Calmodulin in a Heartbeat

Authors

  • Anders B. Sorensen,

    1. Department of Biotechnology, Chemistry and Environmental Engineering, Aalborg University, Denmark
    Search for more papers by this author
    • These authors contributed equally to this work
  • Mads T. Søndergaard,

    1. Department of Biotechnology, Chemistry and Environmental Engineering, Aalborg University, Denmark
    Search for more papers by this author
    • These authors contributed equally to this work
  • Michael T. Overgaard

    Corresponding author
    1. Department of Biotechnology, Chemistry and Environmental Engineering, Aalborg University, Denmark
    • Correspondence

      M. T. Overgaard, Department of Biotechnology, Chemistry and Environmental Engineering, Aalborg University, Sohngaardsholmsvej 49, 9000 Aalborg, Denmark

      Fax: +45 9635 0558

      Tel: +45 9940 8525

      E-mail: mto@bio.aau.dk

      Website: www.bio.aau.dk

    Search for more papers by this author

Abstract

Calmodulin is the primary sensor of intracellular calcium (Ca2+) levels in eukaryotic cells playing a key role in the proper deciphering of Ca2+ signalling. Given the versatility of Ca2+ as a secondary messenger, it is not surprising that calmodulin interacts with a vast number of proteins. Calmodulin is an extraordinarily conserved protein, which has not evolved since the genesis of the vertebrate lineage, and further is encoded by three different non-allelic genes in the human genome. The protein displays a high degree of conformational plasticity, allowing for target proteins to evolve specific modes of calmodulin interaction and regulation during Ca2+ sensing. The recent identification of two calmodulin mutations giving rise to a heart arrhythmia with catecholaminergic polymorphic ventricular tachycardia-like symptoms and sudden cardiac death in young individuals, and the following identification of another three calmodulin mutations linked to recurrent cardiac arrest in infants, is in many ways intriguing. How can mutations result in cardiac-specific phenotypes when calmodulin is fundamental for correct Ca2+ signal interpretation in virtually all cells in vertebrate organisms? Are there specific cardiac target protein interactions that are affected by these mutations? Another challenge is to elucidate how one mutated allele out of six encoding an identical calmodulin protein results in a dominant phenotype. Here we aim to give an overview of components in the cardiac contraction cycle whose function is modulated by calmodulin. In principle, these may all be implicated in the pathogenic molecular mechanism linking calmodulin mutations to cardiac arrhythmia and sudden cardiac death.

Abbreviations
[Ca2+]cyt

cytosolic Ca2+ concentration

[Ca2+]mit

mitochondrial matrix Ca2+ concentration

AC

adenylyl cyclase

AP

action potential

apoCaM

Ca2+ free calmodulin

AV node

atrioventricular node

BrS

Brugada syndrome

CaCaM

Ca2+ bound calmodulin

CALM

human calmodulin gene

CaMBD

calmodulin binding domain

CaM

calmodulin

CaMKII

Ca2+/CaM dependent protein kinase II

CASQ2

calsequestrin 2

CaV1.2

voltage-gated L-type Ca2+ channel

CaV3.1

voltage-gated T-type Ca2+ channel

CDF

Ca2+ dependent facilitation

CDI

Ca2+ dependent inactivation

CICR

calcium induced calcium release

CPVT

catecholaminergic polymorphic ventricular tachycardia

DAD

delayed after depolarization

ECG

surface electrocardiogram

esLQT

early onset severe long QT syndrome

FKBP12.6

FK506 binding protein 12.6

GA

Golgi apparatus

HRC

calumenin and histidine rich Ca2+ binding protein

ICa

Ca2+ current

If

inward pacemaker funny current

IK

K+ current

INa

Na+ current

IP3R

inositol-triphosphate receptor

Kd

dissociation constant

KV

voltage-gated K+ channel

LQT

long QT syndrome

mtNOS

mitochondrial nitric oxide synthetase

mtRYR1

mitochondrial RYR1

Nav

voltage-gated Na+ channel

NCX

Na+-Ca2+ exchanger

nNOS

neuronal nitric oxide synthetase

NScaTE

N-terminal spatial Ca2+ transforming element

PDE

phosphodiesterase

PKA

protein kinase A

PLN

phospholamban

PMCA

plasma membrane Ca2+ ATPase

PP

protein phosphatases

RYR1/2

ryanodine receptor 1/2

SA node

sinoatrial node

SERCA

sarco/endoplasmic reticulum Ca2+ ATPase

SOICR

store overload induced calcium release

SPCA

secretory pathway Ca2+ ATPase

SR

sarcoplasmic reticulum

VDI

voltage dependent inactivation

Introduction

In all eukaryotic cells, the cytosolic concentration of calcium ions ([Ca2+]cyt) is tightly controlled by a complex network of transporters, pumps, channels and binding proteins (reviewed in [1, 2]). Calmodulin (CaM) is probably the most essential intracellular Ca2+ sensor and signalling molecule, the importance of which is underscored by the presence of three independent genes (CALM1–3) encoding identical proteins in the human genome. Also highlighting the pivotal role of CaM is its remarkable degree of conservation with no amino acid changes introduced since the appearance of vertebrates [3, 4]. Ironically, differences in expression levels of the invariant CaM formed the basis of the beak shape differences of Darwin's finches, thus inspiring the very concept of variational evolution between species [5].

CaM is a relatively small 148 amino acid residue α-helical protein composed of N- and C-terminal lobes, each containing two Ca2+ binding EF-hands (Fig. 1). The N- and C-lobes display a 10-fold difference in Ca2+ affinity, with an apparent Kd of approximately 10 and 1 μm, respectively [6, 7]. Combined with both intra- and inter-lobe Ca2+ cooperativity, CaM has evolved into a complex Ca2+ sensing molecule, capable of reacting to changes in [Ca2+]cyt over a wide span. This span is even further expanded by target-specific changes in Ca2+ affinities and kinetics upon target binding [8, 9].

Figure 1.

Structure of CaM in various conformations. The mutated residues are shown in stick representation and coloured as follows: N53I red, D95 blue, N97 green, D129 purple-blue, F141 magenta. Ca2+ is shown in metallic grey as spheres. Top row: (A) apoCaM (1DMO), (B) CaCaM in an extended conformation (1CLL) and (C) CaCaM in a collapsed conformation (1PRW). Bottom row: CaM in complex with different peptides (magenta) derived from CaM interaction partners with IQ motifs (yellow) and hydrophobic anchors (orange); (D) ApoCaM in complex with a NaV1.5 peptide (2L53), (E) CaCaM in complex with a CaV1.2 peptide (2F3Y) and (F) CaCaM in complex with a RYR1 peptide (2BCX). Left side lettering indicates the orientation of CaM N- and C-lobes, respectively.

The fact that the N- and C-lobes display different binding kinetics, the N-terminal lobe binding faster than the high affinity C-lobe, further increases the versatility and possible complexity of how CaM can translate changes in [Ca2+]cyt to target proteins. One striking example of this is CaM regulation of the voltage-gated L-type Ca2+ channels (CaV1.2). CaM tethered to the channel can sense rapid, large changes in the local nano-domain Ca2+ concentrations with the C-lobe, whilst simultaneously sensing slow/longer lasting changes in the lower global [Ca2+]cyt with the N-lobe [10].

Ca2+ binding to CaM induces a conformational change which exposes hydrophobic patches in each lobe, enabling binding to large hydrophobic anchor residues present in canonical Ca2+ loaded CaM (CaCaM) target binding sequences (Fig. 1). The hydrophobic patches contain a large fraction of flexible methionine residues which, combined with a highly dynamic and flexible central linker between the two lobes, allows interaction with a multitude of different target proteins and sequences [11]. The classical CaCaM binding motifs are characterized by a stretch of approximately 20 amino acid residues with a high density of positively charged residues and contain two large hydrophobic anchor residues that are spaced at variable distances (e.g. 1–10 in CaMKII, 1–14 in myosin light-chain kinase, 1–16 in CaM kinase kinase, 1–17 in RYR1/2). By wrapping around these α-helical motifs, the N- and C-lobes of CaCaM can bind to the two hydrophobic anchors. Adding to the complexity of the CaM target interaction, binding to CaCaM binding motifs can be in either a parallel (N-lobe CaM binding N-terminal end of the motif) or anti-parallel orientation. The binding orientation is believed to be largely determined by the charge density of the motif, with the CaCaM C-lobe having a preference for the more positively charged end [11].

