Formation of inactive cAMP-saturated holoenzyme of cAMP-dependent protein kinase under physiological conditions.

The complex of the subunits (RIalpha, Calpha) of cAMP-dependent protein kinase I (cA-PKI) was much more stable (K(d) = 0.25 microm) in the presence of excess cAMP than previously thought. The ternary complex of C subunit with cAMP-saturated RIalpha or RIIalpha was devoid of catalytic activity against either peptide or physiological protein substrates. The ternary complex was destabilized by protein kinase substrate. Extrapolation from the in vitro data suggested about one-fourth of the C subunit to be in ternary complex in maximally cAMP-stimulated cells. Cells overexpressing either RIalpha or RIIalpha showed decreased CRE-dependent gene induction in response to maximal cAMP stimulation. This could be explained by enhanced ternary complex formation. Modulation of ternary complex formation by the level of R subunit may represent a novel way of regulating the cAMP kinase activity in maximally cAMP-stimulated cells.

increased muscle RI␣, showed deficient cAMP-stimulated CREB phosphorylation and CRE-dependent gene expression in muscle (6). We have previously observed relatively more holoenzyme-associated kinase than expected from the tissue cAMP content during the pre-replicative cAMP surge in the regenerating liver, in which both RI and RII were up-regulated (7). These observations suggest the possibility that RI or RII subunits may have a negative effect on cA-PK dissociation even at high cAMP concentrations. We used the CRE-luciferase reporter gene to probe for dissociation of cA-PKI and cA-PKII in intact cells. Nuclear translocation of the C subunit requires cA-PK dissociation and is considered a prerequisite for phosphorylation of the CREB/CREM family of nuclear transcription factors, and hence for cAMP stimulation of CRE-governed reporter gene expression (8 -10). We show that cells overexpressing either hRI␣ or hRII␣, even when maximally cAMP challenged, had decreased cAMP responsive gene induction, suggesting that cAMP produced incomplete dissociation of either isozyme in intact cells. We will also present evidence that the affinity between RI and C subunits in the presence of saturating cAMP is 1 to 2 orders of magnitude higher than hitherto assumed (11).
The presence of a substantial amount of ternary complex of cAMP, R and C in intact cells actualizes the unresolved issue of whether the cA-PK holoenzyme has any catalytic activity in the presence of cAMP (12). Several arguments have been provided in favor of this possibility. The cGMP-dependent protein kinase, which is highly homologous to cA-PK (13), is activated by cGMP without dissociation (12). The RII subunit of cA-PK type II mutagenized in the substrate motif was able to form holoenzyme without blocking the C activity (14). The cAMP-saturated cA-PKII holoenzyme was reported to be fully active (15), and this observation was linked to the fact that most cA-PK anchoring proteins (AKAPs) preferentially bind RII (16,17). In several cases the disruption of R binding to AKAPs blocks the cAMP control of specific substrate proteins (18). This is more easily explained if cA-PK is catalytically active while physically retained in a supramolecular complex with its substrate, than if activation involves dissociation of the C subunit from the AKAP-anchored R subunit. We show that neither cA-PKI nor cA-PKII had significant catalytic activity against synthetic or physiological substrate under near physiological in vitro conditions. Recombinant human phenylalanine 4-monooxygenase (PAH) and tyrosine 4-monooxygenase (TH) was from Dr. Torgeir Flatmark, University of Bergen, Norway. pUC7/RI␣ containing bovine RI␣ cDNA was kindly provided by Dr. Susan Taylor, University of California, San Diego (19). pGEX-KG/hRI␣ and pGEX-KG/hRII␣ containing full-length cDNA of human RI␣ and RII␣ fused to GST, was kindly provided by Dr. Kjetil Taskén, University of Oslo, Norway. His 6 hRI␣ and His 6 hRII␣ were constructed by subcloning the RI␣/RII␣ cDNA into pBAD (CLONTECH). For cell transfection, RI␣ and RII␣ cDNAs were subcloned into pcDNA (Invitrogen). The pCMV5-C␣ plasmid was a kind gift from Dr. Stanley McKnight, University of Washington, Seattle, WA. The luciferase reporter plasmid, pT81-4CRE-Luc was a kind gift from Dr. Marit Bakke, University of Bergen, Norway. The green fluorescent protein plasmid pGFP-C1 was from CLONTECH. Buffer A is a near physiological buffer with respect to pH, ionic strength, potassium, phosphate, and magnesium concentration, and consists of 15 mM Hepes, pH 7.2, 1 mM Na 2 PO 4 , 130 mM KCl, 0.3 mM ATP, 2 mM Mg(CH 3 COO) 2 , 0.3 mM EGTA, 1 mM EDTA, 0.1 mM dithioerythritol.
