Isoform-specific Differences between the Type Iα and IIα Cyclic AMP-dependent Protein Kinase Anchoring Domains Revealed by Solution NMR*

Cyclic AMP dependent protein kinase (PKA) is controlled, in part, by the subcellular localization of the enzyme (1). Discovery of dual specificityanchoring proteins(d-AKAPs) indicates that not only is the type II, but also the type I, enzyme localized (2). It appears that the type I enzyme is localized in a novel, dynamic fashion as opposed to the apparent static localization of the type II enzyme. Recently, the structure of the dimerization/docking (D/D) domain from the type II enzyme was solved (3). This work revealed an X-type four-helix bundle motif with a hydrophobic patch that modulates AKAP interactions. To understand the dynamic versus static localization of PKA, multidimensional NMR techniques were used to investigate the structural features of the type I D/D domain. Our results indicate a conserved helix-turn-helix motif in the type I and type II D/D domains. However, important differences between the two domains are evident in the extreme NH2 terminus: this region is extended in the type II domain, whereas it is helical in the type I protein. The NH2-terminal residues in RIIα contain determinants for anchoring, and the orientation and packing of this helical element in the RIα structure may have profound consequences in the recognition surface presented to the AKAPs.

Extracellular signals are relayed from the plasma membrane to specific intracellular targets with precision and speed. Many signaling pathways do so by altering the phosphorylation state of the target proteins. Kinases and phosphatases have broad substrate specificity, and mechanisms exist to organize and effectively concentrate the correct repertoires of enzymes into distinct signaling cascades. In the case of cAMP-dependent protein kinase (PKA), 1 the importance of signal integration and coordinated assembly of signaling cascades has been established (1). The opposing actions of adenylate cyclase and phosphodiesterases generate localized gradients of cAMP (4), which will exert its greatest influence only when concentrated pools of PKA are colocalized in an inactive conformation (5,6). To account for this, Scott and co-workers (7) have proposed the "targeting hypothesis." This hypothesis states that phosphorylation events are not only controlled by the balance of kinase and phosphatase activity, but also by their respective subcellular localization (7).
It is the regulatory subunits of PKA that mediate subcellular compartmentalization via their binding to A-Kinase anchoring proteins (AKAPs) (6,8). The anchoring of the type II PKA holoenzyme has been extensively investigated, and AKAPs have been found in centrosomes, mitochondria, Golgi, microtubules, filopodia, dendrites, and the plasma membrane (1, 9 -15). More recently, the structural basis for this high affinity interaction was revealed by solution NMR (3). The X-type four helix bundle topology in RII␣-(1-44) provides a hydrophobic interaction surface for an amphipathic helical motif in the AKAPs (3,16).
Two novel AKAPs, D-AKAP1 and D-AKAP2 (2,17), were recently isolated and were designated dual specificity AKAPs, since they interacted with the type I as well as the type II regulatory subunits. RII␣, however, binds this class of proteins with a 25-500-fold higher affinity than RI (2,17). Since the cloning of D-AKAPs, other RI-binding proteins have emerged (18 -21). Recently, a novel Caenorhabditis elegans AKAP that binds C. elegans R-subunit, R CE , has been cloned (21). AKAP CE binds R CE with nanomolar affinity, and neither RII␣ nor RII␤ competitively inhibit formation of AKAP CE -R CE complexes. These studies suggest that RI␣ dimers, like RII␣, have a site available for interactions for anchoring proteins. Although RI␣ is mostly cytoplasmic in certain cell types, it is clearly anchored in vivo (22)(23)(24)(25). The dynamic nature of RI␣ anchoring, in contrast to the static anchoring of RII␣, is of critical importance in vivo and needs to be understood at the molecular as well as the structural level.
