The spatial and temporal complexity of Ca2+ signalling is central to the regulation of a diverse range of cellular processes. The decoding of dynamic Ca2+ signals is, in part, mediated by the ability of Ca2+ to regulate other second messengers, including cyclic AMP (cAMP). A number of kinetic models (including our own) predict that interdependent Ca2+ and cAMP oscillations can be generated. A previous study in Xenopus neurons illustrated prolonged, low-frequency cAMP oscillations during bursts of Ca2+ transients. However, the detection of more dynamic Ca2+ driven changes in cAMP has, until recently, been limited by the availability of suitable cAMP probes with high temporal resolution. We have used a newly developed FRET-based cAMP indicator comprised of the cAMP binding domain of Epac-1 to examine interplay between Ca2+ and cAMP dynamics. This probe was recently used in excitable cells to reveal an inverse relationship between cAMP and Ca2+ oscillations as a consequence of Ca2+-dependent activation of phosphodiesterase 1 (PDE1). Here, we have used human embryonic kidney (HEK293) cells expressing the type 8 adenylyl cyclase (AC8) to examine whether dynamic Ca2+ changes can mediate phasic cAMP oscillations as a consequence of Ca2+-stimulated AC activity. During artificial or agonist-induced Ca2+ oscillations we detected fast, periodic changes in cAMP that depended upon Ca2+ stimulation of AC8 with subsequent PKA-mediated phosphodiesterase 4 (PDE4) activity. Carbachol (10 μM) evoked cAMP transients with a peak frequency of ∼3 minute-1, demonstrating phasic oscillations in cAMP and Ca2+ in response to physiological stimuli. Furthermore, by imposing a range of Ca2+-oscillation frequencies, we demonstrate that AC8 acts as a low-pass filter for high-frequency Ca2+ events, enhancing the regulatory options available to this signalling pathway.
The discovery of intracellular Ca2+ oscillations (Cuthbertson and Cobbold, 1985; Woods et al., 1986) opened the door to new concepts for the dynamic spatial and temporal regulation of cell signalling events. A number of Ca2+-dependent targets (e.g. Ras, PLC and PKC) have been shown to display parallel dynamic changes in activity, mirroring the frequency of Ca2+ transients (Bartlett et al., 2005; Cullen and Lockyer, 2002; Young et al., 2003). However, prime candidates, the cyclic AMP (cAMP)-producing Ca2+-dependent adenylyl cyclases (ACs), have until recently (Gorbunova and Spitzer, 2002) been refractory to the display of such behaviour. Four of the nine known AC isoforms are regulated by submicromolar concentrations of Ca2+. AC1 and AC8 are stimulated by Ca2+/calmodulin; AC5 and AC6 are directly inhibited by Ca2+ (Cooper, 2003). In the intact cell, these Ca2+-sensitive cyclases are targeted to lipid raft regions of the plasma membrane (Fagan et al., 2000b; Smith et al., 2002), where they are regulated by discrete modes of Ca2+ entry (Chiono et al., 1995; Fagan et al., 2000a; Fagan et al., 1998; Fagan et al., 2000b). It is highly likely, at least theoretically, that dynamic Ca2+ changes within these cellular microdomains will produce parallel, dynamic changes in cAMP to elicit an assortment of downstream events.
Regular oscillations in cAMP were first detected over 30 years ago in strips of frog ventricular muscle as a function of the myocardial contraction cycle (Brooker, 1973). The more recent discovery of Ca2+-sensitive ACs allowed the development of theoretical models, which predicted that cAMP oscillations could arise from a minimal feedback loop whereby cAMP stimulates Ca2+ entry, which in turn inhibits a Ca2+-sensitive AC5/AC6 (Cooper et al., 1995; Rapp and Berridge, 1977). A number of other scenarios have also been proposed that would support parallel cAMP and Ca2+ oscillations (Cooper et al., 1995; Gorbunova and Spitzer, 2002; Rich and Karpen, 2002; Yu et al., 2004) including Ca2+-dependent activation of cAMP-hydrolysing phosphodiesterases (PDEs).