Cell Culture and Transfection-HEK 293 cells were seeded at a density of 2 ϫ 10 5 cells/cm 2 in a 6-well plate and transfected 24 h later by calcium phosphate precipitation with reporter plasmid (0.16 g of pT81-4CRE-Luc), various concentrations of pCMV5-C␣, pcDNA-hRI␣, or pcDNA-hRII␣. The total amount of plasmid was kept constant (2 g) by compensating with pCMV5 empty vector. The cells were washed once with phosphate-buffered saline 24 h after transfection, lysed in 80 l of 25 mM Tris, 1 mM EDTA, 10% glycerol, 1% Triton X-100, and 2 mM dithiothreitol, and assayed for luciferase activity. Some of the cultures had received treatment with cAMP elevating agents (30 M forskolin, 250 M isobutylmethylxanthine) and cAMP analogs (1 mM 8-chlorophenylthio-cAMP, 0.7 mM N 6 -monobutyryl-cAMP, and 0.7 mM N 6 -benzoyl-cAMP) the last 3 h before harvesting. The transfection efficiency of the HEK 293 cells was determined by replacing pT81-4CRE-Luc with 0.1 g of pGFP-C1, and visualization of green fluorescent cells 24 h after transfection.
Purification of Proteins-The C␣ subunit was purified from 12 kg of bovine heart. The 10,000 ϫ g supernatant after homogenization in 10 mM KPO 4 , 1 mM EDTA, 0.1 mM dithioerythritol was passed through P1cellulose, applied to DEAE-Sepharose (1.5 liters), washed with 80 liters of 55 mM KPO 4 , pH 6.8, 1 mM EDTA, 0.1 mM dithioerythritol, and C␣ eluted with 0.1 mM cAMP in 50 mM KPO 4 , 1 mM EDTA, 0.1 mM dithioerythritol. Active fractions were diluted 3-fold in water, chromatographed on SP-Sepharose, and eluted with a linear KCl gradient (0 to 500 mM). The 99% pure enzyme was applied on a hydroxylapatite column, and eluted in 5 ml of 600 mM KPO 4 at a concentration of 0.3 mM.
Recombinant human R subunits were harvested from transformed Escherichia coli BL21, bovine R subunit from E. coli E222. Bovine RI was purified by ammonium sulfate precipitation and DEAE-Sepharose chromatography. His-tagged R was purified on Ni 2ϩ -NTA chromatography, and GST-tagged R by GSH affinity chromatography. Thrombin cleavage to separate GST and R was performed with GST-R still bound to resin. For final purification the R subunits were subjected to FPLC size exclusion chromatography on a column equilibrated in buffer A without ATP.
