Selective coupling of type 6 adenylyl cyclase with type 2 IP₃ receptors mediates direct sensitization of IP₃ receptors by cAMP

Cells use a limited repertoire of diffusible intracellular messengers, including cAMP and Ca²⁺, to regulate most aspects of their behavior, yet they succeed in responding appropriately to a barrage of extracellular signals. The versatility of intracellular messengers is increased by their precise spatial and temporal organization. Local increases in Ca²⁺ (Berridge, 1997) or cAMP (Zaccolo and Pozzan, 2002; Mongillo et al., 2006; Smith et al., 2006; Willoughby and Cooper, 2007) can selectively regulate closely associated proteins, whereas more global changes regulate different processes. Both Ca²⁺ and cAMP can also be delivered in precise temporal patterns (Berridge, 1997; Gorbunova and Spitzer, 2002; Dyachok et al., 2006). The frequencies of these messenger spikes are important in determining responses to Ca²⁺ (Dupont et al., 2003) and are also likely also to be important for cAMP spikes. The versatility of both messengers is increased further by interactions between them (Bruce et al., 2003; Werry et al., 2003; Screaton et al., 2004; Willoughby and Cooper, 2007). Ca²⁺, for example, regulates cAMP formation and degradation, and by regulating protein kinases and calcineurin, it modulates many effects of protein kinase A (PKA). Conversely, cAMP regulates formation and phosphorylation of inositol 1,4,5-trisphosphate (IP₃), the activities of Ca²⁺ channels and pumps, and formation of other Ca²⁺-mobilizing messengers. Such interactions endow cells with computational ability (Bray, 1995). Many proteins involved in cAMP and Ca²⁺ signaling exist in multiple isoforms, which may differ in behavior and expression. This diversity adds further to the complexity of the signaling pathways, but its physiological significance is incompletely understood.

Parathyroid hormone (PTH) plays a central role in the regulation of plasma Ca²⁺ concentration, although its receptors are also expressed in tissues unrelated to Ca²⁺ homeostasis. Each of the related PTH receptors (types 1 and 2) belongs to the family of G protein–coupled receptors that includes those for secretin, calcitonin, and glucagon (Jüppner et al., 1991; Behar et al., 1996). PTH receptors share with these an ability to both stimulate adenylyl cyclase (AC) and increase cytosolic Ca²⁺ concentration ([Ca²⁺]_i), but it is unclear how PTH stimulates Ca²⁺ release from intracellular stores. Some evidence suggests that PTH receptors stimulate PLC-β via Gq or Gi, and thus formation of IP₃ (Jüppner et al., 1991; Mahon et al., 2006). PTH receptors might also, like β-adrenoceptors, activate PLC-ε via cAMP (Schmidt et al., 2001). Other evidence suggests that Ca²⁺ release occurs without formation of IP₃ (Seuwen and Boddeke, 1995; Short and Taylor, 2000), and perhaps even in the presence of an antagonist of IP₃ receptors (IP₃R; Seuwen and Boddeke, 1995). We (Short and Taylor, 2000; Tovey et al., 2003) and others (Buckley et al., 2001) have suggested that PTH potentiates IP₃-evoked Ca²⁺ release, but the intracellular signals responsible have not been identified.

Here, we find that PTH, and other receptors that stimulate cAMP formation, communicate with IP₃R via intracellular “cAMP junctions,” wherein cAMP passes directly from a specific isoform of AC (AC6) to a specific subtype of IP₃R (IP₃R2) to increase its IP₃ sensitivity. This identifies IP₃R as a new target for cAMP, an unexpected interaction between two ubiquitous signaling pathways, and a novel binary mode of cAMP signaling that allows robust digital recruitment of a response by graded concentrations of an extracellular stimulus.

Carbachol (CCh), via endogenous muscarinic receptors, stimulates PLC and thereby IP₃ formation and Ca²⁺ release from the intracellular stores of human embryonic kidney (HEK) cells (Fig. 1 A). In the absence of extracellular Ca²⁺, stimulation of HEK cells stably expressing human type 1 PTH receptors (HEK-PR1 cells) with a supramaximal concentration of CCh caused an increase in [Ca²⁺]_i that returned to basal levels within 60–70 s. The addition of PTH, in the continued presence of CCh, evoked a further release of Ca²⁺ (Fig. 1 A). This additional Ca²⁺ release results from sensitization of IP₃R to IP₃ (Fig. S1, A and B; Short and Taylor, 2000; Tovey et al., 2003). Some analogues of PTH recruit other signaling pathways without stimulating AC (Takasu et al., 1999), but only the analogues that stimulate AC potentiated Ca²⁺ release (Fig. 1, E and F). Prostaglandin E₁ (PGE₁; unpublished data) and isoproterenol, which stimulate AC via their endogenous receptors, also potentiated CCh-evoked Ca²⁺ signals. So too did forskolin (FK), which directly stimulates AC (Fig. 1, B–D). However, 1,9-dideoxyforskolin (100 μM), an FK analogue that does not stimulate AC but mimics the AC-independent effects of FK (Zerr et al., 1996), had no effect on [Ca²⁺]_i either alone or with CCh (unpublished data). None of these stimuli significantly increased [Ca²⁺]_i in the absence of CCh (Fig. S1, C and D; Willoughby et al., 2006). Single cell analyses established that in most cells that responded to CCh (98 ± 1%), the Ca²⁺ signal was potentiated by PTH (99 ± 1%), isoproterenol (80 ± 4%), and FK (95 ± 3%).