Recent studies using PKA- or cyclic nucleotide gated channel (CNGC)-based probes to directly measure cAMP have illustrated the spatial and temporal segregation of cAMP signals to sub-cellular compartments (Rich et al., 2001a; Zaccolo and Pozzan, 2002) but they do not address the issue of potential Ca2+ and cAMP interplay. The first study to simultaneously monitor single-cell Ca2+ and cAMP confirmed Ca2+-inhibition of AC6 with some success (DeBernardi and Brooker, 1996). Later, Gorbunova and Spitzer (Gorbunova and Spitzer, 2002) provided compelling evidence for dynamic interactions between these two messengers. When using Xenopus spinal neurones loaded with a Ca2+-sensitive dye (fluo-4) and microinjected with the fluorescence resonance energy transfer (FRET)-based cAMP probe FlCRhR, they showed reciprocity between Ca2+ and cAMP. Modelling of their data predicted that bursts of Ca2+ spikes, with an interval of several minutes between bursts, would produce low-frequency cAMP oscillations. More regular Ca2+ spikes would not mediate higher frequency cAMP transients (Gorbunova and Spitzer, 2002).
It is possible that the slow response time of PKA-based cAMP probes (such as FlCRhR and GFP-tagged PKA subunits) makes them unsuitable for monitoring more rapid kinetics of cAMP signalling (Rich and Karpen, 2002). A FRET-based cAMP indicator comprised of a cAMP-binding domain of an isoform of the exchange protein directly activated by cAMP (Epac-1) has recently been developed with good sensitivity and temporal resolution (Nikolaev et al., 2004). This cAMP probe has since been employed to provide evidence for the periodic activation and inactivation of PDE1 giving rise to cAMP oscillations in insulin-secreting MIN6 cells during depolarization evoked Ca2+ events (Landa et al., 2005). Further, indirect, evidence supporting the likelihood of simultaneous Ca2+ and cAMP oscillations was obtained from cyclic-nucleotide-activated currents in frog olfactory receptor cells (Reisert and Matthews, 2001).
Here, we have used the Epac-1 fluorescent probe to examine the potential for Ca2+-mediated cAMP oscillations in human embryonic kidney (HEK293) cells expressing the Ca2+-stimulated AC8. The high temporal resolution of the catalytically inactive Epac-1 probe has enabled us to detect fast, real-time oscillations in cAMP to imposed and agonist-induced Ca2+ oscillations as a consequence of Ca2+-driven AC8 activity and the subsequent activation of the cAMP-dependent PDE4. Interestingly, higher frequencies of Ca2+ oscillation appeared to be filtered by AC8 to produce a steady rise in cAMP levels.
CCE-induced stimulation of AC8 activity
To assess the suitability of the fluorescently tagged Epac-1 probe as a sensor for Ca2+-regulated cAMP production, we first established that capacitative Ca2+ entry (CCE) induced a detectable decrease in the emission ratio of yellow fluorescent protein to cyan fluorescent protein (YFP:CFP) of the probe (representing an increase in cAMP) in HEK293 cells expressing the Ca2+-stimulable AC8. Previous studies in populations of AC8-transfected HEK293 cells suggest that the cyclase is exclusively stimulated by Ca2+ entering through CCE channels (Fagan et al., 1996). We confirmed that thapsigargin (Tg)- or carbachol (CCh)-induced CCE clearly stimulated AC8-mediated cAMP production at the single-cell level (P<0.001, Fig. 1). Treatment of fura-2-loaded HEK293 cells with 100 nM Tg (Fig. 1Ai) or 500 μM CCh (Fig. 1Bi) in the absence of extracellular Ca2+, induced a rise in cytosolic Ca2+ levels due to release of Ca2+ from the endoplasmic reticulum (ER) stores. Subsequent addition of extracellular Ca2+ (0.2, 1 or 2 mM) produced a dose-dependent, sustained rise in intracellular Ca2+ ([Ca2+]i) levels as a consequence of CCE of ∼300-700 nM. In parallel experiments, using HEK293 cells expressing the fluorescent Epac-1 probe and AC8, Tg-induced CCE was accompanied by a clear stimulation of cAMP production (Fig. 1Aii). A low concentration of the prostanoid receptor agonist prostaglandin E1 (PGE1) was included in the extracellular buffer to provide a relatively weak, priming stimulation of the ACs to allow the cAMP signal to be amplified to values above the sensitivity threshold for the genetically expressed fluorescent probe (see