Calmodulinopathy variants impair Ca_V1.3 and Ca_V2.1 regulation

Calmodulin (CaM) is a small yet indispensable protein encoded by the CALM1, CALM2, and CALM3 genes. It is found in all eukaryotic cells, where it acts as an essential Ca²⁺ sensor (Brehm and Eckert, 1978; Peterson et al., 1999; Qin et al., 1999; Zühlke et al., 1999; Hussey et al., 2023). Each CaM molecule contains four Ca²⁺-binding EF hand domains, which occur in coordinating pairs in each of the two lobes of CaM (Fig. 1 A). Ca²⁺ binding to CaM impacts the function of an array of target proteins (Westerlund and Delemotte, 2018), including voltage-gated calcium channels (VGCCs) where apoCaM preassociates with the C-tail of the channel to act as a resident Ca²⁺ sensor (Peterson et al., 1999; Qin et al., 1999; Zühlke et al., 1999; Ben-Johny and Dick, 2022). Over the past decade, a growing number of missense mutations in CaM have been identified in patients suffering from debilitating and life-threatening cardiac arrhythmias, known as calmodulinopathies (Nyegaard et al., 2012; Crotti et al., 2013; Crotti et al., 2019; Crotti et al., 2023; Limpitikul et al., 2014; George, 2015; Reed et al., 2015; Nyegaard and Overgaard, 2019; Hussey et al., 2023). Disease phenotypes primarily include long-QT syndrome (LQTS) and/or catecholaminergic polymorphic tachycardia (CPVT) (Fig. 1 A) (Nyegaard et al., 2012; Crotti et al., 2023), with some reports describing neurological comorbidities, including developmental delay, seizures, and autism spectrum disorder (ASD) (Crotti et al., 2023). Moreover, some CaM variants have been identified in patients with predominantly neurological symptoms and minimal or no identified cardiac pathology (Crotti et al., 2023), demonstrating a diverse pathogenic phenotype.

Investigations have revealed multiple pathological mechanisms attributable to calmodulinopathy variants, often involving a diminished ability of CaM to bind to Ca²⁺ (Nyegaard et al., 2012; Crotti et al., 2013; Limpitikul et al., 2014). Many of these variants were found to impair the Ca²⁺/CaM-dependent regulation of the VGCC Ca_V1.2, an essential component of cardiac function known to contribute to LQTS (Limpitikul et al., 2014; Limpitikul et al., 2017; Dick et al., 2016; Bamgboye et al., 2022). In these channels, CaM-dependent regulation takes the form of Ca²⁺/CaM-dependent inactivation (CDI) (Brehm and Eckert, 1978; Peterson et al., 1999; Qin et al., 1999; Zühlke et al., 1999), a vital form of negative feedback that tempers Ca²⁺ entry into cells. While each lobe of CaM is independently capable of imparting channel regulation (DeMaria et al., 2001; Ben-Johny and Dick, 2022), the majority of Ca_V1.2 CDI stems from the C-lobe, consistent with the fact that almost all LQTS-associated variants identified thus far are located within the C-lobe of CaM (Crotti et al., 2023; Hussey et al., 2023).

In addition to Ca_V1.2, Ca_V1.3 and Ca_V2.1 also exhibit robust C-lobe CaM-mediated regulation (Lee et al., 1999; Erickson et al., 2001; Ben-Johny and Dick, 2022), making them potential pathogenic targets in calmodulinopathies. As in Ca_V1.2, CaM modulates Ca_V1.3 by conferring CDI, with distinct contributions from each lobe of CaM (DeMaria et al., 2001). Additionally, multiple splice variants of Ca_V1.3 exist with distinct CaM regulatory features. While a short form of the channel (Ca_V1.3_43s) exhibits robust CDI, a long splice variant, Ca_V1.3_L, contains a C-terminal domain, termed inhibitor of CDI (ICDI) (Wahl-Schott et al., 2006) or C-terminal modulatory domain (CTM) (Lieb et al., 2012), that diminishes CDI by competing with CaM binding to the channel (Liu et al., 2010). These splice variants are crucial for adjusting the channel’s activity for distinct cellular requirements, demonstrating the importance of fine-tuned CaM regulation in physiological function (Bock et al., 2011; Ben-Johny and Dick, 2022). For Ca_V2.1, CDI is primarily mediated by the N-lobe of CaM, while the C-lobe drives Ca²⁺-dependent facilitation (CDF), characterized by a transient elevation of whole-cell currents following Ca²⁺ entry through the channel (DeMaria et al., 2001). Each of these channels are widely expressed in multiple brain regions (Hell et al., 1993; Tan et al., 2011; Catterall et al., 2013; Nanou et al., 2018), and disruption of these channels has been associated with ASD (Pinggera et al., 2015; Limpitikul et al., 2016), developmental delay (Folacci et al., 2023), and other neurological disorders (Pietrobon, 2010; Indelicato et al., 2019; Folacci et al., 2023; Kramer et al., 2023; Timothy et al., 2024), making them potentially relevant to the neurological features in calmodulinopathies.

In this study, we examine the biophysical properties of calmodulinopathy variants in the context of Ca_V1.3 and Ca_V2.1, both in terms of channel regulation and binding affinity. We focus on mutations identified within the C-lobe of CaM previously shown to decrease the affinity of Ca²⁺ binding to CaM and reduce the CDI of Ca_V1.2 channels (Crotti et al., 2013; Limpitikul et al., 2014). We find that many of these mutations exhibit profound disruptions of Ca_V1.3 CDI and Ca_V2.1 CDF, demonstrating a potential role for these channels in calmodulinopathy pathogenesis. Comparatively, a CPVT-associated mutation within the N-lobe of CaM demonstrated no significant impact on CDI or CDF of these channels, consistent with a primarily cardiac pathology acting through the ryanodine receptor (RyR2), as previously described (Hwang et al., 2014; Søndergaard et al., 2015a; Søndergaard et al.,2020). Thus, our results demonstrate a potential mechanism by which LQTS-associated CaM variants may impact multiple VGCCs including those outside the heart.