PTH and a high concentration of the membrane-permeant analogue of cAMP, 8-Br-cAMP (30 mM), similarly potentiated CCh-evoked Ca²⁺ signals (Fig. 2, A–F); 8-Br-cGMP (30 mM) was ineffective (not depicted). Much lower (<100 μM) concentrations of 8-Br-cAMP activate PKA (He et al., 2003), but they did not significantly potentiate Ca²⁺ signals (Short and Taylor, 2000; Tovey et al., 2003). The low sensitivity to 8-Br-cAMP (half-maximal effective concentration, EC₅₀ = 0.87 ± 0.37 mM; Fig. 2 C) is important and is discussed further in later sections.

The extent to which the maximal Ca²⁺ release evoked by CCh can be enhanced is limited by the finite size of the intracellular Ca²⁺ stores, but there is no such limit on the extent to which the sensitivity to CCh can be increased. Two independently acting stimuli should cause an additive increase in CCh sensitivity. It is significant, therefore, that maximal concentrations of 8-Br-cAMP and PTH had the same effects on both the sensitivity of the Ca²⁺ signals to CCh and the maximal response (Fig. 2 E), and it is also significant that their effects were not additive (Fig. 2 F). The nonadditive effects of PTH and 8-Br-cAMP on the EC₅₀ for CCh-evoked Ca²⁺ signals establish the fact that cAMP is the only signal linking PTH receptors to sensitization of IP₃R (Fig. 2 G). G protein βγ subunits have been reported to stimulate IP₃R directly and so to cause release of Ca²⁺ (Zeng et al., 2003). PTH cannot work via this pathway because PTH does not alone cause Ca²⁺ release (Fig. S1), nor does somatostatin (1 μM), which activates Gi and thereby the release of βγ subunits (unpublished data; Law et al., 1993; Willoughby et al., 2007). We conclude that cAMP alone mediates the effects of PTH on IP₃R (Fig. 2 G). We provide additional evidence to support this key conclusion later (see Fig. 6).

All three subtypes of IP₃R can be phosphorylated by PKA, although at different sites and with different consequences (Bruce et al., 2003; Soulsby and Wojcikiewicz, 2007). IP₃R2 is the major (46%) IP₃R subtype in HEK cells (Table I), and in other cells expressing mainly IP₃R2, PKA potentiates IP₃-evoked Ca²⁺ release (Burgess et al., 1991; Bruce et al., 2003). In HEK-PR1 cells, PTH caused modest phosphorylation of IP₃R (Figs. 3 A and S2 B), and PKA very slightly (though not to a statistically significant degree) increased the sensitivity of the intracellular Ca²⁺ stores to IP₃ (see Fig. 4 A).

PTH, FK, and 8-Br-cAMP stimulated phosphorylation of many proteins in HEK-PR1 cells, and H89, an inhibitor of PKA, prevented this (Fig. 3 B, and see Fig. 5 F). But neither H89 (Fig. 3, C and E) nor other inhibitors of PKA (Figs. 3 F and S2, C–F) affected potentiation of the CCh-evoked Ca²⁺ signals. A kinase–anchoring proteins (AKAPs) contribute to PKA-mediated inhibition of IP₃R3 by cholecystokinin (Straub et al., 2002) and to many other PKA-mediated events (Wong and Scott, 2004). A membrane-permeant form of ht31 peptide (st-ht31), which uncouples AKAPs from PKA (Willoughby et al., 2006), blocked phosphorylation of many of the proteins phosphorylated in response to PTH and FK without affecting the response to 8-Br-cAMP (Fig. 3 B). But st-ht31 had no effect on the potentiation of CCh-evoked Ca²⁺ signals by FK (not depicted) or PTH (Fig. 3 D). We conclude that phosphorylation of IP₃R by PKA slightly potentiates responses to IP₃, and PTH stimulates modest phosphorylation of IP₃R, but the phosphorylation is not required for PTH to potentiate IP₃-evoked Ca²⁺ release.

Other proteins are also regulated by cAMP (Dremier et al., 2003), notably cyclic nucleotide-gated cation channels (CNGC; Rich et al., 2000) and epacs (Bos, 2003). Indeed, a cAMP-evoked increase in [Ca²⁺]_i has been proposed to result from activation of epac, Rap 2, and PLC-ε (Schmidt et al., 2001). By Western blotting (WB), we detected expression of epac-1 (but not epac-2) in HEK-PR1 cells (unpublished data), but no significant effect of isoproterenol, PTH, or FK alone on [Ca²⁺]_i (Fig. S1; Willoughby et al., 2006), although each potentiated responses to CCh (Figs. 1 B and S1). Nor did we detect formation of IP₃ in response to PTH (Short and Taylor, 2000). Epacs are less sensitive than PKA to cAMP (Bos, 2003), but 8-Br-2′-O-methyladenosine-3′,5′-cAMP selectively activates epacs (Christensen et al., 2003). Even very high concentrations (30 mM) of this analogue, which is more membrane-permeant than 8-Br-cAMP, had no effect on CCh-evoked Ca²⁺ signals, whereas 8-Br-cAMP potentiated them (Fig. 2 C). These results establish that potentiation of Ca²⁺ release by PTH results from sensitization of IP₃R rather than production of IP₃, and neither PKA nor epac is required.

