Structural Insights into the Regulation of Ca²⁺/Calmodulin-Dependent Protein Kinase II (CaMKII)

Abstract

Ca²⁺/calmodulin-dependent protein kinase II (CaMKII) is a highly conserved serine/threonine kinase that is ubiquitously expressed throughout the human body. Specialized isoforms of CaMKII play key roles in neuronal and cardiac signaling. The distinctive holoenzyme architecture of CaMKII, with 12–14 kinase domains attached by flexible linkers to a central hub, poses formidable challenges for structural characterization. Nevertheless, progress in determining the structural mechanisms underlying CaMKII functions has come from studying the kinase domain and the hub separately, as well as from a recent electron microscopic investigation of the intact holoenzyme. In this review, we discuss our current understanding of the structure of CaMKII. We also discuss the intriguing finding that the CaMKII holoenzyme can undergo activation-triggered subunit exchange, a process that has implications for the potentiation and perpetuation of CaMKII activity.

Ca²⁺/calmodulin-dependent protein kinase II (CaMKII) is a multimeric Ser/Thr protein kinase with a unique holoenzyme architecture (Hook and Means 2001; Hudmon and Schulman 2002; Lisman et al. 2002, 2012; Stratton et al. 2013). It is one of the most predominant proteins in the brain, where it is critical for the long-term potentiation of synaptic signaling and in processes underlying the development of learning and memory (Silva et al. 1992; Giese et al. 1998; Elgersma et al. 2002; Hudmon and Schulman 2002; Giese and Mizuno 2013; Lisman and Raghavachari 2015). CaMKII is also expressed in the heart, where it plays a pivotal role in the modulation of excitation–contraction coupling (Maier and Bers 2007; Swaminathan et al. 2012; Erickson 2014).

Most Ca²⁺/CaM-dependent kinases are monomeric (Swulius and Waxham 2008). By contrast, CaMKII is organized into large oligomers, typically with 12 or 14 subunits for the human α-isoform (Fig. 1A). Central to the organization of CaMKII is the hub domain, also known as the association domain, which forms a donut-shaped ring that is the core of the holoenzyme. The kinase domains are tethered to the central hub by the regulatory segments and intrinsically disordered linkers, which we refer to as the kinase-hub linker (Fig. 1A,B; Hudmon and Schulman 2002; Stratton et al. 2013). There are four isoforms, and about 40 splice variants, of CaMKII in humans, and these differ primarily in the length and sequence of the kinase-hub linker (Fig. 1C; Tombes et al. 2003).

Figure 1.

Structural organization of CaM-dependent protein kinase II (CaMKII). (A) The hub domain forms a dodecameric or tetradecameric ring and the kinase domains are flexibly linked to it by the regulatory segment and an intrinsically disordered linker, referred to as the kinase-hub linker. (B) Schematic diagram showing the domain organization of one CaMKII subunit. (C) Schematic depiction of the differences in linker lengths between the four isoforms of human CaMKII. (D) Crystal structure of the autoinhibited kinase from human CaMKII-δ bound to a small molecule inhibitor (PDB: 2VN9) (Rellos et al. 2010). The CaM-binding element of the regulatory segment places the Thr 306 for phosphorylation in cis. (E) Crystal structure of the autoinhibited kinase domain from Caenorhabditis elegans CaMKII shows a helical organization of the entire regulatory segment (PDB: 2BDW) (Rosenberg et al. 2005). (F) The C. elegans CaMKII kinase domains forms a dimer in the crystal, with the CaM-binding element forming an antiparallel coiled-coil (PDB: 2BDW). The dimer prevents cis-phosphorylation of Thr 305 or Thr 306 in each subunit. Thr 286 is held far away from the active site of the other subunit in the dimer.

Most protein kinases are regulated by phosphorylation on the activation loop, a centrally located element that provides a platform for substrate binding (Huse and Kuriyan 2002; Endicott et al. 2012). CaMKII is unusual in that it does not have a phosphorylation site in its activation loop. Instead, like other kinases that are regulated by calcium-bound calmodulin (Ca²⁺/CaM), the activity of the kinase domain in CaMKII is suppressed in the absence of calcium signals by a regulatory segment that blocks the active site. Ca²⁺/CaM binds to a CaM-binding element in the regulatory segment of CaMKII, releasing it from the kinase domain and thereby activating the enzyme. Upon stimulation by Ca²⁺/CaM, CaMKII activity is regulated by phosphorylation on two sets of highly conserved phosphosites in the regulatory segment (Thr 286 and Thr 305/Thr 306; human CaMKIIα numbering is used throughout in the review) (Lai et al. 1986; Lou et al. 1986; Miller and Kennedy 1986; Colbran et al. 1989; Lou and Schulman 1989).

Activation of CaMKII results in downstream phosphorylation of several targets at the synapse, notably the AMPA receptors and transmembrane AMPA-receptor-regulating proteins (Barria et al. 1997; Herring and Nicoll 2016). This contributes to a strengthening of synaptic transmission (Lisman et al. 2002; Wayman et al. 2008). Mutations in CaMKII at the highly conserved threonine residues in the regulatory segment result in impairments in learning and memory (Silva et al. 1992; Giese et al. 1998; Elgersma et al. 2002). In addition to this, mounting genetic evidence indicates that mutations in CaMKII are correlated with cognitive defects and developmental delay in humans (Robison 2014; Küry et al. 2017; Chia et al. 2018).

