Calmodulin Mutations Associated with Long QT Syndrome Prevent Inactivation of Cardiac L-type Ca²⁺ Currents and Promote Proarrhythmic Behavior in Ventricular Myocytes

Abstract

Recent work has identified missense mutations in calmodulin (CaM) that are associated with severe early-onset long-QT syndrome (LQTS), leading to the proposition that altered CaM function may contribute to the molecular etiology of this subset of LQTS. To date, however, no experimental evidence has established these mutations as directly causative of LQTS substrates, nor have the molecular targets of CaM mutants been identified. Here, therefore, we test whether expression of CaM mutants in adult guinea-pig ventricular myocytes (aGPVM) induces action-potential prolongation, and whether affiliated alterations in the Ca²⁺ regulation of L-type Ca²⁺ channels (LTCC) might contribute to such prolongation. In particular, we first overexpressed CaM mutants in aGPVMs, and observed both increased action potential duration (APD) and heightened Ca²⁺ transients. Next, we demonstrated that all LQTS CaM mutants have the potential to strongly suppress Ca²⁺/CaM-dependent inactivation (CDI) of LTCCs, whether channels were heterologously expressed in HEK293 cells, or present in native form within myocytes. This attenuation of CDI is predicted to promote action-potential prolongation and boost Ca²⁺ influx. Finally, we demonstrated how a small fraction of LQTS CaM mutants (as in heterozygous patients) would nonetheless suffice to substantially diminish CDI, and derange electrical and Ca²⁺ profiles. In all, these results highlight LTCCs as a molecular locus for understanding and treating CaM-related LQTS in this group of patients.

1. Introduction

Calmodulin (CaM) is a ubiquitous Ca²⁺-sensor molecule that modulates a vast array of proteins, thereby controlling signaling cascades via Ca²⁺-dependent adjustment of relevant proteins. As such, CaM critically orchestrates numerous functions, including cellular excitability, muscle contraction, memory, and immunological responses [1, 2]. So important are the functions of CaM that it has long been thought that naturally occurring mutations within this molecule would prove lethal, and that such mutations would thereby play little role in disease processes afflicting living individuals.

Yet, a role for CaM in a number of diseases has begun to emerge. Alterations in the overall level of CaM have been implicated in heart failure [3], schizophrenia [4], and Parkinson's disease [5–7]. Outright CaM mutations in Drosophila have been associated with muscle malfunction [8]. Very recently, human genetic studies uncovered de novo and heritable CaM mutations (N54I and N98S, start methionine denoted residue 1) that are associated with 11 cases of catecholaminergic polymorphic ventricular tachycardia (CPVT), where altered CaM-ryanodine receptor function is implicated as a major contributing factor [9]. Further, whole-exome and targeted gene sequencing has revealed an association between three de novo missense CaM mutations and severe long-QT syndrome (LQTS) with recurrent cardiac arrest [10]. The first hints of underlying mechanism can be gleaned by relating the locations of these mutations to the basic structure-function layout of CaM, a 17 kDa protein comprised of N- and C-terminal lobes linked by a flexible helix. Each lobe of CaM contains two EF hands, canonical Ca²⁺ binding motifs, with the N-lobe having slightly lower Ca²⁺ binding affinity. Ca²⁺ binding to these EF hands induces a conformational change that alters function of target molecules to which CaM is bound, thus transducing changes of intracellular Ca²⁺ concentration [11] into modulation of molecular function. Each of the LQTS mutations (D96V, D130G, and F142L, with start methionine denoted residue 1) resides at or near Ca²⁺ coordinating residues within the EF hands of the C-lobe of CaM, and have been shown to decrease affinity for Ca²⁺ binding [10]. By contrast, the reported CPVT mutations in CaM imparted little-to-mild reduction of Ca²⁺ binding affinity [9]. It is perhaps interesting to speculate that the contrasting effects on Ca²⁺ binding may underlie the elaboration of distinguishable LQTS and CPVT phenotypes by these two classes of mutations. At present, however, the mechanisms linking these mutations in CaM to their corresponding disease phenotypes are essentially unknown.

That said, progress towards elucidating these mechanisms will ultimately prove invaluable in devising personalized therapeutics for afflicted individuals, and in gleaning general lessons about LQTS pathogenesis from these single-point-mutation case examples. Among the most prominent mechanistic unknowns are the following. First, do the LQTS CaM mutations actually cause the emergence of LQTS substrates in heart? At present, no experimental evidence directly establishes causality. Second, what are the predominant molecular targets through which CaM mutations exert their actions in heart? Likely cardiac myocyte targets abound, including ryanodine receptors (RyR2), voltage-gated Na channels (Na_V1.5), slowly activating delayed-rectifier K channels (I_Ks), and L-type Ca²⁺ channels [10–12] (Ca_V1.2). All of these contribute to shaping action-potential morphology and thereby represent plausible candidates. Third, the severity of the LQTS fits in a seemingly incongruous fashion with the redundancy of human CaM genes (CALM1, CALM2, and CALM3), each of which encodes for an identical CaM molecule at the protein level. Given the heterozygosity of these LQTS patients [10], this redundancy implies that only one of six alleles of CaM would possess a mutation, yielding only a portion of mutant versus wild-type CaM.

Here, therefore, we acutely introduce LQTS CaM mutants into adult guinea-pig ventricular myocytes (aGPVMs) and demonstrate marked prolongation of action potentials, along with intense disturbance of Ca²⁺ cycling. As these effects are reminiscent of those we observed previously by man-made CaM mutants acting strongly through diminished CaM-mediated regulation of L-type Ca²⁺ channels [13] (LTCCs), we tested directly for the effects of naturally occurring LQTS CaM mutants on these very channels. Indeed, we establish that Ca²⁺ regulation of LTCCs can be strongly suppressed by overexpression of LQTS CaM mutants, posturing altered regulation of these channels as an important contributor to the LQTS phenotype. By contrast, overexpressing CPVT CaM mutants caused weaker or undetectable perturbation of LTCC function and action potentials. Finally, we note the requirement that a single Ca²⁺-free CaM (apoCaM) must first preassociate with LTCCs for subsequent Ca²⁺ regulation to occur [5, 11, 14–17], and substantiate how this feature rationalizes how a limited fraction of LQTS CaM mutants can nonetheless elaborate significant perturbation of channel regulation, sufficient to appreciably prolong action potentials.