For electrophysiological experiments, a plasmid encoding a human α_1C channel variant containing exon 8a was employed for Ca_V1.2 (Z34810) (Dick et al., 2008). Plasmids encoding a human α_1D channel variant containing exons 8a and 42 was employed for Ca_V1.3_L (EU363339), and a variant terminating in exon 43 was employed for Ca_V1.3_43s (GenBank accession no. NM_001128839) (Sang et al., 2021). These plasmids were generously provided by Dr. Jörg Striessnig (University of Innsbruck, Innsbruck, Austria). A plasmid encoding an α_1A channel variant, including exons 43 and 44 and excluding exon 47, originally provided by Dr. Terrance Snutch (University of British Columbia, Vancouver, BC, Canada), was employed for Ca_V2.1 (GenBank accession no. NM_001127221) (Lee et al., 2015). Each of these was within a pcDNA3 expression vector. For these same experiments, human CALM1 cDNA within the pIRES2-EGFP vector was utilized for WT CaM (Clontech Laboratories). From this WT CaM-pIRES-EGFP construct, pathogenic CaM mutations were generated using QuikChange Lightning Site-Directed Mutagenesis (Agilent). 8 μg each of human α₁ cDNA, rat brain β_2a (GenBank accession no. M80545.1) or rat brain β_1b (GenBank accession no. X61394), and rat brain α_2δ (GenBank accession no. NM_012919.3) subunits, and WT or mutant CaM were heterologously co-expressed. For all recordings, β_2a was used unless otherwise specified. Expression was enhanced with 2 μg of simian virus 40 T antigen cDNA. Expression of all plasmids was under the control of a cytomegalovirus (CMV) promoter. HEK293 cells were cultured in 10-cm dishes on glass coverslips, and channels were transfected using an established Ca²⁺ phosphate method (Peterson et al., 1999). Data from each experimental condition were gathered from a minimum of two independent transfections.

Whole-cell voltage-clamp recordings of HEK293 cells were carried out 1–2 days after transfection at ambient temperature. An Axopatch 200B amplifier (Axon Instruments) was utilized to generate the recordings. Noise in these recordings was minimized via a lowpass filter at 2 kHz, followed by digital sampling at 10 kHz. Pipette resistances were 1.5–5 MΩ and P/8 leak subtraction was used. Internal solutions contained (in mM): 114 CsMeSO₃, 5 CsCl, 1 MgCl₂, 4 MgATP, 10 HEPES (pH 7.4), and 10 BAPTA, at 295 mOsm adjusted with CsMeSO₃. External solutions contained (in mM): 140 TEA-MeSO₃, 10 HEPES (pH 7.4), and 40 (or 5 for Ca_V2.1) CaCl₂ or BaCl₂ 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. Note that the use of 10 BAPTA limits Ca²⁺ to only the local signal due to entry through the channel (Dick et al., 2008). For Ca_V2.1, this manipulation restricts CaM regulation to C-lobe–mediated CDF (Ben-Johny and Dick, 2022), reducing the confounding effect of N-lobe–mediated CDI in our CDF measurements. However, it should be noted that this manipulation also reduces the ability to observe N-lobe effects; thus, the N54 mutation is used largely as a negative control under these conditions.

For r₃₀₀ Ca²⁺ data in Ca_V1.2/CaM experiments, a ratio of peak current during a step depolarization to 30 mV to the current remaining following 300 ms was employed for each CaM variant. For each cell, this value was subtracted from the analogous value for Ba²⁺ at 30 mV, resulting in an f₃₀₀ value that quantified CDI. For both Ca_V1.3/CaM datasets, similar comparisons were employed, with the following differences: currents were compared at 10 mV step depolarizations, and the 1,000 ms depolarization time point was used, resulting in f₁₀₀₀ values. For experiments investigating Ca_V1.3/CaM VDI, r₁₀₀₀ values acquired during Ba²⁺ depolarizations were used to estimate VDI.

For Ca_V2.1/CaM experiments, CDF was characterized by comparing Ca²⁺ currents during depolarizing voltage steps to 10 mV in the presence and absence of a prepulse, which was sufficient to partially facilitate the channel (DeMaria et al., 2001; Chaudhuri et al., 2004). To adjust for variations in current density, each sweep is normalized relative to the end of the 50 ms test pulse. This time point is chosen such that CDF is complete and CDI is minimal. ΔQ is quantified as the area between the currents with and without the prepulse. To normalize for the time course of facilitation, ΔQ is divided by the time constant of the curve without the prepulse τ, giving an approximation of the degree of CDF triggered by the prepulse, termed relative facilitation (RF). In cases where the lack of CDF led to an absence of a monoexponential curve to derive τ from, τ was set as the average value from WT dataset (8.5). Data from prepulses to 20 mV were used for statistical comparisons.

To evaluate voltage dependence of activation, Ca²⁺ current densities of each variant were plotted as function of depolarization voltage. Data for each cell were normalized to peak value and averaged across cells. Data points were fit by the curve Y = G_max*(X − V_rev)/{1 + exp[(V₅₀ − X)/Slope]}. The half-maximal voltage of activation was calculated from these curves.

Statistics: For Ca_V1.2, the f₃₀₀ value from each cell was averaged for the mutant condition and compared with the WT data via a Student’s t test. For all patch-clamp experiments where more than one mutant was being compared with WT (Ca_V1.3 and Ca_V2.1), the relevant values were compared via one-way ANOVAs using Dunnett’s multiple comparison corrections, with each mutant being compared with WT. Data from individual cells were excluded from analysis if peak Ca²⁺ currents were below 150 pA, if excessive leak current (>150 pA) was present, or if series resistance exceeded 10 MΩ. For comparisons of half-maximal voltage of activation, a sum-of-squares F test was employed with a threshold P value = 0.05.

