Stamp      Kuriyan Lab Logo      Luke Chao     

A mechanism for tunable autoinhibition in the structure of a human Ca2+/calmodulin-dependent kinase II holoenzyme

Luke H. Chao, Margaret M. Stratton, Il-Hyung Lee, Oren S. Rosenberg, Joshua Levitz, Daniel J. Mandell, Tanja Kortemme, Jay T. Groves, Howard Schulman, and John Kuriyan

Cell 146, 732-745, September 2, 2011 (local copy)

Abstract / Video Abstract (Paperflick) / Figures from the paper / Additional Supplemental Items / Coordinates


Calcium/calmodulin-dependent kinase II (CaMKII) forms a highly conserved dodecameric assembly that is sensitive to the frequency of calcium pulse trains. Neither the structure of the dodecameric assembly nor how it regulates CaMKII are known. We present the crystal structure of an autoinhibited full-length human CaMKII holoenzyme, revealing an unexpected compact arrangement of kinase domains docked against a central hub, with the calmodulin binding sites completely inaccessible. We show that this compact docking is important for the autoinhibition of the kinase domains and for setting the calcium response of the holoenzyme. Comparison of CaMKII isoforms, which differ in the length of the linker between the kinase domain and the hub, demonstrates that these interactions can be strengthened or weakened by changes in linker length. This equilibrium between autoinhibited states provides a simple mechanism for tuning the calcium response without changes in either the hub or the kinase domains.

Video Abstract:

Figures from the paper:

Click on the small image to get a bigger one.

First figure from paper

Figure 1: CaMKII Subunit Architecture and Activation

(A) Domain architecture of an individual CaMKII subunit: the kinase domain (blue) is followed by a regulatory segment (yellow and burgundy), a variable linker region (dark blue), and the hub domain (grey). The regulatory segment is comprised of the R1 element (containing the autophosphorylation site Thr 286), the R2 element (an intramolecular clamp that docks the regulatory segment), and the calmodulin binding region, R3. The calmodulin binding footprint overlaps R3 and a portion of R2 (burgundy).

(B) Activation of subunits in the holoenzyme proceeds via regulatory segment displacement by Ca2+/CaM binding to enable access and presentation of Thr 286 for phosphorylation by other subunits.

(C) Structure of an individual CaMKII subunit in the crystallized holoenzyme. The regulatory segment (yellow) extends from the C-terminus of the kinase domain in an α helix, then dissolves and makes a tight turn to incorporate itself into the β sheet of the hub domain. See also Supplemental Figures S1 and S6.

Illustrator file, local only

First supplimental figure from paper

Figure S1: Sequence Alignment of Four CaMKII Homologues, Related to Figure 1

Four CaMKII isoforms are shown: the crystallization construct, human (Homo sapiens) α with a β7 linker (GenBank: NP_741960.1 and AAD42070.1, referred to in alignment as human β7), mouse (Mus musculus) α (GenBank: NP_037052), C. elegans (Gene ID: 177921), sea urchin (Strongylocentrotus purpuratus) (Gene ID: 373512), and hydra (Hydra magnipapillata) (Gene ID: 100200906). The phosphorylation sites are shown in red circles and the catalytic residues mutated in the crystal structure are shown in green stars. Residues at the β-clip interface and the hub domain spur are highlighted in orange and blue, respectively. The linker region (dark blue) often contains sequences specifying cellular localization (Griffith et al., 2003; Shen et al., 1998).

Illustrator file, local only

Sixth supplimental figure from paper

Figure S6: Structures of the Isolated Ca2+/calmodulin-Dependent Kinase Domains, Related to Figure 1

