Oligomerization states of the association domain and the holoenyzme of Ca 2+ ⁄ CaM kinase II Oren S. Rosenberg 1,3 , Sebastian Deindl 1 , Luis R. Comolli 2 , Andre ´ Hoelz 5 , Kenneth H. Downing 2 , Angus C. Nairn 4 and John Kuriyan 1,2 1 Department of Molecular and Cell Biology and, Department of Chemistry and, Howard Hughes Medical Institute, University of California, Berkeley, CA, USA 2 Life Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA, USA 3 Department of Cell Biology and 4 Department of Psychiatry, Yale University School of Medicine, New Haven, CT, USA 5 The Rockefeller University, New York, NY, USA Many cellular processes are modulated by fluctuations in the cytosolic concentration of calcium ions (Ca 2+ ) [1]. Ca 2+ ⁄ calmodulin (Ca 2+ ⁄ CaM) activated protein kinases (CaMKs) are among the most important intra- cellular transducers of Ca 2+ signals and the multifunc- tional Ca 2+ ⁄ calmodulin activated kinase II (CaMKII) is one of the most abundant kinases of this class [2]. CaMKII is highly conserved throughout the animal kingdom [3] and is found in virtually all mammalian cell types, where it phosphorylates a large array of different substrates, including itself [4]. It has a particularly important and well studied role in the response of neurons and myocytes to Ca 2+ transients (reviewed in [5] and [6], respectively) In these cells, CaMKII has been shown to be important in complex physiological processes such as the generation of long- term potentiation and the regulation of the heartbeat. There are four mammalian isoforms of CaMKII and many different splice variants, but all CaMKII proteins share the same basic architecture (Fig. 1). All of the isoforms assemble into multimeric holoenzymes. Each polypeptide chain in the holoenzyme contains a Keywords association domain; Ca 2+ ⁄ calmodulin dependent protein kinase II; holoenzyme; kinase activation; oligomerization Correspondence J. Kuriyan, University of California, Berkeley, Barker Hall MC 3202, Berkeley, CA 94720- 3202, USA Fax ⁄ Tel: +1 510 643 0137 Fax: +1 510 643 2352 E-mail: kuriyan@berkeley.edu Website: http://jkweb.berkeley.edu (Received 6 September 2005, revised 27 November 2005, accepted 5 December 2005) doi:10.1111/j.1742-4658.2005.05088.x Ca 2+ ⁄ calmodulin activated protein kinase II (CaMKII) is an oligomeric protein kinase with a unique holoenyzme architecture. The subunits of CaMKII are bound together into the holoenzyme by the association domain, a C-terminal region of  140 residues in the CaMKII polypeptide. Single particle analyses of electron micrographs have suggested previously that the holoenyzme forms a dodecamer that contains two stacked 6-fold symmetric rings. In contrast, a recent crystal structure of the isolated associ- ation domain of mouse CaMKIIa has revealed a tetradecameric assembly with two stacked 7-fold symmetric rings. In this study, we have determined the crystal structure of the Caenorhabditis elegans CaMKII association domain and it too forms a tetradecamer. We also show by electron micro- scopy that in its fully assembled form the CaMKII holoenzyme is a dode- camer but without the kinase domains, either from expression of the isolated association domain in bacteria or following their removal by pro- teolysis, the association domains form a tetradecamer. We speculate that the holoenzyme is held in its 6-fold symmetric state by the interactions of the N-terminal  1–335 residues and that the removal of this region allows the association domain to convert into a more stable 7-fold symmetric form. Abbreviations Ca 2+ , calcium; CaMK, Ca 2+ ⁄ calmodulin dependent protein kinase; CaMKII, Ca 2+ ⁄ calmodulin dependent protein kinase II; PEG, polyethylene glycol; TCEP, tri(2-carboxyethyl)phosphine hydrochloride. 682 FEBS Journal 273 (2006) 682–694 ª 2006 The Authors Journal compilation ª 2006 FEBS kinase domain (residues 1–280 in the mouse CaMKIIa numbering is used throughout this article) followed by a regulatory segment (residues 281 to  316) that binds to the kinase domain and inhibits its activity. A linker region follows the regulatory segment, the length of which is variable and depends on the isoform or splice variant. The C-terminal segment of the polypeptide contains the so-called association domain (residues 345–478) that is responsible for oligomerization. Previ- ous work has suggested that the association domains form a ring at the center of the holoenzyme with the kinase domains surrounding the central ring like spokes on a wheel [7–10]. The number of subunits in the CaMKII holo- enyzme has been assessed by many groups using a number of different techniques, and has been estima- ted to be from 4 to 14 [8,9,11–19] but with the con- sensus being that the holoenzyme is a dodecamer. The crystal structure of the isolated association domain (residues 335–478 of the mouse CaMKIIa) has been solved, and found to be a tetradecamer [10]. This crystallization construct is also a tetrade- camer in solution [10]. In the structure, the 14 association domain protomers assemble into two 7- fold symmetric rings. These rings face each