THE JOURNAL
OF
BIOLOGICAL CHEMISTRY
Vol. 279, No. 18, Issue of April 30, pp. 19084 –19090, 2004 Printed in U.S.A.
C Subunits Binding to the Protein Kinase A RI␣ Dimer Induce a Large Conformational Change* Received for publication, December 8, 2003, and in revised form, February 24, 2004 Published, JBC Papers in Press, February 25, 2004, DOI 10.1074/jbc.M313405200
William T. Heller‡§¶, Dominico Vigil¶储**, Simon Brown储, Donald K. Blumenthal‡‡, Susan S. Taylor储, and Jill Trewhella‡§§ From the ‡Bioscience Division, Los Alamos National Laboratory, Los Alamos, New Mexico 87545, the 储Department of Chemistry and Biochemistry and the Howard Hughes Medical Institute, University of California, San Diego, La Jolla, California 92037, and the ‡‡Departments of Pharmacology & Toxicology and Biochemistry, University of Utah, Salt Lake City, Utah 84112
We present structural data on the RI␣ isoform of the cAMP-dependent protein kinase A that reveal, for the first time, a large scale conformational change within the RI␣ homodimer upon catalytic subunit binding. This result infers that the inhibition of catalytic subunit activity is not the result of a simple docking process but rather is a multi-step process involving local conformational changes both in the cAMP-binding domains as well as in the linker region of the regulatory subunit that impact the global structure of the regulatory homodimer. The results were obtained using small-angle neutron scattering with contrast variation and deuterium labeling. From these experiments we derived information on the shapes and dispositions of the catalytic subunits and regulatory homodimer within a holoenzyme reconstituted with a deuterated regulatory subunit. The scattering data also show that, despite extensive sequence homology between the isoforms, the overall structure of the type I␣ holoenzyme is significantly more compact than the type II␣ isoform. We present a model of the type I␣ holoenzyme, built using available high-resolution structures of the component subunits and domains, which best fits the neutron-scattering data. In this model, the type I␣ holoenzyme forms a flattened V shape with the RI␣ dimerization domain at the point of the V and the cAMP-binding domains of the RI␣ subunits with their bound catalytic subunits at the ends.
* This work was performed in part under the auspices of the U. S. Department of Energy Contract W-7405-ENG-36 and in support of the Office of Science/Biological and Environmental Research Oak Ridge Center for Structural Molecular Biology (J. T. and W. T. H.). Support from this work also comes from the University of California Directed Research and Development (UCDRD) Collaborative UC/Los Alamos Research (CULAR) Grant 10005, and National Institutes of Health Grants GM19301 and GM34921 (to S. S. T.). Neutron scattering data were obtained using instrumentation supported by the National Science Foundation under agreement DMR-9423101 at the Cold Neutron Research Facility at the National Institute of Standards and Technology. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. § Present address: Condensed Matter Sciences Division and Center for Structural Molecular Biology, Oak Ridge National Laboratory, Oak Ridge, TN 37831. ¶ These authors contributed equally to this work. ** Supported by a National Science Foundation Graduate Research Fellowship. §§ To whom correspondence should be addressed. Tel.: 505-667-2690; Fax: 505-667-2670; E-mail:
[email protected].
Many cellular signaling pathways in eukaryotes involve protein kinases, which phosphorylate Ser, Thr, and Tyr residues in a variety of target proteins. The cAMP-dependent protein kinase A (PKA1 or protein kinase A) is one of the best-studied members of the Ser/Thr protein kinase family. PKA is known to be involved in the regulation of a large number of cellular processes including metabolism, contractile activity, growth, apoptosis, and ion flux (1). Mutations in PKA can lead to diseases such as Carney complex (2) and lupus (3). The catalytic (C) subunits of PKA are responsible for catalyzing the phospho-transfer reaction, whereas the regulatory (R) subunits serve both to confer cAMP dependence and to localize the holoenzyme to discrete subcellular locations via interactions with A-kinase anchoring proteins (AKAPs) (4). At low cAMP concentrations, PKA is maintained as an inactive tetrameric holoenzyme complex (R2C2) consisting of a homodimeric R2 subunit and two C subunits. When intracellular concentrations of cAMP increase in response to specific cellular stimuli, four cAMP molecules bind to each R2 subunit. This event causes a release of inhibition of C by R, allowing the C subunits to phosphorylate their target proteins. There are four major isoforms of PKA, types I␣, I, II␣, and II, which differ with respect to their R subunits (RI␣, RI, RII␣, and RII, respectively). The isoforms have different biological functions, as determined by genetic studies using mice. For instance, mice lacking the RI␣ gene die in utero (5), whereas mice lacking the RII gene are viable, lean, and resistant to diet-induced obesity (6). Despite their differing biological functions, all of the R isoforms share the same domain organization. At the N terminus of each R subunit is a dimerization/docking (D/D) domain that serves both to dimerize and anchor the R subunits to AKAPs. The C terminus of the R subunit consists of tandem cAMP-binding domains. Between the D/D domain and the cAMP-binding domains is a linker region that contains a pseudo-substrate inhibitor region that binds to the active site of the C subunits. Mutagenesis (7) and hydrogen/deuterium exchange experiments (8) have identified the helical subdomain in the more N-terminal cAMP-binding domain (domain A) of RI␣ as an additional region of contact between the R and C subunits. There is high sequence homology between the isoforms in the D/D domain and the cAMPbinding domains, but the linker regions are quite different in length and sequence (9). High-resolution structures of the C subunit (10), cAMP-bind-
1 The abbreviations used are: PKA, protein kinase A; R, regulator subunit; C, catalytic subunit; RC, R and C heterodimer; AKAP, A-kinase anchoring protein; D/D, dimerization/docking; SANS, small-angle neutron scattering; and SAXS, small-angle x-ray scattering.
