Subcellular Organization of the cAMP Signaling Pathway

A. Gs Protein Coupled Receptors

An individual cell detects the message from a huge number of extracellular cues via a multitude of specific GsPCRs at the plasma membrane, and the unique informational content of each extracellular stimulus is retained even though all these signals converge toward raising intracellular cAMP. Clear evidence shows that the different localization of GsPCRs at the plasma membrane contributes to triggering a stimulus-specific cellular response (Averaimo et al., 2016; Halls, 2019). The localization of GsPCRs within different plasma membrane domains (e.g., rafts versus nonraft domains) can physically couple the receptor with diverse components of the signaling cascade, e.g., specific G proteins and AC isoforms (Insel et al., 2005; Pontier et al., 2008; Magalhaes et al., 2012; Agarwal et al., 2018). For example, in the neuroblastoma cell line PC12, in which AC6 and AC7 are the two predominant cyclase isoforms, the cAMP response to activation of the pituitary adenylate cyclase–activating polypeptide type I receptor by its selective endogenous neuropeptide agonist pituitary adenylate cyclase–activating polypeptide 38 was unaffected by knockdown of AC7 but abolished by knockdown of AC6, indicating that coupling of pituitary adenylate cyclase–activating polypeptide type I receptor and AC6 is an exclusive requirement for pituitary adenylate cyclase–activating polypeptide signaling (Emery et al., 2015). In adult mouse cardiac myocytes, caveolin-3 was found to direct the localization of AC5 predominantly to T-tubular invaginations of the plasmalemma, whereas AC6 was found to reside outside the T-tubules, where it is responsible for β₁-AR–mediated enhancement of the L-type Ca²⁺ current (Timofeyev et al., 2013). Another study found that the cAMP signal generated by activation of β₂-AR is mainly confined to T-tubules, where its localization requires caveolin-3 (Wright et al., 2014), whereas β₁-AR–induced cAMP signals are more diffuse across the entire cell (Nikolaev et al., 2010). The arrangement of receptors at the plasma membrane and their preferential coupling with ACs is not static, and this can significantly contribute to fine-tuning the cellular response to a given stimulus. That activation can affect the localization of GPCRs at specific plasma membrane domains has long been recognized, as studies have shown, for example, that treatment with agonist leads to a marked decline in the abundance of β₂-AR (but not β₁-AR) in caveolae (Rybin et al., 2000). In addition, GPCRs can switch their coupling to ACs in a regulated manner. For example, in uterine smooth muscle cells, α₂-AR stimulation is inhibitory in early pregnancy via predominant Gαi-mediated inhibition of ACs but becomes stimulatory after midterm pregnancy via selective Gβγ activation of AC2 (Zhou et al., 2000, 2007). Evidence also suggests that the spatial arrangement of GPCRs can be disrupted in disease. For example, in chronic heart failure, β₂-ARs redistribute from the T-tubules to nontubular portions of the plasmalemma, leading to loss of compartmentalization of the β₂-AR–dependent cAMP signal (Nikolaev et al., 2010). Selective stimulation of β₁-AR generates a more global cAMP signal and activates proapoptotic pathways and hypertrophy (Engelhardt, 2007), whereas selective stimulation of β₂-AR results in a more confined cAMP signal, is antiapoptotic, and has protective effects on the myocardium (Yano et al., 2008). It is possible that the redistribution of β₂-AR observed in heart failure may contribute to the failing phenotype as a consequence of disrupted cAMP compartmentation, although direct experimental evidence for a causative role is still missing.

Evidence supports confinement of GsPCRs to membrane hot spots defined by actin fibers, microtubules, and clathrin-coated pits where both receptors and G proteins are concentrated and where G protein coupling and signaling preferentially occur (Magalhaes et al., 2012). Activation of the relaxin family peptide receptor 1 (RXFP1) is an example of how differential coupling of GsPCR with distinct signaling pathway components can result in subcellular context-specific responses. Studies have shown that picomolar relaxin can generate cAMP by binding to a preformed complex involving the relaxin receptor RXFP1 and Gαs/AC2. A slower second wave of cAMP is then generated upon binding of nanomolar relaxin to RXFP1; activation of Gα13 leads to cAMP synthesis via a pathway involving Gβγ-phosphoinositide 3-kinase/protein kinase C (PKC) and AC5 (Nguyen and Dessauer, 2005; Halls et al., 2006, 2007; Dschietzig et al., 2011). However, evidence supports distinct functional outcomes downstream of these different RXFP1-activated pathways, with the Gαs-coupled pathway promoting cAMP response element–controlled gene transcription and the Gα13 pathway regulating nuclear factor kappa-light-chain-enhancer of activated B cells (NFκB)-mediated gene transcription (Halls et al., 2007). Interestingly, the expression level of different Gα13 subunits can vary significantly—for example, in disease conditions (Dschietzig et al., 2011)—providing further mechanisms whereby the functional outcome of receptor activation and cAMP signaling depends on the contingent cellular state.

A recent study connects differential expression levels of Gα subunits and sub–plasma membrane location of GPCR as factors that contribute to the response output downstream of a certain receptor. The data show that the level of expression of different Gα subunits affects efficacy and potency of β-AR ligands, a feature that the authors attribute to Gα subunit–directed partitioning of the receptors in different sub–plasma membrane domains (Onfroy et al., 2017). The hypothesis is that the impact of Gα subunits on receptor membrane distribution could account for receptor-biased pharmacology such that a specific ligand could promote different conformations of the GPCR/G protein complexes, with different outputs depending on their compartmentalization at the cell surface. How biased signaling intersects with compartmentalized cAMP signaling is an important question that remains largely unexplored (Shen et al., 2019).

In addition to differential expression and localization at the plasma membrane, the ability of at least some Gαs-coupled receptors to trigger cAMP production after internalization provides additional mechanisms for diversification of the cAMP message at a temporal and spatial level (Irannejad et al., 2013; Gidon et al., 2014). The conventional paradigm is that, upon activation, receptors are internalized as part of the GPCR desensitization process, which involves receptor phosphorylation and recruitment of β-arrestin followed by GPCR endocytosis (Freedman and Lefkowitz, 1996). Receptor internalization and its degradation in the lysosome provides a mechanism that protects the cell from excessive stimulation, and receptor recycling back to the plasma membrane induces resensitization and protects cells from prolonged hormone resistance (Zhang and Eggert, 2013). However, studies combining imaging of real-time cAMP signaling and GPCR trafficking events in the intact cell have challenged the dogma that internalized GPCRs are inactive by showing that sustained cAMP signaling can occur from GsPCR embedded in internalized vesicles (Freedman and Lefkowitz, 1996; Ferrandon et al., 2009; Calebiro et al., 2010; Lan et al., 2012; Irannejad et al., 2013; Stoddart et al., 2015; Namkung et al., 2016; Pavlos and Friedman, 2017), where they couple with ACs that are internalized with the receptor (Lazar et al., 2020).

