Activation-triggered subunit exchange between CaMKII holoenzymes facilitates the spread of kinase activity

  1. Margaret Stratton
  2. Il-Hyung Lee
  3. Moitrayee Bhattacharyya
  4. Sune M Christensen
  5. Luke H Chao
  6. Howard Schulman
  7. Jay T Groves  Is a corresponding author
  8. John Kuriyan  Is a corresponding author
  1. Berkeley, United States
  2. University of California, Berkeley, United States
  3. Howard Hughes Medical Institute, University of California, Berkeley, United States
  4. Allosteros Therapeutics, United States
  5. Lawrence Berkeley National Laboratory, United States
11 figures and 1 video

Figures

CaMKII architecture.

(A) The architecture of a dodecameric CaMKII holoenzyme. The inactive holoenzyme is shown as a more compact configuration. Upon activation by Ca2+/CaM, or phosphorylation of Thr 286 in the regulatory segment (purple circles), the kinase domains are extended from the hub assembly. (B) The domains of a CaMKII subunit. (C) Phosphorylation control in CaMKII. The R1 element of the regulatory segment leads into a helical R2 element that blocks the substrate binding channel of the kinase domain in the inactive form. The R3 element contains the calmodulin-recognition motif, and upon CaM binding, CaMKII is autophosphorylated at Thr 286 in the R1 element. After CaM dissociates, Thr 305 and 306 are phosphorylated if Thr 286 is already phosphorylated.

https://doi.org/10.7554/eLife.01610.003
Figure 2 with 5 supplements
Single-molecule assay for subunit exchange reveals activation-dependent subunit exchange.

(A) A representative single-molecule TIRF image, with red and green channels overlaid (left). For analysis, CaMKII holoenzymes are immobilized on glass slides via biotin/streptavidin interactions (right). (B) The rate of increase in colocalization is significantly faster at 37°C (red) compared to 25°C (blue) when Ca2+/CaM and ATP are added. At 37°C, the unactivated sample (i.e., with no addition of Ca2+/CaM and ATP) shows only a low level of exchange even at long time points (green). (C) Under activating conditions, decreasing the concentration of CaMKII from 8 µM (red) to 1 µM (blue) results in reduction of the rate of colocalization.

https://doi.org/10.7554/eLife.01610.004
Figure 2—figure supplement 1
Dynamic light scattering measurements on CaMKII.

(A) CaMKIIT286D was mixed under the same conditions as in the single-molecule experiments (ATP, 37°C), and samples were examined by dynamic light scattering (DLS) after 1 hr. All experiments were performed in duplicate. The autocorrelation function (left) calculated from the data before ATP addition (blue) is slightly different from the function calculated 1 hr after ATP addition (red). The autocorrelation functions were used to calculate the molecular size distribution in the sample (right). After 1 hr incubation with ATP (red), there exists the same major species corresponding to single holoenzymes, as seen without incubation (blue). In addition, there is a small percentage of larger particles that accumulates over time, but this is a very minor population. (B) Wild-type CaMKII was activated using Ca2+/CaM and ATP and incubated at 37°C for 1 hr. The autocorrelation curves do not change significantly over this time course (left). The major species in this solution also remains the same over time (right). The size of this species is consistent with fully extended CaMKII with CaM bound.

https://doi.org/10.7554/eLife.01610.005
Figure 2—figure supplement 2
Custom particle-tracking program.

Analysis of colocalized particles using a single-molecule particle-tracking program. Two images from the same area are shown, one illuminated with the excitation laser for Alexa 488 and the other with the excitation laser for Alexa 594. The left panel shows the raw images obtained using the two illuminations. The fitting procedure for each particle is based on a threshold subtraction and vector gradient analysis for image processing followed by two-dimensional Gaussian fitting (see ‘Materials and methods’). Particles that are too bright or not bright enough are omitted from analysis. The middle panel shows the results from the particle-tracking program, which locates each particle (red circles) and discounts particles that are not circular (purple circles) or misshaped due to edge effects (green circles), and would not fit properly to a two-dimensional Gaussian curve (double purple circles). The right panel shows the results of overlaying these two images to determine which particles are occupying the same position (red circles). These are counted as the colocalized particles, and percent colocalization is determined by taking the ratio of colocalized particles/total particles in the less populated channel.

https://doi.org/10.7554/eLife.01610.006
Figure 2—figure supplement 3
Intensity distribution analysis of single-molecule images.