19084
This paper is available on line at http://www.jbc.org
Solution Structure of Type I Protein Kinase A ing domains of RI␣ (11) and RII (12), and the D/D domains of RII␣ (13) and RI␣ (14) are available and provide important molecular insights into the activation process. The C subunit is a globular, bilobal protein with the active site in a cleft between the lobes. The cAMP-binding domains of the R subunits each consist of a -barrel region that binds cAMP, and a helical region, whereas the D/D domain consists of an antiparallel four-helix bundle. There are no high-resolution structures of the linker regions, which are thought to be relatively flexible in solution (15). Similarly, there are no high-resolution structures of the full-length R homodimers or holoenzyme complexes, which is likely due to the inherent flexibility of the linker regions. Ultracentrifugation and analytical gel filtration indicate that there are significant structural differences between the isoforms. Specifically, the type II␣ holoenzyme is more elongated than the type I␣. The reported Stokes radii range from 56.8 – 60 Å for the type II␣ holoenzyme, and 47.4 –53.8 Å for the type I␣ holoenzyme (16 –18). This difference in Stokes radii is larger than expected for the relatively small difference in isoform molecular weight. Neutron scattering with contrast variation and deuterium labeling is a useful technique for determining the relative shapes and positions of subunits within a molecular complex. The technique was used previously to investigate the subunit arrangement of the PKA type II␣ holoenzyme (19). The neutron data and associated modeling revealed an extended dumbbell shape for the holoenzyme, with the C subunits located in the lobes of the dumbbell and separated by ⬃120 Å. To further investigate the structural differences between the PKA isoforms and the conformational changes involved in the activation process, we completed small-angle neutron scattering experiments with contrast variation on the type I␣ holoenzyme with deuterated R subunits. To aid in the interpretation of the scattering data, we built models of the RI␣ holoenzyme using the NMR structure of the RI␣ D/D domain (14) and two previously published models of the RC heterodimer (20, 21) to determine the arrangements of subunits that best fit the contrast series data. The models and the structural parameters derived from the scattering data (Rg, dmax, and P(r)) indicate that the RI␣ holoenzyme is significantly more compact than the type II␣ holoenzyme studied previously by small-angle neutron scattering with contrast variation (19) as well as a recombinant RII␣ holoenzyme prepared similarly to the RI␣ holoenzyme studied here. The two C subunits in the RI␣ complex are well separated and do not interact with each other, just as in the type II␣ isoform. However, the R subunit homodimer in the type I␣ complex forms a flattened V-shape in contrast to the more extended dumbbell shape seen for the RII␣ homodimer. Interestingly, the RI␣ homodimer is more extended when bound to the C-subunits than when it is free in solution (22). Our results clearly demonstrate that the interaction of the C subunits with the R subunits plays a critical role in defining the domain organization within the holoenzyme. MATERIALS AND METHODS
Cloning, Protein Expression, and Purification—To create an isopropyl-1-thio--D-galactopyranoside-inducible construct that makes more efficient use of the expensive deuterated media required to produce deuterated R, the wild-type bovine RI␣ gene was subcloned out of a wild-type RI␣ pUC118 vector into an ampicillin-resistant pRSET vector (Invitrogen) using standard methods. After optimizing media deuteration levels to maximize protein yield, large scale expression was performed in 2 liters of 97% deuterated Martek-9 media and unlabeled Martek-9 media at a 7:3 ratio. Recombinant bovine R (RI␣ and RII␣) was purified from Escherichia coli Bl21-DE3 cells using cAMP-agarose resin (23) and eluted with 25 mM cGMP instead of cAMP. Recombinant, non-myristoylated bovine catalytic subunit was purified from E. coli Bl21-DE3 cells as described previously (24). Holoenzymes were then reconstituted as described elsewhere (8).