The first direct evidence that internalized ligand/GPCR complexes can stimulate synthesis of cAMP in mammalian cells came from two independent studies showing that thyroid stimulating hormone receptor (TSHR) and parathyroid hormone receptor, respectively, are able to elicit a “second wave” of Gs protein activation after ligand-induced internalization (Calebiro et al., 2009; Ferrandon et al., 2009). Conformation-specific single-domain antibodies that can directly probe the activated state of β₂-AR were used to demonstrate that the adrenergic agonist isoproterenol promotes an immediate activation of β₂-AR and Gs protein at the plasma membrane and, within several minutes after agonist application, a delayed activation at the membrane of early endosomes (Irannejad et al., 2013). The temporal shift of these two waves of cAMP signals provides opportunities for encoding complex biologic information that the cell can translate into specific cellular effects (Grundmann and Kostenis, 2017). Evidence indicates that signaling by internalized GPCR is functionally relevant and is responsible, for example, for meiosis resumption in the oocyte within the follicle (Lyga et al., 2016) or regulation of transcription of certain cAMP-dependent genes (Tsvetanova and von Zastrow, 2014; Godbole et al., 2017). One study used a bacteria-derived adenylyl cyclase that allows cAMP production to be acutely induced by light and that was selectively targeted either to the plasma membrane or to the membrane of endosomes. cAMP generated on light stimulation at the endosomes, but not at the plasma membrane, triggered expression of the phosphoenolpyruvate carboxykinase 1 gene, a cAMP-regulated enzyme, confirming that information can be encoded in the spatial pattern of cAMP signals and supporting the notion that cAMP synthesis at the endosomes can effectively uncouple transcriptional control from cAMP generated at the plasma membrane (Tsvetanova and von Zastrow, 2014). Different Gs-coupled receptors can adopt distinct modalities of signaling after internalization. For example, whereas internalized parathyrod hormone receptors or β₂-adrenergic receptors appear to signal from the early endosome compartment (Ferrandon et al., 2009; Irannejad et al., 2013) and their retrograde trafficking to the trans Golgi coincides with signal termination (Feinstein et al., 2011), signaling from internalized TSHR occurs only at the Golgi/trans Golgi network (TGN) and not at early endosome membranes (Godbole et al., 2017). For the latter modality, the model is that thyroid stimulating hormone (TSH)/TSHR complexes traffic to the TGN and there activate Gs proteins to induce cAMP synthesis. This leads to activation of a local PKA subset and regulation of gene transcription. Importantly, inhibition of TSHR internalization shows no effect on the overall amplitude of the cAMP response but prevents cAMP/PKA signaling at the Golgi as well as phosphorylation of the cAMP response element binding protein (CREB) and gene transcription. In line with this model, subcellular anchoring of PKA to A-kinase anchoring proteins (AKAPs) appears to be required for TSH-dependent regulation of gene transcription (Calebiro et al., 2006), although the detailed architecture of this Golgi/TGN signalosome and its regulation remain largely to be defined. Thus, receptor-specific signaling after internalization can potentially occur at distinct intracellular sites, providing additional opportunities to achieve response specificity via compartmentalization of cAMP.

A number of studies have also documented cAMP signaling from intracellular GPCRs that are located at intracellular membranes independently from receptor internalization (Kumar et al., 2008; Suofu et al., 2017; Dahl et al., 2018; Nash et al., 2019). For example, studies show that a pre-existing β₁-AR receptor pool that is not delivered from the cell surface resides at the Golgi, where it can be activated by endogenous ligands that enter the cell by facilitated transport via the organic cation transporter 3 (Irannejad et al., 2017). Activation of this internal pool of receptors induces Epac-phospholipase Cε (PLCε)–dependent phosphatidylinositol-4-phosphate (PI4P) hydrolysis in a process that has been linked to catecholamine-dependent cardiomyocyte hypertrophy (Nash et al., 2019).

B. Adenylyl Cyclases

Two classes of mammalian ACs, the nine transmembrane (AC1 to AC9) and the soluble (AC10) enzyme, generate cAMP from ATP either in response to GsPCR activation or bicarbonate, respectively, or via cross talk with other signaling pathways (Hindson, 2003). All transmembrane ACs share the same general architecture, yet they vary considerably outside their catalytic and transmembrane domains and are subject to complex regulation by a number of different modulators, including G proteins, PKA, protein kinase C, and calcium (Sunahara et al., 1996; Dessauer et al., 2017). The specific regulation and subcellular localization of different transmembrane AC isoforms provide an ideal mechanism for compartmentalization of cAMP signaling (Cooper and Tabbasum, 2014; Halls and Cooper, 2017; Johnstone et al., 2018). For example, AC1, AC3, and AC8 are Ca²⁺ activated, whereas AC5 and AC6 are inhibited by Ca²⁺ (Wachten et al., 2010; Halls and Cooper, 2017). Ca²⁺ regulation, however, can be complex, as AC3, for example, is activated by Ca²⁺/calmodulin and inhibited by the Ca²⁺/calmodulin protein kinase CaMKII (Wei et al., 1998). Although both AC5 and AC6 are Ca²⁺-inhibited, disruption of AC5 increases longevity and protects against aging-induced cardiomyopathy (Yan et al., 2007), whereas deletion of AC6 has opposite effects, leading to increased mortality upon β-AR stimulation–induced cardiomyopathy (Tang et al., 2013). These opposite effects have been ascribed to the different localization of these two enzymes at the plasma membrane. In arterial smooth muscle cells, AC6 and, to some extent, AC3 have been associated with vasodilatory pathways via regulation of K⁺ channels (Ostrom et al., 2002; Nelson et al., 2011), whereas AC5 was shown to mediate vasoconstriction in response to high glucose. Super-resolution analysis of arterial myocytes revealed a nanometer proximity between subpopulations of AC5 and the L-type Ca²⁺ subunit Cav1.2. Glucose-induced AC5 activity was found to mediate the activation of an anchored PKA pool, phosphorylation of Cav1.2, potentiation of L-type Ca²⁺ channel activity, increased [Ca²⁺]_i, and vasoconstriction, a chain of events that the authors suggest underpins increased myogenic tone during diabetic hyperglycemia (Syed et al., 2019).