below). Parallel fura-2 experiments confirmed that addition of 10 nM PGE1 to the extracellular buffer did not alter the amplitude or time course of imposed Ca2+ changes (data not shown). Addition of 10 nM PGE1 at 60 seconds produced a very small decrease in the Epac-1 FRET ratio in cells 1 and 2, and a more substantial stimulation of cAMP production in cell 3 (Fig. 1Aii). All three cells exhibited enhanced cAMP production following the addition of 1 mM CaCl2 to the external solution at 300 seconds. CCh-induced CCE produced similar changes in cAMP production (Fig. 1Bii). In cell 3, AC8 was also (unexpectedly) stimulated in response to CCh-induced Ca2+ release from the ER, suggesting that Ca2+ stimulation of AC8 was also mediated by store release in a sub-population of cells (seen in 16 out of 40 cells tested). The dose dependency of CCE-evoked AC8 stimulation following Tg- and CCh-induced store emptying is illustrated (see Fig. 1Aiii and Biii, respectively). These results confirmed a good sensitivity and temporal resolution of the Epac-1-based probe. Although there were individual differences in the responses of each of the 40 cells tested to PGE, CCh, etc., they all showed a clear stimulatory response to CCE. Hence, we were able to use the cAMP sensor, Epac-1, to further characterise the interactions between Ca2+ entry and cAMP in AC8-expressing cells.
Ca2+-driven cAMP oscillations
Cytosolic Ca2+ oscillations arise spontaneously and in response to electrical or agonist stimulation. The periodicity of the oscillations ranges from less than 1 second to more than 10 minutes depending on the cell type (Goldbeter, 1996). Given the sensitivity of AC8 activity to Ca2+ rises, we hypothesized that Ca2+ oscillations are accompanied by oscillations in intracellular cAMP. To address this possibility we imposed a series of repetitive `artificial' Ca2+ transients over a range of frequencies (4×10-3 to 33×10-3 Hz) and durations (10-120 seconds) and monitored single-cell cAMP changes with the Epac-1 probe. This imposed frequency range is within the range of Ca2+ oscillations seen physiologically (Goldbeter, 1996). Cells were pre-treated with 100 nM Tg in the absence of extracellular Ca2+ to deplete the Ca2+ stores, and 10 nM PGE1 was added to stimulate AC8 activity. The extracellular solution was then periodically switched, from saline containing 1 mM CaCl2 to Ca2+-free saline. This protocol induced a repeated rise and fall in cytosolic Ca2+ levels of all fura-2-loaded cells (Fig. 2 shows averaged and calibrated fura-2 signals from more than 50 cells). Imposing a relatively low frequency of Ca2+ oscillation (Fig. 2A; a 2-minute rise in Ca2+ and 2-minute intervals between each CCE event) gave rise to concurrent oscillations in cAMP that remained in-phase with the Ca2+ transients (seen in 15 out of 28 cells tested). The remaining cells showed little or no increase in cAMP, which may reflect limited expression of AC8. The remarkably synchronized frequency of Ca2+ and cAMP changes was most likely due to Ca2+ modulation of AC8 activity, because HEK293 cells expressing only endogenous ACs did not exhibit any detectable change in cAMP levels during the imposed Ca2+ oscillations (data not shown). Decreasing duration and interval of the Ca2+ rises to 1 minute also produced cAMP oscillations that matched the frequency of Ca2+ signals (Fig. 2B; typical of 12 out of 24 cells tested). By further increasing the incidence of Ca2+ transients (Fig. 2C,D), AC8 stimulation was still seen but the single-cell cAMP signals became either more stepped in appearance or a smooth rise in cAMP production was observed. This suggests that either the cAMP signalling system or the cAMP probe was acting as a low-pass filter at the higher frequencies of Ca2+ oscillation. Alternatively, cytosolic Ca2+ levels in the vicinity of the AC may not have decreased substantially during each 20/40-second-period of wash in Ca2+-free buffer (20-100 nM decrease) and AC8 activity remained elevated as a consequence. By contrast, the onset of cAMP recovery during 1-minute and 2-minute Ca2+ oscillations was apparent within the first 10-15 seconds from switching to Ca2+-free buffer because a more dramatic fall in cytosolic Ca2+ levels (from a higher starting point) was seen.