19085
Small-Angle X-ray Scattering (SAXS) Measurements—The SAXS data were collected using the line source instrument at Los Alamos National Laboratory (25). Intensity data were reduced to I(q) versus q using standard procedures to correct for detector sensitivity and background signal, and a smearing procedure was used to correct for the slit geometry of the instrument (25). The SAXS data from the holoenzyme were compared with a lysozyme standard (26) to verify that the samples were free from nonspecific aggregation and the influence of inter-particle interference. Small-Angle Neutron Scattering (SANS) Measurements—Samples of the type I␣ holoenzyme with deuterated R subunit in H2O/D2O mixtures containing 0, 20, 40, 80, and 90% D2O were prepared by mixing stock solutions of the holoenzyme dissolved in H2O and D2O to provide five neutron contrast values. The samples and corresponding solvent blanks with 0, 20, and 40% D2O were measured in 1-mm path length cells, whereas 20-mm path length cells were used for the samples with 80 and 90% D2O buffer. The precise D2O content of each sample was determined by comparing the neutron absorbance, , to that of pure H2O and D2O. The neutron absorbance is obtained from the measured sample transmission using the relationship ⫽ ln(T)/d, where is the neutron absorbance, T is the transmittance, and d is the sample path length. Neutron absorbance varies linearly with hydrogen content of the buffer; thus, using the relation presented below in Equation 1, P ⫽ 1 ⫺ 共共 sam ⫺ D2O兲/共H2O ⫺ D2O兲兲
(Eq. 1)
the solvent deuteration fraction of the sample, P, can be determined. The D2O content of each sample determined in this manner was used throughout the analysis of the data. The SANS data were collected at the Center for Neutron Research at the National Institute of Standards and Technology (Gaithersburg, MD) using the NG-3 30m SANS instrument (27). A neutron wavelength of 5.5 Å was used with a wavelength spread ⌬/ of 0.26 to provide the maximum flux. Sample and background intensities were collected at sample-to-detector distances of 1.5 and 6 m to obtain data over the required q-range. Data reduction to one-dimensional scattered intensity profiles, I(q) versus q, followed standard procedures to correct for detector sensitivity and sample background (27). The reduced intensities from the two detector distances were merged using routines included with the data reduction software provided by the National Institute of Standards and Technology. Small-Angle Scattering Data Analysis—The small-angle scattering intensity profile of monodisperse, identical particles in solution is given by Equation 2, I共q兲 ⫽
冏 冓 冕 共共rជ 兲 ⫺ 兲e s
V
⫺iqជ 䡠rជ
d 3r
冔冏
2
(Eq. 2)
where (ជr) is the scattering length density of the particle as a function of position rជ within the volume V, s is the average scattering length density of the solvent, and qជ is the momentum transfer, having the magnitude 4(sin)/, where 2 is the scattering angle and is the wavelength. The integration is averaged over all conformations and orientations of the particles in solution, because small-angle solution scattering measures the time- and ensemble-averaged information from all of the particles in the sampled volume. As shown below in Equation 3, P共r兲 ⫽
1 22
冕
⬁
dq䡠共qr兲䡠I共q兲sin共qr兲
(Eq. 3)
0
the inverse Fourier transform of I(q) versus q gives P(r) versus r, the probable distribution of vector lengths, r, between scattering centers within the scattering object. P(r) is readily interpreted in terms of the shape of the scattering object. The indirect Fourier transform algorithm originally described by Moore (28) was used to determine P(r) from I(q). The boundary conditions P(r)/r ⫽ 0 at r ⫽ 0 and the maximum linear dimension, dmax, are applied to P(r). The contrast series intensity profiles I(q) can be written as a set of linear equations in the three basic scattering functions, i.e. the scattering functions of the labeled and unlabeled components and a cross-term (29, 30). The P(r) calculated from the cross-term gives the distribution of vector lengths between the labeled and unlabeled components of the complex. A multiple linear regression routine (31) implemented in the C programming language at Los Alamos National Laboratory was used to solve for the three basic scattering functions using the contrast series data. These basic scattering functions were used to derive information
19086
Solution Structure of Type I Protein Kinase A
on the shapes and relative dispositions of deuterated and non-deuterated components in the system. Structural Models—Structural models of the isotopically labeled subunits and the overall complex were determined from the SANS data using an algorithm developed at Los Alamos National Laboratory and implemented in the program CONTRAST (32). This program can take a set of known high-resolution structures, and/or shapes, and find the relative position and orientation of the components that give the best fit to a set of intensity profiles measured in a neutron contrast series. The best fit model structures are generated using a Monte Carlo approach employed previously (19, 25). To evaluate the quality of the fit of each model to the data, CONTRAST uses the fitting parameter F, as shown below in Equation 4, F⫽
1 N pts
冉冘 Npts
冊
共I共q兲 ⫺ Im共q兲兲2 共q兲⫺2
(Eq. 4)
