscillations, III-AC-G cells were treated with forskolin and Ca 2ϩ imaged. Forskolin stimulates adenylyl cyclases through direct interactions with the catalytic subunit and does not require G proteins (Seamon and Daly, 1981). Forskolin, but not its inactive analogue 1,9-dideoxyforskolin, induced Ca 2ϩ oscillations in III-AC-G cells (Fig. 4). These oscillations were of comparable amplitude and frequency to those stimulated by either glucagon or isoproterenol. Treatment of 293-G cells with forskolin caused a single Ca 2ϩ peak (Fig. 7C). These results indicate that Ca 2ϩ oscillations in III-AC-G cells can be initiated by several different activators of III-AC and are not dependent upon hormone stimulation of the enzyme.

Comparison of Intracellular cAMP Increases Stimulated by
Hormones in III-AC-G and 293-G Cells-The data described thus far suggest that elevations in cAMP may stimulate Ca 2ϩ oscillations in III-AC-G cells. The absence of hormone-stimulated Ca 2ϩ oscillations in 293-G cells, which express low levels of endogenous III-AC, might be due to insufficient cAMP increases. Therefore, 293-G, I-AC-G, and III-AC-G cells were treated with 100 nM glucagon (Fig. 5A) or 10 M isoproterenol (Fig. 5B) and intracellular cAMP accumulations were measured. Glucagon-or isoproterenol-stimulated cAMP increases were 2-3-fold greater in III-AC-G cells than in 293-G or I-AC-G cells (Fig. 5, A and B). Forskolin-stimulated increases in cAMP were also significantly greater in III-AC-G cells compared to 293-G cells (Fig. 6). These data suggest that Ca 2ϩ oscillations in HEK-293 cells may require a threshold cAMP increase that is generated in III-AC-G cells, but not in 293-G or I-AC-G cells. However, other data discussed below suggest that increases in cAMP may be necessary but not sufficient for generation of Ca 2ϩ oscillations. If Ca 2ϩ oscillations in HEK-293 cells require a threshold cAMP increase, then it might be possible to stimulate Ca 2ϩ oscillations in 293-G cells using glucagon or forskolin in combination with cAMP phosphodiesterase inhibitors, which increase cAMP signals. The cAMP increases produced in III-AC-G cells by 100 nM glucagon or 10 M isoproterenol are comparable to those produced in 293-G cells by a combination of IBMX and glucagon. Exposure of 293-G cells to either glucagon or forskolin in the absence of IBMX, a phosphodiesterase inhibitor, produced a single Ca 2ϩ peak (Fig. 7, A and C). IBMX alone had no effect on intracellular Ca 2ϩ (data not shown); however, combinations of glucagon and IBMX (Fig. 7B) or forskolin and IBMX (Fig. 7D) resulted in Ca 2ϩ oscillations. The predominant form(s) of endogenous adenylyl cyclase in HEK-293 cells is Ca 2ϩ -inhibitable. For example, isoproterenol stimulated adenyly cyclase activity in HEK-293 control cells approximately 6.0-fold, and this stimulation was inhibited 60% by increasing intracellular free Ca 2ϩ . These data are consistent with the proposal that a minimal cAMP increase is necessary for Ca 2ϩ oscillations.
cAMP Analogues Alone Do Not Produce Ca 2ϩ Oscillations in HEK-293 Cells-If Ca 2ϩ oscillations are solely dependent upon minimal cAMP increases, then it should be possible to generate oscillations with high concentrations of membrane-permeable cAMP analogues such as dibutyryl cAMP or (S p )-cAMP. Incubation of 293-G cells with either 1 mM dibutyryl cAMP (Fig. 8A) or 400 M (S p )-cAMP (Fig. 8B) resulted in a single Ca 2ϩ transient. Secondary challenges with greater concentrations of cAMP analogues (e.g. 5 mM dibutyryl cAMP) caused no further increase in Ca 2ϩ . These levels of (S p )-cAMP or dibutyryl cAMP are sufficient to fully activate PKA in 293 cells . We conclude that elevated cAMP may be necessary but not sufficient for Ca 2ϩ oscillations. Other regulatory properties of III-AC, for example its sensitivity to Ca 2ϩ inhibition, may contribute to this phenomenon.
