Identification of Critical Determinants for Autoinhibition in the Pseudosubstrate Region of Type I (cid:97) cAMP-dependent Protein Kinase*

The consensus substrate site for cAMP-dependent protein kinase (PKA) is Arg-Arg-Xaa-Ser(P)-Xaa and the autoinhibitory domain of the PKA type I (cid:97) regulatory subunit (RI subunit) contains a similar sequence, Arg 92 -Arg-Arg-Arg-Gly-Ala-Ile -Ser-Ala-Glu. The italicized amino acids form a putative pseudosubstrate site (Ser is re- placed with Ala ), which together with adjacent residues could competitively inhibit substrate phosphorylation by the PKA catalytic subunit (C subunit). The present studies determine the contributions of Arg 92–95 , Ile 98 , and Glu 101 to inhibitory potency. Amino-terminal truncation of RI subunit through Arg 92 ( (cid:68) 1–92) or Arg 93 ( (cid:68) 1– 93) had no detectable effect on inhibition of C subunit. Truncation through Arg 94 ( (cid:68) 1–94), or point mutation of Arg 95 within truncated mutants ( (cid:68) 1–93.R95A or (cid:68) 1– 92.R95A), caused a dramatic reduction in inhibitory po- tency. Truncation through Arg 95 ( (cid:68) 1–95) had a greater effect than did replacement or deletion of Arg 94 or Arg 95 alone. Using full-length RI subunit, the inhibitory po- tency was reduced by replacing Ile 98 with Ala, Gly,

same basic domain structure: a short amino-terminal dimerization domain, an autoinhibitory domain located in the aminoterminal segment of the protein, and two cAMP-binding domains toward the carboxyl-terminal end.
The autoinhibitory domain of the R subunit contains a sequence that mimics the consensus phosphorylation sequence of PKA substrates (Arg-Arg-Xaa-Ser(P)-Xaa). In the type II R subunit (RII subunit), this sequence (Arg-Arg-Val-Ser(P)-Val) is a substrate for C subunit, but in the RI subunit, the phosphorylatable Ser is substituted with a non-phosphorylatable Ala residue to form a pseudosubstrate site (Arg-Arg-Xaa-Ala-Xaa). The pseudosubstrate sequence is believed to interact with the C subunit catalytic site to competitively inhibit substrate phosphorylation (3).
The phosphorylation or pseudophosphorylation site is designated as the P residue, while residues located amino-terminal or carboxyl-terminal to P are designated as minus or plus residues, respectively (4). The P Ϫ3 and P Ϫ2 Arg residues are important determinants for phosphorylation of PKA substrates (5,6). Replacement of either residue in peptide substrate analogs profoundly impairs the phosphorylation of the peptide. The P Ϫ3 and P Ϫ2 Arg residues are invariant in all known R subunit pseudosubstrate sequences and are required for potent inhibition of C subunit (3,7,8). They are also found in the pseudosubstrate sequence of the high affinity, heat-stable protein kinase inhibitor of PKA, PKI (Arg-Arg-Gln-Ala-Ile) (9). The importance of these two Arg residues in PKI has also been well documented using peptide analogs (10,11), and more recently by co-crystallization of C subunit with the peptide derived from the inhibitory segment of PKI, PKI-(5-24) (4,12).
Although many natural substrates for PKA conform to the consensus sequence of Arg-Arg-Xaa-Ser(P)-Xaa, many others contain only a single basic residue at either P Ϫ3 or P Ϫ2 , and/or have additional basic residues more amino-terminal to P Ϫ3 (13). Basic residues are frequently found at the P Ϫ4 , P Ϫ6 , or both positions in PKA substrates (13), yet other substrates contain a cluster of four basic residues at the P Ϫ2 through P Ϫ5 positions (14 -16). Both P Ϫ4 and P Ϫ6 basic residues are important recognition factors for phosphorylation of the peptide derived from rabbit skeletal muscle phosphorylase kinase (␤ subunit) (17). A similar pattern of basic amino acid residues is noted in the PKA regulatory proteins. A P Ϫ6 Arg residue is present in the RII subunit substrate site and in the PKI pseudosubstrate site, while the RI subunit has a cluster of four Arg residues at the P Ϫ2 through P Ϫ5 positions. The PKI P Ϫ6 Arg is crucial for potent inhibition of C subunit (4,11). The contribution made toward C subunit inhibition by the RII subunit P Ϫ6 Arg residue, or the RI subunit P Ϫ4 and P Ϫ5 Arg residues, has not been investigated previously. Based on the evidence above, it is possible that each of the four Arg residues in the pseudosubstrate region of RI subunit interacts with an acidic residue(s) in the active site of C subunit.
The P ϩ1 residue is frequently, but not exclusively, a large hydrophobic residue in PKA substrates (13). Studies with synthetic peptide substrates based on the phosphorylation sequence of pyruvate kinase (Arg-Arg-Ala-Ser(P)-Val) suggested that a large hydrophobic residue at the P ϩ1 position (Phe, Leu, or Ile) is a positive determinant for substrate phosphorylation (13). However, small (e.g. Gly), charged (e.g. Lys, Arg, or Glu), or hydrophilic residues (e.g. Ser) at this position are negative determinants for phosphorylation of these peptide substrates by C subunit (13). The P ϩ1 residue is conserved as Ile or Val in the autoinhibitory domains of all known sequences of PKA and PKG, and as Ile in PKI (10). PKI peptide analog studies (18) and the crystal structure of the C subunit⅐PKI-(5-24) complex (4) demonstrated the importance of a large P ϩ1 hydrophobic residue for potent inhibition of C subunit (18). The contribution of the RI subunit P ϩ1 hydrophobic residue toward C subunit inhibition has not been investigated to date.
Using truncation and site-directed mutagenesis, the present report addresses the importance of the P Ϫ5 ,P Ϫ4 , P Ϫ3 , P Ϫ2 , P ϩ1 and P ϩ4 residues for autoinhibition of type I␣ PKA.