Labeling of cAMP-dependent Protein Kinase Subunits, Determination of Fluorescence Energy Transfer (FRET), and Scintillation Proximity Assay-Commercially available C-FITC and R-TRITC had decreased specific kinase activity and decreased affinity for C subunit, respectively. We therefore labeled bovine C␣ subunit with FITC and bovine RI␣ subunit with TRITC according to Adams (2). The C␣ subunit was biotinylated when in holoenzyme complex to protect the R-C interaction face from modification. The cA-PKI holoenzyme (2 mg/ml) was mixed with 0.6 mg/ml EZ-Link Sulfo-NHS-LC-LC-Biotin (Pierce), and reacted for 0.5 h at room temperature. Free biotin-C␣ subunit was obtained by FPLC (Superdex 200) in 50 mM phosphate buffer with 30 M cAMP. The C-FITC and the biotinylated C subunit had the same molar activity (18 s Ϫ1 ) as unmodified C, and similar efficiency in releasing [ 3 H]cAMP bound to RI␣. Cyclic AMP bound to RI-TRITC with the same affinity as to unmodified RI, and the C subunit released [ 3 H]cAMP from complex with RI-TRITC and RI at similar rate (not shown). For determination of FRET, subunits incubated in a 0.5 ml cuvette, were excited (Xenon lamp) at 490 nm, and emission determined at 510 and 580 nm (by PTI photomultiplier) before and after addition of 50 M cAMP. For scintillation proximity assay, streptavidin-coated flashplates (PerkinElmer Life Science) were coated with the biotin-C␣ and washed with buffer A with 5 mg/ml serum albumin. RI␣ ( (22), and 32 P-RII and [ 32 P]TH detected by autoradiography after separation by SDS-PAGE. The level of C and R subunits of cA-PK in cell extracts was determined as described previously (7), except that the total level of R subunits (RI ϩ RII) was determined by the ammonium sulfate precipitation method (23).
Estimation of the Level of R and C Subunits of cA-PK, and of C Subunit Occupation by Substrate in Intact cAMP-stimulated Cells-A thorough study (24) showed equimolar expression of the R (RI ϩ RII) and C subunits of cA-PK in mammalian tissue, averaging 0.3 pmol/mg tissue, wet weight. This translates to an average intracellular concentration of subunits of about 0.5 M, assuming the extracellular space to occupy 15% of the tissue, and the subunits to distribute in 70% of the intracellular space. The presently studied HEK293 cells had 3.3 pmol of R subunit/mg of protein and 3.0 pmol of C subunit/mg of protein (see first paragraph under "Results"). Assuming that the cells contained 10% protein, this value translates to about 0.3 pmol of subunit/mg of cellular wet weight, which means that the untransfected HEK293 cells had an average level of cA-PK subunit expression.
The free C subunit encounters substrates in the cell, but the % substrate saturation is unknown. To obtain an estimate, the rate of phosphorylation of PAH in vitro in the absence of competing substrates, was compared with the rate in intact, cAMP-stimulated hepatocytes (21). The phosphorylation was about 8 times slower in the intact cell, suggesting that one-eighth of the C subunit pool was accessible for PAH phosphorylation, and that at most seven-eights of it (85-90%) was occupied by other substrates.

Enforced Expression of RI␣ or RII␣ Attenuates cAMP-induced CRE-mediated Gene
Induction-HEK 293 cells were cotransfected with pT81-4CRE-Luc and expression vector containing either RI␣ or RII␣, to study the effect of overexpressed RI and RII subunits on gene transcription via the cAMP-responsive element (CRE). The cells were stimulated with agents elevating the endogenous cAMP as well as potent cAMP analogs to ensure full saturation of the R subunits of cA-PK. It was found that cells overexpressing RI␣ or RII␣ had about half as strong luciferase induction by cAMP agonists as control cells (Fig. 1A). A similarly blunted luciferase induction was noted in RI␣ overexpressing cells exposed to potent acetoxy-methylated cAMP analogs (25,26) at 20-fold higher concentration than required for maximal luciferase induction in control cells (not shown).
There was no evidence that RI␣ expression had interfered nonspecifically with luciferase expression, since co-transfection with 400 ng of C␣ expression plasmid could override the effect (Fig. 1B). The effect of expressed exogenous RI and C subunits could be mutually titrated in cells not stimulated with cAMP agonists (Fig. 1, B and C), confirming that the CRE-Luc expression system responded to C␣ subunit and that the C␣ effect could be blocked by R subunit.
The luciferase induction was next compared between cells with near balanced coexpression of exogenous R and C␣ and cells with moderate overexpression of C␣ alone. Again, cells with enforced expression of RI␣ or RII␣ had lower response to cAMP challenge (Fig. 1D). This suggested that overexpressed R could inhibit the ability of exogenous (Fig. 1D) as well as of endogenous (Fig. 1A) C␣ to induce CRE-governed luciferase.