other, creating seven 2-fold axes of symmetry that are per- pendicular to the 7-fold axis. Single particle analysis of electron micrographs suggest that CaMKII forms dodecamers in its fully assembled state [9,11,13] and the reason for, and significance of, this discrepancy between the oligomerization state of the holoenzyme and that of the crystal structure of the association domain is unknown. Although the association domain is necessary and sufficient for formation of oligomeric CaMKII [18,19], it is not known if other parts of the protein play a role in determining the final functional form of the CaMKII holoenzyme. It has been hypothesized that a central function of the CaMKII holoenzyme is to respond, via an increase in persistent kinase activity, to the frequency of incoming Ca 2+ spikes [20,21]. This remarkable property relies on five features of CaMKII. First, its oligomeric structure brings the kinase domains into a defined special rela- tionship with one another [10]. Second, the event that triggers persistent activity of CaMKII is a trans-auto- phosphorylation, where one kinase subunit phosphory- lates another kinase subunit (on Thr286 in the auto-regulatory segment) [22–26]. Third, this auto-phos- phorylation happens only between the subunits within a single holoenzyme, not between separate holoenzymes [27]. Fourth, the trans-auto-phosphorylation event can only occur when both the phosphorylating subunit and the subunit that is the target of the phosphorylation are both bound to CaM [28] and, presumably, only if they can reach each other. Finally, once the trans-auto-phos- phorylation has taken place, CaM becomes ‘trapped’ by the phosphorylated subunit due to an increase of  13 000-fold in the affinity for CaM [29]. The details of the physical distances between the different subunits in the assembled holoenzyme are a crucial determinant of what defines a ‘neighboring’ subunit in the scheme described above. In addition, interactions of the kinase domains with each other around the ring might affect the rate and dwell time of CaM binding. Defining the predominant oligomeric state of the holoenzyme and thus better understanding the molecular dimensions and architecture of the holo- enyzme is a critical step towards a complete molecular characterization of this complex enzyme and its func- tional properties. In this paper we compare the oligomeric states of the isolated association domains of CaMKII from mouse and Caenorhabditis elegans with that of the holoenzyme using hydrodynamic techniques, electron microscopy, and X-ray crystallography. We find that the assembled holoenyzmes are dodecameric, as noted previously by others [8,9,13]. When the kinase domains are removed, either by deleting them from the expres- sion construct or by proteolysis of the holoenzyme, the association domains assemble into a tetradecameric form that we have now visualized for the C. elegans protein by X-ray crystallography as well as for the mammalian enzyme. We hypothesize that additional interactions of the N-termini of the subunits constrain the oligomerization state of the holoenzyme and that when these additional constraints are removed the isolated association domains transform into a 7-fold state. Fig. 1. The domain structure of the CaMKII proteins. All isoforms have the same basic architecture although the different isoforms have variable insertions of between 21 and 178 residues in the lin- ker between the kinase and the association domain. O. S. Rosenberg et al. Oligomerization of CaMKII FEBS Journal 273 (2006) 682–694 ª 2006 The Authors Journal compilation ª 2006 FEBS 683 Results Static light scattering We expressed and purified full-length mouse CaMKIIa and used static light scattering to estimate the number of subunits in the assembly. This estimate should be independent of the shape of the protein complex [30]. The value of the molecular mass given by the analysis of the scattering data is 689 000 kDa (error 1%) (poly- dispersity M w ⁄ M n ¼ 1.002 where M w is the weight- averaged molar mass and M n is the number-averaged molar mass), suggesting 12.6 subunits per holoenzyme (Fig. 2). Values of the molecular mass obtained previ- ously from a similar experiment with the isolated association domain of mouse CaMKIIa indicated a subunit stoichiometry of 14.8, which is consistent with the tetradecameric crystal structure of the association domain [10]. The holoenzyme is 6-fold symmetric in negatively stained samples We examined the full-length mouse CaMKIIa in uranyl acetate stained samples. We picked 2673 particles, aligned them and classified them into 75 classes. All of the particles appeared to be in the same orientation on the grid and thus all of the class averages are very sim- ilar in appearance (Fig. 3Ai,Aii). These class averaged images show a strong inner ring of density with an outer radius of  6 nm and a much weaker outer ring with an outer radius of  12 nm. The first four eigen images show a strong 6-fold modulation (Fig. 3B). As previ- ously done in other electron microscopic analyses of CaMKIIa we interpret the  6 nm ring of