Activation of PKA Is Required for Generation of Ca 2ϩ Oscillations-cAMP can regulate intracellular Ca 2ϩ by several mechanisms including direct interactions with Ca 2ϩ channels or indirectly by activation of PKA, which phosphorylates IP 3 (Nakade et al., 1994) and ryanodine receptors channels Hohenegger, 1993). To determine if PKA activation is required for generation of Ca 2ϩ oscillations in III-AC-G cells, the effects of two PKA inhibitors, H-89 and (R p )-cAMP (Rothermel et al., 1988), were examined. Preincubation of III-AC-G cells for 30 min with 20 M H-89 did not block the initial Ca 2ϩ rise stimulated by glucagon, but it did inhibit Ca 2ϩ oscillations (Fig. 9A). Similarly, treatment of these cells with (R p )-cAMP, prior to addition of isoproterenol, also blocked Ca 2ϩ oscillations

FIG. 2. Representative Ca 2؉ responses to glucagon in III-AC-G cells. Typical examples of Ca 2ϩ
i responses in III-AC-G cells treated with 100 nM glucagon are presented. A, 7% of the cells showed one Ca 2ϩ spike; B, 8% showed a spike plateau; C, 85% showed Ca 2ϩ oscillations. Intracellular Ca 2ϩ in individual cells was measured with Fura-2 as described under "Experimental Procedures." Representative traces from individual cells are presented.
but not the initial Ca 2ϩ response (Fig. 9B). These PKA inhibitors had no significant effect on basal or hormone-stimulated intracellular cAMP levels . Furthermore, basal Ca 2ϩ levels and the magnitude of the first Ca 2ϩ spike were also unaffected by (R p )-cAMP or H-89. The inability of H-89 or (R p )-cAMP to block the initial cAMP induced Ca 2ϩ transients may be due to incomplete inhibition PKA or to a PKA-independent mechanism for mobilization of Ca 2ϩ by cAMP. These data indicate that PKA activity is required for both glucagon-and isoproterenol-stimulated Ca 2ϩ oscillations.

The CaM Kinase Inhibitor KN-62 Blocks Ca 2ϩ
Oscillations-It is our working hypothesis that CaM kinase activity may contribute to Ca 2ϩ oscillations by inhibiting hormone or forskolin stimulation of III-AC as intracellular Ca 2ϩ increases. III-AC is inhibited by elevations in intracellular Ca 2ϩ in HEK-293 cells, and this inhibition is blocked by KN-62, a specific inhibitor of CaM kinases (Wayman et al., 1995). Consequently, Ca 2ϩ oscillations may arise from a cycle that includes hormone activation of III-AC, inhibition of III-AC by CaM kinases as Ca 2ϩ increases, and subsequent decreases in cAMP, followed by sequestration of Ca 2ϩ and reinitiation of the cycle when Ca 2ϩ drops. If this hypothesis is valid, then KN-62 should inhibit glucagon and forskolin stimulation of Ca 2ϩ oscillations.
KN-62 had no effect on either basal or carbachol-stimulated intracellular free Ca 2ϩ (data not shown). Although KN-62 did not block the initial Ca 2ϩ peak stimulated by glucagon and  forskolin, Ca 2ϩ oscillations were inhibited by KN-62 (Fig. 10). However, Ca 2ϩ i did increase approximately 2-fold over the base line and stayed at this level for at least 30 min. These data are consistent with the hypothesis that Ca 2ϩ inhibition of III-AC, by CaM kinases, may contribute to Ca 2ϩ oscillations. Because CaM kinases regulate a number of proteins involved in the regulation of intracellular Ca 2ϩ , they may also be important for the resequestration of intracellular Ca 2ϩ . For example, phospholambin (Xu et al., 1993) and a sarcoplasmic reticulum Ca 2ϩ pump (Hawkins et al., 1994) are both phosphorylated by CaM kinases. Phosphorylation of the sarcoplasmic reticulum Ca 2ϩ ATPase results in a 2-fold increase in catalytic activity. Therefore, inhibition of CaM kinase activity in HEK-293 cells may inhibit the cell's ability to return intracellular free Ca 2ϩ to basal levels.