In permeabilized HEK cells, high concentrations of cAMP (EC₅₀ = 2.7 ± 1.0 mM; Fig. 4, A and B) increased the sensitivity of IP₃-evoked Ca²⁺ release by ∼4.5-fold. This is much greater than the ∼30% increase evoked by PKA (Fig. 4 A) but similar to the three- to fourfold increase in the sensitivity to CCh evoked by PTH or 8-Br-cAMP in intact cells (Fig. 2 E). The responses of permeabilized cells to cAMP and of intact cells to 8-Br-cAMP were unaffected by inhibition of PKA (by 10 μM H89; unpublished data). Potentiation of IP₃-evoked Ca²⁺ release by cAMP was mimicked by 8-Br-cAMP (EC₅₀ = 324 ± 45 μM; Fig. 4, C and D) but not by 5′-AMP, ATP, ADP, or GTP (10 mM; not depicted). This low-affinity interaction between cAMP and IP₃R is consistent with only high concentrations of 8-Br-cAMP mimicking PTH in intact cells. Indeed, the sensitivities of intact (Fig. 2 C) and permeabilized cells (Fig. 4 D) to 8-Br-cAMP are similar. We conclude that cAMP binds directly to a low-affinity site on the IP₃R (or an associated protein) to increase its sensitivity to IP₃. The interaction is not mediated via changes in IP₃ binding because cAMP has no effect on ³H-IP₃ binding to IP₃R2 (unpublished data). The affinity of this site for cAMP is ∼600-times lower than that of epacs (Bos, 2003) and ∼20,000-times lower than that of PKA or CNGC (Wong and Scott, 2004). It is unsurprising, therefore, that there is no known consensus sequence for cAMP binding within IP₃R (Dremier et al., 2003). A novel low-affinity site mediates the effects of cAMP on IP₃R.

Despite the evidence that cAMP entirely mediates sensitization of IP₃R (Figs. 1–4,²³⁴, S1, and S2), there is a massive disparity in the relationship between cAMP and Ca²⁺ for different stimuli (Fig. 5, A and B). This is most clear from the comparison of the amount of cAMP produced when each agonist causes [Ca²⁺]_i to increase by 30 nM (Fig. 5 C). The disparity is inconsistent with global cAMP signals serving as a universal currency, which are used similarly by each stimulus to regulate IP₃R. We note also that although HEK-PR1 cells express similar numbers of PTH receptors and β-adrenoceptors (see Fig. 7 E), maximal activation with PTH causes a much greater stimulation of AC (Fig. 1, C and D), which suggests that an active PTH receptor stimulates AC more effectively than does an active β-adrenoceptor.

Manipulation of cAMP levels by inhibiting either cAMP formation or its degradation by cyclic nucleotide phosphodiesterases (PDE) further highlights the inconsistent relationship between cAMP and potentiation of Ca²⁺ signals. 3-isobutyl-1-methylxanthine (IBMX), which inhibits most isoforms of PDE that degrade cAMP (Fisher et al., 1998), failed to enhance Ca²⁺ signals. Instead, it caused a slight inhibition (Fig. 5 E and S3 B) that became more pronounced with prolonged incubation (Short and Taylor, 2000). This probably results from activation of PKA and loss of Ca²⁺ from intracellular stores via IP₃R made more sensitive to basal levels of IP₃ after their phosphorylation by PKA (Short and Taylor, 2000). More strikingly, massive inhibition of cAMP formation with a combination of inhibitors of AC (9-(tetrahydro-29-furyl)adenine [SQ 22536] and 2′,5′-dideoxyadenosine [DDA], referred to hereafter as SQ/DDA; see the legend for Fig. 5) failed to attenuate the potentiated Ca²⁺ signals evoked by any concentration of FK, isoproterenol (Fig. S3, B and D), or PTH (Fig. 5 E). The latter was the case even though SQ/DDA abolished PTH-evoked protein phosphorylation (Fig. 5 F), and much lower concentrations of DDA blocked epac-mediated stimulation of PLC-ε (Schmidt et al., 2001).

Our results seem paradoxical. The effects of PTH on IP₃R are mediated entirely by cAMP (Figs. 1–4,²³⁴), yet in intact cells, there is no consistent relationship between cAMP and Ca²⁺ (Fig. 5, A–E). This first led us to think that the effect of PTH on Ca²⁺ signals was not mediated by cAMP (Short and Taylor, 2000), but that conclusion is now untenable. Maximally activated receptors often generate more intracellular messenger than needed to evoke a maximal response. This allows downstream events to be more sensitive (lower EC₅₀) than earlier ones to the initial signal (Strickland and Loeb, 1981). Such systems are said to have “spare receptors” because activation of only a fraction of the receptors can evoke a maximal response. Responses to maximal stimulation may then survive even substantial inhibition of downstream signaling. But in our experiments, the EC₅₀ values for cAMP formation and potentiation of Ca²⁺ signals, measured under identical conditions, are similar for each stimulus (Table II). Furthermore, even when maximal stimulation generates surplus signal, a globally distributed messenger cannot have a safety margin when it evokes a submaximal response: reducing the amount of messenger must reduce the response. Yet in our experiments, the Ca²⁺ signals evoked by every concentration of each stimulus were entirely insensitive to substantial inhibition of AC (Figs. 5 E and S3, B and D). How can cAMP mediate graded regulation of IP₃R and yet operate with a substantial safety margin at every level of stimulus intensity?

