CACNA1H variants are not a cause of monogenic epilepsy.

Abstract

CACNA1H genetic variants were originally reported in a childhood absence epilepsy cohort. Subsequently, genetic testing for CACNA1H became available and is currently offered by commercial laboratories. However, the current status of CACNA1H as a monogenic cause of epilepsy is controversial, highlighted by ClinGen’s recent re-classification of CACNA1H as disputed. We analyzed published CACNA1H variants and those reported in ClinVar and found none would be classified as pathogenic or likely pathogenic per the ACMG classification criteria. Moreover, Cacna1h did not modify survival in a Dravet Syndrome mouse model. We observed a mild increase in susceptibility to hyperthermia-induced seizures in mice with reduced Cacna1h expression. Overall, we conclude that there is limited evidence that CACNA1H is a monogenic cause of epilepsy in humans and that this gene should be removed from commercial genetic testing panels to reduce the burden of variants of uncertain significance for healthcare providers, families and patients with epilepsy.

CACNA1H encodes Cav3.2, a member of the T-type calcium channel family. CACNA1H variants were originally reported in a cohort of individuals with childhood absence epilepsy (CAE) (Y. Chen et al., 2003). Since the initial report, a number of studies have reported on CACNA1H variants in the context of epilepsy, some support this original observation, while others fail to find sufficient evidence for an association with seizures (Y. Chen et al., 2003; Chioza et al., 2006; Heron et al., 2004). However, many of these early supportive studies suffered from small cohort sizes and lack of statistical rigor (Y. Chen et al., 2003). In addition, exome sequencing studies in large epilepsy cohorts including patients with developmental and epileptic encephalopathy (DEE), genetic generalized epilepsy (GGE) and non-acquired focal epilepsy (NAFE) have failed to identify an enrichment of de novo or rare CACNA1H variants in patients relative to controls (Epi25 Collaborative. Electronic address & Epi, 2019; Epi & Epilepsy Phenome/Genome, 2017; Epi et al., 2013; Heyne et al., 2018).

Reported pathogenic CACNA1H variants are primarily missense, though one frameshift variant has been identified (Table 1). Functional studies of these CACNA1H missense variants suggested they may lead to altered biophysical properties or protein trafficking (Y. Chen et al., 2003; Heron et al., 2007; Heron et al., 2004; Khosravani et al., 2005; Vitko et al., 2005). However, because these studies were performed in non-excitable heterologous cell systems (i.e. HEK 293), it remains unclear whether the altered channel properties are sufficient to affect firing of individual neurons or result in hyperexcitability at the network level. Further, without assaying a number of missense variants present in the general population, it is difficult to ascertain what level of altered channel function is tolerated. While studies in model organisms can help elucidate a gene’s role in seizure susceptibility, there are few such studies of Cacna1h in rodent or other models. Powell et al. demonstrate that a Cacna1h variant (gcm; R1584P) accounts for ~33% of the variance for seizure frequency and time spent seizing in the spontaneously epileptic rat model of absence seizures (GAERS), in support of Cav3.2 contributing to seizure susceptibility as a modifier (Powell et al., 2009). In contrast, Cacna1h knockout mice exhibit constitutively constricted coronary arterioles and focal myocardial fibrosis, but not seizures (C. C. Chen et al., 2003).

Table 1:

Reanalysis of CACNA1H missense variants from the literature and present in Clinvar.