Construction of RI Subunit Deletion Mutants-A T7 RNA polymerase/promoter system, pT7-7, was used for bacterial overexpression of the type I␣ R subunit. The RI subunit cDNA was subcloned into pT7-7 as described previously (28) to create pT7RNI and pT7R. pT7RNI, an intermediate vector, contained the entire RI subunit cDNA, which was out-of-frame. pT7R was the wild type (WT) RI subunit bacterial expression vector in which the RI subunit start codon was mutated to a NdeI site to allow direct in-frame fusion with the pT7-7 start codon, also encoded by a NdeI restriction site.
KS(Ϫ)R was created for use in oligonucleotide-directed mutagenesis by subcloning the 1118-base pair (bp) SacI-SalI fragment of pT7RNI (encoding all but 52 base pairs from the 5Ј end of the RI subunit cDNA) into the SacI-SalI fragment (2882 bp) of the M13-derived phagemid, pBluescript KS(Ϫ). Synthetic oligonucleotides 1-13 (Table I) were used to mutate KS(Ϫ)R by site-directed mutagenesis using the Kunkel method (29). Oligonucleotides 1-5 were used to mutate the KS(Ϫ)R sequence on the 5Ј side of the Arg 92 , Arg 93 , Arg 94 , Arg 95 , and Gly 96 codons to a NdeI site (CATATG). Oligonucleotides 6 -8 were used to create a NdeI site 5Ј to the Arg 93 or Arg 94 codons, and to mutate the codons for Arg 94 (CGA) or Arg 95 (GCG) to Ala (GCT) in KS(Ϫ)R. Introduction of a NdeI restriction site allowed for direct in-frame fusion with the NdeI restriction site encompassing the pT7R start codon, thus deleting the 5Ј segment of the RI subunit cDNA. Oligonucleotides 9 -13 were used to mutate the Ile 98 codon to Val (GTT), Ala (GCT), Gly (GGT), and Gln (CAG), and the Glu 101 codon to Ala (GCT) in KS(Ϫ)R. A 700-bp PstI fragment, released from the previously described pUC13-based RI subunit expression vector (pUC13R) (30), was subcloned into the PstI site of M13mp8 to create M13.PstR. Oligonucleotide 14 was used to mutate the Glu 101 codon (GAG) to Gln (CAA) in M13PstR using a commercially available mutagenesis kit (Oligonucleotide-directed in Vitro Mutagenesis, version 2, code RPN.1523; Amersham).
The integrity of the entire cDNA segment that was subjected to the mutagenesis procedure, including the cloning and mutation sites, was verified directly in the pT7R expression vectors by dideoxynucleotide sequencing of double-stranded DNA for mutants 1-12 and 14. DNA sequencing of mutants 1-8 and 14 was performed using a commercially available kit (Sequenase version 2.0, U. S. Biochemical Corp.) and mutants 9 -12 were sequenced using an Automated Biosystems, Inc. model 373A DNA sequencer in the Cancer Center DNA Core facility of Vanderbilt University. The cloning sites for mutant 13 were verified by digestion with PvuII and EcoRI, and the mutation was verified by dideoxynucleotide sequencing using the aforementioned kit.
Bacterial Expression-E. coli strain BL21(DE3) cells were transformed with wild-type or mutant pT7R vectors the day prior to expression. Bacterial expression was performed as described in Ref. 28, except

5Ј-GTGTAGACTTGAGCGCTGAT-3Ј
that the media contained 0.1 mg/ml ampicillin and the cells were pelleted at 5000 ϫ g for 15 min each time. The bacterial pellets were stored at Ϫ20°C until homogenization and were viable for more than 1 year when stored in this manner. Initially, the expression cultures yielded ϳ3-5 mg of RI subunit/liter of culture (5-h induction). Due to the necessity for greater quantities of protein, the expression time was later increased to 36 h, yielding 10 -80 mg of RI subunit/liter culture.
Purification of Recombinant RI Subunits-Bacteria were disrupted as described previously (31) except that the homogenization buffer was 10 mM potassium phosphate, pH 6.8, 1 mM EDTA, 2 mM ␤-mercaptoethanol (KPEM) plus 50 mM benzamidine (KPEMB) and 2.3 mM 3Ј,5Јcyclic inosine monophosphate (cIMP). cIMP was included during the homogenization process to promote exchange with the cAMP bound to RI subunits and thus increase the subsequent efficiency of RI subunit binding to N 6 -H 2 N-(CH 2 ) 2 -cAMP Sepharose affinity resin during purification.
Free cyclic nucleotide was removed from the extract by fractionation on a Sephadex G-25 superfine column (2.75 ϫ 26 cm) equilibrated in KPEMB. Protein-containing fractions (ϳ70 ml) were collected and clarified by centrifugation at 12,000 ϫ g for 30 min. The extract was chromatographed on a 1.5 ϫ 3.5-cm N 6 -H 2 N(CH 2 ) 2 -cAMP Sepharose column (equilibrated in KPEMB) at a flow rate of ϳ1 drop/12 s, or was divided and chromatographed over two columns (1.5 ϫ 3.5 cm and 1.5 ϫ 11.5 cm). In most instances, the flow-through was passed over the column(s) two or three times. RI subunits were isolated by the method of Dills (24) with the exceptions noted in Ref. 28. In addition, 30 mM cIMP was used to elute the affinity resin instead of 10 mM. Fractions containing protein (detected by Bio-Rad protein assay) were pooled and concentrated on a Centriprep-30 concentrator (Amicon) to a final volume of approximately 0.6 ml. Full-length and truncated RI subunits were resolved from proteolytic fragments and from free cyclic nucleotide on a Sephadex G-100 superfine column (0.9 ϫ 58.5 cm) equilibrated in KPEM containing 0.2 M NaCl. Fractions were assayed for [ 3 H]cAMP binding activity, and cIMP elution was measured by UV absorbance at 248 nm. The fractions containing RI subunit were pooled and concentrated on Centricon-30 devices (Amicon), followed by a desalting step with KPEM. The full-length and truncated RI subunits were Ͼ95% pure according to SDS-polyacrylamide gel electrophoresis (SDS-PAGE).