2. Methods

2.1 Adult Guinea-pig Ventricular Myocyte Isolation and Adenoviral Transduction

Adult guinea-pig ventricular myocytes (aGPVMs) were isolated from whole hearts of adult guinea pigs (Hartley strain, 3–4 wk old, weight 250–350 g). Hearts were excised after guinea pigs were anesthetized with pentobarbital (35 mg/kg, intraperitoneal injection). Single ventricular myocytes were isolated from both ventricles according to a published protocol [18] and plated on glass coverslips coated with laminin (20 μg/ml overnight at 4 °C). Cells were transduced with adenovirus carrying wild-type or mutant CaM upon plating in the presence of M199 medium supplemented with 20% fetal bovine serum. Expression of wild-type CaM had little effect on action-potential morphology or duration, as compared to uninfected myocytes (Supplementary Figure 1). After 4 hours, the medium was replaced by M199 medium with 0% fetal bovine serum to maintain the phenotype of acutely dissociated myocytes. Cells were maintained at 37 °C and recording was done at room temperature 20–36 hours later.

2.2 Molecular Biology

LQTS CaM mutations were generated using QuikChange ™ site-directed mutagenesis (Agilent) on rat brain CaM (M17069) in the pcDNA3 vector [13] (Invitrogen). CPVT CaM mutations were generated on human CALM1 gene in the pcDNA3 vector (a kind gift from Michael T. Overgaard [9]). For electrophysiological recordings in HEK293 cells, both wild-type and LQTS mutant CaMs were cloned into the pIRES2-EGFP vector (Clontech Laboratories, Inc.) using NheI and BglII. For adenoviral expression in aGPVMs, wild-type and mutant CaMs were cloned into the pAdCiG viral shuttle vector using XhoI and SpeI. Adenovirus was amplified via a standard cre-recombinase method as previously described [13].

The human cardiac α_1C cDNA was constructed by cloning in an ~1.6 kbase upstream fragment of the cardiac (containing exon 8a) channel variant (kind gift from Tuck Wah Soong [19]) into a human α_1C-1 backbone (NM_000719 kindly gifted from Charlie Cohen of Merck Pharmaceuticals) contained within pcDNA3.1, via HindIII and ClaI sites.

For FRET two-hybrid constructs, CaM and CI region of Ca_V1.2 channels (as defined in Figure 5A and described previously [14]) were tagged on their amino termini with fluorophores (cerulean and venus, respectively) with a linker of 3 alanines, and cloned into the pcDNA3.0 (Invitrogen) using KpnI and XbaI.

Figure 5. Mutant CaMs bind to Ca_V1.2 channels at least as well as wild-type CaM.

A, Cartoon depicting CaM interaction domains on the Ca_V1.2 calcium channel (CI segment). Below, construct schematics depicting FRET interaction pairs used for binding assessment.

B, The canonical FRET binding curve between the CI region and CaM_WT. In particular, an acceptor-centric FRET efficiency (E_A) is plotted as a function of the relative free concentration of donor-tagged molecules D_free (cerulean-CaM_WT). K_d/EFF for CaM_WT binding to CI region is 12000 D_free units [5].

C-E, FRET binding curves between the CI region and CaM_D96V (C), CaM_D130G (D) and CaM_F142L (E). K_d/EFF are 4000, 6500, and 1000 D_free units, respectively. The binding curve for CaM_WT is reproduced in gray for reference.

2.3 Transfection of HEK293 Cells

For whole-cell patch clamp experiments, HEK293 cells were cultured on glass coverslips in 10-cm dishes and Ca²⁺ channels were transiently transfected using a standard calcium phosphate method [20]. 8 μg of human cardiac α_1C cDNA (as described above) was co-expressed heterologously with 8 μg of rat brain β_2a (M80545), 8 μg of rat brain α₂δ (NM012919.2) subunits, and 8 μg of wild-type or mutant CaMs, except for mixing experiments (Figure 6) where various molar ratios of wild-type to mutant CaM were transfected. The auxiliary β_2a subunit was chosen so as to minimize the confounding effects of voltage-dependent inactivation on CDI [21]. To increase expression levels, 2 μg of simian virus 40 T antigen cDNA was co-transfected. Expression of all constructs was driven by a cytomegalovirus promoter.

Figure 6. Dose-dependent effect of mutant CaMs.

A, Langmuir equation relating the extent of CDI and wild-type versus mutant CaM expression ratio γ̂.

B, Predicted CDI as a function of γ̂ (black). Solid circles indicate average data for each ratio γ̂; open circles pertain to data for individual cells. Light rose shaded region indicates the deficit in CDI expected for a expression ratio corresponding to heterozygous mutation in CALM2 gene, while dark rose shaded region indicates predicted extent of CDI reduction for homozygous mutation.

C, Exemplar current traces with the colors corresponding to the Langmuir plot in B. For each set of records, the Ca²⁺ trace is the lighter color waveform, and the corresponding Ba²⁺ trace is normalized to the peak of the Ca²⁺ trace for comparison of decay kinetics. Scale bar corresponds to Ca²⁺. Exemplar traces on far right are from a cell expressing CaM_WT only (γ̂ = ∞), as reproduced from Figure 2A for reference.

For FRET two-hybrid experiments, HEK293 cells were cultured on glass-bottom dishes and transfected with polyethylenimine [22] (PEI) before epifluorescence imaging. Whole-cell patch clamp and FRET two-hybrid experiments were performed 1–2 days after transfection.

2.4 Electrophysiology

Whole-cell voltage-clamp recordings of HEK293 cells were done 1–2 days after transfection at room temperature. Recordings were obtained using an Axopatch 200B amplifier (Axon Instruments). Whole-cell voltage-clamp records were lowpass filtered at 2 kHz, and then digitally sampled at 10 kHz. P/8 leak subtraction was used, with series resistances of 1–2 MΩ. For voltage-clamp experiments, internal solutions contained (in mM): CsMeSO_3, 114; CsCl, 5; MgCl₂, 1; MgATP, 4; HEPES (pH 7.3), 10; and either BAPTA, 10 or EGTA, 1; at 295 mOsm adjusted with CsMeSO₃. The free Ca²⁺ concentrations in these BAPTA- and EGTA-containing were respectively estimated to be ~2.4 and 0.45 pM [23], assuming a contaminant Ca²⁺ concentration of 25 μM (standard conversion at http://maxchelator.stanford.edu/). External solutions contained (in mM): TEA-MeSO₃, 140; HEPES (pH 7.4), 10; and CaCl₂ or BaCl₂, 40; at 300 mOsm, adjusted with TEA-MeSO₃. These solutions produced the following uncorrected junction potentials: 10 BAPTA/40 Ca²⁺: 10.5 mV; 10 BAPTA/40 Ba²⁺: 10.2 mV; 1 EGTA/40 Ca²⁺: 11.4 mV; 1 EGTA/40 Ba²⁺: 11.1 mV [24]. Fraction of peak current remaining after 300-ms depolarization (r₃₀₀) to various voltages were measured. The extent of Ca²⁺/CaM-dependent inactivation (CDI) was calculated as f₃₀₀ = (r_300/Ba − r_300/Ca)/r_300/Ba.