where Npts is the number of points in the data set, I(q) and Im(q) are the experimental and model intensities, respectively, and (q) is the experimental uncertainty of I(q). The D/D domain used was the NMR structure of the RI␣ isoform (Protein Data Bank entry 1R2A, Ref. 14). The modeling was done using two different high-resolution models of the ⌬(1–91)RII␣/C heterodimer based on the crystal structures of the RI␣ cAMP binding domain and C subunit (Protein Data Bank entry 1PVK, Ref. 20, and Protein Data Bank entry 1KMU, Ref. 21). To simulate the volume occupied by the linker region, cylinders of 8, 10, 12, and 14 Å radii were generated with a volume calculated from the molecular weight of the linker sequence missing from the structures. The cylinders were filled with random points to the same density of points as used by the other structures. The structure was chained together in the sequence RC heterodimer:linker cylinder:D/D domain:linker cylinder:RC heterodimer. All of these structural components were then allowed to rotate freely with respect to each other about their connection points. The angles were chosen randomly with no constraints applied. A 2-fold axis of symmetry about the natural axis of symmetry of dimerization/docking domain was enforced for all models. Two independent runs of CONTRAST were performed per cylinder length for each of the two RC heterodimer models, each testing in excess of 250,000 possible models. RESULTS
SAXS Data from the Type I␣ and II␣ Holoenzymes—The small-angle x-ray scattering intensity profiles and associated distance distribution functions, P(r), for the type I␣ PKA holoenzyme (in H2O and in D2O) and the type II␣ PKA (in H2O) holoenzyme, which was similarly prepared for this study, are shown in Fig. 1. The P(r) functions of the type I␣ holoenzyme in H2O and D2O are identical within the error of the measurements, indicating that the overall conformation of the holoenzyme is unaltered by D2O. All of the samples were free of nonspecific aggregation and the effects of inter-particle interference, as judged by the lack of an upturn or downturn, respectively, in the low q region of the scattering intensity plot. Comparison of the forward scattering (I(0)) intensity with that from a lysozyme standard also supported this conclusion. The P(r) function for the type I␣ holoenzyme peaks at ⬃40 Å, has a strong shoulder at 112 Å, goes to zero at 150 Å, and is indicative of a bilobal object. The P(r) curve for the type II␣ holoenzyme peaks just below 40 Å with shoulders at 112 and 140 Å and extends out to ⬃210 Å, indicative of a significantly more extended, asymmetric shape. The structural parameters Rg and dmax, derived from the P(r) plots, are summarized in Table I. Two sets of parameters are given for the type II␣ holoenzyme; one is derived from samples prepared for this study, and the other one is from an earlier study (19). Although both data sets show a more extended structure than what we see for the type I␣ isoform, the earlier study found an even more extended structure than what we see here for the type II␣ samples reconstituted using the same methods employed for the type I␣ isoform. The samples used in the earlier study of the type II␣ holoenzyme were reconstituted using native, myristoylated C subunits, and a urea unfolding step was used to
FIG. 1. Small-angle x-ray scattering data from the type I␣ and type II␣ holoenzymes. Shown are small-angle x-ray scattering intensity profiles I(q) versus q (A) and the corresponding P(r) functions (B) for the type I␣ holoenzyme in H2O (f) and D2O (E) and the bacterially expressed RII␣ holoenzyme (⽧). The intensity profiles have been offset on the vertical axis for clarity. arb. units, arbitrary units.
remove cAMP from R in order to reconstitute holoenzyme, whereas this study used a recombinant, unmyristoylated C subunit and no urea unfolding step. These differences in sample preparation are the likely source of differences in the type II holoenzyme structures. However, both studies show the type II␣ holoenzyme to be significantly more extended than the type I␣ holoenzyme. Thus, our observation of the large isoform differences between the two holoenzymes is not an artifact of the sample preparation conditions. SANS Data, with Contrast Variation, from the Type I␣ Holoenzyme—The intensity profiles collected for the neutron contrast variation series on the type I␣ holoenzyme reconstituted with deuterated R are plotted in Fig. 2. The quality of the data is good as judged by the counting statistics. The 40 and 80% D2O data sets are close to the match points of the unlabeled C and deuterium-labeled R components, respectively. As a result, these two intensity profiles closely match the basic scattering functions for the R and C components within the holoenzyme complex, respectively. The basic scattering functions extracted from the full contrast series data and corresponding to the scattering functions for the deuterated and non-deuterated components in the complex, plus a cross-term, are plotted in Fig. 3A. The P(r) functions derived from the basic scattering functions corresponding to the scattering profiles of the deuterated and non-deuterated components in the holoenzyme are shown in Fig. 3B, and the corresponding structural parameters
Solution Structure of Type I Protein Kinase A
19087
TABLE I Scattering parameters Rg a
Type I␣ holoenzyme in H2O Type I␣ holoenzyme in D2O C subunit pair in holoenzyme RI␣ homodimer in holoenzyme Free RI␣ homodimerb Type II␣ holoenzyme with unmyristoylated Cc Type II␣ holoenzyme with myristoylated Cd
dmaxb
Å
Å
51.2 ⫾ 2.5 50.0 ⫾ 2.3 54.3 ⫾ 2.4 44.3 ⫾ 0.3 40.8 ⫾ 1.5 61.6 ⫾ 2.1 73.5 ⫾ 6.3
150 ⫾ 5 150 ⫾ 6 140 ⫾ 4 140 ⫾ 10 117 ⫾ 5 225 ⫾ 10 220 ⫾ 10
a Rg and dmax were determined from the P(r) as discussed under “Materials and Methods.” P(r) was determined using the method of Moore (28). The C subunit pair and R homodimer in holoenzyme values were calculated from the basic scattering functions derived from SANS. b Data taken from x-ray scattering study of the free RI␣ homodimer (22). c R and non-myristoylated C subunits are bacterially expressed, and buffer conditions are very similar to those used for the type I␣ holoenzyme d Data from Zhao et al. (19), which used native myristoylated C subunits and urea-treated R subunits.