We suggest that cAMP sensitizes IP₃R independent of PKA and epac (Figs. 3 and S2). Signaling by cAMP must be local, and within a signaling complex, a receptor must generate substantially more cAMP than needed maximally to sensitize associated IP₃R (Fig. 5 H). Each complex, or cAMP junction, operates with a large safety margin so that even substantial inhibition of AC fails to compromise communication with IP₃R. The concentration-dependent effects of PTH (or FK) come from recruitment of increasing numbers of these hyperactive junctions and not from increased activity within each (Fig. 5 H). For receptors, the safety margin might arise at various steps of the signaling pathway: a single receptor might, for example, sustain the activity of more Gs than needed to stimulate sufficient AC to cause maximal activation of associated IP₃R (Fig. 5 I). But for FK, only cAMP lies between it and the IP₃R, and yet the Ca²⁺ signals evoked by any concentration of FK are also insensitive to substantial inhibition of AC (Fig. S3, A and B). Activation of a single AC can, therefore, deliver more cAMP than needed maximally to sensitize the IP₃R associated with it. This defines the minimal signaling complex: an AC–IP₃R junction (Fig. 5 I, box). The key point is that each active receptor (or AC) saturates the local signaling machinery and so functions as a local on–off switch. The very low affinity of the cAMP-binding site of the IP₃R (Fig. 4 B) effectively insulates each signaling complex from its neighbors because diffusion will reduce the cAMP concentration to below that needed to sensitize IP₃R within just a few nanometers of the complex (Rich et al., 2000). This digital signaling to IP₃R imposes no requirement for individual PTH receptors (or AC) to differ in their sensitivities; PTH binding to a uniform population of PTH receptors, governed by the law of mass action, is sufficient to generate a graded increase in the number of active receptors as the PTH concentration increases.

Our scheme (Fig. 5, H and I) predicts that AC and IP₃R are intimately associated, and that complete inhibition of cAMP formation within some junctions should more effectively attenuate Ca²⁺ signaling than partial inhibition of all junctions. Subsequent experiments confirm these predictions, but first we provide additional evidence that cAMP mediates the effects of PTH.

We established stable HEK-PR1 cell lines in which expression of Gαs was reduced by >95% by means of siRNA (Fig. 6 A). In these cells, CCh-evoked Ca²⁺ signals were normal (Fig. 6 B), FK potentiated them, and the response to FK was unaffected by inhibitors of AC (Fig. 6 C). However, whereas inhibition of PKA again had no effect on the Ca²⁺ signals evoked by PTH, the sensitivity to PTH was reduced by inhibition of AC in cells with reduced expression of Gαs (Fig. 6 D). These results show that when the safety margin for signaling between PTH receptors and IP₃R is eroded by removal of Gαs, the requirement for AC activity is unmasked. This further confirms that cAMP mediates the effects of PTH on IP₃R (Fig. 2 G).

We used immunoprecipitation (IP) to determine whether any of the many isoforms of AC and IP₃R expressed in HEK cells are specifically associated. AC6 accounts for only 5% of the AC expressed in HEK cells (Table I), but an antibody to it (AbAC6) selectively precipitated 30% of IP₃R2 and lesser amounts of IP₃R3 and IP₃R1 (Fig. 7, A and B). An antibody that recognizes all mammalian AC isoforms (AbACc) selectively precipitated the same amount of IP₃R2 as AbAC6 (Fig. 7 B); and after IP with AbAC6, AbACc was unable to precipitate further IP₃R2 (Fig. 7 C). AC3 comprises 56% of the AC expressed in HEK cells (Table I), and an antibody to it (AbAC3) selectively precipitated some IP₃R2, but much less effectively than AbAC6 (Fig. 7, A and B). AC9 is also relatively abundant (10%), but IP₃R were not precipitated by AbAC9 (Fig. 7 B and S4 D). None of the AbAC precipitated the sarcoplasmic reticulum (SR)/ER Ca²⁺-ATPase (SERCA), and nor were IP₃R precipitated by AbSERCA (Fig. 7, A and B). These results establish that the association of IP₃R with AC depends entirely on AC6. The lesser association of IP₃R with AC3 (Fig. 7 B) may result from hetero-oligomerization of AC3 with AC6. Both isoforms appear to dimerize (Fig. S4 B), and AC does form heterodimers (Willoughby and Cooper, 2007; Baragli et al., 2008). We considered whether PTH might itself regulate association of AC6 with IP₃R2, but the ability of AbAC6 to cause IP of IP₃R2 was unaffected by PTH (unpublished data).

Reciprocal IP from native cells using AbIP₃R was impossible because available AC antibodies do not reliably detect endogenous AC in WB (Willoughby and Cooper, 2007). We therefore transiently expressed tagged human AC3 and AC6 in HEK cells, and confirmed that each AbAC selectively precipitated only the appropriate AC (Fig. S4, A–C). In the transiently transfected cells, AbIP₃R2 but not AbIP₃R3 selectively precipitated AC6, but not AC3 (Fig. 7 D). We conclude that AC6 and IP₃R2 are specifically associated. The lesser associations with AC3 and IP₃R3 probably result from interactions between heterotetrameric IP₃R (Fig. S2 A) and dimeric AC (Fig. S4 B) that include the essential IP₃R2 and AC6 subunits (Fig. 7 E).

Our analysis of AC and IP₃R expression (Table I and Fig. 8) suggest that a HEK cell expresses ∼1,500 AC6 and ∼13,300 IP₃R2 subunits (∼3,300 homotetrameric IP₃R2), and at least 30% of IP₃R2 (∼1,000 IP₃R) are associated with AC6 (Fig. 7 B). These estimates cannot reliably define the stoichiometry of the IP₃R2–AC6 complex, but they are consistent with each AC6 associating with a single tetrameric IP₃R2 (Fig. 7 E).