The principles underlying flow cytometric FRET were previously established (Ben-Johny et al., 2016; Lee et al., 2016; Rivas et al., 2021). Venus-CaM plasmids correspond to those described previously (Ben-Johny et al., 2016; Lee et al., 2016; Niu et al., 2018). QuikChange Lightning mutagenesis (Agilent) was employed to create the Venus-CaM mutants of interest. For Ca_V1.3, the C-terminal fragment (DNFDY…VGKYP) was ligated into a pcDNA3 expression vector containing an upstream Cerulean fluorophore. The DNA template encoding the Ca_V1.3 fragment as well as XbaI and NotI restriction digest sites at positions flanking the coding sequence was purchased (Twist Bioscience). Using this template, digestion with these enzymes was carried out, followed by ligation into the expression vector using T3 DNA ligase (New England Biolabs). To ensure expression and read through, the Cerulean fluorophore was preceded by a CMV promotor, and an in-frame short linker sequence of alanine was present between the end of the Cerulean sequence and before the Ca_V1.3 fragment. At the end of the Ca_V1.3 fragment was a stop codon to create the fusion protein suitable for FRET. The same procedure was employed to generate the Cerulean/Ca_V2.1 construct, with the Ca_V2.1 sequence (DNFEY…EOGPP) inserted into the vector. Dimers linking Cerulean and Venus of various lengths (C5V, C32V, C40V, and CTV) were utilized as previously described (Ben-Johny et al., 2016; Lee et al., 2016; Rivas et al., 2021). All constructs were verified by sequencing (Plasmidsaurus).

Cells were seeded in 6-well plates and transfected with FRET constructs and fluorophore pairs using a PEI protocol (Longo et al., 2013). 48 h after transfection, cells were treated with 100 μM cycloheximide for 2–4 h to reduce incidence of immature fluorophores and were then harvested and resuspended in Tyrode’s solution with minimal Ca²⁺. For experiments involving calcified CaM, 10 mM CaCl₂ was added to equal volumes of cell suspension along with 4 μM ionomycin (I0634; Sigma-Aldrich).

To estimate binding curves and K_D,eff for each of the donor–acceptor pairs, fluorescence measurements are acquired from three filters within the flow cytometer, specifically selected to detect donor, acceptor, and FRET signals. This technique, termed 3³-FRET, permits the isolation of FRET signal from donor and acceptor signals (Erickson et al., 2001). This methodology required evaluation of instrument-specific properties during each experimental run, evaluated using a series of linked Cerulean/Venus dimers such that a 1:1 stoichiometry of each fluorophore was ensured (Ben-Johny et al., 2016). Moreover, Venus and Cerulean fluorophores, individually and co-expressed, were measured to permit evaluation of nonspecific fluorescence that could confound FRET analysis. Cells were kept in normal Tyrode’s solution for recording containing (in mM): 138 NaCl, 4 KCl, 1 MgCl₂, 10 HEPES (pH 7.4), at 300 mOsm adjusted with NaOH. External solutions contained (in mM): 140 TEA-MeSO₃, HEPES (pH 7.4). For high-Ca²⁺ conditions, external Ca²⁺ was increased to 10 mM, and 4 µM ionomycin was added to facilitate Ca²⁺ entry into cells prior to recording.

Statistics: Analysis of flow cytometric FRET was performed as previously described (Ben-Johny et al., 2016; Lee et al., 2016; Rivas et al., 2021). Net decreases in donor fluorescence due to FRET were computed from the signals from each of the specific filter cubes utilized. These decreases are analogous to FRET efficiencies acquired from acceptor photobleaching or donor quenching approaches and correspond to the donor-specific FRET measurement, E_D. For each of the >2,000 cells within each sample, E_D was plotted against the concentration of unbound acceptor molecules (A_free) to generate a binding curve, and effective K_D (K_D,eff) was then calculated from this curve. K_D,eff values from each FRET experiment were compared using one-way ANOVAs using Dunnett’s multiple comparison corrections, with each mutant being compared with WT. Data were analyzed by custom MATLAB software (Ben Johny, 2020) (MathWorks).

Peptides representative of the Ca_V1.3 and Ca_V2.1 IQ domains were N-terminally labeled with a 5-TAMRA fluorophore (Ca_V1.3: 5-TAMRA-DEVTVGKFYATFLIQDYFRKFKKRKEQGLVGKYP-NH2, Ca_V2.1:_5-TAMRA-TDLTVGKIYAAMMIMEYYRQSKAKKLQAMREEQD-NH2; ProteoGenix, purity >95%). Using a previously described 384-well plate-based assay (Brohus et al., 2019), CaM was titrated across the 24 columns at eight different free calcium concentrations (3 nM to 400 µM), 1 per row (Brohus et al., 2021). The change in fluorescence anisotropy (FA) of the TAMRA-labeled IQ domains, a measure of their tumbling rate, was monitored as a function of CaM concentration. This setup resulted in eight CaM/IQ domain-binding curves, each representing binding at a specific free Ca²⁺ concentration.

Due to the high CaM/Ca_V1.3 IQ domain affinity at high Ca²⁺ concentrations, K_D values were difficult to determine robustly between replicates. For this reason, all determined K_D values from all replicates were combined to provide enough data points to fit the EC₅₀ model. From the EC₅₀ model, the fitted initial plateau (at low Ca²⁺ concentrations) was extracted to represent the apo-affinity, and the fitted final plateau (at high Ca²⁺ concentrations) was extracted to represent the CaM/Ca_V IQ domain affinity at saturating Ca²⁺ concentrations (Fig. 7 A, lower panel horizontal stapled lines). The fitted EC₅₀ values were extracted to represent the Ca²⁺ concentration that provides the half-maximal change in logK_D (binding energy) between the apo- and Ca²⁺-bound states—a measure of the Ca²⁺ sensitivity of the interaction (Fig. 7 A, lower panel vertical stapled line). In summary, the assay provides information on the Ca²⁺ dependency of the CaM/IQ domain interaction across the physiologically relevant Ca²⁺-concentration span.

Statistics: Curve fitting and data plotting were done using a custom-written script pipeline in R (v. R-4.2.1). Statistical testing was done on log₁₀-transformed affinities and EC₅₀ values in GraphPad Prism 6 (v. 6.07) using a one-way ANOVA (WT as reference) with Dunnett’s post-hoc test for multiple comparisons. Significance levels were indicated by asterisks: **** (P value <0.0001), *** (P value <0.001), and * (P value <0.05).