Another archetypical type of CaM binding motif, initially characterized as apoCaM binding, is the so-called IQ motif (IQXXXRGXXXR). These motifs are highly hydrophobic and basic in nature and as a consequence will often also bind CaCaM. In addition, CaM has been demonstrated to interact with several targets using a range of different binding modes not conforming to the classical CaM binding motifs (reviewed in [11, 12]).

The inherent flexibility and promiscuity of CaM binding allows for interaction with more than 300 reported target proteins [13-15]. Given that there is a limited amount of CaM in the cytosol, it is thus perhaps not surprising that the cytosolic free CaM concentration is an important factor for CaM dependent Ca2+ signalling [16]. Many, if not most, CaM targets are thus likely to require prebound or tethered apoCaM for effective CaM mediated Ca2+ sensing. Hence, a high affinity target–apoCaM interaction can lead to localized domains of Ca2+ signalling [17]. Another possible way for obtaining localized CaM mediated Ca2+ sensing is through pools of CaM mRNA being targeted to specific subcellular locations, by differences in their untranslated regions [18].

Recently, we described two human CALM1 missense mutations linked to a severe, dominantly inherited form of ventricular tachycardia [19]. The two identified CaM mutations, N53I and N97S (throughout this review, we use the mature CaM sequence numbering, i.e. without the initiator Met; see Table 1), are located one in each lobe of the dumbbell shaped CaM molecule. N53 is positioned on the solvent exposed surface of the N-lobe in α-helix 3, while N97 is one of the Ca2+ coordinating residues located in the C-lobe EF-hand loop III (Fig. 1). Initial analysis of these variants demonstrated diverging effects on CaM Ca2+ binding and an aberrant interaction with a CaM binding motif from the ryanodine receptor 2 (RYR2), although the latter only for the N97S variant and only at low physiological [Ca2+]cyt. Three additional CaM mutations have been described, D129G and F141L (both CALM1) and D95V (CALM2), since the identification of the first CALM1 mutations. These mutations were identified in infants with life-threatening ventricular arrhythmias combined variably with epilepsy and delayed neuronal development [20]. Similar to the N97S mutation, the D129G, F141L and D95V mutations showed markedly reduced CaM C-lobe Ca2+ affinity [19, 20].

Table 1. Missense mutations identified in calmodulin genes. SCD, sudden cardiac death; RCA, recurrent cardiac arrest; esLQT, early onset severe LQT; ND, not disclosed; ESP, Exome Sequenching Project [151]
cDNA sequence changeaAmino acid changebCommonly used colloquial nomenclaturecCALM geneSite of mutation (exon)Phenotypic expressionSource
  1. a Numbering with start codon A = +1 from GenBank NM_006888.4 (CALM-1), NM_001743.4 (CALM-2) and NM_005184.2 (CALM-3). b Numbering including the initiator Met residue. c Numbering of mature CaM without the initiator Met residue – used throughout this review.

c.161A>Tp.Asn54IleN53I CALM-1 3CPVT, SCDNyegaard et al. [19]
c.293A>Gp.Asn98SerN97S CALM-1 5CPVT, SCDNyegaard et al. [19]
c.287A>Tp.Asp96ValD95V CALM-1 5RCA, esLQTCrotti et al. [20]
c.389A>Gp.Asp130GlyD129G CALM-2 5RCA, esLQTCrotti et al. [20]
c.426C>Gp.Phe142LeuF141L CALM-1 5RCA, esLQTCrotti et al. [20]
c.29T>Cp.Ile10ThrI9T CALM-1 2NDESP
c.307G>Ap.Ala103ThrA102T CALM-3 5NDESP
c.427G>Cp.Val143LeuV142L CALM-1 5NDESP

Prompted by the observed cardiac phenotype in individuals carrying CALM1 and CALM2 mutations, we here aim to give an overview of the molecular processes underlying the generation of the cardiac action potential (AP) and the translation of this through a chemical Ca2+ signal into cardiac contraction. This overview will be the basis of the following short review of the role CaM plays in regulation of the proteins known to govern the cardiac contraction cycle.

With this review we aim to add to a number of previous reviews which have given excellent characterizations of the interactions between CaM and Ca2+ channels [1, 17, 21] and of the role of the Ca2+/CaM dependent protein kinase II (CaMKII) in cardiac contraction [22].

Calmodulin modulation of the heart action potential

Generation of the action potential

The biomechanical pumping of the human heart is governed by a tightly controlled sequence of coordinated APs leading to the rapid contraction of distinct groups of cardiac cells. The electrical signal is initiated by so-called pacemaker cells located in the sinoatrial (SA) node. These cells undergo spontaneous depolarization and thereby generate a propagating AP (Fig. 2A). The depolarizing current travels from the SA node to the atrial cardiomyocytes through high conductance intercellular gap-junctions and finally culminates in atrial excitation and contraction. The latter can be recognized as the P-wave on the surface electrocardiogram (ECG) (Fig. 2A). The electrical signal then spreads from the atria down to the atrioventricular (AV) node, where an ~ 0.1 s pause allows sufficient time to complete atrial contraction, thereby emptying blood into the ventricles. Leaving the AV node, the electrical signal travels through the bundle of His, moving along the left and right bundle branches and via the Purkinje fibres spreading into the ventricular myocytes. The depolarization and following contraction of the ventricular myocytes is observed as the QRS complex on a surface ECG with the repolarization seen as the T-wave (Fig. 2A) [23-25].

Figure 2.

(A) Illustration of the electrical activity of the human heart with the sinoatrial (SA) node in green, atrial myocyte conduction pathways in orange, atrial ventricular (AV) node in yellow and the conduction pathway of the ventricular myocytes in dark blue. (B) Schematic diagram of phases 0–4 during the ventricular myocyte AP with inward and outward directed ion currents. Punctured line for INa illustrates the role of IpNa and the dotted line for IK,ATP illustrates that the Kir6.1 channel is active only in response to changes in cell ATP levels. (C) Schematic diagram of the pacemaker cell AP for the SA node with outward and inward directed ion currents.

At the cellular level, the cardiac AP is generated by cyclic changes in inward ionic currents during depolarization (INa and ICa) and outward currents during repolarization (IK) (Fig. 2B). Five phases define the AP of the ventricular and atrial myocytes. The resting phase is characterized by high outward current (IK1) of potassium ions (K+) through the Kir2.1 potassium channels by which a stable resting membrane potential is maintained (Fig. 2B, phase 4). As the ventricular myocyte is excited by adjacent cells, an initial depolarization triggers the opening of voltage-gated sodium ion (Na+) channels (NaV1.5) resulting in a rapid transient inward Na+ current (INa) lasting a few milliseconds and initiating the AP upstroke (Fig. 2B, phase 0). This further depolarization of the membrane potential rapidly deactivates the majority of NaV1.5 channels and activates voltage-gated K+ channels (Kv4.3 and Kv1.4, the latter primarily found in the atrium) generating a transient rapid repolarizing current (Ito) (Fig. 2B, phase 1). A fraction of NaV1.5 channels show much slower inactivation kinetics (100–200 ms) remaining active during AP phases 2 and 3. These generate a persistent Na+ current (IpNa, also known as a late Na+ current) contributing to maintaining the observed plateau in the cardiac AP phase 2. Activation of the CaV1.2 permits an inward directed Ca2+ current (ICa,L) which is electrically balanced by ultra-rapidly (mainly in atrial myocytes), rapidly and slowly delayed outward rectifying K+ currents (IKur, IKr and IKs, respectively) through activation of the Kv1.4, Kv11.1 and Kv7.1 channels, respectively, resulting in the observed plateau in the AP (Fig. 2B, phase 2). Closing of the CaV1.2 channels results in a predominance of K+ currents, leading to further repolarization (Fig. 2B, phase 3). Finally, re-initialization of the IK1 current through activation of Kir2.1 returns the membrane potential to the resting level [24-29].