FIG. 2. Fit of the model intensity profiles I(q) versus q to the neutron contrast series data. The fits of the SANS intensity profiles calculated from the model shown in Fig. 4A (curves) are plotted with the neutron contrast series data as follows: (f and solid line, 0% D2O; E and dashed line, 20%; Œ and dotted line, 40%; and dash-dot line, 80%; and 〫 and dash-dot-dot line, 90%. The intensity profiles have been offset on the vertical axis for clarity.
are presented in Table I along with those derived from SAXS data for the free RI␣ homodimer (22). As observed previously for the type II␣ holoenzyme (19), the P(r) curve for the C subunits in the complex shows two well separated peaks, indicating that they are not in contact with each other. The peak at ⬃25 Å resembles the P(r) expected for the isolated C subunit (19) and corresponds to the intermolecular vector lengths within each individual C subunit. The peak at ⬃112 Å corresponds to the inter-atomic distances between the two C subunits in the holoenzyme complex. The areas under the two peaks are the same, as is expected for two identical subunits separated in space. The P(r) of the R dimer in the complex has a peak at ⬃37 Å with a broad shoulder ⬃85–90 Å, indicating a bilobal structure. There is a striking difference between this shape and RI␣ dimer free in solution, which is significantly more compact as evidenced by its P(r), which has a peak at ⬃33 Å, a high shoulder at ⬃80 Å, and goes to zero at 117 Å. Thus, we see that the C subunit binding to the RI␣ homodimer results in increases in Rg by⬃10 Å (25%) and dmax by 33 Å (28%) (Table I). Taken together, the scattering parameters and the P(r) functions indicate that the RI␣ homodimer undergoes a large extension upon binding C subunits. Structural Models of the Type I␣ Holoenzyme—Using highresolution structures of the RI␣ D/D domain (14) and the RC heterodimer model (20, 21) connected to the D/D domain by cylinders to represent the linker regions, we developed models for the holoenzyme and tested possible configurations of the components, as described under “Materials and Methods,” to find the best fit model to the neutron contrast data. Because
there are no structural data available for the linker regions, we evaluated models built with these regions represented as cylinders of varying radii (8 –14 Å; see “Materials and Methods”). The use of RC models instead of the R and C subunits separately vastly reduces the number of models that must be tested computationally. There are two published models for the RC heterodimer (20, 21) that share some similarities but differ in the details of the RC interface. Both models show a similar position for cAMP-binding domain A with respect to C, but rotated by ⬃160°. As a result, in one model the cAMP-binding domain B points away from the C subunit and does not interact (21), whereas there is a small interaction between domain B and the C subunit in the other (20). At this time there is insufficient experimental data to distinguish the two models, and we therefore completed the model calculations with each of them independently. The contrast series data are equally well reproduced by the simulated intensity profiles produced using either one. An example set of simulated intensity profiles from a holoenzyme model using the RC heterodimer model described in Ref. 20 and linker cylinders with a 10 Å radius are plotted as curves in Fig. 2 with the contrast series data. The goodness-offit parameter, F, between the neutron data and the model shown in Fig. 2 is 0.58. It should be noted that even the more complex shapes of the relatively low contrast 80 and 90% D2O intensity profiles are reproduced very well, giving us confidence that the model is a good representation of the overall solution structure of the holoenzyme. The final F values for the models using either of the RC heterodimer models ranged from 0.56 to 0.64, and there was a clear trend of increasing F values with increasing cylinder length and, hence, smaller radius (e.g. average F values went from 0.56 to 0.64 for models with cylinder radii of 8 and 14 Å, respectively). The F values indicate that the models are in excellent agreement with the scattering data and that the linkers with smaller radii more accurately represent the actual scattering density of the linker regions. One main class of model structures was produced by CONTRAST (32) for each RC heterodimer model. A representative model for each of these is presented in Fig. 4. In both models, the R subunit has a somewhat flattened V-shape where the RC dimers project outward from the D/D domain, consistent with the basic scattering function of the R subunit within the holoenzyme. The separation of the centers of mass of the C subunits of both classes of model is ⬃105 Å, which agrees with the position of the second peak of the P(r) derived from the C subunit basic scattering function. Shown in Fig. 4 is a comparison of the CONTRAST-derived models of the RI␣ homodimer within the holoenzyme and free in solution as derived from SAXS data elsewhere (22). The models indicate that the cAMPbinding domains rotate outward in the plane of the V shape from the D/D domain upon binding C subunits, resulting in a
19088
Solution Structure of Type I Protein Kinase A
