Introduction

Intracellular Ca2+ is a quintessential intracellular messenger, and many of its cellular effects are transduced by calmodulin (CaM). The spatiotemporal characteristics of these intracellular Ca2+ signals hinge on a complex interplay between Ca2+ influx and efflux mechanisms, mobile vs. stationary Ca2+-binding proteins (such as CaM) and intracellular localization/partitioning or sequestration. Over the past decades, great strides have been made in understanding the local control mechanisms underlying the global Ca2+ oscillations (ranging from 0.1 to 1 μM) in the beating heart [1]. It has also become clear that multiple Ca2+ signaling events, independent of those governing cardiac contraction, occur simultaneously in the same myocyte and control myriad cellular outcomes [2]. Several microdomain Ca2+ signals leading to altered transcription have now been identified, e.g., caveolae-localized L-type Ca channels provide the Ca2+ source for the Ca2+-CaM-dependent activation of calcineurin-nuclear factor of activated T cell (NFAT) signaling, and inositol triphosphate receptors (InsP3R) that release Ca2+ can activate Ca2+-CaM-dependent kinase (CaMKII) related to histone deacetylase (HDAC) signaling [3, 4]. Here, we report how recent advances in fluorescent biosensors can provide unique insights into the spatiotemporal dynamics of these local Ca2+-CaM signals, how they are sensed and decoded to shape particular cellular responses.

Calmodulin: a tale of specificity and promiscuity

CaM acts as a pivotal transducer of Ca2+ signals for a wealth of cellular functions including excitation–contraction coupling, metabolism, cell survival, and transcriptional regulation. How CaM regulates such a diversity of signaling pathways with distinct spatial and temporal outcomes remains a central question. Nonetheless, key aspects of CaM signal transduction mechanisms can be explained by the biochemical properties of CaM: its Ca2+-binding properties, localization, concentration and mobility inside the myocyte, and also its binding kinetics to target proteins.

CaM interactions

The three-dimensional crystal structure of Ca2+-free CaM (apoCaM) reveals a dumbbell-like shape (Fig. 1): a flexible, central α helix flanked by two globular lobes (the N- and C-lobes) [5, 6]. Each lobe contains two coupled EF-hand motifs with the C-lobe having higher Ca2+ binding affinity (K d ≈ 1 μM) than the N-lobe (K d ≈ 12 μM) [7–9]. The lobes also differ in Ca2+ dissociation kinetics (∼1,000 s−1for the N-lobe and 100 times slower for the C-lobe). Upon Ca2+ binding, CaM undergoes a profound conformational change (becoming a more elongated structure), modulating the interaction with its target proteins [10]. The structural flexibility of CaM explains the abundance and structural heterogeneity of CaM interacting proteins or interactome [11–14]. These CaM-binding targets differ in their preference for apoCaM, and Ca2+-CaM can be activated or inhibited upon CaM binding and can themselves influence Ca2+ sensitivity of CaM [15, 16]. Target proteins of note in the heart include ion channels (L-type Ca channels (LTCC) [17]), ryanodine receptors (RyR) [18], InsP3R [19], the protein phosphatase calcineurin (CaN) [16], CaMKII, nitric oxide synthase, adenylyl cyclase (AC), phosphodiesterase (PDE), myosin light chain kinase (MLCK), etc. [20].

Fig. 1
figure 1

Ca2+-dependent localization of CaM in cardiomyocytes. a Structures of Ca2+-free CaM (apoCaM) and Ca2+-bound CaM where the arrows indicate the Ca2+ binding sites. Image was reproduced from ref [20]. b and c CaM binding increases at the nucleus and Z lines with higher [Ca2+] i . Fluorescent CaM (60 nM) was washed into permeabilized myocytes at 100 and 500 nM [Ca2+] i . Confocal images were taken to monitor CaM binding. After 40 min of fluorescent CaM wash-in (shown in b), fluorescence intensities were quantified at M lines, Z lines, and nucleus (normalized to fluorescence intensity at M lines at 100 nM [Ca2+] i ) for [Ca2+] i of 100 and 500 nM. Image reproduced from ref [26] with permission. d Ca2+ regulates both functional populations of CaM: the dedicated CaM pool in complex with its target channels as well as the promiscuous CaM pool regulating CaMKII and CaN