					gnomAD
Chr-Pos-Ref-Alt	cDNA	Protein	Cons	CADD	AF_Popmax	Num_Het_Indiv	Num_Homoz_Indiv	ACMG	Study
16–1250335-C-T	c.883C>T	p.Arg295Ter	truncation	N/A	NF	0	0	VUS	This study
16–1245503-C-G	c.483C>G	p.Phe161Leu	missense	18.24	0.00001961	2	0	Likely benign	(Y. Chen et al., 2003)
16–1250296-G-A	c.844G>A	p.Glu282Lys	missense	35	0.0002268	4	0	VUS
16–1251816-T-A	c.1366T>A	p.Cys456Ser	missense	24.9	NF	0	0	VUS
16–1251945-G-A	c.1495G>A	p.Gly499Ser	missense	13.14	0.0105	255	3	Likely benign
16–1252393-C-T	c.1943C>T	p.Pro648Leu	missense	15.87	0.00001422	1	0	Likely benign
16–1254238-G-A	c.2231G>A	p.Arg744Gln	missense	0.062	0.00006633	6	0	Likely benign
16–1254250-C-T	c.2243C>T	p.Ala748Val	missense	3.376	0.0006849	16	0	Likely benign
16–1254325-G-A	c.2318G>A	p.Gly773Asp	missense	2.789	0.00297	59	0	Likely benign
16–1254357-G-A	c.2350G>A	p.Gly784Ser	missense	1.238	0.0002643	8	0	Likely benign
16–1255153-G-A	c.2491G>A	p.Val831Met	missense	21.2	0.0001537	4	0	VUS
16–1255204-G-A	c.2542G>A	p.Gly848Ser	missense	17.91	0.0002485	20	0	Likely benign
16–1261517-G-A	c.4387G>A	p.Asp1463Asn	missense	15.74	0.001373	46	0	Likely benign
16–1251888-G-A	c.1438G>A	p.Ala480Thr	missense	9.555	0.00001704	1	0	Likely benign	(Heron et al., 2004)*
16–1252303-C-T	c.1853C>T	p.Pro618Leu	missense	18.95	0.0009646	123	0	Likely benign
16–1252306-CAG-C	N/A	p.Val621GlyfsTer34	frameshift	N/A	0.0002159	13	0	VUS
16–1254271-G-A	c.2264G>A	p.Gly755Asp	missense	7.817	0.0004536	66	0	Likely benign
16–1245509-G-C	c.489G>C	p.Gln163His	missense	17.02	0.01677	1258	14	Likely benign	(Chourasia et al., 2019)
16–1252062-A-T	c.1612A>T	p.Arg538Trp**	missense	13.5	NF	0	0	VUS
16–1255229-C-T	c.2567C>T	p.Pro856Leu	missense	14.07	0.0001025	12	0	Likely benign
16–1268643-C-T	c.5879C>T	p. Thr1960Ile***	missense	9.512	NF	0	0	VUS
16–1270254-G-A	c.6322G>A	p.Ala2108Thr	missense	15.33	0.002292	340	2	VUS
16–1250392-C-T	c.940C>T	p.Pro314Ser	missense	4.308	0.0005333	10	0	Likely benign	(Liang et al., 2006)
16–1251924-C-T	c.1474C>T	p.Pro492Ser	missense	12.85	0.0005567	6	0	Likely benign
16–1248668-G-C	c.697G>C	p.Val233Leu	missense	14.99	NF	0	0	VUS	Clinvar****
16–1252243-C-T	c.1793C>T	p.Ala598Val	missense	9.282	NF	0	0	VUS
16–1254058-C-A	c.2051C>A	p.Pro684His	missense	12.5	NF	0	0	VUS
16–1258106-C-T	c.3248C>T	p.Thr1083Met	missense	21.2	NF	0	0	VUS
16–1262024-A-G	c.4645A>G	p.Met1549Val	missense	21	NF	0	0	VUS
16–1270830-A-G	c.6898A>G	p.Ile2300Val	missense	0.064	0.0002785	16	0	VUS
16–1203912-G-T	c.175G>T	p.Gly59Cys	missense	14.23	0.001147	5	0	Likely benign	(Lee, Lee, & Lee, 2018)
16–1254370-G-A	c.2363G>A	p.Arg788His	missense	14.67	0.0003706	11	0	VUS
16–1270518-C-A	c.6586C>A	p.Pro2196Thr	missense	14.83	0.0001351	3	0	Likely benign
16–1270936-A-G	c.7004A>G	p.Lys2335Arg	missense	10.16	NF	0	0	VUS
16–1259140-G-A	c.3472G>A	p.Gly1158Ser	missense	0.199	0.0002341	13	0	Likely benign	(Becker et al., 2017)
16–1254369-C-T	c.2362C>T	p.Arg788Cys	missense	7.372	0.1249	20068	1083	Benign	(Heron et al., 2007)
16–1256126-G-A	c.2626G>A	p.Ala876Thr	missense	13.09	0.002548	263	0	VUS
16–1268439-G-A	c.5675G>A	p.Arg1892His	missense	15.63	0.001531	55	1	VUS

Abbreviations: cDNA = coding DNA, Cons = consequence, AF_Popmax = allele frequency in the population with highest allele frequency, Num_Het_Indiv = number of heterozygous individuals with this variant present in gnomAD, Num_Homoz_Indiv = number of homozygous individuals with this variant present in gnomAD, ACMG = ACMG reclassification.

^*Manuscript notes none of these four variants segregate with phenotype.

^**Manuscript lists as c.1612C>T / p.Leu538Phe, but amino acid 538 is actually Arg.

^***manuscript lists as c.5879C>T / p. Ala1960Val, but amino acid 1960 is actually Thr.

^****Clinvar pathogenic or likely pathogenic variants.