In contrast to atrial and ventricular myocytes, cells of the SA and AV nodes lack a stable resting potential and instead undergo constant, slow depolarization imparting their pacemaker properties (Fig. 2C, phase 4). This is due to an absence of IK1 in these cells, which allows for the Na+/K+ hyperpolarization-activated cyclic nucleotide-gated channel to conduct an inward pacemaker funny current (If) which slowly depolarizes the membrane. The slow depolarization effect of If inhibits the NaV1.5 channels and rapid membrane depolarization is instead driven by ICaL and ICaT through the CaV1.2 and the T-type Ca2+ channels (CaV3.1), respectively (Fig. 2C, phase 0) [23-25].

The role of CaM involvement in direct or indirect regulation of many of these channels is further detailed below.

Calmodulin modulation of the voltage-gated sodium channels

The NaV1.5 channels are large, heterooligomeric assemblies consisting of a pore-forming α subunit (~ 220 kDa) and one of four auxiliary β subunits (β1–4, 30–35 kDa) involved in channel trafficking and function [30] (Fig. 3). The pore-forming subunit consists of four structurally homologous domains (DI–IV) and each domain contains six putative transmembrane α-helices (S1–S6) linked through alternating extracellular/intracellular loops. The loop between S5 and S6 defines the channel pore (the P loop), thereby controlling ion conductance and selectivity [31, 32]. The positively charged S4 unit accounts for voltage sensitivity and enables rapid activation during membrane depolarization. Following activation, the cardiac NaV1.5 channel displays two distinct forms of inactivation kinetics: fast inactivation in the millisecond range, and a slower form in the second range, the latter observed during prolonged membrane depolarization [33]. Within the ~ 50 residue highly conserved, cytosolic linker between DIII and DIV, a three amino acid (IFM, amino acids 1485–1487) motif is responsible for the observed fast inactivation kinetics and constitutes the inactivation gate. The same linker region in addition confers the observed Ca2+ regulation of the NaV1.5 [34-36].

Figure 3.

Schematic illustration of a ventricular myocyte depicting the ion channels (blue), pumps/exchangers (red) and effectors (yellow) of cardiac contraction. Structures are illustrated at the level of resolution available in the literature. NaV1.5, Kv7.1, Kir2.1, NCX and CaV1.2 are shown with putative transmembrane helices (S1–S6 and TM1–2) located in domains (DI–DIV) or subunits (SUBI–IV). Protein structures including PLN (1ZLL), SERCA (3TLM), CASQ2 (2VAF), CaM (1CLL), CaMKII (3SOA) and RYR2 (emd_1274) are shown in high resolution (www.pdb.org or http://pdbe.org/emdb). The PMCA structure is modified from SERCA with in silico added C-terminal to illustrate CaM binding. White on black P is reported CaMKII phosphorylation and black on white P is PKA phosphorylation. AID, α interaction domain.

CaM modulates the ion conductance of the NaV1.5 through binding to the channel and sensing the [Ca2+]cyt, while Ca2+ may also affect the channel directly by binding to an EF-hand-like motif in the NaV1.5 C-terminal domain [37-41]. The interaction between NaV1.5 and CaM occurs at two sites: an IQ motif located in the C-terminal domain (E1901–S1927) and a CaM binding domain (CaMBD) within the DIII–IV linker (Q1491–F1522, with Y1494 as the anchor residue) (Fig. 3) [36, 42, 43]. At low [Ca2+]cyt, apoCaM binds to the IQ motif with its C-lobe only [11, 44, 45]. At higher [Ca2+]cyt, the CaCaM C-lobe can bind to both the DIII–IV linker CaMBD and the IQ motif; however, the Ca2+ loaded N-lobe is also able to bind the IQ motif, and with greater affinity than the C-lobe. Thus a tripartite model for the CaM interaction with NaV1.5 has been proposed: at low [Ca2+]cyt apoCaM is tethered via the C-lobe to the NaV1.5 C-terminal domain IQ motif, and with increasing [Ca2+]cyt the Ca2+ loaded CaM C-lobe shifts to the DIII–IV linker CaMBD while the Ca2+ loaded N-lobe simultaneously interacts with the IQ motif. CaM interaction with the DIII–IV linker, adjacent to the IFM gate, reduces the gate inactivation ability, thus shifting the NaV1.5 fast inactivation towards more depolarized membrane potentials with increasing [Ca2+]cyt, i.e. increasing the duration of the INa (phase 0 to 1). In support of this model, NaV1.5 mutations in the CaMBD that reduce CaCaM binding result in a lack of Ca2+ regulation, whereas mutations that increase CaCaM binding confer an increased Ca2+ sensitivity of the mutated channel [36]. Several cardiac diseases are linked to mutations in the DIII–IV linker region, including five mutations in the DIII–IV linker CaMBD: M1498T and L1501V causing the NaV1.5 gain-of-function long QT syndrome 3 (LQT3), Y1494N and G1502S causing the NaV1.5 loss-of-function Brugada syndrome (BrS) and ΔK1500 presenting with mixed phenotypes [33, 46-49]. Noteworthy, the residues Y1494, M1498 and L1501 are directly involved in CaCaM binding to Nav1.5. The observed phenotypes may be caused by effects on Na+ channel inactivation as well as dysregulation of CaM mediated Ca2+ sensing.

Apart from direct binding, CaM also affects NaV1.5 via its Ca2+ dependent activation of CaMKII, which phosphorylates regulatory sites in the DI–II linker and the C-terminal region [50-52]. The exact effects of CaMKII phosphorylation remain elusive although a general effect appears to be increased INa, in turn resulting in a higher cytosolic Na+ concentration and prolonged AP [41]. It should be noted that the NaV1.5 subunits are the central part of dynamic protein complexes encompassing many components apart from those mentioned above [30, 53].

Calmodulin affects a subset of the voltage-gated potassium channels

Potassium channels are the most diverse group of ion channels in the mammalian genome, with more than 80 different genes encoding principal subunits of K+ channels [29]. Three distinct subtypes of channels are found in cardiomyocytes each in control of specific K+ currents. The voltage-gated potassium channels KV4.3 and KV1.4 conduct the fast and slow components of the outward K+ transient current (Ito). In ventricular myocytes, the KV11.1 and KV7.1 conduct the delayed rectifier currents (IKr and IKs) while the voltage-insensitive K+ channels Kir2.1 and Kir6.1 conduct the inward rectifier currents (IK1 and IK,ATP), respectively (Fig. 2B). All of these currents play important roles in shaping the myocyte AP and maintaining resting potentials [23].

Like other members of the KV family, the KV4.3 and KV1.4 consist of six helices (S1–S6) homologous to one domain of the NaV1.5 and CaV1.2 channels (Fig. 3). Four of these α subunits form a functional ion channel; additional accessory proteins are needed, however, to gain a native Ito current [54]. Interestingly, Kv4.3 accessory proteins are known to include the KV-interacting proteins (KChIPs) which contain four Ca2+ binding EF-hand motifs, implying a [Ca2+]cyt sensitivity for the KV4.3 [55]. In general the Kv4.x subfamily seems to be regulated by Ca2+ through a mechanism divergent from that of the CaM utilizing CaV1.2 and NaV1.5. Conversely, CaMKII is a common theme in that both Kv4.3 and Kv1.4 are regulated by CaMKII phosphorylation along with other kinases, e.g. protein kinase A (PKA). CaMKII phosphorylation of KV1.4 results in a slowed inactivation and faster recovery of the channel and similar effects have been reported for Kv4.3 [54].

