3′-5′-Cyclic adenosine monophosphate/protein kinase A (cAMP/PKA) signalling is activated by different extracellular stimuli and mediates many diverse processes within the same cell. It is now well established that in order to translate into the appropriate cellular function multiple extracellular inputs, which may act simultaneously on the same cell, the cAMP/PKA signalling pathway is compartmentalised. Multimolecular complexes are organised at specific subcellular sites to generate spatially confined signalosomes, which include effectors, modulators and targets of the pathway. In recent years, it has become evident that mitochondria represent sites of compartmentalised cAMP signalling. However, the exact location and the molecular composition of distinct mitochondria signalosomes and their function remain largely unknown. In this review, we focus on individual components of the cAMP/PKA signalling pathway at distinct mitochondria subdomains represented by the outer and inner mitochondrial membranes, the intermembrane space and the matrix, highlighting some of the questions that remain unanswered.
The second messenger 3′-5′-cyclic adenosine monophosphate (cAMP) mediates a plethora of intracellular functions, most of which rely on the activation of the cAMP-dependent protein kinase A (PKA). PKA is a multitarget enzyme that consists of two regulatory (R) and two catalytic (C) subunits held in a tetramer. When cAMP binds to the R subunits, the C subunits dissociate and phosphorylate serine and threonine residues in a large number of target proteins. In most cell types, cAMP is generated in response to activation of a multiplicity of Gs-coupled receptors and measurements in intact cells show that cAMP can diffuse at high speed from the site of generation at the plasma membrane to the bulk cytosol (diffusion constant values between 50 and 700 µm/s2 have been reported, depending on the cell type) [1,2]. Another aspect of this signalling pathway is that PKA can promote opposing effects on some intracellular processes and functions. For example, in the heart, PKA-dependent phosphorylation of the L-type Ca2+ channel and the ryanodine receptor increases the amplitude of Ca2+ transients and promotes contraction; on the other hand, PKA-mediated phosphorylation of phospholamban (PLN), a negative regulator of sarco/endoplasmic reticulum Ca2+-ATPase (SERCA), induces Ca2+ re-uptake in the sarcoplasmic reticulum (SR), resulting in relaxation. PKA also phosphorylates cardiac troponin I, resulting in reduced affinity of troponin C for Ca2+, which promotes relaxation during diastole . Considering that PKA is not the only effector of cAMP (which also activates the guanine nucleotide exchange protein exchange protein activated by cAMP (EPAC) — isoforms 1 and 2 —  and the cyclic nucleotide gated ion channels ), one wonders how the cAMP signal is transduced specifically from a given receptor to the right target, to achieve the correct functional outcome.
cAMP and microdomains
Research in the past decade has established that cAMP signalling operates via engagement of spatially distinct domains within the cell, each of which leads to the activation of a selected subset of effectors and targets. These domains undergo local regulation, resulting in the generation of a confined increase in cAMP, and have been described for many intracellular compartments including nucleus, SR, plasma membrane and mitochondria . These domains are defined by the presence of ‘signalosomes’, multiprotein complexes that group a distinct array of molecular components that co-operate to integrate the cAMP signal locally. A typical ‘signalosome’ includes PKA, an A-kinase anchoring protein (AKAP) that binds PKA , a PKA target, and modulators: phosphodiesterases (PDEs) that degrade cAMP, and phosphatases, which counteract the PKA-dependent phosphorylation . A classic example of ‘signalosome’ in the cardiovascular system is located at the SR and includes PLN and the PKA-anchoring protein AKAP18δ; PKA-dependent phosphorylation of PLN removes its inhibitory effect on the SERCA, resulting in faster refilling of the SR with calcium from the cytosol. phosphodiesterase 3A (PDE3A), another component of this ‘signalosome’, has been shown to play a key role in cAMP signalling at the SR by controlling the local level of the second messenger . In doing so, PDE3A regulates baseline myocardial contractile function by modulating PKA-dependent PLN phosphorylation and thus both SR Ca2+ loads and Ca2+ transients amplitude; it has though no effect on L-type Ca2+ currents, which are also under control of PKA, but are part of a distinct domain at the plasmalemma, which does not appear to be regulated by PDE3A. The local effect of PDE3A at the SR is a clear example of how the local modulation of cAMP signals is translated into activation of a distinct pathway that leads to selective activation of a specific downstream function.