In HEK cells, phosphorylation by PKA of PDE4D anchored to the plasma membrane by an AKAP (Willoughby et al., 2006) or recruited by arrestin (Willoughby et al., 2007) massively accelerates cAMP degradation. This rapidly attenuates the increase in cAMP immediately beneath the plasma membrane, whether evoked by PGE₁ (Willoughby et al., 2006) or isoproterenol (Willoughby et al., 2007). Our measurements (at 30 s) precede maximal activation of this PDE, which occurs ∼1 min after stimulation with PGE₁ and ∼3 min after stimulation of β₂-adrenoceptors (Willoughby et al., 2006). However, when the interval between PTH and CCh additions was extended to 5 min to allow maximal activation of anchored PDE4D by PKA, the potentiated Ca²⁺ signals remained insensitive to inhibition of AC or PKA at every concentration of PTH (Fig. 7 F). These results demonstrate that even when PDE is massively stimulated, inhibition of AC fails to attenuate cAMP-mediated communication with IP₃R. Nor is communication with IP₃R improved by inhibiting PDE either directly with IBMX (Figs. 5 E and S3, B and D) or indirectly by preventing its recruitment and activation by PKA-mediated phosphorylation (Fig. 3, A and B; and Figs. 7 F and S2). We conclude that the interaction between AC6 and IP₃R2 is more intimate than that between AC and PKA, epac, or CNGC (Willoughby et al., 2006), even though the latter detect only cAMP immediately beneath the plasma membrane. Only the IP₃R allows cAMP to evade PDE as it passes from AC to its target (Fig. 7 G). We conclude that AC6 and IP₃R2 are specifically associated, and their intimacy is sufficient to allow cAMP to pass between them without encountering PDE. This reveals an important distinction between two different modes of cAMP signaling: binary (to IP₃R) and analogue (to PKA and epac). Because epac and PKA bind cAMP with high affinity, they can respond in a graded fashion to cAMP diffusing to them from a distant AC, but they then assemble with PDE to ensure effective local degradation of the messenger (Dodge-Kafka et al., 2005; Willoughby et al., 2006, 2007). In contrast, IP₃R, with their low affinity for cAMP, must be very close to AC, but diffusion alone is enough to terminate the signaling. Both modes of cAMP signaling generate spatially organized responses: the analogue mode used by high-affinity cAMP sensors relies on local degradation of cAMP, whereas the binary mode requires local synthesis of cAMP to deliver it at high concentration to a low-affinity cAMP sensor (Fig. 7 G).

We used siRNA to inhibit selectively the expression of IP₃R2, IP₃R1, AC3, and AC6 (Fig. 9 A and Table III). Although siRNA also reduced IP₃R3 expression, these cells failed to attach to the plates used for Ca²⁺ and cAMP assays, preventing further analysis. Loss of IP₃R1 decreased the Ca²⁺ signals evoked by CCh without affecting their potentiation by PTH. In contrast, loss of IP₃R2 had no affect on the Ca²⁺ signals evoked by CCh alone but attenuated their potentiation by PTH (Fig. 9, B and C). These results suggest that the muscarinic receptors that alone evoke Ca²⁺ release are distributed differently to those that release Ca²⁺ in synergy with cAMP. IP₃R1 mediates the Ca²⁺ release evoked by CCh alone, whereas IP₃R2 mediates the response to CCh with cAMP (Fig. 9 D). Loss of AC3, the major AC isoform, reduced PTH-evoked cAMP formation, though without affecting the ability of PTH to potentiate CCh-evoked Ca²⁺ signals. Conversely, loss of AC6 had no detectable effect on PTH-evoked cAMP formation but attenuated PTH-evoked Ca²⁺ signals (Fig. 9, E and F). The Ca²⁺ release evoked by CCh alone was unaffected by loss of AC3 or AC6 (Fig. S4, E and F). These results confirm that the specific association between IP₃R2 and AC6 revealed by IP (Fig. 7) underlies the ability of PTH to potentiate CCh-evoked Ca²⁺ signals.

Because we propose that each cAMP junction operates with a large safety margin (Fig. 5 H), complete inhibition of AC within some junctions should more effectively inhibit sensitization of IP₃R than partial inhibition of AC within all junctions. SQ 22536 and DDA are low-affinity inhibitors of all AC (Reid et al., 1990) that should rapidly and uniformly distribute their inhibition across all AC. Such uniform inhibition of AC by ∼70% does not inhibit Ca²⁺ signaling (Fig. 5, D and E; and Fig. S3). In contrast, inhibition of AC6 expression by siRNA is expected to cause random loss of AC from individual signaling complexes, each perhaps comprising only a dimeric AC associated with a single IP₃R (Fig. 5, H and I). A 62% loss of AC6 (∼3% of total AC) had no detectable effect on PTH-evoked cAMP formation but inhibited PTH-mediated Ca²⁺ signaling by 30% without affecting the sensitivity to PTH (EC₅₀ = 63 ± 3 nM and 54 ± 5 nM, before and after loss of AC6, respectively; Fig. 9 F). These results confirm our second prediction. When all AC (and thus all AC6) is uniformly inhibited by 60–70% (by SQ/DDA), Ca²⁺ signaling is unaffected. In contrast, the same overall loss of AC6 activity, though distributed in an all-or-nothing fashion between junctions (by reducing AC6 expression), reduces the magnitude of the Ca²⁺ signals without affecting the PTH sensitivity of the remaining junctions (Fig. 9 G).