Rates of change of cAMP and [Ca2+]
During the imposed 2-minute Ca2+ oscillations, changes in fura-2 and Epac-1 signals were clearly matched in frequency but showed some delay between the up- and down-components of the cAMP transients relative to Ca2+ changes. It is unknown whether this lag was due to limited sensitivity and response time of the cAMP biosensor, delayed responses of the cAMP signalling pathway at the site of AC stimulation and the onset of enhanced PDE activity, or the time taken for local Ca2+ levels to reach the necessary thresholds to modulate AC8 activity.
Fig. 3 shows real-time Ca2+-dependent changes in cAMP in the presence of 10 nM PGE1, that fluctuate with the same periodicity in two neighbouring cells expressing the fluorescent Epac-1 probe and AC8. Pseudocolour images of the YFP:CFP emission ratio revealed a global rise and fall in cAMP levels during the CCE-evoked Ca2+ changes (Fig. 3B). During each 2-minute period of Ca2+ entry, global cAMP levels reached a plateau that enabled us to estimate a half-maximal time constant (t1/2) for cAMP production of 27.9±2.4 seconds (n=12, Fig. 3C). This is almost double the time taken for cytosolic Ca2+ levels to reach their peak response globally (15.6±1.4 seconds). Similarly, the apparent decrease in cAMP levels was double that for Ca2+ during each period of wash in Ca2+-free saline solution (t1/2= 35.7±4.3 seconds versus t1/2=17.8±1.5 seconds, respectively). Previous in vitro assessment of a very similar, Epac-2-based, fluorescent cAMP probe (Nikolaev et al., 2004; Nikolaev et al., 2005) provided fast on-off-rates for the sensor (∼2 and 3 seconds, respectively), with the Epac-1 probe exhibiting an even faster cAMP-activation and -dissociation rate than Epac-2 in the intact cell (Nikolaev et al., 2004). Thus, the significantly delayed rise and fall in cAMP signal illustrated in Fig. 3 provides a true reflection of the response-time of AC8 to local Ca2+ changes, and the time-limited activation of cAMP-hydrolysing phosphodiesterases (PDEs) in our HEK293 cells. The essential role of PDE activity in coordinating the dynamic cAMP signals accompanying Ca2+ oscillations is illustrated in Fig. 3D. The green trace demonstrates a typical change in cAMP levels in response to Ca2+ oscillations following pre-treatment with the broad-spectrum PDE inhibitor IBMX (100 μM). The sustained rise in cAMP suggests that temporal patterning of this messenger does not depend on Ca2+ regulation of AC8 activity alone, but that local PDE activity also plays a crucial role.
Essential role of PDE4 in coordinating Ca2+ and cAMP interplay
Further investigation into the role of PDEs examined the effects of selective PDE inhibitors on the cAMP oscillation profile during artificially imposed 2-minute Ca2+ oscillations (Fig. 4). Representative examples of single-cell cAMP changes in the presence of selective PDE inhibitors (Fig. 4B-E) illustrate the relative roles of PDE4, PDE3 and PDE1 in cAMP hydrolysis following repeated CCE-evoked stimulation of AC8 (Fig. 4A). Previous analysis in HEK293 cells has shown that around 70% of PDE activity is provided by PKA-activated PDE4, with the remainder provided by PDE3 (Lynch et al., 2005). Consistent with these findings, we reveal that pre-treatment of our cells with 10 μM rolipram (a selective PDE4 inhibitor) prevented recovery of the Ca2+-induced cAMP transients by at least 90% (Fig. 4E,F, P<0.001, n=23). The PDE3 selective inhibitor cilostamide (10 μM), also produced a small reduction in cAMP hydrolysis as detected by the Epac-1 probe (Fig. 4D,F; P<0.05, n=9). The PDE4 and PDE3 isoforms are unlikely to be directly activated by changes in cytosolic [Ca2+] and reflect PDE activation as a direct consequence of increased cAMP production. The activity of the PDEs, in particular PDE4, is only manifest once Ca2+ levels fall and AC8 activity is significantly decreased. Control experiments were also performed using the PDE1 inhibitor 8-methoxymethyl 3-isobutyl-1-methylxanthine (MMX, 10 μM). The absence of any effect rules out a role for a Ca2+-stimulable PDE1 (Fig. 4C,F, n=16), consistent with evidence that PDE activity in HEK293 cells is entirely mediated by PDE4 and PDE3 (Lynch et al., 2005).
