Blog

By Shehrazade Dahimene, Leonie von Elsner, Tess Holling, Lauren S. Mattas, Jess Pickard, Davor Lessel, Kjara S. Pilch, Ivan Kadurin, Wendy S. Pratt, Igor B. Zhulin, Hongzheng Dai, Maja Hempel, Maura R. Z. Ruzhnikov, Kerstin Kutsche, Annette C. Dolphin

Excerpt from the article published in Brain, awac081, https://doi.org/10.1093/brain/awac081

Editor’s Highlights

• Bi-allelic variants in CACNA2D1, encoding alpha2delta-1 protein, have been found in two unrelated individuals aged 4 years 11 months and 4 years, showing developmental and epileptic encephalopathy (DEE).
• Bi-allelic loss-of-function variants in CACNA2D1 underlie the severe neurodevelopmental disorder in these two patients.
• Results demonstrate the critical importance and non-interchangeability of alpha2delta-1 and other alpha2delta proteins for normal human neuronal development.
• The loss-of-function frameshift nature of the CACNA2D1 demonstrating a loss-of-function effect for the amino-acid substitution p.(Gly209Asp) confirm disease causation of homozygous and compound heterozygous CACNA2D1 variants, while also calling into question a causative role of monoallelic CACNA2D1 variants in intellectual disability, epilepsy and/or inherited arrhythmogenic diseases.

Abstract

Voltage-gated calcium (Ca_V) channels form three sub-families (Ca_V1-3). The Ca_V1 and Ca_V2 channels are heteromeric, consisting of an α₁ pore-forming subunit, associated with auxiliary Ca_Vβ and α₂δ subunits. The α₂δ subunits are encoded in mammals by four genes, CACNA2D1–4. They play important roles in trafficking and function of the Ca_V channel complexes. Here we report biallelic variants in CACNA2D1, encoding the α₂δ-1 protein, in two unrelated individuals showing a developmental and epileptic encephalopathy (DEE). Patient 1 has a homozygous frameshift variant c.818_821dup/p.(Ser275Asnfs*13) resulting in nonsense-mediated mRNA decay of the CACNA2D1 transcripts, and absence of α₂δ-1 protein detected in patient-derived fibroblasts. Patient 2 is compound heterozygous for an early frameshift variant c.13_23dup/p.(Leu9Alafs*5), highly likely representing a null allele, and a missense variant c.626G>A/p.(Gly209Asp). Our functional studies show that this amino-acid change severely impairs the function of α₂δ-1 as a calcium channel subunit, with strongly reduced trafficking of α₂δ-1^G209D to the cell surface, and a complete inability of α₂δ-1^G209D to increase the trafficking and function of Ca_V2 channels. Thus biallelic loss-of-function variants in CACNA2D1 underlie the severe neurodevelopmental disorder in these two patients. Our results demonstrate the critical importance and non-interchangeability of α₂δ-1 and other α₂δ proteins for normal human neuronal development.

Results

Biallelic variants in CACNA2D1 in two unrelated individuals with DEE

Through GeneMatcher (14), we identified the two unrelated male patients, 1 and 2, with a highly consistent phenotype corresponding to DEE. The two affected individuals carried biallelic pathogenic variants in CACNA2D1 (Tables 1 and 2). Patient 1 developed generalized seizures at age 19 months, while onset of epilepsy was at age 11.5 months in patient 2. At last examination, patient 1 was 4 years 11 months and patient 2 was 4 years old. Both were microcephalic, had severe hypotonia, absent speech, spasticity, choreiform movements, orofacial dyskinesia, and cortical visual impairment (Tables 1 and 2, and Case Reports in Supplementary Information). Brain imaging revealed corpus callosum hypoplasia and progressive volume loss in both (Fig. 1A and B). Patient 1 had no cardiac anomalies, while a tiny patent foramen ovale was found in patient 2 (Tables 1 and 2).