density to be due to the association domains [7,9]. The outer ring at  12 nm does not appear to be sufficiently dense to con- tain the kinase domains. This apparent weakening of the density was noted previously by Kanaseki and coworkers, who saw that in uranyl acetate stained sam- ples the central ring of the presumed association domain is easily visualized, but the presumed peripheral kinase domains, which they observed by other electron micros- copy techniques, are not seen [7]. The kinase domains form a second ring around the ring of the association domains In order to attempt to more clearly define the position of the kinase domains in relation to the association domain we examined the holoenzyme embedded in vi- trous ice. We analyzed holoenzymes from two different species: mouse CaMKIIa discussed above and C. ele- gans CaMKII (UNC-43, splice variant K11E8.d). These proteins are 69% identical in sequence, with no gaps lar- ger than four amino acids in the alignment. We picked 3865 particles from images of the Mus musculus protein and 1859 particles from the C. elegans set. Initial classi- fications suggested that, as with the uranyl acetate stained samples, the ice embedded images also showed a preponderance of a single view along the 6-fold axis which we attribute to a nonrandom orientation of the holoenyzmes during electron microscopic grid prepar- ation. We thus classified the images from each sample into five large classes. Representative classes show a clear inner ring of density at radius  6 nm and an outer ring at radius  12 nm (Fig. 4A,B). We interpret this outer ring of density to be the kinase domains. In these images it is possible to see that the association domains and the kinase domains are made up of individual A B Fig. 2. Static light scattering analysis of M. musculus CaMKII holo- enyzme. (A) Laser light scattered in a single direction from the eluant of a gel filtration column. Signal was measured at 0.5 s inter- vals as a function of elution volume (red line, reported in the pri- mary units of the signal, Volts). The relative concentration of protein, as measured by the refractive index, is also shown (blue). (B) A plot of the molar mass predicted from the analysis of the con- centration and scattering data as a function of elution volume (red dots). Superimposed for reference is the same concentration curve shown above (blue line). The area highlighted in yellow is the por- tion of the concentration and scattering curves used in the analysis. Oligomerization of CaMKII O. S. Rosenberg et al. 684 FEBS Journal 273 (2006) 682–694 ª 2006 The Authors Journal compilation ª 2006 FEBS subunits, but it is not possible to determine unambigu- ously the symmetry of this particle in the C. elegans sample. In the mouse CaMKIIa sample the first eigen image appears to be 6-fold symmetric (Fig. 4C). Thus by examining the strong symmetry of the uranyl acetate stained particles and the overall structure of the ice embedded sample we conclude, in agreement with Morris & Torok [9], that the association domain forms a 6-fold symmetric ring of  6 nm outer radius with the kinase domains surrounding this central ring in a second ring of density with an outer radius of  12 nm. Bacterially expressed association domains form a 7-fold symmetric assembly in negatively stained samples In order to understand the discrepancy between the results from light scattering and electron microscopic analysis of the holoenzyme and the crystallographic results, we next examined the truncated association domain of mouse CaMKIIa (residues 336–478), puri- fied from bacteria, using electron microscopy. We picked, aligned and classified 2317 association domain particles. Again, all of the particles appeared to be in the same orientation on the grid (Fig. 5A). The associ- ation domain particles revealed a strong 7-fold sym- metric modulation as seen in the first eigen image (Fig. 5B). Association domains prepared in diverse ways crystallize as 7-fold symmetric rings Over the course of our experiments using full-length mouse CaMKIIa purified from baculovirus-infected insect cells we noticed that the protein would some- times degrade into two distinct fragments (as seen with Fig. 3. Uranyl acetate stained images of the M. musculus holoenyzme reveal a 6-fold symmetry. (A) Electron microscopic images and class averages. (i) Three representative raw images of single particles. (ii) Six repre- sentative class averages, all of which look very similar, suggesting a limited distribution of orientations on the grid. (B) The first four eigen images of the association domain classification. The first eigen image shows a strong 6-fold modulation, as seen in the inset expanded view. O. S. Rosenberg et al. Oligomerization of CaMKII FEBS Journal 273 (2006) 682–694 ª 2006 The Authors Journal compilation ª 2006 FEBS 685 SDS ⁄ PAGE). In order to better define this degradation process we treated the preparation with increasing amounts of trypsin while incubating the samples on ice. At an intermediate concentration of trypsin (0.01 mgÆmL )1 trypsin) and short time (30 min) the protein