(A) The structure of the kinase domain of CaMKII was first determined for the C. elegans enzyme (Rosenberg et al., 2005). This structure shows that the regulatory segment inactivates the kinase domain by blocking the substrate binding channel in a pseudo-substrate manner, as first demonstrated for other calcium-dependent kinases (Goldberg et al., 1996; Hu et al., 1994; Mayans et al., 1998). The R1 element at the base of the regulatory segment, the region where Thr 286 is located, repositions helix αD so as to weaken nucleotide binding. The kinase domains form a dimer in the crystal (other kinase domain in the dimer shown as transparent cartoon), mediated by the Ca2+/calmodulin binding portion of the regulatory segment (the R2 and R3 elements), which forms an antiparallel coiled coil that masks the calmodulin recognition element (Ikura et al., 1992; Rosenberg et al., 2005). The antiparallel nature of the coiled coil dimer positions Thr 286 of each kinase domain far away from the active site of the other, minimizing the chance of autophosphorylation. Direct evidence for the role of this coiled coil dimer in CaMKII function is lacking because the kinase domains do not dimerize in the absence of the hub domain, and a dissection of the role of residues at the dimer interface is difficult because all of the residues that mediate dimerization are within the binding footprint of calmodulin.

(B) Structures of the kinase domain of human CaMKII have also been determined (Rellos et al., 2010). In these structures, the R3 element of the regulatory segment is shortened by deletion and the coiled coil seen in the C. elegans structure cannot therefore be formed.

(C) The structure of death associated protein kinase bound to calmodulin also shows a long extended helix comprising the R1, R2 and R3 elements like that seen in the C. elegans structure (de Diego et al., 2010).

(D) Structures of the C. elegans kinase domain with the calmodulin binding portion deleted entirely (the R2 and R3 elements removed) show how the portion of the regulatory segment bearing Thr 286 is presented in trans to another kinase domain for phosphorylation (Chao et al., 2010).

(E) The structure of calmodulin bound to the human kinase domain shows that Ca2+/calmodulin binding breaks the same set of interactions that are lost upon deletion of the R3 and R2 elements, and strips the regulatory segment from the kinase domain (Rellos et al., 2010). As seen in the structure of the C. elegans enzyme (panel D), this results in the presentation of Thr 286 for phosphorylation in trans.

Illustrator file, local only

Second figure from paper

Figure 2: Domain Architecture of the Dodecameric CaMKII Holoenzyme

The holoenzyme assembly comprises kinase domains tightly arranged about the central hub domain hub. Each kinase domain occupies a position between two hub domain subunits, with its active site pointed towards the center of the assembly. The arrangement forms two separate hexameric rings of kinase petals which fold against the central hub. See also Supplemental Figures S2 and S3.

Illustrator file, local only

Second supplimental figure from paper

Figure S2: Electron Density, Related to Supplemental Table and Figure 2

(A) Electron density for one subunit of CaMKII in the holoenzyme, calculated at 1.0 σ. Color scheme identical to that used in Figure 1.

(B) The density for the β clip residues is clearly visible in initial electron density maps. Unbiased electron density map (2FO-FC) of the β strand region contoured at 2.0 σ with modeled residues shown in burgundy.

(C) Electron density for the helix αC is visible in the initial maps calculated from a molecular replacement solution containing only the C-lobe of the kinase domain and hub domain.

(D) Electron density contoured at 1.5 σ for region of the kinase domain calculated from 4.0 Å data and data elliptically truncated at 3.6 Å in the a* and b* directions, and 4.0 Å in the c* direction.

Illustrator file, local only

Third figure from paper

Figure 3: Autoinhibitory Interactions

(A) The calmodulin recognition elements are completely sequestered in each autoinhibited CaMKII subunit of the holoenzyme. Schematic of an individual kinase subunit with kinase and hub domains. Right, a zoomed-in surface representation of a subunit with the regulatory segment (yellow) embraced by interactions from the kinase and association domains (cyan and grey, respectively).

(B) The β-clip interaction. The central portion of the regulatory segment recognized by calmodulin is incorporated into the hub domain β sheet as a parallel β strand. Thr 306 (mutated to valine in the crystallization construct), interacts with a hydrophobic pocket in the hub domain.

(C) The hub domain spur. The docking interaction between the kinase domain and the adjacent hub domain subunit is mediated by residues at the base of the activation loop (green) and at a spur in the hub domain. See also Supplemental Figures S3.