We can compare quantitatively the relationships between cAMP and Ca²⁺ because each was measured at the same time under identical conditions, all cells responded similarly, and cAMP alone mediates the effects of PTH (Figs. 1–4,²³⁴). The similar sensitivity of intact and permeabilized cells to 8-Br-cAMP (Figs. 2 C and 4 D) confirms the fact that permeabilized cells retain their sensitivity to cAMP. This allows comparisons of the cAMP sensitivity of IP₃R in permeabilized cells with the calculated concentrations of cAMP evoked by PTH if cAMP were uniformly distributed through the cytosol. For PTH, which produces more cAMP than other stimuli (Fig. 1, C and D), the calculated cAMP concentrations are ∼1,000-fold too low to modulate IP₃R. The disparity is even greater when AC is inhibited and the AC–IP₃R junctions operate with a reduced safety margin. Therefore, the half-maximal Ca²⁺ signal evoked by PTH is associated with an estimated global cAMP concentration of ∼500 nM (Fig. 10 A), whereas the half-maximal response in permeabilized cells requires a 5,400-fold higher concentration of cAMP (2.7 mM; Fig. 4 B).

Subplasma membrane microdomains of cAMP occur in many cells, including HEK cells (Rich et al., 2000; Willoughby et al., 2006, 2007). But the highest local concentrations of free cAMP achieved, whether measured or estimated from the activity of AC and cAMP diffusion (Rich et al., 2000), exceed the global cytosolic cAMP concentration by no more than 10-fold, even within a few nanometers of AC. Cytosolic cAMP concentrations, even very close to AC, are far too low to affect IP₃R directly, and restricting the escape of cAMP with internal membranes like the ER (Rich et al., 2000) might further increase the local cAMP concentration by no more than approximately twofold (Olveczky and Verkman, 1998). Cyclic AMP cannot reach IP₃R via the cytosol at concentrations sufficient to cause their sensitization to IP₃. We instead suggest that the intimate association between IP₃R2 and AC6 (Fig. 7) allows cAMP to pass directly from AC to IP₃R in a manner analogous to substrate channeling in enzyme complexes (Fig. 10 B; Huang et al., 2001).

We have shown that receptors that stimulate cAMP formation sensitize IP₃R to IP₃ via a pathway that requires neither PKA nor epac, but which is instead mediated by a novel low-affinity cAMP-binding site on either the IP₃R itself or an associated protein. A complex of AC6 and IP₃R2 is uniquely required for PTH to potentiate IP₃-evoked Ca²⁺ release. Within each AC–IP₃R junction, cAMP is delivered directly to the associated IP₃R at a supersaturating concentration, allowing the junction to work as a robust on–off switch. Graded responses to changes in the intensity of the extracellular stimulus arise from digital recruitment of AC–IP₃R junctions and not from graded activity of individual junctions. The presence of AC6 in these junctions is significant because AC6 is inhibited by Ca²⁺ (Willoughby and Cooper, 2007). It is therefore likely that within the AC6–IP₃R2 junction, there will be a reciprocal interplay between AC6 and IP₃R, with AC6 potentiating IP₃R activity via cAMP, and active IP₃R inhibiting AC6 via Ca²⁺ (Fig. 10 B). These feedback interactions may generate highly localized oscillations in both Ca²⁺ and cAMP similar to those observed in whole cells (Gorbunova and Spitzer, 2002; Dyachok et al., 2006).

The AC–IP₃R junction suggests an analogy with excitation–contraction coupling in striated muscle, where voltage-sensing dihydropyridine receptors (DHPR) in the plasma membrane regulate opening of ryanodine receptors in the SR across a narrow (∼10 nm) junctional complex (Fig. 10 D). Coupling of these proteins has evolved from chemical coupling mediated by Ca²⁺ (cardiac excitation–contraction coupling) to a physical coupling between the two apposed proteins (vertebrate skeletal muscle; Di Biase and Franzini-Armstrong, 2005). Junctional complexes similar to those in striated muscle occur also in nonexcitable cells; they are implicated in Ca²⁺ signaling (Treves et al., 2004; Varnai et al., 2007), and their width (10–15 nm) would allow IP₃R and AC to abut (Fig. 10 C). We speculate that the AC6–IP₃R2 complex may occur within such junctions, and that its evolution from the looser relationship between AC and PKA mediated by soluble cAMP may be similar to the evolution of conformational coupling in skeletal muscle from Ca²⁺-mediated coupling in cardiac muscle (Di Biase and Franzini-Armstrong, 2005). The difference is that AC6–IP₃R2 communication requires not only direct contact but also passage of a constrained messenger, cAMP, between them (Fig. 10 B).

The IP₃R is a new and important target of cAMP. The AC6–IP₃R2 complex allows cAMP to pass directly to IP₃R and increase its sensitivity to IP₃. These cAMP junctions (Fig. 10 B) allow robust signaling to IP₃R because they are hyperactive on–off switches (Fig. 5 H). Increasing stimulus intensities recruit additional junctions, rather than increasing the activity within individual junctions. Signaling by cAMP to IP₃R is binary, requires direct contact between AC6 and IP₃R2, and relies on diffusion of cAMP to terminate the response of each junction. Other sensors with greater affinity for cAMP, like PKA and epac, respond to graded changes in local cAMP concentration (analogue signaling) and rely on PDE activity to shape local signaling events (Figs. 7 G and 10 D). These different modes of signaling considerably increase the versatility of cAMP as an intracellular messenger.

DDA, cAMP, 8-Br-cAMP, 8-Br-cGMP, H89 dihydrochloride, ionomycin, KT5720, adenosine 3′,5′-cyclic phosphorothioate-Rp (RpcAMPS), SQ 22536, and 1,2-bis(o-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid (BAPTA) were obtained from EMD. 8-Br-2'-O-methyladenosine-3′,5′-cAMP was obtained from Biolog. IP₃ was obtained from American Radiolabeled Chemicals. Thapsigargin was obtained from Alomone Laboratories. Human PTH fragments were obtained from Bachem; unless specifically stated, “PTH” denotes human PTH(1–34). Cell culture media, G-418, Fluo-4AM, Fura-2AM, Mag-fluo-4AM, and Pluronic F-127 were obtained from Invitrogen. Acetic anyhydride, CCh, carbonyl cyanide-p-trifluoromethoxyphenylhydrazone, FK, 1,9-dideoxyforskolin, IBMX, (±)-isoproterenol hydrochloride, the porcine catalytic subunit of PKA, and triethylamine were obtained from Sigma-Aldrich. St-ht31 was obtained from Promega.