In other studies, targeted FRET-based sensors were used to detect cAMP in different subcellular domains in GH3B6 cells, a pituitary cell line that secretes prolactin in response to thyrotropin-releasing hormone (TRH) and vasoactive intestinal peptide (VIP). These two hypothalamic hormones act via different signaling pathways: TRH activates a Gq-coupled receptor and signals via inositol 1,4,5-trisphosphate (IP₃)-dependent Ca²⁺ release from the endoplasmic reticulum (ER), whereas VIP activates a G_s-coupled receptor that signals via cAMP. The authors could measure an increase in cAMP at the plasma membrane in response to VIP when using a sensor targeted via a myristylation and palmitoylation motif, but no cAMP signal was detected when using a sensor targeted to the plasma membrane via fusion to AC8, suggesting a coupling of the VIP receptor with ACs other than AC8. In contrast, when the same cells were stimulated with TRH, a dramatic increase in cAMP was registered by an AC8-targeted sensor, whereas the cAMP response at regions of the plasma membrane away from AC8 was significantly smaller (Wachten et al., 2010). The authors found that the different local cAMP response was due to Ca²⁺-dependent activation of AC8 and to Ca²⁺-dependent inhibition of AC5 and AC6 (as well as to concomitant stimulation of cytosolic PDE1 by Ca²⁺) (Wachten et al., 2010). These findings illustrate how different stimuli can generate locally disparate cAMP responses via selective engagement of individual transmembrane ACs, which respond according to their contingent regulatory status.

The structural basis for the ability of different transmembrane ACs to participate in distinct signaling pathways appears to be their unique coupling with signaling elements at the plasma membrane (Fig. 1). A number of transmembrane ACs have been found to be part of multiprotein complexes organized by AKAP (see below) (Bauman et al., 2006; Dessauer, 2009; Li et al., 2012) and including GPCRs/G proteins (Dupré et al., 2007), PDEs, and phosphatases (Crossthwaite et al., 2006) (see below). These complexes provide a local platform for a compartmentalized source of cAMP at the plasma membrane, where ACs can undergo feedback inhibition via PKA (e.g., AC5, 6, and 8) or feed-forward regulation via PKC (e.g., AC2) (Bauman et al., 2006; Willoughby et al., 2012; Efendiev et al., 2013; Shen and Cooper, 2013). In addition, interaction of an AC to an AKAP complex can lead to greater local increases in cAMP. For example, anchoring of an AC to AKAP5 was shown to sensitize the PKA phosphorylation of the anchored transient receptor potential cation channel subfamily V member 1 (TRPV1) channel to the effects of cAMP by shifting the forskolin dose-response curve by ∼100-fold (Efendiev et al., 2013). Similarly, coexpression of AC9 and the AKAP Yotiao was found to sensitize phosphorylation of the α subunit [potassium voltage-gated channel subfamily Q member 1 (KCNQ1)] of the slowly activating delayed rectifier potassium channel I(Ks) to lower concentrations of β-AR stimulation compared with cells expressing only AC9 or Yotiao (Li et al., 2012).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Plasma membrane and primary cilium cAMP signalosomes. A selection of the signalosomes that localize at the plasma membrane and primary cilium are illustrated. PKA-dependent phosphorylation is indicated in yellow. The cellular function that each signalosome controls is indicated by the gray arrows. Ub indicates ubiquitination. The red gradient indicates cAMP. The white halo surrounding the phosphodiesterases indicates reduced local concentration of cAMP. For details of the different protein components, see the main text. I_Ca, calcium current; I_Cl, cloride current; I_Na, sodium current; Hh, hedgehog protein; NHERF1, Na⁺/H⁺ exchanger regulatory factor 1; P, phosphate group; PCM1, pericentriolar material 1; SAP-97, synapse-associated protein 97.

In contrast to AC1–9, AC10 is not regulated by G proteins and is activated by bicarbonate (Chen et al., 2000; Bitterman et al., 2013) and ATP (Zippin et al., 2013) and is stimulated by Mn²⁺ and Ca²⁺ (Braun, 1975; Braun and Dods, 1975). Although originally identified as a soluble enzyme, soluble AC has been subsequently reported to be present at specific subcellular sites, including mitochondria, nucleus, centrioles, mitotic spindles, and midbodies (Zippin et al., 2003; Valsecchi et al., 2014; Wiggins et al., 2018). At these sites, AC10 is thought to provide a local source of cAMP that is triggered in response to changes in cellular metabolic state. A number of cellular functions have been associated to AC10-generated cAMP, including sperm capacitation and fertility (Esposito et al., 2004), ATP production (Acin-Perez et al., 2009; Di Benedetto et al., 2013), regulation of lysosomal acidification (Rahman et al., 2016), apoptosis (Ladilov and Appukuttan, 2014), and the downstream effects from the trafficking corticotropin-releasing hormone receptor (Inda et al., 2016).

C. Protein Kinase A

PKA is the most extensively studied effector of cAMP (Walsh et al., 1968; Taskén and Aandahl, 2004). In its inactive form, PKA is a heterotetramer composed of two regulatory (R) and two catalytic (C) subunits (Walsh et al., 1968; Corbin and Keely, 1977; Sowadski et al., 1985). Four types of R subunits (RIα, RIβ, RIIα, and RIIβ) and three C subunits (Cα, Cβ, and Cγ) have been described (Diskar et al., 2010). In the absence of cAMP, a dimer of R subunits binds to and blocks the activity of two C subunits. Binding of cAMP to the R subunits causes a conformational change that removes their inhibitory action on C (Taylor et al., 2005, 2013), allowing phosphorylation of protein targets preferentially at a [R/K][R/K/X]X[pS/pT] consensus motif (Taylor et al., 1990). The most significant difference between RI and RII isoforms is the autophosphorylation potential of the inhibitor site in RII subunits at S114, which affects the interaction between RII and C subunits and which, when phosphorylated, results in a reduced activation threshold for PKA (Taylor et al., 1990). PKA tetramers containing RI-type subunits are activated at lower concentrations of cAMP compared with those including RII, providing a mechanism for concentration-dependent selective activation of different PKA subsets. (Dostmann et al., 1990; Moll et al., 2008). Genetic studies demonstrate that PKA subunits are nonredundant (Brandon et al., 1995; Cummings et al., 1996; Hensch et al., 1998), and the different sensitivity to cAMP can contribute to signal compartmentalization, as it provides a means for differential activation of PKA isoforms based on the local cAMP concentration. This is achieved via anchoring of PKA to the multiscaffolding proteins AKAPs, which tether PKA at defined subcellular locations in proximity to specific PKA substrates (see below).

Other mechanisms involving PKA can further contribute to compartmentalize the action of this enzyme. For example, the E3 ubiquitin ligase really interesting new gene-H2 (RING-H2) protein praja2, a component of the ubiquitin proteasome system (UPS) (Yu et al., 2002), has been reported to form a stable complex with, and to be phosphorylated by, PKA. cAMP elevation followed by PKA activation leads to praja2 phosphorylation and activation of UPS-mediated degradation of R subunits. By reducing the R-to-C ratio, this positive-feedback mechanism can promote sustained PKA substrate phosphorylation in response to cAMP (Lignitto et al., 2011). Praja2 targets PKA to distinct cell compartments, including plasma membrane, cellular organelles, and perinuclear membrane, thus potentially providing a means for local regulation of PKA activity. The Praja2-dependent effect on the amplitude and duration of PKA signaling was shown to impact the phosphorylation level of the transcription factor CREB and nuclear gene transcription, significantly contributing to synaptic plasticity and long-term memory (Lignitto et al., 2011; Rinaldi et al., 2015). An additional mechanism involving ubiquitination of PKA C subunits has been reported (Eccles et al., 2016). ARHGAP36, a member of the Rho GTPase-activating protein family, was shown to directly bind PKA C via a pseudosubstrate motif and to potently inhibit PKA activity. The interaction of ARHGAP36 with PKA C quickly promotes their ubiquitination and targets them for endolysosomal degradation, leading to effective suppression of PKA signaling. ARHGAP36-mediated endolysosomal degradation of C subunits can therefore provide a mechanism for timely inactivation and selective degradation of a restricted pool of C subunits and, as described for Paja2-dependent ubiquitination of R subunits, may contribute to the compartmentalization of cAMP/PKA signaling (Eccles et al., 2016).