The recent solution NMR structure of the RII␣ dimerization/ docking domain (D/D) revealed an X-type four-helix bundle topology (3). Although one would expect a conserved fold among these family members, the differential affinities to the various AKAPs and other target proteins, and the absence of disulfide bonds in RII suggests that there are potentially some subtle, yet critical, structural differences dictating different mechanisms of homodimerization and AKAP interactions. We bacterially expressed, 13 C/ 15 N isotopically enriched, and purified a 50-residue fragment, corresponding to the D/D domain of RI␣. Residues 12-61 of RI␣ were previously identified as the tryptic core responsible for dimerization (26). This region is highly stable and is capable of binding the D-AKAPs (2,27). We report here the 1 H, 15 N, and 13 C backbone resonance assignments and secondary structure analysis of uniformly 15 N-and 13 C-enriched RI␣ , as determined by multidimensional NMR techniques. Like RII␣- , this region contains two regions of ␣-helical structure separated by a turn. A key difference, however, is observed when the secondary structures are compared. The extreme NH 2 terminus in RII␣-(1-44) is extended, however, in RI␣- , this region is helical. This altered structural element in RI␣-  at the extreme NH 2 terminus has implications for the unique quaternary contacts in RI␣ and affects homodimerization as well as AKAP binding.

EXPERIMENTAL PROCEDURES
Expression of RI␣-(1-61)-cDNA encoding recombinant RI␣-(1-61) was inserted into the EcoRI site of pRSETc (Invitrogen, Inc.). This construct introduced a His 6 tag at the amino terminus and allowed for rapid purification of the expressed protein by Ni 2ϩ affinity chromatography. Residues 12-61, identified previously as the trypsin-resistant core, were then purified to homogeneity by gel filtration and high performance liquid chromatography (26). The gene encoding RI␣-(1-61) contains two rare arginine codons for expression in Escherichia coli that significantly compromised bacterial expression. Once the plasmid encoding the rare Arg codon, dnaY, was supplied, expression was dramatically enhanced (28,29). For obtaining unlabeled RI␣-(12-61) the expression plasmid was transformed into BL21-DE3 (dnaY) cells (100 g/ml ampicillin; 70 g/ml kanamycin). LB media (4 -6 liters) were then inoculated with a growing culture that had reached an A 600 of 0.5-0.6. The culture was grown at 37°C to an A 600 of 0.6 -0.8 prior to induction with 1 mM IPTG. The cells were grown for an additional 5 h and then harvested by centrifugation for 11 min at 5000 rpm.
Maximum expression of 15 N-labeled protein was obtained using growth media that contained M9 salts (30) to which 1 M FeCl 3 , 25 M ZnSO 4 , 0.1 mM CaCl 2 , and 0.5 mM MgSO 4 , and 20 ml/liter vitamin supplements (Life Technologies, Inc.) were added. Cell growth was initiated by inoculating a 50-ml culture of 2XYT with an expression cell stock to mid-log phase (A 600 of 0.4 -0.6) and then harvesting and dissolving in 500 l of minimal media containing antibiotics. This was then used to inoculate 6 -8 liters of prewarmed M9 salts/glucose media (250 l/liter). The culture was grown at 37°C to an A 600 of 0.5-0.6 prior to induction with 1 mM IPTG. The cells were grown for an additional 5 h post-induction and then harvested by centrifugation for 11 min at 5000 rpm. 13 C/ 15 N isotopically labeled protein was expressed in 13 C/ 15 N Celtone medium, Celtone CN (Martek Biosciences, Inc.) in a 2.5-liter fermentor flask using the Bioflo3000 (New Brunswick, Inc.) Cells were grown at 37°C, pH 7.4. pH was controlled with 5 N NaOH and 5 N H 2 SO 4 . Dissolved oxygen level was maintained at 30% by setting an agitation/ dissolved oxygen cascade with an agitation range of 300 -800 rpm. Cells were induced with 1 mM IPTG at an A 600 of 0.9 and dissolved oxygen of 28% and grown for an additional 5 h.
Purification of RI␣-(12-61)-Cells were resuspended in buffer A (50 mM potassium phosphate, 300 mM sodium chloride, pH 8.0, at 4°C) and lysed by passing through a French pressure cell twice. The cellular debris was then removed by centrifugation for 40 min at 15,000 rpm. The supernatant was batch bound to Ni 2ϩ or Co 2ϩ resin (Invitrogen, Inc.; CLONTECH, Inc.) that had been pre-equilibrated with buffer A. The protein was purified as described previously (26).