Purification of cAMP-saturated Native RI Subunit-Native type I␣ R subunit was purified from a bovine lung homogenate that was primarily used for the purification of cGMP-binding, cGMP-specific phosphodiesterase (cG-BPDE) (32) and type I␣ PKG (33) or it was purified from rabbit skeletal muscle according to Ref. 34. When using bovine lung homogenate, protein extract was adsorbed to DEAE-cellulose followed by batch elution with 0.1 M NaCl in 20 mM potassium phosphate, pH 6.8, 1 mM EDTA, and 25 mM 2-mercaptoethanol. After discarding the initial 0.5 liter of eluate, the next 1.5 liters of eluate was collected for cG-BPDE purification, and the following 1 liter was collected for purification of PKG and R subunit. This 1 liter of extract was slowly loaded onto N 6 -hexyl-cAMP-Sepharose (5 ϫ 2.75 cm) equilibrated in KPEM containing 25 mM 2-mercaptoethanol (KPEM25) at 4°C. The affinity resin was washed with 2 liters of KPEM25 containing 2 M NaCl, followed by 600 ml of KPEM25. The column was batch-eluted with 10 mM cGMP in KPEM25. Following elution of the bulk of PKG, monitored by SDS-PAGE of elutions, the affinity resin was batch-eluted with aliquots of KPEM25 containing 10 mM cAMP to elute types I␣ and II␣ R subunit. Fractions containing both types of R subunit were pooled and chromatographed on DEAE Sephacel (0.9 ϫ 5.5 cm) equilibrated in KPEM containing 0.05 M NaCl at 4°C. Types I and II R subunit were resolved by eluting the DEAE column with a linear salt gradient (0.05-0.35 M NaCl) in KPEM. The RI subunit was further purified by sucrose gradient (5-20%) centrifugation to separate it from PKG contaminants. The RI subunit fractions from five sucrose gradients were pooled and concentrated on a Centriprep-30 device. The protein was then chromatographed on a 0.9 ϫ 58.5-cm Sephadex G-100 superfine column (equilibrated in KPEM containing 0.2 M NaCl) to separate full-length RI subunit from proteolytic fragments of RI subunit and PKG. RI subunit fractions were pooled and frozen at Ϫ20°C in 50% KPEM and 50% glycerol. Native RI subunit from bovine lung contained saturating levels of cAMP as determined by assay of cAMP content (25), and contained ϳ5-10% PKG proteolytic fragments as detected by SDS-PAGE.
Preparation of cIMP-saturated Native RI Subunit-Native RI subunit (purified from bovine lung, above) was concentrated and the glycerol was removed by multiple centrifugations and dilutions on a Centricon-30 device. cIMP (at a final concentration of 1.1 mM) was added to the RI subunit solution (10:1 ratio of cIMP to cyclic nucleotide binding sites), and the solution was stored at 4°C for several weeks to facilitate cAMP/cIMP exchange. RI subunit was then chromatographed on a Sephadex G-25 column (0.9 ϫ 13 cm) equilibrated in KPEM to resolve protein from free cyclic nucleotide. cIMP-saturated RI subunit was pooled and concentrated using a Centricon-30 device.
Determination of Stoichiometry for cIMP Bound to RI Subunits-287 g of WT or mutant RI subunit was denatured at 95°C for 2 h in the presence of 13.5 mM HCl (final pH ϳ2-3) to release cyclic nucleotide. The solution was briefly centrifuged at 15-min intervals during incubation. Following neutralization, the sample was chromatographed at 23°C on a Sephadex G-25 superfine column (0.9 ϫ 13 cm) equilibrated in KPEM to purify cyclic nucleotide (25). cIMP elution was quantitated using the cIMP extinction coefficient (⑀ 1 mM ϭ 12.3 absorbance units at 248 nm). The released cyclic nucleotide was verified to be cIMP by a UV absorbance scan at 220 -320 nm.
Protein Sequence Analyses-The NH 2 -terminal sequence of various proteins was determined by sequential Edman degradation. The analyses were performed by the core sequencing facility at the University of Washington, Seattle, WA and by the Peptide Sequencing and Amino Acid Analysis Shared Resource at Vanderbilt University.
Cyclic Nucleotide Binding and Dissociation Assays-The cyclic nucleotide binding assay was used to identify RI subunit fractions eluted from gel filtration columns and to quantitate purified RI subunits.
[ 3 H]cAMP binding activity was measured as described previously (34) except that 1.1 M [ 3 H]cAMP (2800 -5500 cpm/pmol cAMP) was used, and the reaction mixture was diluted with 2.5 ml of ice-cold 10 mM potassium phosphate, pH 6.8, 1 mM EDTA (KPE) and the tubes were each rinsed with 2.5 ml of KPE. RI subunit was quantitated based on a stoichiometry of 2 mol of cAMP/mol of RI subunit monomer.
[ 3 H]cAMP dissociation assays were performed as described previously (35) using 0.2 M [ 3 H]cAMP (3 ϫ 10 4 cpm/pmol) and 4 nM RI subunit in the equilibrium exchange reaction. A 500-fold molar excess of unlabeled cAMP was used in the dissociation reaction.
Inhibition of C Subunit by Native, WT, and Mutant RI Subunits-Immediately before use, C subunit was diluted in Dilution Buffer (50 mM potassium phosphate, pH 6.8, 0.1 mM dithiothreitol, and 1 mg/ml bovine serum albumin (BSA)). A final concentration of 21 pM C subunit was preincubated with varying concentrations of WT or mutant RI subunit (in KPEM) in Kinase Mix (20 mM Tris, pH 7.4, 20 mM magnesium acetate, 0.1 mM ATP, 0.5 mg/ml BSA, and 0.1 mM isobutylmethylxanthine) for 15-30 min at 30°C. The protein kinase assay was initiated by the addition of a final concentration of 81 M Kemptide, 0.071 mM ATP, and [␥-32 P]ATP (1000 cpm/pmol). The total assay volume was 35 l. Following a 60 -120-min incubation at 30°C, the reaction was terminated by spotting 20 l of reaction mixture onto P-81 phosphocellulose paper and immediately placing the papers into 75 mM phosphoric acid. The P-81 papers were washed a minimum of five times in 75 mM phosphoric acid, dried and counted in 10 ml of non-aqueous scintillant, or counted by Cerenkov radiation.
