Structure of the Autoinhibited Kinase Domain of CaMKII and SAXS Analysis of the Holoenzyme
Oren S. Rosenberg, Sebastian Deindl, Rou-Jia Sung, Angus C. Nairn and John Kuriyan
Summary: Ca2+/calmodulin-dependent protein kinase-II (CaMKII) is unique among protein kinases for its dodecameric assembly and its complex response to Ca2+. The crystal structure of the autoinhibited kinase domain of CaMKII, determined at 1.8 Å resolution, reveals an unexpected dimeric organization in which the calmodulin-responsive regulatory segments form a coiled-coil strut that blocks peptide and ATP binding to the otherwise intrinsically active kinase domains. A threonine residue in the regulatory segment, which when phosphorylated renders CaMKII calmodulin independent, is held apart from the catalytic sites by the organization of the dimer. This ensures a strict Ca2+ dependence for initial activation. The structure of the kinase dimer, when combined with small-angle X-ray scattering data for the holoenzyme, suggests that inactive CaMKII forms tightly packed autoinhibited assemblies that convert upon activation into clusters of loosely tethered and independent kinase domains.
A. The domain organization of CaMKII.
B. Two views of the CaMKII dimer, 90° rotated from each other. Molecule A is shown in green and molecule B is shown in blue. The Ca2+/CaM binding residues are shown in magenta and the rest of the regulatory segment is shown in orange. C. Detailed view of the coiled-coil region of the dimer. Shown in spheres are the residues at the interface in a “knobs and holes” packing. Residues at the interface of the coiled coil highlighted in orange. These include the regulatory residue Thr 306 (Colbran, 1993). Also shown is the homologous region of the Drosophila protein Camguk that binds to the residues involved in the coiled coil of CaMKII (Lu et al., 2003) suggesting that other proteins may exploit these residues to form similar coiled-coil structures.
Figure 2: Details of the Interaction between the Kinase Domain and the Regulatory Segment
A. Detailed view of the interactions between the regulatory segment and the main body of the kinase domains. In the top panel, the substrate binding residues Glu 96 and Glu 139 (Yang et al., 1999) are shown interacting with Arg 297 from the regulatory segment. The bottom panel shows the regulatory Thr 286 sitting at the C-terminal end of helix αF. Mutation of residues Phe 98 in helix αD and Phe 292 in the regulatory segment activate the kinase strongly (Yang et al., 1999).
B. Comparison of three Ca2+/CaM-activated kinases, CaMKI (Goldberg et al., 1996), twitchin kinase (Hu et al., 1994), and CaMKII. In CaMKII, neither the N lobe nor the ATP binding pocket make any contact with the regulatory segment.
Fiugre 3: An Allosteric Mechanism Affecting the ATP Binding Site
A. A structural alignment of phosphorylase kinase (PhosK, bound to ATP; Lowe et al., 1997) and CaMKII showing the rotation and translation of helix αD and the interactions of His 282. Note the position of Glu 96 in CaMKII, in a location too distant to interact with ATP. All residue numbering refers to CaMKII.
B. and C. After ∼15 ns of molecular-dynamics simulation without the regulatory segment, helix αD rotates spontaneously into the position seen in PhosK.
B. An overlay of the crystal structure of CaMKII (inactive) with an instantaneous structure, at 15.125 ns, from the simulation without the regulatory segment.
C. The same 15.125 ns instantaneous structure, at 15.125 ns, overlaid with PhosK (active).
Figure 4: The Coiled-Coil Strut Is Part of a Mechanism Preventing Autophosphorylation
This schematic diagram illustrates the consequences of structural fluctuations that release either the kinase domains or the coiled-coil interaction.
If kinase domains prematurely release from the regulatory segment in the absence of Ca2+/CaM as shown in (A), the coiled-coil strut will prevent the active site from interacting with the regulatory Thr 286. If, as shown in (B), the coiled-coil interaction breaks, the interactions with the regulatory segment will also prevent accidental phosphorylation of Thr 286.
Figure 5: Rigid-Body Modeling Scheme for SAXS Analysis
Six autoinhibited kinase-domain dimers were placed around a central association domain as described in the text. The kinases were separated from one another in steps of 5 Å (this variable was defined as “kinase displacement”) and the theoretical solution X-ray scattering was calculated for each resulting model.
A. A value of kinase displacement equal to 0 Å, corresponding to a model where the autoinhibited kinase domains are in the dimer seen in the crystal structure. The resulting model is a disc with a diameter of ∼200 Å and a height of ∼60 Å.
B. A value of kinase displacement equal to 100 Å resulting in a model with diameter of ∼200 Å and a height of ∼200 Å.
C. As the value of kinase displacement increases, the calculated Rg diverges sharply from the experimental value of ∼72 Å.
D. The rings of autoinhibited kinase domains were moved above and below the association domains while simultaneously shrinking the radius of the kinase rings to generate a series of models with essentially equivalent Rg values (matched to the experimental value). The different models can all be described by a single variable, θ, that represents the angle between the midplane of the association domain and a vector connecting the center of an association-domain dimer and the center of mass of the autoinhibited kinase domain.
E. Theoretical X-ray scattering curves were calculated for models corresponding to different values of θ. At a value of θ equal to 0, corresponding to a model in which the individual autoinhibited kinase domains are held together as dimer pairs in an outer ring that is coplanar with the central plane, the curve approximates the actual data much better. This is quantified in (F) where the χ2 statistic calculated by the program Crysol (Svergun et al., 1995), which compares the calculated scattering curve to the experimental one, is optimal at θ = 0.
Figure 6: Ab-Initio Shape Restorations using SAXS from CaMKII
A. The holoenzyme model that resulted from the rigid body modeling described above. In green are the more electron-dense C lobes of the kinases and in beige are the N lobes of the kinases.
B. Shown in blue mesh is the SAXS shape reconstruction. The model shown in (A) was fit into the SAXS shape reconstruction as described in the Experimental Procedures.
Figure 7:Speculation Regarding the Activation of CaMKII by Ca2+/CaM
A. A schematic diagram of the holoenzyme in its fully autoinhibited form.
B. Ca2+/CaM binds to a regulatory segment of a kinase dimer, releasing the other regulatory segment to bind a second Ca2+/CaM. Release of the dimer pair reveals a basic patch, shown inset with the electrostatic surface overlaid on the structure. This basic patch may then interact with an acidic region on the linker, directing the kinase domains above and below the midplane of the association domain.
C. When an adjacent autoinhibited kinase dimer binds to two Ca2+/CaM, the kinase domains will again be directed above and below the midplane. They will now encounter the previously activated kinase domains, and transphosphorylation of Thr 286 can occur.
D. The addition of Ca2+/CaM to CaMKII causes a change in the shape of the solution SAXS curve. The intensity of the curve without Ca2+/CaM was multiplied by 1.7 so as to overlay the graphs. The Guinier region, shown as an inset, is linear for both curves in a region of s∗Rg = .9 − 1.3. By fitting the Guinier equation (with the program PRIMUS), the value of Rg was determined to be 71.8 ± 0.2 without Ca2+/CaM and 97.5 ± 0.4 when Ca2+/CaM is added. The presence of the linear Guinier region in the s∗Rg region < 1.3 indicates that the samples are not aggregated.
E. The P(r) function, as calculated by the program GNOM (Svergun, 1992), for CaMKII with and without the addition of Ca2+/CaM.
PDB Coordinates: 2BDW.pdb