Whole-cell recordings of aGPVMs were performed 20–36 hours post isolation on the same recording setup. Internal solutions for voltage clamp experiments contained, (in mM): CsMeSO₃, 114; CsCl, 5; MgCl₂, 1; MgATP, 4; HEPES (pH 7.3), 10; BAPTA, 10; and ryanodine, 0.005; at 295 mOsm adjusted with CsMeSO₃. External solutions contained (in mM): TEA-MeSO₃, 140; HEPES (pH 7.4), 10; and CaCl₂ or BaCl₂, 5; at 300 mOsm, adjusted with TEA-MeSO₃. These solutions produced an 8.4 mV uncorrected junction potential [24]. For current clamp, experiments, internal solutions contained (in mM): K glutamate, 130; KCI, 9; NaCl,10; MgCl₂, 0.5; EGTA, 0.5; MgATP, 4; HEPES, 10 (pH 7.3 with KOH). External solution (Tyrode's solution) contained (in mM): NaCl, 135; KCl, 5.4; CaCl₂, 1.8; MgCl₂, 0.33; NaH₂PO₄, 0.33; HEPES, 5; glucose, 5 (pH 7.4). Junction potentials for current-clamp solutions were calculated to be only 0.5 mV [24]. The time from upstroke to 80% repolarization (APD₈₀) was used as the metric for action potential duration throughout. SD_cell, the mean standard deviation of APD₈₀ within individual cells, was used to assess the dispersion of APD₈₀ at the same expression level of CaM. Throughout, whole-cell voltage-clamp records were lowpass filtered at 2 kHz, and then digitally sampled at 10 kHz. Current-clamp recordings were filtered at 5 kHz, and sampled at 25 kHz.

2.5 Ratiometric Ca²⁺ Imaging

Single aGPVMs were plated on glass-bottom dishes coated with laminin. Cells were loaded with Indo-1 AM (1 μM) at room temperature for 5 minutes, rinsed, and further incubated for 10 minutes in Tyrode's solution at room temperature to allow for de-esterification of Indo-1 AM. Cells were stimulated by application of an electric field across individual cells using a Grass stimulator (SDD9) and bipolar point platinum electrodes. Recordings were made at room temperature in Tyrode's solution supplemented with 10 μM ascorbic acid [25] to buffer free radicals generated from electrical pacing and exposure to UV light. Fluorescence was measured using 340-nm excitation and 405- to 485-nm emission wavelengths. The intracellular Ca²⁺ concentration ([Ca²⁺]) was calculated as [Ca²⁺] = K_d/Indo · β · (R − R_min)/(R_max − R). R is the ratio of fluorescence signal at 405 and 485 nm. K_d/Indo was determined as 800 nM [26]. R_min was determined to be 0.53 in a 0 mM Ca²⁺ Tyrode's with 5 mM EGTA and 1 μM ionomycin. R_max was determined to be 2.60 in a Na⁺-free Tyrode's (Na⁺ was replaced with choline ion to minimize the action of Na-Ca exchanger) with 10 mM Ca²⁺, 1 μM ionomycin and 10 mM 2,3-butanedione monoxime. β, as defined by the ratio of fluorescence signal at 485 nM under Ca²⁺-free and Ca²⁺-bound conditions, was determined to be 2.33. Cells were stimulated with a single electrical pulse after steady-state pacing at 0.1 Hz. The total amount of Ca²⁺ entry was determined by integration of the area under Ca²⁺-versus-time waveforms. Sarcoplasmic reticulum Ca²⁺ content (SR content) was determined by application of 5 mM caffeine to aGPVMs superfused with a Na⁺-free Tyrode's (Na⁺ was replaced with choline), containing 1.8 mM Ca²⁺ and 10 mM 2,3-butanedione monoxime. The concentration of caffeine was chosen to minimize Indo-1 quenching [27] but was still sufficient to empty the sarcoplasmic reticulum.

2.6 FRET Two-Hybrid Measurement

Three-cube FRET measurements were performed on HEK293 cells cultured on glass-bottom dishes using an inverted fluorescence microscope in modified Tyrode's solution (in mM, NaCl, 138; KCl, 4; CaCl₂, 2; MgCl₂, 1; HEPES, 10; glucose, 10). FRET efficiency (E_A) of individual cells was computed based on a published protocol [15]. Differential expression of test constructs across individual cells allowed decoration of a binding curve. Effective dissociation constants (K_d,EFF) were calculated by fitting the binding curve with the equation E_A = [D]_fee/(K_d,EEF + [D]_free) · E_A,max, where [D]_free is the free concentration of donor molecules.

2.7 Data Analysis and Statistics

All data were analyzed in MATLAB (The MathWorks) using custom-written scripts. For APD₈₀ and Ca²⁺ transient measurements the Wilcox rank sum test was used to assess statistical significance of differences between cells expressing wild-type and mutant CaMs. In addition, variability not due to expression differences was assessed by calculating the standard deviation within each cell (SD_cell) for both APD₈₀ and Ca²⁺ transient measurements. Statistical significance for variability was determined by a student's t-test with the Bonferroni correction for multiple samples as appropriate. Average Ca²⁺ transients are displayed ± SD. Statistical significance for SR content was assessed using a student t-test with a Bonferroni correction for multiple samples. The values are displayed as mean ± SEM. For electrophysiology and FRET two-hybrid measurements, f₃₀₀ and E_A values were expressed as mean ± SEM, and a student's t-test was used to assess statistical significance.

