Phosphatases Control the Duration and Range of cAMP/PKA Microdomains

Abstract

The spatiotemporal interplay between the second messenger cyclic AMP (cAMP) and its main effector, protein kinase A (PKA), is crucial for the pleiotropic nature of this cascade. To maintain a high degree of specificity, the cAMP/PKA axis is organised into functional units called microdomains, precisely distributed within the cell. While the subcellular allocation of PKA is guaranteed by a family of tethers called A-Kinase-anchoring Proteins (AKAPs), the mechanisms underlying the efficient confinement of a microdomain’s functional effects are not fully understood. Here, we used FRET-based sensors to detect cAMP levels and PKA-dependent phosphorylation within specific subcellular compartments. We find that cellular cAMP levels may depend on different mechanisms and are responsible for the activation of local PKA enzymes. On the other hand, the dephosphorylating actions of phosphatases dictate the duration of the microdomain’s effects. To test the range of action of PKA microdomains, we used rigid aminoacidic nanorulers to distance our FRET sensors from their original location for 10 or 30 nm. Interestingly, we established that cAMP levels do not affect the spatial range of the microdomain while on the contrary, phosphatase activity provides a functional boundary for phosphorylated PKA targets. Finally, using the same strategy to distance phosphatases from the mitochondria, we found that enzymes close to the outer mitochondrial membrane produced a fragmented phenotype that was not observed when phosphatases were moved to 30 nm from the organelle’s surface. Our findings contribute to the design of a picture where 2 microdomain-forming events have distinct roles. Cyclic AMP elevations trigger the initial activation of subcellular PKA moieties, while the temporal and spatial extent of the PKA’s actions are regulated by phosphatases.

Graphical Abstract

Graphical Abstract.

Introduction

Protein kinase A (PKA) is a tetrameric enzyme composed by 2 catalytic (PKACs) and 2 regulatory (PKARs) subunits. In humans, PKARs are encoded by 4 genes (RIα, RIβ, RIIα, and RIIβ), while PKACs can be found in 3 isoforms (Cα, Cβ, or Cγ).¹ PKA holoenzymes that contain RI subunits are defined as type I, while type II are the ones that contain RII regulatory subunits. In response to rises in intracellular cyclic AMP (cAMP) levels, the regulatory subunits release the PKAC monomers, which are constitutively active and able to phosphorylate numerous endogenous proteins. Despite the apparent linearity of the signaling pathway and the promiscuous nature of the catalytic subunits, the cAMP/PKA axis exhibits remarkable functional specificity of action. Indeed, PKA can pair extracellular and/or intracellular stimuli to distinct cellular responses, even though cAMP production is the common denominator of all these cues. The precise modalities through which stimuli that increment cellular cAMP levels are coupled to specific outcomes are not fully understood. The leading idea suggests that the cAMP pathway is compartmentalized into functional units called microdomains. Thanks to this organization, changes in cAMP levels triggered by specific stimuli impinge on distinct microdomains generating a sort of spatiotemporal code that allows the cells to interpret and precisely respond to the initial stimulus.^{2,  3}

If distilled to their essence, microdomains could be characterized mainly by 2 parameters, that is, their subcellular localization, which determines the molecular targets within their reach, and their activation/inactivation kinetics, determining the duration of their actions. When it comes to cAMP/PKA microdomains, subcellular distribution is determined by a family of tethers, the A-kinase anchoring proteins (AKAPs).^{4,  5} AKAPs associate specifically with the regulatory subunits of PKA (preferably RIIs) thanks to a docking and dimerization (D/D) domain located at their N-terminus.^{6,  7} The cellular domains where AKAPs constrain PKA holoenzymes are defined by targeting sequences able to localize them and, consequently, their associated proteins to specific regions within cells. Importantly, the roles of AKAPs extend beyond the relocation of PKA. In fact, these proteins can aid the formation of complexes with other microdomain constituents such as phosphatases (PPs), adenylyl cyclases (ACs), and phosphodiesterases (PDEs).^{8,  9} The coordination of these components allows for efficient management of cAMP and potentially PKA activity within microdomains. For instance, ACs can offer a local cAMP source, while PDEs, with the aid of cAMP buffers,¹⁰ are thought to be the enzymes that “shape” intracellular cAMP levels. Managing cAMP availability is considered the principal regulatory mechanism of PKA activity. Nevertheless, plenty of experimental evidence suggests the involvement of other events in determining the spatiotemporal entity of cAMP/PKA microdomains.^11–13 It is expected that in response to cAMP rises and PKA activation, the targets within reach of the microdomain will be phosphorylated. However, lowering the levels of the messenger will result in decreased kinase activity but is not directly linked to the dephosphorylation of the PKA targets. In fact, the definite termination of the functional outcomes of the cascade requires phosphatases, another regulator of microdomains.¹¹ In line with these considerations, we recently found that PKA-dependent phosphorylation persists longer within domains with low phosphatase activity and subsides earlier where phosphatases are more active.^{11,  13} Based on these observations, it would be reasonable to assume that the range of action of a kinase-driven microdomain may depend on the phosphatase activity of its surroundings. This is particularly relevant for cAMP/PKA microdomains, as AKAPs bind only the PKARs while PKACs can be released in response to cAMP elevations. In principle, free PKACs can diffuse and phosphorylate targets along their path; however, this would go against the principle of compartmentalization, unless the diffusion and/or activity of the kinase were controlled. Several modalities that could limit the diffusion of PKACs have been proposed. The first is based on observations suggesting that, at least for AKAP79, even in the presence of cAMP, anchored PKA holoenzymes remain intact and PKACs do not diffuse away from the anchoring sites.¹⁴ According to this hypothesis, low cAMP elevations fail to induce the release of PKACs, whilst allowing the phosphorylation of proximal targets. Subsequent work used a different approach and proposed an alternative mechanism of free PKAC restriction that is based on the excess of regulatory subunits when compared to the catalytic ones.¹⁵ Taking into consideration that the regulatory subunits can exhibit 2 states, one with high affinity for cAMP, and another with high affinity for the catalytic subunits, the abundance of PKARs as compared to the PKACs (up to 17-fold depending on the cell type^{15,  16}) could indeed support high rates of association of PKA holoenzymes and limit their diffusion. However, the efficient restriction of PKACs would also depend on the state of PKARs and the levels of cAMP within the microdomain. In fact, when the cAMP levels are high (eg, during the activation of a microdomain), PKA holoenzymes remain dissociated leaving open the question of how the actions of PKACs are delimited. In addition, it is well established that phosphorylation of type II but not type I regulatory subunits is important for the association–dissociation cycles of PKA RII-based holoenzymes.^17–19 Contrary to RIs, RIIs contain a phosphorylatable serine within their inhibitory domain. This residue is part of a PKA consensus sequence (RRXS) and can be auto phosphorylated resulting in the reduction of the binding affinity of RIIs for the catalytic subunits. Interestingly, this feature opens a possible regulatory link between PKA holoenzyme assembly and phosphatases. In fact, it was calculated that RII subunits dephosphorylated by phosphatases can associate with PKACs nearly 50 times faster than their phosphorylated counterparts.^{17,  19} Finally, N-terminal myristoylation was found to relocate free PKAC by facilitating their membrane association. This modality, however, increases rather than decreases PKA activity near membranes, at least in neurons.¹²

While the restriction of PKAC subunits is a crucial step for controlling the range of a microdomain, we reasoned that another complementary regulatory mechanism would be to control the range of PKACs actions, and that of their phosphorylated targets, through the actions of phosphatases. We recently demonstrated that PKA-dependent phosphorylation persists longer near the membranes due to differential phosphatase actions.^{11,  13} Based on these findings, we hypothesized that, as PKACs diffuse away from a microdomain, the exposure of their phosphorylated targets to phosphatases will vary, facilitating dephosphorylation and consequently limiting the range of action of the cAMP/PKA domain. To test this hypothesis, we compared 2 cellular models that we found display radically different cAMP handling modalities, the cervical cancer cell line (HeLa) and the osteosarcoma cell line (U-2 OS). We used rigid peptide nanorulers of known length (10 and 30 nm)²⁰ to distance from their original sites the mitochondrial and plasma membrane (PM)-targeted versions of the PKA-dependent phosphorylation-sensitive FRET-based sensor AKAR4²¹ or a catalytically active phosphatase.¹¹ We found that the vicinity to membranes is directly associated with more pronounced and persistent PKA-dependent phosphorylation of the sensors and therefore longer lasting local effects. Interestingly, when the distance of the sensors was increased, the effects of phosphatases became more marked, and the rate of dephosphorylation increased independently of the levels of cAMP. In line with these observations, when we targeted catalytically active phosphatases at different distances from the mitochondrial surface, changes in organellar dynamics became less evident as the distance of the enzymes from the outer mitochondrial membrane increased. Our data suggest that PKACs and their substrates, moving away from membrane bound AKAPs, encounter environments with increasing phosphatase pressure that limit their actions and therefore define both the duration and range of PKA microdomains.