D. A-Kinase Anchoring Protein

The AKAPs play a fundamental role in signal compartmentalization, as they are responsible for tethering PKA holoenzyme to specific subcellular locations in physical proximity to PKA targets, increasing the specificity and efficiency of signaling (Lees-Miller and Anderson, 1989; Scott et al., 1990; Hulme et al., 2003; Lygren et al., 2007; Skroblin et al., 2010). All AKAPs identified so far (>30) share a short characteristic motif that serves to anchor PKA R subunits. The interaction between AKAPs and PKA occurs via binding of the dimerization-docking domain, located at the N terminus of the PKA R subunit, to an amphipathic α-helix, generally encompassing 14–18 residues, that is conserved among the AKAP family members. All four R subunit isoforms contain the conserved dimerization-docking domain (Banky et al., 1998; Gold et al., 2006). AKAPs include additional interaction sites that are responsible for their subcellular targeting and mediate interaction with other components of the cAMP pathway, including GPCRs (Malbon et al., 2004; Malbon, 2007), ACs (Piggott et al., 2008; Dessauer, 2009), phosphatases (Coghlan et al., 1995; Dodge-Kafka et al., 2010), PDEs (Dodge et al., 2001; Scott et al., 2013), as well as other kinases, including PKC, mitogen-activated protein kinase kinase (MEK)1/2, extracellular signal-regulated kinases (ERK)1/2, polycystic kidney disease (PKD), and others (Klauck et al., 1996; Tasken et al., 2001; Scott et al., 2013). The diverse subcellular targeting domains present in AKAPs recruit kinase activities to specific subcellular compartments, such as the plasma membrane, nucleus, mitochondria, endo/sarcoplasmic reticulum, the primary cilium, and other organelles (Langeberg and Scott, 2015; Torres-Quesada et al., 2017) (see below; illustrated in Figs. 1–3). Given the ability of each AKAP to bind and compartmentalize a different assortment of kinases, their substrates, and a variety of regulatory enzymes, a large diversity of PKA-focused signaling complexes (signalosomes) can be assembled at different subcellular locations where specific cAMP signaling is implemented locally (Greenwald and Saucerman, 2011).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Mitochondrial cAMP signalosomes. Multiple cAMP signalosomes localize to the mitochondria. Red gradients illustrate cAMP. The MOM is permeable to cytosolic cAMP, whereas the MIM is not. In the matrix, AC10 generates the second messenger locally. For the individual protein components of the signalosomes, see the main text. GK, glucokinase; P, phosphate group; PTPD1, protein tyrosine phosphatase D1; SAM50, sorting and assembly machinery component; StAR, steroidogenesis acute regulatory protein; TSPO, translocator protein; VDAC-1, voltage-dependent anion channel 1.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

cAMP signalosomes at the endoplasmic reticulum, nucleus, and centrosome. For clarity, multiprotein complexes illustrated here may not include all components that have been described in the literature. For identification of individual protein components, please refer to the main text. The red gradient illustrating cAMP highlights the fact that the nuclear membrane is permeable to cAMP, whereas the ER membrane is not. Lighter red halos surrounding PDEs indicate low local cAMP. P, phosphate group; SEC24, secretion 24.

RII subunits show higher affinity for AKAPs than RI subunits (Banky et al., 2003; Kritzer et al., 2012). This seems at odds with the idea of compartmentalized signaling. An untargeted, predominantly cytosolic and freely moving type I kinase that is activated at a lower cAMP concentration than required to activate the AKAP-bound type II PKA, (Taylor et al., 2012) would appear to nullify the advantage of having pools of cAMP and effector PKAII enzymes localised at at the same site. Multiple studies, however, have identified several type I selective AKAPs with high affinity for RI subunits—including sphingosine kinase interacting protein (SKIP) (Kovanich et al., 2010), small membrane AKAP (smAKAPs) (Means et al., 2011; Burgers et al., 2012, 2016), and G protein-coupled receptor 161 (Gpr161) (Bachmann et al., 2016). In addition, a number of so-called dual-specific AKAPs (e.g., AKAP1/2, ezrin, merlin, and AKAP11) can also interact with RI dimers, although typically with lower affinity than with RII dimers (Herberg et al., 2000; Pawson and Scott, 2010; Skroblin et al., 2010; Torres-Quesada et al., 2017). Experiments using fluorescence recovery after photobleaching showed that in cardiomyocytes, RI, similarly to RII, is largely anchored to AKAPs (Di Benedetto et al., 2008). In the same study, cAMP FRET-based sensors selectively targeted to RI- or RII-binding AKAPs were used to demonstrate that PKA type I and type II define distinct intracellular signaling compartments in these cells. Activation of β-AR generates a cAMP response that is selectively confined to AKAP-anchored PKA type II. This cAMP pool leads to phospholamban (PLN) and troponin I phosphorylation and is sensitive to rolipram (a PDE4 selective inhibitor). Conversely, prostaglandin 1 and glucagon stimulation of cardiac myocytes generates a cAMP signal that is selectively confined to AKAP-anchored PKA type I. This pool of cAMP is selectively sensitive to the PDE2A inhibitor erythro-9-Amino-β-hexyl-­α-methyl-9H-purine-9-ethanol (EHNA) and does not affect PLN or troponin I phosphorylation (Di Benedetto et al., 2008).