FIG. 3. Scattering functions for the RI␣ PKA subunits. Basic scattering functions derived from the contrast variation series data are plotted along with the SAXS data for the RI␣ homodimer (A) for comparison and the corresponding P(r) functions (B). The C subunit pair (f), the R homodimer (E), and the cross-term (‚) from the neutron-derived basic scattering functions are shown along with the SAXS data for the free RI␣ homodimer (), which has been offset for clarity. The free RI␣ homodimer SAXS data are taken from another study (22). Simplified versions of the models shown in Fig. 4 are shown on the right to help visualize which P(r) function represents which subunit within the holoenzyme. For the C subunit basic scattering function, the C subunits within the model to the right are highlighted in black, and the R dimer is in light gray. For the R homodimer basic scattering function, the color scheme is reversed.
more flattened V shape in agreement with the scattering parameters and the P(r) functions showing an extension upon binding C subunits. The precise conformation of the RC structures is somewhat speculative because there is no crystal structure of the RC heterodimer, although both of the RC models (20, 21) were derived from a wide range of experimental data. The holoenzyme models obtained using either RC model were indistinguishable in terms of the goodness-of-fit to the scattering data. This result is expected because, within the overall globular shape of the RC pair, small rotations at the RC interface or about the centers of mass of each fixed heterodimer can be compensated for by small adjustments of other features of the holoenzyme structure. Specifically, we see a difference in the angle between the linker cylinders in the two models. DISCUSSION
The present study demonstrates for the first time that a significant conformational change within the RI␣ homodimer may be critical for PKA function. We show that inhibition of C subunits is not just a simple docking process but rather in-
volves significant conformational changes within the RI␣ subunit linker region. Conversely, activation is a multi-step process involving not only local conformational changes in the cAMP-binding domains but also conformational changes in the linker region of the RI␣ subunit that impact the global structure of the R homodimer. These results help us understand the structural basis for the activation of PKA by cAMP. There are at least two sets of interactions between R and C; one set involves the cAMP binding domains of R, and another involves the pseudosubstrate/ inhibitory sequence within the R linker region. Hydrogen/deuterium exchange experiments with RI␣ have shown that cAMP binding to domain A causes a change in solvent accessibility in a helical region known to interact with C (8, 21) without any apparent change in the linker region. Likewise, SAXS data from the isolated RI␣ homodimer show no large change in global structure upon cAMP binding (22). Our most recent SAXS experiments have shown that the addition of excess cAMP to the type I␣ holoenzyme causes only partial dissociation of the R and C subunits, whereas the addition of cAMP
Solution Structure of Type I Protein Kinase A
19089
FIG. 4. RI␣ PKA models generated by CONTRAST. A, two views of a representative model generated using the R/C heterodimer model from Anand et al. (21). Within the R homodimer, the cAMP-binding A domain is turquoise/green, and the cAMP-binding B domain is green. Light gray and dark gray each represent one monomer of the dimerization/docking domain and linker region to illustrate the dimeric property of the subunit. Within the C subunits, the small lobe is orange and the large lobe is red. B, similar views of a representative model produced with the R/C heterodimer model from Tung et al. (20). The colors are the same as those used in A. C, similar views of the free RI␣ homodimer model derived elsewhere (22) to illustrate the conformational changes within the RI␣ homodimer that occur upon binding C subunits.
plus a peptide substrate results in full dissociation.2 The addition of peptide substrate alone did not result in any dissociation of the R and C subunits. Our results lead to a structural model for activation in which the binding of cAMP to the holoenzyme causes a local conformational change in the cAMP-binding domains that releases one set of RC contacts without releasing the pseudosubstrate sequence from the active site of C. In this conformational state, a substrate can compete with the pseudosubstrate for the active site of C. The current study extends this model of PKA activation by showing that a conformational change within the R linker region upon binding C alters the interdomain arrangement of the RI␣ homodimer. We now have a picture in which cAMP binding to the RI␣ holoenzyme, at least when R and C are still bound by the pseudosubstrate sequence, does not cause large conformational changes within the homodimer, allowing for a rapid activation by cAMP in a process where fast kinetics are important. The release of C from R due to the binding of substrate is associated with a large conformational change within the R homodimer. This latter conformational change will be slower than the local conformational change caused by the binding of cAMP alone, thereby ensuring that the re-association of R and C is slow 2 Vigil, D., Blumenthal, D. K., Brown, S., Taylor, S. S., and Trewhella, J. (2004) Biochemistry, in press.
