Activation of PKA in cell requires higher concentration of cAMP than in vitro: implications for compartmentalization of cAMP signalling

cAMP is a ubiquitous second messenger responsible for the cellular effects of multiple hormones and neurotransmitters via activation of its main effector, protein kinase A (PKA). Multiple studies have shown that the basal concentration of cAMP in several cell types is about 1 μM. This value is well above the reported concentration of cAMP required to half-maximally activate PKA, which measures in the 100–300 nM range. Several hypotheses have been suggested to explain this apparent discrepancy including inaccurate measurements of intracellular free cAMP, inaccurate measurement of the apparent activation constant of PKA or shielding of PKA from bulk cytosolic cAMP via localization of the enzyme to microdomains with lower basal cAMP concentration. However, direct experimental evidence in support of any of these models is limited and a firm conclusion is missing. In this study we use multiple FRET-based reporters for the detection of cAMP and PKA activity in intact cells and we establish that the sensitivity of PKA to cAMP is almost twenty times lower when measured in cell than when measured in vitro. Our findings have important implications for the understanding of compartmentalized cAMP signalling.

any PKA-dependent response. This is clearly in contrast with common observations and a number of hypotheses have been put forward to explain such inconsistency. In early studies the finding that the amount of cAMP in tissue homogenates is the same in control and hormone-treated samples, and apparently sufficient to fully activate PKA, led to the hypothesis that the majority of cAMP must be bound or sequestered and that the metabolically active second messenger is only a small fraction of the total amount 12 . More recently, intracellular free cAMP has been directly measured using FRET-based reporters and micromolar basal cAMP has been confirmed in a variety of cell types. An alternative hypothesis put forward is that the apparent activation constant value for PKA determined in vitro is an underestimation due to the vast excess of cAMP over pure enzyme normally used in the artificial in vitro set up 6 . Compartmentalization of PKA in domains with significantly lower basal cAMP than in the bulk cytosol has also been suggested 13 . Direct experimental evidence in support of any of these models is, however, very scarce. In this study we set out to resolve the apparent discrepancy between basal cAMP levels and PKA activation threshold using direct in cell determination of cAMP concentrations and PKA activation.

Results
Determination of free cAMP concentration in the bulk cytosol of intact cells. We first considered the possibility that previously published values for intracellular basal cAMP may be inaccurate due to methodological limitations. Earlier quantification protocols using assays that measure total cAMP in cell lysates or tissue homogenates 8 may significantly overestimate the amount of free cAMP available to trigger PKA activation. Although free cAMP in intact cells has recently been measured using FRET-based reporters 9,14 , the protocols used relied on the assumption that the dynamic range and/or affinity of the reporter is the same in the intact cell as measured in vitro 9,14 . We recently developed a calibration protocol for FRET-based reporters that uses microinfusion of known concentrations of cAMP in intact cells via a patch pipette 15 . This approach, which results in homogeneous intracellular distribution of cAMP (Supplementary Figure S1) gives the opportunity to control parameters such as ionic composition, ionic strength and pH in the patch pipette and to match them closely to physiological intracellular values (see materials and methods). Using this approach, concentration-dependency curves can be generated without any assumption based on data acquired in vitro. With this method the x-crossing of the cAMP concentration-dependency curve (zero FRET change) indicates the equivalence between the micro-infused cAMP concentration and the intracellular basal cAMP. To exclude that the FRET measurements may be influenced by artificial fluorescence changes (for example secondary to cAMP-dependent variations in the intracellular ionic composition that may affect fluorophore emission), we compared dose-dependency curves of two different FRET reporters, CUTie 16 and EPAC-S H187 . cAMP binding to these two sensors results in an opposite change in FRET (see Fig. 1a,b). We calculated cAMP changes as CFP emission/YFP emission for EPAC-S H187 and YFP emission/CFP emission for CUTie. Therefore, any artificial change in fluorescence occurring during the measurements would affect the ratio value determined with these sensors in an opposite manner.
As illustrated in Fig. 2, our measurements show that the x-crossing of the concentration dependency curve is around 1 μM cAMP for both CUTie and EPAC-S H187 . To exclude the possibility that the microinfusion approach may trigger the synthesis of cAMP and affect basal cAMP levels, we measured FRET changes in saponin permealised, EPAC-S H187 expressing CHO cells to allow free diffusion of known concentrations of bath applied cAMP and found matching results (Supplementary Figure S2). Our in cell determination of basal free cAMP is therefore in agreement with previously reported values.