RI␣-(12-61) was Ͼ95% pure as assessed by SDS-polyacrylamide gel electrophoresis. Complete cleavage of the fusion protein was confirmed by electrospray ionization mass spectrometric analysis (ESI) as well as by NH 2 -terminal sequencing. Protein concentration was determined by quantitative amino acid analysis.
Sample Preparation and Experimental Conditions-Optimal sample conditions were established and assessed by analytical gel filtration, dynamic light scattering, as well as two-dimensional homonuclear TOCSY experiments (31). The optimal buffer conditions were determined to be 50 mM sodium acetate, 150 mM sodium chloride at pH 4.0, and the ideal temperature was at 37°C.
NMR Experiments-All NMR experiments were performed at 37°C on either a Bruker DMX500 or a DRX600 spectrometer using a tripleresonance gradient probe. All homonuclear experiments were done using the same sample as described for the two-dimensional TOCSY experiments. Two-dimensional homonuclear NOESY experiments (32) were collected with mixing times of 150 ms. The 1 H spectral widths were 6944 Hz (12 ppm) in t 2 and t 1 . 2000 by 512 points were collected. Water suppression was achieved through a WATERGATE sequence (33). The 1 H chemical shift of water was 4.717 ppm when referenced to sodium 2,2-dimethyl-2-silapentane-5-sulfonate, while 15 N and 13 C were indirectly referenced to NH 3 (liquid) and 2,2-dimethyl-2-silapentane-5sulfonate, respectively, as established by Wishart et al. (34 -37).
The isotopically enriched samples, typically, were prepared at a dimer concentration of 1.2-1.6 mM (2.4 -3.2 mM monomer), 50 mM sodium acetate, and 150 mM sodium chloride at pH 4.0. The twodimensional 1 H-15 N heteronuclear single quantum coherence (HSQC) experiment (38), the 3D 1 H-15 N HSQC-TOCSY (38), the three-dimensional 1 H-15 N HSQC-NOESY (38), and the three-dimensional HNHA experiments (39) were acquired on the same 15 N-labeled sample. In all experiments, the 1 H carrier during the evolution of the 15 N and 1 H dimensions was centered at the water frequency, but was shifted to 7.857 ppm before data acquisition. In all experiments, the 15 N carrier was placed at 118 ppm. A mixing time of 150 ms was used for the HSQC-NOESY, and the HSQC-TOCSY experiment was run at spin lock time of 97 ms for optimal magnetization transfer. In all cases, water suppression was accomplished using the WATERGATE sequence (33,40). Broadband 15 N decoupling for the three 1 H-15 N experiments was done using a WALTZ16 decoupling scheme (41). In the two-dimensional HSQC experiment, 256 real t 1 ( 15 N) and 512 complex t 2 ( 1 H) points were collected with spectral widths of 1277 Hz (21 ppm) and 2315 Hz (4 ppm), respectively. For the three-dimensional 1 H-15 N experiments, a total of 128 complex t 1 ( 1 H), 64 complex t 2 ( 15 N), and 512 complex t 3 ( 1 H) points were collected. The spectral widths were 5109 Hz (8.5 ppm) (t 1 ), 1277 Hz (21 ppm) (t 2 ) and 2404 (4 ppm) (t 3 ). For the three-dimensional HNHA experiment, a total of 32 complex t 1 ( 15 N), 64 complex t 2 ( 1 H), and 512 complex t 3 ( 1 H) points were collected. The spectral widths were 1277 Hz (t 1 ), 6001.3 Hz (10 ppm) (t 2 ), and 2404 Hz (4 ppm) (t 3 ). The HNHA experiment was used to calculate J HN-H␣ coupling constants. Coupling constants were determined based on the intensity ratio of the cross-peak to the diagonal (39,42).
Acquisition and Processing Parameters-Quadrature phase detec-tion in the indirectly detected dimension for all homonuclear experiments and the triple resonance CT-HNCA experiment was obtained via time-proportional phase incrementation (53). The three-dimensional 15 N heteronuclear experiments used sensitivity enhanced gradientdetected echo/antiecho in the 15 N dimension, and the CBCA(CO)NH used STATES (54)in the 13  All experiments were processed using Felix 95.0 software (MSI, San Diego, CA) using an Indigo-2 workstation (Silicon Graphics, Inc.). The data were apodized with a squared sine-bell function shifted by 70°in all dimensions before Fourier transformation. For all three-dimensional experiments, linear prediction to additional one-third of the number of the points collected was applied before apodization in both t 1 and t 2 .