The study of the functional role of PKA-AKAP interactions has been facilitated by the development of several disruptor peptides and small molecules (Carr et al., 1992; Christian et al., 2011). For example, the peptide super A Kinase anchoring protein-in silico (SuperAKAP-IS) is 10,000-fold more selective for RII over RI (Gold et al., 2006), whereas the peptide RI anchoring disruptor shows 1000-fold higher affinity for RI over RII (Gold et al., 2006). Disruptor peptides that displace PKA specifically from individual AKAPs have been more difficult to generate. Some success has been achieved using a screening approach based on high-resolution structural information. In this study, several engineered RII subunit fragments that show some preference in disrupting interaction of RII with specific AKAPs were isolated. These peptides, named R_Select, were shown to have some selectivity for targeting the cognate AKAP in living cells, although their affinity of binding remains lower than endogenous full-length RII subunits, limiting for the moment their applicability (Gold et al., 2013). One recent example that successfully used PKA-AKAP–disrupting peptides to study the multifaceted role that AKAPs play in the regulation of cAMP/PKA signaling has been reported by Gorshkov et al. (2017), who observed cAMP gradients in developing hippocampal neurons, with higher cAMP levels in more distal regions of the axon compared with other regions of the cell; disruption of PKA-AKAP interactions using competing peptides enhanced the cAMP gradient and promoted axon elongation. The data suggest that, during development, AKAPs localized in the elongating axons organize a negative-feedback loop whereby AKAP-bound PKA activates a local PDE, resulting in reduced local cAMP levels (Gorshkov et al., 2017). Thus, depending on the different signaling elements that they assemble locally, AKAPs can play a complex role in the regulation of the cAMP/PKA pathway, in some cases facilitating and in others quenching signaling.

The localization of some AKAPs has been shown to be dynamic and to respond to changes in cellular conditions, resulting in translocation of the cAMP signalosome across the cell (Fields et al., 2016). One example is AKAP79/150, which targets both PKA and protein phosphatase 2B (PP2B)/calcineurin (CaN) to the plasma membrane to regulate the phosphorylation status of glutamate receptor 1. Activation of glutamate receptors results in the redistribution of AKAP79/150 and its anchored PKA-RII, but not of PP2B/CaN, from postsynaptic membranes to the cytoplasm in hippocampal slices (Smith et al., 2006).

Evidence is also emerging that alteration of AKAP-PKA interactions is linked to disease (Suryavanshi et al., 2018). Analysis of large datasets has uncovered genetic variation in AKAPs that correlates with susceptibility to a variety of conditions, including neurologic disorders, hypertension, cardiac arrhythmias, and certain cancers (Kammerer et al., 2003; Chen et al., 2007; Logue et al., 2014; Reggi and Diviani, 2017; Kjällquist et al., 2018). A systematic investigation of the properties of polymorphic variants in AKAPs that elicit changes of single amino acids in the helical region that binds to regulatory subunits type II of PKA has shown that a number of these amino acid substitutions disrupt PKA anchoring. In one example, a valine-to-methionine variant in the PKA-anchoring helix of AKAP18 was sufficient to reduce RII-binding affinity by over 8-fold, with functional consequences on the cAMP-dependent potentiation of L-type Ca²⁺ currents (Smith et al., 2018).

Although a large body of evidence supports a key role of AKAPs in directing context-specific cellular outputs of cAMP/PKA signals, one long-standing question is how, on PKA activation, dissociation of the C subunits from the R subunits dimer does not result in uncontrolled phosphorylation of PKA targets that are distal to the PKA-AKAP complex, thus nullifying the advantage of locally anchoring the kinase. A few recent studies have provided some possible answers. One hypothesis is that, for PKA to phosphorylate its substrates, complete dissociation of C from R subunits is not necessary. This idea was originally suggested after the observation that the rotational mobility of labeled C subunits complexed with dimeric R subunits was unchanged on addition of cAMP (Yang et al., 1995). More recent work by Smith et al. has provided fresh evidence in support of this hypothesis (Smith et al., 2013, 2017). Using pull-down experiments and native mass spectrometry of PKA holoenzyme in complex with AKAPs, the authors observed that, in the presence of cAMP, the AKAP-bound PKA holoenzyme remains largely intact (Smith et al., 2017). Using single-particle electron microscopy analysis of the AKAP-PKA complex, they observed that intrinsically disordered domains within the RII subunits allow a significant level of conformational plasticity within the AKAP-PKA complex (Smith et al., 2013). This molecular plasticity of the R subunit would provide a radius of motion to the associated C subunits of 15–25 nm, constraining phosphorylation to targets that are within the immediate vicinity of the AKAP-anchored PKA (Smith et al., 2013). An alternative hypothesis to explain how activation of PKA results in preferential phosphorylation of local targets is that C subunits that dissociate upon cAMP binding are rapidly recaptured by R subunits that are present at high density in the same neighborhood. A study revealed that the R subunits can be expressed in large molar excess of C subunits: up to 17-fold in the cells analyzed in that investigation (Walker-Gray et al., 2017). Clustering of AKAP complexes has also been reported—for example, at the plasma membrane (Mo et al., 2017)–which would effectively increase the local concentration of PKA R subunits. With increased local buffering capacity provided by free R subunits, the range over which the released C subunit can phosphorylate substrates would depend on its rate of diffusion and the rate of recapture.

In addition to AC species, AKAPs can also bind PDEs, thus providing the signaling nodes with machinery that allows them to both form and shape cAMP microenvironments (Sherpa et al., 2019).

E. Exchange Protein Directly Activated by cAMP

EPAC1 and EPAC2 proteins function as guanine nucleotide exchange factors for both Rap1 and Rap2 (de Rooij et al., 1998, 2000). Rap, a small Ras-like GTPase, belongs to the Ras superfamily of small G proteins, which cycle between an inactive GDP-bound state and an active GTP-bound state. Rap, whose activation by cAMP occurs independently of PKA, is activated by several extracellular stimuli and is involved in a large variety of cellular processes, including cell proliferation and adhesion, actin cytoskeleton dynamics, cell polarity, exocytosis, membrane protein recycling, and cell differentiation (Bos, 2006). Several studies support the involvement of EPAC in cell migration and proliferation in multiple types of cancer, making this protein a potential therapeutic target for cancer treatments (Wang et al., 2017; Kumar et al., 2018).

As for PKA, EPAC proteins are highly compartmentalized by different anchoring mechanisms. In response to cAMP, EPAC1 translocates and anchors to the plasma membrane, where it tethers to phosphatidic acid or to phosphorylated cytoskeleton proteins to induce its downstream effectors (Ponsioen et al., 2009; Gloerich and Bos, 2010; Gloerich et al., 2010; Consonni et al., 2012). EPAC2 is targeted to the plasma membrane after binding to activated Ras proteins independently of its interaction with cAMP (Li et al., 2006). In thyroid cells, EPAC1 and PKA simultaneously bind to radixin, an actin filament binding protein that belongs to the ezrin/radixin/moesin family, and form a functional cAMP-sensing sub–plasma membrane domain in which the two effectors act synergistically to mediate efficient proliferation in response to the mitogenic action of thyroid stimulating hormone (Hochbaum et al., 2011). Epac1 and Epac2 show distinct activation patterns in response to cAMP-raising stimuli, indicating that, similarly to PKA, differently localized EPAC isoforms are under the control of distinct pools of cAMP. For example, in cardiac myocytes, EPAC1 has been shown to localize to the plasmalemma, sarcoplasm, and nuclear/perinuclear region (Cazorla et al., 2009; Pereira et al., 2015). Stimulation of β₁-AR, but not β₂-AR, induces the recruitment of a β-arrestin-EPAC1 signaling complex at the plasma membrane, whereby it activates a pro-hypertrophic signaling cascade involving the small GTPases Rap2, Ras, and CaMKII, triggering histone deacetylase (HDAC) type 4 nuclear export and initiation of a pro-hypertrophic gene program (Berthouze-Duquesnes et al., 2013). By contrast, EPAC2 has been found to localize at the T-tubules (Pereira et al., 2015), and its activation by cAMP in response to β₁-AR stimulation leads to CaMKII-dependent diastolic sarcoplasmic reticulum (SR) calcium release and arrhythmias via a pathway that involves EPAC-phosphoinositide 3-kinases (PI3K)-protein kinase B (Akt)-nitric oxide synthase (NOS1) (Pereira et al., 2013, 2017).

