Interaction of the Regulatory Subunit (RII) of CAMP-dependent Protein Kinase with RII-anchoring Proteins Occurs through an Amphipathic Helix Binding MotiP

The type I1 CAMP-dependent protein kinase is local- ized to specific subcellular environments through the binding of the regulatory subunit (RII) dimer to RII- anchoring proteins. Computer-aided analysis of secondary structure, performed on four RII-anchoring protein sequences (the microtubule-associated protein 2, P150, and two thyroid proteins Ht 21 and Ht 31), has identified common regions of approximately 14 residues which display high probabilities of forming amphipathic helices. The potential amphipathic helix region of Ht 31 (Leu-Ile-Glu-Glu-Ala-Ala-Ser-Arg-Ile-Val-Asp-Ala-Val-Ile) lies between residues 494 and 507. A bacterially expressed 318-amino acid fragment, Ht 31 (418-736), containing the amphipathic helix region, was able to bind RIIa. Site-directed mutagenesis designed to disrupt the secondary structure in the putative binding helix reduced binding dramat-ically. Specifically, substitution of proline for Ala-498 significantly diminished RIIa binding, and similar mutation of Ile-502 or Ile-507 abolished interaction. Mu- tation of Ala-522 to proline, which is located outside the predicted amphipathic helix region, had no effect on RIIa binding.


494-6972.
subunits from the regulatory (R) subunit dimer (4, 5). Although many hormones utilize the CAMP-signaling pathway to activate PKA, individual hormones produce diverse biochemical effects (4). A potential mechanism to explain this phenomenon is that individual hormones promote activation of PKA pools, which are localized to specific subcellular sites (6).
Subcellular localization of PKA has been shown to be directed through the R subunit (7). Two R subunit classes exist, RI and RII, which form the type I and type I1 holoenzymes, respectively (8, 9). Type I1 PKA is present in all cells, whereas the tissue distribution of type I PKA is more restricted (10,11). The subcellular distributions of both PKA subtypes are also distinct. The RI isoforms (RIa and RIP) are primarily cytoplasmic, while certain tissues contain up to 75% of either RII isoform (RIIa or RIIP) in particulate form, associated with either the plasma membrane, cytoskeletal components, secretory granules, or the nuclear membranes (9,(12)(13)(14)(15)(16)(17)(18). Type I1 PKA localization is dictated by the association of RII with specific anchoring proteins (12)(13)(14)(15). Tissue-specific patterns of RII-anchoring proteins have been detected by protein-blotting techniques or by fractionation on RII-Sepharose affinity columns (19,20). Presumably, tissue-specific PKA function is determined by RII-anchoring proteins which tether the holoenzyme to particular subcellular sites. This localization of PKA determines which substrates are most accessible to the C subunit upon activation by cAMP (6).
Cytoskeletal attachment of PKA occurs through interactions between RII and microtubule-associated protein 2 (MAP 2) (21). Recent reports have shown that the interaction between RII and MAP 2 occurs through the first 79 amino acids of RII (22,23). Furthermore, binding to MAP 2 requires sites on both protomers of the RII dimer (22). In complementary studies, the site on MAP 2 that binds RIIP has been localized to a 31-residue peptide (24,25). The RII-binding domain in P150, a murine brain RII-anchoring protein, has been identified but not defined in detail and is believed to involve the last 15 amino acids of the protein (19). Although there is no obvious sequence similarity between MAP 2 and P150, it is quite possible that both molecules contain common secondary structural motifs which permit interaction with RII. Amphipathic helices (an 0-helix with opposing polar and nonpolar faces) are known to be involved in both intra-and intermolecular protein-protein interactions (26). We have used helical wheel analysis to identify potential amphipathic helices in MAP 2, P150, and two RII-anchoring proteins (Ht 21 and Ht 31) we have recently characterized. In this paper, we demonstrate that Ht 31 and truncated proteins of Ht 31, containing

RII Binds to Amphipathic
Helices on RII-anchoring Proteins the amphipathic helix region, bind to RIIa. Furthermore, disruption of the putative helix regions in Ht 31, by mutation of selected residues to proline, diminishes or abolishes RIIa binding activity. Therefore, we propose that RII-anchoring proteins contain amphipathic helices that are necessary for binding to the regulatory subunit of PKA and thereby dictate this kinase's subcellular location.