At The Northwestern Adult Epilepsy Genetics clinic, we encountered an individual with an inherited p.Arg295Ter CACNA1H truncating variant. This patient has myoclonic epilepsy and difficult to control tonic-clonic seizures as well as mild intellectual disability. The CACNA1H variant was identified through clinical genetic testing on a commercially available epilepsy gene panel and was reported as a variant of uncertain significance (VUS). As part of our internal process for variant interpretation, we performed a more thorough evaluation of the evidence that CACNA1H is a monogenic cause of epilepsy. Overall, we identified a total of 37 CACNA1H variants reported either in the literature or ClinVar in individuals with epilepsy (Table 1). These 37 patients were reported to have a range of epilepsies, including CAE, febrile seizures (FS), myoclonic-astatic epilepsy (MAE), temporal lobe epilepsy (TLE), symptomatic generalized epilepsy (SGE), juvenile myoclonic epilepsy (JME), epilepsy with generalized tonic-clonic seizures (EGTCS), GGE and juvenile absence epilepsy (JAE). We used the American College of Medical Genetics (ACMG) criteria for variant classification of these 37 reported variants (Richards et al., 2015). Overall, 17/37 (45.9%) were reclassified as VUSs and 20/37 (54.1%) as likely benign or benign, none were classified as likely pathogenic. In particular many of these variants were present in a sampling of >140,000 individuals in the general population (gnomAD - see URLs) (Lek et al., 2016). Overall, 28/37 (75%) of these published ‘epilepsy-associated’ CACNA1H variants are present in gnomAD, with slightly over 50% (18/37) being present in 10 or more individuals. For instance, CACNA1H p.Ala876Thr is present in over 250 individuals in gnomAD.

We also used Combined Annotation Dependent Depletion (CADD) scores to compare published and ClinVar missense CACNA1H variants to missense CACNA1H variants present in gnomAD. CADD is an in silico measure of the likelihood that a genetic variant may be deleterious, higher scores are more likely deleterious and CADD ≥ 20 represents the top 1% of predicted deleterious variants (Kircher et al., 2014; Rentzsch, Witten, Cooper, Shendure, & Kircher, 2019). CADD scores for published CACNA1H variants (mean=12.8; n=35) showed the same distribution of CADD scores as the variants present in the general population (gnomAD) (mean=12.9; n=1938) (Figure 1A). Moreover, the ‘epilepsy associated’ and population variants were both located throughout the protein and showed no clustering in the pore loops, transmembrane domains, or gating brake (Figure 1D–E). For contrast, we performed the same CADD score analysis for two bona-fide calcium channel genes implicated in epilepsy, CACNA1A and CACNA1E. For both of these genes pathogenic or likely pathogenic missense variants have significantly higher CADD scores compared to missense variants present in gnomAD (Figure 1B–C). Finally, CACNA1H is not under functional constraint (Z_mis=−2.36; pLI=0), suggesting that missense and predicted loss-of-function alleles are tolerated. Conversely, as above with the CADD score analysis, CACNA1A (Z_mis=5.78; pLI=1) and CACNA1E (Z_mis=5.81; pLI=1) are among the most constrained genes in the human genome (Z_mis>3.09; pLI≥0.9).

Figure 1. Analysis of published or Clinvar ‘pathogenic’ missense calcium channel variants and effect of reduced Cacna1h expression in a Scn1a mouse model of Dravet syndrome.

(A) Distribution of CADD scores in CACNA1H. The CACNA1H_path distribution (n=35) is from published missense variants or Clinvar pathogenic or likely pathogenic missense variants as listed in Table 1. The CACNA1H_gnomAD distribution (n=1938) includes all CACNA1H missense variants in gnomAD. Error bars represent standard deviation. p = 0.9758 (Mann-Whitney ranksum). (B) Distribution of CADD scores in CACNA1A. The CACNA1A_path distribution (n=46) includes all Clinvar pathogenic or likely pathogenic missense variants. The CACNA1A_gnomAD distribution (n=910) includes all CACNA1A missense variants in gnomAD. Error bars represent standard deviation. p < 2.2e-16 (Mann-Whitney ranksum). (C) Distribution of CADD scores in CACNA1E. The CACNA1E_path distribution (n=17) is from published missense variants (K. L. Helbig et al., 2018) or Clinvar pathogenic or likely pathogenic missense variants. The CACNA1E_gnomAD distribution (n=821) is from CACNA1E missense variants in gnomAD. Error bars represent standard deviation. p = 0.003252 (Mann-Whitney ranksum). (D) Distribution of published or Clinvar ‘pathogenic’ missense variants in CACNA1H (Protter; see URLs). (E) Distribution of gnomAD missense variants in CACNA1H (Protter; see URLs). (F) Reduced Cacna1h expression does not affect survival in a mouse model of Dravet syndrome. p = 0.978 (Mantel-Cox logrank). (G) Reduced Cacna1h expression increases hyperthermia-induced seizure susceptibility. GTCS = generalized tonic-clonic seizure. p = 0.0332 (Mantel-Cox logrank).

