The Major Catalytic Subunit Isoforms of cAMP-dependent Protein Kinase Have Distinct Biochemical Properties in Vitro and in Vivo *

Two isoforms of the catalytic subunit of cAMP- dependent protein kinase, C (cid:97) and C (cid:98) 1, are known to be widely expressed in mammals. Although much is known about the structure and function of C (cid:97) , few studies have addressed the possibility of a distinct role for the C (cid:98) proteins. The present study is a detailed comparison of the biochemical properties of these two isoforms, which were initially expressed in Escherichia coli and purified to homogeneity. C (cid:98) 1 demonstrated higher K m values for some peptide substrates than did C (cid:97) , but C (cid:98) 1 was insen-sitive to substrate inhibition, a phenomenon that was observed with C (cid:97) at substrate concentrations above 100 (cid:109) M . C (cid:97) and C (cid:98) 1 displayed distinct IC 50 values for the (cid:97) and (cid:98) isoforms of the protein kinase inhibitor, protein kinase inhibitor (5–24) peptide, and the type II (cid:97) regulatory subunit (RII (cid:97) ). Of particular interest, purified type II holoenzyme containing C (cid:98) 1 exhibited a 5-fold lower K a value for cAMP (13 n M ) than did type II holoenzyme containing C (cid:97) (63 n M ). This latter result was extended to in vivo conditions by employing a transcriptional acti- vation assay. In these experiments, luciferase reporter activity in COS-1 cells expressing RII (cid:97) 2 C (cid:98) 1 2 holoen- zyme was half-maximal at 12-fold lower concentrations of 8-(4-chlorophenylthio)-cAMP and 5-fold lower concen- trations of forskolin than in COS-1 cells expressing RII (cid:97) 2 C (cid:97) 2 holoenzyme. These results provide evidence that type II holoenzyme formed with C (cid:98) 1 is preferen-tially activated by cAMP in vivo and suggest that acti- vation of the holoenzyme is determined in part by interactions between the regulatory and catalytic subunits that have not been described previously.

cAMP exerts its effects in mammalian cells primarily through the activation of cAMP-dependent protein kinase (cAK), 1 a tetrameric enzyme consisting of two catalytic (C) and two regulatory (R) subunits. Occupation of the cyclic nucleotide binding sites of the R subunits by cAMP results in relief of R subunit inhibition of the C subunits and dissociation of the holoenzyme complex. The dissociated, active C subunit can then affect the cell physiology via phosphorylation of a wide variety of protein substrates (1)(2)(3).
One possible explanation for the diversity of cellular responses to cAMP is the presence of multiple isoforms of cAK. Three C and four R mammalian subunit isoforms have been described to date: C␣, C␤, C␥, RI␣, RI␤, RII␣, and RII␤. Splice variants of the prototypic C␤ protein (C␤1) also have been reported, giving rise to C␤2 and C␤3 isoforms (4,5). C␣ and C␤1 have been cloned from numerous sources (6 -9), whereas C␥ has only been cloned from human testis and has not been reported in other species (10). The amino acid sequences of C␣ and C␤1 within a given species are 91% identical (7); however, the amino acid identity of C␣ proteins from different species is significantly greater (98 -100%) (7,11), suggesting that each kinase plays a distinct role(s) in cellular regulation (7,12). Furthermore, C␣ is ubiquitously expressed in mammalian tissues, whereas C␤1 is most highly expressed in brain and reproductive tissues (7,13). Despite these differences, no biochemical property specific to either C subunit isoform has been detected to date.
The four R subunit isoforms also have been cloned and shown to possess different tissue distributions (13)(14)(15)(16). In addition, the RII isoforms but not the RI isoforms can be tethered within the cell via interactions with cAK anchoring proteins (17,18). The resulting localization of type II holoenzymes to subcellular compartments may provide a means for selectively activating the type II cAK or influencing substrate availability (18,19). Unlike the C subunit isoforms, intrinsic differences between the R subunits have been shown to confer distinct biochemical properties upon holoenzymes. For example, whereas each RI isoform has an alanine at the phosphoacceptor site in the pseudosubstrate sequence, the RII isoforms have a phosphorylatable serine or threonine at this position (20). Furthermore, cAK holoenzymes containing RI␤ demonstrated a decreased (3-7-fold) K a value for cAMP when compared with holoenzymes formed with RI␣ (21,22). The increased cAMP sensitivity of type I␤ holoenzymes offered a biochemical basis for the functional effects observed in mice that selectively lacked RI␤ (23). Targeted disruption of the gene encoding RI␤ in mice resulted in severe defects in hippocampal long term depression and depotentiation at Schaffer collateral-CA1 synapses (23) and long term potentiation in the mossy fiber pathway (5). These effects were observed despite a compensatory increase in hippocampal RI␣, suggesting that holoenzymes with increased cAMP sensitivity may be required for discrete physiological functions (5,23).