EXPERIMENTAL PROCEDURES
Purification of Proteins-Recombinant RIIa was purified from Escherichia coli as previously described (22). Fragments of the human thyroid RIIa-anchoring protein Ht 31 and related point mutants were expressed in the E. coli expression vector pETkfc2 which is an adaptation of the T7 polymerase system reported by Studier et al. (27). A DNA cassette encoding a 51-residue peptide containing a concensus PKA phosphorylation site, Factor X cleavage site, and a calmodulin-binding peptide was engineered into the BamHI site of the bacterial expression vector pETlld* (Novagen). Expression of proteins using the pETkfc vector system was as described by Studier and colleagues (27) and produced fusion proteins with a carboxylterminal calmodulin-binding affinity tail. Each fusion protein was purified by affinity chromatography on calmodulin-Sepharose by the method of Sharma and colleagues (28). Purified proteins were concentrated by ultrafiltration through Amicon Centricon microconcentrators.
Construction and Expression of Ht 31 Fragments-A 941-base pair NcoI-BamHI fragment encoding Ht 31 was subcloned into the expression vector, pETkfc. The recombinant plasmid, pETHt3lkfc, was propagated in an E. coli host cell line, DH5a. For protein expression, the recombinant plasmid was transformed into the lysogenic strain of E. coli, BLZl(DEB)pLysS, and 1-liter cultures were grown as described by Studier and colleagues (27).
Site-directed Mutagenesis-All mutagenesis was performed using the "altered sitesTM" in vitro mutagenesis system (Promega). The full-length Ht 31 cDNA was subcloned into the BamHI site of the phagemid vector pSELECTTM. Single-stranded pSELECT templates were obtained after infection of JM109 liquid cultures in midlog phase with R408 helper phage. After overnight growth at 37 "C, the bacteriophage were harvested and DNA was isolated. Double primer mutagenesis was performed with the mutagenic oligonucleotides (21mers) as described in the altered sitesTM manual. All mutations were confirmed by nucleotide sequencing of the entire coding region. Mutated DNA was cut from the appropriate pSELECT vector by digestion with NcoI and BamHI and ligated to a similarly cut pETkfc bacterial expression vector. Each mutant protein was expressed and purified using the pETkfc vector system as described above.
Design of Mutants-Mutants were designed to disrupt the putative a-helix conformation in Ht 31 by substituting proline residues for hydrophobic amino acids in this region. Single point mutations in Ht 31 were introduced at Ala-498 (GCT), Ile-502 (&A), Ile-507 (AJ'C), and Ala-522 (GCC) using mutagenic oligonucleotides of 21 bases. Double mutations at Ala-498 and Ile-507 were introduced by simultaneously using two mutagenic oligonucleotides in the mutagenesis reactions. A triple mutation of Ht 31, introducing prolines at positions 498,502, and 507, was performed using the double mutant Ht 31 Pro-498/502 as the template and changing Ile-507 to proline. RZZ Ouerluy Procedure-The RII overlay binding assay was done according to the method of Lohmann and colleagues (20) incorporating those modifications documented by Bergmann et al. (19).
Gel-shift Analysis-The ability of Ht 31 mutants to complex with RIIa was assessed by nondenaturing polyacrylamide gel electrophoresis. The migration rate of the RIIa-Ht 31 complex is slower than either free RIIa or Ht 31 and can be easily distinguished on the gel.
Polyacrylamide gels were prepared as described by Laemmli (29) with the omission of SDS in all reagents. Binding reactions were conducted in buffer containing 5 mM HEPES (pH 7.2), 10 mM dithiothreitol, 1 mM benzamidine HCl, 100 mM KC1, and 0.01% Tween 20. Samples were incubated for 30 min at 20 "C (23 pl total volume), diluted with 7 pl of 100 mM MOPS (pH 6.8), 40% glycerol, and 0.014% bromphenol blue, and immediately loaded on the gel. Free and complexed proteins were separated by electrophoresis on 6% (w/v) polyacrylamide gels at 40 mA for 3 h at 10 "C. Protein bands were detected using Fast StainTM (Zoion Research).

