Biophysical classification of a CACNA1D de novo mutation as a high-risk mutation for a severe neurodevelopmental disorder

Background There is increasing evidence that de novo CACNA1D missense mutations inducing increased Cav1.3 L-type Ca2+-channel-function confer a high risk for neurodevelopmental disorders (autism spectrum disorder with and without neurological and endocrine symptoms). Electrophysiological studies demonstrating the presence or absence of typical gain-of-function gating changes could therefore serve as a tool to distinguish likely disease-causing from non-pathogenic de novo CACNA1D variants in affected individuals. We tested this hypothesis for mutation S652L, which has previously been reported in twins with a severe neurodevelopmental disorder in the Deciphering Developmental Disorder Study, but has not been classified as a novel disease mutation. Methods For functional characterization, wild-type and mutant Cav1.3 channel complexes were expressed in tsA-201 cells and tested for typical gain-of-function gating changes using the whole-cell patch-clamp technique. Results Mutation S652L significantly shifted the voltage-dependence of activation and steady-state inactivation to more negative potentials (~ 13–17 mV) and increased window currents at subthreshold voltages. Moreover, it slowed tail currents and increased Ca2+-levels during action potential-like stimulations, characteristic for gain-of-function changes. To provide evidence that only gain-of-function variants confer high disease risk, we also studied missense variant S652W reported in apparently healthy individuals. S652W shifted activation and inactivation to more positive voltages, compatible with a loss-of-function phenotype. Mutation S652L increased the sensitivity of Cav1.3 for inhibition by the dihydropyridine L-type Ca2+-channel blocker isradipine by 3–4-fold. Conclusions and limitations Our data provide evidence that gain-of-function CACNA1D mutations, such as S652L, but not loss-of-function mutations, such as S652W, cause high risk for neurodevelopmental disorders including autism. This adds CACNA1D to the list of novel disease genes identified in the Deciphering Developmental Disorder Study. Although our study does not provide insight into the cellular mechanisms of pathological Cav1.3 signaling in neurons, we provide a unifying mechanism of gain-of-function CACNA1D mutations as a predictor for disease risk, which may allow the establishment of a more reliable diagnosis of affected individuals. Moreover, the increased sensitivity of S652L to isradipine encourages a therapeutic trial in the two affected individuals. This can address the important question to which extent symptoms are responsive to therapy with Ca2+-channel blockers.


Background
In electrically excitable cells, Ca 2+ inward current through voltage-gated Ca 2+ -channels (VGCCs; Cav) contributes to membrane depolarization and plays a key role in coupling electrical activity to intracellular Ca 2+dependent signaling processes (for review, see [1][2][3]). Therefore, VGCCs control essential physiological processes, such as hormone secretion, muscle contraction, sinoatrial node pacemaking, and sensory function. In the brain, they trigger neurotransmitter release, shape neuronal excitability, and couple excitation to gene expression associated with synaptic plasticity and different types of learning and memory [1,[3][4][5]. VGCC activity is fine-tuned to the specific requirements of cellular functions. This is accomplished by the functional heterogeneity and diverse subcellular targeting of ten poreforming α 1 -subunit isoforms [6], several accessory βand α 2 δ-subunits as well as by alternative splicing and post-translational modification (for a review, see [3]).
L-type Ca 2+ channels (LTCCs; Cav1) form one of the three main families of VGCCs. From the four members (Cav1.1-Cav1.4), Cav1.2 and Cav1.3 are expressed in most electrically excitable cells, including a wide variety of brain regions [3]. Both are located post-synaptically at somatodendritic locations and serve a key role in activity-dependent gene transcription [1,3,5]. Genetic variants in both channels have been associated with neuropsychiatric disease risk. Multiple common intronic single nucleotide polymorphisms (SNPs) in Cav1.2 (CACNA1C gene) LTCCs have consistently been associated with bipolar disorder and schizophrenia, although the consequences of these polymorphisms for Cav1.2 function remain unknown (for review, see [7][8][9]). Very rare de novo CACNA1C missense mutations cause Timothy Syndrome, a severe disease with lethal arrhythmias, facial dysmorphism, syndactyly and autism spectrum disorder (ASD) in surviving patients [10][11][12]. Together these findings have triggered new interest in clinical trials to repurpose LTCC blockers ("Ca 2+ -antagonists"), licensed as antihypertensive drugs since decades, also for the treatment of mood disorders [13].