Ca 2ϩ Oscillations Are Not Due to Hormone Stimulation of IP 3 Turnover-One of the major mechanisms for coupling of hor-mone receptors to mobilization of intracellular Ca 2ϩ is through stimulation of phospholipase C and activation of the inositol trisphosphate cascade (reviewed by Berridge (1993)). Inositol 1,4,5-trisphosphate (IP 3 ) stimulates release of Ca 2ϩ from nonmitochondrial intracellular stores and, in some systems, activation of the IP 3 pathway stimulates Ca 2ϩ oscillations (reviewed by Berridge (1990) and Fewtrell (1993)). Intracellular cAMP has been reported to regulate IP 3 production in either a positive (Horn et al., 1991) or negative fashion (Campbell et al., 1990). To address the role of this mechanism for hormonestimulated Ca 2ϩ oscillations in III-AC-G cells, we examined the effect of glucagon, isoproterenol, and forskolin on IP 3 turnover (Fig. 11). Although the muscarinic agonist carbachol increased phosphoinositide turnover 80%, glucagon, isoproterenol, and forskolin had no significant effect. These data suggest that Ca 2ϩ oscillations induced by hormone or forskolin stimulation of III-AC were not due to stimulation of the IP 3 pathway.

The IP 3 -regulated Ca 2ϩ Pool Is the Primary Source For Ca 2ϩ Oscillations-Elevations in Ca 2ϩ
i can occur by several mechanisms including the opening of plasma membrane Ca 2ϩ channels or the release of Ca 2ϩ from intracellular stores. To identify the Ca 2ϩ pool that contributes to glucagon-stimulated Ca 2ϩ oscillations in III-AC-G cells, we examined the effect of glucagon and forskolin on Ca 2ϩ oscillations in the absence of extracellular Ca 2ϩ . Glucagon stimulated Ca 2ϩ oscillation in the absence of extracellular Ca 2ϩ (Fig. 12A). The amplitude of the initial Ca 2ϩ peak was comparable in the presence and absence of extracellular Ca 2ϩ . However, the amplitude of subsequent Ca 2ϩ peaks decayed relatively rapidly suggesting that internal pools were depleted. Therefore, extracellular Ca 2ϩ is not required for the initiation of oscillations but may be required for maintenance of Ca 2ϩ oscillations over an extended period of time.
If the primary source of Ca 2ϩ for oscillations is an internal Ca 2ϩ pool, then thapsigargin should inhibit glucagon-stimulated Ca 2ϩ oscillations since this drug is an inhibitor of the intracellular sarcoenodplasmic reticulum Ca 2ϩ ATPases (Thastrup et al., 1990;Lytton et al., 1991). Treatment of III-AC-G cells with thapsigargin in Ca 2ϩ -free medium caused a rapid release and depletion of intracellular Ca 2ϩ stores, and intracellular free Ca 2ϩ returned to basal levels within 15 min (Fig.  12B). Glucagon-stimulated Ca 2ϩ oscillations were completely blocked by pretreatment with thapsigargin, suggesting that FIG. 11. Effect of adenylyl cyclase activators on phosphoinositol turnover in III-AC-G cells. III-AC-G or 293-G cells were incubated with 10 M isoproterenol, 100 M forskolin, 100 nM glucagon, or 1 mM carbachol for 30 min. Total intracellular phosphoinositol was then determined as described under "Experimental Procedures." The data are the mean Ϯ S.D. of triplicate assays.

FIG. 12. Extracellular Ca 2؉ is not required for hormone-stimulated Ca 2؉ oscillations in III-AC-G cells.