Determination of basal cAMP levels at AKAP79. Based on computational studies the hypothesis has been put forward that PKA localises within microdomains where the basal concentration of cAMP may be maintained at significantly lower levels than in the bulk cytosol, thus protecting PKA from activation in the absence of extracellular stimuli 13 . To test this hypothesis we used the sensor AKAP79-CUTie 16 (Fig. 1c). In this variant, the sensor is genetically fused to AKAP79 17 , a prototypical scaffolding protein that anchors PKA in a signalosome at the plasma membrane. When expressed in CHO cells, AKAP79-CUTie shows the expected localization (Fig. 3b), unlike CUTie which is homogeneously distributed in the cytosol (Fig. 3a). We previously demonstrated that when fused to the cAMP reporter AKAP79 maintains the ability to anchor PKA 18 and that AKAP79-CUTie is incorporated into the correct signalosome 16 . Therefore, the cAMP concentration detected by this sensor is the level experienced by a PKA subset localized to a defined microdomain. When we microinfused CHO cells stably expressing AKAP79-CUTie with known concentrations of cAMP to generate a cAMP-FRET concentration-dependency curve for this sensor we found that the calculated x-crossing again is around 1 µM (Fig. 3c), indicating that PKA anchored to AKAP79 at the plasma membrane experiences the same basal cAMP concentration as found in the bulk cytosol.
In cell determination of the activation threshold of overexpressed PKA. We next investigated the extent of PKA activation at basal cAMP. For this we used the C9H6 FRET reporter 19 which is based on the regulatory (R) and catalytic (C) subunits of PKA fused to YFP and CFP, respectively (Fig. 1d). With increasing cAMP concentrations, this sensor dissociates in free R-CFP and C-YFP subunits leading to loss of FRET signal. We previously demonstrated that fusion of the PKA subunits to CFP and YFP does not affect cAMP binding to the holoenzyme or its enzymatic activity 11 . When measured in vitro the apparent activation constant (EC 50 value) for C9H6 was 0.3 μM 20 , in agreement with in vitro values reported for native PKA. However, when CHO cells stably expressing C9H6 were microinfused with 1 µM cAMP, a concentration that, based on the in vitro EC 50 value, should almost maximally activate PKA, no FRET change was detected (Fig. 4a). Microinfusion of 3 µM cAMP induced only minor FRET changes, whereas 10 µM cAMP induced a significant increase. Microinfusion of 1 mM cAMP saturated the sensor. The concentration-dependency curve shows that, when measured in cell, the apparent activation constant of C9H6 is 5.2 μM, almost 20 times higher than the value measured in vitro (Fig. 4b).
Scientific RepoRts | 7: 14090 | DOI:10.1038/s41598-017-13021-y Evaluation of endogenous PKA activation threshold. Previous in vitro measurements showed that the apparent activation constant of PKA is influenced by the concentration of the enzyme used in the assay, with increasing concentrations of cAMP being necessary to activate increasing concentrations of the purified enzyme to the same extent 6 . In our experimental conditions we expect the concentration of overexpressed C9H6 to exceed that of endogenous PKA. To assess whether this may explain the unexpectedly high EC 50 values found for C9H6 we used AKAR3 (Fig. 1e), a FRET-based PKA activity reporter, to measure the activity of endogenous PKA in intact cells. AKAR3 is activated by PKA-mediated phosphorylation and de-activated by dephosphorylation (see Fig. 1e). Therefore, a significant basal activity of endogenous PKA would be expected to result in a detectable change in the FRET signal reported by AKAR3 upon cell treatment with the PKA inhibitor H89. We found that while application of 10 µM H89 completely reversed the AKAR3 FRET change induced by activation of adenylyl cyclases with 1 µM forskolin (Fig. 5a), there was no detectable change in the AKAR3 FRET signal when H89 was applied in otherwise untreated cells (Fig. 5a). These results indicate that there is no significant activity of endogenous PKA at basal levels of cAMP. To exclude that the activity of phosphatases may mask any basal PKA-dependent phosphorylation of AKAR3 we treated CHO cells expressing AKAR3 with phosphatase inhibitors. Treatment with either calyculin A (10 nM) or cyclosporin A (200 nM) revealed no significant dephosphorylation of AKAR3 (Fig. 5b).