Our studies revealed that all these genetic variants are not present in healthy parents or unaffected siblings and are absent in 141,456 reference genomes of controls without pediatric disease (gnomAD database, [21]). This is consistent with high penetrance and strongly supports a likely causative role of these mutations. Moreover, electrophysiological analysis of six mutations (from seven of these patients) after expression in HEK-293 cells revealed a highly consistent pattern of functional changes: they all induce gating changes that can enhance Cav1.3 Ca 2+ -current through these channels, in particular at subthreshold voltages. This gain-of-function is evident from a drastic slowing of channel inactivation and/ or by facilitation of channel opening at more negative voltages [16]. Therefore, the demonstration of such typical gain-of-function gating changes in functional studies may allow to distinguish likely pathogenic from nonpathogenic CACNA1D missense variants and help in the genetic diagnosis of individuals with neurodevelopmental disorders. This appears necessary because several genetic studies failed to classify CACNA1D missense variants as high-risk mutations and CACNA1D as a high-risk gene for neurodevelopmental disorders, including ASD [14,15,22,23]. For example, gain-of-function CACNA1D mutation G407R in a patient with ASD has been identified, but has not been classified as high-risk mutation. However, functional analysis revealed typical gain-of-function changes, which strongly support its pathogenic potential [14].
In contrast to de novo gene-disrupting mutations (nonsense, splice site, frameshift), which cause a protein loss-of-function, the prediction of the pathogenic potential of missense variants is more difficult because in most cases their functional consequences cannot be predicted by bioinformatics tools. While our data argue for a high disease risk due to Cav1.3 gain-of-function, heterozygous de novo CACNA1D variants resulting in a loss of Cav1.3 activity are unlikely to cause human disease. This is strongly supported by previous findings both in knockout mice (for a review, see [3]) and Cav1.3-deficient humans with sinoatrial node dysfunction and deafness (SANDD; OMIM #614896 [24,25]), in which functional loss of one or both CACNA1D alleles did not lead to a central nervous system (CNS) disease phenotype. This complicates the classification of new CACNA1D variants as high-risk mutations in genetic studies.
Here, we provide further convincing evidence for the high disease risk of gain-of-function de novo CACNA1D mutations for neurodevelopmental disorders. This is shown for mutation S652L, which has previously been identified in the Deciphering Developmental Disorders study, in a cohort of individuals with a severe developmental disorder of unknown cause [23]). However, in this study, it has not been classified as novel disease mutation with compelling evidence for pathogenicity and therefore CACNA1D has not been included as one of the 12 novel highrisk genes. Moreover, we demonstrate that a rare variant at the same position, S652W, induces a gating defect compatible with a loss-of-function, which explains its presence in apparently healthy individuals. Our data should raise awareness for the pathogenic potential of CACNA1D mutations, especially in patients without additional congenital endocrine symptoms as diagnostic features. De novo CACNA1D missense mutations may be underdiagnosed in clinical practice.
PCR reaction mix for PCR a, b and c contained 5 μl 10x Pfu buffer with 20 mM MgSO 4 (Cat# EP0571; Thermo Fisher Scientific,Waltham, MA, USA), 2 mM dNTP mix (Cat# R0241; Thermo Fisher Scientific, Waltham, MA, USA), 2.5 μl DMSO, 500 ng DNA template (PCR a and b) or 0.5 μl of PCR products a and b (PCR c), 10 pMol/μl forward primer, 10 pMol/μl reverse primer, 0.5 μl Pfu polymerase (native) (2.5 units/μl; Cat# EP0571; Thermo Fisher Scientific, Waltham, MA, USA) and nuclease-free water to a final volume of 50 μl. Following PCR program was performed: initial denaturation at 95°C for 3 min, then 35 cycles of 30 s denaturation at 95°C, 30 s annealing at 50°C, variable extension time at 72°C depending on fragment size (for Pfu polymerase: 2 min/kb) followed by a final elongation step of 72°C for 7 min. The integrity of all cloned constructs was confirmed by restriction site mapping and sequencing (Eurofins MWG Operon, Ebersberg, Germany).

