Dimerization/Docking Domain of the Type Ia Regulatory Subunit of cAMP-dependent Protein Kinase

Based on increasing evidence that the type I R subunits as well as the type II R subunits localize to specific subcellular sites, we have carried out an extensive characterization of the stable dimerization domain at the N terminus of RIα. Deletion mutants as well as alanine scanning mutagenesis were used to delineate critical regions as well as particular amino acids that are required for homodimerization. A set of nested deletion mutants defined a minimum core required for dimerization. Two single site mutations on the C37H template, RIα(F47A) and RIα(F52A), were sufficient to abolish dimerization. In addition to serving as a dimerization motif, this domain also serves as a docking surface for binding to dual specificity anchoring proteins (D-AKAPs) (Huang, L. J., Durick, K., Weiner, J. A., Chun, J., and Taylor, S. S. (1997) J. Biol. Chem. 272, 8057–8064; Huang, L. J., Durick, K., Weiner, J. A., Chun, J., and Taylor, S. S. (1997) Proc. Natl. Acad. Sci. U. S. A.94, 11184–11189). A similar strategy was used to map the sequence requirements for anchoring of RIα to D-AKAP1. Although dimerization appears to be essential for anchoring toD-AKAP1, anchoring can also be abolished by the following single site mutations: C37H, V20A, and I25A. These sites define “hot spots” for the anchoring surface since each of these dimeric proteins are deficient in binding to D-AKAP1. In contrast to earlier predictions, the alignment of the dimerization/docking domains of RIα and RII show striking similarities yet subtle differences not only in their secondary structure (Newlon, M. G., Roy, M., Hausken, Z. E., Scott, J. D., and Jennings. P. A. (1997)J. Biol. Chem. 272, 23637–23644) but also in the distribution of residues important for both docking and dimerization functions.

cAMP-dependent protein kinase (PKA), 1 one of the first pro-tein kinases to be purified and characterized (4), is maintained as an inactive tetrameric holoenzyme complex of regulatory (R) and catalytic (C) subunits, in the absence of cAMP (5). Upon elevation of intracellular cAMP and its subsequent binding to the R subunit, the holoenzyme complex dissociates into an R subunit dimer and two free and active C subunits (5). Although PKA is one of the best understood protein kinases in terms of the structure and function of the R and C subunits (5)(6)(7)(8), a further level of control is introduced by interaction of the R subunits with A-Kinase Anchoring Proteins (AKAPs) that target the enzyme to specific subcellular sites (9 -12).
Until recently the R subunits have been thought of simply as physiological inhibitors of PKA (6,(13)(14)(15). However, like many other proteins involved in signal transduction, these highly modular proteins are also multifunctional (16). The structural and functional independence of the domains within the R subunits were first recognized by limited proteolysis (17). Although the two tandem cAMP-binding sites at the C terminus as well as the inhibitor site are required for high affinity binding to C (18 -21), it is the dimerization domain at the N terminus that is important for interacting with AKAPs (22)(23)(24).
Numerous AKAPs have now been identified, each having distinct subcellular locations (16,(25)(26)(27)(28). Some AKAPs also appear to act as scaffolds for assembling multiprotein complexes. For example, AKAP79 localizes to the actin cytoskeleton and interacts with the calcium and calmodulin-dependent protein phosphatase 2B (calcineurin) and protein kinase C as well as PKA (29,30). Anchoring is thought to be mediated by a small helical segment in the AKAPs that binds specifically to the N-terminal dimerization domain of RII (22)(23)(24)(25).
Although most AKAPs were thought initially to interact exclusively with RII subunits, PKA can also be tethered via RI␣. In T lymphocytes, for example, both RI␣ and RII␣ are expressed. In resting cells, type I holoenzyme is evenly distributed in the cytoplasm, whereas the type II holoenzyme is particulate. Once the T cell is stimulated by antigen-specific activation of the T cell receptor, however, RI␣ is tightly associated with the capped T cell receptor-CD3 complex (31). RI␣ also colocalizes with the antigen receptor in stimulated B cells (32). In human erythrocytes RI␣ is tightly bound to the plasma membrane (33). A recent report also showed that RI␣ localizes at the neuromuscular junction (34). Moreover, RI␣ was shown to interact with the activated EGF receptor through Grb2 (35). The Grb2-binding site is speculated to lie C-terminally to the dimerization domain in a proline-rich sequence typical for recognition by SH2-binding proteins. Therefore, RI␣ potentially contains specific subdomains geared for protein-protein interactions that may fine-tune type I holoenzyme action in vivo.