Trio ES in patient 1 and parents revealed in the proband the homozygous CACNA2D1 (NM_000722.3) frameshift variant c.818_821dupGAAC / p.(Ser275Asnfs*13). The variant is absent from public databases including gnomAD (v.2.1.1 and 3.1.1). The 4-bp duplication was validated in patient 1’s DNA and fibroblast-derived cDNA. His healthy parents were heterozygous carriers (Tables 1 and 2, and Supplementary Fig. 1). In patient 2, trio ES demonstrated compound heterozygosity for the CACNA2D1 variants c.13_23dupTGCCTGCTGGC [p.(Leu9Alafs*5)] and c.626G>A [p.(Gly209Asp)] (Tables 1 and 2, and Supplementary Fig. 1). His healthy mother was heterozygous for the c.13_23dupTGCCTGCTGGC variant and his healthy father for the c.626G>A variant. The 11-b duplication is likely a loss-of-function variant; it has a worldwide minor allele frequency of 0.003% (gnomAD v.2.1.1), while the c.626G>A variant is absent in gnomAD (v.2.1.1 and 3.1.1).
The missense variant c.626G>A/p.(Gly209Asp) in exon-7 is predicted to be damaging by in silico tools detailed in Supplementary Methods.
In fibroblasts of patient 1, CACNA2D1 mRNA level was reduced to 6-9% compared with control fibroblasts, while it was similar in patient 2 and control fibroblasts (Fig. 1C and Supplementary Fig. 2). We next determined α2δ-1 levels in whole-cell lysates from cultured primary fibroblasts of patients 1 and 2. Qualitatively, we detected little full-length α2δ-1 in patient 1 fibroblasts, while it was present in patient 2 and control cells (Fig. 1D and Supplementary Fig. 3).
Quantification of α2δ-1 indicated 10-12% in patient 1 and 31-38% in patient 2 compared to controls (Fig. 1E). Together, the data suggest patient 1 carries biallelic CACNA2D1 loss-of-function alleles and patient 2 harbors at least one CACNA2D1 loss-of-function variant.
Next, we investigated mRNA levels of the other CACNA2D genes in patient fibroblasts to identify possible compensatory effects. While mRNA levels of CACNA2D2 and CACNA2D4 were too low to be quantified, we detected 3- to 7-fold higher CACNA2D3 mRNA levels in patient 2 fibroblasts compared to patient 1 and control cells (Supplementary Fig. 4).

Glycine 209 is invariant in CACNA2D1

The α2δ-1 protein (also denoted as CACNA2D1) contains a von Willebrand factor-A domain and four Ca2+ channel and chemotaxis receptor (Cache) domains (15), organised into two double-Cache (dCache) domains (16). The p.(Gly209Asp) (G209D) amino-acid substitution in the CACNA2D1 gene product of patient 2 is within the gabapentin and amino-acid binding pocket of its dCache_1 domain (16). This Gly residue is important for maintaining a 3-strand beta-sheet stability and simultaneously providing a critical turn in the structure. G209 is absolutely invariant in both CACNA2D1 and CACNA2D2 orthologs in all vertebrates and paralogs and predecessors from low invertebrates (Supplementary Fig. 5).

The p.(Gly209Asp) variant disrupts plasma membrane ⍺2δ-1 expression

We then investigated the in vitro effect of the p.(Gly209Asp) variant on α2δ-1 as a calcium channel subunit. Firstly, we compared cell surface expression of HA-⍺2δ-1 wildtype (WT) and HA-⍺2δ-1 (G20n9D) in non-permeabilised cells (5). As shown by the HA signal, the expression of HA-⍺2δ-1 (G209D) at the cell surface was reduced by ~80% compared to HA-⍺2δ-1 WT (Fig. 2A-B). In agreement, cell surface biotinylated HA-⍺2δ-1(G209D) was decreased, by 86.2%, compared to HA-⍺2δ-1 WT (Fig. 2C-D).

The p.(Gly209Asp) variant abolishes ability of ⍺2δ-1 to promote CaV2.2 currents

CaV2.2 currents were then measured in tsA-201 cells transfected with HA-tagged CaV2.2 with β1b and α2δ-1 WT, ⍺2δ-1(G209D or no α2δ. While α2δ-1 WT increased CaV2.2 currents by ~13fold, ⍺2δ-1G209D produced no increase compared to without α2δ (Fig. 2E–G). Expression of all subunits was confirmed by western blotting and immunocytochemistry (Supplementary Fig. 6).
This effect was not specific to CaV2.2, as ⍺2δ-1 (G209D) also did not increase CaV2.1 currents (Supplementary Fig. 7A-C).
We next investigated the expression of the calcium channel complex at the plasma membrane, using double-tagged GFP_CaV2.2-HA. When GFP_CaV2.2-HA was co-expressed with α2δ-1 WT, this resulted in an increase in its cell surface expression compared with no ⍺2δ control (Fig. 3A-B).
However, this effect was completely absent when CaV2.2 was co-expressed with ⍺2δ-1 (G209D) (Fig. 3A-B).

