Substrate Enhances the Sensitivity of Type I Protein Kinase A to cAMP*

The functional significance of the presence of two major (types I and II) isoforms of the cAMP-dependent protein kinase (PKA) is still enigmatic. The present study showed that peptide substrate enhanced the activation of PKA type I at low, physiologically relevant concentrations of cAMP through competitive displacement of the regulatory RI subunit. The effect was similar whether the substrate was a short peptide or the physiological 60-kDa protein tyrosine hydroxylase. In contrast, substrate failed to affect the cAMP-sensitivity of PKA type II. Size exclusion chromatography confirmed that substrate acted to physically enhance the dissociation of the RIα and Cα subunits of PKA type I, but not the RIIα and Cα subunits of PKA type II. Substrate availability can therefore fine-tune the activation of PKA type I by cAMP, but not PKA type II. The cAMP-dissociated RII and C subunits of PKA type II reassociated much faster than the PKA type I subunits in the presence of substrate peptide. This suggests that only PKA type II is able to rapidly reverse its activation after a burst of cAMP when exposed to high substrate concentration. We propose this as a possible reason why PKA type II is preferentially found in complexes with substrates undergoing rapid phosphorylation cycles.

The functional significance of the presence of two major (types I and II) isoforms of the cAMP-dependent protein kinase (PKA) is still enigmatic. The present study showed that peptide substrate enhanced the activation of PKA type I at low, physiologically relevant concentrations of cAMP through competitive displacement of the regulatory RI subunit. The effect was similar whether the substrate was a short peptide or the physiological 60-kDa protein tyrosine hydroxylase. In contrast, substrate failed to affect the cAMP-sensitivity of PKA type II. Size exclusion chromatography confirmed that substrate acted to physically enhance the dissociation of the RI␣ and C␣ subunits of PKA type I, but not the RII␣ and C␣ subunits of PKA type II. Substrate availability can therefore fine-tune the activation of PKA type I by cAMP, but not PKA type II. The cAMP-dissociated RII and C subunits of PKA type II reassociated much faster than the PKA type I subunits in the presence of substrate peptide. This suggests that only PKA type II is able to rapidly reverse its activation after a burst of cAMP when exposed to high substrate concentration. We propose this as a possible reason why PKA type II is preferentially found in complexes with substrates undergoing rapid phosphorylation cycles.
The cAMP-dependent protein kinase (PKA) 1 is a model for protein kinases because of its universal distribution and its relative simplicity because the catalytic moiety and the autoinhibitory moiety are on different subunits. The mammalian tetrameric PKA consists of two catalytic (C) subunit monomers and a regulatory (R) subunit dimmer. The binding of two molecules of cAMP to each R subunit favors dissociation of the C and R subunits. The kinase exists in two major isoforms with different R subunits, RI and RII, each with ␣and ␤-subforms. The biological significance of the presence of two major isozymes is still uncertain. It was early noticed that RI predominated in many cells with rapid proliferation, rapid growth in cell size, or malignantly transformed cells (1)(2)(3)(4). More recently it has been shown by homologous recombination that RI␣Ϫ/Ϫ mice die in embryonic life, whereas RII␤Ϫ/Ϫ, RII␣Ϫ/Ϫ, and RI␤Ϫ/Ϫ mice have less obvious defects, mainly in differentiation of adipose tissue (RII␤) and neural functions (RI␤) (5). An important difference between RI and RII is their relative affinity for anchoring proteins that confine PKA type II, and in some cases PKA type I, to subcellular compartments (see Refs. 6 and 7 for recent reviews). Another striking difference is the ability of PKA type II, but not type I, to become autophosphorylated (8 -10). The interface between the R and C subunits of PKA is complex. The RI and RII subunits bind in part to non-overlapping areas of the C subunit (11). The R subunits contact the substrate binding site of the C subunit as well as areas distant from this site (12) and are, through their (pseudo)substrate moiety, believed to displace substrate from the C subunit and thereby inhibit the kinase activity (10,12). The substrate could displace the R subunit from the substrate binding cleft of the C subunit and thereby facilitate holoenzyme dissociation. In this way PKA could be activated at lower concentrations of cAMP in the presence of substrate.