Recent data generated from C␤1 knockout mice have provided the first evidence that C␤1 serves an in vivo function distinct from C␣ (5,24). The C␤1-deficient mice had unaltered total brain kinase activity yet lacked both long term depression and depotentiation and showed significant decreases in the late phase of long term potentiation at hippocampal Schaffer col-lateral-CA1 synapses. In addition, mossy fiber long term potentiation was essentially abolished. The phenotypic similarities between the RI␤ and C␤1 mutant mice are striking and may reflect a common biochemical property of holoenzymes possessing either of these subunit isoforms.
To determine whether catalytic or regulatory differences exist between C␣ and C␤1, we expressed the two proteins in Escherichia coli, purified them, and compared their affinities for nucleotide, substrate, and inhibitors. In addition, holoenzymes formed with recombinant RII␣ and C␣ or C␤1 were purified by gel filtration, and K a values for cAMP were obtained. We chose to investigate the type II holoenzyme because an earlier study showed no difference in K a(cAMP) values for type I holoenzymes containing C␣ or C␤1 (21). Significant differences between RII␣ 2 C␣ 2 and RII␣ 2 C␤1 2 revealed by in vitro experiments were then confirmed in an in vivo transfection system. The results show that C␣ and C␤1 have distinct enzymatic properties. Furthermore, C␤1, like RI␤, can increase holoenzyme cAMP sensitivity, most likely through novel R-C interactions distal to the catalytic site. Together, this work comprises the first detailed comparison of the C␣ and C␤ isoforms and demonstrates the first biochemical differences between the two isoforms.
Construction of an E. coli Expression Vector for C␤1-pET9d (Novagen) was digested with BamHI and filled in using the Klenow fragment of E. coli DNA polymerase. The treated pET9d vector was digested further with NcoI and isolated. Next, the vector pMTC␤1(G1A) (25) was digested with NcoI and SspI, yielding a 1210-base pair C␤1 cDNA fragment, which was isolated and ligated into the prepared pET9d vector. The resulting plasmid, pET9d.C␤1(G1A), was sequenced using the dideoxy termination procedure to ensure that the open reading frame was intact (26).
Expression and Purification of C Subunits-The pET9d.C␤1(G1A) plasmid was electroporated into E. coli BL21(DE3)/pLysS in order to express the C␤1 protein. The E. coli cultures containing the C␤1 expression plasmid were grown and induced as described previously (27). Using carboxymethyl-Sepharose and gel filtration chromatography (27), the C␤1 protein was purified to homogeneity as determined by silver staining (28) of SDS-PAGE gels (29). The purification of C␣ is described elsewhere (27).
Steady State Kinetic Analysis-The purified C subunits (3 nM) were assayed for phosphotransferase activity (11) using Kemptide (30), CREBtide (31), or IP 3 Rtide (32) as the phosphoacceptor peptides. The apparent K m values for the peptides were measured using 200 M ATP and varying peptide concentrations, whereas the apparent K m values for ATP were ascertained by varying the ATP concentration in the presence of 100 M Kemptide. The remaining components of the phosphotransferase assay mix were as described (11). One unit in these experiments is defined as 1 mol of phosphate transferred to the peptide substrate per min at 30°C. The apparent Michaelis constants for the peptide substrates and for ATP were graphically determined by double reciprocal plots of the primary data (33).
Inhibition of Phosphotransferase Activity by PKI Peptide and Proteins-C subunits (50 pM) were incubated for 10 min in the presence of increasing concentrations of PKI␣(5-24), MBP-PKI␣ (27), or MBP-PKI␤1 (27) in the phosphotransferase assay mix (11). MBP-PKI␣ and MBP-PKI␤1 are fusion proteins between the E. coli MBP and the coding regions of human PKI␣ or mouse PKI␤1. MBP has been shown to have no effect on the inhibitory activity of PKI (34). After the preincubation, Kemptide was added to a final concentration of 100 M to initiate the assay. The reaction was terminated after a 2-h incubation at 30°C. Control activities were 7.4 units/mg for C␣ and 7.3 units/mg for C␤1.