Two new families of anchoring proteins, referred to as dual specificity AKAPs (D-AKAPs), were identified recently from a yeast two-hybrid screen using an RI␣ fusion protein as bait (1,2). These proteins not only bind RII but also RI␣ subunits. The R-binding region of D-AKAP1 localized to a 125-residue fragment, RPP7. A human homolog of D-AKAP1, AKAP84, was identified in sperm (36).
So far characterization of the various AKAPs and PKA has focused on RII subunits. With the discovery of D-AKAP1, we set out to determine sequence requirements not only for dimerization of RI␣ but also for D-AKAP1 binding. A further goal was to determine whether RI and RII mediate homodimerization and anchoring by similar mechanisms. Deletion mutants, chimeric proteins, and site-specific mutants were used to delineate critical regions as well as particular amino acids involved in homodimerization and also to identify sequence requirements for RI␣ and D-AKAP1 interaction. We have shown that distinct but overlapping determinants are required for dimerization and docking. Furthermore, and in contrast to earlier predictions, we find strong similarities as well as subtle differences in the dimerization domains of RI and RII.

Materials-
The following materials were purchased as indicated: ATP, cAMP, protease inhibitors, Triton X-100, nickel-NTA resin (Qiagen), cAMP-agarose and GST-agarose (Sigma), enzymes used for DNA manipulations (Life Technologies, Inc.), and the DNA sequencing kit (U. S. Biochemical Corp.) All oligonucleotides were synthesized at the Peptide and Oligonucleotide Facility at the University of California, San Diego.
Three classes of specific site mutations were also made. The cysteines in the dimerization domain, Cys 16 and Cys 37 , were mutated singly and in pairs in order to extend our understanding for the role of the cysteines in dimerization and anchoring of D-AKAP1. Leon et al. (39) had previously engineered RI␣(C37H) and shown that the disulfide bonds are not required for dimerization. Two additional single site mutants were made, RI␣(C16A) and RI␣(C37A). A double mutant was also constructed, RI␣(C16A,C37A).
An additional set of mutants was used to assess requirements for dimerization. These mutations were double mutants where Cys 37 had been mutated to either His or Ala. Mutating one of the cysteines was sufficient to remove the two interchain disulfide bonds and allowed us to examine the role of the additional residues for dimerization. Double mutations with C37H were engineered by changing the following residues: V20A, I25A, L29A, L28A, Ile 33 /Val 34 , F52A and F47A.
A final set of mutations, V20A and I25A, was engineered that had only single site mutations. These were used specifically to assess binding to D-AKAP1. Arg 94 and Arg 95 also were both mutated to alanines to make the double mutant, RI␣(R94A,R95A), as described previously (40,41). A summary of all the mutants used for analysis is presented in Fig. 1A.
Construction of Chimeric Proteins-Two chimeric proteins of RI␣ and RII␣ were constructed and expressed as poly(His) fusion proteins. First, an NheI site was engineered just before the initiation Met on pRSETb-RI␣ and pRSETb-RII␣, respectively. An MunI site was then engineered at residue 62 of RI␣ and residue 46 of RII␣ on these two vectors containing the new NheI sites. One protein, designated (His 6 )RI␣/RII␣, was constructed by excising the cDNA encoding the first 62 residues of RI␣ with NheI and MunI and then ligating this oligonucleotide fragment to cDNA encoding residues 47-379 of RII␣. This chimeric protein contained the N-terminal dimerization of RI␣ fused to the C terminus of RII␣. The other protein, designated (His 6 )RII␣/RI␣, was constructed by excising the cDNA encoding the first 45 residues of RII␣ with NheI and MunI and ligating this fragment to the cDNA encoding residues 63-379 of RI␣. This chimeric protein contained the N-terminal dimerization domain of RII␣ fused to the C terminus of RI␣ (Fig. 1B).