Data S1 contains all data underlying each figure.

Calmodulinopathy variants have been associated with a range of symptoms and evaluated for effects on a variety of target proteins (Crotti et al., 2013; Crotti et al., 2019; Hwang et al., 2014; Yin et al., 2014; George, 2015; Jensen et al., 2018; Nyegaard and Overgaard, 2019; Clemens et al., 2020; Ledford et al., 2020; Kang et al., 2021). Here, we focus on several variants that have been strongly implicated in LQTS and sudden death, with variable reports of neurological features (Fig. 1, A and B). Mutations D96V, F142L, D130G, and D132G reside in the C-lobe of CaM and are strongly associated with LQTS. They have each been documented in patients suffering from ASD, developmental delay, and/or epilepsy, with F142L associated with a broader set of reported neurological deficits (Crotti et al., 2013; Kaplanis et al., 2020; Crotti et al., 2023). While D96V has a reported association with developmental delay (Crotti et al., 2013), the neurological effects appear less consistently reported (Crotti et al., 2013; Crotti et al., 2023) and were subsequent to cardiac arrest, making it difficult to rule out hypoxic injury. CaM mutations at residue G114 have previously been associated with sudden death and cardiac arrest; however, neurological symptoms were not reported (Brohus et al., 2021; Crotti et al., 2023). Finally, we include the N-lobe mutation CaM N54I, which has a well-evidenced pathological role in CPVT (Nyegaard et al., 2012) with the effects largely independent of VGCC channels, making it a useful control for nonspecific effects.

In the case of LQTS, CaM mutants are thought to contribute to the disease largely through the disruption of Ca_V1.2 CDI (Limpitikul et al., 2014). The majority of C-lobe mutations focused on in this study have previously been shown to blunt Ca_V1.2 CDI, while the CPVT-associated N54I mutation largely spares CDI (Fig. 1 C) (Limpitikul et al., 2014; Brohus et al., 2021). However, the effect of the CaM mutation D132G has yet to be evaluated on Ca_V1.2 despite its association with severe LQTS (Zahavich et al., 2018) and developmental delay (Kaplanis et al., 2020). We therefore sought to determine the impact of CaM D132G on Ca_V1.2 CDI. To do so, we co-expressed mutant CaM with Ca_V1.2 in HEK293 cells, such that CDI can be seen as the larger decay of the Ca²⁺ current (red), as compared with the Ba²⁺ current (black) in response to a sustained depolarization from −80 to 30 mV (Fig. 1 D, top). CDI is then quantified by comparing the ratio of Ca²⁺ current that remains following 300 ms of depolarization (r₃₀₀) with that of Ba²⁺. In this way, Ba²⁺ currents serve to control for the VDI of the channels, which occurs alongside CDI and is minimized in our recording conditions by the use of the β_2a auxiliary subunit. In the presence of CaM D132G, the Ca²⁺ current is largely sustained following the depolarizing step, demonstrating a loss of CDI, which persists across a range of test potentials (Fig. 1 D, bottom). Thus, CaM D132G displays a similar capacity to reduce Ca_V1.2 CDI as compared with other previously documented LQTS-associated CaM variants (Limpitikul et al., 2014) (Fig. 1 C).

Given that regulation of channel gating by CaM extends beyond Ca_V1.2, we investigated the effects of pathological CaM variants on Ca_V1.3 inactivation, starting with the Ca_V1.3_L splice variant. Ca_V1.3_L channels represent a long splice variant (also called Ca_V1.3₄₂ [Bock et al., 2011]), which includes the ICDI/CTM (Singh et al., 2008; Lieb et al., 2012), resulting in a channel with attenuated CDI. Nonetheless, under conditions where exogenous CaM is overexpressed, robust CDI is observed in the presence of CaM WT (Fig. 2 A). Strong CDI (red) is observed across a range of depolarizing potentials (Fig. 2 A, middle), while Ba²⁺ currents (black) exhibit only modest VDI. While the N-lobe CaM N54I produced no significant deficits in CDI (Fig. 2 B), each of the C-lobe CaM mutations imparted a significant reduction in CDI as compared with CaM WT (Fig. 2, C–G). We also note that the CDI observed in the presence of G114R displayed more rapid inactivation kinetics as compared with the other mutants. As CaM is capable of mediating CDI in Ca_V1.3_L channels through both the C- and N-lobes (Dick et al., 2008), the remaining CDI observed in the calmodulinopathy variants is likely due largely to N-lobe CDI, which is known to exhibit slower inactivation kinetics as compared with C-lobe–mediated CDI (Peterson et al., 1999; Yang et al., 2006; Dick et al., 2008). Consistent with this, an artificially mutated CaM₁₂₃₄, in which Ca²⁺ binding is eliminated in both lobes of CaM, results in complete abolition of CDI in Ca_V1.3_L (Fig. 2 H). Thus, calmodulinopathy variants can impair Ca_V1.3_L CDI in a manner similar to that observed in Ca_V1.2 (Fig. 1).

Analysis of the voltage dependence of activation (Fig. 2, A–H, bottom) revealed no difference across each condition; however, evaluation of Ba²⁺ current decay during the depolarization revealed a slight but significant variability in VDI. Specifically, at higher test potential voltages, CaM G114R imparted greater VDI than other mutants or WT CaM as revealed by Ba²⁺r₁₀₀₀ values (Fig. 2 D, black). Additionally, CaM D132G appeared to follow a similar trend, but this effect did not rise to the level of significance (Fig. 2 F). To investigate this phenomenon further, these mutants were co-expressed with Ca_V1.3_L in conditions optimized to reveal voltage-dependent effects. Employing the β_1b accessory subunit in the transfection permits robust VDI in Ca_V1.3_L channels (Fig. 3 A, top). Once again, the G114R mutant revealed a slight but significantly enhanced VDI as compared with CaM WT or CaM D132G, as revealed by smaller r₁₀₀₀ values following 20 mV depolarizations (Fig. 3, A–C, bottom panel). Thus, it appears that VDI is uniquely different in the CaM G114R recordings.