Electrophysiological recordings in tsA-201 cells
For whole-cell patch-clamp experiments, patch pipettes were pulled in a micropipette puller (Sutter Instrument, Novato, CA, USA) using borosilicate glass capillaries (borosilicate glass; Cat# 64-0792, Warner Instruments, Hamden, CT, USA) and fire-polished using a MF-830 microforge (Narishige Co, Tokyo, Japan). Pipettes with a resistance of 1.5-3 MΩ were backfilled with internal solution containing (in mM): 135 CsCl, 10 Cs-EGTA, 1 MgCl 2 , 10 HEPES, 4 Na 2 ATP adjusted to pH 7.4 with CsOH. The bath solution contained (in mM): 15 CaCl 2 or 15 BaCl 2 , 150 Choline-Cl, 1 MgCl 2 , 10 HEPES, adjusted to pH 7.3 with CsOH. Whole-cell patch-clamp recordings were performed at room temperature (20-23°C) using an Axopatch 200B Amplifier (Molecular Devices, San José, CA, USA). Data were digitized (Digidata, 1322A digitizer, Molecular Devices, San José, CA, USA) at 50 kHz, low-pass filtered at 1-5 kHz and analyzed using pClamp 10.2 software (Molecular Devices, San José, CA, USA). Series resistance was compensated by 60-90% and all voltages were corrected for a liquid junction potential of − 9.3 mV [28]. Currents were leak subtracted either offline using a 50-ms hyperpolarizing voltage step from − 89 to − 99 mV or using an online P/ 4 protocol. Current-voltage (I-V) relationships were measured by applying 50 ms depolarizing square pulses to various test potentials (Δ 5 mV increments) starting from a holding potential (HP) of − 89 mV. I-V curves were fitted to the equation where I is the peak current, G max is the maximum conductance, V is the test potential, V rev is the extrapolated reversal potential, V 0.5 is the halfmaximal activation voltage, and k is the slope factor. The voltage dependence of activation was obtained from the I-V relationship by calculating the conductance (G = I/V − V rev ) followed by normalization (G/G max ) and plotting as a function of voltage. The G-V curve was fitted using the following Boltzmann relationship: The steady-state inactivation was determined by calculating the ratio between current amplitudes of a control versus a test pulse (I/I control ; both 20 ms to V max ) separated by a 5-s conditioning step to various potentials (10 mV increments; 30 s intersweep interval; HP: − 89 mV) and plotting as a function of voltage. Steady-state inactivation curves were fitted using a modified Boltzmann equation: G = (1 -G max )/(1 + exp [(V -V 0.5,inact )/k inact ] + G max where V 0.5inact is the halfmaximal inactivation voltage and k inact is the inactivation slope factor. Channel open probability was estimated by dividing the peak ionic tail current (I tail ) by the integrated "ON" gating charge (Q ON ) at the potential where there is no ionic inward or outward current (V rev ). ONgating currents were filtered at 5 kHz and digitized at 50 kHz. Window currents were obtained by multiplying the steady-state inactivation at a given voltage (Fig. 1c,d, Fig.  6b) with the corresponding current densities (pA/pF) at the given potentials of the I-V relationships (Fig. 1a,b,  Fig. 6a). Pulses to physiologically relevant potentials i.e. − 20 mV which corresponds to ∼ − 35-38 mV in physiological Ca 2+ -concentrations were applied for 5-s. Normalized inward Ca 2+ -currents (I Ca ) were multiplied with the corresponding conductance at − 20 mV (G: WT L , 0.1253; S652L S , 0.4330; WT S , 0.2647; S652L S , 0.6325) and plotted as a function of time. The percentage of inactivation during a 5-s long depolarizing pulse from a HP of − 89 mV to the potential of maximal inward current (V max ) was determined after 50, 100, 250, 500, 1000, and 5000 ms with Ca 2+ or Ba 2+ as a charge carrier. Ca 2+ -dependent inactivation (CDI) was determined over a broad voltage range by analyzing the fraction of remaining currents at the end of 250 ms depolarizations (expressed as fraction of the peak current amplitude, r 250 ) to different test potentials (Δ 10-mV