Expression and Purification of R Subunits-Two methods were used to purify the wild type and mutant R subunits. In method 1, the deletion mutants were expressed in BL21(DE3) cells. The BL21(DE3) cell strain was a gift from Bill Studier (Brookhaven National Laboratories). The R subunits were then purified on a DE52 ion exchange column as described previously (42). Further purification was achieved using MonoQ (Amersham Pharmacia Biotech) ion exchange chromatography. In method 2, the R subunits were expressed in 222 cells (42) and then affinity purified on cAMP-agarose resin (43). Further purification of these mutants was achieved using gel filtration (Superdex 200). Cells were lysed in 20 mM phosphate buffer containing 5 mM ␤-mercaptoethanol and protease inhibitors at pH 6.5 (42). After centrifugation, the supernatant was incubated with cAMP resin for 6 h at 4°C. After extensive washing, the R subunits were eluted in lysis buffer using 50 mM cAMP.
Analytical Gel Filtration-Analytical gel filtration was carried out deletion mutant (⌬1-91)RI␣. To determine the apparent molecular weight and Stokes radii of the various mutants, the column was calibrated using the Amersham Pharmacia Biotech Calibration Kit.
Electrophoresis of Proteins-Samples of wild type and mutant RI␣ subunits were denatured in gel loading buffer and subjected to electrophoresis in polyacrylamide gels as described previously (39). Non-denaturing gel electrophoresis was performed in a 7.5% polyacrylamide gel at 50 V for 4 -6 h (44). All electrophoresis was performed using Mini-Protean II electrophoresis system (Bio-Rad). SDS-PAGE reagents were prepared according to Laemmli (45). Proteins were visualized by staining with Coomassie Blue.
D-AKAP1 Binding Assay-A deletion mutant of D-AKAP1, RPP7 (1) which contains the R-binding region, expressed as either a poly(His)tagged fusion protein or a GST fusion protein, was used in these binding assays. This fragment corresponds to residues 284 -408 in the D-AKAP1 core. Bacterial cell lysates containing His 6 -RPP7 or GST-RPP7 were incubated with either nickel-NTA resin or glutathione resin, respectively, for 2 h at 4°C in phosphate-buffered saline (10 mM potassium phosphate, 150 mM NaCl, pH 7.4) with 0.1% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, 5 mM benzamidine, and 5 mM ␤-mercaptoethanol and then washed extensively with the same buffer. Fulllength and RI mutants (100 -200 mg) were then added to the resin and incubated for 2 h at 4°C. After washing the resin extensively with phosphate-buffered saline, proteins associated with the resin were released by boiling in SDS gel-loading buffer and analyzed by SDS-PAGE.

Minimum Core Requirements for Dimerization of RI␣-
The following nested N-terminal deletion mutants were made, expressed, and purified to homogeneity as described under "Experimental Procedures": (⌬1-17)RI␣, (⌬1-22)RI␣, (⌬1-37)RI␣, and (⌬1-45)RI␣. Once the purity of each protein was assessed by SDS-PAGE ( Fig. 2A), the oligomerization state was analyzed by both analytical gel filtration and electrophoresis under non-denaturing conditions (Fig. 2, B and C). (⌬1-17)RI␣ eluted as a single peak with a retention volume that was identical to the wild type dimer, RI␣. (⌬1-37)RI␣ and (⌬1-45)RI␣ both eluted as a single peak but were monomeric based on their retention volume (Fig. 2B). The apparent molecular weight and Stokes radii of all the mutants were compared with the wild type RI␣ and presented in Fig. 2D.