Comparison of Ca2+-regulated AC8 activity using different AC activators
Combined activation of AC8 with PGE1 and induction of Ca2+ entry through CCE channels provides a robust increase in cAMP production in Epac-1-expressing HEK293 cells (see above). However, data from our laboratory using radioassays to measure cAMP production in populations of HEK293 cells suggest that, CCE stimulated AC8 activity is also seen in the absence of agonist-mediated AC stimulation, although more robust Ca2+ stimulation of AC8 is seen in the presence of the prostanoid receptor agonist PGE1, the β-adrenergic receptor agonist isoprenaline, or during direct activation of ACs with forskolin (our unpublished observations) (Fagan et al., 1996). The enhanced sensitivity of AC8 to Ca2+/calmodulin in the presence of AC agonist provides support for the speculation that Ca2+-sensitive cyclases can act as coincidence detectors (Abrams et al., 1991; Impey et al., 1994). Thus, selectivity in the dual regulation of AC8 by different agonists and local Ca2+ changes was explored (Fig. 5). In the absence of added agonist a very modest, transient stimulation of cAMP levels was seen during 2-minute periods of CCE following ER-store depletion with 100 nM Tg (Fig. 5A). Similar results were seen in two other cells, although the majority of Epac-1-expressing cells (24 out of 27 cells tested) produced no detectable change in cAMP as a consequence of imposed Ca2+ oscillations alone. Although this supports the theory that AC8 acts as a coincidence detector, requiring agonist activation of AC8 in addition to Ca2+ stimulation, the lack of detectable cAMP change might also reflect the resolution limits of the cAMP sensor. When AC8-expressing HEK293 cells were incubated with 10 nM PGE1 (Fig. 5B, representative of 21 out of 34 cells tested), 10 nM isoprenaline (Fig. 5C, representative of 28 out of 46 cells tested) or 10 nM forskolin (Fig. 5D, representative of 13 out of 31 cells tested) the imposed Ca2+ oscillations were accompanied by a concurrent rise and fall in cAMP levels. Interestingly, periods of PDE activity between each Ca2+ load in the presence of isoprenaline or forskolin were less pronounced than that seen during activation with PGE1 suggesting that, isoprenaline and forskolin might target endogenous ACs (not stimulated by Ca2+) more effectively than PGE1, thereby providing a steady baseline rise in cAMP production independent of cytosolic Ca2+ levels. In conclusion, our data suggest that Ca2+-dependent modulation of cAMP production is not specific to the AC activator used. In the absence of agonist, low-frequency (2 minutes) CCE-evoked Ca2+ oscillations produced a modest Ca2+-dependent activation and inactivation of AC8, thereby generating small oscillations in cAMP in three cells out of 27 tested. These results suggest that prior activation of AC8 is not strictly necessary to observe Ca2+-driven cAMP dynamics. Nevertheless, the amplitude of the transients is enhanced when AC8 is stimulated, either directly or through activation of Gs.