F. Phosphodiesterases

The different PDE isoforms have distinct tissue distribution, and each individual cell typically expresses several PDE isoforms. The diversity of binding affinities and regulatory and targeting features of the various PDE isoforms provides a sophisticated system for unique regulation of amplitude and duration of cAMP signals at different sites (Houslay and Milligan, 1997; Houslay, 2001, 2010; Houslay and Adams, 2003; Houslay et al., 2005; Lugnier, 2006; Zaccolo et al., 2006; Conti and Beavo, 2007; Francis et al., 2011a). The large number of PDE isoforms arises from a combination of multiple genes, alternative splicing sites, and multiple transcription starting points. Each member shares a similar overall structure but is functionally unique. The catalytic domain located at the C-terminal region of each PDE enzyme is highly conserved (Conti and Beavo, 2007). The N terminus of these proteins is quite diverse and confers unique regulatory properties and targeting sequences that result in tethering of PDEs to specific subcellular locations (Bender and Beavo, 2006; Zaccolo, 2009; Kokkonen and Kass, 2017). The N-terminal region can carry post-translational modification sites. For example, phosphorylation of PDE3, PDE4, and PDE10 modulates their enzymatic activity and the ability of these enzymes to form complexes with other proteins. In addition, some PDEs can be regulated by cGMP (e.g., PDE2 and PDE3) (Zaccolo and Movsesian, 2007; Kokkonen and Kass, 2017). The UPS is also involved in the regulation of PDE degradation (e.g., PDE4D), which leads to upregulation of cAMP levels and to activation of PKA (Zhu et al., 2010). For long PDE4D isoforms, direct binding of the protein phosphatase calcineurin has been shown to inhibit proteasomal degradation of the PDE. Such degradation may occur in specific cellular compartments, including the nucleus, suggesting a complex regulatory mechanism in which both phosphatases and kinases are involved in the control of cAMP signaling (Zhu et al., 2010; Rinaldi et al., 2015). A detailed discussion of the regulatory mechanisms that control the activity of PDEs is beyond the scope of this article, but excellent reviews are available elsewhere (Conti and Beavo, 2007; Houslay, 2010; Cheepala et al., 2013; Baillie et al., 2019; Chao et al., 2019).

A large body of experimental evidence (Jurevicius and Fischmeister, 1996; Rich et al., 2001; Zaccolo and Pozzan, 2002; Mongillo et al., 2004; Rochais et al., 2004; Houslay et al., 2005, 2007; Lynch et al., 2006; Baillie, 2009; Houslay, 2010; Perera and Nikolaev, 2013; Lefkimmiatis and Zaccolo, 2014; Kokkonen and Kass, 2017) as well as computational studies (Oliveira et al., 2010; Saucerman et al., 2014; Chay et al., 2016; Yang et al., 2016; Lohse et al., 2017) indicate that PDEs play an important role in compartmentalization of cAMP signaling. However, how PDEs can accomplish this is still debated (Feinstein et al., 2012; Saucerman et al., 2014; Lohse et al., 2017). The heart of the problem is in the apparent contradiction between the high diffusivity of cAMP and the somewhat sluggish enzymatic activity of the PDEs (Koschinski and Zaccolo, 2017). The diffusion coefficient of cAMP has been experimentally calculated to be around 40 μm²/s, only approximately one order of magnitude lower than in water (Nikolaev et al., 2004; Richards et al., 2016). Based on the fast intracellular diffusion of cAMP, one would predict fast equilibration of cAMP levels across the cell, in apparent conflict with the notion of cAMP compartmentalization (Terrin et al., 2006; Lefkimmiatis and Zaccolo, 2014; Kokkonen and Kass, 2017). Based on the reported K_M and V_max values for PDEs, it is difficult to envisage how PDEs may be able to maintain the concentration of cAMP below the activation threshold of PKA even at basal cAMP levels, measured to be around 1 μM in a number of cell types (Beavo et al., 1974; Börner et al., 2011; Koschinski and Zaccolo, 2017), let alone contribute to compartmentalization of high concentrations of fast diffusing cAMP generated on hormonal stimulation. Inside a living cell, a number of factors may however come into play that slow down cAMP diffusion, including cAMP buffering (Chen et al., 1999; Lefkimmiatis et al., 2009), cytosol viscosity (Feinstein et al., 2012), and physical barriers (Richards et al., 2016). These factors may be particularly relevant within the restricted volume of a nanodomain, in which the stoichiometry of the molecules involved may be close to equimolar and enzymatic reactions do not follow Michaelis-Menten kinetics. In the confined volume of a nanodomain, local PDEs may be able to effectively reduce the concentration of free cAMP, keeping it below the level achieved a few tens of nanometers farther away (Musheshe et al., 2018; Chao et al., 2019).