The observed effects of enforced expression of R subunits in cAMP-stimulated cells were statistically significant. The Wilcoxon signed-rank test showed significantly less cAMP-stimulated luciferase expression in cells overexpressing RI (p Ͻ 0.010; n ϭ 10) or RII (p Ͻ 0.025; n ϭ 6). When including results from cells coexpressing C␣ subunit, the p values were Ͻ0.001 (n ϭ 18) for RI and Ͻ0.005 (n ϭ 12) for RII. It is concluded that cells with enforced expression of RI␣ or RII␣ subunits of cA-PK had blunted response to maximal cAMP stimulation.
In separate experiments, in which the luciferase reporter gene was replaced by green fluorescent protein (GFP) as reporter, about 25% of the cells were detectably green 24 h after transfection. The transfection efficiency was not affected by co-transfection with RI␣, RII␣, or C␣ plasmids. GFP containing cells co-transfected with RI␣, RII␣, or C␣ plasmids at the concentration used in the experiments shown in panels A and D of Fig. 1 were tested for content of R (RI ϩ RII) subunit and C subunit. The average increase of RI, C, and RII was 2.4, 2.8, and 9.5 pmol/mg protein, respectively. In cells transfected with 400 rather than 4 ng of C␣ plasmid the protein kinase activity was increased 30-fold. The basal level of RI ϩ RII was 3.3 pmol/mg protein, and of C subunit was 3.0 pmol/mg protein.
The moderate overexpression of RI explained why it only pro-tected partially against 4 ng of C␣ plasmid and not at all against 400 ng of C␣ plasmid (Fig. 1B). It also suggested that RI might be more efficient than RII in preventing the C subunit to stimulate gene transcription in cells with strongly increased cAMP level.

cA-PKI Holoenzyme Can Form at Submicromolar Concentrations of RI␣ and C Subunits in the Presence of cAMP-
The results of Fig. 1 suggested that the RI subunit of cA-PKI could sequester the C subunit even in maximally cAMP-stimulated intact cells. This was unexpected in view of previous estimates of the dissociation constant of the complex between cAMPsaturated RI and the C subunit (11,27). It was therefore decided to study in detail the strength of the interaction between cAMP-saturated RI and C under physiologically relevant conditions using isolated protein kinase subunits.
One approach used the ability of C-FITC to enhance the emission of RI-TRITC and RI-TRITC to quench the emission of C-FITC when at close distance (FRET) (2). The fraction of C-FITC in complex with RI-TRITC was calculated from the relative emissions at 510 and 580 nm, as detailed in Table I. Using this method, an apparent equilibrium K d of 0.24 M was determined for the complex between RI-TRITC and C-FITC in the presence of a saturating concentration of cAMP (Fig. 2).
This K d value was surprisingly low in view of previous estimates, and could be due to enhancement of the R-C binding by the introduction of the fluorescent groups. In a second approach we used biotin-labeled C␣ and unlabeled RI subunit. The biotin labeling of the C subunit was performed when the C was in cA-PKI holoenzyme complex and the coupling chemistry was different from that used for the C-FITC labeling. Biotin-C␣ was immobilized to streptavidin-coated wells of a microplate with intrinsic solid scintillant (and scintillation proximity assay). The added RI⅐[ 3 H]cAMP complex was detected only when close to the wall (bound to biotinylated C). The immobilized C subunit was estimated to be half-maximally saturated by 0.16 M RI-[ 3 H]cAMP (Fig. 3), confirming the high affinity between RI-cAMP and C subunit.
A third method relied on unlabeled native R and C subunits. A trace amount of C␣ (50 pM) was injected into size exclusion FPLC columns equilibrated with 100 M cAMP and various concentrations of hRI␣. The position where the C subunit kinase activity eluted was determined. Half-maximal shift of C␣ elution position occurred when the column was equilibrated with 0.25 M hRI␣ (data not shown). It is concluded that the C␣ subunit interacts with cAMP-saturated RI␣ with a K d in the submicromolar range.