Full size image

In vivo, CaM tethering to target proteins can be monitored with fluorescence resonance energy transfer (FRET), even with simultaneous [Ca2+] using Ca2+ indicators [21]. Our group used a variant of this approach in adult cardiac myocytes to map CaM binding to the RyR by monitoring FRET between fluorescently labeled FKBP12.6 bound to RyR and fluorescent CaM [22]. CaM binding was found to stabilize the zipped conformational state of the RyR (reducing RyR activation). Interdomain “unzipping” of the RyR by treating the cells with a peptide DPc10 that mimics one of the zipped regions, also reduced CaM binding and enhanced RyR gating, Ca2+ sparks and waves that favor arrhythmogenic events and unloading Ca2+ from the sarcoplasmic reticulum (SR), thereby mimicking aberrant channel activation as seen in heart failure [23].

CaM (trans)location and availability

Surprisingly, data regarding CaM location and redistribution remain scarce in adult cardiomyocytes. Total CaM concentration is ∼6 μM in isolated cardiac myocytes [24], and at resting [Ca2+] levels, most CaM is apoCaM (not Ca2+ bound). Some of this apoCaM is already prebound to targets as structural elements of the protein or is poised nearby the target awaiting local activation (part of the signaling scaffold). Using null-point titration methods with fluorescently labeled CaM, Wu and Bers found that, as in other cell types, free CaM is around 50–75 nM [25, 26]. That is, only ∼1 % of total CaM is free, which implies that dynamic competition for CaM among its targets is an integral part of the CaM signaling mechanism. In this “limiting CaM” scenario, target proteins with higher Ca2+-CaM affinities (e.g., CaN, PDE, MLCK with K d ≤ 1 nM) will be preferentially activated compared to those targets with lower affinity (e.g., CaMK, AC with K d > 10 nM).

At resting Ca2+ (100 nM [Ca2+] i ), most intracellular CaM is bound to cellular constituents and appears highly localized to the Z line (or transverse tubule) (Fig. 1) [26]. High levels of CaM at the junction between the sarcolemma and SR membrane, where LTCCs and RyRs are both concentrated, facilitate CaM-dependent modulation of both Ca2+ current and SR Ca2+ release. Indeed, such regulatory CaM is already bound to the LTCC and RyR where upon LTCC Ca2+ influx and SR Ca2+ release, locally prebound apoCaM becomes Ca2+-CaM and mediates Ca2+-dependent inactivation of Ca2+ current and favors RyR closure. Thus, Ca-CaM is part of a highly local negative feedback system at the junctional cleft at the heart of excitation–contraction coupling.

Junctional targeting of CaM would likewise facilitate activation of lower CaM affinity targets such as CaMKII, by enhancing the latter’s ability to compete with higher affinity global targets. This may relate to Ca2+-CaM-dependent activation of LTCC via CaMKII activity (Ca2+ current facilitation) in which Ca2+ current amplitude increases (and inactivation slows) in a stepwise manner during increases in stimulation rate [27–29]. This facilitation may help sustain Ca current amplitude at higher heart rates when Ca2+-dependent inactivation might otherwise limit it.

When [Ca2+] is elevated (from 100 to 500 nM), more CaM is bound at the Z lines (where the junctional SR is), and from which CaM dissociation is extremely slow, but there is also a significant translocation of CaM into the nucleus [26]. Normal twitch Ca2+ oscillations did not trigger this nuclear redistribution [4]. This indicated that the kinetics of Ca2+-CaM signaling may be critical for this CaM translocation. Endothelin-1 binding to its G-protein-coupled receptor (GPCR) also causes nuclear translocation of CaM, without triggering a detectable elevation of global [Ca2+] i [4]. Analogous to the RyR environment above, it is possible that CaM at or near InsP3R downstream of endothelin-1 might stimulate this CaM translocation. So, chronic global Ca2+ elevation can trigger nuclear CaM accumulation, but certain local Ca2+ signals can achieve the same effect. Changes in nuclear CaM are likely to be of pathophysiological importance given its regulation of nuclear CaMKII activity (see below), but also because of direct effects on myocyte enhancer factor 2-dependent transcription [30]. In neurons, nuclear CaM was also shown to affect phosphorylation of cAMP response element binding protein [31]. How CaM translocates to the nucleus in cardiomyocytes remains to be identified [32]. In neurons, this nuclear import occurs in association with partner proteins [33], and this is probably also true in the heart given the low free CaM levels (and competition for CaM binding). A more thorough understanding of what type of signals trigger CaM nuclear import could help to clarify things. Another outstanding question is how CaM expression and localization may be altered in cardiac disease. There are conflicting reports regarding CaM expression levels in human heart failure [34, 35] and even less information on CaM localization. Alternatively, association of both apoCaM and Ca2+-CaM with target proteins could be altered. For instance, Ai et al. demonstrated in their rabbit heart failure model that although the total CaM levels were unchanged, CaM association with the RyR was decreased (consistent with increased RyR activity or SR Ca2+ leak) [36]. In that context, if in failing myocytes CaM was found to be redistributed to the nucleus, it could influence gene transcription and alter contractile function in the various etiologies of hypertrophy and heart failure. Gangopadhyay and Ikemoto, albeit in neonatal myocytes, suggested that this might be the case. Here, dantrolene treatment of neonatal myocytes, which ameliorates RyR unzipping and aberrant Ca2+ events following prolonged endothelin-1 exposure, reduced nuclear translocation of CaM and the development of hypertrophy [37].