3. Results

3.1 CaM Mutants Promote Proarrhythmic Electrical and Ca²⁺ Activity in Ventricular Myocytes

CaM mutations have been associated with severe LQTS and recurrent cardiac arrest [10], but to date, no direct evidence exists that these mutations can actually promote proarrhythmic properties in an experimental cardiac model. Accordingly, before investigating specific Ca²⁺ regulatory disturbances relating to the interaction of LQTS CaM mutants and individual molecular targets, we tested whether the expression of these mutants at all perturbed the overall electrical and Ca²⁺ cycling properties of aGPVMs. This particular model was chosen because it features action potentials with a prominent plateau phase reminiscent of that in humans, making this system particularly suitable for understanding long-QT phenomena.

Figure 1A displays the prototypic action potentials of a single such myocyte expressing only wild-type CaM (CaM_WT), obtained at 0.5-Hz stimulation under whole-cell current clamp. The timing of current injection stimuli is shown underneath for orientation. The waveforms are nearly identical from one stimulus to the next, with a mean action potential duration (APD₈₀) of ~300 ms [13]. Population behavior for APD₈₀ is summarized in Figure 1B, which plots the cumulative distribution of durations drawn from 285 responses in 10 cells, where P_APD is the probability that APD₈₀ is less than the value on the abscissa. The sharp rise of the distribution confirms a mean duration of 349.6 ms, with a modest standard deviation of 79.6 ms. Additionally, the mean standard deviation of APDs within individual myocytes (SD_cell, intra-cell standard deviation) was only 21 ms, further indicating relatively homogeneous behavior. By contrast, adenoviral-mediated expression of CaM_D96V induced a strikingly different profile (Figure 1C). Here, action potentials could be enormously elongated (red), exceeding even the inter-stimulus interval of 2 seconds. For reference, the control waveform with only CaM_WT present is reproduced in gray. Population data, displayed in cumulative histogram format (Figure 1D), reveal marked lengthening and dispersion of APD₈₀ values (red), with mean and standard-deviation values of 897.3 and 222.9 ms (P < 0.001). Here, SD_cell increased to 156.3 ms, indicating significant variability within each cell as compared to CaM_WT (P < 0.01). Both of these features furnish the cellular substrates for electrically driven arrhythmias at the tissue and organ level [28]. Similar results were obtained for expression of CaM_D130G (Figures 1E–F, APD₈₀ = 915.3 ± 231.7 msec, P < 0.001; SD_cell = 78.5 msec, P ≤ 0.01) and CaM_F142L (Figures 1G–H, APD₈₀ = 864.9 ± 320.1 msec, P < 0.001; SD_cell = 179.1 msec, P < 0.01). The exemplar for CaM_D130G illustrates the occurrence of alternans (Figure 1E), and that for CaM_F142L exemplifies simple APD prolongation. All these behaviors (Figures 1C, 1E, 1G) could be observed in the presence of any of the CaM mutants and persist at faster pacing rates (Supplementary Figure 2). Detailed parameters for action potential recordings are in Table 1 and Supplementary Table 1.

Figure 1. CaM mutants induce arrhythmia.

A, Exemplar action potentials recorded via current clamp from one-day-old aGPVMs transduced with CaM_WT. The stimulus waveform is depicted below. The dashed horizontal gray line indicates 0 mV, here and throughout.

B, Population data corresponding to A plotted as the cumulative distribution of APD₈₀ (285 responses from 10 cells). All APD₈₀ population data in panels B, D, F, and H from myocytes stimulated at 0.5 Hz. Gray bar in B displays SD_cell. Inset shows confocal image of typical myocyte expressing GFP as a marker of transduction by adenoviral CaM_WT.

C, Transduction of the mutant CaM_D96V induced marked prolongation of action potentials in this exemplar recording (red) as compared to CaM_WT transduction (gray, reproduced from A).

D, The cumulative distribution for APD₈₀ from CaM_D96V transduced aGPVMs (red) demonstrates a dramatic increase in APD₈₀ and much greater APD₈₀ variability (gray bar, P ≤ 0.01), both as compared to myocytes transduced with CaM_WT.

E–H, Similar AP disturbances were induced via transduction of CaM_D130G (E, F) and CaM_F142L (G, H). Example of electrical alternans in aGPVM expressing CaM_D130G (E). Data displayed in E and G were obtained during pacing at 1 Hz.

I, Average Ca²⁺ transient (black) recorded from aGPVMs transduced with CaM_WT after steady-state pacing at 0.1 Hz, n =5 cells. Standard deviation range shown as gray shading. Indo-1 AM was used as the inorganic Ca²⁺-sensitive fluorescent dye.

J–L, Transduction of mutant CaMs resulted in dramatic increases in the amplitude and variability of the Ca²⁺ transients. Solid gray trace reproduces the CaM_WT data for reference, while red and rose depict average and standard deviation of Ca²⁺ transients from aGPVMs expressing the three mutants as labeled. Data averaged from n = 11, 9, and 15 myocytes for respective panels J–L.

M, Population data for CaM_WT (black) and CaM_D96V (red), plotted as cumulative histograms for the area integrated under the Ca²⁺ transient waveform recorded from each myocyte. Horizontal error bars depict average standard deviation of Ca²⁺ entry for transients within the same cells, and is significantly larger for each LQTS CaM mutant (P < 0.05). The significant right shift due to transduction of CaM_D96V demonstrates a remarkable increase in Ca²⁺ entry during single APs, while the slower rise indicates greater heterogeneity across myocytes transduced with CaM mutants.

N, Cumulative histograms of Ca²⁺ entry for CaM_D130G (red, left) and CaM_F142L (red, right), with the fit for CaM_WT reproduced in gray.

Table 1.

Average values for APs recorded at 0.5 Hz pacing

CaM	APD80 (ms)	(dV/dt)max (mV/ms)	Vrest (mV)**	Vmax (mV)
WT	417.0 ± 6	119.9 ± 3.2	−62.0 ± 0.2	55.7 ± 0.4
D130G	824.2 ± 16*	112.5 ± 2.1	−61.3 ± 0.2	48.4 ± 0.6
D96V	973.6 ± 12*	139.5 ± 1.7	−62.6 ± 0.1	56.3 ± 0.3
F142L	874.8 ± 22*	131.7 ± 3.8	−61.8 ± 0.2	51.8 ± 0.5
N54I	391 ± 4.5	128.7 ± 1.4	−64.9 ± 0.3	53.6 ± 0.3
N98S	751.8 ± 5.8*	160.6 ± 1.0	−62.4 ± 0.1	58.1 ± 0.1

^*P < 0.01;

^**Values consistent with those expected for aGPVMs after one day in culture [56, 57].