In addition to its key role in neuronal signaling, CaMKII is a key transducer of Ca²⁺ signaling in the heart. Briefly, factors such as high heart rate, β-adrenergic stimulation, high angiotensin II concentrations, and high blood glucose concentrations in diabetes, lead to hyperactivation of CaMKII in the heart (Erickson et al. 2013, 2008; Erickson 2014). CaMKII hyperactivation promotes cardiac pathology, such as arrhythmias and heart failure, by modulating electrophysiological and mechanical properties of heart muscle cells via phosphorylation of downstream targets (Maier and Bers 2007; Swaminathan et al. 2012). This makes CaMKII an attractive drug target for treating cardiovascular pathologies (Pellicena and Schulman 2014).

In this review, we focus on how structural insights inform our current understanding of CaMKII activation. We do not cover physiological aspects of CaMKII function, which have been reviewed extensively elsewhere (Hudmon and Schulman 2002; Zhang and Brown 2004; Lisman et al. 2012).

AUTOINHIBITION AND Ca²⁺/CaM-DEPENDENT PHOSPHORYLATION CONTROL OF CaMKII

The autoinhibition of CaMKII by the regulatory segment is consistent with a pseudosubstrate mechanism for kinase autoinhibition that was proposed originally for myosin light chain kinase (MLCK) (Kemp et al. 1987, 1994) and first seen in structures of twitchin kinase (Kobe et al. 1996), phosphorylase kinase (Owen et al. 1995), CaMKI (Goldberg et al. 1996), and titin kinase (Mayans et al. 1998). In the autoinhibited state, the CaMKII regulatory segment binds to the substrate-docking groove on each kinase domain (Fig. 1D,E; Rosenberg et al. 2005; Rellos et al. 2010). This interaction of the regulatory segment with the kinase blocks the binding of substrates and also sequesters Thr 286, which is buried at the interface between the regulatory segment and the kinase domain. The interaction between Ca²⁺/CaM and the regulatory segment of CaMKII has been defined by nuclear magnetic resonance (NMR) and crystallographic studies (Ikura et al. 1992; Meador et al. 1992, 1993; Rellos et al. 2010). The CaM-binding element of the regulatory segment adopts a helical conformation and is recognized by a hydrophobic central channel formed between the two Ca²⁺-binding lobes of CaM (Fig. 2A). Binding of Ca²⁺/CaM to the CaM-binding element peels the regulatory segment off the kinase, releases autoinhibition, and results in CaMKII activation (Fig. 3A).

Figure 2.

Autoinhibited CaM-dependent protein kinase II (CaMKII). (A) Crystal structure of the kinase domain of CaMKIIδ bound to Ca²⁺/CaM (PDB: 2WEL). The CaM-binding element adopts a helical conformation and is sandwiched in a hydrophobic channel between the two lobes of calcium-bound calmodulin (Ca²⁺/CaM). Thr 305 and Thr 306 are enclosed by Ca²⁺/CaM in this complex. (B) Model based on negative-stain electron microscopy of the extended conformation of autoinhibited human CaMKIIα, with the kinase domains variably distributed around the hub (PDB: 5U6Y). The blue annular ring represents the overall volume occupied by the kinase domains. (C) Crystal structure of a variant of human CaMKIIα in which the kinase-hub linker is deleted exhibits a compact conformation with the kinase domains symmetrically docked on the hub (PDB: 3SOA). (D) Schematic depiction of the three major conformational states for the autoinhibited CaMKII holoenzyme, showing the kinase-paired, extended, and compact forms of CaMKII. The colors correspond to those in Figure 1B. (Panels B and D are adapted from Myers et al. © 2017, The Author(s), https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5467236, under a Creative Commons Attribution 4.0 International License.)

Figure 3.

Regulation of CaM-dependent protein kinase II (CaMKII) by autophosphorylation. (A) CaMKII regulatory segment binds to the substrate-docking groove on the kinase domain and autoinhibits the kinase. The regulatory segment contains three sites of phosphorylation: Thr 286 and Thr 305/Thr 306. Upon interaction of the regulatory segment with calcium-bound calmodulin (Ca²⁺/CaM), Thr 286 undergoes phosphorylation, which renders the kinase Ca²⁺/CaM-independent. Upon subsequent release of Ca²⁺/CaM, phosphorylation of Thr 305/306 prevents rebinding and further stimulation by Ca²⁺/CaM. The colors correspond to those in Figure 1B. (B) Structural depiction of the substrate-capture model, where the amino-terminal half of the regulatory segment from one kinase (substrate kinase) is docked in the substrate-binding pocket of another kinase (enzyme kinase). This complex optimally positions Thr 286 for phosphorylation in the active site of the enzyme kinase (PDB: 3KK8). A similar mode of interaction is also seen in another structure (PDB: 2WEL).