In contrast to the above mutants, deletion of five additional residues beyond 17 yielded a protein, (⌬1-22)RI␣, that displayed an intermediate behavior (Fig. 3A). It eluted as two distinct peaks, one migrating at the characteristic dimer position, and the other migrating with a retention volume corresponding to a monomer. Based on SDS-PAGE analysis, the protein was Ͼ95% pure, indicating that the two peaks were not due to heterogeneity (Fig. 3B). This behavior is characteristic of two species in equilibrium if they are in slow exchange. To test this possibility, the concentration dependence of the elution profile was tested. After loading (⌬1-22)RI␣ at three different concentrations, the ratio of the dimeric peak to the monomeric peak was monitored. As shown in Fig. 3A, the slower migrating peak, corresponding to the monomer, was more prominent as the loading concentration of the protein decreased, whereas the peak corresponding to the dimer decreased. Moreover, when the first peak was collected, concentrated, and reloaded onto the analytical Superdex 200, two species were generated thus confirming that both species exist in a slowly exchanging equilibrium (Fig. 3B). The presence of the dimeric peak was also salt-dependent. When the column was equilibrated with high salt (1 M NaCl), the ratio of the dimer to monomer changed with the dimer peak being enhanced (data not shown). These combined results indicate that there are critical amino acids between residues 17 and 22 that mediate dimerization. Their removal significantly reduced the equilibrium between the two protomers. Given that the interaction is enhanced at higher ionic strength, the nature of the interaction of the monomers is likely to be predominantly hydrophobic in nature. We thus believe that the main "core" for mediating dimerization of RI␣ lies within the region flanked by residues 17 and 37, and we refer to this disulfide-bonded core subsequently as subdomain 1 (Fig. 4).
Alanine-scanning Mutagenesis of RI␣-The contribution of particular amino acids to dimerization of RI␣ was evaluated by alanine-scanning mutagenesis. Since the two cysteines in the dimerization domain, Cys 16 and Cys 37 , form interchain and anti-parallel disulfide bonds, it was essential to analyze these mutants in the absence of any covalent linkage. Earlier, Leon et al. (39), had shown that the removal of one Cys, namely C37H, abolished both disulfide bonds thus supporting the original prediction for anti-parallel arrangement of the two protomers (39,46). However, the dimer remained intact, demonstrating that the disulfide bonds were not essential for dimerization. All additional mutants for this study of the dimerization determinants were made in addition to the C37H mutation. The mutants were classified into two categories that reflect two "subdomains," the disulfide-bonded core mutants (subdomain 1) and the "C-terminal helix" mutants (subdomain 2) (Fig. 4). Based on the deletion mutants "subdomain 1" is essential for dimerization. Since salt-dependence of dimerization was indicative of hydrophobic interactions, we speculated that hydrophobic amino acids are key players at the dimer interface, and the mutagenesis was set up to test this hypothesis. The C-terminal helix or "subdomain 2" is based on earlier secondary structure predictions done by CDESTIMA as reported earlier (39) as well as sequence alignments and inference from NMR secondary structure analysis that had been done on the dimerization domain of RII (3). The boundaries of the subdomains and the various mutants are summarized in Fig. 4.
Each mutant protein was analyzed for its oligomerization state by analytical gel filtration and non-denaturing gel electrophoresis. Surprisingly, none of the mutations in the core affected the oligomerization state of RI␣. As shown in Fig. 5A all the core mutants as well as the single site cysteine mutants migrated like the wild type RI␣ subunit dimer under native conditions. Moreover, they eluted with the same retention volume on an analytical gel filtration column (Fig. 2D). The only mutations that affected dimerization were RI␣(C37H/F47A) The DD domain of RI␣ is comprised of two subdomains. Subdomain 1 is flanked by the cysteines, and subdomain 2 begins at residue 45 and is predicted to be helically based on secondary structure algorithms. The residues within subdomain 1 that were mutated are marked with asterisks, and those in subdomain 2 are marked with an arrow. F/A marks the two single-site alanine mutant that were monomeric. and RI␣(C37H/F52A). As shown in Fig. 5B their mobility on a non-denaturing gel was clearly shifted as compared with the dimeric RI␣ subunit. Assuming that this mutation has not affected the charge to mass ratio, the mobility is indicative of a change in apparent molecular weight of this protein (44). An-alytical gel filtration analysis of these mutants further confirmed that they are monomeric (Figs. 5C and 2D). Additionally, once these mutants were analyzed on SDS-PAGE, it was confirmed that the integrity of the molecule was retained, and the change seen in apparent molecular weight was not simply due to degradation (Fig. 5D).