Beyond electrical disturbances, Ca²⁺ cycling dysfunction may also drive arrhythmogenesis [29]. Accordingly, we examined the effect of LQTS CaM mutant expression on intracellular Ca²⁺ transients. Figure 1I displays the typical Ca²⁺ waveform in a myocyte expressing only CaM_WT. Ratiometric Indo-1 imaging was used to gauge Ca²⁺ activity, and data are shown as the mean ± SD drawn from multiple cells. The black trace plots the mean, and standard deviation bounds are shown by gray shadows. Upon expression of CaM_D96V, Ca²⁺ transients are markedly amplified and prolonged (Figure 1J, red). Reproduction of the control CaM_WT waveform (gray) serves to emphasize the strong changes in Ca²⁺ activity. Likewise, expression of CaM_D130G and CaM_F142L produced similarly striking increases of Ca²⁺ transients (Figures 1K–L). Representing these data in cumulative histogram format serves to emphasize the increased dispersion of peak Ca²⁺ transient amplitude produced by CaM mutants (Figures 1M–N). Shown here are the cumulative probabilities of the area under Ca²⁺ transients (P_Q) for CaM_WT and CaM mutants as labeled. The precipitous rise of the wild-type distribution confirms the similarity of Ca²⁺ amplitudes among cells (Figure 1M, black relation). By contrast, the sluggish rise of distributions for CaM mutants (Figures 1M and 1N, red relations) reveals marked heterogeneity of Ca²⁺ transients among cells, as confirmed by significantly larger intra-cell standard deviations (gray bars, P < 0.05). Additionally, both diastolic Ca²⁺ concentrations and SR Ca²⁺ content were significantly elevated by overexpressing LQTS CaM mutants (Supplementary Figure 3). In all, CaM mutants furnish the cellular substrates for Ca²⁺-driven arrhythmias [29], by increasing amplitude and dispersion of Ca²⁺ transients, heightening diastolic Ca²⁺ concentration, and augmenting SR Ca²⁺ content.

3.2 LQTS Calmodulin Mutants Suppress Ca²⁺/CaM-mediated Inactivation of Ca_V1.2 Ca²⁺ Channels

The ability of naturally occurring LQTS CaM mutants to prolong and disperse action potentials was reminiscent of effects we and others observed previously under expression of man-made CaM mutants in the same and similar model systems [13, 30]. There, many of the action potential effects could be attributed to the suppression of a Ca²⁺/CaM-mediated inactivation (CDI) of Ca_V1.2 Ca²⁺ channels. We therefore tested for the effects of the naturally occurring LQTS-related CaM mutants on Ca_V1.2 CDI, heterologously expressed in HEK293 cells for maximal biophysical resolution. In this regard, Ca_V1.2 expression here included the use of an auxiliary β_2a subunit to better visualize CDI effects by minimizing voltage-dependent inactivation [21].

Figure 2A displays exemplar currents of Ca_V1.2 channels coexpressed with CaM_WT. The sharp decay of Ca²⁺ current (red) evoked by a 30-mV depolarizing step is the well-known result of the CDI process. As confirmation, Ba²⁺ current (black) evoked in the same cell hardly decays, as Ba²⁺ binds poorly to CaM. Population data shown below (Figures 2B–C) rounds out characterization of the baseline behavior of channels in the presence of CaM_WT. Figure 2B displays the average of the peak normalized Ba²⁺ current as a function of step potentials, and Figure 2C plots the fraction of peak current remaining after 300-ms depolarization to various voltages (r₃₀₀). The U-shaped Ca²⁺ r₃₀₀ relation (red) recapitulates the classic hallmark of CDI [31, 32], while the flat Ba²⁺r₃₀₀ relation (black) confirms the lack of appreciable inactivation without activation of CaM. Hence, the difference between Ba²⁺ and Ca²⁺ r₃₀₀ relations at 30 mV, as normalized by the corresponding Ba²⁺ r₃₀₀ value, formally gauges the extent of CDI (f₃₀₀ = 0.690 ± 0.028).

Figure 2. LQTS CaM mutants diminish CDI in HEK293 cells.

A, Exemplar currents evoked by 30-mV voltage step (top) in cells co-transfected with Ca_V1.2 and CaM_WT. CDI manifests as the stronger decay in Ca²⁺ (red) current as compared to Ba²⁺ (black). Ba²⁺ trace is scaled downward to match the peak of the Ca²⁺ trace, thus facilitating comparison of decay kinetics, and the scale bar for current references the Ca²⁺ trace, here and throughout.

B, Average normalized peak current versus voltage relation obtained with Ba²⁺ for the same cells as in A. Data are plotted as mean ± SEM here and throughout.

C, Population data for CDI across voltages. r₃₀₀ measures the current remaining after 300 ms, after normalization to peak current. f₃₀₀ is the difference between Ca²⁺ and Ba²⁺ at 30 mV, after normalization by the Ba²⁺ r₃₀₀ value.

D, Expression of CaM_D96V severely blunts CDI of Ca_V1.2.

E, The current-voltage relation for the CaM_D96V scenario remains unaltered.

F, Population data bears out the CaM_D96V reduction of CDI across voltages. For reference, the Ca²⁺ r₃₀₀ curve for CaM_WT is reproduced in gray.

G–L, CaM_D130G and CaMF142L also induce dramatic CDI deficits. Format as in D–F.

Upon coexpressing Ca_V1.2 channels with mutant CaM_D96V, a starkly different functional profile is observed (Figures 2D–F). Here, CDI is strongly suppressed (f₃₀₀ = −0.009 ± 0.008, P < 0.001), without shift in the voltage activation profile (Figure 2E). Similarly, coexpression of channels with CaM_D130G or CaM_F142L also sharply diminished CDI (Figures 2G–I and Figures 2J–L, f₃₀₀ = −0.002 ± 0.011, P < 0.001 and 0.065 ± 0.005, P < 0.001, respectively). The above results were obtained with strong intracellular Ca²⁺ buffering by 10 mM BAPTA, to restrict Ca²⁺ elevations to those in the nanodomains of individual channels, and thereby minimize cell-to-cell variations owing to differences in current amplitudes. Importantly, however, under more physiological Ca²⁺ buffering (1 mM EGTA) that allows global elevation of Ca²⁺, strong but incomplete blunting of CDI was produced by the CaM mutants (Supplementary Figure 4). This residual CDI can be attributed to signaling through the N-terminal lobe of CaM (largely unaffected in LQTS CaM mutants), which is sensitive to sustained global elevation of calcium [33, 34]. Overall, the naturally occurring CaM mutants suppressed Ca_V1.2 channel CDI, in a manner indistinguishable from that of a man-made mutant CaM₃₄ molecules that selectively eliminate Ca²⁺ binding to the C-but not N-terminal lobe of this molecule [20, 33, 34].