KV11.1 and KV7.1 share structural homology with KV4.3 and KV1.4, and the KV11.1 conducted IKr is the main contributor to repolarization in most cardiac cells [24]. Similar to the KV4.3 and KV1.4, no direct regulation of KV11.1 by CaM has been shown; however, the KV7.1 channel appears to be an exception [56-58]. CaM interacts with two non-continuous regions in the KV7.1 C-terminal domain, one being an IQ-like motif (A371–S389) and the other a 1-5-10 motif (E508–R533). CaCaM binds both regions, whereas apoCaM is limited to the 1-5-10 motif. Electrophysiological experiments show that the effect of CaCaM binding to KV7.1 is an increase in IKs via altering of the channel voltage dependent activation, resulting in more open channels at a given membrane potential [56, 57, 59]. Furthermore, an entirely different mode of CaM regulation may affect the KV7.1. Studies have identified a possible role for CaM in facilitating proper channel expression and folding, thereby increasing the number of functional channels at the cell surface. Both of these roles for CaM were shown to be vital for proper channel function, as mutations causing LQT1 located near and in the KV7.1 CaM binding motifs affect either proper channel gating, channel trafficking to the cell membrane or both [56, 57].

To our knowledge no studies have investigated Ca2+ or CaM regulation of the cardiomyocyte Kir 2.1 and Kir 6.1.

Voltage-gated calcium channels are intimately controlled by calmodulin

The CaV1.2 channels are arranged as heterotrimers with a main pore forming αC1 subunit and two accessory subunits α2/δ and β, both involved in modulating properties of the αC1 subunit. In striking resemblance to the NaV1.5, the αC1 subunit has four homologous domains (DI–IV) connected by cytosolic loops, each consisting of six putative transmembrane helices (S1–S6) and, as for the NaV1.5, with S4 being the voltage sensor (Figs 3 and 4). The loop between S5 and S6 (P loop) in each domain forms the channel pore, each domain contributing one glutamate residue conferring the Ca2+ selectivity [60]. The β subunit is cytosolic and associates with the linker region between DI and DII, modulating CaV1.2 voltage dependent activation and inactivation [26, 61, 62]. The CaV3.1 channels of the pacemaker cells share the same overall structural makeup, although they have an αG1 pore-forming subunit with distinct hyperpolarized activation potentials and fast inactivation kinetics relative to αC1.

Figure 4.

(A) Detailed view of the dyadic cleft in Fig. 3 during CICR and cardiomyocyte contraction: (1) Ca2+ entry through CaV1.2; (2) initial Ca2+ influx triggers CICR and RYR2 releases SR stored Ca2+; (3) increased [Ca2+]cyt relieves myofilaments for contraction; (4) cytosolic Ca2+ is recycled to the SR, extracellular space and mitochondrial matrix. Green arrows show release of Ca2+ and red arrows recycling and extrusion. Arrow thickness represents approximate flux contributions. (B) Cell global events of CICR illustrating the spatial aspect of Ca2+ signalling in heart contraction. For numbering see (A).

The heart ICaL currents have two important functions: one is prolongation of the AP and the other triggering sarcoplasmic reticulum (SR) Ca2+ release (see below) [63]. The CaV1.2 channel requires highly depolarized potentials for activation and has a slow voltage dependent inactivation (VDI). Furthermore, the magnitude of both activation and VDI depend on several regulatory proteins [64, 65]. In addition to the observed voltage dependent regulation, CaV1.2 channels are regulated by the permeating Ca2+ ions, enabling control of the Ca2+ influx at the point of entry. The channel open probability is affected by two oppositely directed [Ca2+]cyt dependent processes: a strong Ca2+ dependent inactivation (CDI) and a so-called Ca2+ dependent facilitation (CDF) process [66]. CDI is a major determinant of the duration of the cardiac AP ensuring a functional balance between Ca2+ influx and K+ efflux (AP phases 2–3). As an example, disruption of CDI increases the total ion flux resulting in a prolonged plateau phase and ultimately a prolonged cardiac AP [22]. In the case of CDF, increased [Ca2+]cyt or frequent depolarizations shift the CaV1.2 channel into a state of higher open probability, allowing entry of a larger amount of Ca2+ per heartbeat and thus contributing to an increased force–frequency relationship, e.g. during exercise [67]. Interestingly, it is CaM binding to CaV1.2 and the CaM Ca2+ sensing that govern both CDI and CDF [68-73]. The main CaM interaction with the CaV1.2 centres around a region in the C-terminal cytosolic loop, CT1 (CaV1.2 amino acids 1538–1692), containing a CaV1.2 EF-hand and three CaM binding sites termed the A-region, C-region and the CaV1.2 IQ motif (Fig. 4) [66].

Reminiscent of its interaction with the NaV1.5, apoCaM binds the CaV1.2 IQ motif tethering CaM at the site of Ca2+ influx. This interaction is mainly mediated by the CaM C-lobe, with the N-lobe having no or little interaction with the IQ motif [74]. In this setting, apoCaM is in an optimal position for Ca2+ dependent regulation of the CaV1.2 [68, 71, 75]. CaCaM interaction with the IQ motif is mainly via the CaCaM C-lobe (Kd ~ 2.6 nm) and accounts for CDI, whereas the CaCaM N-lobe engages in a slightly more promiscuous interaction (Kd ~ 57.6 nm) involved in CDF, possibly leaving it available for interaction with other binding partners in this process [69]. Apart from the IQ motif, CaCaM binds to the C-region through a C-lobe interaction with moderate affinity (Kd ~ 400 nm compared with ~ 2.5 nm for the IQ motif) and to the A-region with a weak to non-specific CaCaM N-lobe interaction, again leaving the N-lobe free to interact with other elements of the CaV1.2 [76, 77]. ApoCaM binding to the A–C region remains controversial and may depend on Ca2+ binding to the CaV1.2 endogenous EF-hand [66, 78]. Disruption of CaCaM C-lobe binding to the C-region has no effect on CDI; however, a reduction of CDF is observed [76, 79]. In studies using synthetic peptides, CaCaM binding to the IQ motif and C-lobe binding to the C-region have been shown to coexist, resulting in two CaM molecules bound to the CaV1.2 CT1 region [76, 77, 80]. The functional effect of this dual binding mode has yet to be resolved but is reminiscent of how CaM is organized on myosin V and the plasma membrane Ca2+ ATPase (PMCA), possibly lending credence to the idea of two CaM molecules individually regulating the CDI and CDF processes of CaV1.2 [9, 76, 81].

Intriguingly, two different regions of the αC1 subunit have been suggested to participate in a bridging interaction with the promiscuous CaCaM N-lobe from CaM otherwise bound to the C-region and IQ motif, respectively. One αC1 region is the N-terminal αC1 NScaTE region (N-terminal spatial Ca2+ transforming element) and the other possibly a region on the I–II linker in complex with the β2 subunit [71, 76, 77, 82]. The CaM–NScaTE interaction was shown to accelerate CDI, possibly via binding of CaM shifting the CaV1.2 N-terminus into a position occluding the ion pore [82]. The CaM-I–II linker interaction affects VDI, CDI and CDF by unknown mechanisms. Apart from CaM directly mediating CDI and CDF, the two processes are also modulated by CaMKII, through phosphorylation of key residues in both the αC1 C-terminal (T1622, S1535 and S1585) and β2a (T498) subunits [83-86]. The molecular effects are not yet determined; however, the T1622 phosphorylation was shown to increase the stoichiometry of CaM binding to the CaV1.2 C-terminal, thus allowing two CaM per CaV1.2 [80].

CaV1.2 is also regulated by the β-adrenergic pathway through the β-adrenergic receptor/cAMP to PKA pathway [87]. PKA is responsible for phosphorylation of both the αC1 and β subunits which results in increased open probability by a decrease in VDI and hence a larger influx of Ca2+ [88-90]. This heightened Ca2+ influx through CaV1.2 is the main contributor to the increased heart contractility during β-adrenergic stimulation [62].