Mitochondrial cAMP–PKA signalling
cAMP/PKA signalling regulates several aspects of mitochondrial physiology, including mitochondrial morphology, mitochondria-mediated cell death and mitochondrial respiration. Over the past few years, it has become increasingly evident that these organelles host multiple cAMP/PKA signalosomes, and multiple components of the PKA signalling pathway have been identified within different suborganelle compartments (see Figure 1). Catalytic and regulatory subunits of PKA have been isolated from purified mitochondria , and it is reported that mitochondria contain several PKA substrates. These include, among others, cytochrome c oxidase and NDUSF4 involved in mitochondria respiration and localised at the inner mitochondrial membrane (IMM); Bcl2 associated death promoter and dynamin related protein 1 (Drp1), which are key players in apoptosis and mitochondrial fission, respectively, and localise at the outer mitochondrial membrane (OMM); ChChd3, a coiled-coil protein that localises at the intermembrane space (IMS) and is involved in the regulation of cristae morphology [11–14].
The mitochondrial signalosome.
An analysis of mitochondrial phosphopeptides by mass spectrometry identified over 100 potential PKA targets among proteins that are resident in the mitochondria . Most of these phosphorylations have not been investigated yet; however, these findings suggest that the rate of reversible phosphorylation at these organelles may be quite high, and open intriguing questions regarding the nature of the stimuli activating mitochondrial signalosomes, how the local signalling is regulated and integrated with signalling at other intracellular sites and whether some targets can shift from one compartment to another after phosphorylation.
Many mitochondrial AKAPs have also been identified. AKAP1 (or its isoforms AKAP121, AKAP149 and S-AKAP84, depending on the species) was the first to be discovered; its N-terminus domain contains a functional tubulin-binding motif essential for mitochondrial localisation . The importance of this AKAP in mitochondrial physiology is demonstrated by the fact that a point mutation that disrupts its binding to PKA associates with increased apoptotic cell death . Conversely, it has been shown that the overexpression of wtAKAP1, via recruitment of PKA to the OMM and increased Drp1 phosphorylation at Serine 637, traps Drp1 in large, slowly recycling polymers, which promote mitochondrial fusion and protect the cell from apoptosis .
D-AKAP2, which can anchor both RII and RI subunits of PKA, has been shown with immunostaining to co-localise with cytochrome c , suggesting that this AKAP localises in the IMS although the absence of detectable D-AKAP2 in the mitochondria at the neuromuscular junction indicates that this localisation may be cell type-dependent . D-AKAP2 can also interact via its C-terminus with PDZ domain proteins, like NHERF1, therefore promoting the formation of macromolecular complexes.
The AKAP Rab32 was originally found to be involved in intracellular trafficking of mitochondria ; in further studies, it was shown to localise at the interface between mitochondria and endoplasmic reticulum and has been suggested to be involved in the modulation of endoplasmic reticulum (ER) calcium handling. Overexpression of Rab32 in HeLa cells resulted in enhanced association of PKA with ER membranes . Conversely, the expression of a mutant form unable to bind PKA induced a relocalisation of the kinase into the cytosol. Moreover, Rab32 was found to promote Drp1 phosphorylation and to delay apoptosis, suggesting a role in the regulation of cell death . Another AKAP, named SKIP, has been shown to localise to the mitochondrial IMS, where it is thought to interact with PKA type I and to phosphorylate ChChd3 .