In the present study we found that PKA type I, but not type II, was sensitized to cAMP by peptide substrate or by macromolecular substrates like tyrosine hydroxylase. This novel observation suggests that the activity of PKA type I is determined not only by the cAMP level but also by the availability of substrate. Substrate inhibited the rate of reassociation of the cAMP-dissociated subunits of PKA type I, but not type II. This suggests that high substrate availability may prolong the duration of the activation of PKA type I after a burst of cAMP.
We have shown previously that physiologically relevant concentrations of cAMP-saturated R subunit of PKA types I and II can form inactive holoenzyme (13). Using physical methods to determine R and C subunit interaction we could confirm this and show that substrate facilitated the dissociation of PKA type I, but not type II. We propose therefore that substrate can selectively enhance PKA type I activity through three mechanisms: by facilitating the activation by low cAMP concentration, by prolonging the activation after a burst of cAMP, and by the dissociation of the holoenzyme at saturating cAMP concentration. medicine, University of Bergen, and was produced recombinantly from a vector obtained from Dr. Aitken, Department of Biomedical Sciences, University of Edinburgh, Scotland. For mutagenesis of RI (hRI␣ G325D) the QuikChange site-directed mutagenesis kit (Stratagene) was used. The RI double stranded plasmid and two synthetic oligonucleotide primers with the desired mutations were annealed and extended by means of the Pfu DNA polymerase. After temperature cycling, the parental DNA template was digested using DpnI. The mutated DNA was transformed into Epicurian Coli® XL1-Blue supercompetent cells. The mutation was confirmed by sequencing. RII␣ (1 M) was stoichiometrically phosphorylated by the C subunit (0.2 M) of PKA for 5 h at 25°C in buffer A with 30 M cAMP and 100 M ATP. The bulk of C␣ was separated from the phosphorylated RII␣ by batch incubation with SP-Sepharose in buffer A with 4 M urea and 50 M cAMP. The remaining trace of C␣ was removed by fast protein liquid chromatography size exclusion chromatography (Superdex 200, 24 ml total volume) equilibrated with buffer A. To prevent reassociation of R and C subunits, 2.5 ml of buffer A with 4 M urea and 50 M cAMP was injected before the sample containing P-II␣. The P-RII␣ used had 0.002 mol of C subunit/mol of RII monomer. The P-RII␣ showed complete conversion to a more slowly migrating form (15) by standard SDSpolyacrylamide gel electrophoresis (not shown).
Phosphotransferase Assays-The phosphorylation of kemptide was determined essentially as described in Ref. 13. The RI and C subunits were preincubated for 30 min in buffer A, and the reaction was initiated by the addition of [␥-32 P]ATP (to a final concentration of 0.3 mM), various concentrations of substrate peptide, and cAMP. The phosphorylation of tyrosine 4-monooxygenase or glutathione S-transferase-SR-RASVGL-14 -3-3␥ was determined as for kemptide except that the reaction was started by the addition of [␥-32 P]ATP just after the addition of substrate protein and the reaction was terminated by adding two volumes of stop solution (1.5 mM ATP, 0.08% SDS, and 60 mM EDTA). Samples (40 l) from this mixture were spotted on 3MM cellulose filters and subjected to extensive washing in 5% trichloroacetic acid before drying and scintillation counting.   (16) Scintillation Proximity Assay-The association of RI with bound [ 3 H]cAMP to immobilized C subunit was determined in buffer A with 0.3 mM ATP or ADP by scintillation proximity assay as described previously (13).
Separation of the Free and Holoenzyme-associated C Subunit by Size Exclusion Chromatography-A size exclusion column (Superdex 200, PC 3.2/30; Amersham Biosciences) was equilibrated with various concentrations of R subunit in buffer A containing one or more of the following substances: 0.3 mM ATP, 0.6 mM ADP, 1.2 mM AMP, 0.03 mM cAMP, 0.1 mM kemptide. The sample (50 l) contained C subunit and bovine hemoglobin as endogenous standard. Fractions (0.1 ml) were collected and analyzed for kinase activity in the presence of 10 M cAMP and for absorbance at 405 nm (hemoglobin). The fraction of C complexed with R in a particular sample was determined based on the elution position of the kinase activity in that sample compared with the elution position of free C subunit and the holoenzyme form of the kinase. If the kinase activity eluted midway between the holoenzyme and free C subunit, the sample was considered to contain 50% holoenzyme and 50% free C subunit.