Inhibition of Phosphotransferase Activity by RII␣-The titrations of RII␣ activity were determined as before (35) with the modification of using 100 M Kemptide in the assay. The C subunit concentration in the assay was 50 pM.
Purification of Type II Holoenzymes and Determination of Their K a Values for cAMP-RII␣ 2 C␣ 2 and RII␣ 2 C␤1 2 holoenzymes were formed and purified by gel filtration as described by Yang et al. (36) using 8 g of purified RII␣ and a 2-fold excess of either purified C␣ or C␤1. Under these conditions, holoenzyme was formed equally well in the presence or the absence of 4 mM ATP and 6 mM magnesium acetate. Maximal cAMP-dependent kinase activity eluted from a Sephacryl S-200 column (Pharmacia Biotech Inc.) at the same volume for both C␣and C␤1containing type II holoenzyme. Peak cAMP-independent kinase activity, representing free C subunit, eluted at a significantly higher and identical volume for C␣ and C␤1. Purified RII␣ 2 C␣ 2 and RII␣ 2 C␤1 2 were assayed separately in the presence of 200 M ATP, 11 nM [␥-32 P]ATP (specific activity, 200 cpm/pmol), magnesium acetate (10 mM), Kemptide (30 or 100 M), and varying concentrations of cAMP. Holoenzymes were formed and purified immediately prior to each assay, and the amount of purified holoenzyme used in these experiments (2 l of a 1-ml fraction) was within the linear range of the assays. The reaction was terminated after 30 min, and radioactivity incorporated into Kemptide was determined by scintillation counting. The K a value for cAMP is defined as the concentration of cAMP required to achieve one-half the V max value. Control activities for RII␣ 2 C␣ 2 and RII␣ 2 C␤1 2 were 5.0 and 4.2 units/ml, respectively.
Expression of C Subunit Isoforms in COS-1 Cells-The plasmid pC-MV.RII␣ was constructed by subcloning the RII␣ cDNA insert from pET16b.RII␣ (35) into the BglII site of pCMV.Neo (25). The pCMV.C␣, pCMV.C␤1, and pSV 2 .␤gal expression vectors have been described previously (25). The human chorionic gonadotropin-luciferase expression vector used as a cAMP-responsive reporter (37) was a gift of Dr. Stan McKnight (University of Washington). COS-1 cells were transfected at 20% confluency on 10-cm plates with a total of 30 g of plasmid DNA using the calcium phosphate method (38). The experiments shown used 6 g of pCMV.C␣ or pCMV.C␤1, 0.5 g of human chorionic gonadotropin-luciferase, 2.0 g of pSV 2 .␤gal, and the indicated amounts of pCMV.RII␣. The total amount of plasmid DNA transfected was always brought to 30 g with the parental pCMV.Neo plasmid. 24 h after transfection the cells were washed and treated with forskolin or 8-(4chlorophenylthio)-cAMP for another 24 h before harvesting. Extracts were prepared by sonicating the cells in 500 l of a 100 mM potassium phosphate buffer (pH 7.2) containing 1 mM dithiothreitol, after which the extracts were assayed for ␤-galactosidase and luciferase activities as described (39).

RESULTS
In an earlier study, the ␣ and ␤1 isoforms of the catalytic subunit of cAMP-dependent protein kinase were purified from stably transfected cell lines (12). The purification process, however, yielded only microgram amounts of protein, and the V max values (12) were lower than other published reports (30,40). In order to investigate whether the C subunit isoforms have unique properties, we took advantage of a system already used to express and purify C␣ from E. coli to obtain large amounts of soluble C␣ and C␤1 for biochemical characterization (41,42).
Expression and Purification of C␤1-Murine C␤1 was expressed in E. coli and purified using ion-exchange and gel filtration chromatography (see "Experimental Procedures") (27). In a previous report, the expression of C␤1 at 37°C in the bacterial host strain BL21(DE3) resulted in mostly insoluble protein (43). However, this was not observed when C␤1 was produced at room temperature in BL21(DE3)/pLysS bacteria, an E. coli strain that allowed minimal promoter leakage (see "Experimental Procedures"). Also, the expression of C␣ at room temperature resulted in significantly more soluble C subunit than when cultures were grown at 37°C (41,42). The purified, soluble C␤1 migrates slower than C␣ during SDS-PAGE (Fig.  1A). This migration difference also was observed in immunoblots of C␣ and C␤1 from stably expressed cell lines using C subunit antibodies (11) and with pure preparations of C␣ and C␤1 (12). The calculated M r of unmodified C␤1(G1A) (40,590 Da) is only 122 Da larger than C␣(G1A) (40,468 Da), a differ-ence that should not be detectable under most conditions of SDS-PAGE. Because the C subunits are phosphoproteins, it was possible that the slower migration of C␤1 was due to differential phosphorylation. However, phosphoamino acid analysis did not reveal any difference in phosphothreonine or phosphoserine levels between C␣ and C␤1 (data not shown), and the phosphorylation of C␣ was similar to that reported previously (27).