Upon activation by Ca²⁺/CaM, CaMKII undergoes trans-autophosphorylation on Thr 286 in the regulatory segment (Fig. 3A; Lai et al. 1986; Lou et al. 1986; Miller and Kennedy 1986). Thr 286 is located at the base of the kinase domain, ∼30 Å from the active site. As a consequence, phosphorylation of Thr 286 can only occur in trans. Once phosphorylated on Thr 286, the regulatory segment is prevented from docking on the hydrophobic pocket in the kinase domain even when Ca²⁺/CaM is absent. In this way, activation by Ca²⁺/CaM can lead to Ca²⁺/CaM-independent activity, a phenomenon referred to as autonomy (Miller et al. 1988; Thiel et al. 1988; Lou and Schulman 1989).

When Ca²⁺/CaM is not bound to CaMKII, autophosphorylation ensues at two sites, Thr 305 and Thr 306, in the CaM-binding element (Fig. 3A; Schworer et al. 1988; Colbran et al. 1989; Lou and Schulman 1989). Phosphorylation of Thr 305 and Thr 306 can occur in cis, as suggested by crystal structures of human CaMKII kinase domain (Fig. 1D; e.g., PDB: 2VN9) (Rellos et al. 2010). Phosphorylation at these sites prevents the binding of Ca²⁺/CaM to CaMKII and such a kinase cannot be further stimulated (Hashimoto et al. 1987; Lickteig et al. 1988; Lou and Schulman 1989; Meador et al. 1993; Mukherji and Soderling 1994). For this reason, Thr 286 phosphorylation is referred to as “activating,” and Thr 305 and Thr 306 phosphorylation is referred to as “inhibitory” (Lou and Schulman 1989; Hanson and Schulman 1992).

The temporal order of phosphorylation at the activating and inhibitory sites results in different outcomes with respect to CaMKII activity. Autophosphorylation at Thr 305 and Thr 306 before phosphorylation of Thr 286 makes the kinase nonresponsive to Ca²⁺/CaM stimulus and such a kinase cannot be activated. On the other hand, if Thr 286 gets autophosphorylated first, it leads to a holoenzyme in which Thr 305 and Thr 306 are protected by Ca²⁺/CaM, and cannot be phosphorylated (Meador et al. 1993; Rellos et al. 2010). Thus, the phosphorylation status of CaMKII permits it to toggle between states in which Thr 286, or alternatively, Thr 305 and Thr 306, is phosphorylated. The functional consequences of this toggle are not understood. Replacement of Thr 286, Thr 305, and Thr 306 by residues that cannot be phosphorylated, or are phosphomimics, affects synaptic plasticity and elicit severe impairments in learning and memory (Giese et al. 1998; Elgersma et al. 2002; Giese and Mizuno 2013).

Although the mechanism of autoinhibition in CaMKII resembles that of other Ca²⁺-regulated kinases in general terms, the regulation by autophosphorylation of the regulatory segment (on Thr 286) and the lack of phosphorylation in the activation loop are unique to CaMKII. CaMKII genes have been found in the genomes of the choanoflagellates, which are unicellular organisms that diverged from the metazoan lineage about 600 million years ago (Peterson and Butterfield 2005; Carr et al. 2008; King et al. 2008; Richter and King 2013). The kinase domains of choanoflagellate CaMKII are ∼50% identical in sequence to that of human CaMKII. Like human CaMKII, choanoflagellate CaMKII proteins lack a phosphorylation site in the activation loop, and are regulated by phosphorylation on Thr 286. In addition, the crystal structure of the autoinhibited form of a choanoflagellate CaMKII closely resembles that of human CaMKII (Bhattacharyya et al. 2016). Intriguingly, the phosphorylation sites corresponding to Thr 305 and Thr 306 are missing in choanoflagellate CaMKII sequences. Thus, the unique phosphorylation control of CaMKII that enables the enzyme to acquire Ca²⁺/CaM-independent activity emerged before the origin of the metazoan lineage. However, the ability to regulate CaM-binding through phosphorylation is a feature that emerged in the metazoan lineage.

FLEXIBLE LINKERS TETHER THE KINASE DOMAIN TO A CENTRAL HUB ASSEMBLY IN CaMKII

There are four CaMKII isoforms in humans, denoted CaMKIIα, CaMKIIβ, CaMKIIγ, and CaMKIIδ (Fig. 1C). Alternative splicing generates ∼40 distinct splice variants of these isoforms (Tombes et al. 2003). The α and β isoforms of CaMKII are the predominant isoforms in the brain, and CaMKIIδ is predominant in the heart (Tobimatsu and Fujisawa 1989). CaMKIIγ is expressed throughout the body. These isoforms and splice variants share a high sequence identity in the kinase domains (∼95%) and the hub (∼80%). The kinase-hub linkers are divergent in both length and sequence.

CaMKIIα and CaMKIIδ both have 30 residue-long linkers. CaMKIIβ and CaMKIIγ have much longer linkers (200 and 110 residues, respectively) (Fig. 1C). In contrast to the divergence in the linker region across different isoforms, the CaMKII linker residues show a high sequence conservation within each isoform. Normally, residues in the intrinsically disordered linkers are often less evolutionarily conserved when compared to their structured counterparts (Brown et al. 2011). This isoform-specific conservation of linker residues points to a functional importance of the kinase-hub linker that is not understood.