Circular Dichroism Spectroscopy-Circular dichroism (CD) spectra were measured in 50 mM sodium acetate and 150 mM sodium chloride at pH 4.0 as well as in 50 mM potassium phosphate, 150 mM sodium chloride, pH 6.5 at 37°C and 25°C on an Aviv model 202 Spectrometer

RESULTS
Optimizing NMR Conditions-Full-length RI␣ is a dimeric protein of 98 kDa molecular mass (55). The NH 2 -terminal 61 residues are responsible for maintaining the integrity of the dimer as well as D-AKAP interactions (2,27). Initial two-dimensional homonuclear TOCSY experiments (41) of RI␣-(12-61) were acquired in 50 mM potassium phosphate at pH 6.5. The efficiency of TOCSY transfer, under these conditions, is extremely poor (data not shown). The protein had an apparent molecular mass of 18 kDa at pH 6.5 (actual molecular mass of the dimeric fragment ϭ 10 kDa), as determined by gel filtration chromatography, suggesting a higher order association. Moreover, dynamic light scattering indicated a heterogenously dispersed sample (data not shown). A variety of buffer conditions were screened to determine the optimal sample conditions. The ideal sample conditions were observed at a pH of 4.0 and 150 mM sodium chloride (Fig. 1A). Under these conditions, the extent of magnetization transfer in the two-dimensional TOCSY was dramatically improved. These conditions were thus chosen for future NMR studies.
Even though working at low pH conditions is preferred for NMR studies due to lower amide exchange rates, there is a potential risk of compromising the structural integrity of the protein under study. To address this issue, RI␣-(12-61) was analyzed by circular CD to monitor the overall secondary structure. At both pH 6.5 and pH 4.0, the spectra show the classical double minima at 210 and 222 nm expected for helical structures in solution. In addition, the spectra are similar under the two different buffer conditions, suggesting the absence of any gross structural changes due to the different solution conditions (Fig. 1B). Additionally, the functionality of this domain is preserved at these buffer conditions, as indicated by analysis of peptide binding by NMR (61).
Resonance Assignments-The two-dimensional 1 H-15 N HSQC of RI␣-(12-61) is shown in Fig. 2. All 47 expected nonproline backbone resonances, excluding the NH 2 -terminal ser- ine were observed. The spectrum is well dispersed and well resolved. Even though RI␣-(12-61) is a disulfide-bonded covalent dimer, the number of resonances observed is equivalent to the number of backbone resonances expected for a monomeric RI␣- . Therefore, it is inferred that RI␣-(12-61) is indeed in a stable, symmetric homodimeric conformation. In a dimeric system it can be difficult to distinguish interfrom intraresidue NOEs (56,57). Analysis of sequential connectivities is simplified using triple resonance experiments. Comparison of the intraresidue 1 H N , 15 N to 13 C␣/ 13 C␤ connectivities in the CT-HNCA and the HNCACB with sequential intraresidue 1 H N , 15 N to 13 C␣/ 13 C␤ connectivities observed in the CBCA(CO)NH allowed complete identification of the sequential connectivities (Fig. 3). Side chain assignments were obtained using TOCSY-HSQC, HCCH-TOCSY, HC(CO)NH-TOCSY, and C(CO)NH-TOCSY. All 1 H N , 15 N, C␣, C␤, and over 80% of the other side chain assignments were completed.
Secondary Structure Analysis-Carbon chemical shift, short and medium range NOEs, amide proton exchange rates, as well as spin-spin coupling constants ( 3 J HN␣ ) were used collectively to assess the secondary structure conformation in RI␣- . Initial evaluation of secondary structure was done by determining the 13 C␣ chemical shifts and subsequently comparing them to random coil chemical shift information. According to the chemical shift index method (CSI), the , angles of the polypeptide backbone change in ␣-helix and ␤-sheet configuration, which leads to characteristic shifts in the 13 C␣ chemical shifts (34, 36, 37). The secondary shifts as a function of residue number are plotted in Fig. 3. These data indicates that RI␣-(12-61), contains two helices separated by one turn. The first helix, however, does not appear "classical," since it is interrupted at His 23 , Asn 24 , and Ile 25 . This discontinuity in secondary shifts may be explained by the intermolecular disulfide bond that could cause a contortion of the polypeptide backbone leading to unusual , angles not normally expected in an ␣-helix.