Preparation of PKA Holoenzyme from WT RI Subunit and Native C Subunit-cIMP-saturated WT RI subunit (3.8 M) and native C subunit (14.7 M) were incubated for 30 min at 30°C in Kinase Mix containing 0.23 mM ATP and 10 mg/ml BSA (final volume ϭ 15 l). 1 l of reaction was diluted into 19 l of Dilution Buffer and saved for subsequent assay of kinase activity (Ϯ cAMP). The remainder of the reaction was cooled on ice and then loaded onto a Sephadex CM50 column (0.9 ϫ 0.5 cm) equilibrated in KPEM at 4°C. Approximately 1 ml of KPEM was added once the sample had entered the resin, and 1-drop fractions (ϳ23 l) were collected. The fractions were assayed for phosphotransferase activity using the peptide substrate Kemptide (LRRASLG) (81 M) in the presence of Kinase Mix containing 0.3 mM ATP and 10 mg/ml BSA plus 40 M Kemptide, [␥-32 P]ATP (ϳ33 cpm/pmol), with or without 2 M cAMP. Fractions collected between 47 and 138 l (ϳdrops 3-6) contained ϳ70% of the total holoenzyme, and the kinase activity was 87% cAMP-dependent. Fractions collected between 139 and 230 l (ϳdrops 7-10) contained ϳ18% of the total holoenzyme, and the kinase activity was 73% cAMP-dependent.
Determination of cIMP K a for Cyclic Nucleotide-depleted Holoenzyme (WT RI Subunit and Native C Subunit)-The activation constant (K a ) of cIMP for PKA holoenzyme was determined by measuring the transfer of phosphate to Kemptide in the presence of cIMP (0 -20 M). Holoenzyme was diluted to 24 nM in Dilution Buffer. 10 nM holoenzyme was combined with 0.1 mg/ml Kemptide (118 M) and cIMP (0 -20 M) in Kinase Mix containing 0.2 mM ATP and 10 mg/ml BSA plus [␥-32 P]ATP (ϳ30 cpm/pmol). The 24-l reaction was incubated for 10 min at 30°C before spotting 15 l onto P-81 phosphocellulose paper. The papers were washed five times in 75 mM phosphoric acid, dried, and counted in scintillant. A 20% basal kinase activity was subtracted from each activity determination prior to calculating the K a . Preparation

RESULTS AND DISCUSSION
Amino-terminal truncation and site-specific mutants (Table  II) of the RI subunit were employed to investigate the contributions that conserved residues within and near the pseudosubstrate region make toward the potency with which RI subunit inhibits C subunit kinase activity. The individual roles of Arg 92-95 were assessed by creating truncation mutants of RI subunit that were sequentially deleted through each of residues 91 through 95. Amino-terminal truncation mutants were also prepared in which Arg 94 or Arg 95 was replaced with Ala by site-directed mutagenesis. To examine a potential role for additional residues in the putative autoinhibitory domain, two other highly conserved residues (Ile 98 and Glu 101 ) were also investigated utilizing site-specific mutants of full-length RI subunit. Ile 98 was mutated to Val, Ala, Gly, and Gln; Glu 101 was changed to Ala and Gln. Finally, to provide a more comprehensive study, and to minimize certain pitfalls inherent to each form of RI subunit, inhibition of C subunit was examined using four forms of RI subunit: 1) cAMP-saturated, 2) cIMPsaturated, 3) cyclic nucleotide-saturated with excess cyclic nucleotide in the solution, and 4) cyclic nucleotide-free.

Confirmation of the Structural Integrity of Recombinant RI Subunits
All segments of RI subunit cDNA subjected to a mutagenesis reaction were sequenced to verify the integrity of the mutation and subcloning sites before overexpression in bacteria. The recombinant RI subunits were purified to apparent homogeneity as shown in Fig. 1 (see "Experimental Procedures"). Consistent with previous results (36), the full-length RI subunit migrated at M r ϳ 49,000 when analyzed by SDS-PAGE, considerably larger than the molecular weight calculated by the amino acid sequence for bovine skeletal muscle RI subunit (42,804) (37). The truncated RI subunits migrated at M r ϳ35,000 on SDS-PAGE, but were calculated to have molecular weights of ϳ31,790 -32,414 by amino acid sequence. The puri-fied WT, E101Q, and all truncated RI subunits were subjected to amino acid sequence analysis to verify the predicted primary structure at the amino termini. The initiator Met was present in proteins in which the penultimate residue was Arg, but it was absent from those in which this residue was Ala (WT, E101Q, ⌬1-94*) or Gly (⌬1-95). These results were consistent with the previous finding that cleavage of the initiator Met from proteins during post-translational modification is dependent on the size of the adjoining residue (38). Protein expressed from the ⌬1-93.R94A cDNA (two preparations) did not have the expected amino-terminal sequence. This mutant lacked the initiator Met as well as the Ala that was intended to replace Arg 94 . Since the product of ⌬1-93.R94A differed from ⌬1-94 only in the absence of the initiator Met residue it was named ⌬1-94*. Presumably Ala 94 was removed from the protein during cleavage of the initiator Met or during protein purification by an exoproteinase. The cAMP dissociation characteristics of the truncated and full-length mutants were compared to that of WT RI subunit to verify that the mutations did not impair the overall viability of the proteins. The cAMP dissociation rates for all mutant RI subunits were comparable to that of WT. Fig.  2 displays the characteristic biphasic cAMP dissociation (exchange) pattern of WT, and those of representative mutants: ⌬1-94, E101A, and I98A.

Inhibition of C Subunit by cAMP-saturated RI Subunits
In the absence of cAMP, PKA is maintained as an inactive holoenzyme complex R 2 C 2 with an affinity of ϳ0.2 nM between R and C subunit (39). cAMP binding to the R 2 in the R 2 C 2 complex decreases its affinity for C subunit by 4 -5 orders of magnitude (40 -42), resulting in dissociation of the R and C subunits and subsequent activation of the kinase (1) (Reaction 1). A ternary complex involving the R and C subunits and cAMP (exemplified as R 2 cAMP 4 C 2 ) is formed as an intermediate during both activation and inactivation of PKA (40,(43)(44)(45). Other types of ternary complexes (e.g. R 2 cAMP 2 C) are also formed, but are not shown in this simplified Reaction 1 (46,47).