HEK-PR1 cells were cultured as described previously (Short and Taylor, 2000). For siRNA-mediated inhibition of AC or IP₃R expression, HEK-PR1 cells were seeded into 6-well plates (2.5 × 10⁵ cells per well). After 24 h, the medium was replaced (2.5 ml) and Stealth siRNA duplex (30 pmol; Invitrogen) was added in Opti-MEM 1 + Glutamax (0.5 ml) containing 1% lipofectamine RNAimax (Invitrogen). The siRNA sequences (5′–3′) were: IP₃R1 (GGCCUGAGAGUUACGUGGCAGAAAU), IP₃R2 (GAGAAGGCUCGAUGCUGAGACUUGA), AC3 (CCUCUGAAGAUGAGCACGAGCUCAA), and AC6 (CCAGCAUCUUCCUGCUGCUGCUAAU). After 24 h, cells were reseeded into 96-well (2,500 cells per well) or 6-well plates (10⁵ cells per well), and after a further 48 h, the transfection with siRNA was repeated. Cells were incubated for a further 72 h before use. A modified pSUPER vector (Brummelkamp et al., 2002) was used to generate short hairpin RNA–mediated knockdown of Gαs in HEK-PR1 cells. Transfected cells were selected using 10 μg/ml blasticidin and 800 μg/ml G418, and clonally isolated cells were used to establish the seven cell lines used for analyses of cAMP and Ca²⁺ signaling.

The open reading frames for human AC3 (AK122926) in pME18SFL3 (National Institute of Technology and Evaluation Biological Resource Center) and AC6 (BC064923) in pCMV-Sport 6 (GeneService) were cloned as EcoRI–XhoI fragments into pENTR1a (Invitrogen). A Flag (AC3) or Myc (AC6) tag was introduced at the N terminal by PCR. Tagged and untagged AC constructs were subcloned into the pEGFP-C2 vector (Clontech Laboratories, Inc.) to generate AC constructs N-terminally tagged with EGFP (Fig. S4 A). HEK-PR1 cells were transfected using lipofectamine and used after 2 d.

Near confluent cultures of HEK-PR1 cells grown in 96-well plates and loaded with Fluo-4 were used for measurements of [Ca²⁺]_i in cell populations using a FlexStation 96-well fluorescence spectrometer (MDS Analytical Technologies; Tovey et al., 2006). The results reported herein span >4 yr, but all direct comparisons between stimuli or between cAMP and Ca²⁺ were made in parallel experiments. Single cell imaging of [Ca²⁺]_i was performed as described previously (Tovey et al., 2003), and after correction for autofluorescence, fluorescence ratios were calibrated to [Ca²⁺] using Ca²⁺ standard solutions (Invitrogen). CCh stimulation was always in Ca²⁺-free Hepes-buffered saline (HBS; except in Fig. S1, C and D). A low-affinity luminal Ca²⁺ indicator was used to measure IP₃-evoked Ca²⁺ release from permeabilized cells (Tovey et al., 2006).

HEK-PR1 cells in 24-well plates were cultured for 2–3 d (as described in a preceding section). They were washed twice in HBS (135 mM NaCl, 5.9 mM KCl, 1.2 mM MgCl₂, 1.5 mM CaCl₂, 11.6 mM Hepes, and 11.5 mM glucose, pH 7.3) and then incubated under identical conditions (including addition of CCh) to those used for measurements of [Ca²⁺]_i; the only difference was the omission of Fluo-4-AM and Pluronic F-127 from the 1-h incubation. After stimulation, the medium was aspirated, and the cells were lysed by the addition of acidified ice-cold ethanol (10 mM HCl in absolute ethanol). Dried extracts in 500 μl of medium (50 mM sodium acetate, 1 mM theophylline, and 1 mg/ml BSA, pH 5.0) were acetylated by rapid sequential addition of 10 μl triethylamine and 5 μl acetic anhydride. Acetylated cAMP content was determined by RIA using acetylated standards prepared from a cAMP stock calibrated by its UV absorption (ε₂₅₈ = 14,100). For RIA (Brooker et al., 1979), rabbit antiserum to acetylated cAMP (P.D. Marley, University of Melbourne, Melbourne, Australia; Marley et al., 1991) was used at a final concentration of 1:9,000, and 8.4 TBq/mmol adenosine 3′,5′-cyclic phosphoric acid, 2'-O-succinyl[¹²⁵]iodotyrosine methyl ester (PerkinElmer) was used as tracer. Assays (300 μl) included 100 μl of sample. After incubation for 48 h at 4°C, 2 mg/ml BSA and 2 mg/ml Norit A charcoal were added in 50 mM sodium phosphate buffer at pH 7.4, and the free and antibody-bound tracer were separated (1,000 g for 10 min).