Steady State Kinetic Analysis-To determine if C␤1 purified from E. coli is kinetically competent and to assess whether C␤1 and C␣ have different affinities for substrate, the apparent Michaelis constants were determined for various peptide substrates and for ATP. The peptides used in these experiments were derived from cAK phosphorylation sites in pyruvate kinase (Kemptide, LRRASLG) (30), the cAMP response element binding protein (CREBtide, KRREILSRRPSYR) (31), or the inositol 1,4,5-trisphosphate receptor (IP 3 Rtide, GRR-ESLTSFG) (32). The K m values of C␣ for Kemptide and IP 3 Rtide are 2.4-and 5.0-fold lower than those of C␤1, respectively, whereas the K m values of the two isoforms for CREBtide are not statistically different ( Fig. 2A and Table I). The V max value of C␣ for IP 3 Rtide is slightly (1.7-fold) higher than that of C␤1, although the V max values of C␣ and C␤1 for Kemptide and CREBtide are essentially identical ( Fig. 2A and Table I). These results raise the possibility that C␣ and C␤1 may exhibit distinct but subtle substrate selectivities.
As has been observed with C␣ previously (12,27,40,44), Kemptide inhibited C␣ activity half-maximally at 1.8 mM in these assays (Fig. 2B). Substrate inhibition of C␣ was also evident at similar concentrations of CREBtide and IP 3 Rtide (data not shown) using two independent preparations of recombinant enzyme. Interestingly, no inhibition of C␤1 activity was observed at Kemptide concentrations up to 3.6 mM (Fig. 2B) or CREBtide or IP 3 Rtide concentrations up to 9.0 mM (data not shown), suggesting that the C subunit isoforms have different interactions with these peptide substrates. Contrasting the differences observed with substrates, both the apparent K m and V max values of C␣ and C␤1 for ATP were very similar (Table I).
PKI Inhibition of Phosphotransferase Activity-The observation that C␣ and C␤1 have distinct affinities for some peptide substrates raised the possibility that they also could have unequal affinities for the two major protein kinase inhibitor isoforms, PKI␣ and PKI␤1. These PKI proteins are small, potent, and specific pseudosubstrate inhibitors of C subunit (45)(46)(47) that differ in their relative affinities for C␣ (34). Peptides derived from the amino terminus of PKI␣ (e.g., PKI␣(5-24)) remain potent inhibitors of C subunit and provide useful tools for examining cAK effects in vivo (47). To evaluate its interactions with C␤1 more accurately, PKI␣(5-24) was assayed for its ability to block phosphorylation of Kemptide by the purified recombinant C subunits. PKI␣(5-24) possessed a 2.6-fold lower IC 50 value for C␣ as compared with C␤1 (Fig. 3A, Table II).
Despite its low IC 50 , PKI␣(5-24) is still 10-fold less efficacious than full-length PKI␣ as an inhibitor of C␣ (48,49). This observation led to the proposal that the carboxyl-terminal region of PKI plays a part in mediating inhibition of the C subunit (47). If so, C subunit isoforms may participate in distinct interactions with full-length PKI protein isoforms. Therefore, fusion proteins between E. coli maltose binding protein and PKI␣ (27) and PKI␤1 (27) were tested for their ability to inhibit the kinase activity of the C subunit isoforms. PKI␣ and PKI␤1 had 2.8-and 1.7-fold lower IC 50 values with C␣ than with C␤1, respectively (Table II). Overall, these data suggest that differences exist between C␣ and C␤1 in the manner in which they interact with the PKI isoforms. In addition, both isoforms exhibit weaker interactions with PKI␤1 than with PKI␣.