Early evidence of a role for PDEs in compartmentalization of cAMP was provided by Hohl and Li, who showed that although approximately 45% of the cAMP generated upon isoproterenol stimulation of cardiac myocytes is found in the cellular particulate fraction, combining isoproterenol stimulation with inhibition of PDEs results in a decline to less than 20% of the proportion of total cAMP residing in the particulate fraction, in spite of a 3.4-fold total increase in cAMP. The authors concluded that this was a result of compartmentalized cAMP in the particulate fractions due to high PDE activity in the cytosol (Hohl and Li, 1991; Méry et al., 1991). cAMP-sensitive FRET-based sensors confirmed the central role of PDEs in shaping the intracellular cAMP nanodomains (Zaccolo and Pozzan, 2002; Nikolaev et al., 2004; Leroy et al., 2008; Surdo et al., 2017). PDE family–selective inhibitors are available and have contributed significantly to our understanding of the role of different PDEs in the compartmentalization of cAMP (Houslay et al., 2005; Lugnier, 2006; Francis et al., 2011b; Page and Spina, 2012; Kokkonen and Kass, 2017). For example, in rat cardiac myocytes, PDE3 and PDE4 enzymes show a very distinct localization, and inhibition of PDE4, but not inhibition of PDE3, results in significant increase in cAMP levels generated by β-agonists (Mongillo et al., 2004; Bedioune et al., 2017; Conti, 2017). Currently available pharmacological inhibitors do not discriminate between different isoforms within the same PDE family. To dissect the role of individual PDE isoforms in compartmentalization, an approach that has been used successfully relies on using isoform point mutants that are catalytically inactive. When overexpressed in the cell, the mutant PDE has a dominant negative effect by displacing the endogenous active enzyme from its anchor sites, effectively resulting in local ablation of cAMP hydrolytic activity (Lynch et al., 2007). Studies using this approach have shown that different isoforms of the same PDE family have a nonredundant and specific role in shaping local cAMP gradients, and their displacement can effectively disrupt local cAMP nanodomains (Baillie et al., 2003; Lynch et al., 2005; McCahill et al., 2005; Zoccarato et al., 2015). For example, expression of the catalytically inactive mutant PDE4D5-D556A in cardiac myocytes was used to demonstrate that β-AR activation recruits a β-arrestin–sequestered subpopulation of the PDE4D5 isoform that specifically regulates the phosphorylation of the receptor by PKA and induces a switch of the receptor from G_s to G_i coupling (Baillie et al., 2003). In another study, overexpression in neonatal rat cardiac myocytes of PDE2A2, PDE3A2, and PDE4D3 catalytically inactive mutants showed that displacement of endogenous PDE4D3 and PDE3A2 results in cardiac myocyte hypertrophy, whereas displacement of PDE2A2 counteracts hypertrophic growth induced by chronic β-AR stimulation (Zoccarato et al., 2015).

G. Phosphatases

The phosphorylation state of the PKA targets in a subcellular domain depends on the dynamic balance between kinase activity and phosphatase activity available at that specific site (Virshup and Shenolikar, 2009), and phosphoprotein phosphatases (PPPs) are the last checkpoint control for termination of the cAMP/PKA signaling cascade (Heijman et al., 2013). A key defining feature of all PPPs is that they are multimeric enzymes in which the catalytic subunit is associated with numerous regulatory subunits. These regulators are key determinants of PPP subcellular localization, substrate specificity, and enzymatic activity, conferring high complexity to the phosphatase system and the ability to exert a highly diversified local action (Virshup and Shenolikar, 2009). Dephosphorylation of PKA targets is accomplished by the three most abundant serine/threonine phosphatases–namely, protein phosphatase 1 (PP1), PP2A, and CaN (or PP2B) (Virshup and Shenolikar, 2009). In cardiac myocytes, for example, PP1, PP2A, and CaN account for >90% of the total phosphatase activity (Heijman et al., 2013).

PP1 is ubiquitously expressed. It includes three types of catalytic subunits that are encoded in different genes (PP1α, PP1γ, and PP1δ, also known as PP1β) (Herzig and Neumann, 2000). Each PP1α and PP1γ isoform has two splice variants (PP1α1 and PP1α2; PP1γ1 and PP1γ2), and different isoforms localize to distinct cellular regions. For example, in the heart, the myofibrillar fraction contains mainly PP1α (Herzig and Neumann, 2000), whereas PP1β predominantly localizes to the longitudinal SR, and PP1γ is enriched in the junctional SR (Aoyama et al., 2011). PP1 catalytic subunits can dimerize with more than 200 different regulatory subunits (Heroes et al., 2013). Notably, PP1 has been found to be associated with AKAPs, in which it can locally modulate the phosphorylation status of PKA targets within the same AKAP complex. For example, in cardiac myocytes, PP1 interacts with the AKAP Yotiao and regulates the PKA-mediated phosphorylation of the KCNQ1 subunit of the slowly activating potassium channel I(Ks) (Marx et al., 2002). PP1 interaction with AKAP11 (also called AKAP220) has been suggested to play a role in regulating the activity of glycogen synthase kinase 3b (Schillace et al., 2001), whereas anchoring of PP1 to AKAP1 has been proposed to be involved in efficient nuclear envelope assembly after mitosis (Steen et al., 2000). PP1 also interacts with AKAP7 (also called AKAP18), although indirectly. At this AKAP complex, PP1 binds Inhibitor-1 to promote its phosphorylation by PKA, and the net effect of this later phosphorylation event is to promote local inhibition of PP1c (Singh et al., 2011).

PP2A can form a dimer containing a catalytic (PP2A-C) and a structural scaffold (PP2A-A) subunit or a trimer consisting of scaffold, catalytic, and regulatory subunits. The scaffold and catalytic subunits consist of two isoforms each (PP2A-Aα and PP2A-Aβ; PP2A-Cα and PP2A-Cβ), whereas the regulatory PP2A subunits are organized into four families (PP2A-B, PP2A-B′, PP2A-B′′, and PP2A-B′′′), most of which contain multiple isoforms with different splice variants (Shi, 2009; DeGrande et al., 2013; Heijman et al., 2013). Cells can assemble more than 200 distinct protein complexes containing different combinations of PP2A-A, PP2A-B, PP2A-C, and other proteins (Virshup and Shenolikar, 2009). Although regulatory subunit α and β are mainly cytosolic, the γ subunit is enriched in the cytoskeletal fraction, conferring different subcellular localizations to the PP2A holoenzyme (Strack et al., 1998). In addition, PP2A localization can be mediated by interaction with AKAPs. For example, in striated muscle, PP2A is part of a complex organized by muscle-specific A-kinase anchoring protein (AKAP6 or mAKAP) and also includes multiple other signaling components (Dodge-Kafka et al., 2010).

CaN is a Ca²⁺-dependent phosphatase consisting of one catalytic (CNA) and one regulatory (CNB) subunit. CNA is expressed by three genes (CNAα, CNAβ, and CNAγ), whereas CNB is expressed by two genes (CNBα and CNBβ) (Shi, 2009). Regulation of CaN activity toward specific substrates is accomplished primarily through protein-protein interactions, which notably include AKAP5 (AKAP79/150) (Dell'Acqua et al., 2002) and AKAP6 (mAKAP) (Li et al., 2010). For example, in the hippocampus, AKAP5-bound CaN modulates the phosphorylation state of Ser845 of the glutamate A1 (GluA1) subunit of the amino-3-hydroxy-5-methylisoxazole-4-propionate (AMPA) receptor and modulates synaptic plasticity (Woolfrey and Dell'Acqua, 2015). In the heart, the localization of CaN at the AKAP6 complex appears to be involved in efficient dephosphorylation of the nuclear factor of activated T-cells (NFAT), leading to increased gene transcription (Li et al., 2010).

A study comparing the spatiotemporal dynamics of PKA target phosphorylation in the bulk cytosol and at the mitochondrial outer membrane (MOM) using FRET reporters for cAMP and PKA activity showed that the bulk cytosol contains significantly higher phosphatase activity than the MOM (Burdyga et al., 2018). Based on this observation, the authors suggest that the excess of phosphatases in the cytosol may play a role in the compartmentalization of cAMP/PKA signaling by efficient suppression of accidental phosphorylation of PKA targets that may occur outside the designated signaling domain.