Calmodulin involvement in calcium-induced calcium release

The initial influx of Ca2+ through CaV1.2 leads to a localized rise in [Ca2+]cyt in the so-called dyadic cleft (Fig. 4A, 1; 4B). The dyadic cleft constitutes the cytoplasmic space between the sarcolemma T-tubules containing CaV1.2 and clusters of RYR2 Ca2+ channels embedded in the opposing SR junctional cisternae (Figs 3 and 4) [91-93]. Within each dyadic cleft, 10–25 CaV1.2 and 100–200 RYR2 are arranged into a functional unit, a couplon, with RYR2 clustering at the site of CaV1.2 Ca2+ influx [94]. A subpart of the RYR2 in the dyadic cleft responds to the local rise in [Ca2+]cyt by releasing SR stored Ca2+, which in turn triggers neighbouring RYR2 until the entire couplon releases Ca2+ in a so-called Ca2+ spark (Fig. 4A, 2; 4B) [95, 96]. This process, termed Ca2+ induced Ca2+ release (CICR), gives a further increase in [Ca2+]cyt, and as adjacent couplons are CaV1.2 triggered the propagation of a [Ca2+]cyt transient wave across the cardiomyocyte cytosol is initiated (Fig. 4A,B) [94]. Being a transient wave, this Ca2+ signal has a very distinct spatial and temporal distribution determined by cell morphology and the activity of the Ca2+ trafficking components [92, 95]. As will be outlined in the following, CaM is a pivotal component in sensing and shaping this Ca2+ transient. The transient wave yields muscle contraction when Ca2+ reaches troponin C that upon Ca2+ binding relieves the myofilaments for contraction, i.e. ultimately linking AP to muscle contraction (excitation–contraction coupling) (Fig. 4A, 3; 4B) [97]. In the consideration of CICR and CaV1.2 Ca2+ influx, [Ca2+]cyt refers to highly localized concentrations of cytosolic Ca2+ since the cell cytosolic average has little meaning in the context of a transient wave or a RYR2/Cav1.2 localized burst. Indeed, CaV1.2 and RYR2 Ca2+ signalling events occur in literal nano-domains; e.g. the volume of the dyadic cleft has been measured to 4.39 × 105 nm3 (4.39 × 10−10 nL) (Fig. 4) [92]. Like the AP, CICR is a tightly controlled signalling pathway regulated at multiple levels. Ca2+ itself is a regulator of CICR in that both RYR2 and CaV1.2 have their own [Ca2+]cyt sensing Ca2+ binding sites [98, 99]. As described above, CaM modulates CaV1.2 and so is the case for the RYR2, where CaM adds another Ca2+ sensing and regulatory mechanism to the CICR pathway.

Calmodulin modulation of ryanodine receptor 2 activity

Four RYR2 monomers (each ~ 565 kDa) come together to form the functional Ca2+ channel along an axis of symmetry parallel to the ion flow. These large molecular structures are transmembrane embedded in the SR with the bulk of the molecular mass extruding into the cytosol (Figs 3 and 4) [100]. Across the RYR2 surface are various interaction sites as SR luminal, SR membrane and cytosolic molecules bind and regulate RYR2 assemblies. At the cytosolic face, FK506 binding protein 12.6 (FKBP12.6), CaM, Mg2+ and Ca2+ are the main binding partners [99, 101]. The exact role of FKBP12.6 is unclear, although constitutive tethering appears relevant for the coordinated conformational change between open and closed states of RYR2 monomers while channel activity is regulated by, for example, CaM, Mg2+ and Ca2+ [101-103]. Furthermore, phosphorylation by CaMKII and/or PKA as well as dephosphorylation by protein phosphatases PP1 and PP2A occur at the cytosolic face [103]. The integral membrane proteins triadin and junctin bind both RYR2 and the SR luminal Ca2+ binding protein calsequestrin 2 (CASQ2), thereby forming a channel macro complex [102, 104]. CASQ2 is the major SR Ca2+ buffer regulating SR free Ca2+; however, the SR Ca2+ binding proteins calumenin and histidine rich Ca2+ binding protein (HRC) also interact with the RYR2 luminal surface, and these three proteins have all been shown to affect RYR2 Ca2+ gating. [102, 103, 105].

The RYR2 open probability shows a bell-shaped dependence on [Ca2+]cyt probably owing to multiple endogenous Ca2+ binding sites of various affinities [98, 106]. As a result, RYR2 is inactive at nm, active at low μm and again inactive at high μm [Ca2+]cyt, respectively. The RYR2 sensitivity to [Ca2+]cyt is tuned by the binding of CaM [107-109]. One molecule of CaM binds each of the monomers in the massive RYR2 tetrameric ion channel, and both apoCaM and CaCaM bind the same protein region [107, 110].

Similar to CaM interaction with NaV1.5 and CaV1.2, interaction of CaM with RYR2 appears to be lobe-specific with the CaM C-lobe responsible for the apoCaM binding to an RYR2 CaMBD localized to amino acids 3581–3610 [107]. Meanwhile, the apoCaM N-lobe is less tightly bound potentially leaving it free to interact with other parts of RYR2 [107, 111-113]. CaM Ca2+ binding increases both N- and C-lobe affinities for the RYR2 CaMBD; however, from studies of the skeletal muscle isoform of the ryanodine receptor (RYR1) there is evidence that the CaCaM N-lobe interacts with a second site at higher [Ca2+]cyt. This second site is not contiguous with the apoCaM binding site although still structurally close [110-113]. Cryo-electron microscopy studies of RYR show that CaCaM binding is structurally distinct from that of apoCaM [110]. The most recent cryo-electron microscopy studies furthermore indicate that apoCaM binds RYR2 in a region associated with CaCaM binding and channel inhibition in RYR1, thus supporting the inhibitory role of apoCaM on RYR2 [114].

Studies of intact RYR2 demonstrate that both apoCaM and CaCaM binding lowers RYR2 open probability and thus inhibits SR Ca2+ release [107, 109]. Accordingly, the effect of CaM on RYR2 throughout CICR is one of restricting SR Ca2+ release and finally ending the individual RYR2 Ca2+ sparks. From the CaM lobe-specific binding it may be speculated that the apoCaM C-lobe is mainly responsible for the inhibition of RYR2 at low [Ca2+]cyt, while the sensing of increased Ca2+ of both CaM lobes is relevant for increased channel inhibition at elevated [Ca2+]cyt [111]. Furthermore, the CaM dependent regulation of RYR2 may be structurally and energetically different at low compared with high [Ca2+]free, although the final effects on RYR2 are similar. The suggested CaM–RYR2 interaction would be in keeping with the function of CaM as an associated Ca2+ sensor highly similar to that observed for CaV1.2, NaV1.5 and the CaMKII. As an example, the CaM regulation of RYR2 is analogous to the role of CaM in CaV1.2 CDI. In both cases, CaM sensing of [Ca2+]cyt serves as a negative feedback for the AP initiated CICR which in turn limits [Ca2+]cyt increase and the event duration. Conversely, CaCaM binding to CaMKII (see below) facilitates the CaMKII auto-phosphorylation required for interaction with the CaV1.2 leading to CDF, i.e. sustaining and increasing Ca2+ influx. It is important to note that channel or kinase bound CaM responds to the spatial and temporal Ca2+ signals and transfers its Ca2+ binding into a regulation of the binding target. The simultaneous presence of CaM bound to RYR2, CaMKII and CaV1.2, respectively, at elevated heart rhythm demonstrates the intimate linking of CaM Ca2+ sensing to the regulation of CICR and cardiomyocyte transients.

Calcium (re-)cycling and calmodulin

Having received an initiating amount of Ca2+ from the extracellular space, and released even more from the internal SR storage, the cardiomyocyte recycles the Ca2+ to maintain Ca2+ homeostasis and in preparation for a new excitation event. The Na+–Ca2+ exchangers (NCXs) and PMCA pumps transport cytosolic Ca2+ to the extracellular space, with the NCX being the major contributor of Ca2+ efflux (Fig. 4A, 5) [95, 97]. While these sarcolemma ion exchangers and pumps remove some cytosolic Ca2+ (~ 28%), the majority (~ 70%) goes to replenishing the SR Ca2+ storage via the sarco/endoplasmic reticulum Ca2+ ATPase (SERCA) (Fig. 4A, 4) [95].