The forward reaction is exemplified in Fig. 3A, which shows that holoenzyme was activated by cAMP in a concentration-dependent manner. Activation was evident when the cAMP concentration exceeded 2 nM, and holoenzyme was fully activated at ϳ75-100 nM cAMP. Half-maximal activation of WT holoenzyme (K a ) was 21 nM as determined by a Hill plot. The reverse of Reaction 1, inhibition of C subunit by formation of the R 2 C 2 complex, is facilitated in vivo by the action of cyclic nucleotide phosphodiesterases which degrade cAMP (48,49). Because of the high affinity with which cAMP is bound to R subunit in vitro, and because of its dissociation-reassociation characteristics, cAMP remains associated with R subunit throughout extensive molecular sieve chromatography, ion exchange chromatography 2 and dialysis (3). In fact, a significant portion (ϳ20%) of intracellular R subunit contains bound cAMP even in the basal state (50). Because much of the in vitro characterization of native R subunit has utilized the cAMPbound form of the protein, the reverse reaction was initially examined by incubating cAMP-saturated native RI subunit (R 2 cAMP 4 ) with the purified C subunit (Fig. 3B). C subunit was inhibited by RI subunit in a concentration-dependent manner until the concentration of bound cAMP added to the reaction exceeded 2 nM, at which point the inhibition by RI subunit was reversed, presumably because the equilibrium of Reaction 1 shifted back toward the right. These results are consistent with the results presented in Fig. 3A, which showed that the inactive holoenzyme complex predominated until the added cAMP reached 2 nM, whereupon the equilibrium shifted to the right, toward cAMP saturation of the R subunit and dissociation of the holoenzyme complex. The concentration of cAMP in the reaction at the point at which the equilibrium shifted to the right (2 nM) correlated well with the mean equilibrium binding constant (K D ) of cAMP for RI subunit (ϳ1.4 nM) (51). When the concentration of cAMP in the reaction exceeded the K D (Fig. 3,  A and B), the equilibrium of cAMP binding to RI subunit shifted toward formation of the R 2 cAMP 4 complex. These considerations demonstrate the technical difficulties involved in quantitating the potency with which R subunit inhibits C subunit kinase activity when using R subunit maintained in the physiological, cAMP-bound form. Therefore, attempts were made to remove cAMP from RI subunit either by exchanging it for a lower affinity cyclic nucleotide or by urea denaturation as described in the following sections.

Inhibition of C Subunit by cIMP-Saturated RI Subunits
In theory, a quantitative measure of C subunit inhibition would be possible if the RI subunit preparation were bound with a low affinity cyclic nucleotide instead of with cAMP. Thus, cAMP was replaced with cIMP, which exhibits an affinity that is ϳ20-fold lower than that of cAMP for the type I␣ RI subunit cyclic nucleotide binding sites (52). Since expression of RI subunit in E. coli produces the cAMP-saturated form of the enzyme, the exchange of cIMP for cAMP was performed during the preparation of bacterial extracts (see "Experimental Procedures"). This facilitated subsequent purification of RI subunits using N 6 -H 2 N-(CH 2 ) 2 -cAMP Sepharose affinity resin. RI subunit was eluted from the affinity resin with cIMP, and the excess cIMP was separated from RI subunit by molecular sieve chromatography, producing cIMP-saturated recombinant RI subunits (ϳ2 mol of cIMP/mol of RI subunit monomer) (see "Experimental Procedures"). The cIMP-saturated form of type I␣ native RI subunit (from bovine lung) was also prepared. In this case, the purified preparation of cAMP-saturated native RI subunit was incubated with cIMP to allow exchange, followed by molecular sieve chromatography (see "Experimental Procedures"). The retention of cIMP by the RI subunits was unexpected since removing bound cAMP by exchanging for a lower affinity cyclic nucleotide (e.g. cGMP), followed by a procedure such as ion exchange chromatography (30), molecular sieve chromatography (41) or dialysis (53) to remove the lower affinity cyclic nucleotide, is a strategy commonly employed by investigators, including our own laboratory (30).
Unlike the cAMP-saturated RI subunit, the cIMP-saturated WT and native RI subunits completely inhibited C subunit in a concentration-dependent manner; both WT and native proteins exhibited similar IC 50 values, 0.36 nM and 0.40 nM, respectively (Fig. 3C). The IC 50 values were determined by Hill plots and represent the concentration of RI subunit required to inhibit C subunit kinase activity by 50%. The cIMP concentrations introduced into these assays via the R 2 cIMP 4 complex (50 nM) were calculated to be well below the K a of cIMP for WT holoenzyme, which was ϳ750 nM as determined by a Hill plot (data not shown). The inhibition curves for selected cIMP-saturated RI subunit mutants are illustrated in Fig. 4; complete or nearly complete inhibition of C subunit was achieved with each of the RI subunit mutants.