Inhibition of Kinase Activity by RII␣-The intracellular association of R subunit with C subunit is based in part on the  Purified C subunits were assayed for phosphotransferase activity using increasing concentrations of the peptide substrate Kemptide (see "Experimental Procedures"). Double reciprocal plots of the data for the C subunit isoforms are shown in A. Note the sharp rise of the reciprocal data points for C␣ near the ordinate axis, indicative of substrate inhibition. The substrate inhibition of C␣ is shown in more detail in a primary plot of the data from Fig. 2A (B). Activity was determined as described under "Experimental Procedures" and is expressed as the percentage of the highest cAK specific activity obtained. The curves were fitted using the average values of triplicate assay points from a representative experiment, and the error bars depict the standard deviation from the mean. The experiments were performed three times and average K m and V max values and measurements of error are reported in Table I. The symbols for the proteins are ⅜ for C␣ and q for C␤1. high affinity interaction between the two proteins. Although it appears that there are no differences between C␣ and C␤ in their ability to recombine with RI isoforms (21), the interaction of the type II R subunit and the C␤ isoform has not been investigated. In order to analyze RII␣-C association, a decahistidine-tagged RII␣ protein (35) was used to inhibit kinase activity. Similar to the assays of PKI inhibition of the C subunit isoforms, C␣ had a 2.4-fold lower IC 50 for RII␣ than did C␤1 (Fig. 3B and Table II), providing further evidence that C␤1 participates in fewer and/or weaker interactions with inhibitors of C subunit.
Estimation of K a(cAMP) Values of RII␣ 2 C␣ 2 and RII␣ 2 C␤1 2 -Another important measure of the R-C interaction is the cAMP activation constant. Cadd et al. (21) demonstrated that C␣ and C␤1 had no influence on the K a(cAMP) values of type I holoenzymes but that RI␤ holoenzymes were more sensitive to cAMP dissociation than RI␣ holoenzymes. However, RII␣ has a sub-stantially divergent amino acid sequence and, unlike RI, is autophosphorylated and does not require ATP for association with C subunit (50). Type II holoenzymes were formed by combining purified, recombinant RII␣ with a 2-fold excess of either C␣ or C␤1 prior to each assay. Tetrameric RII␣ 2 C␣ 2 and RII␣ 2 C␤1 2 holoenzymes were separated from free C subunit using gel filtration chromatography (see "Experimental Procedures"). RII␣ 2 C␤1 2 displayed a significantly lower K a(cAMP) value (13 nM) than did RII␣ 2 C␣ 2 (63 nM) ( Fig. 4A and Table III), indicating that RII␣ 2 C␤1 2 is 5-fold more sensitive to dissociation by cAMP than is RII␣ 2 C␣ 2 (average of four experiments). In separate experiments, the cAMP analog 8-(4-chlorophenylthio)-cAMP (CPT-cAMP) also displayed a 5-fold lower K a value for RII␣ 2 C␤1 2 than RII␣ 2 C␣ 2 (data not shown). Similar results were obtained in assays of the cAMP dependence of RII␣ autophosphorylation by C␣or C␤1-containing type II holoenzymes (data not shown).
In addition to the difference in K a(cAMP) values, RII␣ 2 C␤1 2 exhibited an over 5-fold higher basal activity than RII␣ 2 C␣ 2 using Kemptide ( Fig. 4B and Table III) or RII␣ (data not shown) as the phosphoacceptor. Therefore, even without full occupation of the RII␣ cAMP binding sites, a relatively large proportion of RII␣ 2 C␤1 2 may be active. Combined, these in vitro data raise the possibility that cellular cAK effects could be altered by incorporating different C subunit isoforms into type II holoenzymes in vivo.