Materials and Methods

Reagents

Forskolin (FSK, Cat. No. 11018), 3-isobutyl-1-methylxantine (IBMX, Cat. No. 13347), and H-89 dihydrochloride (H-89, Cat. No. 10010556) were purchased from Cayman Chemical (USA). Dimethyl sulfoxide (DMSO, Cat. No. D8418), phosphate-buffered saline (PBS, Cat. No. P4417), Tween 20 (Cat. No. P1379), bovine serum albumin (BSA, Cat. No. A9647), and skim milk powder (Cat. No. 70166) were purchased from Merck (Germany).

Cell Culture and Transfection

U-2 OS and HeLa cells were grown at 37°C in a 5% CO₂ atmosphere in Dulbecco’s modified Eagle’s medium (DMEM) with high glucose (Cat. No. ECM0728, Euroclone, Italy) supplemented with 10% fetal bovine serum (FBS; Cat. No. ECS5000L, Euroclone, Italy), 100 U/mL penicillin, 100 μg/mL streptomycin (Cat. No. ECB3001D, Euroclone, Italy), and 2 mm L-Glutamine (Cat. No. ECB3000S, Euroclone, Italy). Cells were split every 2–3 days at a confluence of 80%–90%. Twenty-four hours before transfection, cells were plated onto custom 15 mm diameter circular glass coverslips (thickness 0.16 mm, purchased from Vemi S.r.l., Italy) and were allowed to grow to 50% of confluence. Cells were transfected with Lipofectamine 2000^® reagent (Cat. No. 11668019, Thermo Fisher Scientific, USA) following manufacturer’s instructions.

Plasmids and Cloning Strategies

The construct mCherry2-C1 was a gift from Michael Davidson (Addgene plasmid #54 563; http://n2t.net/addgene:54563; RRID:Addgene_54563), pcDNA3_Epac-S-H187 (Addgene plasmid #170348; http://n2t.net/addgene:170348; RRID:Addgene_170348), as well as pcDNA3-H30 and pcDNA3-H84, which were a gift from Kees Jalink, pcDNA3-AKAR4 was a gift from Jin Zhang (Addgene plasmid #61619; http://n2t.net/addgene:61619; RRID:Addgene_61619), and pcDNA3-Lyn-AKAR4 was a gift from Jin Zhang (Addgene plasmid #61620; http://n2t.net/addgene:61620; RRID:Addgene_61620) all above constructs were purchased from Addgene (USA). pcDNA3-H30-NLS was generated by adding a general nuclear localization signal to the C-terminus of pcDNA3-H30. The mitochondria-targeted RFP construct (mtRFP) was a gift from Prof. Luca Scorrano. The pcDNA3-OMM-AKAR4 sensor and pcDNA3-PKA-mCherry were previously generated in our laboratory.¹¹ The pcDNA3-OMM-SAH10-AKAR4 and pcDNA3-OMM-SAH30-AKAR4 constructs were obtained by PCR amplification of SAH10 and SAH30, respectively, and then cloned in pcDNA3-OMM-AKAR4 using NheI and BamHI restriction sites. SAH10 and SAH30 were amplified from pcDNA3-Epac1-camps-SAH10-PDE4A1 and pcDNA3-Epac1-camps-SAH30-PDE2Cat kindly gifted by Professor Martin J. Lohse, Dr. Andreas Bock, and Dr. Isabella Maiellaro. The pcDNA3-SAH10-AKAR4 and pcDNA3-SAH30-AKAR4 were obtained from pcDNA3-OMM-SAH10-AKAR4 and pcDNA3-OMM-SAH10-AKAR4, respectively, by excision of the OMM targeting sequence using the double HindIII restriction site. The constructs pcDNA3-Lyn-SAH10-AKAR4 and pcDNA3-Lyn-SAH30-AKAR4 were obtained from pcDNA3-OMM-SAH10-AKAR4 and pcDNA3-OMM-SAH30-AKAR4, respectively. The OMM targeting sequence was excised by HindIII digestion. Primers containing the Lyn sequence and the HindIII restriction site were then used as inserts in the ligation reaction. The construct encoding for PP2ACα (PPP2ACα) was kindly gifted by Professor Ajay Singh, while its mitochondrial targeted version OMM-PP2ACα was previously generated in our laboratory.¹¹  OMM-SAH10-PP2ACα and OMM-SAH30-PP2ACα constructs were obtained by replacing AKAR4 of the OMM-SAH10-AKAR4 and OMM-SAH30-AKAR4 with PP2ACα using BamHI and EcoRI restriction enzymes.

FRET-Based Imaging

U-2 OS and HeLa cells plated onto 15 mm diameter circular glass coverslips (thickness 0.16 mm, purchased from Vemi S.r.l., Italy) and transfected with different FRET sensors were mounted onto an open perfusion chamber RC-25F (Warner Instruments, USA). The cells were bathed in Ringer’s modified buffer: 125 mm NaCl; 5 mm KCl; 1 mm Na₃PO₄; 1 mm MgSO₄; 20 mm Hepes; 5.5 mm D-(+)-glucose; 1 mm CaCl₂; pH adjusted to 7.4 using 1 m NaOH. The experiments were performed on an Olympus IX83 inverted microscope (Olympus, Japan) equipped with Filter Wheel mode with Optospin (Cairn Research Ltd, England) and a CCD camera Retiga R^™ Series R1 (Teledyne Photometrics, USA). The cyan fluorescent proteins (Cerulean for AKAR4 or mTurquoise for H187) were excited for 200 ms at 430 nm, while the emission fluorescence was collected every 15 s for both donor and acceptor fluorophores at 480 and 535 nm, respectively. Automatic image collection and preliminary analysis were performed using the MetaFluor^® 7.10.5.476 (Molecular Devices, USA). Raw data were transferred to Excel (Microsoft, USA) for background subtraction and generation of the ratios. Graphs were generated with Excel (Microsoft, USA) or GraphPad Prism 10 software (GraphPad, USA).

Western Blotting

The cells were lysed in cold Lysis buffer supplemented with cOmplete mini EDTA-free protease inhibitor cocktail (Cat. No. 11836170001, Roche, Switzerland) and phosphatases inhibitor PhosSTOP™ (Cat. No. 04906837001, Roche, Switzerland). The cell lysates were cleared at 10.000 × g at 4°C for 10 min, and 40 μg of proteins were loaded onto 10% polyacrylamide gel (Tris-Glycine Gels, Thermo Fisher Scientific, USA) and then transferred onto polyvinylidene fluoride membranes (Cat. No. 10600023, Cytiva, UK). Subsequently, the membranes were blocked for 1 h at room temperature with 10% (w/v) non-fat-dry milk in 0.2% TBS-Tween (TBS-T) and then incubated overnight at 4°C with primary antibody (1:1000) in 5% non-fat-dry milk-TBS-T. The day after, membranes were washed 3 times with TBST at room temperature and then incubated for 1 h at room temperature with peroxidase-conjugated secondary antibodies. Peroxidase activity was detected with enhanced chemiluminescence (Luminata Crescendo Western HRP, Cat. No. WBLUR0500, Merck, Germany). Membranes were stripped using Restore Western Blot Stripping Buffer (Cat. No. 46430, Thermo Fisher Scientific, USA) for 15 min at room temperature and thoroughly washed with TBS-T.

Anti-PKA Catα (C-20) (Cat. No. sc-903; Santa Cruz Biotechnologies, USA), anti-PKA Reg Iα/β (G-6) (Cat. No. sc-271446; Santa Cruz Biotechnologies, USA), anti-PKA Reg IIβ (L-16) (Cat. No. sc-26803; Santa Cruz Biotechnologies, USA), anti-PP1 (E-9) (Cat. No. sc-7482; Santa Cruz Biotechnologies, USA), anti PP2Aα/β (Cat. No. ab32141; Abcam, UK), and anti β-actin (Cat. No. A3854; Merck, Germany), anti PKA phospho-substrate (Cat. No. 9624; Cell Signaling Technology, USA). Anti-Mouse (Cat. No. 115-035-003) and Anti-Rabbit (Cat. No. 111-035-003) peroxidase-conjugated secondary antibodies were purchased from Jackson ImmunoResearch Europe (UK). Anti-goat peroxidase-conjugated secondary antibody (Cat. No. sc-2033) was purchased from Santa Cruz Biotechnology (USA). Quantification of the bands was performed by densitometric scanning with the iBright analysis software (Thermo Fisher Scientific, USA). Local background correction volume (LB Corr. Vol.) is represented by the sum of each grayscale pixel intensity in the band, minus the local background intensity. The housekeeping protein (β-actin) was used as normalization factor. Results are presented as the mean ± SEM of at least 7 independent experiments.