The intensity distribution of each image over the course of an isolated experiment reveals a population of particles that have similar brightness, which is exemplified by a single population distribution on the histogram. Importantly, this distribution does not change over time. Data are shown for wild-type CaMKII activated by Ca2+/CaM and ATP (A) and CaMKIIT286D mixed with ATP (B).

https://doi.org/10.7554/eLife.01610.007
Figure 2—figure supplement 4
Comparison of activation methods and subunit exchange.

The extent of colocalization is calculated for two separate samples of wild-type CaMKII at 25°C. In one sample, CaMKII separately labeled with red or green dye was mixed together and then activated by Ca2+/CaM and ATP during the mixing reaction (purple). In the second sample, red CaMKII and green CaMKII were separately activated by Ca2+/CaM and ATP for 10 min at 25°C and then mixed together (red). Samples were mixed at 25°C for 1 hr and then analyzed for colocalization. It is clear that pre-activating the samples does not affect the final level of colocalization.

https://doi.org/10.7554/eLife.01610.008
Figure 2—figure supplement 5
FRET mixing experiments corroborate single-molecule results.

(A) The crystal structure of the dodecameric hub assembly of human CaMKIIγ (PDB code: 2UX0). The sites of fluorophore labeling are shown as purple spheres (distances: a ∼33 Å [bright red subunit], b ∼58 Å [light red subunit]). (B) Two separate CaMKII samples, one labeled with Alexa 488 (green) and the other labeled with Alexa 594 (red), are mixed under various conditions and the FRET signal is measured, as defined by the ratio of the acceptor fluorescence to the donor fluorescence. (C) The extent of subunit exchange is monitored by FRET. At each time point, a fluorescence scan is taken with excitation at the donor wavelength. CaMKII is activated by Ca2+/CaM and ATP in one mixed sample (blue squares) while the other mixed sample is unactivated (red circles).

https://doi.org/10.7554/eLife.01610.009
Analysis of the exchange process.

(A) Single-molecule experiments show that fusion of CaMKII to a hexameric protein (Hcp1) slows the rate of colocalization. All samples are activated with Ca2+/CaM and ATP and mixing is done at 37°C. Mixing activated wild-type CaMKII yields about 70% of maximal colocalization (blue). Mixing wild-type and CaMKII-Hcp1 shows a marked decrease in colocalization (red). Mixing CaMKII-Hcp1 species results in nearly no colocalization (green). (B) The isolated hub assembly does not result in colocalization when labeled subunits are mixed. (C) Deletion of the variable linker region does not affect colocalization significantly. Comparison of the short-linker construct (red), short-linker construct mutated at the hub–kinase interface (green), and wild-type CaMKII (blue) shows minimal differences in colocalization.

https://doi.org/10.7554/eLife.01610.010
Figure 4 with 1 supplement
Phosphorylation of the calmodulin-recognition element is crucial for exchange.

(A) Mixing CaMKIIT286D in the absence of Ca2+/CaM results in robust colocalization (red). Mutating the CaM-recognition element (R3) reduces colocalization significantly (blue). (B) Both mixing experiments shown use wild-type CaMKII in the presence of Ca2+/CaM. The addition of a kinase inhibitor and 1 µM ATP significantly reduces colocalization (green) compared to a condition with full kinase activity (250 µM ATP, no kinase inhibitor) (red). (C) All species are mixed with CaMKIIT286D in the absence of Ca2+/CaM. Compared to the colocalization resulting from mixing CaMKIIT286D (red), mixing CaMKIIT286D in the presence of a kinase inhibitor results in reduced colocalization (blue). Replacement of either Thr 305 or Thr 306 by alanine also results in a reduction in colocalization (pink and green, respectively).

https://doi.org/10.7554/eLife.01610.011
Figure 4—figure supplement 1
Kinase activity is crucial for exchange.

Solution FRET experiments show that the addition of the kinase inhibitor, bosutinib, in the presence of Ca2+/CaM with no ATP significantly reduces the FRET signal (green squares) compared to activated CaMKII (red circles).

https://doi.org/10.7554/eLife.01610.012
Figure 5 with 1 supplement
Evidence for the spread of phosphorylation into unactivated holoenzymes.