enough to allow a termination of the initial cAMP signal before the reformation of the holoenzyme. The net result would be more specific control of PKA signaling. The pseudosubstrate inhibitor sequence of the RI␣ isoform was shown previously by time-resolved fluorescence anisotropy to be more ordered than expected for random coil, but still somewhat flexible (14). Our results indicate that the linker must undergo a significant conformational change upon binding C subunits to allow simultaneous contact of the cAMPbinding domains of R with the C subunits and the pseudosubstrate inhibitor sequence with the catalytic cleft of C. The backbone conformation of the pseudosubstrate sequence bound to the active site of C is very likely to be similar to that of PKI bound to C (33, 34). As the pseudosubstrate region adopts this conformation, an additional conformational change could be induced within other parts of the linker region that, in turn, affect the interdomain orientation of the homodimer. Likewise, the interactions of the cAMP-binding domains with the C subunit could also induce conformational changes within the linker region. The formation of these two interactions between R and C thus contributes to the extension of the R dimer structure observed in the holoenzyme. This extended structure could lead to the formation of binding sites for additional PKA binding partners. For example, it is interesting to consider the possibility that a particular AKAP might only bind the holoen-
19090
Solution Structure of Type I Protein Kinase A
zyme and not the free R homodimer, leading to dissociation of the R subunit from the AKAP subsequent to cAMP activation of the holoenzyme and the release of C subunits. Such a mechanism would be an efficient way to effect long term termination of a cAMP signal by dissociating both PKA subunits from a signaling complex. Our low resolution representation of the linker region as a simple cylinder does not allow us to see direct interaction of the pseudosubstrate sequence with the active site of C, but the models do show that the linkers are in close enough proximity to the C subunits for the two to interact. The models also indicate that the cAMP-binding sites are well exposed to the surrounding solvent, allowing facile access to intra-cellular cAMP. Additionally, the D/D domain is free from steric hindrances that might inhibit its binding to AKAPs. Thus, the models produced by CONTRAST both fit the scattering data and are reasonable with regard to the locations of structures within the complex required for the various known binding interactions. The models generated here from the scattering data provide templates with which to address the specific nature of the conformational changes occurring in the linker regions that results from binding of the pseudosubstrate sequence of RI␣ to the active site of C. It will be interesting to investigate further the biophysical properties of the linker region to achieve a more detailed understanding of the activation process. A high-resolution RC heterodimer crystal or NMR structure would clearly be of greatest value in further refining the holoenzyme models, and strategies to overcome the impediments to crystallization or size limitations for NMR analysis are critical. In the mean time, additional experimental information such as any distance constraints from NMR, fluorescence resonance energy transfer, or cross-linking would help in further refining these structural models. Comparison of our results for the type I␣ and type II␣ holoenzymes demonstrate that differences in the sequences of the R subunit linker regions have a profound influence on the global structure of the holoenzyme, in agreement with hydrodynamic studies of chimeric R subunits in which the linker regions of RI␣ and RII␣ were swapped.3 This large difference is much more than one would expect based on their molecular mass differences (167 kDa for type I␣ holoenzyme versus 171 kDa for type II␣, based on the amino acid sequences of the bovine proteins). A more comprehensive discussion of the structural variations among the PKA R subunit isoforms, along with the functional implications, is presented elsewhere (22). Small-angle scattering, especially with neutron contrast variation, provides important information on the overall structural properties of protein complexes in solution and the individual components within them. The extremely powerful approach of comparing a subunit structure free in solution as seen by x-ray scattering, with its conformation within a complex as revealed by neutron contrast variation, provides infor3