Effects of RII␣ and C␣ or C␤1 Expression on Transcriptional Activation in COS-1 Cells-To test the previous in vitro results
in an in vivo system, expression plasmids for RII␣ (pCMV-.RII␣), C␣ (pCMV.C␣), and/or C␤1 (pCMV.C␤1) were introduced into COS-1 cells along with a luciferase reporter construct driven by a cAMP-responsive human chorionic gonadotropin promoter (37) (see "Experimental Procedures"). Transient transfection of 6 g of pCMV.C␣ or pCMV.C␤1 produced an 8-fold increase in luciferase activity (data not shown). The addition of varying amounts of pCMV.RII␣ demonstrated that 2-3-fold less pCMV.RII␣ was required to half-maximally inhibit luciferase activity in cells expressing C␣ than in cells expressing C␤1 (Fig. 5A). This difference was not due to the weaker expression of C␣, because the control activities in the pCMV.C␣ and pCMV.C␤1 transfected COS-1 cells were identical (data not shown). Subsequent transfections used an amount of pCMV.RII␣ (8 g) sufficient to fully inhibit the activation of luciferase gene transcription by C␣ or C␤1. Thus, all expressed C subunit was incorporated into type II holoenzyme under these conditions.   3. Inhibition of C␣ and C␤1 by PKI(5-24) peptide and RII␣. C␣ and C␤1 were incubated in the presence of 100 M Kemptide and varying concentrations of PKI(5-24) (A) or RII␣ (B) (see "Experimental Procedures"). Activity is expressed as the percentage C␣ or C␤1 specific activity in the absence of inhibitor. The curves were fitted using the average values of triplicate assay points from representative experiments, and the error bars depict the standard deviation from the mean. The experiments were performed three times for each inhibitor. Average IC 50 values and measurements of error are reported in Table II. The symbols for the proteins are ⅜ for C␣ and q for C␤1. COS-1 cells expressing RII␣ 2 C␣ 2 or RII␣ 2 C␤1 2 were exposed for 24 h to varying concentrations of CPT-cAMP, a membranepermeable cAMP analog, or forskolin, a membrane-permeable stimulator of adenylate cyclase. COS-1 cells expressing RII␣ 2 C␤1 2 half-maximally stimulated luciferase production at 300 M CPT-cAMP (Fig. 5B). As predicted from the in vitro experiments, RII␣ 2 C␣ 2 required a significantly greater (12fold) extracellular concentration of CPT-cAMP (3600 M) to stimulate luciferase production to the same extent, although full activation of RII␣ 2 C␣ 2 was never reached due to CPT-cAMP solubility constraints. Similarly, transiently expressed RII␣ 2 C␤1 2 was activated at 5-fold lower concentrations of forskolin (0.9 M) than was RII␣ 2 C␣ 2 (4.6 M) (Fig. 5C). Lastly, RII␣ 2 C␤1 2 partially purified from COS-1 cell extracts by gel filtration chromatography exhibited a lower K a(cAMP) value (3-4-fold) than partially purified RII␣ 2 C␣ 2 (data not shown).
COS-1 cells expressing RII␣ 2 C␤1 2 also had a higher basal luciferase activity (approximately 2-fold) than those expressing RII␣ 2 C␣ 2 (Fig. 5, B and C). However, the in vivo difference in the basal kinase activities of RII␣ 2 C␣ 2 and RII␣ 2 C␤1 2 is small compared with the analogous difference observed in vitro ( Fig.  4B and Table III). This discrepancy may be due in part to either the presence of phosphatases and/or excess RII␣ in the COS-1

FIG. 4. Activation of type II cAK isoforms by cAMP in vitro.
RII␣ 2 C␣ 2 and RII␣ 2 C␤1 2 holoenzymes were formed and purified as described in "Experimental Procedures." The activities of RII␣ 2 C␣ 2 (⅜) and RII␣ 2 C␤1 2 (q) were measured in the presence of varying concentrations of cAMP. Double reciprocal plots of the data for RII␣ 2 C␣ 2 and RII␣ 2 C␤1 2 are shown in A. B, the difference in basal kinase activity between RII␣ 2 C␣ 2 and RII␣ 2 C␤1 2 is evident in a primary plot of the data from the experiment shown in A. Activity was determined as described under "Experimental Procedures" and is expressed as the percentage of the highest cAK activity obtained. The curves were fitted using the average values of triplicate assay points from a representative experiment, and the error bars depict the standard deviation from the mean. The experiments were performed four times and average K a(cAMP) values, average percent basal activity values, and measurements of error are reported in Table III. Note that, for simplicity, the legends refer only to the C subunit isoform present in the holoenzyme.  