Indirect Immunofluorescence and Confocal Imaging

HeLa cells were seeded onto 15 mm diameter circular glass coverslips (thickness 0.16 mm, purchased from Vemi S.r.l., Italy) and transfected after 20-24 h. The day after, cells were fixated in 4% paraformaldehyde-PBS (Cat. No. sc-281692, Santa Cruz Biotechnology, USA) for 10 min, then washed 5 times in PBS and permeabilized in 0.5% Triton X-100 (Cat. No. X100, Merck, Germany)-TBS for 10 min. During blocking, cells were incubated at room temperature in 0.1% Triton-X100, 2% BSA, and 0.1% Sodium Azide (Cat. No. 71289, Merck, Germany)-TBS. The primary antibody anti-Citrate Synthase (D7V8B), purchased from Cell Signaling Technology (USA, Cat. No. 14309S) diluted 1:300 in the same blocking solution as well as the secondary antibody Alexa Fluor 647 (Cat. No. A21244, Life Technologies, USA). Plasma membranes were stained with wheat germ agglutinin (WGA) conjugated with Alexa Fluor 647 (Cat. No. W32466, Life Technologies, USA) following manufacturer instructions. After the immunostaining coverslips were mounted with the Fluoroshield^™ mounting medium supplemented with DAPI (Cat. No. F6057, Merck, Germany). Cells immunostained for Citrate Synthase were acquired on a Leica TCS SP8 confocal scanning microscope using oil immersion 63x (HC PL APO 63x/1.40 Oil CS2, Leica, Germany) objective. Cells stained with WGA were acquired with a Zeiss LSM900 Confocal upright confocal scanning microscope (Carl Zeiss Microscopy, Germany) using an oil immersion 63x (Zeiss Plan-Apochromat 63x/1.4, Carl Zeiss Microscopy, Germany) objective. Cells transfected with OMM-PP2A, OMM-SAH10-PP2A, or OMM-SAH30-PP2A and co-transfected with mtRFP that did not require immunostaining were directly washed and mounted after permeabilization.

Statistics, Data Fitting, and Data Analysis

FRET experiments were independently repeated at least 3 times with similar results. The cell number analyzed for each experiment is reported in the figure legends. The overall FRET graphs are shown as the average of single-cell traces ± SEM. Data were analysed with GraphPad Prism 10 (GraphPad Software, USA). In Figures 1 , 2, and 4, the maximal FRET responses were quantified by calculating the difference (Δ) between the maximal FRET ratio (480 nm/535 nm for cAMP sensors and 535 nm/480 nm for PKA sensors) (R_max) for each response and its pre-stimulus FRET ratio (R_basal). In addition, for the experiments in Figures 1, 2, and 4, where data fitting was not possible due to the lack of clear plateau phases, the rate of FRET change was calculated as the ratio difference between the maximum ratio reached by a stimulus minus the ratio at the time of the treatment over the time of the response (points were chosen manually). Error bars on the aligned dot plots indicate the mean ± SEM. The Gaussian distribution of the datasets was assessed with a Shapiro–Wilk normality test. For data normally distributed unpaired sample t-tests were performed, while the Mann–Whitney test was performed in non-parametric conditions. In Figures 5, 6, and 7 datapoints following the addition of FSK and IMBX were fitted using a linear function to obtain a slope (dratio/dt). To perform the exponential decay fit, individual traces were subtracted for the slope of a linear fit performed on the last 50 points to obtain a flat baseline, allowing a reliable single exponential decay fitting using the following equation: Y=(Y0–Plateau) × exp(−K × X) + Plateau, (initial values equal to 1, 1000 iterations), the time constant tau (τ) is obtained. Slopes and exponential decay data were analysed using GraphPad Prism (version 10.0.2 for Mac, GraphPad Software, USA), performing a Shapiro–Wilk normality test followed by the ROUT outlier identification test (Q = 1%). The statistical analysis performed is an ordinary one-way ANOVA and a Tukey’s multiple comparison test. Data are expressed as mean ± SEM. Western blotting results were quantified and statistically analysed with the non-parametric Mann–Whitney test. For confocal fluorescence experiments cells were acquired across 2 or 3 independent experiments. The fluorescence intensity was calculated with Fiji (NIH, USA), and the graphs were generated with GraphPad Prism 8 software (GraphPad Software, USA). Analysis of the grade of fusion of the mitochondrial network in Figure 7D, we used the open-source image analysis platform MitoGraph¹ (https://github.com/vianamp/MitoGraph), which performs automated segmentation and skeletonization of mitochondrial networks from fluorescence microscopy images.²² The data obtained from MitoGraph were processed and analyzed in R Studio using scripts from the MitoGraph-Contrib-RScripts repository (https://github.com/Hill-Lab/MitoGraph-Contrib-RScripts), following the steps outlined by the developers. Among the parameters calculated by the software, the comprehensive MitoGraph Connectivity Score was considered to compare the grade of fusion of the different samples.

Figure 1.

HeLa and U-2 OS exhibit different PDE-dependent cyclic AMP (cAMP) hydrolysis potential. HeLa cells (A) or U-2 OS cells (B) expressing the cAMP-sensitive FRET-based sensor H187 were challenged with low doses (5 µm) of forskolin (FSK) alone or combined with IBMX (500 µm) to inhibit phosphodiesterases (PDEs), followed by the addition of high doses of FSK (20 µm) to achieve saturation of the sensor. Contrary to U-2 OS, HeLa cells did not respond to low FSK unless PDEs were blocked, suggesting that these cells harbor high cAMP-hydrolysing activity. Traces are expressed as the average ± SEM; n = 36 HeLa cells; n = 22 U-2 OS. (C) Quantification of maximum FRET changes (i) and rate of FRET change (ii) for each treatment. (D) HeLa cells or U-2 OS cells (E) were challenged with FSK (20 µm) combined to IBMX (500 µm), resulting in the saturation of the sensor. Subsequent rinsing of IBMX, to unmask the IBMX-sensitive PDE activity resulted in the recovery of the FRET signal. In line with B and C in HeLa cells FRET recovery, which mirrors the rate of cAMP hydrolysis, was significantly faster, suggesting higher PDE activity as compared to U-2 OS. Traces are expressed as the average ± SEM; n = 31 HeLa cells; n = 16 U-2 OS. (F) Quantification of maximum FRET changes (i) and rate of FRET change (ii) for each treatment. For P-value calculation, sample distribution was assessed using a Shapiro–Wilk test (α = 0.05), samples with normal distribution were subject to unpaired t-test (α = 0.05). While samples with non-Gaussian distribution were subject to Mann–Whitney tests (α = 0.05).

Figure 2.

In HeLa but not U-2 OS phosphodiesterase (PDE) activity shapes the sub-plasma membrane (PM) cyclic AMP (cAMP) levels. HeLa cells (A) or U-2 OS cells (B) expressing a PM targeted version of the cAMP-sensitive FRET-based sensor EPAC-SH30 (PM-H30) were challenged with low doses (5 µm) of forskolin (FSK) alone or combined with IBMX (500 µm) to inhibit PDEs, followed by the addition of high doses of FSK (20 µm) to achieve saturation of the sensor. In U-2 OS, inhibition of PDEs had only marginal effects, on the contrary, in HeLa cells addition of IBMX resulted in a significant FRET response. Traces are expressed as the average ± SEM; n = 41 HeLa cells; n = 29 U-2 OS. (C) Quantification of maximum FRET changes (i) and rate of FRET change (ii) for each treatment. (D) HeLa or U-2 OS cells (E) expressing PM-H30 were treated with saturating concentrations of FSK (20 µm) and IBMX (500 µm). Rinsing of IBMX to release PDE activity had no effect in U-2 OS cells but resulted in a fast decrease of the FRET signal in HeLa cells. Traces are expressed as the average ± SEM; n = 22 HeLa cells; n = 26 U-2 OS. (F) Quantification of maximum FRET changes (i) and rate of FRET change (ii) for each treatment. For P-value calculation, sample distribution was assessed using a Shapiro–Wilk test (α = 0.05), samples with normal distribution were subject to unpaired t-test (α = 0.05). While samples with non-Gaussian distribution were subject to Mann–Whitney tests (α = 0.05).

Figure 4.

HeLa and U-2 OS cells have similar protein kinase A (PKA) substrate dephosphorylation kinetics. (A) Schematic representation of the experimental strategy used. (B) HeLa cells (dark traces) and U-2 OS cells (light traces) expressing AKAR4, OMM-AKAR4 (C), or Lyn-AKAR4 (D). Cells were treated with high concentrations of forskolin (FSK) (20 µm) and IBMX (500 µm) until all sensors reached saturation. At this point, the PKA inhibitor H-89 (60 µm) was also added to the experimental medium. The rate of dephosphorylation within different compartments was virtually overlapping independently of the cell model. Traces are expressed as the average ± SEM; AKAR4: n = 30 HeLa cells and n = 16 U-2 OS; OMM-AKAR4: n = 33 HeLa cells and n = 10 U-2 OS; Lyn-AKAR4: n = 27 HeLa cells and n = 14 U-2 OS. Insets: (ii) quantification of maximum FRET changes and rate of FRET change (iii) for each treatment. P-values were obtained from unpaired t-test (α = 0.05) performed after normality verification using Shapiro–Wilk test (α = 0.05).