Cyclic AMP Saturated cA-PKI and II Lack Demonstrable Protein Kinase Activity-Since the ternary cAMP⅐cA-PKI complex (cAMP 4 , RI␣ 2 , and C␣ 2 ) apparently could form in intact cells (Fig. 1) and at submicromolar concentrations of cA-PKI subunits (Figs. 2 and 3), the question of the catalytic activity of the ternary complex has biological importance. We investigated this possibility using several substrates. In the first series of experiments the C␣-catalyzed phosphorylation of the physiological substrate phenylalanine-4-monooxygenase was studied at various concentrations of RI␣ in the presence of a very high concentration (100 M) of cAMP. At the highest concentration (30 M) of cAMP-saturated RI␣ studied the phosphorylation rate was inhibited more than 99% (Fig. 4A). This shows that C␣ in the ternary complex with cAMP-saturated RI expressed less than 1% of its potential phosphotransferase activity, and possibly had no activity at all. Half-maximal inhibition of the phosphorylation was observed at 0.32 M cAMP-saturated RI␣ (Fig. 4B), in line with the physicochemical data indicating a submicromolar K d for the complex between C␣ and cAMPsaturated RI␣ (see above). Qualitatively similar results were  4 g), or control vector. The increment due to cAMP challenge is shown by hatching. Panel B shows the luciferase activity in cells transfected with C␣ (4, 10, or 400 ng) with or without RI␣ vector (1.4 g). Panel C shows the titration of luciferase activity induced by C␣ (4 ng) by increasing RI␣ vector (0.4, 1.0, and 1.7 g). Panel D is similar to panel A, except that all cells were transfected with C␣ vector (4 ng). All cells were transfected with 0.16 g of pT81-4CRE-Luc in addition to plasmids as indicated in the panels and above. Nineteen hours thereafter they were exposed to cAMP challenge (30 M forskolin, 0.25 mM 3-isobutyl-1-methylxanthine, 1 mM 8-chlorophenylthio-cAMP, 0.7 mM N 6 -monobutyryl-cAMP, and 0.7 mM N 6 -benzoyl-cAMP) or vehicle, and harvested 3 h thereafter for determination of luciferase activity. The data are given as mean Ϯ S.E. (n ϭ 6) except for panel C which shows the mean Ϯ range (n ϭ 2-4). obtained when the C␣ activity was determined with tyrosine-4-monooxygenase as the substrate, using autoradiography to detect the phosphorylated protein (Fig. 5). The phosphorylation of tyrosine-4-monooxygenase in the presence of cAMP (100 M) was inhibited by at least 95% also by RII␣ (Fig. 5).
RI␣ and RII␣ (10 M) inhibited the kinase to a similar degree whether the free cAMP concentration was 50, 100, 300, or 600 M. This suggested that 50 -100 M cAMP, as routinely used, could saturate cA-PKI and II holoenzyme nearly completely.
Although cAMP-saturated cA-PKI or II holoenzyme had insignificant activity toward large physiological protein substrates (Figs. 4 and 5), the possibility remained that they could phosphorylate smaller substrates (15). In our hands, RI␣ (Fig.  6), as well as RII␣ (Fig. 7), could inhibit Kemptide phosphorylation nearly completely in the presence of cAMP. The inhibition was reversible and similar whether bovine or human RI␣ was used, and whether the R subunit was a GST fusion protein, a thrombin-cleaved product of the latter, or was polyhistidinetagged (not shown).  (Fig. 6). At 100 M Kemptide about 1 M RII␣ was required to half-maximally inhibit the kinase (Fig. 7). The K m value for Kemptide was determined to be 11 M, and at 100 M Kemptide the activity was about 90% of V max , suggesting 90% occupation of the C subunit by substrate (not shown). It is concluded that 1 M cAMP-saturated RI␣ or RII␣ is sufficient to achieve 50% kinase inhibition, even at substrate concentrations 10 times above the K m value. The data were obtained with RII␣ subunit in the dephospho-form.

Estimation of the Stability of the Ternary Complexes of cAMP⅐RI␣⅐C␣ and cAMP⅐RII␣⅐C␣ at Assumed Physiological
In separate experiments using RII␣ phosphorylated by the C subunit, a higher concentration of cAMP-saturated RII was required to inhibit the C subunit (not shown). From previously published data it can be estimated that the C subunit in cAMPstimulated hepatocytes is at most 90% saturated with substrate, and that the concentrations of C subunit and R subunit (RI ϩ RII) is about 0.5 M in the average cell (see the last paragraph under "Experimental Procedures").