We next considered the impact of the CaM mutants on a short isoform of Ca_V1.3, Ca_V1.3_43s, which lacks the modulatory ICDI/CTM domain and thus displays full CDI (Bock et al., 2011). The CaM mutants had generally similar impacts on CDI in Ca_V1.3_43s as they did in Ca_V1.3_L (Fig. 4, A–H). As expected, the N54I N-lobe CPVT mutant was not statistically different from WT (Fig. 4 B), while mutations D96V, D130G, D132G, and F142L all significantly impaired CDI (Fig. 4, B, C, and E–G). The artificial mutation CaM₁₂₃₄ produced a near total lack of Ca_V1.3_43s CDI (Fig. 4 H). In contrast, CaM G114R, which displayed diminished CDI as compared with overexpressed CaM WT in Ca_V1.3_L, showed no effect on Ca_V1.3_43s CDI (Fig. 4 D). Additionally, no effects on Ca_V1.3_43s VDI were observed.

While most C-lobe CaM mutants conferred Ca_V1.3_L and Ca_V1.3_43s CDI deficits in a similar fashion to their action on Ca_V1.2, data from CaM G114R diverge from this paradigm. It is therefore reasonable to question how CaM G114R interacts with Ca_V1.3. Notably, the results for each Ca_V1.3 splice variant would be consistent with a lack of CaM G114R binding to the channel. Thus, the possibility emerges that CaM G114R does not interact with either Ca_V1.3 variant, and the results demonstrate the expected CDI under endogenous CaM WT levels (Liu et al., 2010). Moreover, any mutant CaM must bind to the channel at least as well as CaM WT in the apo state in order for the variant CaM to preassociate with a significant portion of channels in the context of mixed WT and mutant CaM expression (Erickson et al., 2003; Limpitikul et al., 2014; Badone et al., 2018; Hussey et al., 2023). We therefore sought to explore whether each mutant CaM is capable of binding the channel with a similar affinity as compared with CaM WT, utilizing FRET. To quantify binding activity, we employed a previously validated (Ben-Johny et al., 2016; Lee et al., 2016; Rivas et al., 2021) live-cell flow cytometric methodology to construct FRET-based binding curves and estimate the effective binding affinity (K_D,Eff) for each channel/CaM pair (Table 1). A peptide spanning known C-terminal CaM-binding regions of Ca_V1.3 was ligated to a donor fluorophore, while CaM was joined to an acceptor fluorophore (Fig. 5 A), and the two were expressed in HEK293 cells under basal Ca²⁺ conditions. Exemplar binding curves (Fig. 5, D–I) of CaM variants N54I, D96V, D130G, D132G, and F142L show broad similarity with the binding curve of CaM WT, and the average results from each of these mutants demonstrate no significant alterations to K_D,Eff relative to WT (Fig. 5 B and Table 1). However, CaM G114R demonstrates a profoundly diminished K_D,Eff (Fig. 5, B and F; and Table 1), where maximal saturation could not be achieved under the experimental condition. Thus, CaM G114R appears to exhibit reduced preassociation with the channel in the apo state.

Given that CaM binding can be impacted by its association with Ca²⁺ (Brohus et al., 2021), we investigated whether CaM G114R/Ca_V1.3 binding could be recovered under conditions sufficient to calcify CaM. Subsequent testing revealed that, in the presence of an ionophore and high external Ca²⁺, binding activity could be partially rescued (Fig. 6, A–C). These data indicate that calcified CaM G114R maintains the ability to bind Ca_V1.3 at high Ca²⁺, but preassociation in the apo-state is likely reduced. Given this result, the modified CDI observed in Ca_V1.3_L (Fig. 2D) as compared with overexpressed CaM WT (Fig. 2 A) is likely due to the known competition between CaM and ICDI/CTM for binding to the Ca_V1.3 IQ (Fig. 6 D). At low, endogenous levels of CaM WT, ICDI/CTM is able to out-compete CaM for binding to the channel, resulting in a significant reduction of CDI as previously described (Singh et al., 2006; Wahl-Schott et al., 2006; Liu et al., 2010; Lieb et al., 2012; Sang et al., 2021). Upon overexpression of CaM WT, CDI is restored due to mass action. Without sufficient apoCaM binding, CaM G114R does not alter this scheme. Moreover, in Ca_V1.3_43s lacking ICDI/CTM, endogenous CaM is sufficient to drive robust CDI, and the lack of CaM G114R binding causes no apparent difference in CDI.

In addition to investigating the interaction between CaM and the Ca_V1.3 IQ in low and high Ca²⁺ conditions, we explored the Ca²⁺ dependence of the interaction of each CaM variant with the channel domain. This was done by monitoring the change in FA of the 5-TAMRA–labeled Ca_V1.3 IQ during titration with CaM at eight different Ca²⁺ concentrations (Fig. 7), as previously done for the Ca_V1.2 IQ (Wang et al., 2018; Wang et al., 2020; Brohus et al., 2021). The affinity of CaM WT for Ca_V1.3 IQ dramatically increased as a function of Ca²⁺-concentration (Fig. 7 A). Likewise, the affinity of all other CaM variants increased with Ca²⁺ concentration (Fig. 7, B–G). As expected, the Ca²⁺-insensitive variant, CaM₁₂₃₄, retained an apo-affinity for the Ca_V1.3 IQ across the entire range of Ca²⁺ concentrations tested, demonstrating its inability to bind Ca²⁺ ions (Fig. 7 H). To summarize the effects of the CaM mutations on the interaction with the Ca_V1.3 IQ, the complex affinity in apo- and Ca²⁺-saturating conditions, and its Ca²⁺ sensitivity were estimated from the EC₅₀ fits (Fig. 7 A, bottom panel dashed lines). First, the complex affinities in apo- and Ca²⁺-saturating conditions were estimated by the initial and final plateaus, respectively (Fig. 7 A, horizontal dashed lines and Fig. 7, I and J). In support of the FRET experiments and the hypothesis of reduced channel preassociation, CaM G114R had the largest effect on the apo-affinity (defined by the initial plateau), causing a significant 10-fold reduction compared with CaM WT (Fig. 7 I). CaM D130G and D132G significantly reduced the apo-affinity by twofold compared with CaM WT. Of note, only CaM F142L caused an increase in the apo-affinity (13-fold compared with CaM WT), in agreement with what has been observed for the Ca_V1.2 IQ (Wang et al., 2018). In Ca²⁺-saturating conditions, defined by the final plateau of the EC₅₀ curve, the generally high affinity caused a more variable determination of K_D values, in turn causing a less well-defined final plateau compared with the initial plateau (apo-conditions). Despite the higher variability at high Ca²⁺ concentrations, we tested for significant differences in the CaM/Ca_V1.3 IQ-domain affinities. CaM D130G reduced the IQ-domain affinity 5.5-fold compared with CaM WT, whereas CaM D132G and F142L caused a 3.5-fold and 2-fold increase, respectively (Fig. 7 J). As a measure of the Ca²⁺ sensitivity of CaM when in a complex with the Ca_V1.3 IQ domain, we determined the Ca²⁺ concentration causing a half-maximal change in log(K_D) (representing binding energy) between the apo- and Ca²⁺-bound states (EC₅₀) (Fig. 7 A, dotted vertical line and Fig. 7 K). The EC₅₀ value for CaM WT/Ca_V1.3 IQ is 400 nM free Ca²⁺. All C-lobe mutations impaired the Ca²⁺ sensing of CaM when complexed to the Ca_V1.3 IQ (2- to 11-fold) in line with their reduction of C-lobe Ca²⁺-affinity (Fig. 7 K) and consistent with their effect in impairment of Ca_V1.3 CDI.