␣-Helices, ␤-strands, and ␤-turns display characteristic NOE cross-peaks that help distinguish among them fairly conclusively (58). Analysis of the three-dimensional HSQC-NOESY experiment generated a pattern of NOE cross-peaks for RI␣-(12-61) that confirmed the secondary structure analysis of the 13 C␣ secondary shift data presented in Fig. 3.
␣-Helical Structures-RI␣-(12-61) demonstrates a high proportion of ␣-helix when analyzed by circular dichroism as well as NMR spectroscopy. The first ␣-helical segment is subdivided in two segments, since according to many of the NMR parameters, it cannot be defined as a long uninterrupted helical segment. The first helical segment spans residues Arg 14 through Asn 24 as observed from strong d NN , medium d ␤N(i,iϩ1) cross-peaks. In addition, weak intensity d ␣N(i,iϩ3) and d ␣␤(i,iϩ3) and 13 C secondary shift analysis further indicate this region to be helical. This helix is disrupted at Cys 16 , according to 13 C␣ chemical shift analysis, which could be explained easily by the disulfide bonding that has resulted in unusual torsion angles. Additional indication for the helical nature of this region comes from analysis of coupling constants ( 3 J HN␣ ). Residues Arg 14 through Gln 22 have 3 J HN␣ values less than 5 Hz, indicative of an ␣-helix. Despite the presence of a medium intensity d ␣N(i,iϩ3) between Gln 21 and Asn 24 , and d ␣␤(i,iϩ3) between Asn 24 and Ala 27 , His 23 , and Asn 24 have J HN␣ values of 9.4 and 6.6 Hz, respectively. This region could be kinked or bulged, resulting in aberrant torsion angles. Furthermore, the 13 C secondary shifts for these residues are zero.
Helix I spans Ile 25 through Ala 39 and is defined by strong d NN cross-peaks, medium intensity d ␣N(i,iϩ1) , as well as weak d ␣N(i,iϩ3) , and d ␣␤(i,iϩ3) . Analysis of 13 C secondary shifts also indicates this region to be helical (Fig. 3). Additionally, residues Ile 25 through Thr 38 have 3 J HN␣ values less than the 5.5 Hz characteristic of ␣-helical structures. Residue Arg 40 is not included in the COOH-terminal end of Helix I because of the lack of d ␣N(i,iϩ3) as well as d ␣␤(i,iϩ3) . Above all, this residue has J HN␣ Ͼ5 Hz. Finally, Helix II spans residues Met 45 through Glu 58 by similar criteria. This helix is defined by characteristic ␣-helical NOEs, coupling constants, and 13 C␣ secondary shift analysis.
Hydrogen exchange experiments were used to detect slowly exchanging amide protons in RI␣- . Amide protons that exchange slowly are either solvent inaccessible and thus buried within the protein or are involved in hydrogen bonding interactions, implying stable secondary structure elements. The slowly exchanging protons in RI␣-  lie in the regions defined as ␣-helical from both the NOESY and CSI method, as well as coupling constant analysis (Fig. 3). The main exception is the amide exchange rates for the first helical moiety. These residues show fast exchange rates, despite their helical nature, suggesting that perhaps this helix does not participate in numerous quaternary contacts.