CaMKII is not the only kinase in which the catalytic domain is linked to a scaffold by a long flexible linker. In the protein kinase A (PKA) holoenzyme, the kinase subunit is bound to regulatory subunits that contain intrinsically disordered segments that are about the same length as those in CaMKIIα (Smith et al. 2013). The amino-terminal domains of the regulatory subunits in PKA dimerize, and are bound to A-kinase anchoring proteins (AKAPs). This unit is linked to two kinase domains through the intrinsically disordered segments. In PKA, the length of the flexible linker can modulate cAMP signaling in cells (Smith et al. 2013). Negative-stain electron microscope (EM) reconstruction reveals that the two kinase domains in the holoenzyme are arranged randomly around the AKAP dimer with a radial separation consistent with the length of the disordered linker (∼40 residues in the human type IIα subunit) (Smith et al. 2013).

The design of the CaMKII holoenzyme takes this principle one step further, with 12 or 14 kinase domains distributed around the core assembly formed by the hub domain (Fig. 1A; Woodgett et al. 1983). Molecular models for autoinhibited CaMKII based on early analysis of negative stain electron micrographs had the kinase domains distributed symmetrically above and below the equatorial plane of the hub (Kolodziej et al. 2000) or in the equatorial plane (Morris and Török 2001). A generally similar model was derived from small-angle X-ray scattering (SAXS) data for Caenorhabditis elegans CaMKII (Rosenberg et al. 2005).

A recent analysis of the autoinhibited CaMKIIα, using negative-stain EM without the imposition of symmetry, provides new insights into the organization of full-length CaMKII holoenzyme (Myers et al. 2017). Instead of generating a unique model for the holoenzyme, the hub and the kinase domains were treated separately in this analysis. A pseudoatomic model of CaMKIIα based on these EM data reveals that the kinase domains are flexibly arranged at variable distances around a well-defined central hub. This flexible organization of the kinase domains at variable spacing from the hub leads to a dynamic organization of the resulting CaMKII holoenzyme, with individual holoenzyme particles varying from 15 to 35 nm in diameter (Fig. 2B–D; Myers et al. 2017).

The majority of the single particles that were analyzed (∼97%) correspond to CaMKIIα holoenzymes with 12 subunits, with only a small fraction (∼4%) having 14 subunits (Myers et al. 2017). This contrasts with the results of an earlier analysis of CaMKIIα by negative-stain EM, which showed roughly equal proportions of 12 and 14 subunits (Bhattacharyya et al. 2016). The constructs and expression systems used in the two studies were different. The analysis of Myers et al. used intact, wild-type CaMKIIα expressed in insect cells. Bhattacharyya et al. used a modified construct in which the last three residues of CaMKIIα were replaced by a nine-residue linker followed by a biotinylation-tag, and the protein was expressed in Escherichia coli. The origin of the difference in stoichiometry is unclear but, as discussed later, even the substitution of a small number of residues in the hub can shift its stoichiometric balance (McSpadden et al. 2019).

A significant fraction (∼20%) of the kinase domains in the CaMKII holoenzyme appear to be dimeric (Fig. 2D; Myers et al. 2017). This observation is interesting in light of a crystal structure of the autoinhibited kinase domain from C. elegans CaMKII, which shows a dimeric organization with the CaM-binding element of the regulatory segment forming an antiparallel coiled-coil (Fig. 1E,F; Rosenberg et al. 2005). The structure of this dimer is intriguing because it appears to stabilize the autoinhibited form of the kinase domain while holding Thr 286 in both kinases far away from the active site of the partner kinase, preventing trans-autophosphorylation. In addition, the dimer prevents cis-autophosphorylation on Thr 305 and Thr 306 because those residues are buried at the interface between the two helices in the coiled-coil.

Kinase dimers may explain the cooperativity of binding of Ca²⁺/CaM to CaMKII. The binding of the first Ca²⁺/CaM would disrupt the coiled-coil, which facilitates the binding of subsequent Ca²⁺/CaM to the released regulatory segment of the partner in the dimer. Note, however, that there are other explanations for the cooperativity of CaM binding to CaMKII (Chao et al. 2010). The dimerization of the kinase domains of CaMKII has been reported by other studies (Thaler et al. 2009; Rellos et al. 2010; Nguyen et al. 2012, 2015), but direct evidence for the coiled-coil dimer seen in the C. elegans kinase structure is lacking. Such a dimer may only form in the context of the CaMKII holoenzyme, where the local concentrations of the kinase domains are much higher as compared to isolated kinase domains (Rosenberg et al. 2005).

The presence of the intrinsically disordered kinase-hub linker makes CaMKII a very challenging target for crystallization. A crystal structure of an intact and autoinhibited CaMKII holoenzyme was obtained by making three modifications to human CaMKIIα. First, the kinase-hub linker was deleted, yielding a construct that is analogous to a naturally occurring splice-variant of CaMKIIβ. Second, the catalytic activity of CaMKII was impaired by replacing Lys 42 and Asp 135 with methionine and asparagine, respectively. Third, Thr 306 was replaced by valine to reduce phosphorylation heterogeneity in the sample. This variant of CaMKIIα was crystallized and the structure was determined at 3.5 Å resolution (Chao et al. 2011). In this structure, the holoenzyme adopts a compact configuration, with the kinase domains packed tightly around the hub (Fig. 2C,D). The CaM-binding region of the regulatory segment in each kinase domain is integrated into the tertiary structure of each hub subunit, enforcing a tight docking of the kinase domain on the hub.