Localization of the RI␣-binding Surface-To localize further the D-AKAP1-binding site on RI␣, the N-terminal deletion mutants discussed above were tested for their ability to bind the R-binding region of D-AKAP1(RPP7). His 6 -RPP7 was able to pull down (⌬1-17)RI␣ and (⌬1-22)RI␣. However, (⌬1-45)RI␣ and (⌬1-91)RI␣ were not pulled down in the binding assay. Residues 22-45 are thus necessary for interaction with D-AKAP1. Since deletion of the first 45 residues also abolished dimerization, it is likely that a dimer is necessary for interaction with D-AKAP1 (Fig. 6, A and B). To characterize further the anchoring properties of RI␣, a series of site-directed mutants were also tested to identify specific amino acids that are essential for interaction with D-AKAP1. The Cys residues that form disulfide bonds and some of the hydrophobic amino acids were targeted in particular. The single site mutants were then characterized for their ability to bind His 6 -RPP7.
The anti-parallel disulfide bonding between Cys 16 and Cys 37 is one of the features that distinguishes RI␣ from RII␣. Although these disulfide bonds are not essential for dimerization, they are extremely stable even under highly reducing conditions (39). To test if they contributed to D-AKAP1 interaction, Cys 16 was mutated to His, and Cys 37 was mutated to both His and Ala. A double mutant with both Cys 16 and Cys 37 changed to Ala was also engineered. Although RI␣(C16H) showed no reduction in binding to D-AKAP1, D-AKAP1 binding was nearly abolished when Cys 37 was replaced with His. Therefore, even though Cys 16 forms an intermolecular disulfide bond with Cys 37 and neither mutation interfered with dimerization, the mutations had different effects upon D-AKAP1 binding. As shown in Fig. 6C, RI␣(C37A) and the double mutant RI␣(C16A,C37A) also were unable to bind tightly to D-AKAP1.
Since long chain aliphatic residues were shown to be essential for RII-AKAP interaction (44), Val 20 and Ile 25 were also mutated to Ala individually and tested for their ability to interact with D-AKAP1. As shown in Fig. 6C, both mutations resulted in the total loss of ability to associate with D-AKAP1.
D-AKAP1 Interaction with Chimeric Proteins of RI␣ and RII␣-Although D-AKAP1 tethers both RI␣ and RII␣ through their N termini, these domains are among the least conserved regions in the two R subunits. The dimerization domain boundaries are defined within residues 12-61 of RI␣ and the first 45 residues of RII␣ (39,47,48). To characterize further the requirements for conveying D-AKAP1 binding at the N terminus of the two classes of R subunits and to determine whether the dimerization domain was sufficient for anchoring, we tested the ability of two chimeric proteins, (His 6 )RI␣/RII␣ and (His 6 )RII␣/RI␣, to bind GST-RPP7. These two proteins were engineered so that only the N-terminal dimerization domains were swapped. Specifically, the RI␣/RII␣ chimera contained the dimerization domain of RI␣ and the C terminus of RII␣, whereas the RII␣/RI␣ chimera contained the dimerization domain of RII␣ and the C terminus of RI␣ (Fig. 1B). As shown in Fig. 6D, GST-RPP7 bound to both chimeras but had lower affinity for His 6 RI␣/RII␣. Based on these preliminary qualitative findings, the first 61 amino acids of RI␣ are probably not sufficient to generate an optimal binding site for D-AKAP1.
Effect of Mutations at the Inhibitor Site of RI␣ for Anchoring D-AKAP1-The inhibitor site in RI␣ that is responsible for recognition by the C subunit is localized C-terminally to the dimerization domain. This site, Arg 94 -Arg 95 -Gly-Ala-Ile 98 , is thought to bind in the active site of the C subunit in the holoenzyme, therefore rendering the enzyme inactive. A double mutant with the two arginines changed to Ala was also tested for its ability to bind D-AKAP1. RI␣(R94A,R95A) showed decreased affinity for D-AKAP1 relative to wild type RI␣. These data are schematically summarized in Fig. 6E.