Figure 5.

Select overview of ventricular cardiac diseases related to reported mutations in proteins of the cardiomyocyte excitation–contraction cycle. Proteins are divided into cellular locations, i.e. sarcolemma, cytosol and sarcoplasmic reticulum (SR) proteins. SQT, short QT syndrome; JLN, Jervell and Lange–Nielsen syndrome. Gene names for all proteins are given in parentheses after the protein abbreviation. Relevant ion currents are given for sarcolemma ion channels. A comprehensive overview of general cardiac diseases related to proteins involved in cardiomyocyte excitation–contraction can be found in [29, 33, 63].

The cardiomyocyte SERCA isoforms (2a and 2b) are not present in the dyadic cleft but take up Ca2+ from the cytosol at other positions along the SR, where the closely associated membrane protein phospholamban (PLN) functions as an inhibitor (Figs 3 and 4A) [115, 116]. CaM does not directly affect SERCA; however, the PLN inhibition depends on its phosphorylation state which, among other kinases, is regulated by CaMKII [117]. Phosphorylation of PLN relieves SERCA inhibition; thus CaMKII activation by CaM can lead to increased SERCA activity via PLN [118]. Increased Ca2+ recycling through SERCA generally promotes a higher frequency of cardiomyocyte contraction, as the cell Ca2+ cycling is sped up.

PMCA is considered of less importance for the Ca2+ transients than SERCA; however, recent research has highlighted its importance in controlling the Ca2+ exposure of the neuronal nitric oxide synthetase (nNOS) [119]. PMCA1 and PMCA4 are the cardiomyocyte isoforms which regulate nNOS both via physical interaction and extruding Ca2+ and reducing local [Ca2+]cyt availability for the Ca2+-sensitive nNOS [119-121]. Nitric oxide (NO) binds and regulates NaV1.5, CaV1.2, RYR2 and SERCA, but also affects cell cGMP and cAMP levels [119, 121, 122]. Furthermore, PMCA4 is found associated with the NaV1.5 macro complex and, interestingly, dysfunctional regulation of the PMCA4 nNOS inhibition has been implicated in the molecular mechanism of arrhythmogenic diseases such as some forms of LQT, supporting its relevance in regulation of the heartbeat [123, 124]. A CaMBD is located in the PMCA cytosolic C-terminal tail, a region which auto-inhibits the ion pump through contacts with other PMCA regions [121]. At cell resting [Ca2+]cyt this auto-inhibition renders the PMCA pump basically inactive, and activation by, for example, CaCaM binding is required for increasing the PMCA Ca2+ affinity to physiological significant levels [125]. ApoCaM does not bind PMCA noticeably, and hence local CaM Ca2+ sensing is a regulator of PMCA activity [9, 125, 126].

Secondary effectors of the cardiomyocyte contraction

The β-adrenergic pathway

By virtue of the β-adrenergic pathway the sympathetic nervous system can mobilize the cardiomyocyte excitation–contraction, greatly enhancing the Ca2+ circulating mechanisms described above (Fig. 4) [95]. Epinephrine (and norepinephrine) binds the β1- or β2-adrenergic receptors in the cardiomyocyte sarcolemma and ultimately leads to phosphorylation of NaV1.5, CaV1.2, RYR2 and PLN [32, 95]. β-adrenergic receptors are G-protein coupled receptors, and ligand binding induces activation of adenylyl cyclase (AC) resulting in production of cAMP. The rise in cytosolic cAMP relieves the regulatory subunit of the PKA, the final intracellular effector of the β-adrenergic pathway. PKA phosphorylation of NaV1.5 leads to an increased amplitude of INa and similarly deregulates the auto-inhibition of CaV1.2, i.e. increases Ca2+ influx [127]. Further, PKA phosphorylation increases RYR2 activity with a quicker response in both opening and closing, and the simultaneous increased SERCA activity from PKA phosphorylated PLN release results in overall quickened SR Ca2+ release and uptake. Thus the net result of β-adrenergic stimulation is that cardiomyocyte Ca2+ channels and pumps accelerate their Ca2+ transport activities thereby increasing the overall amplitude and frequency of Ca2+ transients.

Calmodulin modulation of cyclic nucleotide signalling

ACs are a family of membrane bound enzymes with AC5 and AC6 being the main heart tissue isoforms [128-130]. While some AC isoforms are CaM regulated (AC1, AC3 and AC8), the AC5 and AC6 are not, although they are Ca2+ sensitive and inhibited by elevated [Ca2+]cyt [130]. As already exemplified in the cAMP regulation of PKA, the cellular concentrations of the secondary messengers cAMP and cGMP also regulate cardiomyocyte Ca2+ signalling. The functional antagonists of the ACs are phosphodiesterases (PDEs), as they degrade cyclic nucleotides and subsequently lower, for example, the PKA activity. Mammalian PDEs comprise 21 genes with multiple splice variants; however, from the combined CaM and cardiomyocyte viewpoint, three PDE1 gene products (PDE1A–C) appear particularly interesting as PDE1A–C are all CaM regulated and found in heart tissue. CaCaM binding to N-terminal sites in these PDE1 isoforms stimulates their activity and thus increases cell cAMP and cGMP degradation [131]. Studies show that the PDE1C1 variant in particular is a major contributor of cAMP and cGMP turnover in human cardiomyocytes [132]. Degradation of cyclic nucleotides generally signals an easing of cardiomyocyte contractility, e.g. by inhibiting the cAMP dependent phosphorylation of CaV1.2 and RYR2 in the β-adrenergic pathway. Further complicating the PDE regulation is the fact that CaMKII regulates PDE1B while PKA regulates many PDEs [130, 131]. The details of these interactions are outside the scope of this review, however (see [131, 133] for encompassing reviews).

The ubiquitous calcium/calmodulin dependent kinase II

As mentioned in the sections above, CaMKII has a recurring function as an ion channel modulator, and the properties of CaMKII and in particular its regulation by CaM also affect the cardiomyocyte Ca2+ transients. The major isoform in cardiomyocytes is CaMKIIδ which has three distinct parts: an N-terminal catalytic domain, a regulatory domain and a C-terminal tail. The C-terminal tail is responsible for the CaMKII monomers arranging into 6–12 oligomer wheel-like structures that in turn may dock and interact with ion channels like NaV1.5, CaV1.2 and RYR2 [41, 134] (Fig. 3). Within the regulatory domain, an autoregulatory region acts as a pseudo-substrate arresting the CaMKII in an inactive state [22, 135]. CaCaM binding to the autoregulatory region relieves this auto-inhibition. Accordingly, the CaMKII activation is strictly Ca2+ dependent via CaM Ca2+ binding. Following CaCaM activation, adjacent CaMKII monomers trans-phosphorylate Thr286 close to the autoregulatory domains, finally leaving the catalytic domains in an auto-activated state [41, 135]. Thr286 phosphorylation increases CaCaM affinity 1000-fold and effectively traps a CaCaM with each activated CaMKII monomer. Furthermore, structural studies have indicated that the activated, functional unit of CaMKII is a dimer [22, 135, 136]. In general, the functional unit closely associates with, if not tethers to, ion channels where the kinase activity contributes to their regulation, although the exact mechanism is channel type dependent [41]. CaMKII's return to the auto-inhibited state requires dissociation of CaCaM and subsequent dephosphorylation of Thr286 by PPs, and includes a particularly slow (hundreds of seconds) dissociation rate for CaCaM [41, 136]. This slow dissociation of CaCaM allows for the CaMKII activity to be sustained and integrated across repeated cardiomyocyte Ca2+ transients leading to, for example, CaV1.2 CDF [41, 70, 83, 137]. In this sense, the combined CaMKII activity level constitutes a chemical memory of heartbeats, adapting the cardiomyocyte Ca2+ cycling to, for example, sustained physical exercise, and while the kinase is the ion channel effector, CaCaM is the pivotal mediator of the underlying Ca2+ signalling.