increments) with Ca 2+ or Ba 2+ as a charge carrier resulting in typical U-shaped dependence of voltage. The difference between Ca 2+ -and Ba 2+ -curves indicates the strength of CDI and is defined as parameter f for each voltage step and plotted as a function of voltage. The fractional Ca 2+dependent component of inactivation after 250 ms was calculated as CDI = 1 − r Ca /r Ba , where r Ca and r Ba is the fraction of current remaining at a given time point of inactivation, respectively. Persistent currents were determined after 5-s long depolarizations to different potentials expressed as fractional persistent current (%) normalized to the peak current amplitude measured by a 20-ms pre-pulse to the V max in the same sweep. Tail currents were obtained from a transition from + 80 mV to − 60 mV or − 40 mV and normalized tail currents were fitted to a bi-exponential equation. In general, experiments with currents < 100 pA and > 1000 pA (range was defined prospectively) were excluded from analysis to avoid errors in the measurements of V 0.5,act , which tends to become more hyperpolarized at larger current amplitudes. For pharmacological experiments, cells were depolarized using a 100-ms square pulse to the V max of each individual cell (0.1 Hz; HP: − 89 mV). Cells were perfused using an air pressure-driven perfusion system (BPS-8 Valve Control System, ALA Scientific Instruments) with external bath solution (15 mM Ca 2+ ) in the presence or absence of isradipine (Fisher scientific, 50-850-70001) with a flow rate of 0.5 ml/min. Isradipine stocks were prepared in DMSO and freshly diluted 1: 1000 in bath solution to the final concentration prior to the experiment. On each recording day, individual control recordings with bath solution only were performed using the same tubes subsequently used for isradipine experiments. Drug application was started after at least three constant control sweeps during perfusion with bath solution. Drug effects were corrected for linear current decay ("run-down") measured in control cells.  3 splice variants recorded in parallel on the same day using 50-ms depolarizing square pulses to various test potentials from a holding potential (HP) of -89 mV. Inset: Representative I Ca traces upon depolarization to the potential of maximal inward current (V max ). Statistics: two-way ANOVA followed by Bonferroni post hoc test, *p < 0.05, **p < 0.01, ***p < 0.001. c, d Normalized steady-state activation and inactivation curves of WT L vs S652L L (c) and WT S vs S652L S (d). Data are presented as mean ± SEM; for parameters, statistics and number of experiments see Table 1. e, f Window currents of WT L vs S652L L (E) and WT S vs S652L S (F). Data were obtained by multiplying the steady-state inactivation (c, d) at a given potential with the corresponding current densities of the I-V-relationships (a, b). Statistics: Student's t-test (multiple comparison adjusted),***p < 0.001, **p < 0.01. Data are represented as mean ± SEM for the n-numbers indicated. Data were collected from > 3 independent transfections trains were applied at a frequency of 10 Hz and 300 sweeps per run. The APW protocol was elicited from -80-mV HP composed of 3 voltage ramps: step: − 80 to − 60 mV for 2.5 ms, 1st ramp: − 60 to + 20 mV in 1 ms, 2nd ramp: + 20 to -70 mV in 1.5 ms, 3rd ramp: − 70 to − 60 mV in 5 ms (afterhyperpolarization), step: − 60 mV for 90 ms. Fluorescent signals were normalized to baseline fluorescence (F 0 ) and current density (pA/pF) determined in a ramp before the start of the train. Ca 2+charge of WT L and S652L L was obtained by integrating the area of I tail normalized to maximum I Ca determined in a ramp before the start of the train.
The absolute expression level of α 1 -subunit protein varies between membrane preparations of individual transfections. Therefore, transfections, membrane preparations, and Western blot analysis were always carried out in parallel for WT and mutant channels.