A, the effect of glucagon on intracellular free Ca 2ϩ in the absence of extracellular Ca 2ϩ (no extracellular Ca 2ϩ , 1 mM EGTA) was monitored. B, the effect of thapsigargin on glucagon-stimulated Ca 2ϩ oscillations was examined. III-AC-G cells were pretreated with 100 nM thapsigargin in the absence of extracellular Ca 2ϩ (Ca 2ϩ -free, 1 mM EGTA), followed by 100 nM glucagon as indicated. Intracellular free Ca 2ϩ was measured and determined as described under "Experimental Procedures." Representative traces from individual cells are presented. Ca 2ϩ oscillations were dependent upon intracellular Ca 2ϩ pools.
Two of the major intracellular Ca 2ϩ pools are the ryanodineand IP 3 -sensitive pools, both of which are regulated by cAMP through PKA (Bird et al. 1993;Nakade et al., 1994;Yoshida et al., 1992). High concentrations of ryanodine inhibits the release of Ca 2ϩ from the ryanodine-sensitive Ca 2ϩ pool. Treatment of III-AC-G cells with ryanodine (1-50 M) had no effect on glucagon-stimulated Ca 2ϩ oscillations in III-AC-G cells, indicating that the ryanodine-sensitive Ca 2ϩ pool does not contribute to this phenomenon (Fig. 13).
HEK-293 cells express muscarinic receptors, which are coupled to mobilization of intracellular free Ca 2ϩ through the phospholipase C/IP 3 pathway. The muscarinic agonist carbachol increases IP 3 turnover and intracellular Ca 2ϩ in these cells. Furthermore, PKA phosphorylation of IP 3 receptors stim-ulates Ca 2ϩ release from intracellular stores (Burgess et al., 1991;Bird et al., 1993;Joseph and Ryan, 1993;Nakade et al., 1994) and could account for the cAMP-generated Ca 2ϩ transients caused by forskolin, glucagon, or isoproterenol in III-AC-G cells. If the IP 3 -sensitive Ca 2ϩ pool contributes to the Ca 2ϩ oscillations stimulated by forskolin, then pretreatment of III-AC-G cells with carbachol in the absence of extracellular Ca 2ϩ should exhaust the IP 3 -sensitive pool and inhibit forskolin-stimulated Ca 2ϩ oscillations. Incubation of III-AC-G cells with carbachol in Ca 2ϩ -free media gave a single Ca 2ϩ transient, and subsequent addition of forskolin did not stimulate Ca 2ϩ oscillations (Fig. 14). Furthermore, pretreatment of these cells with forskolin for 5 min, in the absence of external Ca 2ϩ , diminished the Ca 2ϩ increase caused by subsequent application of carbachol, indicating that both reagents stimulated Ca 2ϩ release from a common pool. Collectively, these data suggest that the major Ca 2ϩ pool contributing to glucagon and forskolin-stimulated Ca 2ϩ oscillations was the IP 3 -sensitive pool.

DISCUSSION
There is increasing interest in molecular mechanisms for generation of Ca 2ϩ oscillations in non-excitable cells (reviewed by Fewtrell (1993) and Berridge (1990Berridge ( , 1992). Presumably Ca 2ϩ oscillations provide enhanced Ca 2ϩ signals averaged over an extended period of time without the toxicity associated with persistently elevated Ca 2ϩ i . One of the most extensively characterized mechanism for generation of Ca 2ϩ oscillations is through stimulation of the phospholipase C/IP 3 pathway. For example, Bird et al. (1993) have proposed that IP 3 -generated sinusoidal oscillations in intracellular Ca 2ϩ require negative feedback regulation of phospholipase C by protein kinase C. In this study we describe a new mechanism for generation of Ca 2ϩ oscillations that is based upon the unique regulatory features of III-AC, an enzyme that is stimulated by G s -coupled receptors in vivo but inhibited by elevated Ca 2ϩ i . Forskolin, glucagon, and isoproterenol stimulated Ca 2ϩ oscillations in HEK-293 cells that were stably transfected with III-AC. Control HEK-293 cells did not show Ca 2ϩ oscillations unless glucagon or forskolin were applied with IBMX. Since HEK-293 cells express III-AC activity (Xia et al., 1993) and hormone stimulation of endogenous adenylyl cyclase activity is FIG. 14. Carbachol pretreatment inhibits forskolin-stimulated Ca 2؉ oscillations in III-AC-G cells. A, III-AC-G cells were pretreated with 1 mM carbachol and no extracellular Ca 2ϩ followed by 100 M forskolin as indicated, and intracellular free Ca 2ϩ was measured. B, III-AC-G cells were pretreated with 100 M forskolin followed by 1 mM carbachol as indicated, and intracellular free Ca 2ϩ was measured. Intracellular free Ca 2ϩ was measured and determined as described under "Experimental Procedures." Representative traces from individual cells are presented.