Isoform Specificity for D-AKAP1-To investigate further the specificity of D-AKAP1 for the various R subunit isoforms, RI␣, His 6 RI␤, His 6 -RII␣, and RII␤ were tested in parallel for their ability to bind to the R-binding fragment of D-AKAP1, RPP7. As shown in Fig. 7, RI␣, His 6 -RII␣, and RII␤ were all pulled down by the GST-fused R-binding fragment of D-AKAP1 and GST-RPP7, but His 6 -RI␤ was not.

DISCUSSION
The dimerization domain at the N terminus of RI␣ is a very stable domain that is also essential for RI␣ binding to a novel family of dual specificity AKAPs (1,39). To emphasize the multiple functions of this domain, it will subsequently be referred to as the dimerization/docking (DD) domain. A series of mutants were engineered to probe the structural features of this domain and to identify regions as well as residues that contribute to dimerization and/or docking. By using deletion and site-specific mutants, the boundaries required for dimerization were defined. Deletion of up to 17 residues had no effect on dimerization, whereas the deletion of five more residues (⌬1-22) generated a concentration-dependent dimer. Since dimerization of (⌬1-22)RI␣ was enhanced by high salt, the hydrophobic residues in this region, Leu 18 , Tyr 19 , and Val 20 , are potential contributors to the hydrophobic dimer interface. Single site replacement of Val 20 to Ala, however, did not abolish dimer formation. Deletion of the first 37 residues abolished dimerization. Therefore, the region flanked by the two cysteines, the "disulfide-bonded core" or subdomain 1, is essential for dimerization of RI␣.
Single site mutations of Cys 16 or Cys 37 to Ala or His in the disulfide-bonded core were not sufficient to abolish dimerization; however, two other mutations did abolish dimerization, RI␣(C37H,F47A) and RI␣(C37H,F52A). These aromatic residues, both localized in subdomain 2, are clearly contributing to the dimer interface either directly or by folding back on to subdomain 1. Thus, although residues 17-37 are essential for dimerization, they are not sufficient.
The anchoring of RI␣ to D-AKAP1 also depended on the dimerization state of RI␣. Deletion of the region between residues 22 and 45 abolished both dimerization and anchoring, whereas (⌬1-22)RI␣ was able to bind D-AKAP1. Thus, the surface created by the dimer is probably as critical for anchoring to D-AKAP1 as it is for anchoring to other AKAPs (48,49). Single site mutations, C37H, V20A, and I25A identified at least three critical residues for D-AKAP1 interaction. Although these mutations did not abolish dimerization, they did reduce the affinity of RI␣ for D-AKAP1, suggesting that the requirements for dimerization and docking are distinct yet overlapping. The effect of the C37H mutation was particularly striking since the C16H replacement had no effect on binding to D-AKAP1 and suggested that the specific surface flanking Cys 37 maybe a "hot spot" for RI␣ and D-AKAP1 interaction. Furthermore, even though the two cysteines are involved in an interchain disulfide bond, their role in anchoring appears to be asymmetric: Cys 37 is essential and Cys 16 is not.
The construction of RI␣/RII␣ chimeras shed further light on the differences in anchoring requirements between RI␣ and RII␣. The apparent affinity of D-AKAP1 for RII␣ (50 nM) is higher than that for RI␣(1 M) based on surface plasmon resonance (50). Since the RII␣/RI␣ chimera that contained only the DD domain of RII␣ also bound tightly to D-AKAP1, this domain is sufficient for conveying high affinity binding to D-AKAP1. In contrast, the parallel chimera that contained only the DD domain of RI␣ did not bind tightly to D-AKAP1, indicating that additional regions peripheral to the DD domain are required for RI␣ to bind optimally to D-AKAP1. Further evidence to support this hypothesis came from another RI␣ mutant where the inhibitor site that docks to the active site cleft of the C subunit in the holoenzyme complex was mutated. Specifically, Arg 94 and Arg 95 that dock to the P-3 and P-2 recognition sites in the catalytic subunit were replaced with Ala. This mutant RI␣(R94A,R95A) also did not bind efficiently to D-AKAP1, suggesting that these residues either interact directly with D-AKAP1 or contribute to the domain organization of RI␣ so that the DD domain is poised for interaction with D-AKAP1. Quantitation of binding is necessary to determine whether full-length RII binds more tightly to D-AKAP1 than the DD domain alone. Based on our results, however, the dimerization domain of RII␣ alone, but not RI␣, is necessary and sufficient for conveying high affinity binding to AKAPs.