Homology modeling
We predicted the structure of the WT Cav1.3 α 1 -subunit and two mutants by developing a homology model based on the cryo-electron microscopy (EM) structure of the Cav1.1 α 1 -subunit in the closed (potentially inactivated) state [30]. The high sequence conservation of the Cav1.1 α 1 -subunit to the Cav1.3 α 1 -subunit (~75% similarity and~62% identity measured by MOE) allowed us to predict a reliable structure model. Homology modeling has been performed using Rosetta and MOE (Molecular Operating Environment, version 2018.08, Molecular Computing Group Inc., Montreal, Canada). Additionally, ab initio Rosetta was used to generate structures for loops that were not resolved in the original Cav1.1 α 1 -subunit template. The structures for the mutants were derived from the WT model by replacing the mutated residue and carrying out a local energy minimization using MOE. The C-terminal and Nterminal parts of each domain were capped with acetylamide (ACE) and N-methylamide to avoid perturbations by free charged functional groups. The structure model was embedded in a plasma membrane consisting of POPC (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine) and cholesterol in a 3:1 ratio, using the CHARMM-GUI Membrane Builder. Water molecules and 0.15 M KCl were included in the simulation box. Energy minimizations of WT and mutant structures in the membrane environment were performed. The topology was generated with the LEaP tool of the AmberTools18, using force fields for proteins and lipids, ff14SBonlysc and Lipid14, respectively. The WT and mutant structures were gradually heated from 0 to 300 K in two steps, keeping the lipids fixed, and then equilibrated over 1 ns. Then molecular dynamics simulations were performed for 10 ns, with time steps of 2 fs, at 300 K and in anisotropic pressure scaling conditions. Van der Waals and short-range electrostatic interactions were cut off at 10 Å, whereas long-range electrostatics were calculated by the Particle Mesh Ewald (PME) method. MOE was used to visualize the key interactions and point out differences in the WT and mutant structures.

Ethics approval
The Deciphering Developmental Disorders Study [23] has UK Research Ethics Committee approval (10/ H0305/83, granted by the Cambridge South REC, and GEN/284/12 granted by the Republic of Ireland REC). Parental informed consent has been obtained for updated clinical information of the probands.

Statistics
Data were analyzed using Clampfit 10.2 (Axon Instruments) and Sigma Plot 11 (Systat Software, Chicago, IL). For statistical analysis Graph Pad Prism 5.01 software (GraphPad Software, La Jolla, CA) was used. Significance of group differences between two groups was determined using unpaired Student's t test for normally distributed data or Mann-Whitney U test for non-normally distributed data. Significance of group differences between three and more groups was determined using one-way analyses of variance (ANOVA) or two-way ANOVA for normally distributed data (with Bonferroni post-test as indicated). All data are represented as mean ± SEM. Significance level was set to α-error lower than p < 0.05 (*), p < 0.01 (**) and p < 0.001 (***). All original datasets are available from the corresponding author on reasonable request.

Results
S652 L, a novel CACNA1D de novo mutation in monozygotic twins with a severe neurodevelopmental disorder and ASD The novel CACNA1D variant has been reported in the Deciphering Developmental Disorders Study [23]. In a large unbiased genotype-driven approach this study analyzed 1,133 children with severe, undiagnosed developmental disorders and their parents using a combination of exome sequencing and array-based detection of chromosomal rearrangements [23]. The CACNA1D variant (chr3: 53757881 C>T, human reference genome hg19) was predicted to cause a p.Ser652Leu (S652L) mutation (reference sequence NM_001128839). However, it was not considered a novel disease mutation in this study. The patients (DECIPHER database individuals #262954 and #262956; decipher.sanger.ac.uk) are male monozygotic twins, thirteen years old at present, and both harbor one copy of the mutation. Their clinical phenotype has been updated recently. It manifests as a severe neurodevelopmental phenotype with delayed speech and language development and a global developmental delay. Both individuals show self-injurious behaviors and have been diagnosed with ASD two years ago. One patient is tall (5 ft 5 in.) and heavy (57 kg) for his age, has undescended testes and suffered from seizures at the age of two without recurrence. In addition, he shows challenging behavior with attention deficit hyperactivity disorder-like symptoms. No abnormal blood pressure has so far been reported and both are currently not treated with any medication. In addition, facial dysmorphism characterized by epicanthus, abnormality of the nose, microtia, a small vermillion border, and widely spaced teeth was noted. Prediction of the possible impact of this amino acid substitution on protein structure and function using bioinformatics prediction tools Poly-Phen2, SIFT and MutationTaster indicated a probably damaging (score: 1.00), deleterious (100%) or diseasecausing (probability: 0.999) effect, respectively. This variant is not reported in the gnomAD database (a reference database that lists exomes and genomes from a total of 141,456 unrelated individuals harboring mutations without pediatric disease [21]).