Illustrator file, local only

Third supplimental figure from paper

Figure S3: The trans Interaction Between Kinase Domains and Hub Domains Resembles the Docking of the Protein Kinase A Regulatory Subunit Against its Kinase Domain, and Linker Insertion Sites, Related to Figures 2 and 3

(A) The kinase domain of protein kinase A (PKA) (green cartoon representation) bound by its regulatory subunit (tan surface representation) (Kim et al., 2007). A zoom in of the region illustrates the residues involved at the docking interface (shown as sticks and dots representation), which occupy the base of the activation loop. A similar docking is observed in p21 associated protein kinase (Lei et al., 2000; Pirruccello et al., 2006).

(B) The CaMKII kinase domain B in the holoenzyme docks against a spur in hub domain subunit A. Several hydrophobic residues at the base if its activation loop are shown. The structure of the regulatory subunit of PKA is unrelated to that of the CaMKII hub domain, but the docking sites on the kinase domain are similar.

(C) The region for variable linker insertions is presented at the equatorial plane of the holoenzyme assembly. Kinase domain subunits (light and dark cyan) are arranged at positions above and below the central hub midplane in the autoinhibited structure, exposing the sites for variable linker insertions (red spheres). The loss of cooperativity when mutations that affect kinase docking are introduced into the hub suggests that neither kinase domain dimers nor the substrate capture mechanism contribute significantly to the cooperativity of activation by Ca2+/calmodulin in the short linker form. The distances between adjacent points of connection to the hub domain are 24Å and 33Å in the holoenzyme hub. The minimal linker present in the short linker construct might destabilize the coiled coil seen in the C. elegans crystal structure because the ends of the helices are ~40Å apart in the coiled coil (Rosenberg et al., 2005). It is more difficult to understand why the substrate capture mechanism does not contribute to cooperativity in the short linker holoenzyme, because the transphosphorylation reaction clearly occurs in the short linker construct. The contribution to cooperativity from substrate capture originates from the high affinity of the R1 element for the active site of the kinase domain (Chao et al., 2010). In the structure of the calmodulin-bound kinase domain, the distance between the start of the R1 element of the 'enzyme' kinase and the end of the regulatory segment of the substrate kinase is ~80Å (Rellos et al., 2010). Geometrical constraints in the short linker holoenzyme could generate an entropic penalty that reduces the apparent affinity of the R1 element for the kinase domain.

Illustrator file, local only

Fourth figure from paper

Figure 4: Human Short-Linker CaMKII Activation and Conformation

(A) Central hub mutation results in a decreased EC50 and a reduced nH. The human short-linker CaMKII shows high cooperativity and a μM EC50 value, while a human short-linker CaMKII with central hub mutation shows non-cooperative activation and a nM EC50 value. Error bars and ± terms expressed are s.e.m.

(B) Autonomous activity of the long-linker CaMKII increases with the frequency of stimuli (100 ms pulse durations). Error bars and ± terms expressed are s.e.m.

(C) Short-linker human CaMKII autonomous activity (red) compared with autonomous activity of short-linker CaMKII with a mutation at the β clip docking interaction (green) (350 ms pulse duration, at 2 and 3 μM CaM). Error bars and ± terms expressed are s.e.m.

(D) SAXS shows conversion from a compact to an extended state upon central hub mutation (green). Shape reconstructions are unchanged upon T306V mutation and addition of bosutinib (blue), and match the dimensions of the crystallized holoenzyme (transparent red envelope), when compared to the crystal structure (orange). See also Supplemental Figures S4 and S5.

Illustrator file, local only

Fifth figure from paper

Figure 5: Long-Linker CaMKII Conformation and Activity

(A) SAXS shows conversion between an extended and compact state. C. elegans CaMKII has a long linker, and displays an extended SAXS reconstruction (purple, left). Deletion of this linker results in a compact shape reconstruction (cyan) similar to that observed for human short-linker constructs (red). Human long-linker CaMKII also shows an extended state (purple, right). Previous SAXS and FRET analysis indicate that Ca2+/CaM binding further unravels the complex to an even more extended form (Rosenberg et al., 2005; Thaler et al., 2009).

(B) Human long-linker CaMKII activates cooperatively with a nM EC50 value, and these properties are unaffected by mutation of the central hub. Error bars and ± terms expressed are s.e.m.