Since previous AKAPs have been shown to interact specifically with RII␣, it was of particular interest to compare the anchoring properties of D-AKAP1 to the different isoforms of the R subunits. D-AKAP1 binds RI␣, RII␣, and RII␤, but it does not bind RI␤ despite the sequence similarities between RI␣ and RI␤. Sequence comparison of the two indicates that within the region flanked by the two cysteines, subdomain 1, there are two nonconservative mutations (K22L and A24G). Given that these mutations lie within a region predicted to be critical for D-AKAP1 interaction, it might be sufficient that the absence of a charged group or a polar residue is enough to perturb the binding surface required for D-AKAP1 binding. In the subdomain 2, there are four non-conservative mutations (T34I, A39S, A46K, and A60N). These mutations would also perturb the binding surface indirectly by altering the dimer interface. Moreover, the actual boundaries of dimerization and the nature of disulfide linkages have not been mapped out in RI␤; there are two additional cysteines that, depending on their oxidation states, could affect the structural orientation of the dimeric interface and subsequently the binding surface for D-AKAP1.
The mutations described here also allow for a rigorous comparison of RI␣ and RII␣ subunits in terms of both functions of the N-terminal domain, namely dimerization and docking. The alignment shown in Fig. 8 reveals that the similarities within this domain are more extensive than previously thought. Functional similarities are further supported by the mutations when our results are compared with previous biochemical and functional studies on RII␣ and RII␤ (43,44).
Based on site-directed mutagenesis and deletions, the dimerization domain of RII, encompassing the first 45 residues, is necessary for its interaction with previously identified AKAPs. Dimerization, in fact, is a prerequisite for AKAP binding (22)(23)(24). In analogy, Huang et al. (1) also showed that the DD domain of RI␣ is sufficient for D-AKAP1 interaction. Since all regulatory subunits of PKA are dimers, except that of Dictyostelium (51), it is likely that the dimerization of this protein has evolved in an effort to add another level of regulation to PKA action through its interaction with AKAPs. Dimerization thus creates an oligomeric template for protein-protein interaction.
Leu 13 and Phe 36 were shown to be required for dimerization of RII; however, substitution of Ala for Ile 3 , Ile 5 , Leu 12 , Val 20 , Leu 21 , Phe 31 , Leu 33 , or Leu 39 did not impair dimerization (43,44). According to the alignment shown in Fig. 8, there are striking similarities and differences in the mechanism of homodimerization between RI and RII. Even though in RII␤, a mutation within subdomain 1, RII(L13A), was sufficient to abolish dimerization, the analogous mutation, RI␣(C37H/L29A) had no effect on the RI␣ oligomerization state. No other mutations in subdomain 1 were sufficient to abolish dimerization of RI␣.
Moreover, even though RI␣(C37H,F47A) abolished dimerization, a similar mutation in RII␤, RII␤(F31A) did not abolish dimerization. However, when the RII␤(F36A) equivalent mutation was made in RI␣, RI␣(C37H,F52A), as in RII␤, the dimerization of RI␣ was compromised. Therefore, there are not only differences in subdomain 1 but also in subdomain 2 in terms of requirements for dimerization of RI and RII. Newlon et al. (3) have characterized subdomains 1 and 2 within RII using multidimensional NMR and have shown that these subdomains are helical. More recently the secondary structure of RI(DD) has been probed by similar techniques. 2 An understanding of these mutations in the context of their role in the secondary structures of these domains and ultimately tertiary and quaternary structure will help delineate the exact mechanisms of homodimerization as well as anchoring of RI and RII.