A frameshift mutation (c.1934_1935insG (p.Glu646Glyf-sTer)) in the KIF22 gene (chr16: 29816479 T>TG, human reference genome hg19) was also identified in both patients. This gene is highly expressed in bones, cartilage, skin, ligaments and joint capsules [31]. Mutations in the KIF22 gene have so far not been associated with neurodevelopmental disorders but result in a syndrome called spondyloepimetaphyseal dysplasia with joint laxity (SEMDJL; OMIM #603213), with malformations of the spine, skeletal dysplasia and malalignment of limbs but no intellectual impairment. Complete knock-out of Kif22 in mice results in premature intrauterine death but surviving Kif22 −/− embryos develop into healthy adult mice [32]. Moreover, protein loss-of-function variants (stop gained, frameshift mutations) are reported in more than 40 controls in the gnomAD database. Therefore, the KIF22 mutation is unlikely to explain the severe neurodevelopmental phenotype of the two patients. In contrast, CACNA1D mutation S652L has also been reported as a somatic aldosterone-producing-adenoma (APA) mutation [33] in a patient suffering from resistant hypertension, providing additional indirect evidence for a pathogenic role of this novel germline CACNA1D variant.

Mutation S652L changes the voltage-dependence of activation and inactivation
Based on recent findings of unique gating changes induced by pathogenic CACNA1D de novo mutations [14-18, 20, 34], we therefore hypothesized that S652 L could also explain the neurodevelopmental phenotype in both patients. For a detailed biophysical characterization in tsA-201 cells we introduced this mutation into the biophysically distinct C-terminally long (WT L , S652L L ) and short (WT S , S652L S ) Cav1.3 splice variants [26]. Both splice variants are abundantly expressed in the brain and differ considerably with respect to their biophysical properties, with higher voltage-sensitivity and faster Ca 2+dependent inactivation of WT S [3,16]. Mutant α 1 -subunit proteins were expressed as intact proteins with the expected molecular mass (Additional file 1: Figure S1).
Mutation S652L induced pronounced gating changes. It significantly shifted the voltage-dependence of activation (Fig. 1a, b) and steady-state inactivation to more negative potentials indicating a phenotype that can support a channel gain-of-function in both splice variants (Fig. 1c, d; for parameters see Table 1) by promoting Ca 2+ -inward currents (I Ca ) at negative voltages. As a consequence the mutation induced a higher window current at subthreshold potentials (− 50 and − 40 mV) compared to WT in the short Cav1.3 splice variant (Fig. 1e, f). These variants comprise about half of the Cav1.3 α 1 -subunits in the brain [35]. At − 50 mV significant window current was only measurable in S652 L S but not in WT S , and it was two times larger in the mutant at − 40 mV. Whereas current amplitudes were larger at negative voltages, I Ca above the potential of maximal inward current (V max ) were significantly smaller (Fig. 1a, b). Since gating currents, which are a measure for the number of active channels at the cell surface, were not different in the mutant channels (Q ON [mean ± SEM; pA*ms)]: WT L , 158.9 ± 26.3, n = 23; S652 L L , 140.3 ± 25.1, n = 21; Mann-Whitney test), reduced current density is likely due to a decreased open probability (P O ). This is further supported by a significantly decreased slope of the I tail /Q ON relationship for S652 L L channels (I tail /Q ON [linear regression slopes, mean ± SEM; ms −1 ]: WT L : − 7.22 ± 0.916, r 2 = 0.72, n = 26; S652 L L : − 4.24 ± 0.657, r 2 = 0.72, n = 25; slopes are significantly different: F = 6.43, p = 0.015, F test; Additional file 2: Figure  S2).