Concept of “dedicated” vs. “promiscuous” CaM

An emerging theme for CaM modulation of ion channels in cardiac myocytes is the multiple modes of regulation mediated by distinct functional populations of CaM (Fig. 1): a dedicated pool of CaM typically tethered to the channels, which directly regulates channel activity (most often involved in inactivation), and a more promiscuous pool of CaM poised nearby to transduce local Ca2+ signals via activation of downstream effectors such as CaN and CaMKII (providing a negative or positive feedback loop). These intricate levels of CaM regulation of channel function were recently reviewed in detail [38] for the L-type Ca2+ channel, the ryanodine receptor, and the InsP3 receptor but are beyond the scope of this review. In general, the mechanisms of promiscuous CaM regulation are less clear, but from an imaging perspective, several of the remaining questions can be addressed such as: is this promiscuous CaM activated to act locally or does it diffuse from more distant sources? Are the Ca2+signals transduced by the dedicated (tethered) and promiscuous (more global) CaM the same? What is the reservoir/source of this promiscuous CaM pool (e.g., which proteins release CaM with Ca2+ influx)? Several of these events are likely happening within the single resolvable volume of traditional confocal microscopy. The technical challenge to overcome here is the spatial resolution limitation, and this may require novel experimental approaches or a combination thereof: total internal reflection fluorescence microscopy which limits the fluorescence signal to ∼100 nm from the coverslip, FRET imaging of dynamic protein interactions, subcellular targeting of biosensors, super resolution fluorescence microscopy (PALM and STORM).

Dynamics of CaM signals

As mentioned above, CaM target proteins vary greatly in their affinity for CaM (with dissociation constants ranging from 0.1 to >100 nM [39–41]. The Ca2+-CaM affinity of the target protein will determine its Ca2+ dependence (greater for those that bind Ca2+-CaM tightly). The Persechini group developed FRET-based biosensors (BsCaMs) for dynamic measurements in vivo of Ca2+-CaM interaction with CaM target proteins in response to Ca2+ [32, 42–45]. Changes in FRET are detected between the CFP and YFP fluorophores which flank a linker sequence derived from the Ca2+-CaM binding domain of the smooth muscle MLCK (Fig. 2). Site mutations of the linker sequence confer different Ca2+-CaM sensitivities to the sensors [42]. In cardiomyocytes, these sensors can be used to measure the real-time modulation of CaM targets by various Ca2+signals including the beat-to-beat oscillations [24]. Sensors with higher affinity (like CaN) bind Ca2+-CaM more tightly and therefore integrate the rhythmic Ca2+ signals, whereas sensors with lower affinity (like CaMKII) turn on and off more completely with each beat and display only minor integration [46] (Fig. 2). These different behaviors could also be predicted quantitatively from a myocyte model [46, 47] which implies that differences in Ca2+-CaM affinities of CaM target proteins provide an important mechanism for accomplishing the CaM signaling diversity and complexity: Ca-CaM affinity confers specific Ca2+-decoding ability to the CaM target protein thereby fine-tuning the cellular response.