To assess the activation threshold of endogenous PKA we next generated a concentration-dependency curve for AKAR3 phosphorylation using the cAMP microinfusion method. As shown in Fig. 5c,d, AKAR3 shows a steep cAMP concentration-dependency and the cAMP concentration required for half-maximal PKA-dependent phosphorylation of AKAR3 is about 2.1 µM. Extrapolating from this curve, the level of AKAR3 phosphorylation at 1 µM cAMP can be estimated to be about 10% of maximum, confirming minimal activity of endogenous PKA.
In support of a significant difference in the activation threshold of PKA in vitro and in cell, when we assessed PKA-dependent phosphorylation in lysates obtained from CHO cells we found that addition of 1 µM cAMP to the lysate was sufficient to maximally phosphorylate the PKA targets troponin I and CREB (Fig. 6).
In line with the above findings, when we measured the FRET change generated by the PKA-based FRET sensor C9H6 in a cell lysate rather than in intact cells, we found that application of 1 µM cAMP results in 80% maximal FRET change. Further application of 10 µM cAMP only slightly increased the FRET signal and no additional FRET change was observed at 100 or 1000 µM cAMP (Fig. 7).

Discussion
In this study we demonstrate that the activation threshold of PKA in the intact intracellular environment is significantly higher (at least one order of magnitude) than previously thought based on in vitro measurements. Our findings resolve the incongruity of basal cAMP concentration being sufficient to almost maximally activate PKA.
The possibility that considerable differences may exist between the control of enzyme reaction rates in vitro and in vivo has long been appreciated [21][22][23] . This has been ascribed, at least in part, to the fact that in vitro measurements are usually conducted with a large excess of substrate over enzyme, a condition that does not necessarily apply to enzymes in their intracellular environment. For example, in vitro measurements showed that while 0.3 μM cAMP is necessary to half-maximally activate 0.009 μM PKA, five-fold higher concentration of cAMP (about 1.5 μM) is necessary to activate to the same extent 0.15 μM PKA, a concentration of the enzyme that is closer to the estimated intracellular concentration of PKA (0.23 μM in skeletal muscle) 6 .
Here, by directly measuring cAMP levels and PKA activation in intact cells we directly assessed the apparent activation constant of PKA in cell and found that it is about twenty-fold higher than the value previously determined in vitro. A number of other factors, in addition to enzyme concentration, may contribute to the inaccuracy of the in vitro measurements, including artificially low ionic strength or low pH, both known to affect PKA activation 24 , as well as lack of physiological cofactors that may be present in the cellular environment but missing in the in vitro settings 6 . Although our study concerns PKA, it is reasonable to expect that a similar fault may apply to in vitro measurements reported for other enzymes.
Our findings have important implications for the understanding of cAMP signalling and the model of cAMP/ PKA compartmentalised signalling. A large body of evidence supports a role of PDEs in limiting cAMP homogeneous diffusion inside the cell. Several studies from different laboratories demonstrate that inhibition of PDEs disrupts the boundaries between cAMP pools and results in equilibration of cAMP levels across the cell, indicating that the ability of PDEs to degrade cAMP contributes to the compartmentalization of the second messenger [25][26][27][28][29][30] . One criticism of this model has been that when taking into account the apparent activation constant of PKA as determined in vitro (in the 0.1-0.3 μM range) and the reported K M and V max values for PDEs (see Table 1), it is difficult to envisage how PDEs may be able to maintain the concentration of cAMP below the activation threshold of PKA even at basal cAMP levels, let alone contribute to compartmentalization of the cAMP response to hormonal stimulation, when the levels of cAMP significantly increase 13,31 . Based on these considerations, it has been argued that PDEs cannot dictate what subset of PKA is activated in response to a given stimulus, simply because their activity is inadequate to reduce cAMP levels below the PKA activation threshold.
The problem is illustrated in Fig. 8 which shows the concentration dependency curves for selected PDEs and for PKA calculated from in vitro values found in the literature (for conversion and standardization of values see Materials and Methods). Synthesis rate of cAMP by adenylyl cyclases is also shown. The graph shows that, for example, only at a cAMP concentration of about 2 µM would one molecule of PDE3A or PDE2A be able to degrade cAMP rapidly enough to compensate for the production of cAMP by one active AC. Comparison of these curves shows clearly that none of the PDEs would be able to degrade cAMP fast enough to maintain the level of cAMP below the activation threshold of PKA as determined in vitro. If we consider an intracellular concentration of PKA of 0.23 µM (as measured in some cells types 6 ), to achieve effective degradation of cAMP a concentration of PDEs several order of magnitude higher would be required. However, estimates based on literature values and our own measurements indicate that the overall PDE concentration in cardiac myocytes is in the range of 0.22 µM (for estimation see Methods), insufficient to effectively degrade cAMP to the required level. In contrast, PDEs would Scientific RepoRts | 7: 14090 | DOI:10.1038/s41598-017-13021-y be able to maintain basal cAMP below the activation threshold of PKA as determined in cell (shown in blue in Fig. 8), even when cAMP synthesis is activated.