Peptide Substrate Competitively Inhibited the Formation of
Inactive cAMP-saturated PKA Type I Holoenzyme but Failed to Affect PKA Type II Holoenzyme Formation-In line with our previous report (13), the PKA activity was inhibited at submicromolar concentrations of cAMP-saturated RI or RII subunit. The IC 50 for cAMP-saturated RII was 0.8 M irrespective of the kemptide concentration. On the other hand, several-fold more RI was required to form inactive holoenzyme in the presence of high than low kemptide concentration (Fig. 1, A and B). To know more about the mechanistic background for this striking difference between the isozymes, we studied the kinase activity at various concentrations of kemptide and at constant concentration of cAMP-saturated R subunit. The RII subunit inhibited the kinase activity in a non-competitive manner (not shown), whereas the RI subunit was a competitive inhibitor of kemptide (K i ϭ 0.19 M), as shown by a double inverse plot (Fig. 2). The K i value was also similar when calculated from experiments where the concentration of RI␣ was 1 or 3.8 M (not shown).
To know whether kemptide could also modulate the holoenzyme formation for RI subunit monosaturated with cAMP, we studied the RI(G325D) mutant. This protein did not bind cAMP to site B but retained high affinity cAMP binding to site A (not shown). It could form inactive holoenzyme at subnanomolar concentration in the presence of excess cAMP (Fig. 1C). The concentration of kemptide required to modulate holoenzyme formation was similar for RI(G325D) and RI wild type (Fig. 1). This suggested a competitive relationship between RI and kemptide whether one or both cAMP binding sites of RI were occupied.
The Physical Formation of PKA Type I and Type II Holoenzyme Complex Was Differentially Modulated by Mg/ATP, Mg/ ADP, and Substrate Peptide-The experiments described above (Figs. 1 and 2) relied on kinase activity as a marker of holoenzyme formation and therefore had to be conducted under conditions allowing substrate phosphorylation by the C subunit of the kinase. Each catalytic cycle of the kinase is associated with the formation of C subunit-bound Mg/ADP, whose slow dissociation is the rate-limiting step of the phosphorylation cycle (17). Under phosphorylation conditions the C subunit can therefore exist in complex not only with Mg/ATP or Mg/ATP and kemptide but also with Mg/ADP or Mg/ADP and kemptide or phosphokemptide. One possible explanation for the puzzling difference between PKA types I and II regarding kemptide effect (see above) could be differential affinity of RI and RII subunits for the Mg/ADP-liganded C subunit formed in the presence of kemptide rather than a differential affinity for the kemptide-liganded C subunit itself. To test this possibility we used biophysical methods to determine holoenzyme formation (a) in the presence of Mg/ADP and Mg/ATP in the absence of kemptide, and (b) in the presence of Mg/ADP and absence and presence of kemptide. In this way we could study the effects of Mg/ATP, Mg/ADP, and kemptide without interference from any reaction products formed due to phosphorylation of substrate.
The scintillation proximity assay allows the selective determination of isotope close to the immobilized scintillant. We attached biotinylated C subunit of PKA to the scintillant-containing vial wall and titrated its binding of R with bound [ 3 H]cAMP (13) (Fig. 3A). We found that the apparent K d of formation of PKA type I holoenzyme (ternary complex of C, RI, and [ 3 H]cAMP) rose from 0.16 M in the presence of Mg/ATP to 1.3 M in the presence of Mg/ADP (Fig. 3A). The formation of Mg/ADP bound to the C subunit can therefore be one mechanism by which substrate can destabilize the PKA type I holoenzyme.
To know whether kemptide could have a direct destabilizing effect on the RI-C complex, as suggested by the competitive nature of its action (Fig. 2), we studied the formation of PKA holoenzyme in the absence and presence of kemptide. The formation of holoenzyme was determined based on the retention of C subunit injected into a size exclusion chromatography column equilibrated with various concentrations of cAMP-saturated RI subunit. We found that kemptide (100 M) increased the retention, indicating less holoenzyme formation, whether Mg/ATP (Fig. 3B) or Mg/ADP (Fig. 3C) was present. The effect of kemptide in the presence of Mg/ADP suggests strongly that kemptide can also inhibit holoenzyme formation independently of the conversion of Mg/ATP to Mg/ADP during the phosphorylation cycle.