Figure 5.

Phosphatase-dependent dephosphorylation of protein kinase A (PKA) substrates is affected by the distance from the plasma membrane (PM) (A) HeLa cells expressing AKAR4 (green trace), Lyn-SAH30-AKAR4 (violet trace), Lyn-SAH10-AKAR4 (blue trace), or Lyn-AKAR4 (red trace). All sensors were saturated by treatment with forskolin (FSK) (20 µm) combined to IBMX (500 µm). Addition of H-89 (60 µm) in the presence of FSK/IBMX, resulted in different dephosphorylation kinetics as measured by the recovery of the FRET signals. Traces are expressed as the average ± SEM. (B) The exponential decay of each sensor curve was assessed using tau, τ. Data are presented as aligned dot plot of individual data points and mean ± SEM; AKAR4: n = 34; Lyn-AKAR4: n = 37; Lyn-SAH10-AKAR4: n = 35; Lyn-SAH30-AKAR4: n = 25. Similar experiments were performed in U-2 OS cells (C), and tau descriptor of the exponential decay of the curves is shown in (D). Traces are expressed as the average ± SEM. Data are presented as aligned dot plot of individual data points and mean ± SEM; AKAR4: n =11; Lyn-AKAR4: n = 33; Lyn-SAH10-AKAR4: n = 14; Lyn-SAH30-AKAR4: n = 24. In both cellular models, tau indicated that increases in phosphatase activity reached a plateau already after 10 nm from the PM. Seconds (s).

Figure 6.

Phosphatase-dependent dephosphorylation of protein kinase A (PKA) substrates increases proportionally to the distance from mitochondria (A) HeLa cells expressing AKAR4 (green trace), OMM-SAH30-AKAR4 (violet trace), OMM-SAH10-AKAR4 (blue trace), or OMM-AKAR4 (red trace). All sensors were saturated by treatment with FSK (20 µm) combined to IBMX (500 µm). Addition of H-89 (60 µm) in the presence of FSK/IBMX, resulted in different dephosphorylation kinetics as measured by the recovery of the FRET signals. Traces are expressed as the average ± SEM. (B) The exponential decay of each sensor curve was assessed using tau. Data are presented as aligned dot plot of individual data points and mean ± SEM; AKAR4: n = 34; OMM-AKAR4: n = 39; OMM-SAH10-AKAR4: n = 30; OMM-SAH30-AKAR4: n = 21. P-values were obtained from one-way ANOVA and Tukey’s multiple comparison test (α = 0.05). Similar experiments were performed in U-2 OS cells (C), and tau descriptor of the exponential decay of the curves is shown in (D). In both cellular models, tau decreases were inversely proportional to the distance from the outer mitochondrial membrane, indicating increasing phosphatase pressure moving away from the organelles. Traces are expressed as the average ± SEM. Data are presented as aligned dot plot of individual data points and mean ± SEM; AKAR4: n = 11; OMM-AKAR4: n = 11; OMM-SAH10-AKAR4: n = 13; OMM-SAH30-AKAR4: n = 17). P-values were obtained from one-way ANOVA and Tukey’s multiple comparison test (α = 0.05). Seconds (s).

Figure 7.

Overexpression of Phosphatases targeted at different distances from the OMM induces distinct signaling and functional outcomes. (A) HeLa cells expressing OMM-AKAR4 alone (Scr.) (black trace), or together with soluble PP2A (green trace), OMM-PP2A (red trace), OMM-SAH10-PP2A (blue trace), or OMM-SAH30-PP2A (violet trace). All sensors responded to treatment with forskolin (FSK) (20 µm) and were saturated by treatment with FSK (20 µm) combined to IBMX (500 µm). Addition of H-89 (60 µm) without rinsing FSK/IBMX from the medium, resulted in different dephosphorylation kinetics as measured by the recovery of the FRET signals. Traces are expressed as the average ± SEM. (B) The exponential decay of each sensor curve was assessed using tau. Data are presented as aligned dot plot of individual data points and mean ± SEM; OMM-AKAR4: n = 21; PP2A: n = 31; OMM-PP2A: n = 45; OMM-SAH10-PP2A: n = 31; OMM-SAH30-PP2A: n = 38. P-values were obtained from one-way ANOVA and Tukey’s multiple comparison test (α = 0.05). (C) Confocal photomicrographs of HeLa cells expressing a mitochondrial red fluorescent protein (mtRFP) alone (Scr.) or together with OMM-PP2A, OMM-SAH10-PP2A, or OMM-SAH30-PP2A. (D) Evaluation of fusion/fission status by calculating the connectivity score of each treatment using the MitoGraph algorithm. Connectivity scores are multiparametric and are calculated from the sum of metrics elevated in a fused mitochondrial state divided by the sum of metrics elevated in a fragmented state. Data are presented as aligned dot plot of individual data points and mean ± SEM; mtRFP (Scr.): n = 23; OMM-PP2A: n = 21; OMM-SAH10-PP2A: n = 16; OMM-SAH30-PP2A: n = 19. P-values were obtained from one-way ANOVA and Tukey’s multiple comparison test (α = 0.05). Seconds (s).

Results

HeLa and U-2 OS Cells Display Distinct cAMP Hydrolyzation Kinetics

The intracellular levels of cAMP are defined by a complex multiparametric equilibrium established between intracellular buffers (regulatory subunits and effectors), the production machinery (membrane receptors and adenylyl cyclases (ACs)), and the hydrolases responsible for its degradation (PDEs).¹⁰ The main intracellular buffers are the free PKA regulatory subunits, which are more abundant as compared to the catalytic¹⁵ and, in their free conformation, display significantly higher affinity for the messenger than the holoenzymes.^{10,  23,  24} While the role of cAMP buffers in the maintenance of low intracellular cAMP levels at basal conditions may be relevant, their high affinity suggests that during cAMP-generating events buffers will be fast saturated, allowing the holoenzymes to be activated. Based on these considerations, whether a physiological signal will reach the cAMP concentration threshold to activate PKA mainly depends on its ability to tilt the balance in favor of production.²⁵ To study the involvement of PDE activity in cAMP/PKA compartmentalization, we employed the cAMP-sensitive FRET-based sensor EPAC-S^H187 (H187).^{26,  27} We performed a preliminary screening on different cell lines and chose to continue our investigation using HeLa and U-2 OS cells based on their wide use in the field^{11,  26} and differences in cAMP handling.

To design a protocol aiming to dissect and compare the actions of PDEs, it is imperative to achieve the maximum messenger production in the 2 cell lines. However, it is inherently challenging to account for all PDE isoforms and ACs expressed by the cell models of our study. To overcome this limitation, we used forskolin (FSK),²⁸ a labdane diterpene that activates all transmembrane ACs, except the soluble AC10 (sAC)²⁹ and/or the general PDE inhibitor 8-methoxymethyl-3-isobutyl-1-methylxanthine (IBMX). We reasoned that FSK would provide better control of AC activity as compared to the use of G-Protein Coupled Receptor agonists that, albeit more physiological, result in highly variable cell-type-dependent responses that depend on the expression of the receptors, their G proteins, compartmentalization, and associated signaling machinery.^{30,  31} Another valid method for controlling endogenous cAMP levels would be to employ cAMP analogs;³² however, this approach also presents limitations. For instance, cAMP analogs are not freely permeable and are differentially degraded by PDEs; therefore, their use in different cell lines may introduce unexpected variabilities. Based on these considerations and aiming to test the balance between cAMP production and degradation, we designed the protocols depicted in Figure S1A and B, where we challenged cells expressing FRET-based cAMP-sensitive sensors with different dosages and combinations of FSK and IBMX. Based on previous experiences, we were aware that HeLa cells exhibit high PDE activity. For this reason, and to ensure that the majority of PDEs were inhibited, we performed an IBMX dose–response experiment. As shown in Figure S1C, cells expressing the H187 FRET sensor displayed residual PDE activity after treatment with 250 µm IBMX, for this reason, for our experiments, we used 500 µm. As shown in Figure 1A, HeLa cells expressing H187 barely responded to low doses of FSK, while the addition of IBMX, to block PDEs, produced fast responses, suggesting that cAMP hydrolysis due to IBMX-sensitive PDEs is very pronounced in these cells. As previously demonstrated, increasing the levels of FSK (20 µm) in the presence of IBMX resulted in the saturation of the sensor.^{11,  27} We used the same protocol in U-2 OS cells (Figure 1B), and low doses of FSK alone resulted in larger FRET changes, while PDE inhibition (IBMX) had only a minor and relatively slow effect on the FRET signal that was further enhanced only by higher FSK doses. Quantifications of the max FRET ratio change for each treatment as well as the rate of FRET changes (Figure 1C) confirmed that U-2 OS cells responded better than HeLa to low doses of FSK (5 µm), while the latter had larger and faster responses upon PDE inhibition. Finally, the maximal FRET signal reached using high levels of FSK (20 µm) together with IBMX (500 µm) was higher in HeLa cells (Figure 1C). These experiments suggested that PDEs impinge to a greater extent on cAMP signals in HeLa cells, while they are less active in U-2 OS. To directly test the cAMP hydrolysing potential of HeLa and U-2 OS cells, we used an experimental protocol where high doses of FSK in the presence of IBMX were used to induce the saturation of the sensor, followed by rinsing of IBMX (whilst keeping the levels of FSK constant) to release PDE activity but maintain the high production levels (Figure S1B). As shown in Figure 1D and E, combined treatment with high doses of FSK and IBMX produced fast, saturating responses in both cell lines. On the other hand, rinsing IBMX, to release PDE activity, resulted in drastically different kinetics in the 2 models. In line with our previous results, in HeLa cells the FRET signal reversed with very fast kinetics (Figure 1F), despite the persistent FSK-dependent activation of ACs (maximal cAMP production), confirming high PDE activity in these cells. On the contrary, the same manoeuvre (IBMX removal) resulted in a minimal response in U-2 OS cells, confirming the evidence that the release of PDE activity was not enough to overcome the AC-dependent cAMP production (Figure 1F). In fact, in U-2 OS cells the FRET signal recovered only when FSK was rinsed away from the experimental medium, while the same manoeuvre produced a small recovery in HeLa cells (Figure 1D, E, and F).