RESULTS
Identification of Putative Amphipathic Helices in Ht 31, Ht 21, P150, and MAP 2-By using the modified RII overlay procedure of Bergmann and colleagues (19) a human thyroid Xgtll expression library was screened for RII-anchoring proteins. Two cDNA clones encoding RII-anchoring proteins were isolated and designated H t 21 and Ht 31.3 The nucleotide sequence of H t 21 contains a continuous open reading frame of 1,281 base pairs which encodes a protein of 427 amino acids. The COOH-terminal107 amino acids are 79% identical to the murine RII-anchoring protein P150 (19). S' mce conserved structure often reflects conserved function, it seemed likely that this region contained the RII-anchoring domain. Furthermore, previous studies suggest that the last 15 amino acids of P150 are required for RII anchoring (19). The RIIanchoring site of MAP 2 has been localized to a region of 31 residues in the amino terminus of the molecule (24,25). Comparison of the Ht 21 and P150 sequences with the RIIanchoring site of MAP 2 revealed no striking sequence homology. However, as has been observed for other classes of functionally similar proteins, e.g. calmodulin-binding proteins, it is possible that the RII-binding site in each protein is conserved in its secondary structure.
Computer-aided secondary structure predictions for H t 31, H t 21, P150, and MAP 2, performed by the methods of Eisenberg and colleagues (30), were used to identify a region in each protein which displays a high probability for amphipathic helix formation. These regions are up to 14 residues in length and consist of alternating pairs of hydrophobic and hydrophilic amino acids (Fig. L4). Sequence comparison of the four regions revealed little homology, with only one amino acid, a glutamic acid a t position 3, conserved in all the sequences (Fig. 1A). However, comparison of polar and nonpolar residues in this region reveals a striking alignment of hydrophilic and hydrophobic residue types (Fig. 1B). When each sequence was drawn as a helical wheel with 3.6 residues/ turn, the hydrophilic and hydrophobic residues became grouped on opposite faces (Fig. 1B). No other sequences in these proteins could be similarly aligned, suggesting these are functionally relevant regions in each molecule.
Identification of Putative Amphipathic Helices and the Minimum Region of H t 31 Required for RII Binding-The analysis of the protein sequence of Ht 31 (1035 amino acids) for potential amphipathic helices identified a region between residues 494 and 507 (Leu-Ile-Glu-Glu-Ala-Ala-Ser-Arg-Ile-Val-Asp-Ala-Val-Ile) (Fig. 1A). This region of Ht 31 is 43% identical in sequence to the binding domain of the RIIanchoring protein MAP 2 (Fig. 1A) and is also 43% identical to Ht 21/P150, although different amino acids are identical.
To determine whether residues 494-507 of Ht 31 are involved in RII binding, a 319-amino acid fragment of H t 31, representing residues 418-736 of the protein, was expressed in E. coli. H t 31 (418-736) bound RIIa as assessed by solid-phase binding (Fig. 2). Furthermore, Ht 31 (486-717), a 232-residue Ht 31 fragment which contained the amphipathic region at the extreme amino terminus, also bound RII (data not shown). These results are compatible with our hypothesis that the amphipathic helix region of Ht 31 is necessary for RII binding, Disruption of RII-Ht 31 Interaction by Mutagenesis within the Amphipathic Helix Region-To determine if an intact ahelix is required for RII binding, a family of H t 31 point mutants was produced in the Ht 31 (418-736) fragment (Fig.  3A). The introduction of proline, especially toward the COOH terminus of an a-helix conformation, often disrupts the sec-  Each sequence was drawn as an tr-helix with 3.6 amino acid residues/ turn. Shndrd awns identify the hydrophobic face of each amphipathic helix. Amino acid residues are represented hv single-letter codes.
ondary structure of the region and can result in a 20" bend in the helix. Hydrophobic residues, located within the second and third turns (helix residues 5 and 9) and at the COOHterminal end of t,he predicted helix region (residue 14), were mutated to proline (Fig. 3 A ) . Mutants were expressed in the pETkfc vector system, and each protein was analyzed for the ability to bind RII by solid-phase binding assay (Fig. 3, R and   C ) . Furthermore, each mutant protein was assessed for RIIn interaction by gel-shift analysis in polyacrylamide gels (Fig.  4). This technique monitors protein-protein interactions under nondenaturing conditions and was performed to ensure that results obtained by solid-phase binding were not an artifact of the denaturing conditions used.