(C) Ca2+/CaM-dependent activation of human long-linker CaMKII shows increased cooperativity when measured under conditions of molecular crowding, which is lost upon mutation of the docking interactions. See also Supplemental Figures S4 and S5. Error bars and ± terms expressed are s.e.m.

Illustrator file, local only

Fourth supplemental figure from paper

Figure S4: Ca2+/Calmodulin-Dependent Activity Assays, Related to Figure 4 and 5

(A) Mutation of the hub domain spur also abrogates cooperative activation and shifts the EC50 to lower calmodulin concentrations. Show in the center and right panel, mutation of Thr 340 or Leu 362 both reduce the Hill coefficient to ~1, and shifts the EC50 value to nM calmodulin concentrations.

(B) Human long linker CaMKII demonstrates an increased Hill coefficient dependent on the concentration of lysozyme crowding agent. Increasing the concentration of lysozyme also shifts the EC50 value to high calmodulin concentrations.

Illustrator file, local only

Fifth supplimental figure from paper

Figure S5: SAXS Data, Related to Figure 4 and 5

(A) Left panel, representative averaged and scaled solvent-subtracted scattering curves (I(Q)) for human short linker CaMKII (red) and the crystallization construct (human short linker CaMKII with mutation T306V and bosutinib) (blue). The theoretical scattering curve calculated from crystal structure using CRYSOL is shown in orange. Center insert, Guinier region of respective samples, with calculated Rg values. At right, the respective interatomic distance distribution (P(r)) plots, with Dmax values indicated.

(B) Mutation at the kinase domain central hub docking site (I321E) in the human short linker CaMKII (green) results in scattering curves more similar to that of the C. elegans long linker CaMKII (purple). Mutations at the hub domain spur show a similar effect (data not shown).

(C) Shortening of the C. elegans linker (cyan) results in scattering curves which match the human short linker CaMKII (red).

Illustrator file, local only

Sixth figure from paper

Figure 6: A Stochastic Computational Model for the CaMKII Frequency Response

(A) Scheme for simulations of CaMKII's frequency response (see Extended Experimental Procedures and Figure S7).

(B) Results for a simulation in which a popped-out subunit increases the popping-out probability of the adjacent subunit five-fold, and increases Ca2+/CaM binding to the adjacent subunit five-fold (each simulation was run for 30 seconds). Data for simulations in which cooperativity was absent are shown in Figure S7C. See also Supplemental Figures S7.

Illustrator file, local only

Seventh supplemental figure from paper

Figure S7: Results of Stochastic Simulations for the Frequency-Dependent Activation of CaMKII by Ca2+/CaM, Related to Figure 6

(A) Plots illustrating the time evolution of different subunit species (docked, calmodulin-bound and Thr 286 phosphorylated) while varying the frequency of calcium stimulation.

(B) Accumulation of Thr 286 phosphorylation upon varying the values of kpop and k-pop to change the pop-out equilibrium constant, K from 10-2 to 102, at three different calmodulin concentrations.

(C) The effect on Thr 286 phosphorylation of altering the factor χ1, a coupling factor for pop-out probability for adjacent subunits, and the χ2 factor for increased calmodulin binding probability to adjacent subunits.

Illustrator file, local only

Seventh figure from paper

Figure 7: An Equilibrium Between Compact and Open Autoinhibited States Sets the Frequency Threshold

The compact autoinhibited state is completely inaccessible to calmodulin due to docking against the central hub and incorporation of the regulatory segment into the hub domain. The compact state is in equilibrium with an extended form, where the calmodulin recognition element (shown in burgundy) is accessible. Both states are autoinhibited. Linker length alters the strength of kinase-central hub autoinhibitory interactions. Shortening the linker shifts the equilbrium towards the compact state, setting the threshold frequency to higher values. Lengthening the linker shifts the equilibrium to the extended state, setting the threshold calcium frequency to lower values.

Illustrator file, local only

Additional Supplemental Items

Table S1

Extended Experimental Procedures

Supplemental Figure References


PDB coordinates 3SOA

Coordinates for dodecameric holoenzyme