The H187 FRET sensor bears a Q270E point mutation in the EPAC module, which increases its affinity for cAMP approximately 2.5-fold, making this construct particularly fit for measuring low levels of the messenger.^{25,  26} High-affinity sensors constitute also potential buffers and are less optimal for measuring variations in cells that can retain higher basal cAMP levels. To exclude an effect of H187 overexpression, we repeated the experiments shown in Figure S1A and B, using a different cAMP-sensitive FRET-based sensor that lacks the Q270E substitution called EPAC-S^H84 (H84)³³ and therefore displays lower affinity for cAMP. As shown in Figure S1D and E, albeit an overall lower dynamic range, H84 displayed kinetics that virtually overlap with those of H187 in both HeLa and U-2 OS. Taken together, these experiments indicate that in HeLa cells PDE activity strongly impacts on cytosolic cAMP levels contrary to U-2 OS, where the PDE-dependent hydrolysis is less dominant and cAMP production by ACs is scarcely contrasted.

Different Mechanisms Control Intracellular cAMP Levels in HeLa and U-2 OS

In our experiments, cAMP production was initiated at the PM by FSK. Under these conditions, it is expected that cAMP will be higher at the site of production, and its free levels will likely decrease due to buffering and PDE activity as it diffuses toward cytosolic sites. The main intracellular cAMP buffers are the free PKA regulatory subunits.^{20,  34} As shown in Figure S2A, we performed Western Blotting experiments to test the expression of PKA signaling components in the 2 cell lines. As quantified in Figure S2B, in HeLa cells the expression of PKA regulatory subunits type I (RI) is more pronounced than in the U-2 OS cells, while the expression of RII subunits is higher in the latter. These data, together with the differences in PDE activity observed in Figure 1, would suggest that the cAMP signal “shaping” capacity will be more pronounced in HeLa than U-2 OS. To test whether the dynamics of cAMP signaling at the proximity of PM are different than those observed in the cytosol, we repeated the experimental protocols depicted in Figure S1A using sensors targeted to the PM. Due to its high affinity, the sensor H187 was deemed less optimal for measurements at the PM where cAMP levels are expected to be high. We employed a PM-targeted version of the FRET-based sensor EPAC-S^H30 (H30) that displays similar kinetics and sensitivity to H84 as demonstrated in Figure S3A and B. Using cells expressing PM-EPAC-S^H30 (PM-H30), we were unable to observe different kinetics near the PM than those measured in the cytosol in any of the 2 cell models. More specifically, as shown in Figure 2A, in HeLa cells low levels of FSK (5 µm) resulted in marginal cAMP elevations (virtually overlapping with the cytosolic responses of the soluble sensors H30, H84 and H187), while PDE inhibition with high doses of IBMX (500 µm) strongly affected the FRET signal, suggesting that PDEs impinge on cAMP production on site. On the other hand, in U-2 OS expressing PM-H30, treatment with FSK 5 µm triggered a significant change in the FRET signal while the effects of PDE inhibition remain limited, as also observed in the cytosol (Figure 2B). Comparison of the maximum FRET change and FRET rate in the 2 cell lines confirmed the differences previously observed in the cytosol (Figure 2C).

To estimate the PDE activity in the vicinity of the PM, we used the protocol of Figure S1B in HeLa and U-2 OS expressing PM-H30, and FRET changes specifically due to cAMP hydrolysis were registered. As shown in Figure 2D and E and quantified in Figure 2F, IBMX removal resulted in fast responses in HeLa cells but had nearly no effect in U-2 OS, which signals reversed only when FSK was removed from the experimental medium with relatively slow kinetics, which is in line with low endogenous PDE activity.

Taken together, these experiments suggested that PDEs differentially impinge on both the kinetics and the levels of free cAMP at the PM. In fact, in cells with high (HeLa) or low (U-2 OS) PDE activity, AC activation using low levels of FSK produced drastically different cAMP responses both as intensity and velocity (Figure 1). Moreover, when we compared cAMP dynamics in different compartments of HeLa and U-2 OS, we found that cAMP kinetics in the cytosol mirrored those measured at the PM in both cell models. In our experiments, we did not observe any cAMP gradients independently of the cAMP hydrolysing ability of each cell line. This observation is in line with recent computational simulations suggesting that PDE activity alone is not sufficient for the generation of intracellular cAMP gradients.³⁵

PKA Activation in Specific Compartments Mirrors Intracellular cAMP Levels

Among the cAMP effectors, PKA is mostly involved in compartmentalization of the pathway. In fact, tethering of this kinase to specific subcellular locations by AKAPs is at the basis of the generation of cAMP microdomains. Both the affinity for cAMP and AKAPs of the PKA holoenzymes depend on their regulatory subunits. Type II regulatory subunits display lower cAMP affinity but higher tendency to complex with AKAPs, while RIs are more sensitive to the messenger but less prone to complex with the tethers.³⁶ In addition, both types of regulatory subunits can act as cAMP buffers and could contribute to its compartmentalization.^{34,  37} Based on these considerations, we checked the expression levels of the cAMP/PKA signaling components in HeLa and U-2 OS. Western Blotting experiments showed that both cell models express RI and RII subunits as well as the main catalytic subunit (PKACα) and the main phosphatases (Figure S2A and B). To determine whether the different cAMP kinetics observed in HeLa and U-2 OS cells resulted in the activation of distinct cAMP/PKA domains, we measured the kinetics of PKA activation in different subcellular locations using targeted versions of the FRET-based sensor AKAR4,²¹ which provides a good approximation of the PKA-dependent substrate phosphorylation dynamics. We opted for 2 different compartments, the mitochondria and PM, and as a control, we used the soluble AKAR4 parent sensor. The outer mitochondrial membrane was chosen because it hosts a cAMP/PKA microdomain built around D-AKAP2³⁸ and subject to specific regulatory modalities.¹¹ Moreover, it was recently proposed that PKA microdomains build around D-AKAP2 may be important for cAMP buffering and diffusion.¹⁰ On the other hand, the PM compartment was chosen because it hosts most of the ACs and therefore is expected to be the domain with the highest cAMP levels both at rest and during stimulation. We used 2 well-calibrated sensors, OMM-AKAR4¹¹targeted by a generic tag to the outer mitochondria membrane, and the lipid raft-binding Lyn-AKAR4 as a PM-targeted sensor.²¹