Having evaluated the impacts of pathological CaM mutants on Ca_V1.3, we next turned to investigate the P/Q-type channel, Ca_V2.1. To do so, we utilized a similar experimental setup as before with Ca_V2.1 transfected into HEK293 cells along with the mutant CaMs. For Ca_V2.1, C-lobe–mediated CaM regulation takes the form of CDF (DeMaria et al., 2001). To evaluate CDF, we employed a step depolarization, with and without a prepulse at varying potentials. Ca²⁺ entry during the prepulse initiates CDF, resulting in a larger initial current during the test pulse. To quantify the amount of CDF, the area between the two curves is integrated to give the RF value (Fig. 8) (DeMaria et al., 2001; Chaudhuri et al., 2004). As expected for a C-lobe–mediated process, the N-lobe CaM mutant N54I did not significantly impair Ca_V2.1 CDF as compared with WT (Fig. 8 B). On the other hand, the C-lobe mutants G114R, D130G, D132G, and F142L all significantly reduced CDF, with G114R and F142L displaying somewhat more moderate effects (Fig. 8, D–G). The artificial mutant CaM₁₂₃₄ demonstrates the impact of a complete loss of Ca²⁺ binding to CaM (Fig. 8 H). Interestingly, the C-lobe mutant D96V did not display impaired CDF (Fig. 8 C) despite the fact that the mutated residue resides in the third Ca²⁺ binding EF hand and has been shown to dramatically reduce Ca²⁺ binding (Limpitikul et al., 2014; Søndergaard et al., 2015b). Finally, evaluation of the normalized current–voltage relationship demonstrated small (<10 mV) shifts in the half-maximal voltage of activation of Ca²⁺ currents for G114R, D132G, and F142L variants relative to WT (Fig. 8, bottom panels).

Given the lack of CDF effect for D96V, we wondered if the affinity of the D96V mutant for Ca_V2.1 might be impaired as seen for G114R for Ca_V1.3. As before, we evaluated the capacity of mutant CaMs to bind via FRET. Using a C-terminal fragment of Ca_V2.1 coupled to a cerulean fluorophore, the donor molecule of the FRET pair was constructed. Similarly, each mutant CaM was ligated to Venus as the acceptor fluorophore. Evaluation of the FRET-binding curves to estimate K_D,eff under basal Ca²⁺ conditions revealed that CaM G114R alone displayed significantly altered binding affinity (Fig. 9, A–G; and Table 2), although the effect was much milder as compared with the binding deficits observed in Ca_V1.3. Interestingly, CaM₁₂₃₄ displayed a similarly mild but significant increase in K_D,eff (Fig. 9 J and Table 2), implying that even in the apo state, this form of CaM has diminished binding affinity.

To explore the Ca²⁺-dependency of the CaM/Ca_V2.1 IQ interaction, we once again used the FA assay (Fig. 10). The affinity of CaM WT for the Ca_V2.1 IQ dramatically increased with increasing Ca²⁺ concentration (Fig. 10 A). All other CaM variants, except CaM₁₂₃₄, followed the same pattern of increasing complex affinity with Ca²⁺ concentration (Fig. 10, B–H). In the low Ca²⁺ apo-state, all CaM C-lobe mutations displayed a small but significant reduction in binding affinity for the Ca_V2.1 IQ domain (1.2- to 2.7-fold compared with CaM WT) (Fig. 10 I). Of note, the magnitude of reduction in apo-affinity for CaM G114R toward the Ca_V2.1 IQ was less than that toward the Ca_V1.3 IQ (2.7-fold reduction for Ca_V2.1 IQ vs. 10-fold reduction for Ca_V1.3 IQ). This difference in binding affinity between the two channels supports a lack of apoCaM G114R preassociation with Ca_V1.3 and thus a lack of effect on CDI, while retaining an effect on Ca_V2.1 CDF. Similarly to the observation for the apo-affinities, all C-lobe mutations caused a reduction in the affinity of the Ca²⁺-saturated complex (2-fold to 14-fold compared with CaM WT) (Fig. 10 J). All C-lobe mutations reduced the Ca²⁺-sensitivity of the complex (1.8- to 4.3-fold compared with CaM WT), in line with their reduction of C-lobe Ca²⁺ affinity (Fig. 7 K) and consistent with their general impairment of Ca_V2.1 CDF. These results, however, do not explain the distinct effects of D96V observed in functional assays. It is possible that the relatively small Ca_V2.1 peptide used here and in our FRET assay is not sufficient to elucidate all CaM binding effects.