Fig. 2
figure 2

Visualizing Ca2+-CaM signals in cardiomyocytes. a Principle of the FRET-based Ca2+-CaM biosensors that mimic either high- (BsCaM2) vs. low (BsCaM45)-affinity CaM targets such as CaN and CaMKII, respectively. b and c BsCaM FRET changes (F CFP/F YFP) in response to pacing reveal that both types of CaM targets can track the beat-to-beat intracellular Ca2+ changes, but in different ways: BsCaM2 integrates the oscillatory Ca2+signals more strongly, whereas BsCaM45 turns on and off more completely with each Ca2+ transient. Reproduced with permission from ref [32]

Full size image

Targeted versions of these biosensors are currently being used by our group to verify model projections of how local CaM signals might regulate CaMKII and CaN [47]. While large Ca2+-CaM oscillations are predicted to occur at the dyadic cleft on a beat-to-beat basis, the extensive CaM buffering and lower cytosolic Ca2+ would minimize the Ca-CaM fluctuations there. For CaMKII, this would mean not only beat-dependent oscillations of activity in the dyadic cleft but also very little activation of CaMKII in the bulk cytosol because of the limited CaM signals in that compartment. It is hard to reconcile this prediction with known CaMKII-dependent phosphorylation of cytosolic targets such as phospholamban, although this may be why CaMKII is routinely anchored nearby its targets—and might have a nearby pool of promiscuous CaM. CaN activity on the other hand is predicted to be frequency dependent in the cytosol and saturated in the dyad. This is consistent with observations that CaN/NFAT signaling correlates closely to global Ca2+ oscillation [48, 49]. Since CaN is tethered to LTCCs [50], it will be important to assess whether CaN activity at the dyad is indeed constitutively active or remains somewhat dynamically regulated by Ca2+ signals.

CaMKII: a multifunctional kinase

While CaMKII is uniquely adapted to sense, integrate, and transduce cellular Ca2+ (and Ca2+-CaM) signals, it is much more than merely a CaM effector. CaMKII is a key regulator of cardiac excitation–contraction coupling via its effects on ion channels and Ca2+ handling proteins, but is also important for more chronic cardiac responses via regulation of gene expression. Aside from this physiological role, CaMKII is also critically involved in the development of arrhythmias and heart failure. Several reviews have highlighted the role of CaMKII in various pathophysiological conditions and its potential as a therapeutic target [51–58].

Biochemistry of CaMKII

CaMKII is unique among protein kinases because it forms a dodecameric holoenzyme. It is encoded by four different genes (α, β, δ, and γ) [59, 60]. Of these, CaMKIIδ is the predominant cardiac isoform, and two splice variants are known to be present in the heart (δB and δC). Each subunit of the holoenzyme consists of a serine/threonine-specific catalytic domain followed by a regulatory segment that binds Ca2+-CaM (Fig. 3). The latter is followed by a flexible linker of variable length (isoform-specific) that connects to the association or hub domain. The association domains assemble the subunits into two stacked hexameric rings. FRET experiments, where the catalytic domain was replaced with CFP or YFP, showed that CaMKII freely hetero-oligomerizes and that the variable linker influences the distance between N-termini [61]. The recent crystal structure of the auto-inhibited CaMKII holoenzyme indicated that this variable linker also provides a fine-tuning mechanism of the holoenzyme Ca2+ response: the variable linker alters the equilibrium between compact and more open docking states of the kinase domains against the central hub (with the CaM binding sites more accessible for the longer linkers) [62]. Indeed, different isoforms display slightly altered sensitivities to Ca2+ [63], but how this translates to signal diversity in the cardiomyocyte remains to be established.

Fig. 3
figure 3

Visualizing CaMKII signaling in cardiomyocytes. a Structural organization of the CaMKII holoenzyme (modified from ref [20]). The CaMKII monomer consists of three different domains: the association domain for assembly into the holoenzyme, the catalytic domain, and the regulatory domain. CaMKII activation requires Ca2+-CaM binding to the regulatory domain relieving the autoinhibition of the kinase. The associated conformational change now permits activity-sustaining modifications of the kinase such as phosphorylation (T286) and oxidation (MM281/282). b The FRET sensor Camui consists of the full-length CaMKII sandwiched between a fluorophore FRET pair. Activating conformational changes result in a reduction of FRET. c Localization of Camui in cardiomyocytes using CFP and YFP emission signals confirms targeting of Camui to Z and M lines (left panel). The right panel is a FRET ratio image of Camui showing spatial differences in CaMKII activation (in the nucleus, perinucleus, and cytosol). Panels b and c were from ref [89] with permission

