Mitochondria are highly dynamic organelles comprising at least three distinct areas, the OMM (outer mitochondrial membrane), the IMS (intermembrane space) and the mitochondrial matrix. Physical compartmentalization allows these organelles to host different functional domains and therefore participate in a variety of important cellular actions such as ATP synthesis and programmed cell death. In a surprising homology, it is now widely accepted that the ubiquitous second messenger cAMP uses the same stratagem, compartmentalization, in order to achieve the characteristic functional pleiotropy of its pathway. Accumulating evidence suggests that all the main mitochondrial compartments contain segregated cAMP cascades; however, the regulatory properties and functional significance of such domains are not fully understood and often remain controversial issues. The present mini-review discusses our current knowledge of how the marriage between mitochondrial and cAMP compartmentalization is achieved and its effects on the biology of the cell.
Mitochondria are marvels of compartmentalization, and this property is intimately connected with their function. Compartmentalization of these organelles is achieved thanks to two structures with distinct characteristics: the IMM (inner mitochondrial membrane) and the OMM (outer mitochondrial membrane). For instance, thanks to the IMM, which is notably impermeable and separates the matrix from the external environment, mitochondria are able to generate the proton electrochemical gradient that drives ATP production and also protect their genetic material (mtDNA) from external insults . In contrast with the IMM, the OMM is permeable to solutes up to 5 kDa  and allows communication of the IMS (intermembrane space), the compartment defined by the IMM and OMM, with the cytosol. On the other hand, the OMM is tight enough to keep cytochrome c segregated in the IMS and released only when the OMM integrity is compromised in response to apoptotic stimuli, resulting in cellular death .
The distinct permeability of the OMM and IMM and the presence of specific mitochondrial transporters determine the intramitochondrial availability of cytosolic molecules, including the two second messengers calcium (Ca2+) and cAMP. In contrast with Ca2+, which permeates the IMM thanks to a transporter protein called the MCU (mitochondrial calcium uniporter) [4,5], extramitochondrial cAMP can cross the OMM and reach the IMS; however, the IMM appears to be impermeable to cAMP, therefore cytosolic messenger cannot reach the mitochondrial matrix [6–8]. This differential permeability to cAMP could promote the existence of submitochondrial compartments with distinct cAMP concentrations, a feature that is the hallmark of compartmentalized cAMP cascades or microdomains. Thanks to the development of novel tools based on FRET that allow the direct measurement of cAMP and its actions in different cellular compartments [7–9], we are beginning to appreciate the complexity of localized cAMP signalling. It is becoming increasingly clear that mitochondria could host multiple cAMP microdomains. However, the distinct regulatory and functional characteristics of these pathways as well as their effectors and targets remain largely unknown. In the present article, I provide an overview of our knowledge of the mitochondrial cAMP domains, highlighting the debated points and discussing the cAMP targets for each mitochondrial compartment.
The cAMP signalling machinery
cAMP is generated by a set of enzymes called adenylate cyclases. Ten distinct genes encode nine TmACs (transmembrane adenylate cyclases) and one sAC (soluble adenylate cyclase), all sharing the ability to generate cAMP, but differing in their modes of regulation . Ordinarily, TmACs are activated by the GPCR (G-protein-coupled receptor)–heterotrimeric G-proteins axis in response to extracellular stimuli. In contrast with TmACs, sAC is insensitive to G-proteins and responds to intracellular cues such as bicarbonate, Ca2+ and ATP [11–13]. Additionally, sAC is the only adenylate cyclase proposed to associate with mitochondria . It is important to note, however, that the cytosol-facing mitochondrial compartments (IMS and OMM) are reached by cAMP produced at the plasma membrane, suggesting a functional link between TmACs and these domains.
cAMP signals are transduced via three receptor proteins: the CNG (cyclic-nucleotide-gated) channels, Epacs (exchange proteins directly activated by cAMP) and PKA (protein kinase A). Both Epac and PKA have been reported to associate with mitochondria; however, to date, only PKA was shown to affect mitochondrial function and is therefore considered the main effector of mitochondrial cAMP signalling [15,16]. PKA in its inactive form is a tetramer made by two PKA-Cs (catalytic PKA subunits) and two PKA-Rs (regulatory PKA subunits). Binding of cAMP to the PKA-Rs releases the constitutively active PKA-Cs and allows them to phosphorylate their targets . Three genes encoding PKA-C have been characterized (Cα Cβ and Cγ) and all of them can complex with the four types of PKA-Rs (RIα, RIβ, RIIα and RIIβ) to form cAMP-responsive heterotetramers. Despite its soluble nature, PKA can be targeted to specific subcellular compartments (including mitochondria) via a family of proteins called AKAPs (A-kinase-anchoring proteins) .