While our experiments utilize overexpression, physiologically each calmodulinopathy mutation would be expressed in a single of six CALM genes, resulting in a mix of WT and mutant CaM protein. To address the question of whether a mutant CaM would modulate the channel in the context of mixed expression with CaM WT, we co-expressed CaM D130G in a 1:1 ratio with CaM WT and evaluated the CDI and CDF of Ca_V1.3 and Ca_V2.1, respectively (Fig. 11). Indeed, this configuration led to an intermediate decrease in CDI of Ca_V1.3_L (Fig. 11 A) and Ca_V1.3_43S (Fig. 11 B), as compared with overexpressed D130G (Figs. 2, and 4). Likewise, mixed expression of CaM WT and D130G resulted in intermediate deficits in CDF of Ca_V2.1 (Fig. 11 C vs. 8). These results are consistent with prior data showing that the effect of mutant CaM on Ca_V1.2 scaled with the expression of mutant to WT protein (Limpitikul et al., 2014).

CaM is known to modulate numerous target proteins, making the pathogenesis of calmodulinopathies complex and multifaceted (Ghosh et al., 2006; Nyegaard et al., 2012; Fukuda et al., 2014; Nomikos et al., 2014; Yin et al., 2014; George, 2015; Boczek et al., 2016). Multiple studies have focused on the pathogenic mechanisms underlying the cardiac effects of calmodulinopathies; however, many patients show noncardiac effects including profound neurological symptoms. Thus, the pathogenesis of this disorder may include targets outside the heart. Prior work has demonstrated that many CaM variants can impair Ca_V1.2 channel regulation, leading to LQTS (Limpitikul et al., 2014; Limpitikul et al., 2017; Limpitikul et al., 2018). However, Ca_V1.2 channels, and indeed other Ca_V channel subtypes, are also widely expressed throughout the body. Importantly, the preassociation with apoCaM that was demonstrated to be important for Ca_V1.2 pathogenesis (Limpitikul et al., 2014) is shared by other VGCCs, making them particularly vulnerable to CaM mutations that alter Ca²⁺ binding. Thus, multiple VGCCs may be implicated in the pathogenesis of calmodulinopathies. Here, we show that many CaM mutations can impair the regulation of Ca_V1.2, Ca_V1.3, and Ca_V2.1.

We find that a number of previously identified CaM mutants impair the CDI of Ca_V1.3 across distinct splice backgrounds. In particular, CaM mutants decrease CDI in Ca_V1.3_L, which possess the modulatory ICDI/CTM domain and is expressed abundantly in the heart and brain. Likewise, these CaM variants impair CDI in Ca_V1.3_43s, a variant lacking ICDI/CTM and expressed at higher levels in the brain as compared with the heart (Bock et al., 2011). In the context of Ca_V2.1, these same CaM mutations modulate CDF, demonstrating a largely conserved mechanism of VGCC dysregulation by calmodulinopathy mutations. Given the critical role of these channels in the brain (Pietrobon, 2010; Catterall et al., 2013; Limpitikul et al., 2016; Pinggera et al., 2017; Pinggera et al., 2018; Nanou et al., 2018), such disruption could contribute to the neurological features of these patients.

The N54I mutation, located in the N-lobe of CaM, provided a useful control for nonspecific effects. Our binding (FRET and FA) data (Figs. 5, 7, 9, and 10) demonstrate that it is able to readily associate with Ca_V1.3 and Ca_V2.1. However, it fails to impair CDI or CDF (Figs. 2, 4, and 8). Interestingly, Ca_V1.3 channels exhibit both N- and C-lobe–mediated CDI; thus, the N-lobe locus of N54I does not exclude a strong effect on the channel. However, the preservation of CDI and CDF is consistent with prior data showing that N54I has a milder impact on CaM binding to Ca²⁺ than LQTS-associated mutations shown to act on Ca_V1.2 (Søndergaard et al., 2015a; Vassilakopoulou et al., 2015). Moreover, this lack of VGCC effects aligns with the CPVT phenotype in these patients (Crotti et al., 2023).

The CaM mutations D130G and D132G follow a pattern predicted by prior results in Ca_V1.2: both are located within or just adjacent to an EF-hand in the C-lobe of CaM and severely impair the Ca²⁺-dependent regulation of Ca_V1.3 and Ca_V2.1. F142L, on the other hand, resides outside the EF-hand coordinating residues but is critical to the packing and folding of the C-lobe of CaM (Søndergaard et al., 2017). Interestingly, this mutation exhibits a comparable impact on CDI in the context of Ca_V1.3 (Figs. 2, and 4) but exhibits a lesser effect on Ca_V2.1 CDF as compared with the EF-hand mutations (Fig. 8). Thus, the detailed mechanism underlying the loss of CaM regulation may be distinct for F142L. In addition, all three of these mutant CaMs show strong affinity for Ca_V1.3 and Ca_V2.1 channel peptides at low Ca²⁺ levels, indicating preservation of the preassociation feature. However, F142L again differs in that binding to Ca_V1.3 (Figs. 5, and 7) and Ca_V1.2 (Limpitikul et al., 2014; Wang et al., 2018) IQ region is enhanced at low Ca²⁺ levels (Crotti et al., 2013; Crotti et al., 2023).