Full size image

At baseline, CaMKII is in an auto-inhibited state where the catalytic domain is constrained by the pseudosubstrate region within the regulatory domain [64] (Fig. 3a–b). When Ca2+ rises, Ca2+-CaM binds to the regulatory domain of CaMKII and triggers a conformational change that disrupts association with the catalytic domain and exposes an otherwise inaccessible phosphorylation site (T286) [64]. CaMKII deactivates again upon Ca2+-CaM dissociation. During prolonged [Ca2+] elevation, autophosphorylation at T286 prevents reassociation of the regulatory and catalytic domains, resulting in autonomous activity that persists in the absence of Ca2+-CaM [65, 66]. Precisely how CaMKII activity responds to the intensity, duration, and frequency of Ca2+ transients or how acute Ca2+ signaling and GPCR-linked Ca2+ signals are integrated on CaMKII are ongoing research foci. Reactive oxygen species (ROS) can also trigger autonomous CaMKII activity by oxidizing M280/281 residues after initial activation by Ca2+-CaM [67]. Both mishandling of Ca2+ and changes in ROS have been linked to cardiac disease, positioning CaMKII at a key interface [53].

Localization of CaMKII

CaMKII isoforms differ in subcellular targeting as well as interacting partners [68–70] and is purported as a key aspect of isoform signal specificity. In the heart, the CaMKII δB and δC subtypes differ only by the presence of a nuclear localization sequence and are thought to preferentially localize accordingly: δB mainly nuclear and δC mainly cytosolic [71–73]. Different functions had been attributed to each isoform, with nuclear δB involved in hypertrophic gene regulation and cytosolic δC in regulation of ion channels and Ca2+ handling. Initial findings in transgenic mice supported this since CaMKII δB transgenics develop cardiac hypertrophy, but this hypertrophy rapidly transitions to heart failure in the CaMKIIδC mice (with extensive Ca2+ handling dysfunction) [73–77]. However, all CaMKII isoforms (α–δ, including these splice variants) appear to hetero-multimerize [61] which can explain in part why CaMKIIδB is not restricted to the nucleus and CaMKIIδC is not exclusively cytosolic [78]. Indeed, CaMKIIδB and δC were found to have similar effects on MEF2 modulation and gene expression, but initially, this was attributed to direct effects of CaMKIIδB and δC subtypes on HDAC in the nucleus and cytosol, respectively [79]. More recently, Mishra et al. [78] upset this traditional dogma of isoform-specific location and postulated that spatial and functional specificity of CaMKIIδ activation is obtained by mobilization of different Ca2+ stores [78]. This was based on their findings that even when expressed in CaMKIIδ null background, CaMKIIδB and δC subtype targeting remained nonexclusive, and both subtypes had compartment-specific activation profiles: a phenylephrine stimulus triggered nuclear CaMKIIδ activation and HDAC phosphorylation, whereas caffeine elicited cytosolic CaMKIIδ activation and phospholamban phosphorylation. To what extent these findings will impact the notion of CaMKII δB as having a cardioprotective role and δC as having a more deleterious role in cardiomyocyte survival and heart disease [80–83] remains to be established.

It is clear that CaMKII location matters and several direct interactions of CaMKII with target proteins have been described such as with the RyR [74, 84], InsP3R [85], LTCC subunits [28, 29], and mitochondria [86]. Indeed, CaMKII overexpression potentiates SR Ca2+ release and also enhances mitochondrial Ca2+ uptake that can contribute to cell death [87]. Localization poises CaMKII near both local Ca2+ pools and a signaling microdomain, thereby achieving signal specificity (e.g., the InsP3R-CaMKII-HDAC signaling pathway [4]). This direct interaction pattern of CaMKII is in contrast to the scaffolding protein interactions found for PKA (A-kinase anchoring proteins) [88]). However, α kinase anchoring protein was found to serve as an “anchoring” protein for SERCA and CaMKII, positioning CaMKII for modulating PLB phosphorylation [89]. To what extent these interacting proteins modulate CaMKII function is still unclear. In neurons, α-actinin can act as a Ca2+-independent surrogate for CaM binding to CaMKII and also modulates CaMKII phosphorylation of specific substrates (such as glutamate receptor) [90]. In essence, α-actinin acted as a “phosphostat” for a subset of CaMKII targets, maintaining their basal phosphorylation. Other outstanding questions include to what extent CaMKII is translocated in cardiac myocytes and whether CaMKII interactions, localization, and translocation are altered in cardiac disease.

