For generations, a large Swedish family presented with repeated episodes of syncope and cardiac arrest in response to exercise or emotional stress. Several family members were diagnosed with the inherited disorder catecholaminergic polymorphic ventricular tachycardia (CPVT), which is often fatal due to a high risk of ventricular fibrillation and sudden cardiac death (SCD). Indeed, two of the 13 affected individuals died from SCD, both at a young age. In 2012, we linked the disease to a mutation in the gene CALM1, which encodes the calcium (Ca²⁺) sensor calmodulin (CaM) (Nyegaard et al., 2012). The identification of a human CaM missense mutation came as a dramatic surprise to the CaM research field; CaM is exceptionally conserved across species with all vertebrate CALM genes encoding identical proteins, and human mutations had not previously been reported. The slow evolution of CaM emphasizes the strong selection pressure against even minor changes in the protein sequence (Halling et al., 2016). Further, CaM regulates more than 300 intracellular targets, each interaction with unique facets of binding sites, Ca²⁺-dependency, target affinity, and functionality (Shen et al., 2005; O’Connell et al., 2010). With this versatility in mind, it was believed that mutations in CaM could not be tolerated.

After our initial finding, a number of CaM mutations have been identified in patients with severe cardiac arrhythmia disorders involving recurrent syncope, ventricular fibrillation, and in some instances SCD under adrenergic stimulation (Table 1). The vast majority of these mutations are de novo and carriers present with disease phenotypes early or very early in childhood, in some cases even before birth. In addition to CPVT, carriers suffer from long QT syndrome (LQTS), and one individual was diagnosed with idiopathic ventricular fibrillation (IVF). The link between CaM mutations and these arrhythmias has primarily been attributed to impaired regulation of the cardiac ryanodine receptor isoform 2 (RyR2), and the cardiac L-type voltage gated Ca²⁺ channel isoform 1.2 (Ca_V1.2) (Table 1).

Since CaM is encoded by three active genes and expressed in all cells, the CaM field is faced with intriguing questions and paradoxes at the genetic and phenotypic level. First, how can a single mutation in one of six CaM-encoding alleles dominantly cause SCD? Second, how can identical missense mutations cause LQTS in one patient and CPVT in another? Third, with an increasing number of new rare CaM missense mutations identified in sequencing databases of large human cohorts, could there be other phenotypes associated with CaM mutations? Improved understanding of the functional impact of CaM mutations may enable predictions of the genotype–phenotype relationship for variants in any of the three CALM genes.

In this review, we summarize and discuss the current knowledge on CaM mutations and their impact on the regulation of Ca_V1.2 and RyR2, and address the few studies that suggest an involvement of other targets. Finally, we discuss the special genetic context of CaM and the implications for future studies.

Fast and compound changes in cytosolic Ca²⁺ concentration is the foundation for a wide number of cellular responses, including muscle contraction and neuronal firing (Clapham, 2007). Thus, at rest, the cytosolic Ca²⁺ concentration is maintained at ∼100 nM, but can rapidly increase to more than 100 μM, when Ca²⁺ channels open in the plasma membrane or in internal stores such as the sarcoplasmic reticulum (SR). Detection of this steep change in Ca²⁺ concentration depends on Ca²⁺ binding proteins. CaM is one of the major Ca²⁺ sensors that relay information on Ca²⁺ concentration to functionally modulate target proteins (known as calmodulation). CaM is synthesized as a 149 amino acid protein, however, the initiator Met residue is removed upon translation, leaving 148 amino acid residues in the mature protein (Sasagawa et al., 1982). This has led to some confusion in the numbering of CaM variant positions. The Human Genome Variant Sequence (HGVS) nomenclature (den Dunnen et al., 2016) recommends to count the initiator Met as residue number 1, while the CaM protein community tends to leave the residue out as it has no functional role in the mature protein. Throughout this paper we will use the HGVS nomenclature.