5. Activation of luciferase gene transcription by type II cAK isoforms in vivo. Six micrograms of an expression vector for C␣ (⅜) or C␤1 (q) and 0.5 g of a cAMP-responsive luciferase reporter construct were transiently transfected into COS-1 cells along with indicated amounts of an RII␣ expression vector (see "Experimental Procedures") (A). An amount of RII␣ expression vector (8 g) that ensured complete formation of RII␣ 2 C␣ 2 and RII␣ 2 C␤1 2 was used in all subsequent transfection experiments. COS-1 cells expressing RII␣ 2 C␣ 2 (⅜) or RII␣ 2 C␤1 2 (q) were treated for 24 h with varying concentrations of CPT-cAMP (B) or forskolin (C), harvested, and assayed for luciferase activity (see "Experimental Procedures"). In all of these experiments, luciferase activity is expressed as the percentage of maximal relative light units (RLU) attained. The curves were fitted using the average values of triplicate assay points from a representative experiment, and the error bars depict the standard deviation from the mean. Each experiment was performed three or four times. Note that for simplicity, the legends in B and C refer only to the C subunit isoform present in the expressed holoenzyme. cells. All of the in vivo results described here are representative of three or four experiments. Furthermore, identical results were obtained using a luciferase reporter construct driven by a cAMP-responsive human enkephalin promoter (25) (data not shown). Together, these data suggest that the in vitro differences between purified type II holoenzymes are relevant in vivo and therefore are likely to be physiologically significant. DISCUSSION cAMP-dependent protein kinase has been shown to play an important role in a wide range of cellular processes, including transcription (25,51), metabolism (52,53), cell cycle progression (54,55), apoptosis (56), and hippocampal long term potentiation (57,58). However, such functional diversity necessitates the existence of mechanisms that ensure cAK activity elicits specific and appropriate physiological responses in discrete cell types. One possible mechanism for increasing the specificity of the cAMP signaling pathway is through the compartmentalization of cAK via binding to A kinase anchoring protein. By localizing type II cAK near target substrates or within specialized environments, A kinase anchoring proteins may mediate the functional consequence(s) of cAK activation (18,19). Additionally, the expression of distinct R and C subunit isoforms may alter cellular responses to changes in cAMP levels. Different combinations of R and C subunit isoforms could affect cAK interactions with substrates, inhibitors, and cyclic nucleotide. Differential cAK isoform expression could also contribute to the varying responses of different tissues to stimuli that alter intracellular concentrations of cAMP.
The R subunit isoforms are known to impart distinct biochemical properties upon holoenzymes (20,21). Conversely, the available data on the two major C subunit isoforms, C␣ and C␤1, suggest that they possess similar substrate specificities (59) and have little or no effect on the cAMP sensitivity of type I holoenzymes (21). In addition, both isoforms are able to increase enkephalin and prolactin promoter activity with equal efficiency (25,60). However, difficulties producing large amounts of soluble C␤1 have prevented detailed comparisons of the enzymatic properties of C␣ and C␤1 prior to this report (43).
In the present study, recombinant C␣ did demonstrate lower K m values for the substrates Kemptide and IP 3 Rtide than did recombinant C␤1. This likely reflects subtle structural differences between C␣ and C␤1, because critical amino acids in C␣ that contact residues within the consensus substrate phosphorylation site are conserved in C␤1 (61,62). In addition, the isoform-specific resistance of C␤1 to substrate inhibition at high peptide concentrations suggests that the kinetic mechanisms of the C subunit isoforms may also differ. Results from earlier experiments with C␣ are consistent with a steady state random kinetic model in which ATP binds prior to substrate, followed by phosphotransfer and the ordered release of phosphopeptide before ADP (63,64). The inhibition of C␣ at relatively high Kemptide concentrations was proposed to result from the nonproductive process of binding Kemptide before ATP (44). Because C␤1 is not inhibited by millimolar concentrations of substrate, it is possible that differences in the kinetic mechanisms of C␣ and C␤1 exist. One plausible explanation is that the K d(substrate) value for C␤1 is much greater than for C␣, which prevents C␤1 from entering the nonproductive kinetic pathway. The resistance of C␤1 to substrate inhibition may cause it to be a more effective kinase than C␣ under certain subcellular conditions. C␣ also possessed lower IC 50 values for the cAK inhibitors PKI␣ and PKI␤1 than did C␤1. This difference may not be physiologically relevant, because both PKI isoforms are able to inhibit both C isoforms with great potency. However, the 2-3-fold lower concentration of RII␣ that was required in vitro to half-maximally inhibit C␣ relative to C␤1 was also observed in in vivo transfection experiments. These results suggest the possibility that endogenous inhibitors or substrates preferentially bind the C␣ isoform within cells.