The high apparent PKA activation constant determined in cell has significant impact on the quantitative evaluation of cAMP/PKA signal transduction and it will be important to include this value in future computational studies that model cAMP signalling. In this respect, it is interesting to note that some in silico analyses 32 , including our recent study 16 , already make the assumption that the affinity of cAMP for PKA is about ten times lower than the in vitro reported values. Our findings also indicate that efforts should be made in the future to determine in cell the enzyme reaction rates of other components of the cAMP/PKA signalling pathway. For example, it would be of key importance to establish kinetic parameters for the PDEs in the intracellular environment. At the same time, caution should be used when drawing conclusions on the basis of parameters constrained by experimental data acquired in vitro.
Stable cell lines were generated using TransIT ® -LT1 transfection reagent (MIR 2300, Mirus) and plasmidic DNA Clones were selected using the infinite dilution method. AKAR3 33 and EPAC-S H187 34 FRET sensors were kind gifts from Jin Zhang (University of California, San Diego) and Kees Jalink (The Netherlands Cancer Institute, Amsterdam), respectively.

In cell calibration procedure by microinfusion. CHO-cells stably expressing the FRET sensors
were patch-clamped and simultaneously observed under FRET-excitation. After establishment of a tight seal between cell-membrane and patch-pipette ("Gigaseal") the membrane under the pipette-tip was ruptured and the whole-cell configuration was established. This configuration provides direct access from the pipette to the cytoplasm and cAMP can either diffuse from the pipette into the cell or vice versa, depending on the cAMP-concentration in the patch-pipette solution. FRET-ratio changes were monitored for different cAMP concentrations in the patch pipette and computed into concentration-dependency curves. Seal-and whole-cell resistances were monitored in parallel to ensure a permanently tight seal between pipette and cell-membrane. Seal  FRET measurements were performed with a Nikon Eclipse FN-1, equipped with an Opto-LED fluorescent light source (Cairn-Research), a Dual-View beam splitter (Optical Insights) and a CoolSnap HQ 2 camera (Photometrix). All images were acquired with a 40x/0.8 numerical aperture, long distance water dipping objective (Nikon). Excitation wavelength was 436 ± 25 nm, excitation/emission dichroic was 455 nm long pass. Emission light was split by a 505 nm longpass dichroic and filtered at 480 ± 15 nm for CFP emission and 535 ± 20 nm for YFP emission. Acquisition and analysis was performed using Optofluor (Cairn Research). All FRET measurements were background-subtracted and, if necessary, corrected for baseline drifts. FRET changes were determined at steady state indicated by a plateau of the ratio change.
As especially the YFP fluorophores are prone to quenching effects due to pH or Cl − , for a correct determination of basal cAMP-values it was crucial to determine the intracellular pH and the influence of different ionic conditions on the FRET signal to correct for artificial shifts. For determination of the intracellular pH, cells expressing the FRET reporter were placed in a "high K + " extracellular buffer (KCl: 140 mM, NaCl: 4 mM, MgCl 2 : 2 mM, CaCl 2 : 2 mM, Glucose: 15 mM, HEPES: 10 mM), adjusted to different pHs (range 6.8 to 8.0) with KOH or HCl. Nigericin (10 µM) and Valinomycin (5 µM) were added to permeabilize the membranes to H + and K + 35 and the change in FRET signal was recorded over time. pH-dependency curves were generated and a pH of 7.55 to 7.64 (depending on the different cell clones) was identified as the pH value that did not affect the FRET signal, indicating that this pH reflects the mean intracellular pH of the CHO cellline (adapted from 35 ). All intracellular buffers were subsequently matched to the pH of the cell clone under study.