The structure of CaM reflects its refined Ca²⁺ sensing abilities (Kretsinger et al., 1986; Chattopadhyaya et al., 1992). CaM is a 16.7 kDa protein consisting of two lobes connected by a flexible and unstructured or α-helical linker (Figures 1A,B). Each lobe has two EF-hands, which can each coordinate one Ca²⁺ ion (Figures 1A,B, gray spheres). The C-terminal lobe of CaM binds Ca²⁺ with six times higher affinity (K_D 2.5 μM) than the N-terminal lobe (K_D 16 μM), allowing CaM to sense Ca²⁺ across a wide concentration range (Linse et al., 1991; Søndergaard et al., 2015a). Hydrophobic patches on the inside of each lobe recognize binding motifs on interaction partners, and thereby facilitate CaM binding and target regulation (Tidow and Nissen, 2013). Ca²⁺ binding to CaM and CaM binding to target proteins allosterically affect the affinity of each other, and targets specifically modulate the conformation of CaM. In this way, the small CaM protein displays a plethora of binding and regulation properties.

Despite the remarkable conservation of CaM, 26 cases of arrhythmogenic mutations have now been identified in humans. Their positions in CaM are indicated on a CaM structure in Figure 1A and on the CaM sequence in Figure 1C. Strikingly, the mutations are primarily found in the C-terminal lobe and most affect residues involved in Ca²⁺ coordination, dramatically reducing Ca²⁺ affinity (Table 1). One interesting exemption is the mutation initially identified in the large Swedish family, CaM-N54I. This mutation is unique since it (1) resides in the N-terminal lobe and (2) neither coordinates Ca²⁺ nor is part of the hydrophobic target binding patches. Biochemical and cellular experiments have been employed to model and explain how CaM mutations lead to arrhythmic phenotypes. The results from these studies are discussed in the following.

In humans, CaM is encoded by three different and independent loci; on chromosome 2 (CALM2), 14 (CALM1), and 19 (CALM3). Although there are differences in the genomic sequence, the three different transcripts are translated into the exact same protein (Fischer et al., 1988). During the last six years (2012–2018), 26 cases of pathogenic mutations in CaM have been reported, and all three CALM genes are now established major genes for both CPVT and LQTS (Table 1). All identified pathogenic CaM mutations cluster in the C-lobe, except the CPVT-causing variant N54I. Interestingly, this is the mutation with the mildest effect on biophysical parameters of CaM, including Ca²⁺ binding affinity. It is also the only variant found in a large family. Further, all LQTS-causing CaM mutations strongly reduce the C-lobe Ca²⁺ affinity, and, except for the F142L mutation, are all located in Ca²⁺-coordinating residues. This suggests that CaM mutations that strongly affect C-lobe Ca²⁺ affinity lead to LQTS (Table 1).

To date, there are more published cases of pathogenic mutations in CALM1 (11) and CALM2 (11) compared to CALM3 (4). This could either reflect that the first published mutations were found in CALM1 and CALM2 and thus these two genes were included in genetic screening panels before CALM3. Or, it may be due to some subtle functional differences between the three genes. Quantitative PCR have shown that the three genes are not expressed at equal levels in cardiomyocytes, but the relative levels are not clear: whereas one study found higher levels of CALM3 transcripts in human hearts (Crotti et al., 2013), the CALM1 transcript was the most abundant in human stem cell-derived cardiomyocytes (Rocchetti et al., 2017).

As the number of published pathogenic CaM mutations has increased, several conclusions about genotype–phenotype relationships begin to form. Of particular interest is the CaM-D130G mutation, which has been identified in four unrelated individuals; two carrying the mutation in CALM1, one in CALM2, and one in CALM3, and all four suffering from LQTS. Similarly, the CaM-F142L mutation was found in both CALM1 and CALM3, and all carriers suffered from LQTS. These observations imply that the amino acid position and type of change is important for the phenotype, and not the genetic origin of the transcript (CALM1, −2, or −3). One intriguing observation does, however, challenge this simple genotype–phenotype conclusion. The CaM-N98S mutation was found in CALM1 in one individual, and in three other individuals in CALM2. Interestingly, these four patients present with different phenotypes – either CPVT or LQTS or both – suggesting that we still do not fully understand the underlying mechanisms determining the disease phenotype. These may involve other genomic variants able to shape the phenotype, complex protein regulatory effects, or environmental factors.