Agonist-evoked Ca2+ and cAMP oscillations
The ability of artificially imposed relatively low-frequency Ca2+ oscillations to produce dynamic changes in global cAMP production strengthens modelling predictions of simultaneous oscillations in the two messengers. However, we wanted to see whether more physiological Ca2+ transients were equally capable of inducing oscillations in cAMP. For these experiments, we used the muscarinic receptor agonist CCh (10 μM). At this concentration, CCh induces regular oscillations in cytosolic Ca2+ over a frequency range of 0.3 to 5.5 oscillations minute-1 in approximately 35% of HEK293 cells (e.g. see Fig. 6Aii). These Ca2+ oscillations are likely to arise from a combination of Ca2+-store release, intracellular Ca2+ buffering and/or extrusion, and CCE (Bird and Putney, 2005). In the remaining cells, CCh produced a large initial Ca2+ transient followed by a plateau (Fig. 6Ai). Switching to Ca2+-free extracellular media slowed and eventually stopped the Ca2+ oscillations, or terminated the plateau phase. Numerous Epac-1 probe experiments were performed to see whether CCh-evoked Ca2+ oscillations could give rise to similar temporal patterns in cAMP production in AC8-expressing HEK293 cells. As before, these experiments were conducted in the presence of 10 nM PGE1 to enhance AC8 activity. In 308 out of 502 cells tested, 10 μM CCh produced a sustained rise in cAMP levels, as represented by the trace in Fig. 6Bi. Switching to Ca2+-free buffer produced a subsequent decrease in cAMP levels. In nine other cells, the application of 10 μM CCh produced an initial large rise in cAMP levels, followed by regular cAMP oscillations (Fig. 6Bii). The frequency of the cAMP oscillations ranged from approximately 1.3 to 3.0 cAMP transients minute-1 (1.8±0.2 transients minute-1 on average), consistent with the frequency of Ca2+ oscillations seen in parallel fura-2 measurements (0.3 to 5.5 transients minute-1). Upon removal of extracellular Ca2+ the cAMP oscillations slowed and then stopped. These findings support the hypothesis that dynamic interplay between Ca2+ entry and cAMP production can give rise to cAMP signals that encode signalling information in their frequency in addition to the spatial localization and amplitude of cAMP signals.
Methodologies for measuring real-time cAMP changes are slowly gaining pace with the development of a range of sensors based on downstream cAMP targets including PKA (Adams et al., 1991; Zaccolo et al., 2000; Zhang et al., 2001), cyclic-nucleotide-gated channels (Fagan et al., 2001; Rich et al., 2001b) and, most recently, Epacs (DiPilato et al., 2004; Nikolaev et al., 2004; Ponsioen et al., 2004). The PKA-based cAMP sensor FlCRhR has previously been used to illustrate the presence of low frequency neuronal cAMP oscillations during bursts of Ca2+ spikes (Gorbunova and Spitzer, 2002). However, the response kinetics, signal-to-noise ratio and potential buffering by endogenous and fluorescent PKA-subunits may limit its ability to detect smaller and/or more rapid changes in cAMP in response to a single Ca2+ transient. Biosensors for cAMP based on CNGC mutants that monitor cAMP production using either patch-clamping or fura-2 measurements of Ca2+ entry through the CNG channels (Fagan et al., 2001; Rich et al., 2001b) provide greater temporal resolution of cAMP signals with minimal buffering. However, they are compromised when investigating the interplay between Ca2+ and cAMP production because the channels conduct, and can be inhibited by, Ca2+ (Kaupp and Seifert, 2002). The recent introduction of a cAMP probe comprised of an Epac-1-binding domain tethered to CFP and YFP has provided the opportunity to assess real-time changes in cAMP with good sensitivity (EC50=2.35 μM) and relatively high temporal resolution (on- and off-rates ∼2 and 3 seconds, respectively) (Nikolaev et al., 2004).
Direct interactions between Ca2+ and cAMP have provided the basis for models predicting their interdependent oscillations in excitable cells (Cooper et al., 1995; Gorbunova and Spitzer, 2002; Rapp and Berridge, 1977; Yu et al., 2004). Central to all of these models are the actions of Ca2+ on Ca2+-inhibited and Ca2+-stimulated ACs respectively. Other key features include the feedback of cAMP onto Ca2+ channels and Ca2+/calmodulin-dependent PDE activity. Evidence of a dynamic interplay between Ca2+ and cAMP as a consequence of Ca2+-dependent PDE activity (PDE1C) was presented in a recent paper by Landa and colleagues (Landa et al., 2005). In this report, we reveal that cAMP oscillations can arise in non-excitable cells lacking both VGCCs and Ca2+-stimulated PDE activity. Artificially imposed or CCh-induced Ca2+ oscillations were shown to generate simple oscillations in cAMP matching the Ca2+ oscillation frequency. These oscillations were due to the stimulation of a Ca2+-sensitive AC (AC8), and subsequent PDE4 activity driven by the oscillations in cAMP (and activation of PKA). Simultaneous Ca2+ and cAMP oscillations were more readily observed in the presence of a low dose of AC8 agonist (PGE1, isoprenaline or forskolin) although CCE-evoked changes in Ca2+ alone (in the absence of GS-coupled receptor activation) were capable of mediating detectable cAMP transients. The enhanced sensitivity of agonist-activated AC8 to Ca2+ changes is consistent with the enzyme acting as a coincidence detector for the convergence of Gs- and Ca2+/calmodulin-mediated signals. The dual regulation of Ca2+-stimulated cyclases in this manner is hypothesized to underlie their role in learning and memory (Abrams et al., 1991; Impey et al., 1994).