Not surprisingly, mouse model studies have shown that increased CaMKII activity leads to severe heart diseases such as dilated cardiomyopathy and cardiac hypertrophy, and is likely to be arrythmogenic due to altered cardiomyocyte Ca2+ handling [138-140]. Specifically, over-expression of CaMKIIδ increases SR Ca2+ release during CICR and simultaneously the frequency of spontaneous RYR2 Ca2+ sparks, i.e. untimely couplon Ca2+ release [139]. As CaMKIIδ is shown to associate with RYR2, these studies exemplify the potentially lethal effects of erroneous CaMKII regulation leading to alterations of the RYR2 phosphorylation state [139, 141, 142].

Secondary cardiomyocyte Ca2+ stores

Cardiomyocytes have two major Ca2+ storing organelles other than the SR. Mitochondria and the Golgi apparatus (GA) are organelles of considerable Ca2+ storage and handling owing to their own sets of ion channels and pumps [143-145]. Mitochondria respond to the [Ca2+]cyt and the resulting mitochondrial matrix Ca2+ concentration ([Ca2+]mit) modulates cell energy metabolism [144]. Although it remains inconclusive whether mitochondria respond quickly enough for beat-to-beat effects, an integration of [Ca2+]cyt is seen in [Ca2+]mit. Furthermore, the close proximity of mitochondria to the CICR events allows mitochondria to contribute to Ca2+ re-uptake. [Ca2+]mit rises with cardiomyocyte workload and/or β-adrenergic stimulation and thereby increases energy production. Also mitochondrial RYR1 (mtRYR1) and NOS (mtNOS) in the inner mitochondrial membrane are sensitive to [Ca2+]cyt. Like other NOS isoforms the mtNOS is regulated by CaM, and there is evidence that the mtRYR1 is highly similar if not identical to the RYR1 and hence also CaM regulated. mtNOS produces NO in a localized domain on the matrix side of the inner mitochondrial membrane, while mtRYR1 functions as a Ca2+ uptake channel trafficking Ca2+ into the mitochondrial matrix [92, 146-148]. Dysfunction of the mitochondrial Ca2+ handling components has been implicated in cardiomyopathy highlighting their importance for heart Ca2+ homeostasis [149].

The cardiomyocyte GA has a broad set of Ca2+ handling proteins including luminal Ca2+ binding proteins and membrane proteins such as SERCA2, RYR, secretory pathway Ca2+ ATPases (SPCA1–2) and the inositol-triphosphate receptor (IP3R). The Ca2+ concentrations of the various GA compartments affect protein trafficking and organelle morphology highlighting the central role of Ca2+ for regulating GA function. Similarly, studies of neonatal cardiomyocytes have shown GA stored Ca2+ release via GA RYRs in response to CICR. The latter implies a role for the GA in excitation–contraction Ca2+ transients [145]. As seen for the SR, CaM is a ubiquitous regulator of the organelle Ca2+ handling in that RYR and IP3R are regulated by CaM [150].

Calmodulin mutations and heart arrhythmia

Given the target dependent and differential effects that CaM exerts on the regulation of the AP and cardiac contraction, it is difficult to predict the impact of any CaM mutation, as also detailed in the sections above. Missense mutations in some of the genes and proteins described have been linked to specific cardiac arrhythmias. A combined overview of selected ventricular-arrhythmia-linked genes, the proteins they encode, their relevant ion currents and whether CaM is known to bind and regulate their function is given in Fig. 5 (for a more complete description of cardiac arrhythmia genes see [29, 33, 63]). CaM binding can be inhibitory (CDI) or facilitating (CDF) for the CaV1.2 channels' inward currents, and facilitating for both the inward INa and outward IKs currents through regulation of the NaV1.5 and KV7.1 channel, respectively (Fig. 5). CaM mutations resulting in loss of CaM function could thus lead to a cardiac phenotype characterized by a prolonged AP phase 2, from loss of CaV1.2 CDI or KV7.1 facilitation finally resulting in LQT. Or, they may lead to a shortened QT interval through loss of CaV1.2 CDF or NaV1.5 fast inactivation. Thus, prediction of cardiac phenotypic effects of a CaM gene mutation is not simple and is perhaps best illustrated by the observation that missense mutations directly in the CaMBD of NaV1.5 can lead to both gain-of-function (prolonged AP LQT) or loss-of-function (BrS) effects. In this context, the finding that individuals carrying one of two identified CaM mutations (N53I or N97S) display a catecholaminergic polymorphic ventricular tachycardia (CPVT)-like specific phenotype is highly intriguing [19]. The following identification of three other CaM mutations (D95V, F141L and D129L) in infants with early onset severe LQT (esLQT), a ventricular arrhythmic disorder mechanistically different from CPVT, highlights the complexity of CaM mediated arrhythmic disorders [20]. Interestingly, three CaM missense mutations have been identified in the Exome Sequencing Project; however, since no phenotypic details of the individual carrying these mutations have been published, they will not be included in our discussion [151]. A combined overview of the identified CaM mutations is given in Table 1.

Arrhythmogenic mechanisms in catecholaminergic polymorphic ventricular tachycardia

CPVT is characterized by episodic syncope and/or sudden cardiac arrest induced by adrenergic stimulation such as during exercise or acute emotional stress [152, 153]. The ECG usually appears within normal limits at rest, but typically displaying prominent U waves which at times of high adrenergic activation may give way to ventricular arrhythmias [154, 155]. At the cellular level, arrhythmias in CPVT are believed to be due to spontaneous SR Ca2+ release not coupled to changes in the AP. These events trigger untimely Ca2+ export by NCXs resulting in a net inward current from antiported Na+ and ultimately lead to a delayed after depolarization (DAD) of the sarcolemma. Under high adrenergic stimulation, the DAD amplitude and frequency may reach a threshold for activation of NaV channels, and erroneously triggered APs are initiated which generate extra-systolic arrhythmic beats when propagating to the entire ventricle [63].

Prior to identification of the link between CALM1 mutations and CPVT, two forms of the disease had been described: CPVT1 is an autosomal dominant disease caused by mutations in RYR2 and CPVT2 is an autosomal recessive form caused by mutations in the CASQ2 gene [156-159]. RYR2 mutations are gain-of-function mutations, many of which have been shown to result in leaky SR Ca2+ channels. The molecular mechanism underlying this leaking behaviour is still somewhat debated, and two different views have been put forward. In one model, the CPVT linked RYR2 mutations confer increased open probability of the channel, as a result of increased [Ca2+]cyt sensitivity, which has been demonstrated for mutated, isolated single channels [160, 161]. In the alternative model, the RYR2 sensitivity to SR luminal Ca2+ levels is increased, resulting in store overload induced calcium release (SOICR) [162, 163]. SOICR occurs if the SR luminal Ca2+ concentration becomes higher than the threshold value for channel opening, and the CPVT RYR2 mutations investigated have been shown to decrease this threshold concentration, thus increasing the RYR2 SR luminal Ca2+ sensitivity. Whether it is the RYR2 cytosolic or SR Ca2+ sensitivity, or a combination of both, that is pathologically disturbed may very well depend on the specific mutation. The CASQ2 mutations are also believed to alter the RYR2 SR Ca2+ sensitivity, leading to SOICR [164]. Recently, a new form of CPVT was identified in individuals homozygous for null mutations in triadin (TRDN), an SR membrane protein associated with RYR2 and CASQ2, thus adding to the evidence that dysregulation of the RYR2 channel macro complex is the predominant molecular mechanism of CPVT (Fig. 5) [165].