To test how cAMP production and hydrolysis affect PKA-dependent phosphorylation in different compartments, we subjected HeLa and U-2 OS cells to the experimental protocol depicted in Figure 3A. As shown in Figure 3B, the soluble AKAR4 sensor responded differently in HeLa and U-2 OS cells, faithfully mirroring the cAMP kinetics previously measured by H187 in the 2 cell lines (Figure 1A, B, and C). Similarly, treatment with low doses of FSK (5 µm) resulted in increased PKA-dependent phosphorylation at the outer mitochondrial membrane only in U-2 OS cells (Figure 3C). Finally, the addition of IBMX to the experimental medium resulted in strong (saturating) activation of both AKAR4 and OMM-AKAR4 (Figure 3B and C). These results confirm that PKA activation in the cytosol and at the outer mitochondria membrane depends on the cAMP levels and is in line with previous reports indicating that these compartments do not present differences in their cAMP managing ability.^{11,  13} Contrary to the soluble and OMM sensors, Lyn-AKAR4 responded to FSK 5 µm in both HeLa and U-2 OS cells (Figure 3D). Interestingly, in U-2 OS, saturation of Lyn-AKAR4 was achieved independently of PDE inhibition, while in HeLa cells, the addition of IBMX induced a further response of Lyn-AKAR4, confirming the superior hydrolysing ability of these cells even in the vicinity of the production site. These experiments confirmed the importance of cAMP managing for the activation of compartmentalized PKA. Indeed, in HeLa cells, where cAMP is kept under strict control by PDEs, the addition of IBMX was necessary for fully releasing PKA-dependent phosphorylation induced by low doses of FSK. Interestingly, this was evident at the outer mitochondrial membrane and cytosol but not the PM, where the penetrance of PDE activity was evident albeit partial, probably due to the vicinity to the cAMP producing sites. Indeed, it is important to note that activation of PKA at the PM in response to 5 µm FSK was not predicted by measurements of the cAMP levels in this compartment (Figure 2A). This discrepancy may be due to different localization of the 2 sensors used or could denote the presence at the PM of less phosphatase pressure or high cAMP affinity RI-based PKA holoenzymes with lower activation thresholds. On the other hand, in U-2 OS cells, where PDE activity appears negligible, all treatments resulted in saturating PKA responses further confirming the importance of cAMP hydrolysis in PKA activation.

Figure 3.

Protein kinase A (PKA)-dependent phosphorylation of different domains is dictated by global cyclic AMP (cAMP) levels in HeLa and U-2 OS cells. (A) Schematic representation of the experimental strategy used. (B) HeLa cells (dark traces) and U-2 OS cells (light traces) expressing the PKA-dependent phosphorylation FRET-based sensor AKAR4, its mitochondrial targeted version OMM-AKAR4 (C), or its Lipid raft targeted version Lyn-AKAR4 (D), were treated with low doses (5 µm) of forskolin (FSK) alone or combined with IBMX (500 µm), followed by the addition of FSK (20 µm) to saturate the sensor. PKA-dependent phosphorylation kinetics mirrored those of the cAMP levels, except for the plasma membrane (PM) domain in HeLa cells. Traces are expressed as the average ± SEM; AKAR4: n = 43 HeLa cells and n = 18 U-2 OS; OMM-AKAR4: n = 21 HeLa cells and n = 50 U-2 OS; Lyn-AKAR4: n = 39 HeLa cells and n = 20 U-2 OS).

PKA Termination Dynamics are Similar in HeLa and U-2 OS Cells

PKA microdomains are defined in subcellular locations where the kinase is tethered by AKAPs in the vicinity of both its targets and the machinery that controls its activation kinetics and range of action.³⁹ Our experiments suggest that cAMP increases drive PKA activation, which is then translated to target phosphorylation and eventual cellular responses. However, the inverse does not apply, since PKA deactivation in response to cAMP hydrolysis limits the ex-novo phosphorylation of targets but does not affect those already phosphorylated. In fact, to terminate the cellular responses induced by PKA is necessary to dephosphorylate its targets through the actions of phosphatases.^{11,  13,  40} Based on these considerations, it could be suggested that PKA substrate dephosphorylation is a complex process, initiated by the inactivation of the kinase, but completed by the actions of phosphatases. Thus the rate of dephosphorylation could define the extent of cellular responses induced by cAMP/PKA microdomains.

To test the role of phosphatases on the termination of local PKA-dependent phosphorylation, we employed the experimental protocol shown in Figure 4A. HeLa and U-2 OS cells expressing different versions of the AKAR4 sensor were challenged with high doses of FSK and IBMX, a treatment that leads to the saturation of all sensors. We then supplemented the experimental medium with H-89, an ATP analog that is well known to inhibit PKA and was expected to tilt the balance in favour of phosphatases. As shown in Figure 4B, AKAR4 responded to treatment with FSK (20 µm) and IBMX (500 µm) with an increase in FRET that reached similar levels of saturation in both cell lines (Figure 4Bii) and was slightly faster in HeLa cells (Figure 4Biii). Addition of H-89 (60 µm), which strongly inhibited endogenous PKA catalytic subunits, resulted in the dephosphorylation of the sensor, which was fast in both HeLa and U-2 OS and once more slightly faster in the former (Figure 4Biii). We repeated this protocol in cells expressing OMM-AKAR4. As shown in Figure 4Ci, the 2 sensors reached different saturation levels (Figure 4Cii), and HeLa cells once more displayed faster FRET rates of change (Figure 4Ciii). However, despite these differences, the dephosphorylation kinetics observed in both cell lines in response to H-89 were drastically slower than the ones measured by the soluble sensor and had similar kinetics, which we quantify in Figure 4Ciii. Interestingly, the kinetics of dephosphorylation observed at the lipid rafts under the PM using Lyn-AKAR4 in HeLa and U-2 OS cells were slightly different, with U-2 OS displaying slightly faster dephosphorylation of the sensor. The overall kinetics fell in between the rates measured for AKAR4 and OMM-AKAR4 (Figure 4D and quantified in Figure 4Diii). For Lyn-AKAR4, treatment with FSK and IBMX produced very similar results in the 2 cell lines both in terms of saturation levels and kinetics (Figure 4Di and quantified in Figure 4Dii and iii). Taken together, these experiments suggest that the activation kinetics of PKA depend on the cAMP levels. On the other hand, the dephosphorylation of PKA substrates both targeted and soluble, which determines the duration of the kinase’s actions, depends on phosphatase activity, and is decoupled from the messenger levels. In fact, the termination kinetics at the mitochondria were overlapping in both HeLa and U-2 OS despite their different cAMP managing modalities (Figures 1 and 2) and the different saturation levels reached (Figure 4C).

Phosphatases Impinge on the Range of Sub-PM PKA-dependent Events

Our previous experiments suggested that different compartments could display various degrees of phosphatase activity as measured by the decay rate of the signals measured by the targeted AKAR4 constructs. These data point to the phosphatases as important regulators of the duration of PKA-dependent events.

To establish the range within which phosphatases could influence the distribution of phosphorylated PKA substrates, we designed a series of AKAR4-based sensors located at different distances from their resident sites exploiting rigid spacers of known length.²⁰ As shown in Figure S4A, we sandwiched the SAH10 (10 nm) and SAH30 (30 nm) peptides between the targeting sequence and the sensor. Contrary to the outer mitochondrial membrane, where cAMP kinetics faithfully mirror those in the cytosol,¹¹ domains in the vicinity of PM display higher than expected PKA responses (Figure 3D) and different kinetics. These considerations prompted us to test the effects of phosphatases on the range of PM-PKA-dependent events. We generated and validated modified versions of the Lyn-AKAR4 containing the SAH10 and SAH30 spacers (Lyn-SAH10-AKAR4, Lyn-SAH30-AKAR4) (Figure S4Bi and validated in Figure S4C and D). As shown in Figure 5A, in HeLa cells Lyn-AKAR4 responded strongly to FSK but only reached saturation in response to IBMX. Interestingly, the responses of Lyn-SAH10-AKAR4 and Lyn-SAH30-AKAR4 to FSK were not as marked as those of their parent sensor, while IBMX addition resulted in a more evident increase in PKA-dependent phosphorylation, suggesting a stronger involvement of PDEs. Inhibition of PKACs by H-89 caused the recovery of the FRET signal in all sensors. To quantify the differences observed, we calculated the tau-parameter which is the time constant of the exponential decay of a curve for each sensor. As summarized in Figure 5B, the distance from the PM is associated with the rate of dephosphorylation of the sensors (tau value, τ). However, adding SAH10 or SAH30 resulted in a faster dephosphorylation rate when compared to Lyn-AKAR4, but these 2 constructs gave comparable results (Figure 5B). Similar kinetics were obtained also in U-2 OS cells, as shown in Figure 5C and summarized in Figure 5D.