(A) Both mixing experiments shown use wild-type CaMKII in the presence of Ca2+/CaM and ATP. CaMKIIT286D mixed with unactivated wild-type CaMKII results in high colocalization (red). This colocalization is suppressed by the addition of the Hcp1 module (CaMKII-Hcp1; blue). (B) Levels of pThr286 labeling in each mixing experiment from (A). The pThr286 antibody is modified with Alexa 647, which is then added to the mixed samples. Subsequent analysis is for 3-color colocalization between Alexa dyes 488, 594, and 647. The phosphorylation spreads significantly more in the CaMKIIT286D sample (red) compared to the sample mixed with CaMKIIT286D-Hcp1 (blue). (C) There is colocalization of the pThr286 antibody with particles that have both Alexa 488 and 594, indicating that the antibody is only binding to those CaMKII holoenzymes that have already exchanged subunits. The graph shows the fraction of antibody label that is colocalized to particles that contain both the red and green fluorophores. Note that this fraction is close to 100%. (D) Kinase activity against a peptide substrate (syntide) was monitored in solution using the ADP Quest assay, where the fluorescence of resorufin is an indicator of ATP consumption. The reaction rates are plotted for three separate samples. First, 3 µl of an unactivated CaMKII sample, then 1 µl of an activated CaMKII sample (both are at the same final protein concentration), and finally a mixture of these components. The value of the reaction rate is indicated above each bar. It is clear that the reaction rate in the mixture is higher than just the addition of the rates of the individual components.

https://doi.org/10.7554/eLife.01610.013
Figure 5—figure supplement 1
Controls for the pThr286 antibody.

Labeling activated wild-type CaMKII (in which Thr 286 is expected to be phosphorylated) with the pThr286 antibody results in significant colocalization of the antibody with CaMKII (∼78%). Labeling CaMKIIT286D with the antibody yields very little signal corresponding to the antibody label, within the noise of the experiment (<4%).

https://doi.org/10.7554/eLife.01610.014
Schematic of a potential mechanism for subunit exchange in CaMKII.

(A) In the unactivated state of CaMKII, the dodecameric hub domain undergoes fluctuations, which leads to transient lateral openings between vertical dimeric units. (B) Upon Ca2+/CaM binding, Thr 286 is trans-phosphorylated. For simplicity, just one kinase domain is depicted. When calcium levels drop, CaM falls off and Thr 305 and Thr 306 are subsequently phosphorylated. The now-released regulatory segment is free to bind the open β sheet of its own hub domain. This brings residues 305/306 in close proximity to the hub cavity (blue triangle), which houses the three conserved Arg residues. Binding of the regulatory segment induces a crack to open in the hub domain, which exposes the Arg residues to the phosphate groups. We reason that phosphorylation of the regulatory segment leads to an interaction between the R3 element and the hub domain that weakens the lateral association between hub domain dimers, leading to their release from one holoenzyme. (C) Fluctuations in the autoinhibited holoenzyme create a hub assembly that resembles a ‘C’ shaped structure. This fluctuation may allow the capture of a vertical dimer that has been released from an active holoenzyme. The drawing depicts the adoption of a tetradecameric structure upon docking of an incoming vertical dimer, which may be an intermediate in the exchange process. As discussed in the main text, we have no direct experimental evidence at present concerning the exchange process. We cannot, therefore, rule out alternative mechanisms, such as those involving a transient aggregation of holoenzymes prior to exchange.

https://doi.org/10.7554/eLife.01610.015
Figure 7 with 1 supplement
A vertical dimeric unit of the CaMKII assembly may be the unit of exchange.