α2δ-1 (G209D) does not promote CaV2.2 cell surface expression or trafficking in hippocampal neurons

CaV2.2 is a neuronal calcium channel, and we, therefore, investigated the effect of α2δ-1(G209D) on CaV2.2 trafficking in neurons, as previously described (13). We first analyzed cell surface expression of GFP_CaV2.2-HA in cultured hippocampal cell bodies. As expected, in the presence of α2δ-1 WT, GFP_CaV2.2-HA was strongly expressed at the cell surface (HA signal, Fig. 3C-D). In contrast, in the presence of α2δ-1 (G209D), GFP_CaV2.2-HA could not be detected at the cell surface, similar to no α2δ (Fig. 3C-D).
The neurites of these cells were then imaged. As expected, GFP_CaV2.2-HA showed strong expression when co-expressed with α2δ-1 WT; this was observed for both HA (cell surface) and GFP (total CaV2.2) (Fig. 3E-F). In contrast, α2δ-1 (G209D) did not promote trafficking of CaV2.2 into hippocampal neurites (Fig. 3E-F). This is indicated by the finding that both HA (cell surface CaV2.2) and GFP (total CaV2.2) signals were reduced in parallel.

α2δ-1G209D shows reduced complex formation with CaV2.2 and limited proteolytic cleavage

To examine whether the lack of ability of ⍺2δ-1G209D to promote calcium channel function was due to reduced interaction with CaV2.2, we performed co-immunoprecipitation experiments, using GFP_CaV2.2-HA (Fig. 4A). For α2δ-1 WT, robust interaction was shown by the presence of HA-α2δ-1 WT, co-immunoprecipitated by GFP_CaV2.2-HA, using anti-GFP antibody (Fig. 4A, lane 6). In contrast, very weak co-immunoprecipitation was observed for HA-α2δ-1(G209D) (Fig. 4A, lane 7), quantified in Fig. 4B. As a control, there was no co-immunoprecipitation of HA-α2δ-1 WT using CaV2.2-HA without a GFP tag (Fig. 4A, lane 5).
Interestingly, co-immunoprecipitated HA-α2δ-1(G209D) (Fig. 7A, lane 7, arrow) had a noticeably higher apparent molecular weight compared to HA-α2δ-1 WT (lane 6), and we found this was due to almost complete lack of proteolytic cleavage of HA-α2δ-1(G209) into α2 and δ (Supplementary Results, Supplementary Fig. 8).
In summary, these results show that α2δ-1(G209D) remains largely as the uncleaved immature form, indicating that it probably remains in the endoplasmic reticulum. In agreement with our previous results for uncleaved α2δ-1(17), it shows much less complex formation with CaV2.2. This result suggested that α2δ-1(G209D) would be unlikely to interfere with other α2δ proteins interacting with CaV2.2. In agreement with this, we found that α2δ-1(G209D) did not affect the ability of α2δ-3 to enhance CaV2.2 currents (Supplementary Fig. 9A-C). This result underscores that the p.(Gly209Asp) variant has a loss-of-function effect.