Since the protein products from all CALM genes are identical, it is tempting to speculate if a deletion of one allele (equivalent to a loss-of-function mutation) is less pathogenic than missense mutations. This idea immediately poses a therapeutic solution to patients carrying a CaM missense mutation, for example using the CRISPR/Cas9 technology to delete the pathogenic allele. Two studies specifically silenced the mutated CaM allele in patient-derived pluripotent stem cells differentiated into cardiomyocytes. Here, the CaM-D130G and -N98S mutations were silenced in CALM2 with a partial or almost full restoration of Ca_V1.2 regulation (Limpitikul et al., 2017; Yamamoto et al., 2017). These experiments are proof-of-principle that removal of the diseased allele may be a therapeutic solution. Also, these studies suggest that potential frameshift mutations causing premature stop codons or protein degradation may not be as detrimental as missense mutations. It still needs to be determined, however, if individuals carrying a loss-of-function mutation in a CALM gene are in fact unaffected from disease. Interestingly, a large exome sequencing study [Exome Aggregate Consortium (ExAC)] (Lek et al., 2016), found that CALM1 and CALM2 are intolerant to loss-of-function mutations (pLI = 0.89 and 0.86 respectively).

Given the ubiquitous role of CaM, it is striking that all mutations identified are associated with a strong cardiac phenotype. We speculate whether these cases were discovered because of their unusual severity and because cardiologists and geneticists have specifically screened for CALM mutations in populations with cardiac disorders. Looking for CaM mutations in other patient groups may reveal new aspects and consequences of these mutations.

In a database containing variants from a large sequencing effort of almost 140,000 individuals (GnomAD, Lek et al., 2016), additional rare CaM missense mutations are reported (Figures 1B,C, gray residues and circles). The number of coding variants for all three genes is much lower than expected by chance. However, the cumulative frequency of additional rare CaM mutations suggests that CaM variants do not exclusively cause severe cardiac arrhythmias. At present, there is no overlap between variants identified in GnomAD (database variants) and the published pathogenic mutations. Further, the GnomAD missense variants are distributed throughout the entire protein, and more evenly distributed on the three CALM genes (9, 9, and 12 mutations in CALM1, −2, and −3, respectively), compared to the arrhythmogenic CaM variants. Also, all GnomAD variants except two, fall outside Ca²⁺-coordinating residues. Taken together, we therefore speculate that some of these uncharacterized variants are associated with unknown traits not involving cardiac arrhythmia. Sequencing results from large cohorts with known phenotypes are required to confirm this hypothesis.

We propose that studies of tissues other than cardiac are warranted for future research on the effects of CaM mutations. In particular, CaM expression is high in excitable neuronal cells. Also Ca_V1.2 is widely expressed in neuronal tissues. Here, Ca_V1.2 plays a role in cellular firing as well as in gene regulation, and mutations in Ca_V1.2 have been attributed to psychiatric diseases (Nyegaard et al., 2010; Nanou and Catterall, 2018). Neurons express a number of other Ca_V channel isoforms, including Ca_V1.3 and Ca_V2 variants, which are also regulated by CaM. RyR2 plays a less prominent role in neurons, where the inositol triphosphate receptor (IP₃R), which is also regulated by CaM, is the dominating intracellular Ca²⁺ release channel. Mild neuronal defects have been observed in some patients with CaM mutations, but these effects were suggested to be secondary, resulting from the frequent and severe episodes of syncope or cardiac arrest (Crotti et al., 2013; Boczek et al., 2016; Pipilas et al., 2016).

Within the last six years, CaM mutations have emerged as a novel cause of human diseases, the calmodulinopathies. All described pathogenic mutations have been identified in patients suffering from severe arrhythmic disorders, and biochemical as well as cellular studies have demonstrated that particularly the regulation of the Ca²⁺ channels Ca_V1.2 and RyR2 are affected by these mutations. Currently, there is a strong correlation between LQTS-causing CaM mutations and Cav1.2 dysregulation, whereas all mutations affect RyR2 function. Given the ubiquitous role of CaM in a vast number of cellular processes, we predict that yet other targets may be affected. Our database search revealed a number of uncharacterized CaM missense mutations with unknown phenotypic consequences present in the population. Future studies will reveal whether other protein targets as well as other disease phenotypes can be assigned to mutations in CaM.

HJ wrote the first draft of the manuscript. HJ, MB, MN, and MO contributed to the content and writing. HJ and MB prepared the figure and table. All authors have read and approved the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.