(A) The CaMKII hub assembly can be described as a set of six vertical dimers, and each dimer is labeled as A/B; C/D; etc. One of these dimers (A/B) is highlighted in blue/magenta (black dashed line). The lateral interfaces and equatorial interface for this dimer are indicated by orange and black lines, respectively. (B) A schematic diagram that indicates how one vertical dimer may be released from the holoenzyme. (C) The structure of the vertical hub dimer from CaMKII is shown in comparison to a dimer of NTF2 (PDB codes: 2UX0 and 1OUN, respectively). The notation for the secondary structural elements of the CaMKII hub domain are shown. (D) A molecular dynamics simulation was started from the dodecameric crystal structure of the hub domain (PDB code: 2UX0). The starting crystal structure is overlaid with an instantaneous structure from the molecular dynamics trajectory at 100 ns, aligning onto just one subunit (indicated on the schematic). It is clear that the vertical dimeric unit is relatively stable (blue/magenta), but there is a significant change in the relative positioning of the blue/magenta dimer with respect to the yellow/green dimer. This indicates that the lateral interfaces are more dynamic than the equatorial interfaces.

https://doi.org/10.7554/eLife.01610.016
Figure 7—figure supplement 1
Molecular dynamics simulations suggest that the contacts across equatorial interfaces are stronger than those across the lateral interfaces.

Molecular dynamics simulations provide a clear indication that the lateral interfaces in the hub assembly are more likely to be disrupted than the equatorial ones. We generated a 100 nanosecond (ns) trajectory of the hub domain of human CaMKIIγ (PDB code: 2UX0). We aligned each instantaneous structure from the trajectory onto the initial structure by using each subunit, one at a time, and plotted the root mean square deviations in the positions of the Cα atoms in the two subunits across the lateral interfaces (red, blue) and the one subunit across the equatorial interface (green). For comparisons across the equatorial interfaces, the rms displacements in the neighboring subunits are ∼2 Å across the ring. In contrast, for comparison across the lateral interfaces, the rms displacements of the neighboring subunits have shifted by as much as 6 to 10 Å for several of the interfaces, consistent with the idea that the lateral interfaces are more flexible than the equatorial ones.

https://doi.org/10.7554/eLife.01610.017
Strain associated with the closed ring formed by the dodecameric hub assembly of CaMKII.

During the simulation of the CaMKIIγ dodecamer, the sixfold symmetry of the hub assembly breaks down due to the tightening of some of the lateral interfaces and loosening at others. (A) A view of one of the lateral interfaces, with a close-up view in (B). At this interface, the residues on helix αD in one subunit and the β4 and β5 strands on the other are splayed apart after 100 ns of simulation, so that the sidechains of Phe 364 and Phe 367, which are buried in the crystal structure (left) and more stable interfaces, are now partially solvent exposed (right). These results suggest that the constraint of ring closure prevents the simultaneous optimization of all of the lateral interfaces.

https://doi.org/10.7554/eLife.01610.018
Figure 9 with 2 supplements
Molecular dynamics simulations of an open-ringed (decameric) hub assembly.

(A) The dodecamer consists of six vertical dimers, denoted A:B, C:D…K:L. The decamer used in the simulations is created by removing the A:B dimer. (B) We initiated two independent molecular dynamics trajectories from this decameric structure and the results for one are shown in this diagram. Instantaneous structures from this simulation are shown overlaid with the crystal structure for either the dodecameric (PDB code: 2UX0) or tetradecameric (PDB code: 1HKX) hub assembly. At 29 ns, it is clear that the decamer has relaxed to the tetradecameric conformation, with further opening evident at 91 ns. (C) There are two internal vertical interfaces in the decamer, between the E:F and G:H vertical dimers and between the G:H and I:J vertical dimers (colored in the structural diagram shown at the top). To demonstrate the convergence of the two vertical interfaces to an arrangement that is distinct from the interfaces in the crystal structure, we calculated the displacement of atomic positions in interfacial subunits after the E:F/G:H and G:H/I:J interfaces were brought into spatial alignment using only one subunit, for a series of instantaneous structures extracted from the trajectories. To do this, we made a copy of the trajectory. The subunits in the original are labeled C through K, and in the copy they are labeled C′ through K′. The two internal interfaces (E:F/G:H and G:H/I:J) are brought into alignment by superimposing subunit E from the original onto subunit G′ of the copy of the trajectory, for pairs of structures at the same point in the trajectory. The overlaid structures are shown at the bottom left. The close overlap shows that the two vertical interfaces in this instantaneous structure are similar. We then aligned subunit E of the instantaneous structure with subunit G of the crystal structure (bottom right). The poor overlap between the crystal structure and the instantaneous structure is evident.

https://doi.org/10.7554/eLife.01610.019
Figure 9—figure supplement 1
Relieving strain in the dodecameric ring tends towards a tetradecameric conformation.