Phosphatases Can Define the Range of OMM-related PKA-Dependent Events

Contrary to the domains in the vicinity of PM, the kinetics of cAMP at the outer mitochondrial membrane faithfully mirror those of the soluble compartment.¹¹ Previous work from our group in primary neonatal cardiac myocytes showed that PKA substrates at the OMM are less accessible to phosphatases and therefore stay phosphorylated for longer periods as compared to cytosolic targets.¹¹ To test whether phosphatases impinge on the range of OMM-based microdomains, we used SAH10 (10 nm) and SAH30 (30 nm) to generate, respectively, OMM-SAH10-AKAR4 and OMM-SAH30-AKAR4 (Figure S4A). Adding the 2 peptides had no effect on the targeting of OMM-AKAR4, and both new sensors localized strictly at the mitochondria (Figure S4Bii). All versions of the sensors retained their ability to measure PKA-dependent phosphorylation and dephosphorylation kinetics, as demonstrated in experiments using HeLa cells co-expressing OMM-AKAR4, OMM-SAH10-AKAR4, or OMM-SAH30-AKAR4 together with an mCherry-tagged active PKA catalytic subunit. As shown in Figure S4C, control cells expressing just mCherry responded to FSK/IBMX treatment reaching saturation (Figure S4Ci), while H-89 treatment showed no basal, unsolicited PKA activity to all constructs except Lyn-SAH10-AKAR4 and Lyn-SAH30-AKAR4 (Figure S4Cii). When PKAC-mCherry was expressed, cells did not respond to FSK/IBMX, likely due to saturation (Figure S4Ciii), while H-89 treatment generated large responses (Figure S4Civ), demonstrating their sensitivity to endogenous phosphatases. Finally, as summarized in Figure S4D, all sensors responded with FRET-ratio changes to PKAC overexpression, suggesting that the characteristics of the parent sensor were not significantly altered by the addition of spacers. The ability of AKAR4-based sensors to measure phosphatase-dependent dephosphorylation has been previously documented.^{11,  41,  42} To test the efficiency of CalA (50 nm) in inhibiting phosphatases, we treated cells expressing the soluble AKAR4 or its mitochondrial-targeted version OMM-AKAR4 with CalA, followed by FSK (20 µm) and IBMX (500 µm). Once the saturation levels were reached, we inhibited PKA with high doses of H-89. As shown in Figure S4E, the recovery of AKAR4, which is under strong pressure by phosphatases, became extremely slow but was not abolished indicating that 50 nm CalA did not suffice to inhibit all phosphatases. On the other hand, the recovery of OMM-AKAR4 was completely abolished as expected from a sensor that is under low phosphatase pressure.¹¹ When we attempted to increase the concentration of CalA to 100 nm cell stress was too great; therefore, we decided to use 50 nm for our experiments.

The responses to H-89 treatment obtained with Lyn-SAH10-AKAR4 and Lyn-SAH30-AKAR4 (Figure S4Cii) evidenced that, at least in the vicinity of the PM, phosphatases may participate in maintaining the basal PKA activity levels. To test whether this was true also for other compartments, we performed 2 sets of experiments. First, we treated unsolicited HeLa and U-2 OS cells with H-89 and performed Western Blotting using an antibody that recognizes PKA-phosphorylated substrates.¹¹ As shown in Figure S5A and quantified in Figure S5B, treatment with H-89 produced a detectable but very small effect as compared to vehicle treated cells. Treatment with lambda protein phosphatase (λ-PP) largely abolished the phospho-antigens (negative control), while treatment with saturating levels of FSK (20 µm) together with IBMX (500 µm) resulted in high antibody signal (positive control). We next performed a set of experiments aiming to directly measure the involvement of phosphatases in the regulation of basal PKA activity at different compartments. Unstimulated HeLa cells expressing different targeted sensors were treated with the phosphatase inhibitor CalA (50 nm). As shown in Figure S5C left panel, the PM-targeted sensor Lyn-AKAR4 did not respond to this treatment, while its SAH-based versions and the soluble AKAR4 produced a very small and slow response. On the other hand, none of the mitochondrial sensors (neither OMM-AKAR4 nor its SAH-based versions) responded to acute CalA treatment (Figure 5C right panel). These experiments suggest that in unsolicited HeLa and U-2 OS cells the basal levels of PKA are generally low, except in the vicinity of the PM, where a small but detectable PKA activity was present. Interestingly, the non-responsiveness of Lyn-AKAR4 could be explained by the presence of local phosphatases. Our data suggest against a broad role of phosphatases in the maintenance of basal PKA activity. This is not surprising, in fact most likely the basal PKA activity depends on the efficient hydrolysis of cAMP by PDEs, which maintains the levels of the messenger under the activation threshold of PKA.

Next, we used our sensors to estimate the reach of OMM-based cAMP/PKA microdomains in cells with high (HeLa) and low (U-2 OS) PDE activity. To determine the eventual contribution of ACs, PDEs, and phosphatases, we designed a protocol based on sequential manoeuvres of AC activation, PDE inhibition, and PKA inhibition (Figure S5D). As expected, in HeLa cells (Figure 6A), activation of ACs using high doses of forskolin produced small responses that were greatly increased after PDE inhibition using IBMX, independently of the sensor. Once saturation was reached, we added H-89 to the experimental medium that shifted the balance in favor of the phosphatases and resulted in the recovery of the FRET signal. Thanks to this manoeuvre, we ensured that cAMP levels remained high and only PKA catalytic subunits were inhibited. Interestingly, the recovery kinetics of different sensors were not the same, with soluble AKAR4 being the fastest, the OMM-AKAR4 being the slowest, while OMM-SAH10-AKAR4 and OMM-SAH30-AKAR4 were somewhat in between. As summarized in Figure 6B, we found that the dephosphorylation kinetics of OMM-SAH10-AKAR4 and OMM-SAH30-AKAR4 became faster (calculated by tau) as the distance from the mitochondria increased. We next employed the same protocol in U-2 OS cells expressing the different versions of the AKAR4-based sensors. In line with low PDE activity, FSK treatment suffices to saturate all the sensors (Figure 6C), while H-89 addition induced the recovery of the FRET signals with kinetics that were the fastest in the soluble fraction (AKAR4) and the slowest at the mitochondria (OMM-AKAR4), while becoming progressively faster by increasing the distance from the mitochondria (Figure 6D). All together, these experiments link the activity of phosphatases to the range of PKA-dependent phosphorylation around a specific domain and suggest that dynamic dephosphorylation of PKA substrates can define both the duration and range of PKA-dependent microdomains at the outer mitochondrial membrane.

Phosphatases Targeted at Increasing Distances from the OMM Trigger Distinct Signaling and Functional Outcomes

Our previous experiments suggested that the range of action of OMM-based microdomains would be contained by phosphatases at approximately 30 nm from the mitochondrial surface. We reasoned that, by targeting active phosphatases at different distances from the OMM, we could finely perturb the range of PKA microdomains. We next used the SAH10 and SAH30 spacers to target phosphatases at different distances from the outer mitochondrial membrane (OMM-SAH10-PP2A and OMM-SAH30-PP2A). To functionally test these constructs, HeLa cells expressing OMM-AKAR4 were co-transfected with differently spaced-out versions of catalytically active PP2A¹¹ or its soluble version as control and FRET experiments were performed using the protocol of Figure S5D. As shown in Figure 7A and quantified in Figure 7B, co-expression of both soluble PP2A or OMM-PP2A (no spacer added¹¹) accelerated the dephosphorylation rate of OMM-AKAR4. On the other hand, co-expression of OMM-SAH10-PP2A had a smaller effect on the dephosphorylation rate of OMM-AKAR4 as compared to OMM-PP2A. Finally, co-expression of OMM-SAH30-PP2A did not affect the dephosphorylation kinetics of OMM-AKAR4 (Figure 7A and quantified in Figure 7B). Interestingly, when we measured the effects of the expression of targeted PP2A constructs on the kinetics of OMM-AKAR4 in response to FSK (Figure S6A), we found that overexpression of OMM-SAH30-PP2A and OMM-PP2A had different “on” kinetics when compared, while in response to FSK/IBMX, the difference was evident between OMM-SAH10-PP2A and OMM-PP2A (Figure S6B). Based on these experiments, mitochondrial PKA microdomains are contained within a 30 nm distance from the organelle. To determine whether this range is functionally relevant, we tested mitochondrial fragmentation in HeLa cells co-transfected with the OMM-targeted PP2A constructs together with a mitochondrial red fluorescent protein (mtRFP). It is well established that PKA-dependent phosphorylation inhibits the mitochondrial fission regulator Dynamin related protein 1 (Drp1) and results in elongated organelles.^{11,  43,  44} As shown in Figure 7C and quantified in Figure 7D, the mitochondria of HeLa cells transfected with mtRFP were elongated and presented no fragmentation. On the other hand, when OMM-PP2A was overexpressed, we observed high levels of mitochondrial fragmentation. Interestingly, neither OMM-SAH10-PP2A nor OMM-SAH30-PP2A expression resulted in mitochondrial fragmentation. These experiments are in line with the signaling measures obtained in Figure 7A, where OMM-SAH10-PP2A and OMM-SAH30-PP2A had very weak effects on PKA signaling at the OMM. All together, our results further corroborate the functional constrain of mitochondrial PKA by phosphatases.