Many studies have dealt with the source of cAMP that activates mitochondrial PKA ‘signalosomes’. Although earlier reports suggested that cAMP generated by plasma membrane adenylyl cyclases may be transported inside mitochondria , the mechanism has remained undefined, and this model has been refuted by recent studies showing that, while the OMM is highly permeable to cAMP and allows diffusion of the second messenger into the IMS, the IMM is impermeable to it . As targets of PKA have been identified in the mitochondrial matrix, it has been hypothesised that a source of cAMP must be present in this compartment. In support of this hypothesis, recent studies have reported evidence of a calcium and bicarbonate-regulated soluble adenylyl cyclase (sAC) localised inside the matrix and involved in the regulation of oxidative phosphorylation [26,27]. However, the current view on cAMP signalling in the matrix remains a matter of controversy. Although reports have suggested the presence of PKA in this compartment on the basis of measurements of adenosine triphosphate (ATP) consumption upon activation or inhibition of the kinase , experiments using fluorescence resonance energy transfer (FRET)-based biosensors targeted to specific mitochondrial subdomains failed to detect PKA activity in the matrix . It should be noted that in many studies aimed at monitoring PKA activation in specific mitochondrial subcompartments, synthetic analogues of cAMP were used. A recent investigation indicates that different membrane permeable cAMP analogues, such as brome-cAMP or 8-CPT-cAMP, or treatment with PKA inhibitors, are ineffective in altering respiration in mitochondria from pig hearts . These findings argue against the hypothesis that oxidative phosphorylation is mediated or enhanced by cAMP/PKA activity within mitochondria, as previously reported . The authors also warn about the use of these lipophilic compounds as they are found to induce non-specific effects, including inhibition of ATP synthesis, which are not related to PKA activity .
Another recent study suggests the existence of a mitochondrial signalling pathway localised inside the matrix and involving Epac1 as an effector of cAMP. The study reports that in rat cardiac myocytes, cAMP generated by sAC is responsible for the activation of Epac1, and this protects mitochondria from calcium overload, mitochondrial permeability transition and, ultimately, cell death . The authors use two FRET-based reporters, one cytosolic and one targeted to the mitochondrial matrix, and stimulate the cells with either (to activate sAC) or forskolin (FSK) (which activates the transmembrane adenylyl cyclase). They find that the response to FSK is higher in the cytosol when compared with the mitochondria, whereas elicits a cAMP increase significantly larger at the mitochondria than in the cytosol, and conclude that the second messenger must therefore be generated locally within the organelles .
Role of PDEs in mitochondrial signalosomes
PDEs are recognised to be key modulators of local cAMP signalling  and have been found to be associated with multiple cAMP/PKA signalosomes. PDEs are classified in 11 families with different substrate specificity and kinetic properties. Some of them are selective for either cAMP (PDE4, 7, 8) or cyclic guanosine triphosphate (PDE5, 6, 9), whereas the others hydrolyse both cyclic nucleotides (PDE1, 2, 3, 10, 11). Different PDE families include multiple isoforms that are either encoded by different genes or are generated through alternative splicing . Different isoforms of the same PDE can give a diverse contribution to the compartmentalised signalling by binding to different intracellular subcompartments . Studies have shown that PDEs are found in both the soluble and the particulate fractions of tissue homogenates, indicating that they can anchor to organelles or membranes, in some cases via interaction with AKAPs.
PDEs play a major role in regulating cAMP compartmentalisation, by restricting the propagation of the second messenger outside individual domains and by regulating cAMP levels within each domain. The general idea of PDEs as means to restrict intracellular diffusion of cyclic nucleotides originated for the first time after fractionation analysis in canine ventricular myocytes  and subsequent measurement of the total and particulate cAMP content generated upon incubation with the non-selective PDE inhibitor isobutyl metyl xanthine (IBMX). Years later, FRET imaging experiments in intact cells in rat ventricular myocytes challenged with β-adrenergic receptor agonists, allowed us to directly visualise the accumulation of cAMP in the well-defined sarcomeric structures of these cells, and confirmed that compartmentalisation of the second messenger was abolished on application of IBMX . Many subsequent studies using selective pharmacological inhibitors of PDEs or a dominant-negative approach where displacement of endogenous PDE isoforms is achieved by overexpression of their catalytically inactive mutant have established that individual PDEs play distinct functional roles in the cell. For example, in cardiac ventriculocytes, it has been shown that PDE2 controls a cAMP pool generated specifically upon β3-AR stimulation at the membrane, whereas PDE4 is mainly associated with β2-receptor , or that PDE3A, but not PDE3B, protects the heart from ischemia/reperfusion injury .