By contrast, overexpressing CPVT CaM mutants had weaker effects on Ca_V1.2 channel CDI. CaM_N54I yielded no appreciable change in CDI compared to CaM_WT (Figures 3A–C, f₃₀₀ = 0.583 ± 0.043, P > 0.01). On the other hand, CaM_N98S managed only to partially diminish CDI (Figures 3F–H, f₃₀₀ = 0.367 ± 0.023, P < 0.001). Both results in Figure 3 were obtained under high Ca²⁺ buffering conditions (10 mM BAPTA). Under more physiological Ca²⁺ buffering (1 mM EGTA), we observed a similar trend wherein CaM_N54I and CaM_N98S exerted at most modest diminution of CDI (Supplementary Figure 5). To assess further the more integrative consequences of these CDI profiles (Figures 3A–C, F–H), we investigated the effects of these CPVT CaM mutants within aGPVMs. As might be expected, action potentials in the presence of CaM_N54I were nearly identical to those with CaM_WT (Figures 3D–E). On the other hand, CaM_N98S significantly prolonged action potentials (Figures 3I–J, red P < 0.01) as compared to CaM_WT (gray). Additionally, intra-cell standard deviation (Figure 3J, gray bar) was also larger than CaM_WT (P < 0.05), positioning CaM_N98S for moderate LQTS and affiliated arrhythmias. For CaM_N54I, the nearly complete lack of effect on CDI helps explain why this mutation was not associated with LQTS. Interestingly, the intermediate effects of CaM_N98S on CDI and action potentials match well with reports of LQTS in an unrelated patient [35].

Figure 3. CPVT CaM mutants exert weaker effects on CDI.

A, Exemplar currents evoked by 30-mV voltage step (top) in HEK293 cells co-transfected with Ca_V1.2 and CaM_N54I. Here, coexpression of CaM_N54I did not disrupt CDI compared to CaM_WT (Figure 2A).

B, Average normalized peak current-voltage relationship for same cells as in A. Compared to CaM_WT (Figure 2B), there is no shift in voltage activation.

C, Voltage dependence of r₃₀₀ values for Ca²⁺ (red) and Ba²⁺ (black). Ca²⁺ relation for CaM_WT configuration reproduced in gray for reference. No significant alteration of CDI across all test voltages.

D-E, Coexpression of CaM_N54I does not appreciably effect action potentials (red) as compared to co-expression of CaM_WT (gray). Format as in Figure 1. Data for panel E from 679 action potentials drawn from 6 myocytes.

F, Expression of CaM_N98S, however, modestly diminishes CDI, without affecting voltage activation (G).

H, Population data of r₃₀₀ values confirm a small, but significant reduction of CDI across voltages.

I-J, Overexpression of CaM_N98S has the ability to lengthen and increase heterogeneity of action potentials (P < 0.01). Format as in Figure 1B. Horizontal error bars in panels E and J display standard deviation of APD₈₀ within cells. Data for panel J from 1100 action potentials drawn from 6 myocytes.

Thus far, we have demonstrated the ability of the naturally occurring LQTS CaM mutants to markedly attenuate CDI of Ca_V1.2 channels heterologously expressed in HEK293 cells, to facilitate biophysical resolution. Nonetheless, we next wondered whether similar effects would be observed in native L-type Ca²⁺ currents, as present in the same aGPVMs as used in Figure 1. Figure 4A displays exemplar L-type currents evoked under whole-cell voltage clamp, using 10 mM BAPTA as the intracellular Ca²⁺ buffer, so as to mimic the condition of Figure 2. Ryanodine (5 μM) was included in the intracellular dialyzate to eliminate phasic Ca²⁺ release from the sarcoplasmic reticulum, and limit CDI to that driven by Ca²⁺ entry through individual L-type Ca²⁺ channels [36]. We again observed strong CDI when Ca²⁺ was used as the charge carrier (red) as compared to a limited amount of voltage-dependent inactivation (VDI) seen in the Ba²⁺ current (black). This additional VDI component is expected in this native setting due to a mix of endogenous beta subunits [37, 38], compared to the pure population of β_2a subunits utilized in HEK293 cell experiments. That said, the baseline f₃₀₀ value estimating isolated CDI in control myocytes (Figures 4A–C) was nonetheless 0.67 ± 0.04 (obtained at 20-mV step), which is quite similar to that obtained in recombinant channel expression experiments (Figure 2C). Likewise, population data shows a similar current-voltage relationship and U-shaped Ca²⁺ r₃₀₀ curve (Figures 4B–C). Importantly, expression of mutant CaM_D96V essentially abolished CDI in this native setting (f₃₀₀= −0.18 ± 0.03, P < 0.001, Figures 4D–F) and so did CaM_D130G and CaM_F142L (f₃₀₀= 0.02 ± 0.09, P < 0.001, Figures 4G–I and −0.09 ± 0.03, P < 0.001, Figures 4J–L, respectively), supporting a strong mechanistic link between Ca_V1.2 channel deficits and the LQTS effects seen in patients carrying the CaM mutations.

Figure 4. CaM mutants diminish CDI in aGPVMs.

A, CDI in native LTCCs recorded from one-day-old aGPVMs transduced with CaM_WT. Exemplar current traces elicited by a 20-mV voltage step (top) display strong CDI with Ca²⁺ (red), and a small amount VDI with Ba²⁺ (black).

B, Average normalized peak current-voltage relation obtained in Ba²⁺ for the same cells as in A

C, Population data for CDI across voltages. f₃₀₀ is measured at 20 mV.

D, Expression of CaM_D96V severely blunts CDI in the native setting.

E, Current-voltage relation in the presence of CaM_D96V remains unaltered.

F, Population data confirms the CaM_D96V reduction of CDI across voltages. For reference, the Ca²⁺ r₃₀₀ curve for CaM_WT is reproduced in gray.

G–L, CaM_D130G and CaM_F142L also induce dramatic CDI deficits. Format as in D-F.