It is tempting to speculate that the molecular mechanism for CPVT4 (CaM associated CPVT) is likewise primarily through dysregulation of RYR2 activity as CaM directly interacts with and regulates RYR2 activity. Indeed, the N97S variant demonstrated a defective interaction with a 31 residue peptide corresponding to the RYR2 CaMBD, although only observed within a narrow, low [Ca2+]cyt range [19]. However, using the same peptide no obvious difference in binding to the N53I variant was observed. Thus, either the N53I variant expresses its phenotype through a disturbed interaction with a different target or a potential aberrant RYR2–CaM N53I interaction would be mechanistically different at the molecular level and yet have a similar effect on intact RYR2 function. Given the incomplete understanding of how CaM regulates RYR2, it is not implausible that the N53I missense mutation affects RYR2 function; however, further studies are needed to clarify this. Interestingly, the CaM CPVT mutations do not significantly impair the ability to regulate the skeletal isoform of RYR (RYR1), even though the RYR1and RYR2 CaMBDs are highly similar (two Lys to Arg, and one Ala to Thr substitutions) [113]. One explanation would be differences in CaM regulation of RYR2 and RYR1, as CaM increases the RYR1 channel open probability at low Ca2+ concentration [108]. Or it may be that the dysregulation of the RYR is only seen in the stressed adrenergically stimulated fast beating heart.

Although RYR2 dysregulation seems probable, it cannot be ruled out that the CPVT4 phenotype is a combination of molecular mechanisms, such as suggested for the ankyrin B (ANK2) linked type 4 LQT and Andersen–Tawil syndrome, where some individuals have been reported to display CPVT-like arrhythmias [166, 167]. Ankyrin B is a structural protein linking and coordinating several of the ion channels in the sarcolemma to the cytoskeleton, thus similar to CaM in affecting the function of multiple ion channels. Also, the CaM regulated CaMKII has been suggested to be involved in the molecular pathogenic mechanism for several ventricular arrhythmias, including CPVT [168]. Dysregulation of CaMKII by mutated CaM may also be a contributor to CPVT4.

Arrhythmogenic mechanisms in early onset severe long QT syndrome

LQT arises from an elongation of the action potential plateau phase (phase 2) (Fig. 2). Thus disturbing the outward current of K+ (loss-of-function mutations in K+ channel genes) or increasing the inward current of either Na+ or Ca2+ (gain-of-function mutations in Na+ or Ca2+ channels) have been linked to LQT (Fig. 5) [29, 33]. Since the esLQT CaM mutations display a severely reduced Ca2+ affinity specifically in the C-lobe, we suggest that the most likely molecular mechanism linking CaM mutations with esLQT is primarily through a lack of CDI of the CaV1.2, resulting in a gain-of-function effect on the Ca2+ channel [69, 71]. The resulting anticipated increased Ca2+ concentration may in addition activate the NCX resulting in a further net inward current from exchanging three Na+ for two Ca2+. In addition, a decreased interaction with the KV7.1 channels would reduce the K+ inward current and be expected to contribute to an increased QT interval. A diminished CaM interaction with the NaV1.5 channels would be expected to lead to loss of function and thus have a shortening effect on the QT interval (Fig. 5) [36, 56, 57, 59]. The difference in the phenotypic expression of the esLQT compared to the CPVT mutations may originate from a more severe reduction in the C-lobe Ca2+ affinity, resulting in dramatic reduction in CaV1.2 bound CaM Ca2+ sensitivity. However, as detailed above for CPVT4, the molecular mechanism for CaM esLQT may also involve dysregulation of multiple other CaM binding proteins, and indeed some of the individuals with esLQT CaM mutations displayed ventricular arrhythmias with a polymorphic nature, resembling the CPVT phenotype observed in the initially described CaM mutations [19, 20].

Dominant inherited disease with only one of six CALM alleles mutated

For the RYR2 associated CPVT, the dominant inheritance may be explained by the RYR2 channel being a tetramer; thus with half of RYR2 alleles mutated, 15/16 channels will have at least one RYR2 monomer with a mutation, assuming identical expression and stability of the native and mutated channels. One can easily imagine that just one mutated monomer would destabilize the tetrameric channel, resulting in untimely leaking of Ca2+ from the SR. Thus, such mutations would display the expected gain-of-function effect on the Ca2+ current. For CaM linked disease, it is perhaps more puzzling that it expresses in a dominant inheritance pattern, as only one of six alleles encoding identical proteins is mutated. If the molecular mechanism is through dysregulation of the RYR2 channel, however, one can envision that just one mutated CaM is needed for destabilizing and introducing leakiness in the tetrameric channel. In the case of equal expression ratio for the three CaM genes, and no significant change in RYR2 affinity, one can expect just over half (671 out of 1296 possible) of the CaM–RYR channel complexes to contain at least one mutated CaM. Using RT-QPCR, Crotti et al. reported the CALM13 expression levels in human heart to vary slightly from infancy to adulthood with CALM2 and CALM3 levels two and five times that of CALM1, respectively [20]. Our unpublished observations from DeepSAGE expression analysis of human ventricular muscle, however, show that the relative expression levels of CALM1, CALM2 and CALM3 are 5 to 4 to 1, respectively. Further, as detailed above, the RYR2 channels are organized in couplons, containing assemblies of 100–200 units, and leakage of a few RYR2 channels within one couplon may release enough Ca2+ to trigger the CICR of the entire couplon. Thus, it is not implausible that one mutated out of six alleles can express itself in a dominant negative manner giving rise to CaM linked CPVT disease and demonstrate dominant inheritance. One possible explanation for the dominant inheritance in the esLQT CaM individuals may be that, in the couplon arrangement of 10–25 CaV1.2 adjacent to 100–200 RYR2 channels, only a few CaV1.2 channels having a gain of function (or lack of CDI) could elicit sufficient RYR2 Ca2+ sparks to trigger the entire couplon and eventually activate the NCX.

Conclusions

A large portion of our current understanding of gene function in human physiology comes from the phenotypic observations of individuals carrying missense mutations. When these are linked to a particular disease, the following biophysical, biochemical, cell biological and animal model studies of the molecular, cellular and organismal effects of these mutations often allow deduction of the particular molecular disease mechanism. This novel understanding paves the way for effective and targeted pharmaceutical intervention. The recent identification of the first two human CaM mutations, both linked to the lethal cardiac arrhythmia CPVT, and the following three CaM mutations associated with recurrent cardiac arrest in infants will, it is hoped, allow further insight into the delicate role of CaM Ca2+ handling and sensing in the cardiac contraction cycle.

We suggest that the fact that only one out of six CaM encoding alleles mutated gives rise to a dominant inheritance pattern favours a molecular disease mechanism where dysregulation of multimeric CaM binding protein complexes is involved. Since the only other gene where missense mutations are linked to dominant inheritance of a similar arrhythmic phenotype is RYR2, an altered CaM regulation of this channel is the most obvious candidate for CaM linked CPVT. As is evident from this review, however, CaM regulates the activity of many other crucial components of the cardiac contraction cycle, and the actual mechanistic impact may be far more complicated than through a single aberrant interaction. Indeed, the recent identification of individuals carrying one of three other CaM C-lobe mutations presenting with esLQT illustrates that the effect of an impaired CaM function may take multiple expressions.

The fact that no other apparent phenotypic manifestations of the CPVT CaM mutations are evident and that all five reported CaM mutations present with cardiac arrhythmias implies that in most cells and tissues there are sufficient compensatory mechanisms for the observed CaM mutations. The effects of the mutations are apparently not being fully compensated in the adrenergically activated heart with its highly stressed Ca2+ cycling system. The esLQT linked CaM mutations may be associated with impairment of CaM mediated signalling in neurons [20, 169]. However, as Crotti and co-workers point out, these effects could also be the result of oxygen deprivation from multiple severe cardiac arrests [20].

Very few CaM mutations are likely to be tolerated at all, given the extraordinary conservation of the CALM genes, which has not allowed for any deviation in the protein sequence since the appearance of vertebrates. However, we hope that the identification of CaM mutations linked to cardiac arrhythmia will spur cardiologists to specifically examine all three CALM genes in individuals negative for known cardiac arrhythmia gene mutations. Further biophysical and functional analysis of CaM variants linked to cardiac disease has the potential to help unravel the multifaceted role of CaM mediated Ca2+ signalling in a heartbeat.

Conflict of interest

The authors declare no conflict of interest.

Ancillary