The crystal structure represents a maximally autoinhibited state of CaMKII, because the CaM-binding element is completely buried at the interface between the kinase and the hub. It was shown, using SAXS, that CaMKIIα does not adopt this compact conformation in solution, which is consistent with the recent negative stain EM analysis (Myers et al. 2017). It was suggested that molecular crowding conditions, such as in the cytoplasm, would favor the more compact conformation (Chao et al. 2011), but the validity of this hypothesis has been questioned (Myers et al. 2017). The length and sequence of the kinase-hub linkers vary considerably in different CaMKII isoforms and splice variants. It remains to be seen whether the compact conformation is favored more in particular CaMKII variants over others or whether it is only relevant for CaMKII variants with very short kinase-hub linkers. It is also possible that only a small fraction of the kinase domains in a holoenzyme are docked on the hub at any given time.

CALMODULIN TRAPPING BY CaMKII

An interesting consequence of autophosphorylation at Thr 286 is an increase in the affinity of Ca²⁺/CaM for the CaM-binding element in CaMKII, by about 10,000-fold. This phenomenon is termed “CaM-trapping,” and it can result in the maintenance of high CaMKII activity even at subthreshold levels of cellular Ca²⁺ (Meyer et al. 1992; Putkey and Waxham 1996; Singla et al. 2001). Both Ca²⁺/CaM-independent constitutive activity and CaM-trapping enable retention of the memory of past activation stimuli by CaMKII.

An interesting utilization of the CaM-trapping mechanism has been recently demonstrated for the nuclear transport of Ca²⁺/CaM from the surface membrane, in complex with CaMKIIγ (Ma et al. 2014). Phosphorylation of CaMKIIγ at Thr 287 (equivalent to Thr 286 in CaMKIIα) increases its affinity for CaM. This tight interaction between Ca²⁺/CaM and CaMKIIγ in the complex ensures the delivery of Ca²⁺/CaM to the nucleus, initiating a cascade of signaling events that eventually lead to transcription of CRE-regulated genes (Ma et al. 2014).

A SUBSTRATE-CAPTURE MECHANISM FOR FACILITATING Thr 286 PHOSPHORYLATION

Phosphorylation at Thr 286 is the critical regulatory step, since it confers Ca²⁺/CaM independence. Not surprisingly, CaMKII has a mechanism to specifically promote this phosphorylation. Structural studies have shown that the phosphorylation of Thr 286 is potentiated by a “substrate-capture” mechanism in which an activated kinase domain can promote the release of the regulatory segment from an unactivated kinase domain and subsequent phosphorylation on Thr 286 (Chao et al. 2010; Rellos et al. 2010).

In one study, a construct of CaMKII containing the kinase domain was cocrystallized with Ca²⁺/CaM, providing striking visualization of how Ca²⁺/CaM activates the kinase (Rellos et al. 2010). In another study, a construct of CaMKII was used in which the regulatory segment was truncated, deleting the CaM-binding element. This construct mimicked the effect of activation by Ca²⁺/CaM (Chao et al. 2010). In both structures, the amino-terminal half of the regulatory segment, including Thr 286, from one kinase (the “substrate kinase”) is docked in the substrate-binding pocket of another kinase (the “enzyme” kinase) (Fig. 3B). This enzyme–substrate complex optimally positions Thr 286 for phosphorylation in the active site of the enzyme kinase. Indeed, one of the structures shows a phosphorylated Thr 286 residue positioned in the active site of the enzyme kinase (Chao et al. 2010). This substrate-capture mechanism is supported by biochemical data, and can contribute to ultrasensitivity in the response of CaMKII to Ca²⁺/CaM (Chao et al. 2010).

The architecture of the CaMKII holoenzyme plays a crucial role in the Ca²⁺/CaM response, by bringing the substrate and enzyme kinase in close proximity with a high local concentration. This cooperative positive feedback mechanism within CaMKII holoenzymes in combination with CaM trapping allows it to act as a frequency decoder of cellular Ca²⁺ transients and oscillations and integrator of Ca²⁺ signaling in cells (De Koninck and Schulman 1998; Eshete and Fields 2001).

SENSITIVITY OF CaMKII TO THE FREQUENCY OF CALCIUM (Ca²⁺) PULSES

The acquisition of Ca²⁺/CaM-independence by CaMKII (autonomy) is sensitive to the frequency of Ca²⁺ pulses (De Koninck and Schulman 1998). When exposed to low-frequency Ca²⁺ pulses (e.g., <1 Hz), CaMKII reverts to a dormant state between the Ca²⁺ pulses. In contrast, when the frequency of Ca²⁺ pulses is high (e.g., >10 Hz), CaMKII acquires Ca²⁺/CaM-independent constitutive activity that can be sustained even when Ca²⁺ levels fall (De Koninck and Schulman 1998). The arrival of further Ca²⁺ pulses then increases the degree of autonomy. In this way, CaMKII can serve as an integrator of Ca²⁺ signals, provided that the Ca²⁺ pulses are of sufficiently high frequency.