Intracellular cAMP concentrations are regulated and shaped by a superfamily of proteins called PDEs (phosphodiesterases) [19,20]. There are 11 families of PDEs and eight of them can degrade cAMP. Each of these eight families contains multiple genes, some of which can undergo alternative splicing, resulting in a staggering number of functional enzymes . PDE-dependent cAMP degradation prompts PKA inactivation allowing the dephosphorylation of its targets via the counteracting protein phosphatases, effectively terminating the cAMP signalling cascade .
In its simplest form, i.e. generation of messenger in response to extracellular stimuli followed by activation of its effectors and termination by PDEs, the cAMP signalling cascade is not sufficient to explain how this molecule achieves its specificity of action. The answer to this paradox is that cAMP regulates many diverse cellular functions thanks to the spatial confinement of elements of its pathway to form segregated functional domains (microdomains) that are under the control of specific cAMP signals . Although several mitochondrial actions are regulated by the cAMP/PKA axis, the precise compartmentalization of this signalling cascade within these organelles remains a controversial and active field of study. In the present article, I briefly present the existing evidence in support of compartmentalized cAMP domains in the three mitochondrial compartments.
Whether or not the innermost mitochondrial compartment contains a functional cAMP signalling circuit is highly debated. It has been proposed that the matrix contains all of the components of the cAMP signalling cascade including a cAMP source (sAC) [6,14], an effector (PKA)  and a PDE . However, the experimental evidence in support of a functional cAMP microdomain contained in the matrix is not always unequivocal and the lack of molecular tools to directly measure cAMP and PKA activity in the matrix has been, until recently, a significant obstacle [7,8,26].
The first attempts to measure cAMP in the mitochondria using FRET-based cAMP probes concluded that cAMP produced in the cytosol could reach the OMM as well as the mitochondrial matrix . These findings were challenged by a subsequent biochemical study proposing that, in isolated mitochondria, cAMP could not enter the matrix, but was locally produced by a matrix-resident sAC . These findings were confirmed recently by two independent studies that used cAMP-sensitive FRET sensors tightly targeted to the matrix and showed that cytosolic cAMP cannot reach this compartment [7,8]. These studies also provided evidence for sAC-dependent cAMP production in response to bicarbonate and Ca2+, both in HeLa and primary rat neonatal cardiac myocytes [7,8]. Another interesting conclusion was that, although the IMM is not permeable to cAMP, during MPT (mitochondrial permeability transition), a phenomenon during which the IMM forms pores allowing non-specific flux of molecules , cAMP is free to diffuse into and out of the mitochondrial matrix . MPT can be the irreversible phenomenon that drives cell death, or can act as a transient ‘safety valve’, allowing mitochondria to limit stress-induced damage . It is therefore tempting to question whether cytosolic cAMP entering the matrix during transient MPT events might have a role in the downstream recovery of these organelles.
What is the functional significance of an intramitochondrial cAMP cascade? Acin-Perez et al. [6,30] presented evidence of a model where, in response to metabolically produced CO2, sAC generates cAMP and activates a matrix-contained pool of PKA, resulting in phosphorylation of COXIV-1 (cytochrome c oxidase subunit IV isoform 1) with consequent enhancement of OXPHOS (oxidative phosphorylation), contrasting previous data that suggested PKA-dependent OXPHOS inhibition . According to these authors, matrix CAs (carbonic anhydrases) convert CO2 into bicarbonate, which in turn activates the sAC/cAMP/PKA cascade to phosphorylate COXIV-1. The presence of a functional independent cAMP pathway in the mitochondrial matrix is very appealing, and is supported by the presence of sAC in this compartment; however, there are several debated points in this model. For instance, although there are reports suggesting the presence of PKA in the matrix [24,26], there is no evidence of a possible import mechanism for this enzyme . The AKAPs could provide an alternative mechanism for targeting PKA in the matrix; however, to date, no AKAP has been identified in this compartment [18,26].