(A) A schematic is shown for the calculation of the angle of opening within hub assemblies. This inter-subunit angle was calculated over the course of the trajectory for the decameric CaMKII as defined by the center of mass of one subunit at the apex and between the centers of mass of two subunits across the ring. (B) The value of the inter-subunit angle is graphed over the time course of the trajectory. In a short amount of time (∼10 ns), the value of the inter-subunit angle for the decameric CaMKII changes from that corresponding to a dodecamer to a value closer to that for a tetradecamer, and does not extend significantly beyond this point.

https://doi.org/10.7554/eLife.01610.020
Figure 9—figure supplement 2
Structural changes during molecular dynamics, showing the relaxation of the decamer to a specific interfacial arrangement.

(A) Two vertical interfaces within one instantaneous structure were aligned as described in the legend for Figure 9C. After the alignment, the rms displacement in Cα positions between the E subunit in the original and the G′ subunit in the copy is calculated, and the process is repeated for each time point along the trajectory (top). Relative convergence of orientations is evident from the low deviations in Cα positions between the E and G′ subunits (2–3 Å) towards the end of the simulation. We also calculated the deviations in Cα positions across two replicate trajectories for pairs of structures at the same time point, revealing a similar trend (bottom). (B) When the trajectory is compared in the same way to the crystal structure of the dodecameric hub assembly, the rms displacements in Cα positions are calculated between the E subunit from the trajectory and G subunit from the crystal structure. These values are significantly higher (∼6 Å) towards the end of the trajectory.

https://doi.org/10.7554/eLife.01610.021
Side entrance to the hub cavity and docking of the regulatory segment onto the hub.

(A) One vertical dimer unit has been removed from the side view of the decameric hub assembly (shown as a schematic on the left). There is limited access to this cavity in the context of the holoenzyme. A close up of the G/H dimer from the crystal structure of CaMKIIγ (PDB code: 2UX0) is highlighted in blue/pink (right). There are three arginine residues that are located deep within the hub cavity (residues 403, 423, and 439), and these are highly conserved in CaMKII. (B) Shown is a vertical dimer from the mouse CaMKIIα crystal structure (PDB code: 1HKX). The crystal structure is shown in cartoon representation (left) with the regulatory segment (magenta) docked onto β3 (yellow). On the right, the crystal structure is shown as a surface representation where the hub cavity that contains the arginine residues (blue) is apparent through a small crack (between αA and β3), which we refer to as the lateral opening. (C) The regulatory segment (magenta) was docked onto β3 in the hub assembly (yellow). This docked model was used to generate three independent molecular dynamics trajectories (100 ns each). In each of these trajectories the R3 element retains the hydrogen bonding pattern (right) that is consistent with its incorporation into the β sheet of the hub domain and the phosphate group maintains its proximity to the sidechain of Arg 403, suggesting that this interaction is plausible. Each trajectory was sampled every 2 ns and each dot represents the formation of a hydrogen bond (<3.5 Å). (D) Peptide binding disrupts lateral contacts within the hub domain. Docking of the regulatory segment peptide (magenta) prevents the crack between αA (yellow) and β3 (yellow) from closing. This is coupled to changes in the loops that make lateral contacts with the adjacent vertical dimers. This disruption may be sufficient to allow the release of a vertical dimer unit from the hub assembly and facilitate the exchange of subunits between holoenzymes.

https://doi.org/10.7554/eLife.01610.022
Author response image 1

Author response image 1: Sample traces

Videos

Video 1
Lateral opening in the hub domain.

Video of a 1 μs molecular dynamics simulation of a vertical hub domain dimer that was extracted from the crystal structure of mouse CaMKIIα (PDB code: 1HKX). This simulation revealed transient fluctuations in the hub domain structure that opened the interface between helix αA and the underlying β sheet substantially. These fluctuations provide access to the hub cavity.

https://doi.org/10.7554/eLife.01610.023

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Margaret Stratton
  2. Il-Hyung Lee
  3. Moitrayee Bhattacharyya
  4. Sune M Christensen
  5. Luke H Chao
  6. Howard Schulman
  7. Jay T Groves
  8. John Kuriyan
(2014)
Activation-triggered subunit exchange between CaMKII holoenzymes facilitates the spread of kinase activity
eLife 3:e01610.
https://doi.org/10.7554/eLife.01610