Discussion

PKA compartmentalization is a complex task that requires the coordination of 2 types of machineries, one that guarantees the vicinity of the kinase to its designated targets and another in charge of its activation/deactivation kinetics. Cyclic AMP is generally produced at the PM and must reach intracellular PKA moieties by diffusion. The levels of cAMP able to reach a specific microdomain depend on multiple factors, notably PDEs²⁰ and intracellular buffers.¹⁰ PKARs are the most relevant cAMP buffers, as they can bind cAMP without the consequent activation of the holoenzymes, because they are produced in excess (∼17-fold) compared to the PKACs.^{15,  16} Interestingly, type I regulatory subunits in their free state display higher affinity for cAMP as compared to the same subunits bound to the catalytic ones²⁵ and may redistribute to generate biomolecular condensates that act as highly localized cAMP sinks.³⁴

In the general notion of PKA microdomains, the structural basis of PKA segregation within specific compartments represents a paradox. In fact, while AKAPs offer a structural platform for the tethering of PKA tetramers, they do so by complexing with the regulatory subunits of the tetramer. This feature, however, does not ensure the strict confinement of the PKACs that, in response to cAMP elevations, are released from the regulatory subunits and become free to diffuse away from their site of activation. Recently, it was suggested that AKAP-bound PKA tetramers remain intact independently of their activation status.¹⁴ However, this model and the general idea that PKACs remain linked to PKARs when cAMP levels are high were challenged by subsequent studies.^{12,  15} Several mechanisms that limit PKAC diffusion have been proposed, including myristoylation,¹² binding to excess regulatory subunits,^{15,  37} and autophosphorylation of the inhibitory domain to regulate the affinity of RIIs for the catalytic subunits.^17–19 The latter appears to be a nodal regulatory point of crosstalk between PKA and phosphatases. Indeed, when dephosphorylated, the affinity of RII subunits for the catalytic ones is significantly increased. Interestingly, PKA can also interact with the regulation of phosphatases. Particularly interesting is that PKA phosphorylates the protein dopamine and cyclic AMP-regulated phospho-protein, relative molecular mass 32 000 (DARPP-32) at threonine-34, converting it into a potent inhibitor of PP1. Decreased activity of PP1 allows PKA-mediated phosphorylation events to persist, amplifying downstream signaling cascades. Conversely, PP2A dephosphorylates DARPP-32 at threonine-34, reversing its inhibitory effect on PP1 and restoring PP1 activity. Interestingly, PKA is part of a regulatory mechanism where it phosphorylates the regulatory subunit B56δ of PP2A increasing its activity and consequently the dephosphorylation of DARPP-32.^{45,  46} This crosstalk highlights the importance of dynamic phosphatase and kinase interactions in fine-tuning cellular responses to extracellular signals. Despite these mechanisms, it is unequivocal that, once released from the tetramer, the degree of freedom of PKACs increases and, consequently, the probability to encounter and phosphorylate substrates that are not strictly part of the specific microdomain. In this manuscript, we investigated the existence of mechanisms that can limit the range of action of PKA microdomains by acting on the targets of PKACs rather than on their physical confinement and the phosphorylation status of their substrates. We reasoned that the regulation of cAMP/PKA microdomains goes beyond their activation/inactivation kinetics and set out to understand the distinct contributions of cAMP levels and phosphatases in determining the temporal and spatial range of a microdomain’s actions.

We used 2 cellular models, HeLa and U-2 OS, and found that cAMP kinetics measured at the production site were faithfully mirrored in the cytosol, independently of the cell type and therefore of the regulatory mechanism. Our data suggest that generalized cAMP increases/decreases were the sole determinants of PKA activation/inactivation in all the subcellular compartments investigated. Using soluble and targeted versions of AKAR4, a FRET-based sensor that mirrors the phosphorylation/dephosphorylation kinetics of PKA substrates,¹¹ we found that the termination pace of PKA-dependent events was likely determined by the accessibility of phosphatases within the microdomain. In addition, measurements of PKA-dependent phosphorylation persistence at known distances from specific domains (mitochondria and PM) suggested that phosphatase pressure increases in direct relationship to the distance from the microdomain. Our data, together with recently published evidence^{11,  20,} support a model where the levels of PDEs and buffers in the vicinity of a microdomain will determine its activation by controlling the levels that the messenger can reach. Indeed, the higher the hydrolysing capacity of a location, the more efficiently cAMP levels will drop and PKA will return to its inactive tetrameric state. On the other hand, translating of cAMP-induced PKA activation to cell function requires, and is driven by, the reversible phosphorylation of substrates. Interestingly, once a target is phosphorylated, its regulation is decoupled from the levels of the messenger and the status of the kinase. In fact, cellular responses triggered by PKA activation will subside only after dephosphorylation of the substrates, thanks to the actions of phosphatases. Indeed, once a target is phosphorylated by PKA, the only possible subsequent action is dephosphorylation by a phosphatase. This reflects a binary regulatory mechanism in the kinase-phosphatase equilibrium that can be cyclical and highlights the irreversible and mutually exclusive nature of phosphorylation and dephosphorylation. In the case of cAMP/PKA microdomains additional levels of complexity exist. For instance, the rates of PKA and phosphatase activity can vary, and targets may encounter PKA or phosphatases under specific spatial and temporal conditions. The latter is particularly relevant in the case of microdomains, the “core” of which are richer in PKA thanks to the AKAP (except for AKAPs that can recruit PPs), while in the periphery more phosphatases are encountered. Our data suggest that the rate of target dephosphorylation depends on the spatial distribution of phosphatases and likely determines both the temporal and spatial range of a microdomain’s actions.

From a functional point of view, we provide evidence that a contest where our findings would be physiologically relevant is the regulation of mitochondrial dynamics. The OMM is the interface of mitochondria with the cytosol and is the domain where resident and cytosolic proteins, many of which are PKA substrates, participate to important processes for the maintenance of mitochondrial homeostasis.^{47,  48} Dynamin-related protein 1 (Drp1) is a protein involved in mitochondrial dynamics as a major regulator of mitochondrial fission.⁴⁹ The actions of Drp1 are strictly regulated by several post-translational modifications. In particular, its pro-fission activity is inhibited by PKA-dependent phosphorylation at serine 637,^{43,  50} while is enhanced by the dephosphorylating actions of calcineurin.⁵¹ Based on our findings, dephosphorylated Drp1 molecules will tend to engage mitochondria to promote fission; however, as they reach the OMM phosphatase, pressure decreases, and consequently the likelihood of Drp1 being phosphorylated by OMM-resident PKA enzymes increases. Conversely, phosphorylated Drp1 released from the OMM and migrating toward the cytosol will encounter increasing phosphatase activity and likely revert to its unphosphorylated, active state. Considering the importance of compartmentalization in the specificity of the action of several kinases and phosphatases, it is reasonable to speculate that the mechanism described here may be relevant for the spatiotemporal regulation of other kinases as well. The lower phosphatase activity at the outer mitochondrial membrane could be explained by several mechanisms. Our present and previous data¹¹ do not provide evidence for the involvement of OMM-targeted phosphatases, in fact, slow dephosphorylation of PKA targets at the OMM indicates the opposite. Therefore, we reasoned that interactions between phosphatases and phosphoproteins can be described as diffusion-limited reactions. Based on this hypothesis, immobilized PKA targets at the OMM are expected to exhibit reduced rates of dephosphorylation for several reasons. First, in the absence of local enzymes, dephosphorylation at the OMM would require cytosolic phosphatases to diffuse to this compartment. Second, an immobilized substrate is expected to have restricted rotational freedom, which would limit the rate of viable interactions with phosphatases. Third, phosphatases may have reduced accessibility to the OMM due to steric or physicochemical hindrance. Finally, our experiments cannot exclude the presence of a local phosphatase inhibitor at the mitochondria.

Author Contributions

Filippo Conca (Conceptualization, Data curation, Investigation, Methodology, Writing – review & editing), Doruk Kaan Bayburtlu (Conceptualization, Data curation, Formal analysis, Methodology, Software, Writing – review & editing), Mauro Vismara (Formal analysis, Investigation), Nicoletta C. Surdo (Data curation, Investigation), Alessandra Tavoni (Data curation, Investigation), Leonardo Nogara (Formal analysis), Adamo Sarra (Investigation), Stefano Ciciliot (Investigation), Giulietta Di Benedetto (Investigation), Liliana F. Iannucci (Conceptualization, Formal analysis, Investigation, Writing – review & editing), and Konstantinos Lefkimmiatis (Conceptualization, Funding acquisition, Project administration, Supervision, Writing – original draft)

Funding

We are grateful for funding from the Human Frontier Science Program Research Grant (HSFP #RG0024/2022), the AIRC Foundation for Cancer Research (Grant# IG2021 ID 26140), and the AFM Telethon Research Grant (#25089) to K.L.

Conflict of Interest

None declared.

Data Availability

Data presented in this article were generated in the authors’ laboratories and can be made available upon request.