the fractional association of RI␣-TRITC (RI␣) and C␣-FITC (C␣) by fluorescence resonance energy transfer analysis
The FITC-labeled bC␣ subunit (30 nM) and the TRITC-labeled bRI␣ subunit (50 or 500 nM) were excited at 490 nm, and the emission monitored (cps) at 510 and 580 nm. The emission at 510 nm was due to C␣ alone, since RI␣ did not emit at this wavelength. At 580-nm C␣ had significant emission (15.6% of that at 510 nm) which was subtracted to obtain the emission due to RI␣ only (column 3 from the right). The ratio of emission 510/580 nm (C␣/RI␣) is shown for the separated subunits (r diss ), and for combined subunits in the presence (r x ) and absence (r holo ) of 50 M cAMP. The r diss represents the completely dissociated state (fractional association ϭ 0), and r holo the completely associated holoenzyme state (fractional association ϭ 1.0). The fractional association observed in the presence of cAMP was determined as shown in the right hand column. The experiment was conducted at 25°C in buffer A.  2. The association of FITC-labeled C␣ subunit with cAMP saturated TRITC-labeled RI␣ subunit of cAMP kinase determined by fluorescence resonance energy transfer. The data shown are from a typical experiment conducted like that shown in Table  I, with constant concentration (30 nM) of C␣-FITC and increasing concentrations of RI␣-TRITC at 50 M cAMP. The fractional association was determined as described in Table I , it follows that about one-fourth of the C subunit can exist as ternary complex in an average maximally cAMP-stimulated cell if the R subunit is RI␣ or dephospho-RII. The equation predicts half of the kinase to be in ternary complex if the R subunit concentration is increased from 0.5 to 1.25 M.

DISCUSSION
The ternary (cAMP saturated) cA-PK holoenzyme complex has received little attention, presumably because it traditionally has been considered too unstable to exist at appreciable concentrations in living cells (11). We used the CRE-luciferase reporter gene to probe for dissociation of cA-PK in intact cells. To our initial surprise, cells with enforced expression of RI␣ or RII␣ showed decreased expression of luciferase, even at extremely high concentrations of cAMP analogs and agents raising the endogenous cAMP level. This suggested that formation of cAMP-saturated cA-PKI and cA-PKII significantly impeded nuclear influx of C subunit. The intact cell data were supported by results obtained with isolated cA-PKI subunits under near physiological conditions in the presence of cAMP. As judged by three independent biophysical methods, formation of ternary complex of C␣ with cAMP-saturated RI␣ subunit occurred with an apparent K d in the submicromolar range.
Since the ternary complex was more stable than previously recognized, the question of its catalytic activity becomes biologically important. An intriguing possibility is that R-C holoenzyme associated with scaffolding protein (AKAP), and thereby in immediate contact with AKAP-tethered substrate (16,17), may be retained in an active holoenzyme complex close to the substrate upon cAMP activation (15). The present study failed to demonstrate any significant kinase activity of C␣ in ternary complex with cAMP and RI␣ toward the heptapeptide Kemptide or physiological substrates (phenylalanine-4-monooxygenase and tyrosine-4-monooxygenase). Neither did cAMP-saturated cA-PKII holoenzyme show demonstrable catalytic activity. These results were reproduced with a number of different preparations of R and C subunits, and under a variety of conditions, including close to physiological with respect to temperature, pH, and ionic strength. We conclude that dissociation is a prerequisite for both cA-PKII and cA-PKI to catalyze substrate phosphorylation, at least under physiologically relevant conditions in vitro. Obviously, cAMP-saturated cA-PKI and cA-PKII differ from cyclic nucleotide-saturated cGMP-dependent protein kinase in this respect (12). An early report (11) suggested that the ternary complex between cAMP, RII, and C subunit had about 15% of the activity of the free C subunit. A more recent study (15) found the ternary complex between cAMP, RII, and fluorescein-labeled C subunit to have full catalytic activity toward Kemptide. A possible explanation of this discrepancy may be that C subunit labeled with fluorescein-succinimidyl ester (15) has poor ability to interact with RI subunit (4) and may have subtly altered interaction with RII as well, allowing cA-PKII holoenzyme formation without kinase inhibition. It is known that point mutations of C␣ can affect binding to either RI or RII, without interfering with the catalytic activity (28).