Here, we have not addressed potential feedback effects of cAMP on the Ca2+ transients themselves; however, this can not be excluded from the catalogue of events that shape the frequency, amplitude and duration of cAMP and Ca2+ signals. The incidence of sustained cAMP oscillations in AC8 expressing cells during agonist application was low (seen in 2% of cells) compared with the number of cells in which we observed robust cAMP oscillations during more prolonged, artificially-imposed Ca2+ transients (60% of cells). However, this is not surprising given the number of factors that may influence the detection of more physiological oscillations in cAMP under our experimental conditions. These include; the respective levels of expression of AC8 and PDE in individual cells, the presence and/or frequency of CCh-induced Ca2+ oscillations reflecting differences in Ca2+ buffering and receptor expression, and localization of the cAMP probe to sub-cellular sites where AC8 is active. Almost 40% of cells gave no response to CCh-evoked Ca2+ changes, suggesting that sufficient expression or localization of AC8 or Epac probe, respectively, in any individual cell may account for the lack of effect seen in a significant proportion of the cells. This value is comparable to the number of cells that did not yield detectable cAMP increases during the low frequency, artificially imposed Ca2+ changes, despite clear Ca2+ changes in all cells in parallel fura-2 measurements. It seems reasonable to suggest that the occurrence of cAMP oscillations in cells endogenously expressing Ca2+-regulated ACs is likely to be much higher in response to a stimulus producing Ca2+ oscillations, providing they occur near the AC and that the AC is active. Furthermore, higher frequency Ca2+ oscillations may be filtered to produce a steady rise in cAMP, as seen for the more rapid artificially imposed Ca2+ transients. Here, agonist evoked cAMP oscillations were detected at a maximum rate of approximately three oscillations per minute (0.05 Hz). This is significantly faster than cAMP oscillations previously reported in Xenopus neurones that occurred at a rate of just four oscillations per hour and lasted for several minutes (Gorbunova and Spitzer, 2002), though much faster oscillations are predicted to exist (Cooper et al., 1995; Yu et al., 2004). We propose that the upper rate for cAMP oscillations of ∼0.05 Hz reported here provide an accurate indication of the temporal limits of cAMP production and hydrolysis that can occur physiologically, although the association and/or dissociation rate of the biosensor with cAMP are crucial when determining cAMP dynamics. Ca2+ oscillations have been shown to occur at even higher frequencies, approaching 4 Hz in excitable cells (Allen et al., 1984). Our findings predict that high-frequency Ca2+ oscillations would be accompanied by a smooth rise in cAMP production, with cAMP oscillations only being generated if there are regular intervals between the high-frequency Ca2+ events.
In conclusion, we demonstrate that oscillations in cytosolic Ca2+ can produce simultaneous dynamic changes in cAMP mediated by the stimulatory effects of Ca2+ on AC8, combined with the activation of PDE4 (and perhaps PDE3). These oscillations had a peak frequency of ∼3 cAMP transients minute-1, with the cAMP signalling system acting as a low pass filter for more regular Ca2+-mediated events. The ability of cells to shape the temporal characteristics of cAMP signals provides an important means of expanding the diversity of downstream signalling events mediated by this ubiquitous messenger. Furthermore, the regional distribution of Ca2+-sensitive ACs to lipid rafts may give rise to spatially restricted cAMP signals, in response to local Ca2+ changes. It is now becoming evident that the organization of numerous cellular activities can be controlled by the exquisite interplay between these two messengers to provide signals that are unique for a given stimulus, in terms of their frequency, amplitude, duration and sub-cellular location. We are inclined to speculate that the Ca2+ sensitivity of AC8 (and other ACs) can decode relatively low-frequency Ca2+ oscillations. This is reminiscent of calmodulin kinase II which is considered to be susceptible to different frequencies of Ca2+ oscillation as a part of its role in LTP and LTD (De Koninck and Schulman, 1998). cAMP oscillations, like those shown here, in response to Ca2+ oscillations may themselves be differentially decoded by downstream targets such as PKA, Epac, cAMP-binding protein (CREB) and Rap to initiate events such as gene expression and cell differentiation (Montminy, 1997; York et al., 1998). Future challenges include the real-time measurement of Ca2+ and cAMP changes within sub-cellular microdomains using probes with high spatial and temporal resolution. We predict that local cAMP signals at sites of Ca2+ entry within the cell can be even more dynamic than the global cAMP changes reported here.