We also investigated the cAMP sensitivity of type II holoenzymes formed in vitro with recombinant RII␣ and C␣ or C␤1. Unlike the type I holoenzyme (21), the cAMP sensitivity of the type II holoenzyme is affected by the presence of different C subunit isoforms. Furthermore, the 5-fold difference in the K a(cAMP) values of type II cAK formed with C␤1 or C␣ is similar to the 3-7-fold K a(cAMP) difference previously observed with type I cAK containing RI␤ as opposed to RI␣ (21,22). The distinct cAMP sensitivities of recombinant RII␣ 2 C␤1 2 and RII␣ 2 C␣ 2 were also detectable in assays of luciferase activity in transiently transfected COS-1 cells. It appears possible, then, that the transcriptional response of cells to changes in cAMP levels may be dependent in part upon the isoforms of R and C subunit expressed in the cell at that time.
The observation that cAK subunit isoforms are biochemically distinct has little significance unless these differences can potentially alter the function of cAK within the cell. Recent gene ablation studies in mice have provided strong evidence that specific R and C subunit isoforms are required for certain critical physiological events. Specifically, expression of RI␤ and C␤1 are essential for long term potentiation, long term depression, and depotentiation in discrete hippocampal pathways (5,23,24). Because RI␤ lowers the K a(cAMP) value of type I holoenzyme (21), it was speculated that cellular processes that normally occur in response to small fluctuations in cAMP levels may require cAK isoforms that have increased sensitivity to cAMP (23,24). In support of this hypothesis, the addition of forskolin and 3-isobutyl-1-methylxanthine to hippocampal slices from the RI␤ or C␤1 knockout mice restored mossy fiber long term potentiation (5). The concentration of forskolin used in these experiments (50 M) was sufficient to stimulate all the adenylate cyclase in the cell and establish cAMP levels capable of activating the remaining cAK isoforms. Thus, other endogenous cAK isoforms can compensate for the loss of RI␤ or C␤1, provided that intracellular cAMP concentrations are high enough to activate these less sensitive isoforms. These studies suggest that mechanisms exist within neuronal cells to activate particular cAK isoforms that in turn mediate distinct cellular processes.
In addition to its increased cAMP sensitivity, RII␣ 2 C␤1 2 exhibited higher basal activity in vitro and in vivo than did RII␣ 2 C␣ 2 . Basal cAK activity has been shown to regulate the expression of a number of genes in mouse Y1 adrenal cells (65) and the cystic fibrosis transmembrane conductance regulator gene in the human colon carcinoma T84 cell line (66). In the JEG-3 human placental cell line, transcription from distinct cAMP-regulated promoters was found to be influenced by basal cAK activity as well (67). Therefore, selective expression of RII␣ 2 C␤1 2 or RII␣ 2 C␣ 2 may also dictate basal transcriptional activity of some genes in response to unstimulated levels of cAMP.
Because C subunit isoforms have distinct influences only on the activation of type II holoenzyme, nonconserved amino acids in C␣ and C␤1 likely have important and specific interactions with RII␣. However, all C subunit residues identified to date that are involved in binding R subunit are conserved between C␣ and C␤1 (7,68). Even so, comparison of C␣ and C␤1 primary structures between species reveals an apparent evolutionary pressure to maintain the amino acids that differ between them (7,11). The majority of the nonconserved residues in the two isoforms are near the amino termini of the proteins and, based upon the x-ray crystal structure of C␣, map to a surface on C subunit distal to the catalytic site (7,69). Mutagenesis of the pseudophosphorylation sites of RI suggests that interactions outside the catalytic cleft also may account for most of the discrepancy in K a(cAMP) values between RI␣-and RI␤-containing holoenzymes (21). Therefore, it seems likely that the various R and C subunit isoforms exist in part because their unique primary structures produce physiologically relevant differences in their biochemical properties.
Although a model based on selective holoenzyme activation provides a potential explanation for the effects observed in the RI␤ and C␤1 gene ablation studies, additional aspects of this model need to be considered. For instance, it is currently not known whether endogenous C␣ and C␤1 form composite C␣C␤1R 2 holoenzymes or if cellular mechanisms exist to form C␣ 2 R 2 or C␤1 2 R 2 holoenzymes exclusively. Recently, the ability of RI isoforms to form heterodimeric complexes was demonstrated in a human neoblastic B cell line (70), although a detailed biochemical characterization of a mixed type I␣/I␤ holoenzyme has not been reported to date. If mixed C␣/C␤1 holoenzymes are present in vivo, it is unclear whether C␤1 would dissociate from C␣C␤1RII␣ 2 at the same cAMP concentrations as C␤1 2 RII␣ 2 or if higher cAMP levels would be required. However, the results presented here should provide a foundation for further studies of the physiological roles of the various holoenzyme complexes in vitro and in vivo.