The mechanism behind frequency dependence and Ca²⁺ signal integration by CaMKII has been investigated by simulations of reaction kinetics (Fig. 4A,B; Pepke et al. 2010). When stimulated by low-frequency Ca²⁺ pulses, the CaMKII–CaM complex formed during one Ca²⁺ pulse dissociates completely before the next pulse, making the kinase quiescent during the interstimulus interval (Fig. 4A). In contrast, when stimulated by high-frequency pulses, the CaMKII–CaM complex formed during one Ca²⁺ pulse does not dissociate completely before the next pulse arrives (Fig. 4B). This leads to a cumulative accumulation of CaMKII–CaM complexes with each pulse, resulting in an increased phosphorylation at Thr 286 (Pepke et al. 2010). A critical aspect of this mechanism is the ability of CaM that is only partially occupied by Ca²⁺ to retain an affinity for CaMKII.

Figure 4.

Frequency dependence and Ca²⁺ signal integration by CaM-dependent protein kinase II (CaMKII). (A) Results of simulations of the kinetics of CaMKII activation by calcium-bound calmodulin (Ca²⁺/CaM). At low stimulation frequency, the CaMKII-Ca²⁺/CaM complex formed during one Ca²⁺ pulse dissociates completely before the next one. (B) Under high-frequency pulses, the CaMKII-Ca²⁺/CaM complex formed during one Ca²⁺ pulse does not fully dissociate before the next pulse. This leads to a cumulative accumulation of CaMKII- Ca²⁺/CaM complexes. (Panels A and B are adapted from Pepke et al. © 2010 under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction, provided the original author and source are properly credited.)

It was proposed that simultaneous binding of two CaM on neighboring subunits in CaMKII holoenzyme underlies the decoding of Ca²⁺ frequency by CaMKII (Hanson et al. 1994; Michelson and Schulman 1994). In a different explanation, the simulation study referred to earlier suggests that the frequency dependence arises from the kinetics of interaction of Ca²⁺ pulses with CaMKII–CaM complexes (Pepke et al. 2010). Another model has evoked the importance of the phosphatase in generating a frequency dependence in the response of CaMKII to Ca²⁺ pulses (Dosemeci and Albers 1996). However, the in vitro experiments to demonstrate the frequency dependence were performed in the absence of phosphatase by applying transient Ca²⁺ pulses (De Koninck and Schulman 1998), thereby pointing to a mechanism that is intrinsic to CaMKII holoenzyme (Kubota and Bower 2001). In addition, it has been suggested that CaMKII isoforms and splice variants have differential sensitivity toward the frequency of Ca²⁺ pulses (Bayer et al. 2002). The frequency response of CaMKII is a phenomenon that merits further study.

REGULATION OF CaMKII BY OTHER POSTTRANSLATIONAL MODIFICATIONS

In addition to phosphorylation, other posttranslational modifications, such as methionine oxidation and serine glycosylation, have been reported at sites clustered near the critical stimulatory phosphosite (Thr 286) in CaMKII. Similar to phosphorylation of Thr 286, these modifications impart Ca²⁺/CaM-independent constitutive activity to CaMKII.

Oxidation of CaMKIIδ at Met 281/282 (residue numbering in the CaMKIIα sequence is offset by -1 with respect to that in CaMKIIβ/γ/δ, the corresponding residues in CaMKIIα being Cys 280/Met 281) leads to constitutive activation that is independent of the phosphorylation at Thr 286 (Erickson et al. 2008). Such a mode of regulation couples the generation of reactive oxygen species to the activation of CaMKII in cardiac pathologies (Purohit et al. 2013; Luczak and Anderson 2014). Covalent modification of Ser 279 (Ser 278 in CaMKIIα) by O-linked N-acetylglucosamine (O-GlcNAc) leads to constitutively active CaMKII (Erickson et al. 2013). This modification of CaMKII has been reported under hyperglycemic conditions of diabetes mellitus, and may be a key contributor of cardiopathological conditions associated with diabetes (Hegyi et al. 2019).

VARIATIONS IN THE STOICHIOMETRY AND GEOMETRY OF CaMKII HUB DOMAIN

The hub domain gives the CaMKII holoenzyme its distinctive oligomeric organization and shape. The basic building blocks of the CaMKII hub are wedge-shaped “vertical dimer” units, so named because they span the equatorial plane of the hub assembly (Fig. 1A). The vertical dimers are the unit of oligomeric assembly (Bhattacharyya et al. 2016).

The hub is related structurally to proteins such as nuclear transport factor 2 (NTF2) and ketosteroid isomerases, despite sharing no significant sequence identity with these proteins (Wu et al. 1997; Stewart et al. 1998; Hoelz et al. 2003). Each hub monomer shares the fold of NTF2 and ketosteroid isomerases, comprised of two layers: an inner β-sheet and an outer layer of α-helices (Fig. 5A). There is a central cavity encapsulated by these two layers (Hoelz et al. 2003; Bhattacharyya et al. 2016). This cavity is lined by several highly conserved charged residues and is solvated, imparting a flexibility to the hub subunits and its readiness to accommodate a change in shape (Fig. 5A; Bhattacharyya et al. 2016). The vertical dimer that is the unit of CaMKII hub assembly is also seen in NTF2 and the ketosteroid isomerases, although the ring-shaped oligomeric assembly is unique to CaMKII.

Figure 5.