was digested into two bands, as visualized by SDS ⁄ PAGE (Fig. 6A). These two populations of pro- teins were separated into two peaks on an HPLC column and characterized by electrospray ion trap mass spectrometry. The first peak contained only a 35 198 kDa protein corresponding to residues 2–311 of CaMKII (which encompasses the kinase domain and the auto-inhibitory segment, and is acetylated on the N-terminus). The other peak contained a mixture of association domain fragments of masses 18 255, 17 811, 17 683, and 17 170 kDa, corresponding to resi- dues 318–478, 323–478, 324–478, and 329–478, respect- ively. We noticed that given sufficient time (overnight at 20 °C) the kinase domain fragment was digested com- pletely by the trypsin; the association domain fragment was resistant, however, to further degradation at this concentration of protease. After the extended incuba- tion, all of the protein was degraded to a single band on an SDS ⁄ PAGE gel. As the protein preparation prior to the trypsin treatment was shown by electron microscopy to be 6-fold symmetric, we reasoned that that the association domain assemblies in the proteo- lyzed sample should also be 6-fold symmetric. With the aim of obtaining the crystal structure of a 6-fold symmetric association domain assembly, the proteo- lysed protein solution was used in crystallization trials, resulting in large diamond-shaped crystals in mother liquor with a pH of 4.6 (Fig. 6B). We analyzed the crystals by mass spectrometry, as above, and found them to contain a mixture of fragments 340–478 and 341–478. It is interesting to note that these are not the tryptic fragments that are present in the original mix- ture, indicating that additional cleavage took place during crystallization. These crystals diffract X-rays to 3.7 A ˚ with a tetragonal lattice of a ¼ b ¼ 166.4 A ˚ ,c¼ 192.4 A ˚ (Table 1) that is distinct from the monoclinic lattice seen earlier for the bacterially expressed association domain [10]. A single association domain dimer from the previously determined crystal structure was used as a search model for molecular replacement. We allowed the program phaser [31] to find a solution by placing monomers sequentially into the model, with each placement substantially increasing the significance of the solution (as measured by the Z-score). phaser placed seven dimers automatically into the unit cell, reproducing essentially the structure that has been determined earlier for the mouse CaMKIIa association domain. We next used the original association domain structure as a model for molecular replacement and phaser found a highly significant solution (Z-score for the rotation function equal to 11 and for the transla- tion function equal to 46), indicating that the contents of the unit cell are very well described by the 7-fold symmetric model; the electron density maps (Fig. 6C) produced from this model demonstrate unambiguously the presence of 7-fold symmetry in the crystals. Because this structure has already been well described, further analysis of this crystal form was abandoned. We next investigated whether the tetradecameric oligomeric state of the association domain was an iso- form-specific feature of mouse CaMKIIa. We repeated the proteolysis and crystallization protocol with the C. elegans CaMKII holoenzyme expressed in baculo- virus-infected insect cells. We found that the C. elegans association domain produced in this way also crystal- lized from a mother liquor with a pH of 6.4 and 5% polyethylene glycol (PEG) 400 in an orthorhombic, centered lattice. The self-rotation function revealed the three 2-fold axes of the space group as well as six addi- tional 2-fold axes; that is, there are seven independent 2-fold axes arrayed in the b-c plane, indicating the Fig. 4. Cryo-electronmicroscopy reveals the position of the kinase domains of CaMKII. (A) A representative class average (1 of 5) from the C. elegans holoenyzme micrographs. The central ring of density of  10 nm radius is presumed to be the association domain. The outer ring of density of  22 nm radius is presumed to be the kin- ase domains. (B) As in (A) but from the mouse CaMKIIa isoform micrographs. (C) The first eigen image of the mouse CaMKIIa iso- form dataset. Oligomerization of CaMKII O. S. Rosenberg et al. 686 FEBS Journal 273 (2006) 682–694 ª 2006 The Authors Journal compilation ª 2006 FEBS presence of 7-fold symmetry in the asymmetric unit (Fig. 7A). We carried out molecular replacement using X-ray data to 2.7 A ˚ with a single monomer of the associ- ation domain in the search model using the program phaser as described above. The asymmetric unit defined by molecular replacement solution consists of seven protomers which are related by a crystallo- graphic 2-fold axis along the a axis to another seven protomers, forming a ring of dimeric association domains. The symmetry of the complex intersects with the symmetry of the space group along one of the 2-fold axes of the complex such that the non- crystallographic 2-fold axis of the complex sits on top of the crystallographic space group 2-fold. This explains