Materials and Methods
Cell culture and transfection
HEK293 cells (European Collection of Cell Cultures, Porton Down, UK) were grown in minimum essential medium supplemented with 10% (v/v) foetal bovine serum and maintained at 37°C in a humidified atmosphere of 95% air and 5% CO2. To produce cells stably expressing Epac-1, the HEK293 cells were plated on 100-mm dishes at ∼60% confluence 1 day prior to transfection of 2 μg of Epac-1 probe cDNA with the Ca2+ phosphate method. One day after transfection, the culture media was replaced with fresh media containing 400 μg/ml genetecin (G418) to select untransfected cells. The Epac-1-expressing HEK293 cells were re-fed every 3 days. Stable Epac-1-expressing HEK293 cells were established from around 2 weeks following transfection. These cells were plated onto 25-mm poly-L-lysine-coated coverslips 24 hours prior to transient transfection of 1 μg of AC8 cDNA with the Ca2+ phosphate transfection method. All cAMP measurements were made at ∼48 hours post AC8 transfection.
Single-cell Ca2+ measurements
Cells were plated onto 25-mm poly-L-lysine-coated coverslips 24 hours prior to loading with 4 μM fura-2 AM and 0.02% Pluronic F-127 (Molecular Probes, Leiden) for 45 minutes at room temperature in extracellular buffer containing (in mM): 140 NaCl, 4 KCl, 1 CaCl2, 0.2 MgCl2, 11 D-glucose, 10 HEPES pH 7.4. After loading, cells were washed several times and then imaged using a CoolSNAP-HQ CCD camera (Photometrics) and monochromator system (Cairn Research, Kent, UK) attached to a Nikon TMD microscope (×40 objective). Emission images (D510/80M) at 340 nm and 380 nm excitation were collected at 1 Hz with MetaFluor software (Universal Imaging). Fluorescence ratio changes were calibrated in vitro using a Ca2+ imaging calibration kit (Molecular Probes, Leiden). For Ca2+-free buffers, the constituents were the same as for extracellular buffer but 1 mM CaCl2 was omitted and replaced by 0.1 mM EGTA.
Fluorescent imaging of Epac-1-expressing HEK293 cells was performed using a CoolSNAP-HQ CCD camera (Photometrics) and an Optosplit (505DC) to separate CFP (470 nm) and YFP (535 nm) emission images (Cairn Research, Kent). For dual emission-ratio imaging, cells were excited at 436 nm with a monochromator (Cairn Research) and 51017 filter set (Chroma, VT, USA) attached to a Nikon TMD microscope (×40 objective). Emission images at 470 nm and 535 nm were collected every 2-5 seconds (200-400 millisecond integration time) and then background-subtracted and analysed with Metamorph imaging software (Universal imaging). Certain criteria were used when selecting cells for analysis. First, CFP and YFP fluorescence intensity had to be at least twice the background fluorescence. Second, cells with excessive expression of the fluorescent probe were excluded. FRET data are plotted at changes in 535 nm versus 470 nm (YFP:CFP) emission ratio for each individual cell. Changes in rates of cAMP production in Fig. 1Aiii,Biii (YFP:CFP min-1 pre:post CCE) were calculated by dividing the 535/470 nm ratio change during the 1 minute period immediately prior to addition of extracellular Ca2+ (induction of CCE), by the ratio change recorded over the 1-minute period following the addition of Ca2+. Simultaneous cAMP and Ca2+ measurements using the cAMP probe and fura-2 were not possible with our imaging system constraints of fixed excitation and emission filter positions.
cDNA for the fluorescently tagged Epac-1 cAMP sensor was a kind gift from Martin J. Lohse. This work was supported by The Wellcome Trust and NIH grant NS 28389.
- Accepted November 15, 2005.
- © The Company of Biologists Limited 2006