Experiments similar to those shown for type I holoenzyme were next conducted for type II holoenzyme. We found (Fig. 4A) that the K d for formation of holoenzyme complex between cAMP-saturated RII and the C subunit was only slightly higher in the presence of Mg/ADP (1.5 M) than in the presence of Mg/ATP (0.8 M). In the presence of Mg/AMP the K d of the cAMP-saturated holoenzyme complex was 2.4 M (Fig. 4A). Kemptide failed to affect the retention of C subunit injected into a column equilibrated with 0.5 M RII in the presence of Mg/ADP (not shown) or Mg/ATP (Fig. 4B). We conclude that kemptide substrate did not have a direct effect on holoenzyme formation between RII and C subunit, unlike the findings for RI and C subunits. Furthermore, the substrate-dependent formation of Mg/ADP would be far less effective in dissociating PKA type II (Fig. 4A) than type I (Fig. 3A).

Substrate Peptide Counteracted the C Subunit-induced Dissociation of the RI␣-[ 3 H]cAMP Complex but Not the RII␣-[ 3 H]cAMP
Complex-A major difference between PKA type I and type II is that only the latter is capable of undergoing autophosphorylation at a residue (Ser-95 in human RII) in the (pseudo)substrate part of the linker region between the N-and C-terminal domains of the R subunit (8 -10). The autophosphorylated form of RII is known to bind less tightly to the C subunit (18 -20). Kemptide can inhibit the autophosphorylation of RII (21). One can therefore hypothesize that kemptide could facilitate type II kinase dissociation but that this effect is masked by decreased autophosphorylation, leading to little or no effect of substrate on type II kinase dissociation, as observed ( Figs. 1, 3, and 4). To know whether the different response to kemptide for the isozymes was due to any effect of kemptide on the degree of autophosphorylation, we conducted experiments (Fig. 5) in which the phosphorylation state of RII was unaltered.
The C subunit-induced dissociation of labeled cAMP from its complex with RII was studied in a buffer with Mg/ADP or Mg/AMP instead of Mg/ATP to avoid any possibility of RII phosphorylation. We found no effect of kemptide on the rate of the C subunit-induced dissociation of the complex of RII and [ 3 H]cAMP. This was true for RII incubated with either Mg/ADP (Fig. 5A) or Mg/AMP (Fig. 5B). In contrast, kemptide nearly abolished the C subunit-induced dissociation of [ 3 H]cAMP from its complex with RI incubated with Mg/ADP (Fig. 5D) even at 1 M C subunit (Fig. 5E).
To study phospho-RII, we incubated with Mg/AMP because the C subunit can catalyze dephosphorylation of phospho-RII in the presence of Mg/ADP (22) or in the presence of Mg/ATP and kemptide (21). We found no effect of kemptide on the rate of the C subunit-induced dissociation of the complex of phosphorylated RII and [ 3 H]cAMP (Fig. 5C). This shows that kemptide acted differently on the association of C to RI and RII also under conditions when autophosphorylation of RII or dephosphorylation of phospho-RII could be ruled out.
An inhibitor peptide, representing the 20-residue core of the heat-stable inhibitor of PKA (14) binds to the substrate peptide binding site as well as to an adjacent hydrophobic area on the C subunit (23). Mutagenesis of residues in this area of the C subunit inhibited the interaction with RII (11). It was therefore of interest from a mechanistic point of view to know whether the peptide could modulate the interaction of C subunit and the complex of RII and [ 3 H]cAMP. This peptide at 1 M inhibited the dissociation of [ 3 H]cAMP from RI (Fig. 5D) and RII. Even at 50 nM it was able to counteract the C subunit-induced dissoci-ation of cAMP from RII (Fig. 5A). This showed that the assay was able to pick up effects of peptides competing with RII for binding to the C subunit.
Substrate Sensitized PKA Type I, but Not Type II, toward Activation by cAMP-To know whether peptide substrate could sensitize PKA toward cAMP, we studied the cAMP-dependence of PKA types I and II in the presence of various concentrations of kemptide. Whereas 0.074 M cAMP was sufficient to halfmaximally activate PKA type I at near saturating concentration of substrate peptide, 0.21 M cAMP was required at low peptide concentration (Fig. 6A). Little sensitization by substrate was noted for PKA type II, which was 50% activated at 0.4 -0.5 M cAMP (Fig. 6B).