Variations in the stoichiometry and geometry of CaM-dependent protein kinase II (CaMKII) hub. (A) Schematic representation showing β-sheet layer and the α-helices (left panel). These two layers encapsulate a highly solvated cavity lined by three highly conserved Arg residues (right panel). (B) Structural comparison of the hub assemblies for the closed-ring human CaMKIIα and spiral Salpingoeca rosetta CaMKII. The six vertical dimeric units are labeled 1–6. (C) Structural comparison of tetradecameric and dodecameric closed-ring hubs from human CaMKII. Changes in the curvature of the β-sheet with respect to the α-helices allow the preservation of the hub interface packing between adjacent subunits. (D) Structural comparison between a subunit of the S. rosetta spiral hub and a human dodecameric hub. A drastic twist in the β-sheet by about 30° results in opening of the closed-ring hub, despite maintaining the hub interfacial packing. (Panels C and D are adapted from Bhattacharyya et al. © 2016 under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided the original author and source are credited.)

The CaMKII hub domain has considerable structural flexibility (Bhattacharyya et al. 2016). Isolated CaMKII hub assemblies from human, mouse, and C. elegans have been crystallized in both dodecameric and tetradecameric forms (Hoelz et al. 2003; Rosenberg et al. 2006; Rellos et al. 2010; Bhattacharyya et al. 2016). Native mass spectrometric studies reveal that the isolated human hub assemblies exist as a ∼1:1 mixture of dodecameric and tetradecameric species (Bhattacharyya et al. 2016).

Further evidence of the structural plasticity in the hub is evident from structural studies of isolated hub assemblies from divergent metazoan and nonmetazoan species. The sea anemone Nematostella vectensis, an early metazoan species, has two CaMKII variants that are 55% (isoform A) and 53% (isoform B) identical to human CaMKIIα, respectively (Bhattacharyya et al. 2016). The hub domains of these isoforms (A and B) crystallized as an open-ring and a closed-ring tetradecamer, respectively. The native mass spectra of the N. vectensis CaMKII hub reveal the existence of a small population of hexadecameric species along with a predominant population of tetradecamers, the first indication of a CaMKII hub domain that shows higher stoichiometry beyond a tetradecamer. The hub domain of CaMKII from choanoflagellate Salpingoeca rosetta is 41% identical in sequence to that of human CaMKIIα. The hub assembly of S. rosetta crystallized as an open-ring spiral structure with sixfold screw symmetry (Fig. 5B; Bhattacharyya et al. 2016).

True CaMKII proteins are only found in metazoans and the choanoflagellates. Unexpectedly, several bacteria and green algae contain genes that code for proteins with ∼50% sequence similarity to the human CaMKIIα hub (McSpadden et al. 2019). These bacterial and algal proteins lack the corresponding kinase domains, so the function of these “CaMKII hub” proteins is unclear. The structure of the CaMKII hub assembly from the green algae Chlamydomonas reinhardtii revealed that the hub forms a closed ring-like structure similar to the human oligomeric hub, but with 18 subunits in the oligomeric ring assembly (McSpadden et al. 2019). CaMKII hubs from two other green algae, Volvox carteri f. nagarensis and Gonium pectoral were inspected by native mass spectrometry. They form 18-subunit and 20-subunit assemblies, in an 8:1 ratio, and 16-subunit and 18-subunit assemblies, in a 5:2 ratio, respectively. The CaMKII hub from the bacterium Pirellula sp., which has 52% sequence identity to the human CaMKII hub, forms 14-subunit and 4-subunit assemblies, where only the 14-subunit assembly is expected to form the closed ring.

All of the CaMKII hub assemblies that have been characterized so far have highly conserved interfaces between adjacent subunits. An interesting question is how the hub can alter its oligomeric assembly or geometry while maintaining the interfacial interactions that hold the assembly together. Changes in the oligomeric state of the hub assembly are correlated with an increased curvature of the β-sheet in the hub domain with respect to the α-helical layer (Fig. 5A; Bhattacharyya et al. 2016). Interfacial interactions in the hub occur between the β-sheet in one subunit and helix D in an adjacent subunit. When the β-sheet curvature changes, helix D of the adjacent subunit has to undergo a concerted motion to conserve the interface packing. This change in the orientation of β-sheet and helix D leads to changes in the quaternary assembly and geometry of the hub.

Such conformational changes can explain the transition from the dodecameric to the tetradecameric closed-ring forms of the human CaMKII hub (Fig. 5C; Bhattacharyya et al. 2016). The spiral architecture of the S. rosetta CaMKII hub is also the result of a drastic bending of this β-sheet (Fig. 5D). The higher-order oligomeric state of the Chlamydomonas hub is correlated with increased hydrogen bonding within the hub domain, which leads to an increased curvature of the β-sheet and thus stabilization of the 18-subunit closed-ring state (McSpadden et al. 2019). The introduction of six Chlamydomonas residues into the human CaMKIIα hub allows it to mimic this increased hydrogen-bonding network seen in Chlamydomonas. Interestingly, this mutant human hub adopts 14-subunit and 16-subunit assemblies in a 7:1 ratio as determined using native mass spectrometry, suggesting that small variations in sequence can alter the hub stoichiometry (McSpadden et al. 2019).