the presence of six noncrystallographic 2-fold axes of symmetry in the self-rotation function. We have refined the model to R-values of 24.5% ⁄ 29.6% (working and free, respectively). The structure of the association domain of the C. elegans CaMKII is in general very similar to that of the mouse CaMKIIa association domain (PDB code: 1HKX) with a root mean square deviation in Ca atom positions of 1.8 A ˚ (Fig. 7B,C) over all  143 residues in the monomer. The structure of CaMKII has been hypothesized to be sensitive to pH [32,33]. All of the crystallization conditions of the association domain found so far were at a pH < 7. We wondered whether perhaps pH could affect the oligomerization state of the complex seen in the crystal. We screened for new crystallization conditions for the mouse CaMKIIa association domain construct (expressed in bacteria) at higher pH and found a new crystal form growing in 25% (w⁄ v) PEG 3350 at a pH of 8.0. The crystals diffract X-rays to  3.7 A ˚ resolution and are in the orthorhombic space group P2 1 2 1 2 1 with unit cell dimensions of a ¼ 118.5 A ˚ , b ¼ 56.8 A ˚ and c ¼ 374.9 A ˚ . Although Fig. 5. The CaMKIIa association domain (residue 335–478) expressed in bacteria forms 7-fold rings. (A) Electron microscopic images and class averages. (i) Three repre- sentative raw images from the micrographs of the bacterially expressed CaMKIIa associ- ation domain. (ii) Six representative class averages that look very similar, suggesting a limited distribution of orientations on the grid. (B) The first four eigen images of the association domain classification. The first eigen image shows a strong 7-fold modula- tion. O. S. Rosenberg et al. Oligomerization of CaMKII FEBS Journal 273 (2006) 682–694 ª 2006 The Authors Journal compilation ª 2006 FEBS 687 the crystals are grown at a pH that is 3.7 units higher than the pH used to obtain the original crystal form found by Hoelz and coworkers [10], molecular replace- ment, carried out essentially as described above, showed that these crystals also contain a tetradeca- meric ring (not shown). We therefore conclude that the tetradecameric assemby is a stable state of the isolated association domain. Discussion and Conclusion Determination of the crystal structure of the associ- ation domain of mouse CaMKIIa [10] was an import- ant step towards the ultimate goal of understanding the organization of the kinase holoenzyme. The unex- pected oligomeric state of the crystal structures, which contain 14 subunits in an assembly with 7-fold sym- metry, was surprising. The consistent picture that emerged from previous electron microscopic analyses was that the CaMKII holoenzyme has 6-fold symmetry [9,13,34]. In this work we have shown that the 7-fold symmetry of the association domain crystal structure is a consequence of removing the residues N-terminal to the association domain. Our analyses of the CaMKII holoenzyme by light scattering and by electron micros- copy show 6-fold symmetry in the assembly, consistent with the earlier work of others. An interesting result that emerges from our work is that the 7-fold symmetry of the isolated associ- ation domain ring appears to be a conserved prop- erty of these domains across species. The crystal structure of the association domain of C. elegans AB C Fig. 6. Proteolysis of the mouse CaMKIIa holoenzyme leads to a 7-fold symmetric association domain structure. (A) Proteolysis of the M. musculus holoenyzme with tryp- sin leads to the production of two bands. Subsequent mass spectrometric analysis showed these two bands to be the kinase domain and the association domain as indi- cated in the figure. (B) Crystallization trials with the mixture shown in (A) produces large diamond like crystals. (C) Molecular replacement with the dimer of association domains reveal a 7-fold ring structure in the electron density of the crystals produced from the proteolyzed material. Oligomerization of CaMKII O. S. Rosenberg et al. 688 FEBS Journal 273 (2006) 682–694 ª 2006 The Authors Journal compilation ª 2006 FEBS CaMKII has been determined as part of this work and it forms a tetradecamer with 7-fold symmetry (Fig. 7). Likewise, a high pH form of the mouse CaMKIIa association domain, as well as new low pH crystal forms of the same domain, all show 7-fold symmetry. One important consequence of this tendency of the isolated association domain to form 14-membered assemblies is the inability to obtain high-resolution crystal structures of the relevant 6-fold symmetric association domain assembly. It appears that a direct view of this assembly at high resolution will have to await the determination of the crystal structure of an intact CaMKII holo- enyzme. We have, in the meantime, found it useful to model the 6-fold symmetric association domain assembly by removing a pair of subunits from the 7-fold symmetric crystal structure. The resulting gap has been closed computationally by applying a con- straint that closes the ring while maintaining each interface to be as close as possible to those seen in the 7-fold structure (Fig. 8C). This model is only an