The first evidence of the presence of PDEs at mitochondria dates back to 1982, when Cercek and Houslay, using cell fractionation experiments and PDE activity assays, found two different active enzymes at these organelles, associated with the OMM and the IMM, respectively . As mentioned before, the IMM has been shown to be completely impermeable to cAMP; but the finding that cAMP can be generated inside the matrix by sAC leads to the assumption that there must be a PDE that can degrade the second messenger at this site. Evidence for the existence of a PDE2A2 resident in the mitochondrial matrix in liver and brain extracts was recently reported . Targeting of PDE2A2 to the matrix was shown to rely on the amphipathic helix formed by 19 amino acids at the N-terminus of the enzyme, and selective inhibition of PDE2 was linked to enhanced oxidative phosphorylation and ATP synthesis .
A mitochondrial PDE — named Prune — was subsequently discovered in Drosophila . The authors used a bimolecular fluorescence complementation approach and co-expressed Prune fused to the N-terminal half of a green fluorescence protein (GFP) molecule, together with a known mitochondrial matrix protein — superoxide dismutase 2 (SOD2) — fused to the C-terminal half of GFP; they found that Prune–nGFP complemented SOD2–cGFP to reconstitute functional GFP, suggesting its localisation in the matrix. To test the involvement of Prune in the control of mitochondrial cAMP signalling, a genetically encoded FRET reporter containing a sequence of SOD2 was expressed in Drosophila-cultured cells. The results showed that upon simultaneous application of forskolin and IBMX, there is an increase in cAMP both in the cytosol and the mitochondrial fraction. These findings support the presence of Prune in the mitochondrial matrix, and the authors speculate that it may represent a non-mammalian counterpart of the PDE2A described in mammalian cells . One aspect that is overlooked in these studies is the fact that the IMM is impermeable to cAMP generated at the plasma membrane upon forskolin stimulation, and it is difficult to understand how a signal generated outside the organelles can be detected inside the matrix.
PDE2 is not the only PDE localised in mitochondria. Two isoforms — A and B — of PDE8 were found to be highly expressed in testicular Leydig cells. Interestingly, immunofluorescence staining with specific antibodies showed that, while PDE8B is distributed in the cytosol, PDE8A co-localises with cytochrome P450 and ATP synthase, which are known mitochondrial markers , indicating a selective enrichment of PDE8A to or very near to the mitochondria. Whether PDE8 isoforms are present at these organelles in other cell types is presently unknown.
Mitochondria are important organelles not only for their role as energy factories but also for their implication in other vital processes including cell death, differentiation and aging, and many pathological conditions are associated with deregulation of mitochondria dynamics. Compartmentalisation of cAMP/PKA signalling is now recognised to underpin the ability of this pathway to specifically modulate multiple cellular functions with high fidelity, and an increasing body of evidence supports the existence of multiple cAMP/PKA signalling domains localised at mitochondria. While several studies have described cAMP, PKA and AKAPs in the matrix, IMS, IMM and OMM, less is known about PDEs at these organelles. Important questions remain unanswered about the localisation of these signalosomes at each mitochondrial subcompartment, the exact identity of their individual molecular components and the specific mitochondrial function they regulate. Uncovering of these details in the future may open the way to the development of novel therapeutics with potential applications in a wide variety of pathological conditions where mitochondrial dysfunction plays a role.
A-kinase anchoring protein
Bcl2 associated death promoter
brome cyclic AMP
3′-5′-cyclic adenosine monophosphate
cyclic guanosine triphosphate
dynamin related protein 1
Exchange Protein Activated by cAMP
fluorescence resonance energy transfer
green fluorescence protein
isobutyl metyl xanthine
inner mitochondrial membrane
outer mitochondrial membrane
Protein Kinase A
soluble adenylyl cyclase
sarco/endoplasmic reticulum Ca2+-ATPase
superoxide dismutase 2
transmembrane adenylyl cyclase
The present study was supported by the British Heart Foundation [PG/10/75/28537 and RG/12/3/29423] and the BHF Centre of Research Excellence, Oxford [RE/08/004] to M.Z.
The Authors declare that there are no competing interests associated with the manuscript.