3.3 Limited Expression of LQTS CaM Molecules Still Affects Ca_V1.2 Channel CDI

We have so far demonstrated that strong overexpression of LQTS-associated CaM molecules in myocytes can produce both strongly dysfunctional electrical and Ca²⁺ cycling, and potently diminished CDI. However, in the actual related patient population, only one of six alleles encodes a mutant CaM, while the other alleles would elaborate wild-type CaM. Accordingly, we would anticipate that only a limited fraction of CaM molecules would bear the pathogenic mutation [10]. How then could the significant cardiac deficits encountered by patients be rationalized? Previous mechanistic studies of L-type channel CDI offer a potential explanation. In particular, it has been shown that for CDI to occur, channels must initially preassociate with a Ca²⁺-free CaM (apoCaM), to which subsequent Ca²⁺ binding triggers CDI [15, 39]. That is, bulk CaM in the cytoplasm does not appreciably trigger CDI. Thus, if LQTS-associated mutant CaM molecules can still preassociate on par with wild-type CaM, then a sizeable fraction of channels would be bound to mutant CaM, and thus unable to undergo strong CDI (Figures 2–3). Thus, the overall decrease in CDI should be appreciable, reflecting the aggregate fractional presence of mutant CaM in cells.

Accordingly, we utilized a well-established live-cell FRET two-hybrid binding assay to determine whether mutant CaMs can still interact with Ca_V1.2, in a manner similar to wild-type CaM. Figure 5A (top) cartoons the relevant sites of apoCaM interaction with Ca_V1.2 channels, in particular the CI region of the channel carboxy tail. Our FRET assay therefore paired the Ca_V1.2 CI region with CaM (Figure 5A, bottom). As baseline reference, Figure 5B shows the canonical binding curve between the CI region and CaM_WT, where this plot displays the acceptor-centric FRET efficiency of interaction (E_A) as a function of the relative free concentration of donor-tagged molecules D_free (cerulean-CaM_WT). The curve resembles a typical binding reaction, and the D_free that produces half-maximal E_A yields an effective dissociation constant (K_d,EFF) of 12,000 D_free units [5]. Reassuringly all three mutant CaM molecules bind at least as well as wild-type CaM (Figures 5B–5E), demonstrating that mixed expression of mutant and wild-type CaM will result in some fraction of channels bound to mutant CaM.

To test this notion quantitatively, we first devised simple means to control the expression ratio of wild-type to mutant CaM molecules (γ̂) (Supplementary Note 1.6). Then, we performed whole-cell electrophysiology experiments to test explicitly whether the strength of CDI in Ca_V1.2 channels would be graded by different γ̂ values, just as anticipated by the relative binding affinities of channels for mutant versus wild-type apoCaM (Figures 5B–E). Here, our approach was to strongly overexpress variable ratios of such molecules so that the contribution of endogenous CaM would be negligible. If such a scenario were to hold true, we could quantitatively predict that the aggregate CDI strength (CDI) as a function of the protein expression ratio of wild-type to mutant CaM, as the Langmuir equation in Figure 6A (Supplementary Note 1.7). CDI_WT is the full-strength CDI measured with only wild-type CaM strongly overexpressed, and K_d/WT and K_d/MUT are the dissociation constants for channel preassociation with wild-type and mutant apoCaM, as specified in Figures 5B–E. Figure 6B plots this relation explicitly as the smooth black curve. Colored data symbols, with corresponding exemplar traces in Figure 6C, nicely decorate this Langmuir function, as do data from numerous other cells (open symbols in Figure 6B). Similar results were obtained for the other two LQTS-associated CaM mutants (Supplementary Figure 9). Thus, mixtures of wild-type and mutant CaM would weaken L-type channel CDI as predicted by the relative channel affinities for these two molecules in their Ca²⁺-free form. Based on the relative expression profile of each CALM during infancy [10], heterozygous D96V mutation on CALM2 gene yields γ̂ ~ 7, predicting the substantial decrement of CDI indicated by the light rose shading in Figure 6B, likely sufficient to appreciably prolong APDs [30, 40]. Interestingly, the corresponding prediction for a hypothetical homozygous scenario (γ̂ ~ 3, dark rose shading) suggests a severe reduction in CDI, potentially incompatible with life. This would perhaps predict the absence of living homozygous individuals.

In all, we would argue that the electrical and calcium dysfunction affiliated with LQTS-associated CaM mutations arises as summarized in Figure 7. Mutant CaM elaborated by a single allele among three CALM genes would yield a mixture of wild-type and mutant CaM molecules, as specified by the expression ratio γ̂. Because channels must first preassociate with apoCaM to undergo subsequent CDI, this fractional expression of mutant CaM would produce graded reduction of overall CDI in myocytes, as demonstrated in Figure 6. This decrement of Ca²⁺ feedback inhibition would elaborate abnormally long action potentials and QT intervals [13], likely in a cell-specific manner dependent on both the precise value of γ̂, and complex interactions with the configuration of other ion-channel and Ca²⁺-cycling molecules present. The latter interaction factors likely contribute to the impressive dispersion of properties documented in Figure 1. Given the variable propensity for action potential prolongation and calcium augmentation within different cells, arrhythmogenic behavior at the tissue and organ level could thus result. Although other effects of mutant CaM molecules are likely to contribute to overall pathogenesis (Figure 7, gray pathway with arrow), this study furnishes strong evidence that a major underlying mechanism concerns the attenuation of L-type calcium channel CDI by the presence of LQTS-associated mutant CaM molecules. This outcome furnishes at least one major molecular target that merits scrutiny for potential therapeutics.

Figure 7. Proposed mechanism of electrical and calcium dysfunction for LQTS CaM mutants.

Flow diagram schematizing how heterozygous CALM2 mutation might lead to fractional decrease in CDI, yielding action potential prolongation and ultimately long QT phenomena.