approximation to the true structure, and the details of the interatomic contacts are incorrect at the inter- faces. Nevertheless, the model is a useful guide to the overall geometry and architecture of the associ- ation domain. To facilitate its use by others the model is made available as Supplementary material. Previous analysis, using electron microscopy, of a somewhat longer construct of the isolated association domain (residues 317–478) of CaMKIIa expressed in baculovirus resulted in the conclusion that it is 6-fold symmetric [8]. It is possible that the region between residues 317–335, which is not included as part of our bacterial expression constructs and is removed by pro- teolyic digestion of the baculovirus-expressed intact protein, makes a difference to the oligomerization state. Nevertheless, we conclude from the results pre- sented in this study that some region of the holo- enzyme outside of the association domain affects the final assembled structure of the holoenzyme. Whether this involves the whole auto-inhibited kinase domain or a limited region of residues 1–335 is not clear at the present time. Our results raise the intriguing possibility that the kinase domains somehow prevent the association domains from relaxing from a 6-fold ring into a more stable 7-fold form. Implicit in this idea is that the N-terminal kinase domains interact with each other around the ring of the association domains. The binding of Ca 2+ ⁄ CaM to the holoenzyme is highly cooperative [11,35]. Both of these studies obtained a Hill coefficient for binding of around or above 2, suggesting that interactions between adjacent kinase domains impede the binding of Ca 2+ ⁄ CaM. These interactions may induce a strain in the ring that is released in the absence of the kinase domains allow- ing the association domains to relax into the 7-fold state (Fig. 8). Given the conservation of the 7-fold symmetry in the association domain ring, and the imposition of Fig. 7. Proteolysed C. elegans CaMKII also crystallizes as 7-fold association domain rings. (A) The self rotation function reveals seven 2-fold axes of rotation in the association domain crystals, although one of the 2-fold axes coincides with a crystallographic 2-fold axis. (B,C) The structure of the monomer (B) and the 7-fold association domain ring (C) produced from the proteolysis of the C. elegans CaMKII holoenzyme. O. S. Rosenberg et al. Oligomerization of CaMKII FEBS Journal 273 (2006) 682–694 ª 2006 The Authors Journal compilation ª 2006 FEBS 689 6-fold symmetry in the holoenyzme, one can specu- late whether this oligomerization state reflects some important aspect of the function of CaMKII. CaMKII is thought to play a role in the storage of long-term memories [5], based on its ability to behave in a switch-like manner such that Ca 2+ con- centrations above a certain threshold maintain the holoenyzme in a phosphorylated ‘on’ state despite the action of phosphatases and protein turnover. Recent studies have shown that the association domain monomers do not exchange between different rings when the association domain is expressed in isolation [36]. Our results suggest that there may be a period of instability before the formation of the stable tetradecameric form observed in the crystal structure when exchange between rings is possible. Perhaps some change in ring tension after activation could allow for swapping of damaged holoenzyme subunits, which would then be rapidly phosphorylated by the active subunits in the holoenzyme, maintaining the activation state of the holoenzyme for long peri- ods of time. The present study does not provide any information about the energetics of such instability, but understanding the mechanics of the transition from 6- to 7-fold symmetry is an important goal for future studies. Table 1. Data collection statistics for crystallography. Highest resolution bins are indicated in parentheses alongside the resolution of each dataset. Mouse CaMKIIa (trypsinized) C. elegans CaMKII (trypsinized) CaMKIIa (bacterially expressed) (high pH) Space Group P4 C222 1 P2 1 2 1 2 1 Unit cell (A ˚ ) a ¼ b ¼ 166.4 c ¼ 192.5 a ¼ 70.9 b ¼ 186.9 c ¼ 182.8 a ¼ 56.9 b ¼ 115.4 c ¼ 371.2 Wavelength (A ˚ ) 0.95 0.953 0.953 Resolution Range (A ˚ ) 50–3.88 50–2.64 50–3.74 I ⁄ r 5.06 (2.15) 17.9 (1.82) 5.65 (2.14) Completeness 75.5% (59.0%) 99.4% (99.4%) 95.5% (98.3%) R sym 12.0% (33.0%) 13.3% (10.5%) 34.6% (84%) Fig. 8. Models of holoenzyme. In the holo- enzyme the kinase domains may constrain the ring so as to maintain a dodecameric assembly (A). When the kinase domains are absent this constraint is released allowing the ring to relax into the 7-fold symmetric assembly (B). (C) A 6-fold symmetric associ- ation domain model in contrast with (D), the 7-fold association domain crystal structure from proteolyzed C. elegans full-length CaMKII. Oligomerization of CaMKII O. S. Rosenberg et al. 690 FEBS Journal 273 (2006) 682–694 ª 2006 The Authors Journal compilation ª 2006 FEBS Experimental procedures Protein purification The full-length C. elegans CaMKII (UNC-43 splice variant K11E8.d) was subcloned