Residues 1-91, Arg 92-95 Tetrad- Table III summarizes  subunit mutants (RcIMP 2 ) that were used to examine the contribution of the amino-terminal segment of RI subunit, and each residue of the Arg 92-95 tetrad. The cIMP-saturated truncated RI subunits are represented as RcIMP 2 instead of R 2 cIMP 4 since the dimerization domain has been eliminated by truncating the protein. Truncation of the amino-terminal segment of RI subunit through Gly 91 (⌬1-91) did not impair the inhibitory potency of this mutant compared to that of WT. In the crystal structure of the C subunit⅐PKI-(5-24) complex, the PKI P Ϫ6 Arg forms a hydrogen bond with Glu 203 , the P Ϫ3 Arg forms hydrogen bonds with Glu 127 and Glu 331 and the P Ϫ2 Arg hydrogen bonds with Glu 230 and Glu 170 (4). Since the RI subunit pseudosubstrate region contains four arginine residues (P Ϫ5 ,P Ϫ4 ,P Ϫ3 ,P Ϫ2 ), it is possible that each Arg could interact with one or more of these Glu residues in C subunit, and thus contribute to the high affinity interaction with C subunit. However, truncation through Arg 92 (P Ϫ5 ) or Arg 93 (P Ϫ4 ) did not detectably impair the inhibitory potency of either of these mutants (⌬1-92 and ⌬1-93). The importance of the P Ϫ3 residue (Arg 94 ) was investigated using two RI subunit mutants (⌬1-94 and ⌬1-94*) which were truncated through Arg 94 . The inhibitory potency of ⌬1-94 was ϳ17000-fold lower than that of WT, while the ⌬1-94* mutant, which lacks the initiator Met, was ϳ21000-fold less potent than WT. Thus, the initiator Met was , or I98G (f) RI subunits as described in Fig. 3B. The reactions were initiated by the addition of Kemptide substrate, and the incubation proceeded for an additional 60 -120 min at 30°C. The reactions were terminated, and the amount of Kemptide phosphorylation was determined. 100% catalytic activity was determined to be ϳ6 pmol/min/ng of C in the absence of RI subunit and cAMP. Each curve represents the mean of Ն4 assays for the Arg 92-95 mutants, 2 assays for the Ile 98 mutants, and 31 assays for WT RI subunit. not detrimental to the interaction of ⌬1-94 with C subunit. The requirement for a P Ϫ2 Arg (Arg 95 ) was examined using the truncated ⌬1-93.R95A mutant. Substitution of Arg 95 with Ala within ⌬1-93.R95A profoundly reduced the inhibitory potency of this truncated mutant (ϳ31000-fold). Removal of all of the amino-terminal Arg residues, ⌬1-95, reduced the inhibitory potency of RI subunit for C subunit by ϳ54000-fold. This was an additive decrease in effectiveness compared to that observed when Arg 94 or Arg 95 were deleted or mutated individually.
These results demonstrated that both Arg 94 and Arg 95 are critical determinants for potent inhibition of C subunit, but that neither Arg 92 nor Arg 93 is detectably involved. The results also indicated that the amino-terminal 91 residues are not involved in autoinhibition, despite the established role of the PKI residues which correspond to residues 79 -87 in RI subunit (4,18). The results are in agreement with similar studies in full-length RI (8) and RII subunits (7), and with PKI peptide analog studies (10,11), in which the P Ϫ3 and P Ϫ2 Arg residues were shown to be required for C subunit inhibition. The data presented here suggested that the requirement for a basic residue is approximately the same for the P Ϫ3 or P Ϫ2 position, with perhaps a slightly greater requirement at the P Ϫ2 position. This conclusion is consistent with findings showing that a basic residue is of slightly greater significance at the P Ϫ2 position than at the P Ϫ3 position of the peptide substrate Kemptide (6), but contrary to findings using PKI peptide analogs (11).
Although neither Arg 92 (P Ϫ5 ) nor Arg 93 (P Ϫ4 ) is required for potent inhibition of C subunit, it is possible that these highly conserved residues make minor contacts with C subunit that would be more evident in the absence of the P Ϫ3 or P Ϫ2 inhibitory residues (Arg 94 or Arg 95 ). It is also plausible that some flexibility exists in the interaction between pseudosubstrate residues and C subunit that would permit a shift in the spatial location so that the P Ϫ4 and P Ϫ3 Arg residues could serve the same role as the P Ϫ3 and P Ϫ2 Arg residues. The P Ϫ4 -P Ϫ3 arrangement of basic residues is seen in naturally occurring PKA substrates such as phosphorylase kinase (␤ subunit) (54) and glycogen synthase (site 1b) (55). These substrates have a P Ϫ4 Lys and a P Ϫ3 Arg residue, but do not possess an Arg residue at the P Ϫ2 position. The ⌬1-92.R95A mutant, in which an Ala residue was substituted for the P Ϫ2 Arg, was used to examine the interaction between the RI subunit P Ϫ4 -P Ϫ3 Arg residues and the active site of C subunit. The ⌬1-92.R95A mutant was an extremely poor inhibitor of C subunit, comparable to ⌬1-93.R95A (Table III), clearly demonstrating that the RI and C subunit interaction is not flexible enough to permit spatial repositioning of the Arg residues by even a single residue.
Effects of Substitutions at Ile 98 -The P ϩ1 position is highly conserved as Ile or Val in the autoinhibitory domains of all species of PKA, PKG and in PKI. Substitution of Ile 22 (P ϩ1 ) with Gly in PKI peptide analogs reduced the inhibitory potency of the peptide 150-fold (18). In the crystal structure of the C subunit⅐PKI-(5-24) complex, Ile 22 (P ϩ1 ) is situated in a hydrophobic pocket on C subunit comprising Leu 198 , Pro 202 , and Leu 205 , indicating that a P ϩ1 hydrophobic residue is essential to a high affinity interaction between PKI and C subunit (4). Since a role in C subunit inhibition has not been previously demonstrated for the R subunit P ϩ1 residue, mutants of the RI subunit P ϩ1 residue (I98V, I98A, I98G, and I98Q) were used to investigate the importance of aliphatic side chain length and steric constraints on the ability of RI subunit to potently inhibit C subunit. The IC 50 values for the cIMP-saturated, full-length RI subunits (R 2 cIMP 4 ) mutated at the P ϩ1 position (Ile 98 ) are presented in Table IV. Conservative replacement of Ile 98 with Val (I98V) in full-length RI subunit did not impair the IC 50 value of I98V compared to that of WT. When the aliphatic side chain length at the P ϩ1 position was reduced the IC 50 values were increased by ϳ5300-fold and 14000-fold for I98A and I98G, respectively, compared to that of WT. When the side chain length was increased, as in the I98Q mutant, the IC 50 value was ϳ4200-fold greater than that of WT. The loss of inhibitory potency was greatest for I98G, but substitution of Gly at the P ϩ1 position could cause a conformational change that alters the position of other critical residues in the pseudosubstrate region. The increased length of the Gln side chain and its hydrophilic nature are the most probable factors causing the reduced inhibitory potency for the I98Q mutant. The results suggested that the RI subunit P ϩ1 hydrophobic residue is quite important for a potent interaction with C subunit, and that the length of the aliphatic side chain is crucial. Strict steric constraints at the P ϩ1 site are also indicated by Kemptide analog studies. Substitution of the P ϩ1 residue in Kemptide with Pro (LRRASPG) resulted in a significant reduction in substrate capacity (56), but introduction of N ␣ -methyl Leu in place of Leu at the P ϩ1 position resulted in only a moderate loss of substrate capacity (57).