Turns-A turn defined by residues Pro 41 -Glu 42 -Arg 43 -Pro 44 separates Helix I from Helix II. Side chain identification of the prolines was done using the HCCH-TOCSY as well as the three-dimensional 13 C-edited HMQC-NOESY experiments. NOE cross-peaks, 13 C␣ secondary shifts, dihedral angles measurements derived from the HNHA experiment, and H N exchange data suggest a type I conformation. The second and FIG. 3. A, representation of NOE contacts, 13 C␣ chemical shifts, NH exchange, and 3J H␣N coupling constants relevant for secondary structure analysis. The chemical shift differences for 13 C␣ are reported. They are based on the CSI method; a value of ϩ1 is assigned to the residue if it falls downfield of the random coil value 0.7 ppm (helical), and a value of Ϫ1 is assigned to the residue if it falls upfield of the random coil value ϩ/0.7 ppm (␤-sheet). Regions corresponding to each element of secondary structure are indicated. The symbols above the sequence indicate 3 J H␣N coupling constants. Red circles denote coupling constants of less than 5 Hz; gray circles represent coupling constants of Ͼ5 Hz, and the open circles are for those with coupling constants Ͼ8 Hz. Symbols below the sequence represent NH exchange data. Filled circles, open circles, and filled diamonds represent those amide protons present after 2, 7, and 24 h, respectively. B, secondary structure comparison of RI␣ and RII␣ DD domain. 13 C␣ secondary shift analysis of RI␣ and RII␣ are compared. third residues in the turn, Glu 42 and Arg 43 , have 3 J HN␣ values of 4.7 and 9.4 Hz, respectively. These values correspond to those expected for the second and third residues in a Type I turn (58). NOE cross-peaks involved in determining this turn include a medium d NN(i,iϩ1) between Glu 42 and Arg 43 , a weak d ␣N(i, iϩ1) between Pro 41 and Glu 42 , and finally a weak d ␣N(i,iϩ1) between Glu 42 and Arg 43 .
Data obtained from the three-dimensional 13 C-edited HMQC-NOESY experiment reveals that the two proline residues are, in fact, in trans-peptide bonds with their respective preceding residues. In addition, Pro 41 and Pro 44 , show medium d ␣␦(i,iϩ1) cross-peaks, which are indicative of trans-peptide bonds. d ␣ ␣ (i,iϩ1) NOEs, which would indicate a cis-peptide bond for the sequence Xaa i -Pro iϩ1 were not observed. DISCUSSION The identification of a family of anchoring proteins that bind RI␣ has offered novel insights into the unique and dynamic nature of PKA anchoring via RI␣. By analogy with RII␣, it is the D/D domain of RI␣ that is responsible for interaction with AKAPs. There are isoform-specific differences in the nature of these interactions that need to be addressed at the structural and molecular level. In addition, the type I R-subunits are maintained as a dimer by two interchain disulfide bonds, whereas the type II R-subunit are noncovalent dimers (55). We, therefore, pursued structural characterization of the RI␣ D/D domain by NMR to address the potential differences in the molecular basis for homodimerization as well as anchoring. Newlon et al. (3) recently determined the solution structure of RII␣-(1-44) by multidimensional NMR. Given the overall conservation in function among the two R-subunits, one would speculate a conserved fold for the NH 2 terminus. However, there is biochemical evidence to expect differences in quaternary structure. We expressed isotopically enriched RI␣-  with 15 N and 13 C nuclei and purified it to homogeneity. This region is necessary and sufficient for binding D-AKAPs (2,27).
Secondary Structure Comparison with RII␣-Secondary structure analysis of RI␣ and RII␣ shows that elements of secondary structure are conserved among the two domains. Both contain two conserved regions of helicity separated by a turn. As speculated earlier in our mutagenesis studies of RI␣ and demonstrated in the NMR solution structure of the RII␣ D/D domain, these two helices correspond to two functional subdomains each serving a distinct role (3,27,31). Helix I, or subdomain 1, contains determinants for anchoring, whereas Helix II, or subdomain 2, contains determinants for dimerization. In contrast to these conserved helices there are major differences in the secondary structure at the extreme NH 2 terminus in the small segment that precedes Helix I. 13 C secondary shift analysis shows that instead of an extended strand for the first 9 residues in RII␣, RI␣ has a helical segment (Fig.  3B). We propose that the region corresponding to the extended ␤-strand in RII␣, and a helical region in RI␣, are isoformspecific, variant in sequence and secondary structure, and may provide additional requirements for docking. In fact, the two critical isoleucines (Ile 3 and Ile 5 ) that have been implicated in AKAP binding to RII␣ reside in this region (3,59). The importance of this region in mediating anchoring interactions will be addressed eventually in the solution structure of RI␣-  in complex with an anchoring protein.