into pFastBac-1 (Gibco, Grand Island, NY, USA). Aspartate 135 was mutated to aspara- gine using the QuikChange mutagenesis kit (Stratagene, La Jolla, CA, USA) to inactivate the kinase domain and pre- vent autophosphorylation. Recombinant bacmid DNA was prepared according to the manufacturer’s instructions (Bac- to-Bac expression system, Gibco) and transfected into Sf9 cells. Baculovirus obtained from the transfection was used to infect Sf9 cells grown in suspension to a density of 2.5 · 10 6 per ml at a multiplicity of infection of approxi- mately 10. Cells were grown for 48 h, centrifuged and resuspended in 50 mm Hepes pH 7.4, 50 mm KCl, and 10% (v ⁄ v) glycerol. Cells (4 L) were lysed with a French press and centrifuged in an ultracentrifuge at 100 000 g to remove cellular debris. The protein was purified with HiTrap SP Sepharose Fast Flow (SP, Amersham Bio- sciences, Piscataway, NJ, USA), HiTrap Q Sepharose Fast Flow (Q, Amersham Biosciences) and size exclusion chro- matography (Superose 6 Prep Grade, Amersham Bio- sciences). The final buffer from the gel filtration was 20 mm Tris pH 8.0, 150 mm KCl, 1 mm dithiothreitol. The purified protein was more than  95% pure and its identity was confirmed with complete trypsin digestion and identification of peptides by mass spectrometry. The full-length M. musculus CaMKIIa was subcloned into a pFastBac-1 plasmid (Gibco) modified to contain a C-terminal, 6-histidine tag. Aspartate 135 in the kinase domain was mutated to asparagine as above. The protein was expressed and purified as described for the C. elegans full-length protein except that an additional nickel affinity column was added after the initial SP column. The final buffer from the gel filtration was 20 mm Tris pH 8.3, 200 mm KCl, 5% (v ⁄ v) glycerol. The protein was  99% pure and its identity was confirmed with complete trypsin digest and identification of peptides by mass spectrometry. The association domain of mouse CaMKIIa (residues 336–478) was expressed and purified as described [10]. Static light scattering Protein (20 lm) was injected onto a Superdex 200 H10 ⁄ 30 size exclusion chromatography column equilibrated in 20 mm Tris pH 7.4, 200 mm KCl, 10 mm MgCl 2 , and 1 mm tri(2-carboxyethyl)phosphine hydrochloride (TCEP). The column was coupled to an 18-angle light scattering detector (DAWN EOS) and refractive index detector (Optilab DSP) (Wyatt Technology, Santa Barbara, CA, USA). The col- umn was run at 0.4 mLÆmin )1 and data were collected every 0.5 s. The data were analyzed using the program package astra (Wyatt Technology). Electron microscopy Sample preparation and data collection For negative stain microscopy, a 5 lL sample of protein (15–30 lgÆmL )1 )in20mm Tris pH 8.0, 200 mm KCl and 1mm TCEP was placed on the carbon side of a glow-dis- charged, Formvar ⁄ carbon 300 mesh copper grid (Ted Pella, Redding, CA, USA), and the excess was removed by wick- ing with filter paper. The bound particles were stained with 5 lLof2%(w⁄ v) uranyl acetate for 30 s; the excess stain was removed by blotting with filter paper. Images of stained full-length CaMKIIa were recorded on Kodak SO-163 film with a Philips CM200 transmission electron microscope (FEI, Hillsboro, OR, USA) at 200 kV using a magnification of 66K. Negatively stained association domain was examined using a JEOL-3100-SFF transmis- sion electron microscope (JEOL, Peabody, MA, USA) equipped with a field emission gun operating at 300 kV, at a magnification of 70K (nominal value of 50K with post- column magnification of ·1.4). Images were recorded on a 2048 · 2048 slow-scan CCD camera (Gatan, Pleasanton, CA, USA) with defocus values between 2 lm and 3 lm. For cryo-electron microscopy, a 5 lL sample of CaMKIIa (150 lgmL )1 )in20mm Tris pH 8.0, 200 mm KCl and 1 mm TCEP was deposited on the glow-dis- charged carbon side of a lacey Formvar ⁄ carbon 300 mesh copper grid (Ted Pella), and the excess was removed by blotting with filter paper. The grid was rapidly cooled by plunging into liquid ethane; specimens were stored in liquid nitrogen (77 K) and kept below )170 °C. Images were acquired in a JEOL-3100-SFF electron microscope, des- cribed before, also using an in-column Omega energy filter with a slit width of 32 ev. Images were recorded on Kodak SO-163 film at a magnification of 60K and defocus values  2–5 lm. All data were acquired under low dose condi- tions, allowing a maximum dose per image of 20e ⁄ A 2 . Digitization and particle extraction Micrographs were digitized using a Nikon Super CoolScan 8000 scanner (Nikon USA, Melville, NY, USA) at a step size of 6.35 lm per pixel, and subsequently averaged to yield a final pixel size corresponding to 2.12 A ˚ (frozen hydrated specimen) and 1.93 A ˚ (negatively stained particles) on the specimen scale. CCD images of the association domain were recorded with a final pixel size of 4.45 A ˚ on the specimen scale. Micrographs showing significant frost, astigmatism, or drift were rejected. Particles were selected in 160 · 160 (frozen hydrated full-length), 180 · 180 (negative stained full-length) and 80 · 80 (negative stained association domain)-pixel boxes using the boxer procedure of the eman software package [37,38]. The particle data sets consisted of 3865 CaMKIIa (frozen hydrated), 2673 CaMKIIa (stain), and 2317 association domain (stain) particle images. O. S. Rosenberg et al. Oligomerization of CaMKII FEBS Journal 273 (2006) 682–694 ª 2006 The Authors Journal compilation ª 2006 FEBS 691 [...]