Effects of Substitutions at Glu 101 -Glu 101 , at the P ϩ4 position in RI subunit, is invariant in the regulatory domain of all known sequences of PKA and PKG. To examine the putative electrostatic interaction between Glu 101 and Lys residues of C subunit (see Introduction), the acidic charge of Glu 101 was eliminated by substitution of Glu 101 with Gln (E101Q) or Ala (E101A) in full-length RI subunit. The IC 50 values for the cIMP-saturated Glu 101 mutants (Table IV) were similar to that      of WT RI subunit. Despite the highly conserved nature of the P ϩ4 Glu and its proximity to the pseudosubstrate site, Glu 101 does not appear to play an important role in C subunit inhibition, although it could serve some other function.

Influence of cIMP on Inhibition of C Subunit by RI Subunits
Although complete or nearly complete inhibition of C subunit was achieved using each of the cIMP-saturated RI subunit mutants, concentrations as high as 20 -50 M were required by most RI subunit mutants to inhibit C subunit completely, compared to 0.01 M for WT (Fig. 4). C subunit was inhibited in those reactions containing 50 M mutant RI subunit (100 M cIMP), despite the fact that the cIMP concentration in the reaction was calculated to exceed the K a by Ͼ100-fold. The results of studies with cAMP-saturated RI subunit (Fig. 3B) would have predicted less C subunit inhibition as the concentration of the R 2 cIMP 4 or RcIMP 2 complex approached the K a of cIMP for holoenzyme (750 nM). When using cIMP-saturated WT RI subunit, cIMP would not be expected to interfere with holoenzyme formation since the assays are performed using very dilute RI subunit (25 nM) where the concentration of cIMP (50 nM) would be far below the K a (750 nM). It is suggested that the large rightward shift exhibited in the inhibition curves by certain RI subunit mutants must reflect the following considerations. 1) The decreased affinity between the pseudosubstrate site of the mutant RI subunit and the catalytic site of C subunit causes a rightward shift in the equilibrium of Reaction 1 (toward R 2 cIMP 4 (or RcIMP 2 ) and active C subunit); 2) the necessity of using mutant R 2 cIMP 4 (or RcIMP 2 ) at concentrations that vastly exceed the mean equilibrium binding constant of cIMP for RI subunit (ϳ28 nM) (51, 52) causes a dramatic shift in the equilibrium of cIMP binding to RI subunit (toward formation of the R 2 cIMP 4 (or RcIMP 2 ) complex). The second consideration causes a shift of the equilibrium of Reaction 1 even further to the right than that caused by mutation of RI subunit alone, thereby resulting in the high IC 50 values determined for the cIMP-saturated RI subunit mutants. This is supported by the observation that the equilibrium of Reaction 1 was progressively shifted to the right when cIMP-saturated ⌬1-94* (9.8 M RI subunit:19.6 M cIMP) and C subunit (21 pM) were incubated in the presence of increasing concentrations of exogenous cIMP (up to 20 M) (data not shown).

Inhibition of C subunit by R 2 cAMP 4 or R 2 cIMP 4 in the Presence of Excess Cyclic Nucleotide
Under physiological conditions the PKA holoenzyme complex (R 2 C 2 ) is inactive (see Reaction 1). The PKA ternary complex (R 2 cAMP 4 C 2 ) is also largely inactive (IC 50 ϳ 15 M for type I enzyme) when the reverse reaction of Reaction 1 is measured using pharmacological concentrations of R 2 cAMP 4 and a vast excess of cAMP, such that the RI subunit remains in the R 2 cAMP 4 form (40). Since the IC 50 values obtained for several cIMP-saturated RI subunit mutants approximate the IC 50 value obtained for this native ternary complex, it was necessary to determine if these values might reflect the inhibited mutant ternary complexes instead of the R 2 C 2 (full-length RI subunit mutants) or RC (truncated RI subunit mutants) holoenzyme complexes.
The IC 50 of the WT ternary complex was measured by incubating C subunit with either WT-R 2 cIMP 4 or WT-R 2 cAMP 4 in the presence of a large excess (500 M) of cIMP or cAMP (Fig.  5). The WT-R 2 cIMP 4 C 2 and WT-R 2 cAMP 4 C 2 ternary complexes exhibited IC 50 values similar to each other (IC 50 ϳ10 -12 M), and to that published for the native ternary complex (40). These values were ϳ4 -5 orders of magnitude higher than the IC 50 value measured for WT holoenzyme (R 2 C 2 ) (0.36 nM; Table  III). Inhibition of C subunit by a truncated mutant, cIMPsaturated ⌬1-93.R95A RI subunit, was also examined in the presence of excess cIMP to measure inhibition of the ⌬1-93.R95A ternary complex (RcIMP 2 C). Only 20% of C subunit activity was inhibited by 50 M cIMP-saturated ⌬1-93.R95A when assayed in the presence of 500 M cIMP (Fig. 5), whereas cIMP-saturated ⌬1-93.R95A exhibited an IC 50 value of 11.2 M when assayed in the absence of excess cIMP (Table III). Similarly, only 16% of C subunit activity was inhibited by 5 M cIMP-saturated I98A when assayed in the presence of 500 M cIMP (data not shown), whereas cIMP-saturated I98A exhibited an IC 50 value of 1.9 M when assayed in the absence of excess cIMP (Table IV). Assuming that there is no RC or R 2 C 2 complex present when 500 M cIMP is included in the reaction, these results suggest that the IC 50 values obtained for the cIMP-saturated mutant RI subunits in the absence of excess cyclic nucleotide are indeed a measure of the inhibited RC or R 2 C 2 holoenzyme complexes, and do not reflect the inhibited ternary complex. The results also indicate that the interactions between C subunit and WT-or mutant-R 2 cIMP 4 or mutant-RcIMP 2 are specific interactions and not due simply to the addition of a high concentration of RI subunit; the interaction of C subunit with either ⌬1-93.R95A-RcIMP 2 or I98A-R 2 cIMP 4 was profoundly reduced compared to that interaction with WT-R 2 cIMP 4 when measured in the absence or presence of a large excess of cIMP. This also suggests that the same crucial resi-  5. Effect of excess cyclic nucleotide on the inhibition of C subunit by RI subunit. C subunit (21 pM) was preincubated with cIMP-saturated WT RI subunit in the presence of 500 M cIMP (E) or 500 M cAMP (É), or C subunit was preincubated with ⌬1-93.R95A (Ⅺ) in the presence of 500 M cIMP at 30°C for 19 min under the same conditions as described in Fig. 3B. The reactions were initiated by the addition of Kemptide substrate, and the incubations proceeded for an additional 120 min at 30°C. The reactions were terminated, and the amount of Kemptide phosphorylation was determined.