DISCUSSION
Computer-aided searches for potential secondary structure in four RII-anchoring proteins have identified a common region in each molecule which displays n high probability for amphipathic helix formation. The putative amphipathic helix in M A P 2 is located within the 31-amino ncid RII-binding domain (24,25). while the amphipathic helix region in 1'150 is located just upstream of the COOH-terminal region which, if removed, prevents binding to RII (19). The data presented in this study strongly suggest that a similar amphipathic helix in the thyroid RII-anchoring protein Ht 31 is required for binding RIIn. Mutations which should disrupt the helicity of this region prevent RIItr binding. These findings have led us to propose that amphipathic helices, on the surface of R I Ianchoring proteins, form the sites for interaction with Rlltr.
The amphipathic/amphiphilic n-helix is a common secondary structure motif found in several biologically active peptides and proteins (26). It is defined as an tr-helix with opposing hydrophilic and hydrophobic faces orientated down the long axis of the helix. To test our h-ypothesis that an ( 1helix conformation between residues 494 and 507 of Ht 31 was required for interaction with RIIn, we disrupted the helicity of the region by substituting selected hvdrophohic residues with proline. Proline is regarded as sterically incompatible with normal tu-helical conformations hecause of the imino peptide linkage. For example, the recently solved crystal structure of the E. coli DNA-binding protein, FIS, contains a 20-residue a-helix which is distorted by 20" along its axis due to a proline at position 13 (31). Therefore, the substitution of proline a t residues 498, 502, and 507 likely distorts the secondary structure a t three distinct positions within the predicted a-helix, with only minor alterations to the amphipathic or hydrophobic nature of the region. Since each point mutation diminished or abolished RII binding, it is reasonable to assume that an a-helical conformation between residues 494 and 507 of H t 31 is necessary for interaction with RII. A characteristic of the amphipathic helix motif is the ordered placement of alternating pairs of hydrophobic and hydrophilic amino acids within the linear sequence of a protein (26). In addition to this criterion, each RII-anchoring protein contains acidic amino acids distributed over the hy- drophilic face of the helix. In particular, an invariant glutamic acid at position 3 is located within the first turn of the amphipathic helix (Fig. 1B). Therefore, the overall sequence characteristics of an RII-anchoring site appear to be approximately a 14-residue region of sequence rich in acidic and hydrophobic amino acids with a high probability for amphipathic helix formation. The lack of a consensus RII binding sequence is consistent with the hypothesis that each molecule contains a common secondary structure. Another example of conserved secondary structure is the family of 20 or so calmodulin-binding peptides, which exhibit little or no conservation in primary structure other than an abundance of basic and hydrophobic amino acids (32). The basic faces of these amphipathic peptides form ionic interactions with the acidic central helix of calmodulin (32). In fact, synthetic peptides consisting of alternating pairs of leucine and lysine residues have nanomolar binding affinities for calmodulin (33). Future experiments are planned to synthesize RII-anchoring peptides and determine whether acidic residues within the hydrophilic face of the amphipathic helix region are essential for RII anchoring. The identification of the RII-anchoring site in H t 31 has provided structural information which can be tested on H t 21 and may allow the development of peptide antagonists of RII anchoring. These reagents will be introduced into cells in order to disrupt the normal subcellular localization of PKA in vivo.
H t 31 and MAP 2 are 43% identical over the 14-residue amphipathic helix region, and Ht 21 and P150 have 93% sequence identity (Fig. lA), suggesting that there may be two classes of RII-anchoring proteins. It is of interest to note that MAP 2 has a 7-fold higher affinity for the RIIa isoform than RIIP (12). It will be important to establish whether H t 21 and H t 31 preferentially bind to a particular RII isoform. The differential subcellular distribution of RIIa and RIIP in neurons, demonstrated by electron microscope immunocytochem-

RII Binds
to Amphipathic Helices on RII-anchoring Proteins istry (34), may reflect, in part, a preferential interaction with distinct RII-anchoring proteins. Structural and functional analysis of additional RII-anchoring proteins is required to establish if, in fact, there are two classes of RII-anchoring sites and if they preferentially associate with a specific RII isoform. Nevertheless, the preferential interaction of RII isoforms with specific classes of RII-anchoring proteins is an attractive mechanism to govern the subcellular localization of different PKA isoforms.
Our previous studies have shown that the RII dimer is required for anchoring protein interaction (22). Furthermore, we and others have shown that MAP 2 and a fragment of P75 (the bovine homolog of P150) bind at the same or overlapping sites on RII (22, 23). It is likely, therefore, that the amphipathic helix of the RII-anchoring protein interacts with sites on both RII subunits. The nature of these interactions is unclear but may involve ionic interactions between the acidic face of the amphipathic helix and basic residues located on both RII protomers. Whereas the amphipathic helix region is responsible for RII association, it is clear that other regions in each anchoring protein are responsible for specific interaction with membranes, organelles, or the cytoskeleton. This %fold binding could potentially determine the subcellular location of the PKA (6). Presumably, these interactions anchor PKA in close proximity to its preferred substrates (6). Recently, a splice variant of the bovine PKA catalytic subunit, CP2, has been identified which contains a putative amphipathic helix at the amino terminus, replacing the normal myristylation signal (35). The role of this amphipathic helix is unknown, but the authors suggest it may function to tether PKA catalytic subunit to membranes. Potentially, amphipathic helices could be responsible for anchoring both RII and CP2 subunits, such that the active catalytic subunit was immobilized and remained close to its regulatory subunit upon cAMP activation. Under these conditions, the accessibility of PKA to particular substrates would be tightly controlled and kinase activity would be highly regulated, since rapid reassociation of the holoenzyme complex would occur after cAMP concentrations returned to basal levels.