... reconstructions of calcium ⁄ calmodulin-dependent (CaM) kinase IIa and truncated CaM kinase IIa reveal a unique organization for its structural core and functional domains J Biol Chem 275, 14354–14359 35 Rosenberg OS, Deindl S, Sung R, Nairn AC & Kuriyan J (2005) Structure of the auto-inhibited kinase domain of CaMKII and SAXS analysis of the holoenzyme Cell 123, 849–860 36 Lantsman K & Tombes RM (2005) CaMK -II oligomerization. .. from the structure A list of all the distances between a single subunit and all other subunits was calculated with the program cns [41] The values in this list were used as very strong restraints on the distances between the individual subunits – including the ones on either side of the gap where the two subunits were removed – in a standard rigid body refinement After 50 steps of minimization the ring... Identification and characterization of delta B-CaM kinase and delta C-CaM kinase from rat heart, two new multifunctional Ca2+ ⁄ calmodulin-dependent protein kinase isoforms Biochim Biophys Acta 1221, 89–101 Oligomerization of CaMKII 17 Caran N, Johnson LD, Jenkins KJ & Tombes RM (2001) Cytosolic targeting domains of gamma and delta calmodulin-dependent protein kinase II J Biol Chem 276, 42514–42519 18... all six interfaces by this procedure No further optimization of the structure was carried out Trypsin digestion and crystallization The coordinates for the structure of the C elegans CaMKII association domain prepared by proteolysis have been deposited in the PDB and given the code 2F86 Acknowledgements We thank Robert Glaeser, Pietro de Camilli, Karin Reinisch and Cori Bargmann for helpful discussions... were processed using the program package hkl2000 [40] The structures were solved with molecular replacement as described in the text using the program phaser [31] For the M musculus CaMKIIa association domain, a model phased 692 map of electron density was generated with the program cns [41] Inspection of the maps showed that the structure was essentially indistinguishable from the higher resolution... frequency of Ca2+ oscillations Science 279, 227–230 21 Hudmon A & Schulman H (2002) Structure-function of the multifunctional Ca2+ ⁄ calmodulin-dependent protein kinase II Biochem J 364, 593–611 22 Thiel G, Czernik AJ, Gorelick F, Nairn AC & Greengard P (1988) Ca2+ ⁄ calmodulin-dependent protein kinase II: identification of threonine-286 as the autophosphorylation site in the alpha subunit associated with the. .. unique organization for its structural core and functional domains J Biol Chem 275, 14354–14359 9 Morris EP & Torok K (2001) Oligomeric structure of alpha-calmodulin-dependent protein kinase II J Mol Biol 308, 1–8 10 Hoelz A, Nairn AC & Kuriyan J (2003) Crystal structure of a tetradecameric assembly of the association domain of Ca2+ ⁄ calmodulin-dependent kinase II Mol Cell 11, 1241–1251 11 Gaertner TR,... Zhou ZH, Ando S, Furutsuka D & Ikebe M (1995) Characterization of Ca2+ ⁄ calmodulin-dependent protein kinase II from smooth muscle Biochem J 310, 517– 525 15 Dosemeci A, Reese TS, Petersen JD, Choi C & Beushausen S (1999) Localization of the Linker Domain of Ca2+ ⁄ Calmodulin-Dependent Protein Kinase II Biochem Biophys Res Commun 263, 657–662 16 Edman CF & Schulman H (1994) Identification and characterization... Sondermann, Cori Ralston, David King and Arnie Falick for invaluable technical expertise The cDNA for UNC-43 splice variant K11E8.d was a kind gift of Dr Min Han Oren Rosenberg is supported by the Yale Medical Scientist Training Grant Parts of this work have been supported by the Director, Of ce of Science, Of ce of Basic Energy Sciences, of the U.S Department of Energy under Contract No DE-AC02-05CH11231... masks of ˚ ˚ ˚ radius 64 A or 128 A (full-length) and 80 A (association domain) , followed by calculation of class averages Trypsin digestion and crystallization For analytical studies, trypsin (0.5 lg, Sigma, St Louis, MO, USA) was added to 50 lL of protein (either M musculus CaMKIIa or C elegans CaMKII ⁄ UNC-43) solution at 0.55 mm on ice for 30 min The reaction was stopped by adjustment of the solution . state of the holoenzyme and that of the crystal structure of the association domain is unknown. Although the association domain is necessary and sufficient for formation of oligomeric CaMKII [18,19],. Conclusion Determination of the crystal structure of the associ- ation domain of mouse CaMKIIa [10] was an import- ant step towards the ultimate goal of understanding the organization of the kinase holoenzyme. The. forming a ring of dimeric association domains. The symmetry of the complex intersects with the symmetry of the space group along one of the 2-fold axes of the complex such that the non- crystallographic