All procedures were performed at 4°C. A confluent 175-cm² flask of HEK-PR1 cells was washed in 50 ml PBS supplemented with protease inhibitors (one tablet of Roche complete protease inhibitor per 50 ml). Cells were scraped into 1.5 ml solubilization medium, sonicated (three times for 10 s), and incubated with rotation for 1 h. Solubilization medium had the following composition: 140 mM NaCl, 5 mM NaF, 10 mM Tris, 1 mM Na₄P₂O₇, 0.4 mM Na₃VO₄, 1% Triton X-100, pH 7.4, and one mini-tablet of Roche protease cocktail inhibitor per 10 ml. Cell debris was removed by centrifugation (5,000 g for 10 min) and the supernatant was cleared by incubation (1 h) with 5 μl of preimmune rabbit serum. Protein agarose Ig A/G beads (10 μl; Autogen Bioclear) were added, and after 2 h, immunoprecipitated material was removed by centrifugation (1,000 g for 1 min). An aliquot (300 μl) of the cleared lysate was incubated with the primary antiserum (Table S1) for 1 h before addition of protein A/G–agarose beads (50 μl) and incubation for 12 h. After centrifugation (1,000 g for 1 min), the supernatant and pellet (washed three times) were used for WB. WB used 3–8% Tris acetate NuPAGE gels (Invitrogen), and blots were quantified using a GeneGnome (SynGene) with gel loadings adjusted to ensure that intensities scaled linearly with protein loading. Antibodies are listed in Table S1.

cDNA was synthesized in a final volume of 20 μl directly from cell lysate using Fastlane cell cDNA kit (QIAGEN) according to the manufacturer's instructions. For QPCR, each reaction included primers specific for AC (Ludwig and Seuwen, 2002) or IP₃R subtypes (Quantitect Primer Assay; QIAGEN), and for calibration, primers for a house-keeping gene (GAPDH; Ogunwobi et al., 2006). Each reaction (25 μl) contained 1 μl of the HEK-PR1 reverse-transcription product, 0.5 μM of the AC- or IP₃R-specific primers, and Quantace Sensimix according to the manufacturer's instructions (Bioline). For PCR (Rotor-Gene 6000; Corbett Life Sciences), an initial denaturation at 93°C for 10 min was followed by 40 cycles of amplification (93°C for 10 s, 50–65°C for 15 s, and 72°C for 20 s). Expression levels were calculated as described previously (Moneer et al., 2005).

HEK-PR1 cells were washed in HBS supplemented with a complete protease inhibitor cocktail (Roche) and resuspended (∼4 × 10⁶ cells per milliliter) in TM (50 mM Tris and 1 mM EDTA, pH 8.3) for ³H-IP₃ binding, in DM (75 mM Tris, 1 mM EDTA, 12.5 mM MgCl₂, and 0.2% BSA, pH 7.4) for ³H-dihydroalprenolol (DHA) binding, and in FM (50 mM Tris and 10 mM MgCl₂, pH 7.4) for ³H-FK binding. Cells were lysed (Polytron, Inc.), and the lysates were used for equilibrium competition binding assays. Incubations, in a final volume of 500 μl, included ³H-IP₃ (1.5 nM, 21 Ci/mmol; PerkinElmer), ³H-DHA (0.165 nM, 97 Ci/mmol; GE Healthcare), or ³H-FK (10 nM, 27 Ci/mmol; PerkinElmer; with 100 μM p[NH]ppG, 10 μM isoproterenol, and 40 μM cytochalasin; Emala et al., 2000); each with appropriate concentrations of unlabeled competing ligand. After equilibrium had been attained (5 min on ice for ³H-IP₃, 90 min at 25°C with shaking for ³H-DHA, and 10 min at 25°C for ³H-FK), incubations were terminated by centrifugation (20,000 g for 5 min at 4°C) and the pellet was washed with cold buffer, then resuspended in 200 μl of buffer for liquid scintillation counting. For ¹²⁵I-PTH binding, cells in 24-well plates (2–5 × 10⁵ cells per well) were incubated in PM (Dulbecco's modified Eagles medium with 20 mM Hepes and 0.1% BSA) with ¹²⁵I-PTH (44 pM, 1559 Ci/mmol; Bachem) for 5 h at 4°C (Chauvin et al., 2002). Cells were washed twice with ice-cold PBS and lysed with 0.8 M NaOH, then their radioactivity was determined. Equilibrium competition binding curves were fitted to four parameter logistic equations (GraphPad Prism; GraphPad Software) from which IC₅₀ (and thence K_D) and B_max values were calculated. Expression of Gαs was quantified by WB, calibrated using recombinant Gαs (EMD).

Concentration-effect relationships were individually fitted to four-parameter logistic equations by nonlinear curve-fitting (Prism 4; GraphPad Software), and the results from each experiment (EC₅₀, Hill coefficient h, and maximal response) were pooled for analyses. For simplicity, EC₅₀ values are given as means ± SEM, although log EC₅₀ values were used for statistical analysis. Student's t test, or a one-way analysis of variance test followed by a post hoc Bonferroni's test were used as was appropriate.

Fig. S1 shows that PTH potentiates Ca²⁺ signals by sensitizing IP₃R. Fig. S2 shows that potentiation of Ca²⁺ responses by isoproterenol and FK does not require PKA. Fig. S3 shows that inhibition of cAMP formation or degradation does not directly effect potentiation of Ca²⁺ responses by isoproterenol or FK. Fig. S4 shows the selectivity of the AC antibodies and the lack of effect of AC3 or AC6 siRNA on CCh-evoked Ca²⁺ release. Table S1 lists the antibodies used.

We thank P. Marley, S. Yarwood, and G. Milligan for antisera to acetylated-cAMP, epac-1, and Gαs, respectively; and M. Falcke (Berlin) for advice on diffusion.

This work was supported by the Wellcome Trust (072084).