A. Plasma Membrane

The plasma membrane has been shown to host multiple cAMP signaling domains, which are often organized by large multiscaffolding protein complexes. The specific localization at the plasma membrane of individual G_s-coupled receptors affects the subcellular compartment that experiences the rise in cAMP and determines, at least in part, the functional response to receptor activation. This is supported, for example, by experiments in which membrane rafts are disrupted via cholesterol depletion (Calaghan et al., 2008; DiPilato and Zhang, 2009; Wright et al., 2018). How the arrangement of the receptors at the plasma membrane can affect compartmentalized cAMP signaling is not completely understood, but a contributing factor appears to be the coupling of a given receptor with a specific AC.

B. Primary Cilium

The primary cilium, a nonmotile, microtubule-based protrusion of the plasma membrane present in most cells in the body, functions as a highly specialized sensory structure with a primary role in cell fate specification and cell proliferation (Ingham et al., 2011). In normal proliferating cells, it is transiently observed in G1 phase and is reabsorbed when the cell enters the cell cycle (Plotnikova et al., 2008). The primary cilium uses cAMP to transduce multiple signals, and given the spatially confined nature of this structure, it is not surprising to find that it hosts compartmentalized cAMP/PKA signaling modules (Moore et al., 2016; Mukherjee et al., 2016; Jiang et al., 2019; Sherpa et al., 2019).

C. Mitochondria

Mitochondria have been increasingly recognized as organelles with tightly regulated cAMP signaling. Indeed, every mitochondrial compartment, including the MOM, the mitochondrial intermembrane space (MIS), and the mitochondrial matrix, harbors one (or more) spatially and functionally distinct cAMP signalosome. cAMP signaling at the mitochondria contributes to the extreme adaptability of these organelles in response to changes in metabolic demands (Carlucci et al., 2008; Reggi and Diviani, 2017) via direct modulation of a variety of functions, including mitochondria biogenesis and morphology, gene expression, metabolism, steroidogenesis, and survival (Zhang and Xu, 2016). Several components of the cAMP signaling pathway, including source of cAMP, effectors, terminators, and targets, have been identified in mitochondria (Monterisi and Zaccolo, 2017; Rinaldi et al., 2018).

PKA R and C subunits were isolated from purified mitochondria over 30 years ago (Schwoch et al., 1990; Papa, 1999). Mass spectrometry analysis of the mitochondrial phosphoproteome identified more than 100 potential PKA targets in the mitochondrial phosphopeptide pool (Grimsrud et al., 2012). Most of these phosphorylation sites are yet to be validated, and it is therefore possible that the mitochondria may harbor additional cAMP signalosomes that remain to be characterized (Monterisi and Zaccolo, 2017). At least five different AKAPs have been found to associate with the MOM: AKAP1 (also known as AKAP121, AKAP149, and spermatid (S)-AKAP84, depending on the species) (Ma and Taylor, 2002; Affaitati et al., 2003), AKAP2 (Wang et al., 2001), Rab32 (Alto et al., 2002), peripheral-type benzodiazepine receptor–associated protein 7 (PAP7) (Li et al., 2001), and Wiskott-Aldrich syndrome protein verprolin homologous (WAVE) (Danial et al., 2003) (Fig. 2). The presence of several AKAPs at the MOM implies that this structure hosts several cAMP signaling domains with different roles in the regulation of mitochondrial function.

D. Golgi Apparatus

Several components of the cAMP signaling pathway have been found to localize at the Golgi, and several studies support compartmentalized cAMP/PKA signaling at this organelle (Godbole et al., 2017; Nash et al., 2018). For example, Yang et al. showed that in cardiac myocytes, where the Golgi encircles or is adjacent to the nucleus, activation of β₁-AR stimulates Ca²⁺ release from this organelle. Release of Ca²⁺ from the Golgi seems to require activation of both EPAC and PKA and results in enhanced trafficking of proteins to the cell surface. The perinuclear release of Ca²⁺ appears to be distinct from the release of Ca²⁺ from the SR, an event that is also under the control of β₁-AR/cAMP and that regulates myocyte contraction. The amount of agonist required to induce release of Ca²⁺ from the Golgi is larger than the amount that promotes Ca²⁺ release from the SR. However, such difference is obliterated on inhibition of PDE3 or PDE4 (Yang et al., 2015), two enzymes that have also been shown to be associated with the Golgi fraction (Pooley et al., 1997; Geoffroy et al., 2001), but not by inhibition of PDE2 (Yang et al., 2015), which instead affects the amplitude of the Ca²⁺ transient in response to catecholamines (Mongillo et al., 2006). These findings are consistent with compartmentalization of cAMP signaling in the microenvironment surrounding the Golgi apparatus attained by local degradation of cAMP by PDE3 and PDE4 (Yang et al., 2015). Yang et al. did not investigate whether an AKAP-dependent signalosome is involved in the regulation of Ca²⁺ release from the Golgi. Interestingly, AKAP9 [AKAP350/AKAP450/centrosome and golgi localized protein kinase N-associated protein (CG-NAP)/Yotiao] has been found to localize to the Golgi apparatus in cerebellar granule cells (Shanks et al., 2002), where it brings PKA-RIIβ in close proximity to the IP₃ receptors (Collado-Hilly and Coquil, 2009). Several studies, predominantly in cell lines, have shown that Ca²⁺ efflux from the Golgi is mediated by IP₃ receptors (Pinton et al., 1998), and PKA is known to phosphorylate and activate IP₃ receptors and enhance Ca²⁺ release from the ER (Betzenhauser and Yule, 2010). However, in cardiac myocytes, Ca²⁺ release from the Golgi appears to be mediated by ryanodine receptors rather than IP₃ receptors (Yang et al., 2015).

E. Centrosome

The centrosome is the major microtubule-nucleating organelle in animal cells and is usually composed of a pair of centrioles surrounded by a protein matrix (Doxsey et al., 1994). In mitosis, the centrosome ensures fidelity of chromosome segregation through the formation of a bipolar spindle, whereas in differentiating cells, the centrosome acts as a platform for the formation of the primary cilium (Joukov and De Nicolo, 2019). Centrosomes contain all the components of the cAMP signaling pathway, including PKA (Nigg et al., 1985), at least two AKAPs (Schmidt et al., 1999; Takahashi et al., 1999; Witczak et al., 1999; Diviani et al., 2000), PDEs, and targets of PKA regulation, such as dynein, MAP2, and centrin, which are necessary for spindle formation, chromosome movement, and cytokinesis (Diviani and Scott, 2001).

F. Endo/Sarcoplasmic Reticulum

G. Nucleus

Multiple components of the cAMP signaling cascade have been shown to localize to the nucleus and/or nuclear membrane, including GPCRs, G proteins, ACs, AKAPs, and PKA (Vaniotis et al., 2011).