FIG. 15. Mechanism for hormonestimulated Ca 2؉ oscillations in III-AC-G cells.
It is hypothesized that stimulation of III-AC-G by hormones or forskolin leads to activation of PKA, stimulation of IP 3 receptors, and increases in intracellular Ca 2ϩ . As intracellular Ca 2ϩ increases, III-AC activity is inhibited and cAMP levels are decreased by cAMP phosphodiesterases. When cAMP drops below a threshold level, Ca 2ϩ is resequestered and the cycle is repeated as long as activators of III-AC are present. R s , adenylyl cyclase stimulatory receptor; III-AC, type III adenylyl cyclase; CaM, calmodulin; PKC, protein kinase C; PKA, cAMP-dependent protein kinase; PLC, phospholipase C; DAG, diacylglycerol; IP 3 , inositol 1,4,5-trisphosphate; IP 3 R, IP 3 receptor/ channel; CaMK II/IV, CaM kinase type II or IV; PDE, cAMP phosphodiesterase. also Ca 2ϩ -inhibitable, it seems likely that this enzyme contributed to Ca 2ϩ oscillations in 293-G cells when endogenous cAMP phosphodiesterase activity was inhibited by IBMX. Hormonestimulated Ca 2ϩ oscillations in III-AC-G cells were dependent upon PKA activity, the IP 3 -sensitive Ca 2ϩ pool, and they were inhibited by KN-62, a CaM kinase inhibitor. High levels of cAMP analogues, that were sufficient to generation single Ca 2ϩ peaks, did not stimulate Ca 2ϩ oscillations suggesting that elevated cAMP was necessary but not sufficient to account for Ca 2ϩ oscillations. Glucagon, forskolin and isoproterenol did not generate Ca 2ϩ oscillations by stimulating IP 3 turnover.
What is the mechanism for hormone-stimulated Ca 2ϩ oscillations in III-AC-G cells? Our data are most consistent with the following model (Fig. 15). When III-AC is activated by hormones, cAMP stimulates PKA, which phosphorylates and activates IP 3 receptors. As intracellular Ca 2ϩ rises, III-AC activity is attenuated by CaM kinase(s) and intracellular cAMP levels decrease because of cAMP phosphodiesterases. When cAMP levels drops below a threshold point and the IP 3 receptor is dephosphorylated, Ca 2ϩ is resequestered and the cycle can be repeated if III-AC is chronically exposed to an activator such as forskolin or glucagon. In fact, Ca 2ϩ oscillations do not occur with a single exposure to the hormone or forskolin; the adenylyl cyclase activator has to be constantly present for the Ca 2ϩ oscillations to persist. Interestingly, it has been hypothesized that feedback inhibition of adenylyl cyclase activity by intracellular Ca 2ϩ may lead to Ca 2ϩ oscillations (Rapp and Berridge, 1977;Cooper et al., 1995). The data described in this report are the first evidence that Ca 2ϩ inhibition of adenylyl cyclase activity can lead to Ca 2ϩ oscillations in animal cells.