Recently, a FRET-based PKA activity sensor was targeted to the mitochondrial matrix. This probe failed to detect any endogenous PKA activity in response to membrane-permeant cAMP analogues or bicarbonate (to activate sAC); however, it displayed saturating FRET changes when co-expressed with a matrix-targeted PKA, whereas the FRET signal recovered to basal levels when a PKA inhibitor was used, demonstrating that the probe was functional . These data suggest against the presence of PKA in the mitochondrial matrix and are in contrast with a previous study which found that, in isolated mitochondria, the matrix contained high levels of PKA activity . It is noteworthy that overexpression of a matrix-targeted PKA induced specific phosphorylation patterns compared with cytosolic PKA activated by a combination of the non-selective AC activator forskolin and PDE inhibitor IBMX (3-isobutylmethylxanthine) and both treatments increased the cellular ATP content suggesting a level of functional interchangeability between cytosolic and mitochondrial cAMP cascades .
The IMS is the second aqueous compartment of mitochondria (the other being the matrix). The IMS is defined by the OMM and IMM and has emerged as a significant regulatory domain in the co-ordination of mitochondrial and cellular function . The position of the IMS at the interface between cytosol and mitochondrial matrix influences the small-molecule composition of this domain. For example, the OMM is permeable to molecules with molecular mass lower than 5 kDa , therefore cAMP (~329 kDa) is likely to permeate the OMM and its concentration within the IMS is expected to mirror that of the cytosol. Additionally, more than 30 years ago, the IMS was found to contain significant PDE activity, supporting the idea that an autonomously regulated cAMP cascade might be present in that domain .
Concrete evidence for PKA-dependent phosphorylation in the IMM/IMS was provided by the identification of the protein ChChd3 (coiled-coil–helix–coiled-coil–helix-domain-containing 3) as a PKA target . This protein was shown to reside in the IMM facing the IMS using biochemical techniques and conventional electron microscopy ; however, a recent study that used a combination of MS and EM provided evidence for matrix localization of ChChd3 . Additional support for the ChChd3 IMS-facing conformation was provided by the identification of OMM, IMM and IMS proteins (Sam50, OPA1 and mitofilin) as its binding partners . ChChd3 is a protein of the Chch family and was shown to be a key regulator of cristae maintenance . It will be interesting to understand whether PKA-dependent phosphorylation of ChChd3 affects cristae architecture; however, experimental evidence in support of this is lacking.
The answer to whether the PKA pool responsible for ChChd3 phosphorylation migrates from other sites or is resident in the IMS was given when an AKAP was found in the IMM/IMS fractions of murine heart mitochondria . In contrast with the majority of AKAPs, which bind preferentially to type II PKA regulatory subunits , the IMM/IMS-resident AKAP SKIP (sphingosine kinase-interacting protein) interacts specifically with type I regulatory subunits [39,40]. Means et al.  provided compelling evidence showing that SKIP is able to bind two PKA holoenzymes that were responsible for the PKA-dependent phosphorylation of ChChd3. On the basis of these findings, it can be proposed that the IMS contains a cAMP microdomain built around SKIP and relying on PKA type I. Compared with type II, type I regulatory subunits display higher affinity to cAMP , thus the presence of PKA type I could confer on the IMS-cAMP cascade the ability to respond to low levels of messenger that perhaps would not elicit PKA activity in other mitochondrial domains.
Outer mitochondrial membrane
The OMM is the barrier that separates the IMS from the cytosol. Thanks to this structure, solutes smaller than 5 kDa can permeate the IMS, whereas larger molecules cannot . In terms of cAMP signalling, the OMM is widely accepted as a site hosting significant cAMP/PKA activity . At least three different AKAPs can tether PKA to the OMM: AKAP1 and its isoforms (AKAP 149, AKAP121 and D-AKAP) , Rab32  and AKAP2 . AKAP2 is a protein of dual specificity and can associate equally with PKA type I and type II. Rab32 binds only PKA type II and AKAP1 can associate with both, but its affinity is much higher for the type II enzymes .
PKA tethered to the OMM regulates important cellular functions such as apoptosis, but also mitochondrial morphology [26,43]. For instance, PKA at the OMM can phosphorylate Drp1 (dynamin-1-like protein), a large GTPase involved in the mitochondrial fission–fusion balance. Unphosphorylated Drp1 is recruited to the mitochondria where it polymerizes and promotes mitochondrial fission. PKA-dependent phosphorylation at Ser637 inhibits the formation of Drp1 complexes and promotes mitochondrial fusion. The AKAP121/PKA domain promoted mitochondrial fusion, ameliorated mitochondrial function and promoted cell survival in a neuronal cell model , whereas, in a different study, was shown to regulate mitochondrial adaptation to hypoxia . OMM-bound PKA was also shown to phosphorylate the BH3 (Bcl-2 homology 3) pro-apoptotic protein BAD. Under stress conditions, BAD translocates to the OMM where it binds and inhibits the anti-apoptotic protein Bcl-2, promoting cytochrome c release from the IMS, leading to apoptosis. Phosphorylation by OMM PKA blocks BAD association with Bcl-2 and results in retention of cytochrome c in the IMS and cell survival .