ACTIVATION-TRIGGERED SUBUNIT EXCHANGE IN CaMKII

CaMKII has been shown recently to display an interesting phenomenon in which activation leads to the colocalization, or mixing, of subunits from different holoenzymes (Stratton et al. 2014; Bhattacharyya et al. 2016). Subunit exchange entails the exchange of activated subunits between two activated, or an activated and an unactivated holoenzyme, as observed by colocalization of differently labeled CaMKII subunits after mixing of activated holoenzymes (Fig. 6A,B). Such colocalization was minimal when CaMKII was not activated.

Figure 6.

Activation-triggered subunit exchange in CaM-dependent protein kinase II (CaMKII). (A) Schematic diagram depicting the phenomenon of the activation-triggered subunit exchange in CaMKII. (B) A representative single-molecule total internal reflection fluorescence image showing colocalization of subunits, with red and green channels overlaid. (Panel B created from data in Stratton et al. 2013.) (C) Linker residues in a Nematostella vectensis CaMKII hub (PDB: 5IG5) docked on the hub interface extending the β-sheet of the hub subunit. (D) The cartoon depicts the distortion seen in the structure of the N. vectensis CaMKII hub upon linker docking at the interface. The tetradecameric spiral arrangement of the subunits is shown in the diagram, with each subunit labeled. (Panels A, C, and D are adapted from Bhattacharyya et al. © 2016 under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided the original author and source are credited.) (E) Schematic depiction of the three-way competition for the interaction with the CaM-binding element.

This phenomenon may provide a mechanism of input signal amplification, whereby the activation signal can propagate from a small population of activated species to a large population of unactivated species (Lisman 1985; Miller and Kennedy 1986; Lisman and Raghavachari 2015). A recent computational analysis has elucidated the potential importance of subunit exchange in enhancing information retention by CaMKII, suggesting that this phenomenon may play a role in learning and memory consolidation (Singh and Bhalla 2018).

The natural variation in the stoichiometry and geometry of the oligomeric assembly in CaMKII holoenzyme provides a plausible molecular mechanism for activation-triggered subunit exchange (Bhattacharyya et al. 2016). The dodecameric and tetradecameric forms of human CaMKII may interconvert by the reversible addition and release of vertical dimeric subunits, possibly mediated by a ring-open spiral assembly (Bhattacharyya et al. 2016). The S. rosetta CaMKII can capture and release dimeric subunits and undergo spontaneous subunit exchange without activation (Bhattacharyya et al. 2016).

Recent structural studies of the CaMKII hub from a sea anemone, N. vectensis (PDB: 5IG5) have revealed that a peptide segment can extend the open β-sheet at the hub interface, thereby destabilizing the integrity of the hub and opening the closed ring (Fig. 6C,D). This peptide-mediated β-sheet extension is proposed as a generalized mechanism for hub ring opening and disassembly in CaMKII (Bhattacharyya et al. 2016). Mutations in the CaM-binding element block subunit exchange in constitutively active CaMKII (T286D), suggesting that the CaM-binding element might be responsible for destabilizing the hub when released from the kinase domain (Stratton et al. 2014). A peptide corresponding to the CaM-binding element binds to the hub with a dissociation constant of ∼50–100 μm (Bhattacharyya et al. 2016), but a direct demonstration of hub destabilization by such a peptide remains to be established.

These recent findings suggest that there is a three-way competition for the CaM-binding element of CaMKII (Fig. 6E). In the absence of Ca²⁺/CaM and phosphorylation, the CaMKII regulatory segment is bound to the kinase domain (Chao et al. 2010). Upon activation by a Ca²⁺ transient, the CaM-binding element on the regulatory segment interacts with Ca²⁺/CaM (Rellos et al. 2010). This is followed by autophosphorylation at Thr 286, which eliminates the rebinding of the regulatory segment on the kinase domain even in the absence of Ca²⁺/CaM. Withdrawal of Ca²⁺/CaM after the passage of the Ca²⁺ transient and autophosphorylation on Thr 305 and Thr 306 prevents the rebinding of Ca²⁺/CaM to the regulatory segment. Under this condition, the only interaction partner for this peptide element is the low affinity binding site on the hub interface (Fig. 6E; Bhattacharyya et al. 2016). This three-way competition for the engagement of the CaM-binding element is hypothesized to be crucial for the regulation of activation-triggered disassembly and subunit exchange in CaMKII.

CONCLUDING REMARKS

A detailed structural understanding of the overall architecture of autoinhibited CaMKII holoenzyme has emerged in the past two decades, from studies that consider the kinase domains and the hub assemblies separately, as well as studies of the CaMKII holoenzyme. These structural studies have mainly focused on the autoinhibited forms of the CaMKII holoenzyme. Understanding the structural changes that are associated with the activation of the CaMKII holoenzyme awaits future studies. The role of the intrinsically disordered kinase-hub linkers, which distinguishes different CaMKII isoforms and splice variants, in regulating CaMKII activation and function is not well understood. Identification of in vivo substrates of CaMKII, notably anchoring proteins, and their role in docking CaMKII at functionally important sites (such as synapses) is another area that warrants further investigation. Studying the molecular mechanism of activation-triggered subunit exchange and establishing its physiological relevance is yet another important avenue for future investigation.