The effects of G114R on Ca_V1.3 deviates from the other mutations. This C-lobe mutation has been shown to impair Ca²⁺ binding in a similar fashion to other pathogenic mutations, although with less severity as compared with D96V, D130G, D132G, or F142L (Brohus et al., 2021). CaM G114R has also been shown to impair Ca_V1.2 CDI, although to a lesser extent as compared with the LQTS-associated mutants described in this study (Brohus et al., 2021). While these facts might predict a similar impairment of Ca_V1.3, our data suggest a different result. In Ca_V1.3_L, CaM G114R appears to diminish CDI and increase VDI as compared with overexpressed CaM WT (Figs. 2, and 3). However, in Ca_V1.3_43s, the mutation has no effect on CDI or VDI (Fig. 4). In addition, our FRET results indicate that CaM G114R does not bind Ca_V1.3 effectively in the apo state, a result supported by FA data which specifically showed a prominent reduction of binding to the Ca_V1.3 IQ for G114R at low Ca²⁺ concentrations (Fig. 7). Thus, we postulate that CaM G114R does not preassociate with either Ca_V1.3 splice variant at low Ca²⁺ concentrations, and the results reflect the lower expression of endogenous CaM as compared with our overexpressed CaM WT control. Accordingly, the apparent CDI and VDI changes seen in Ca_V1.3_L can be explained by a mass action effect (Fig. 6 D). Importantly, this mechanism also explains the rapid CDI kinetics of the channel with CaM G114R (Fig. 2 D) and the VDI effect (Fig. 3). With both lobes of endogenous CaM available to initiate CDI, the kinetics reflect both the rapid C-lobe–mediated CDI combined with the slower N-lobe component (Yang et al., 2002). Moreover, the presence of ICDI/CTM has been shown to enhance channel VDI, explaining the increase in VDI observed in the presence of CaM G114R as the effect of ICDI/CTM binding (Singh et al., 2008; Bock et al., 2011). Similarly, the strong CDI seen in Ca_V1.3_43s, which lacks the ICDI/CTM domain, are consistent with the function of endogenous WT CaM (Singh et al., 2008; Sang et al., 2021) in the presence of unbound CaM G114R.

Interestingly, CaM G114R does impair Ca_V2.1 CDF under our experimental conditions of heterologous overexpression, although to a lesser extent as compared with other C-lobe mutants (Fig. 8 D). This result is consistent with the modest reduction in binding affinity with the channel IQ (Figs. 8 and 9), as well as the more moderate effect of this mutation on Ca²⁺ binding (Brohus et al., 2021). This is also reflected in a more moderate effect on Ca²⁺ sensing in the presence of the Ca_V2.1 IQ domain, as shown by our FA experiments (Fig. 10 K). Thus, CaM G114R may not preassociate to the same extent as CaM WT, and CaM G114R retains some Ca²⁺ sensitivity while bound to the channel, enabling some CDF (Fig. 8). Therefore, our data indicate that CaM G114R likely has minimal impacts on Ca_V1.3 channel regulation and a modest potential to impair Ca_V2.1 CDF, highlighting the differential impact a single CaM mutant can have on Ca_V2.1 vs. Ca_V1.3.

D96V has been shown to severely impair the ability of CaM to bind Ca²⁺ (Vassilakopoulou et al., 2015) and has a significant impact on CDI in Ca_V1.2 (Limpitikul et al., 2014) (Fig. 1) and Ca_V1.3 (Figs. 2 and 4). Although it binds well with both Ca_V1.3 and Ca_V2.1 channel fragments in our FRET assay, it fails to impair Ca_V2.1 CDF (Fig. 7). The unusual discrepancy between Ca_V1.3 and Ca_V2.1 channel regulation by CaM D96V is not made clearer by our FA results, where it shows a similar profile of Ca²⁺-dependent binding to a Ca_V2.1 channel peptide as the other mutants. It is therefore tempting to speculate that this mutation may have an effect on CaM structure that imparts a distinct effect in the context of Ca_V1.2/1.3 vs. Ca_V2.1. While such an effect remains to be elucidated, D96V has a documented effect on CaM structure, resulting in an unusual protein conformation that resembles neither apoCaM nor the fully calcified form of the protein (Wang et al., 2018). Interestingly, the structural changes of this mutation appear to be less severe as compared with those of the D130G mutation (Wang et al., 2020), indicating that perhaps Ca_V2.1 CDF is more amenable to certain structural changes as compared with Ca_V1.2 or Ca_V1.3 channels.

Finally, while deficits in Ca_V channel regulation are frequently characterized as “gain-of-function” (GoF) or “loss-of-function” (LoF) in terms of their impact on Ca²⁺ influx, the deficits imparted by the CaM mutations we have characterized are not so readily delineated into one of these categories. The same CaM mutation that impairs Ca_V1.3 CDI, therefore permitting excess Ca²⁺ influx in what could be termed a GoF, can also impair Ca_V2.1 CDF, resulting in diminished cytosolic Ca²⁺ and therefore a LoF. Due to these divergent impacts on two prevalent CaM targets, the effect of CaM mutations on overall Ca²⁺ signaling is likely to be multifaceted.

Overall, this study presents evidence that mutations in CaM can impact multiple VGCCs, expanding our knowledge of the potential pathogenic mechanisms underlying calmodulinopathies. The impact of CaM mutations on VGCC function incorporates many of the same principles identified in the disruption of Ca_V1.2 CDI leading to severe LQTS, where VGCCs likely represent vulnerable targets due to preassociation with CaM at low Ca²⁺ levels. Overall, disruption of VGCCs may represent an important component of the pathogenic mechanisms underlying calmodulinopathies both within and beyond the heart.

All data are available in the article itself and its supplementary materials.

Christopher J. Lingle served as editor.

We thank all members of Dr. Dick’s lab for helpful suggestions and discussion of the project. We also thank Josiah Owoyemi and Emma Gudmundsson for dedicated technical support.

This work was supported by grants from the National Institutes of Neurological Disorders and Stroke (R21NS127294 to I.E. Dick) and the National Heart, Lung, and Blood Institute (R01HL149926 to I.E. Dick).

Author contributions: John W. Hussey: conceptualization, data curation, formal analysis, investigation, resources, software, validation, visualization, and writing—original draft, review, and editing. Emily DeMarco: data curation and formal analysis. Deborah DiSilvestre: formal analysis, investigation, methodology, project administration, resources, supervision, and validation. Malene Brohus: data curation, formal analysis, investigation, methodology, resources, visualization, and writing—original draft, review, and editing. Ana-Octavia Busuioc: investigation. Emil D. Iversen: formal analysis and investigation. Helene H. Jensen: conceptualization, supervision, and writing—review and editing. Mette Nyegaard: data curation and writing—review and editing. Michael T. Overgaard: conceptualization, data curation, project administration, resources, supervision, validation, and writing—review and editing. Manu Ben-Johny: formal analysis and software. Ivy E. Dick: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, software, supervision, validation, visualization, and writing—original draft, review, and editing.