dues interact with C subunit regardless of whether the measured inhibition is that of the ternary complex or of the holoenzyme complex. It follows from the above considerations that since a significant portion of R subunit is bound with cAMP under physiological conditions (50), a natural biological mutation of one of the pseudosubstrate Arg residues or the Ile residue would form constitutively active PKA.

Inhibition of C Subunit by Cyclic Nucleotide-free RI Subunit Mutants
For the sake of comparison, it seemed prudent to examine the effect of RI subunit mutation(s) on inhibitory potency using the cyclic nucleotide-free form of RI subunit, whereby the interaction of R and C subunit is measured directly without the interference of cAMP or cIMP (Reaction 2).
Investigators have previously used R subunit that was made cyclic nucleotide-free by urea denaturation (followed by renaturation) to examine the effect of R subunit mutation(s) on the inhibition of kinase activity (8,58). Since the renatured R subunit has been shown to have the same affinity for C subunit as that of non-denatured R subunit (59), this method was selected for the preparation of cyclic nucleotide-free RI subunit in the present study. R subunits that have been treated with urea, however, have an increased rate of cyclic nucleotidedissociation from the binding sites (i.e. decreased affinity for cyclic nucleotide) (60). Since the effect on cyclic nucleotide dissociation was found to be directly related to the length of time that the R subunits were exposed to urea (60), the exposure of the RI subunits to urea was temporally minimized as described under "Experimental Procedures." The IC 50 values for the cyclic nucleotide-free RI subunits were determined by Hill plots and are summarized in Tables  III and IV. Urea treatment of WT RI subunit had minimal effect on the potency of inhibition of C subunit activity compared to that of untreated WT RI subunit (0.72 nM and 0.36 nM, respectively). The IC 50 values obtained for the cyclic nucleotide-free RI subunit mutants were considerably lower than those for the cIMP-saturated mutant RI subunits, but within the same range as those previously reported for urea-treated R subunit mutants (8,58). Deletion of the first 91 amino acid residues (⌬1-91) of RI subunit did not impair the inhibitory potency of this mutant toward C subunit kinase activity. Compared to WT, the IC 50 values for the ⌬1-94 and ⌬1-94* mutants were 43-and 10-fold greater, respectively, while those for ⌬1-93.R95A and ⌬1-92.R95A were 61-and 51-fold greater. When the P Ϫ3 and P Ϫ2 Arg residues were both deleted (⌬1-95), the IC 50 value increased by 746-fold over that of WT, which was considerably more than additive when compared with the IC 50 of mutants in which these residues were deleted or mutated individually. These results suggested that Arg 94 and Arg 95 interact in a synergistic manner with the active site on C subunit. The inhibitory potency of a single cyclic nucleotidefree Ile 98 mutant, I98A, was also measured and was 4-fold less potent toward C subunit than was WT.
Four different forms of RI subunit (cAMP-bound, cIMPbound, cIMP-bound plus excess cyclic nucleotide, and cyclic nucleotide-free) have been used in the present study. Each form provides useful and sometimes distinct information about R subunit/C subunit interactions. The cAMP-saturated form of R subunit provides insights into cellular interactions, but presents a technical impediment in quantitatively assessing the interaction between R and C subunit. However, since a signif-icant portion of R subunit is bound with cAMP under physiological conditions (50), the results of mutating the RI subunit reveal that a natural mutation of one of the pseudosubstrate Arg residues or the Ile residue would form constitutively active PKA in body tissues. The RI subunit that is saturated with the lower affinity cyclic nucleotide, cIMP, can be better used to determine the relative contribution of particular residues toward the inhibition of C subunit. However, a true quantitative measurement of this interaction is hindered even with this RI subunit since mutation of critical pseudosubstrate site residues reduces the R-C subunit affinity and necessitates the use of RcIMP 2 or R 2 cIMP 4 concentrations that far exceed the cIMP K D for RI subunit. This causes an exaggerated increase in the IC 50 value. However, because the IC 50 values are exaggerated, this form of R subunit is ideal for detecting small differences in the contribution of particular residues toward C subunit inhibition that might otherwise have been discounted. When assayed in the presence of excess cyclic nucleotide, the cIMPsaturated RI subunits are specific, but extremely weak inhibitors of C subunit. It is suggested that a more quantitative assessment of the R-C interaction is obtained using RI subunit that is made cyclic nucleotide-free by urea denaturation. Furthermore, considerably less of this RI subunit is required for inhibition than is required when using cIMP-saturated RI subunit, thus facilitating most experimental protocols. Although evidence suggests that the R and C subunit interaction is not impaired as a result of urea treatment of the RI subunit, this possibility cannot be ruled out since urea denaturation alters the affinity of R subunit for cyclic nucleotide (60).
In summary, the results of this study establish that both the P Ϫ2 and P Ϫ3 Arg, and, to a lesser extent, the P ϩ1 Ile residues are critical for potent inhibition of C subunit by RI subunit. Neither the P Ϫ4 Arg, P Ϫ5 Arg, nor the P ϩ4 Glu residue has a detectable effect. The results indicate that the interactions between the pseudosubstrate site Arg residues and the catalytic site acidic residues are not flexible enough to tolerate spatial re-positioning of the two required Arg residues from the P Ϫ2 -P Ϫ3 to the P Ϫ3 -P Ϫ4 positions.