P. Banky and S. S. Taylor, unpublished results.
mation about conformational changes not available by other techniques. Combined with structural modeling approaches using high-resolution structures of individual domains, global structural properties and important conformational changes can be characterized, providing a molecular framework to understand structural changes important for cellular signaling. Acknowledgments—We thank Celina Juliano for purifying the C subunit and Brian MacDonald for assistance with the x-ray instrumentation. REFERENCES 1. Shabb, J. B. (2001) Chem. Rev. 101, 2381–2411 2. Kirschner, L. S., Carney, J. A., Pack, S. D., Taymans, S. E., Giatzakis, C., Cho, Y. S., Cho-Chung, Y. S., and Stratakis, C. A. (2000) Nat. Genet. 26, 89 –92 3. Laxminarayana, D., and Kammer, G. M. (2000) Int. Immunol. 12, 1521–1529 4. Ska¨ lhegg, B. S., and Tasken, K. (2000) Front. Biosci. 5, D678 –D693 5. Amieux, P. S., and McKnight, G. S. (2002) Ann. N. Y. Acad. Sci. 968, 75–92 6. Cummings, D. E., Brandon, E. P., Planas, J. V., Motamed, K., Idzerda, R. L., and McKnight, G. S. (1996) Nature 382, 622– 626 7. Gibson, R. M., Ji-Buechler, Y., and Taylor, S. S. (1997) J. Biol. Chem. 272, 16343–16350 8. Anand, G. S., Hughes, C. A., Jones, J. M., Taylor, S. S., and Komives, E. A. (2002) J. Mol. Biol. 323, 377–386 9. Canaves, J. M., and Taylor, S. S. (2002) J. Mol. Evol. 54, 17–29 10. Knighton, D. R., Zheng, J. H., Ten Eyck, L. F., Ashford, V. A., Xuong, N. H., Taylor, S. S., and Sowadski, J. M. (1991) Science 253, 407– 414 11. Su, Y., Dostmann, W. R., Herberg, F. W., Durick, K., Xuong, N.-H., Ten Eyck, L., Taylor, S. S., and Varughese, K. I. (1995) Science 269, 807– 813 12. Diller, T. C., Madhusudan, Xuong, N.-H., and Taylor, S. S. (2001) Structure 9, 73– 82 13. Newlon, M. G., Roy, M., Morikis, D., Hausken, Z. E., Coghlan, V., Scott, J. D., and Jennings P. A. (1999) Nat. Struct. Biol. 6, 222–227 14. Banky, P., Roy, M., Newlon, M. G., Morikis, D., Haste, N. M., Taylor, S. S., and Jennings, P. A. (2003) J. Mol. Biol. 330, 1117–1129 15. Li, F., Gangal, M., Jones, J. M., Deich, J., Lovett, K. E., Taylor, S. S., and Johnson, D. A. (2000) Biochemistry 39, 15626 –15632 16. Erlichman, J., Rubin, C. S., and Rosen, O. M. (1973) J. Biol. Chem. 248, 7607–7609 17. Zoller, M. J., Kerlavage, A. R., and Taylor, S. S. (1979) J. Biol. Chem. 254, 2408 –2412 18. Herberg, F. W., Dostmann, W. R. G., Zorn, M., Davis, S. J., and Taylor, S. S. (1994) Biochemistry 33, 7485–7494 19. Zhao J. K., Hoye, E., Boylan, S. Walsh, D. A., and Trewhella, J. (1998) J. Biol. Chem. 273, 30448 –30459 20. Tung, C. S., Walsh, D. A., and Trewhella, J. (2002) J. Biol. Chem. 277, 12423–12431 21. Anand, G. S., Law, D., Mandell, J. G., Snead, A. N., Tsigelny, I., Taylor, S. S., Ten Eyck, L. F., and Komives, E. A. (2003) Proc. Natl. Acad. Sci. 100, 13264 –13269 22. Vigil, D., Blumenthal, D. K., Heller, W. T., Brown, S., Canaves, J. M., Taylor, S. S., and Trewhella, J. (2004) J. Mol. Biol. 337, 1183–1194 23. Diller, T. C., Xuong, N.-H., and Taylor, S. S. (2000) Protein Expr. Purif. 20, 357–364 24. Slice, L. W., and Taylor, S. S. (1989) J. Biol. Chem. 264, 20940 –20946 25. Heidorn, D. B., and Trewhella, J. (1998) Biochemistry 27, 909 –915 26. Krigbaum, W. R., and Kugler, F. R. (1970) Biochemistry 9, 1216 –1223 27. Hammouda, B., Barker, J. G., and Krueger, S. (1996) Small Angle Neutron Scattering Manuals, National Institute of Standards and Technology, Gaithersburg, MD 28. Moore, P. B. (1980) J. Appl. Crystallogr. 13, 168 –175 29. Olah, G. A., Rokop, S. E., Wang, C.-L. A., Blechner, S. L., and Trewhella, J. (1994) Biochemistry 33, 8233– 8239 30. Ibel, K., and Stuhrmann H. B. (1975) J. Mol. Biol. 93, 255–265 31. Bevington, P. R. (1969) Data Reduction and Error Analysis for the Physical Sciences, pp. 164 –176, New York, McGraw-Hill 32. Heller, W. T., Abusamhadneh, E., Finley, N., Rosevear, P. R., and Trewhella, J. (2002) Biochemistry 41, 15654 –15663 33. Olah, G. A., Mitchell, R. D., Sosnick, T. R., Walsh, D. A., and Trewhella, J. (1993) Biochemistry 32, 3649 –3657 34. Knighton D. R., Zheng, J. H., Ten Eyck, L. F., Xuong, N. H., Taylor, S. S., and Sowadski, J. M. (1991) Science 253, 414 – 420