A concern was that kemptide failed to sensitize the type II isozyme because of its small physical size and that a macromolecular substrate would also sensitize PKA type II. To test this possibility we used as substrate a macromolecule where a PKA substrate consensus sequence (SRRASVGL) was incorporated as a linker between the two macromolecules glutathione Stransferase and 14 -3-3␥. We also studied the physiological and large substrate tyrosine hydroxylase, which exists as a tetramer of four subunits of 60 kDa (24). Tyrosine hydroxylase is a physiological substrate whose activity is regulated by phosphorylation by PKA (25). We found that either of these macromolecule substrates, like kemptide, sensitized PKA type I, but not type II, toward activation by cAMP (Fig. 7). We conclude FIG. 4. The effect of Mg/AXP and kemptide on the complex formation between (RII␣) 2 (cAMP) 4 and C␣. The experimental conditions were as described in the legend to Fig. 3, B and C, except that RII␣ was present instead of RI␣. A, the percentage of injected C␣ subunit associated with (RII␣) 2 (cAMP) 4   that physiologically relevant macromolecular PKA substrates also facilitate cAMP activation of PKA type I, but not type II. DISCUSSION The biological significance of the existence of two major isozymes (PKA types I and II) of the cAMP-dependent protein kinase with distinct regulatory subunits (RI, RII) is still enigmatic. The amino acid sequence of the RI and RII subunits differs most in the N-terminal region, which interacts with the A kinase-anchoring protein family of scaffolding proteins. It is therefore likely that one teleological purpose of the existence of two isozymes is to obtain differential subcellular localization of the kinase depending on which isozyme is expressed.

FIG. 5. Effect of kemptide and kinase inhibitor peptide on the C subunit-induced release of [ 3 H]cAMP from ([ 3 H]cAMP) 4 (RI␣) 2 or from ([ 3 H]cAMP) 4 (RII␣) 2 . RII␣, P-RII␣, or RI␣, all saturated with [ 3 H]cAMP and present at 2 nM, was incubated at 25°C in buffer
The present study demonstrates another striking difference. The presence of substrate peptide sensitized PKA type I to the activating effect of cAMP. This was true whether the substrate was a minimal heptapeptide or a macromolecular substrate like tyrosine hydroxylase (Figs. 6 and 7). This means that for PKA type I the degree of activation by a moderate cAMP increase would depend on the concentration of available substrate in the immediate vicinity. We believe that such substrate activation is a new way in which PKA type I can be activated without altering the cAMP level. The substrate histone was already reported in 1973 to induce dissociation and sensitize PKA type II toward cAMP (26). We found little sensitizing effect of substrate toward type II kinase in the present study, using a number of substrates. Presumably, the effect of histone can be explained by nonspecific electrostatic interac-tion with RII not related to its ability to act as substrate (27).
PKA type II is anchored close to ion channels that are PKA substrates (28,29). For the L-type Ca 2ϩ channels in heart there is a requirement for rapid activation and inactivation during the cardiac contraction cycle (reviewed in Refs. 30 and 31). A functional requirement for rapid kinase inactivation following an abrupt decrease of [cAMP] may also be present for other channels regulated by anchored PKA type II (7). The rate with which PKA is reinactivated after a burst of cAMP depends on the rate of recombination between the C subunit and the cAMP-saturated R subunit. Each phosphorylation cycle generates C(Mg/ADP), which is believed to be more abundant than C(Mg/ATP) under steady state conditions of active substrate phosphorylation, because each catalytic cycle of the kinase is associated with the formation of C subunit-bound Mg/ADP, whose slow dissociation is the rate-limiting step of the phosphorylation cycle (17,21,32). We found that C(Mg/ADP) reassociated ϳ15-fold more rapidly with cAMP-saturated RII␣ than RI␣. In the presence of kemptide the difference was even larger (Fig. 5). It appears therefore that a short-lived burst of cAMP is likely to lead to a prolonged activation (feed-forward reaction) for PKA type I, but not type II. This may provide a functional explanation for why PKA type II is preferred in complexes with substrates undergoing rapid phosphorylation cycles.