Imaging CaMKII activity and regulation

Until recently, the tools for measuring CaMKII activation in cardiomyocytes were very limited and destructive. These included snapshot measurements of CaMKII modifications (i.e., its autophosphorylation or oxidation state using Western blotting or immunocytochemistry) or CaMKII target phosphorylation and kinase assays in cell lysate (32P incorporations or fluorescence quenching). The Hayashi group developed Camui to study CaMKII activation in neurons [91]. This FRET-based biosensor consists of the full-length kinase flanked by the CFP/YFP fluorophores and takes advantage of the conformational changes of the kinase associated with activation. It is a powerful tool for detecting subtle, subcellular differences in CaMKII activation, so our group adapted it for use in cardiomyocytes [92]. The donor/acceptor fluorescence ratio (F CFP/F YFP) of Camui is a good reporter of CaMKII activation by Ca2+-CaM, autophosphorylation, and oxidation (since all result in an opening of the regulatory and catalytic CaMKII domains). That is, Camui is more than a readout of Ca2+-CaM binding dynamics to cellular CaMKII (which could be inferred from the BsCaM responses), it is a direct readout of the active conformation of the kinase.

It should be noted that although the Camui response is fairly linear over a wide range for autonomous CaMKII activity (via phosphorylation or oxidation), direct Ca2+-CaM activation produces a stronger FRET change for a given amount of enzyme activity. A second limitation is that the Camui FRET is not simple bimolecular FRET but incorporates a structural element (the CFP on the catalytic domain can interact with more than one YFP on the association domain). This may contribute to some nonlinearity in F CFP/F YFP increase as catalytic domains become activated. Nonetheless, Camui is a good proxy for CaMKII activity. It can track both temporal and spatial changes in CaMKII activity in vivo and has the added benefit that the specific contributions of autophosphorylation and oxidation to CaMKII activation can be examined using Camui constructs lacking the target sites for these modifications. This makes Camui uniquely suited to study the subcellular mechanisms driving CaMKII signaling pathways in cardiomyocytes. Indeed, initial experiments showed that while there is a significant role for oxidative activation of CaMKII in response to angiotensin II and endothelin-1 exposure, the isoproterenol- and phenylephrine-triggered activation of CaMKII was largely Ca-CaM and somewhat autophosphorylation dependent.

Camui studies examining the sensitivity of CaMKII function to varying beat-to-beat global Ca2+ signals are already ongoing, as are experiments assessing how CaMKII responds to multiple spatially unique Ca2+ signal inputs (global cytosolic Ca2+ and GPCR-linked signals). Improvements to the Camui sensor also continue to be made [93, 94]: alternate CaMKII isoforms, improved FRET pairs, etc. An alternative to the Camui-style biosensors is a surrogate substrate-based reporter (Kinase Activity Reporter family (KAR), [95]) where a forkhead association domain triggers a molecular switch upon phosphorylation by the target kinase, and this would induce changes in fluorescence intensities of the FRET fluorophores. These KARs have the added benefit that they can be subcellularly targeted to signaling microdomains providing even more detailed spatial information of kinase function. Future studies combining Camui and various models of cardiac disease are likely to provide extensive mechanistic insight into how CaMKII signaling contributes to arrhythmias and heart failure.

Concluding remarks and future perspectives

Numerous studies have shown that Ca2+-CaM-CaMKII signaling has a critical role in the regulation of both acute excitation–contraction and chronic excitation–transcription coupling in the heart. There is also abundant evidence that CaMKII has a ubiquitous role in pathophysiological conditions of the heart. Several fundamental questions regarding the signal processing and spatial intricacies of CaM/CaMKII signaling were highlighted here and deserve further investigation. The new generation of biosensors also provides unique opportunities to delineate the role and regulatory mechanisms of CaM and CaMKII in normal and diseased hearts. This will greatly benefit future efforts to design effective therapeutics targeted at cardiac CaMKII.