Structural Implications-The RII␣-(1-44) dimer folds as an X-type four helix bundle that encompasses an extended hydrophobic dimerization interface as well as an AKAP binding surface (3). Extensive, well ordered hydrophobic interactions define the dimerization interface in RII␣. Residues important in dimer contacts reside in Helix I as well as Helix II as shown in Fig. 4A. Sequence alignment with RI␣ shows that these residues are mostly conserved. The conserved residues among the two isoforms are depicted on the RII␣ structure in green space filling models (Fig. 4B). Given the preponderance of the hydrophobic interactions in dictating the dimerization interface in RII␣ and the conservation of these residues in RI␣, it is likely that the two proteins have a similar quaternary fold. The extreme NH 2 terminus that is extended in RII␣ is colored red in Fig. 4B. This is the region that corresponds to the first helical segment in RI␣. Once the first 23 residues are deleted in the full-length RI␣, the stability of the dimer is compromised due to the shift in the monomer/dimer equilibrium to the monomeric species, although the dimeric species is still present (27). This deletion mutation corresponds to deleting Helix N-1. This helix is most likely contributing to the stability of the dimerization interface, and its quaternary contacts could affect the dimer core in RI␣.
RI␣ contains two interchain disulfide bonds: Cys 16 , in the extreme NH 2 -terminal region, is linked in an antiparallel orientation to Cys 37 in Helix I (55). Mapping these sites on the RII␣-(1-44) structure would suggest that in RI␣, the region that parallels the extended region in RII␣-(1-44) must somehow contact Helix I to accommodate the interchain disulfide linkage with the correct geometry.
Anchoring Interactions-In addition to the hydrophobic core, RII␣ also presents an extended, solvent accessible hydrophobic surface (3). This surface is formed by the antiparallel arrangement of helices I and IЈ (3). Residues that show direct contact with Ht31 are shown in Fig. 5. Only three of these residues are conserved between RI␣ and RII␣, which could explain the differences in the affinities of these domain for the various AKAPs. It is important to stress that two critical residues for AKAP binding in RII␣, Ile 3 and Ile 5 , lie within the ␤-strand extended region. These determinants are missing in RI␣ and may account for the differences in D-AKAP binding affinity. Moreover, the position and orientation of the helical segment in RI␣ and its quaternary contacts could have an important consequence in terms of the surface presented for AKAP binding. Due to the antiparallel linkage of Cys 16 and Cys 37 , the helical segment in RI␣ must somehow fold back onto Helix I and IЈ and potentially occlude the binding interface that is accessible in the case of RII␣- . The three-dimensional solution structure of RI␣ is in progress to specifically address the nature of the peptide binding surface present in RI␣.
Conclusions-The different isoforms of the R-subunits do not have redundant functions in vivo and, in fact, contribute distinct regulation of PKA activity. Targeted disruption of the RII␣ gene results in viable mice with no physiological defects, whereas RI␣ knockout mice are embryonically lethal (60). The potentiation of the L-type Ca 2ϩ channel in skeletal muscle cells from RII␣ knockout mice is retained, and, in fact, immunocytochemical studies show that the C subunit of the type I holoenzyme is colocalized in transverse tubules with the L-type Ca 2ϩ channels. In RI␣ and RII␣ knockout mice, there is an increase in the level of RI␣ protein in tissues that normally express the ␤ isoforms (60). Compensation by RI␣ represents a crucial biological mechanism for safeguarding the cells from unregulated PKA activity. Understanding the molecular and structural differences in isoform-specific PKA anchoring interactions is thus of utmost importance.
We report the first structural characterization of the RI␣ D/D domain. Comparison of this domain with the solution structure of the RII␣ D/D domain indicates that they both share a conserved helical scaffold contributing to two distinct functions, namely dimerization and anchoring. However, the extreme NH 2 terminus is distinct structurally. This region will contribute to the dimerization interface, since it contains one of the cysteines involved in the antiparallel disulfide linkage. More importantly this region in RII␣ contains determinants for anchoring, and the orientation and packing of this helical element in RI␣ will have consequences in terms of the recognition surface presented to AKAPs. The solution structure of RI␣ alone and in complex with an anchoring peptide will help delineate the contributions of this unique element of secondary structure in RI␣ that sets it apart from RII␣ structurally and functionally.