4. Discussion

Our experiments demonstrate that CaM bearing LQTS mutations induce the cellular substrates that would favor a LQTS phenotype. Acute introduction of LQTS mutant CaMs into aGPVMs lead to: (1) electrical disturbances including prolonged APD and electrical alternans, as well as (2) Ca²⁺ cycling disturbances, such as increased Ca²⁺ transients and SR Ca²⁺ load. Importantly, such alterations manifested in a highly dispersed fashion across and within cells, thus furnishing a critical ingredient for arrhythmogenesis at the tissue and organ levels [28]. The present study also clearly indicates that a key contributor to these effects involves the disruption of L-type channel CDI by LQTS CaM mutants. Such CDI attenuation would elaborate increased Ca²⁺ current during phases 2 and 3 of the action potentials, thus prolonging APD and increasing SR Ca²⁺ load. Finally, we have established one scenario by which a small fraction of CaM mutants would suffice to create an appreciable prolongation of action potentials. Preassociation of apoCaM to the Ca_V1.2 channels plays a critical role, enabling a fraction of channels to be occupied by the CaM mutants with resulting failure to undergo CDI.

Interestingly, CaM mutants commonly affiliated with CPVT exhibited negligible or weaker effects on action potential duration and L-type channel CDI. The complete lack of effect of CaM_N54I on CDI and action potential duration is well explained by its near wild-type Ca²⁺ binding affinity [41], and these molecular and cellular outcomes fit nicely with the lack of appreciable QT prolongation in corresponding probands [9]. This CPVT-associated mutant could nonetheless interact with other targets like RyR2 calcium release channels to potentially contribute to pathogenesis [41]. On the other hand, the CPVT CaM mutant N98S is capable of producing either CPVT [9, 41] or LQTS in patients [35]. This dual effect may well arise from the overlapping effects of these mutations on multiple CaM targets in the heart. Indeed, CaM_N98S turned out to both reduce L-type channel CDI and moderately prolong cardiac action potentials (Figure 3). The intermediate effects of this CaM mutant thus rationalize how LQTS or CPVT may become the more prominent clinical phenotype, perhaps as a function of differing expression levels among patients.

In addition to L-type (Ca_V1.2) channels, other molecular targets of CaM remain as potential contributors to LQTS pathogenesis. Focusing in particular on targets that preassociate with Ca²⁺-free CaM, voltage-gated Na channels [42] (Na_V1.5) and slow delayed rectifier K channels [43] (I_Ks) loom among likely targets. In Na_V1.5 channels, Ca²⁺/CaM is proposed to both facilitate initial opening and stabilize the inactivated state [42]. However, a recent study reports that LQTS CaM mutants lacked significant effects on most splice variants of Na_V1.5 channels, though the CaM_D130G mutant appeared to moderately enhance persistent current in one fetal splice variant [44]. For I_Ks, Ca²⁺-free CaM may help traffic channels to plasmalemma [45], and Ca²⁺/CaM is believed to facilitate opening. In fact, mutations in I_Ks that disrupt CaM binding result in decreased K current, thus causing LQTS [43, 46]. More broadly, because CaM regulates many other Ca²⁺ channel subtypes, including those predominate in neurons and immune cells, disruption of CDI could lead to a multi-system disorder similar to Timothy syndrome [47–49]. It may well be that extra-cardiac effects are also present in patients possessing LQTS CaM mutants, but that these effects were not recognized in the face of immediately life-threatening cardiac-related sequelae. For other CaM-modulated signaling molecules that do not preassociate with Ca²⁺-free CaM, the present study would suggest that a limited fraction of LQTS CaM mutants would matter little. Only when the fraction of CaM mutants approaches unity would this class of targets be predicted to exhibit altered function. Key members of this class of CaM targets in cardiac myocytes would include Ca²⁺/CaM-dependent kinase II (CaMKII) and calcineurin (CN). CaMKII has been argued to influence the electrical properties of cardiomyocytes by phosphorylation of ryanodine receptors, phospholamban, SERCA, and L-type Ca²⁺ channels, all of which could alter electrical and Ca²⁺ function [50]. By contrast, the Ca²⁺/CaM-activated phosphatase CN dephosphorylates numerous targets including the transcription factor NFAT, implicated in regulating expression levels of numerous ion channels in heart [51]. Nonetheless, if our insights are correct regarding the necessary role of target preassociation with apoCaM to amplify the effects of a limited fraction of CaM mutants, molecules like CaMKII and CN may play little role in the LQTS phenotype at hand.

Even before testing for a role of the additional target molecules alluded to above, potential targeted therapeutic strategies in patients expressing LQTS CaM mutants are suggested by our finding that LTCC dysfunction likely contributes in this particular setting. In addition to beta-adrenergic blockade, as per the general standard of care for LQTS patients, immediate benefits may arise by seeking appropriate modulators of LTCCs such as roscovitine, which has demonstrated beneficial effects within certain in vitro models of LTCC-related LQTS [52]. Additionally, a recent study implicates a non-linear threshold effect between the extent of CDI diminution in LTCCs and onset of outright arrhythmias [53], rather than a continuously graded interrelation. Accordingly, only a few-fold decrease in the fraction of CaM mutants (γ̂) may yield marked improvement of electrical stability and decrease in the incidence of cardiac arrest. The limited alteration of γ̂ potentially required to bring about these benefits may considerably improve the feasibility of devising novel therapies towards this end.

Although the prevalence of diseases caused by de novo CaM mutations is limited, investigating their pathogenesis may offer revealing opportunities to expand our basic knowledge of LQTS-related arrhythmogenesis. Moreover, additional discoveries of CaM mutations will help expand our database of related genotype-phenotype correlations, lending further resources for understanding. Indeed, following the first discoveries of CaM mutations, three more recent preliminary studies [35, 54, 55] have uncovered further CaM-affiliated arrhythmias. These include the following, listed according to gene and syndrome: D134H (CALM2; LQTS), N98S (CALM2; LQTS), D132F (CALM2; LQTS and CPVT), N54I (CALM1; LQTS and/or sudden unexplained death in the young (SUDY)), A103V (CALM3; CPVT and/or SUDY), F90L (CALM1; ventricular fibrillation). These exciting discoveries suggest that a small yet substantial population of patients with CaM mutations is emerging, thus necessitating the inclusion of CALM genes in genetic test panels for LQTS and CPVT, and providing added motivation for the discovery of new therapies. In this light, it may be warranted to dub this expanding group of CaM-related disorders as calmodulinopathies.

Nonstandard Abbreviations and Acronyms

CaM

calmodulin

CPVT

catecholaminergic polymorphic ventricular tachycardia

LQT(S)

long-QT (syndrome)

aGPVM(s)

adult guinea-pig ventricular myocyte(s)

LTCC(s)

L-type Ca²⁺ channels

apoCaM

Ca²⁺-free calmodulin

APD

action potential duration

CDI

Ca²⁺/CaM-dependent inactivation