For some measurements a KCl-based intracellular patch-clamp buffer containing 162 mM Cl − was used. Whereas the normal intracellular Cl − -concentration of CHO-cells should be around 22 mM 36 ,we performed FRET-experiments microinfusing CHO-cells with 22 or 162 mM Cl − , respectively, at various Figure 8. Comparison of the concentration-dependent activities of PKA (right ordinate) and the absolute activities of selected PDEs (left ordinate). "High" and "low" denote high and low affinity states of the respective PDE. Note that the activity of all PDE4 isoforms, except PDE4D in its high affinity state, are too low to be distinguishable at this scaling factor. All PDE curves are calculated according to the corresponding K M values reported in the literature (see Table 1) and assuming, as a first approximation, that K d ≈ K M ≅ EC 50 . The respective absolute activity was calculated from V max and used to set the "top" of the curves (see Table 1). "Bottom" was set to zero. As cAMP-degradation by PDEs is a simple dual molecule reaction without cooperativity a maximal Hill-coefficient of 1 was assumed. In vitro PKA curves (green) are recalculated according to 20 , in cell PKA activity (blue) is calculated according to values found in this study. Also shown is the production rate of adenylyl cyclase 39 . All activities are calculated as number of cAMP-molecules degraded/ generated in one second by one PDE/AC molecule.
cAMP-concentrations to control for possible quenching effects. These experiments showed that at high Cl − the cAMP-dependent FRET signal is shifted to 1.7% more negative values (mean of 4 different cAMP concentrations). FRET changes recorded in conditions of high Cl − were therefore corrected for this shift. The saponin permeabilisation protocol was as in 37 except for that the "intracellular buffer" was used in the bath.

Normalization of data from the literature
Conversion of V max into absolute activity. To allow comparison of the cAMP-degrading activity of different PDE's and to relate PDE activity to the production of cAMP by ACs we calculated the "absolute activity" as molecule(s) cAMP per molecule enzyme per second. As typically V max is expressed as "µmol substrate × mg enzyme −1 × minute −1 " and PDEs differ in their weight significantly, we first calculated the number of molecules per mg according to equation 1, using the conversion 1 KDa ≅ 10 6 mg.
The values for the molecular weight, K M , V max and the calculated absolute activities for selected PDEs are listed in Table 1. As the activity values are derived from V max they served as maximum ("top") values to calculate the corresponding curves in Fig. 8.
Assumptions for the conversion of K M into EC 50 . The Michaelis constant K M = (k cat + k r )/ k f differs from the dissociation constant K d = k r /k f regarding the influence of k cat . However, it might be assumed that for most enzymes K d will be close to the value of K M , as in most cases k cat is small compared to k r and k f . Therefore, as a first approximation, we can assume that a cAMP concentration leading to half-maximal binding will also lead to half-maximal reaction velocity (K M ≈ K d ). Thus, for PDEs K d is considered as equivalent to EC 50 .
Estimation of absolute numbers of selected components of the cAMP-signalling pathway. As some data were not available for CHO cells, we used data from neonatal cardiomyocytes for our estimations: Volume of neonatal cardiomyocytes: 819 µm 3 (1 day in culture) to 1532 µm 3 (3 days in culture) 38 . As our cardiomyocytes are typically measured after 2 days in culture, we assumed 1000 µm 3 , corresponding to 1 pl (1·10 −12 l).
Estimation of the number of PDE's per cell and effective degradation rate: Mongillo et al. 11 determined the total PDE activity of neonatal cardiomyocytes (in lysate assays) to be 102 ± 8 pmol/minute/mg protein. From our own experiments we can calculate that 1 mg protein is equivalent to about 4.285 million cells. This enables us to calculate the effective degradation rate for the overall PDE population in one cell to be 240,000 cAMP molecules per second.
If we assume that the majority of PDEs in neonatal cardiomyocytes consists of about 30% PDE3 and 60% PDE4 11 and we assign mean degradation activities of about 10 × s −1 for the PDE3 fraction (6.25-12.5 for PDE3A, 17.4 for PDE3B), 1.5 × s −1 for the PDE4 fraction (PDE4D, 60% of the PDE4 population, turnover 0.045 for non-activated, up to 2.37 for activated, all other PDE4s only 0.09 to 0.41) and 1 × s −1 for all other PDEs, we can estimate that the population of PDEs consists of about 7,200 PDE3 molecules, 96,000 PDE4 molecules (thereoff 1,800-2,400 PDE4B molecules 39 ) and 24,000 molecules for all other cAMP degrading PDEs. All activities are calculated from V max values as shown in Table 1. This means a total number of about 130,000 PDEs per cell, equivalent to a concentration of about 0.216 µM.
Estimation of the number of PKA-molecules per cell: Beavo et al. 6 calculated a concentration of up to 0.23 µM PKA in rabbit skeletal muscle cells. Assuming a similar concentration in cardiac myocytes and the cell volume to be 1 pl, this would be equivalent to about 138,000 PKA molecules per cell. Data availability. All datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.