PKA at the OMM ordinarily responds to cytosolic cAMP variations; however, it was proposed that, during ischaemia or acidosis, sAC might translocate to the mitochondria [47,48]. These authors used cardiac myocytes as well as coronary endothelial cells and showed that, during ischaemic stress, sAC moved to the mitochondria (probably the OMM) where it presumably activated a local pool of PKA, enhancing the activity of the pro-apoptotic BAX protein and resulting in cell death [47,48]. The presence of sAC at the OMM would provide an alternative local source of cAMP able to respond to intracellular cues. It would be interesting to understand whether sAC at the OMM can be regulated by CAs, as it was proposed for the matrix. Interestingly, recent work performed in adult rat cardiac myocytes showed that, in these cells, the mitochondrial matrix contained negligible CA activity, whereas the bulk of these enzymes was localized in the proximity of the OMM .
Despite many studies supporting the notion that cAMP at the OMM drives a key regulatory cascade, very little is known about the kinetics and regulation of this localized pathway. A large number of OMM-bound AKAPs would suggest that this compartment is rich in PKA; however, when PKA activity was assessed in isolated mitochondria, the OMM was found to contain only 9% of the total mitochondrial PKA activity . In a recent study, FRET-based probes for the detection of cAMP and PKA activity were targeted to the OMM of HeLa and HEK (human embryonic kidney)-293 cells . Combinations of these tools were used in a co-culture protocol to perform parallel measures of PKA activity and cAMP kinetics in the OMM and cytosol. Thanks to these experiments, it was found that cAMP kinetics at the OMM mirror those of the cytosol. The OMM and cytosolic PKA responded equally to cAMP elevations; however, when the cAMP signal was terminated (both by rinsing away the agonists or blocking PKA with its inhibitor H89) the OMM PKA activity persisted significantly longer than the cytosol. Further experiments demonstrated that this differential regulation was independent of PDEs and probably depended on lower phosphatase activity at the OMM compared with the cytosol .
Historically, mitochondria were recognized solely as the ‘powerhouses’ of the cell, responsible for the generation of nearly 90% of cellular ATP. However, this view has changed drastically, and now these organelles are recognized to be crucial players in important cellular processes such as apoptosis, reactive oxygen species production and Ca2+ buffering . In order to synchronize their function to the cellular needs, mitochondria evolved in signalling centres able to receive, integrate and execute cellular cues. The second messenger cAMP appears to be one of these cues; in fact, it is becoming increasingly clear that all mitochondrial compartments contain proteins able to sense and respond to cAMP. Most of the particulars of how cAMP signalling in specific mitochondrial compartments regulates different aspects of mitochondrial physiology remain to be elucidated. However, it is obvious that localized cAMP signalling can affect many important mitochondrial functions. The future challenge is to identify the cellular stimuli that activate each submitochondrial cAMP microdomain and the downstream targets of each of these cascades. Our increasing understanding on mitochondrial cAMP signalling might provide an effective way to fine-tune various organellar functions by using targeted manipulation of cAMP signals at specific mitochondrial compartments, which could pave the way towards novel therapeutic interventions for mitochondria-related diseases.
Targeting cAMP Signalling to Combat Cardiovascular Disease: A Biochemical Society Hot Topic Event held at Charles Darwin House, London, U.K., 9 December 2013. Organized and Edited by Mark Bond (University of Bristol, U.K.), Tim Palmer (University of Glasgow, U.K.) and Stephen Yarwood (University of Glasgow, U.K.)
cytochrome c oxidase subunit IV isoform 1
exchange protein directly activated by cAMP
inner mitochondrial membrane
mitochondrial permeability transition
outer mitochondrial membrane
protein kinase A
catalytic PKA subunit
regulatory PKA subunit
soluble adenylate cyclase
sphingosine kinase-interacting protein
transmembrane adenylate cyclase
I thank Dr Oliver Lomas and Dr Alex Burdyga for a critical reading of the paper before submission.
This work was supported by a British Heart Foundation, Centre of Research Excellence Intermediate Fellowship [award RE/08/004].