A valid argument could be made that these other monogenic causes, CACNA1A and CACNA1E, are generally associated with much more severe epilepsy disorders, including DEE; while the CACNA1H associated variants are reported in individuals with milder epilepsies. In this instance genetic association is more appropriate, and indeed CACNA1H variants may convey some risk to these milder epilepsies. However, as outlined above, this association is disputed in the literature and original reports were not sufficiently statistically rigorous. Moreover, a recent GWAS in 15,000 individuals with various epilepsy syndromes, including CAE, GGE and JME failed to identify any risk loci in or near CACAN1H (International League Against Epilepsy Consortium on Complex, 2018). Moreover, exome sequencing in the common epilepsies, including GGE, have failed to identify CACNA1H as being enriched for rare variants using a gene burden analysis (Epi25 Collaborative. Electronic address & Epi, 2019; Epi & Epilepsy Phenome/Genome, 2017). Conversely, CACNA1G is the most highly enriched gene for deleterious variants in individuals with GGE and is the only calcium channel gene found to be putatively associated with epilepsy in this large exome sequencing study.

We previously published Cacna1g as a modifier of both Scn1a and Scn2a mouse models of epilepsy (Calhoun, Hawkins, Zachwieja, & Kearney, 2016, 2017). Given that Cacna1g and Cacna1h are both members of the T-type calcium channel gene family and CACNA1G is enriched for rare variants in GGE, we sought to determine whether altered Cacna1h expression would also modulate phenotype severity in a mouse model of Dravet syndrome. Cacna1h^KO/+ (1H^KO/+) mice were crossed with Scn1a^tm1Kea (1A^KO/+) mice to generate double transgenic (1A^KO/+;1H^KO/+) mice and single mutant littermate controls. We found that mice with reduced Cacna1h expression (1A^KO/+;1H^KO/+) exhibited similar survival over 8 weeks relative to their control littermates (1A^KO/+) (Fig. 2A). 1A^KO/+;1H^KO/+ mice were slightly more susceptible to hyperthermia-induced seizures relative to control 1A^KO/+ littermates (Fig. 2B). However, the magnitude of this effect is much smaller than the effect of antiepileptic drugs (AEDs) such as clobazam, levetiracetam, phenobarbital, or lamotrigine suggesting that, while statistically significant, the result is unlikely to be biologically relevant (Hawkins et al., 2017). Based on a lack of effect on survival and a small but statistically significant increase in hyperthermia-induced seizure susceptibility, we conclude that Cacna1h is not a robust modifier of Scn1a-associated seizures.

In summary, there is currently no human genetic evidence to support the association of CACNA1H with epilepsy, nor does Cacna1h meaningfully modify seizure susceptibility in an Scn1a mouse model. Similar to our findings here, the ClinGen working group has undertaken an effort to curate genes appearing on panels (I. Helbig et al., 2018) and has classified CACNA1H as a disputed gene based on poor genetic support, moderate experimental support, and lack of replication over time. Despite minimal evidence for CACNA1H as a monogenic cause of epilepsy, it currently appears on at least 20 independent commercial clinical genetic tests. While CACNA1H-associated variants could potentially be a risk factor for milder epilepsies, the inclusion of ‘risk factor’ genes is completely inappropriate for current clinical genetic testing, as identification of variants has no diagnostic value. New CACNA1H variant epilepsy case reports continue to appear in the literature, which may lead to confusion among clinicians and genetic counselors when interpreting VUS in CACNA1H, as was the case initially in our Adult Epilepsy Genetic Clinic (Chourasia, Osso-Rivera, Ghosh, Von Allmen, & Koenig, 2019). There a large number of VUS being reported in this gene, data from one commercial company alone has reported 407 CACNA1H VUSs in 4859 individuals (almost 10% of reports) in recent years from their epilepsy panel (Truty et al., 2019). These VUSs place a large burden on clinicians and genetic counselors to properly counsel patients regarding variants that are unlikely to be a primary contributor to their phenotype. Our findings here are not limited to CACNA1H, and the ClinGen working group are reevaluating genes to ensure they are removed from clinical testing panels.