The ternary complex of C subunit with cAMP and R was destabilized by protein kinase substrate (Fig. 6). We envisage therefore that substrate depletion due to phosphorylation, by allowing more C subunit to be sequestered in ternary complex, may act as a negative feedback mechanism of kinase activity. Experimental verification of this possibility is hampered by the instability of the FRET signal used to monitor cA-PK dissociation in intact cells, due to photobleaching and nuclear translocation of the C subunit. We note, however, that Zaccolo et al. (3) reported the FRET signal typical of cA-PKII dissociation to decrease with time in some Chinese hamster ovary cells continuously exposed to a maximal cAMP stimulus.
Extrapolation from the in vitro data suggests that about one-fourth of the C subunit can be sequestered as inactive, cAMP-saturated cA-PK (ternary complex) in the average cAMP-stimulated cell. This allows a novel avenue for control of the cA-PK activity in maximally cAMP stimulated cells, through regulation of the R subunit level. Up-regulation of R subunit will sequester more C subunit as inactive complex, and down-regulation of R will release kinase from the ternary complex. Such regulation will not be possible if it is assumed that all cA-PK is dissociated in maximally cAMP-stimulated cells. The decreased CRE-mediated gene expression in cells overexpressing RI␣ or RII␣ (Fig. 1) is not readily explained without invoking ternary complex formation. Since overexpressed R subunit is not taken up by nuclei (3,29), the effect is best explained by assuming sequestration of C subunit in a ternary complex in the cytoplasm. In retrospect, ternary complex formation can explain our previous observation of relatively more holoenzyme-associated kinase than expected from the tissue cAMP level during the pre-replicative cAMP surge in the regenerating liver, in which both RI and RII were up-regulated (7). Similarly, ternary complex formation may explain the attenuated cAMP-stimulated gene induction in C/EBP␤ null mice, which have increased RI and RII (5). It may also explain the deficient cAMP-stimulated CREB phosphorylation and CREdependent gene expression in protein kinase inhibitor null mice, which have 50% increase of RI␣.
Selective degradation of the free cAMP-complexed R subunit (30) will, in contrast to the above examples of R up-regulation, serve to further enhance the kinase activity in maximally cAMP-stimulated cells. An intriguing example is found in Aplysia neurons exposed to 5-hydroxytryptamine. In these cells the cAMP level remains elevated for 2 h, during which time R subunit degradation (about 20%) must occur to ensure enough CRE-dependent gene transcription to establish long term potentiation (31). The current view is that degradation of R renders the Aplysia cA-PK independent of cAMP (31,32). This may be an oversimplification, since R p -cAMPS, which acts by promoting R-C association by displacing cAMP (12,33), blocked potentiation, even when given several hours after the cAMP stimulus (34). Aplysia and mammalian cA-PK appear basically similar (31,35). The possibility should therefore be considered that R degradation in Aplysia amplifies cAMP action by allowing more C subunit to escape ternary complex formation both during the acute cAMP stimulation and thereafter.
In conclusion, the ternary complex of C␣ with cAMP-saturated RI␣ or RII␣ was devoid of catalytic activity against relevant substrates. The ternary complexes had higher in vitro stability than previously recognized. Extrapolation from in vitro data predicted a significant proportion of C subunit to be in the cytoplasmic ternary complex in maximally cAMP stimulated normal cells, and an even higher proportion in cells overexpressing RI␣ or RII␣. Ternary complex formation is predicted to decrease nuclear accumulation of C subunit and thereby CRE-dependent gene expression, and offered the only rational explanation of the experimentally observed lowered CRE-luciferase expression in cells overexpressing RI␣ or RII␣. Modulation of ternary complex formation represents, to our knowledge, the only way of tuning cAMP-dependent protein kinase activity in maximally cAMP-stimulated cells. Such modulation may occur through altered substrate availability or regulation of the cellular content of R subunit.