Discussion

In the current study, we show that biallelic loss-of-function variants in CACNA2D1 underlie DEE. In patient 1 the homozygous frameshift variant p.(Ser275Asnfs*13) causes nonsense-mediated mRNA decay of mutated CACNA2D1 transcripts and absence of α2δ-1 in patient-derived fibroblasts. The variants p.(Leu9Alafs*5) and p.(Gly209Asp) in patient 2 are a combination of a very early frameshift and a missense variant in trans, with the latter severely affecting CaV2 calcium channel function.
Our electrophysiological, biochemical and immunocytochemistry data show that α2δ-1G209D completely non-functional, in that, unlike wild-type α2δ-1(5), it traffics extremely poorly to the cell surface, and does not enhance the function or trafficking of CaV2 channels in both non-neuronal cells and hippocampal neurons. Furthermore, α2δ-1 (G209D) shows markedly reduced cleavage into α2 and δ, an enzymatic process that normally begins in the Golgi apparatus13,17. This suggests that α2δ-1 (G209D) does not traffic beyond the endoplasmic reticulum. Our previous finding that an uncleavable mutant α2δ-1 shows lower association with the CaV2.2 α1 subunit than the mature cleaved α2δ-1 (17), indicates that the lack of proteolytic cleavage of α2δ-1 (G209D) will likely contribute to the observed reduction in interaction of α2δ-1 (G209D) with the CaV2.2 α1 subunit. This demonstrates the importance of a detailed understanding of α2δ-1 processing and function, in order to identify the basis for such deleterious variants.
Variants in CACNA2D1 have previously been associated with cardiac phenotypes, in both humans and mice. Homozygous knockout of Cacna2d1 in mice resulted in a mild cardiac phenotype and reduced ventricular myocyte calcium current density (18). These mice also showed peripheral sensory deficits and delayed development of neuropathic pain-related responses (19).
Relevant to this, both patients showed insensibility to pain (Table 2). Transgenic mice constitutively over-expressing α2δ-1 have no gross nervous system defects (20). However, they show spontaneous epileptiform EEG abnormalities and behavioral arrest (21), suggesting that not only the spatial and temporal expression but also the expression strength of α2δ-1 is critical for proper functioning of the mouse brain. Indeed, α2δ-1 is the major α2δ isoform in rodent cerebral cortex (22). Furthermore, an auto-antibody recognizing α2δ-1 is found in cases of autoimmune encephalitis (23) and amyotrophic lateral sclerosis associated with type 2 diabetes (24). 
In humans, heterozygous variants in CACNA2D1 have previously been associated with inherited arrhythmogenic disease, including Brugada (25) and short QT (26) syndromes, as well as infantile spasms (27) and intellectual disability and epilepsy (28). Reevaluation of these monoallelic variants, together with genetic data presented here give rise to reasonable doubt about an association of these CACNA2D1 variants with disease (Supplementary Information).
Pathogenic variants in genes encoding several voltage-gated calcium channels have been associated with neurological diseases in humans, ranging from early-onset severe spinocerebellar ataxia with neurodevelopmental deficits to DEE (see Supplementary Information). In CACNA2D2, rare biallelic loss-of-function variation has been reported in individuals with DEE, corpus callosum hypoplasia, cerebellar atrophy and ataxia (9,10). The two unrelated patients reported here show considerable clinical overlap with individuals carrying homozygous CACNA2D2 loss-of-function variants, such as global developmental delay and/or intellectual disability, epilepsy and hypoplasia of the corpus callosum. However, atrophy of the brain affects the cerebrum in the two affected individuals with CACNA2D1 variants, whereas cerebellar atrophy was consistently reported in subjects with CACNA2D2 variants (29). These data suggest that loss of α2δ-1 or α2δ-2 cannot be compensated by any of the other α2δ subunits during development. In agreement with this, important, non-overlapping roles for specific α2δ proteins in synapse formation in vitro have been identified recently, some of which may be calcium channel-independent (30).
In conclusion, our data demonstrate that biallelic loss-of-function variants in CACNA2D1 underlie early-onset DEE characterized by microcephaly, profound developmental delay, seizures, visual impairment, truncal hypotonia, limb spasticity, and movement disorder. These clinical features are similar to those in previously reported individuals with homozygous CACNA2D2 null alleles. Individuals with biallelic CACNA2D1 or CACNA2D2 variants all have corpus callosum hypoplasia, while patients with CACNA2D1 variants show progressive cerebral atrophy whereas subjects harboring CACNA2D2 variants have cerebellar atrophy. The loss-of-function frameshift nature of two of the three identified CACNA2D1 variants together with our functional studies demonstrating a loss-of-function effect for the amino-acid substitution p.(Gly209Asp) confirm disease causation of homozygous and compound heterozygous CACNA2D1 variants, while also calling into question a causative role of monoallelic CACNA2D1 variants in intellectual disability, epilepsy and/or inherited arrhythmogenic diseases.