The cAMP responsiveness of PKA type I in the intact cell is enigmatic. There is a general notion that PKA type I is more responsive than PKA type II to slight increases of cAMP (33,34). The higher cAMP sensitivity of PKA type I than type II in the presence of abundant substrate (Figs. 6 and 7) can explain the preferential activation of PKA type I by slight cAMP stimulation. On the other hand, RI␣ is known as the "tissue-specific extinguisher" of cAMP-induced gene activation (35,36), implying that it can counteract cAMP-dependent transcription more efficiently than other proteins, including RII. In line with this, less expression of RI␣ than of RII␣ was required to inhibit cAMP-response element-dependent gene transcription in maximally cAMP-challenged HEK-293 cells (13), and less cAMPinduced gene expression has been noted in mice expressing increased levels of RI␣ (37,38). The cAMP control of gene transcription is complex and offers several possible avenues for regulation by PKA (39 -42). We believe that one avenue by which RI␣ can attenuate gene regulation induced by strong cAMP stimulation is through the formation of PKA type I holoenzyme even at saturating cAMP level (Fig. 1) (13). In this way cAMP-liganded RI␣ can act as a clamp on the maximal PKA activation and thereby limit the amount of C subunit that translocates to the nucleus and stimulates gene transcription. We found that the cAMP-saturated RI␣ and C␣ interacted with a K i of 0.19 M when peptide substrate was absent (Figs. 2 and  3), against an IC 50 of above 1 M (Fig. 1) in the presence of a high concentration of kemptide. The physiological concentration of RI subunit is in the range of 0.1 to 1 M (see Ref. 13 for references), suggesting that a significant proportion of the C subunit can be clamped as inactive holoenzyme in cAMP-stimulated cells, particularly at low substrate availability. We believe that substrate facilitation of the dissociation of type I kinase is of relevance in the intact cell because it was observed with RI␣ subunit concentrations from low nM (Fig. 6) to above 1 M (Fig. 1), i.e. encompassing the physiological range of RI␣ concentrations.
We noted that the K d for C-and cAMP-saturated RI␣ decreased ϳ10-fold (Fig. 3), against only 2-fold for C and cAMPsaturated RII␣ (Fig. 4) when Mg/ADP replaced Mg/ATP. This means that RI can discriminate better than RII not only between C subunit with bound kemptide (Figs. 1 and 3-7) but also between C(Mg/ADP) and C(Mg/ATP). This lends further support to the notion (12) that RI and RII differ not only in their N-terminal anchoring domains but also in the C-terminal domains responsible for interacting with the C subunit. This observation can also help explain at least partly the difference between our previous study (13) and Vigil et al. (43). They reported that PKA type I (1.5 M with respect to RI␣) was nearly completely dissociated in the presence of cAMP, 0.2 mM Mg/ATP, and 1 mM kemptide, whereas the type II isozyme was less dissociated by kemptide. With a catalytic turnover number of ϳ25 s-1 (Ref. 17 and Fig. 6), the C subunit at concentration of 1.5 M will generate up to 30 M ADP from ATP/second of incubation. It is therefore more instructive to compare the data of Vigil et al. with those obtained by us in the presence of Mg/ADP, where we noted a marked inhibition by 0.3 mM kemptide of the rate of C subunit-induced dissociation of cAMP from its complex with RI (Fig. 5, D and E). We noted also that kemptide led to increased physical dissociation of the type I holoenzyme in the presence of Mg/ADP (Fig. 3).
In conclusion, we have presented evidence that substrate can enhance the dissociation of PKA type I at low, physiologically relevant concentrations of cAMP. Substrate availability can therefore fine-tune the activation of PKA type I by cAMP in the intact cell. In contrast, substrate failed to affect the interaction of the subunits of type II kinase. This may explain why type I kinase has been reported to be more sensitive than type II kinase to low concentrations of cAMP in live cells (33,34). We also found other distinct differences between the isozymes, particularly that the cAMP-saturated RII subunit reassociated much more rapidly than the RI subunit with C(Mg/ADP) in the presence of substrate peptide. In the intact cell this would make PKA type II more suitable for a rapidly reversible activation after a burst of cAMP. This may provide a functional explanation for why PKA type II is preferred in compartmentalized complexes with substrates undergoing rapid phosphorylation cycles.