Note that short Cav1.3 splice variants have very small, non-measurable ON-gating currents [26] and were therefore not further analyzed here.

Mutation S652L accelerates voltage-dependent inactivation but has opposing effects on Ca 2+ -dependent inactivation
Since either acceleration or slowing of the inactivation time course of Cav1.3 is also a hallmark of pathogenic CACNA1D mutations, we studied voltage-(VDI) and Ca 2+ -dependent inactivation (CDI). Mutation S652L significantly accelerated inactivation kinetics (Fig. 2) during 5-s depolarizations to V max with both Ba 2+ (which reports VDI) and Ca 2+ (which, in addition, induces CDI) as charge carriers in both Cterminally long (Fig. 2a) and short ( Fig. 2b; for statistics, see Table 2) splice variants. It also significantly reduced I Ca during long-lasting depolarizations as shown as the percentage of remaining current after 5s depolarizations to various test potentials in both Cav1.3 splice variants (Fig. 2c, d). By comparing the difference of fractional inactivation between I Ca and inward Ba 2+ -current (I Ba ) after 250-ms depolarizations to different test potentials (f value, see Fig. 3) mutational effects on the voltage-dependence of CDI could also be determined. Whereas maximal CDI was unchanged in the long Cav1.3 splice variant (Fig. 3a, c), it was significantly reduced in the short variant (Fig.  3b, d). Therefore, faster inactivation of I Ca must be due to the acceleration of VDI, despite being partially Table 1 Steady-state activation and inactivation parameters of mutation S652L Activation Inactivation   Table 2. c, d Persistent currents were determined after 5-s long depolarizations to different potentials and are expressed as % of the peak current amplitude measured by a preceding 20 ms pre-pulse to V max . Statistics: two-way ANOVA followed by Bonferroni post hoc test, *p < 0.05, ***p < 0.001. Data are represented as mean ± SEM for the n-numbers indicated. r values represent the percent of remaining I Ca or I Ba after 50, 100, 250, 500, 1000 and 5000 ms (5-s depolarization from a HP of − 89 mV to V max ). All values are presented as mean ± SEM (> 3 independent transfections). Statistics: unpaired Student's t test of r values, *p < 0.05, **p < 0.01, ***p < 0.001 in comparison to respective WT. n, number of recordings; WT, wild-type compensated by reduced CDI in WT S . Assuming that CDI and VDI are independent processes, we also calculated the fractional Ca 2+ -dependent component of inactivation as previously described [36] for WT S vs S652L S from the data shown in Table 2. After 250 ms of inactivation there was no difference between WT S (0.778 ± 0.027) and S652L S (0.805 ± 0.021; p = 0.40, n = 21; unpaired Student's t test) and this was also true for all other time points. This further confirms that S652L can promote Cav1.3 inactivation largely by affecting VDI.

Mutation S652L increases Ca 2+ -signaling during stimulation protocols simulating neuronal firing patterns
In order to predict the consequences of mutation S652L on Ca 2+ -influx during neuronal activity patterns, we simulated its activity during sustained upstate potentials and during action potential firing. At low potentials, Cav1.3 channels can contribute to the formation of plateau potentials due to their known negative activation range (for a review, see [3]). This was described in medium spiny neurons when transient upstate potentials were induced by glutamatergic excitatory input [37]. To quantify changes of I Ca amplitudes induced by the mutation at sustained subthreshold depolarizations we mimicked plateau potentials by prolonged depolarizations to − 20 mV with 15 mM Ca 2+ as charge carrier. This voltage would correspond to~− 35-38 mV at physiological Ca 2+ -concentrations [26] (Fig. 4a, b). The resulting I Ca traces were normalized to maximal I Ca at V max in individual cells and corrected for splicing-and mutationdependent differences in V 0.5 by multiplying with the corresponding conductance measured for the different WT and mutant constructs at − 20 mV (derived from activation curves in Fig. 1c, d). As shown in Fig. 4a, b, mutated channels caused a marked and highly significant increase of I Ca over the first 300 ms of the depolarization compatible with enhanced channel activity suitable to support upstate potentials during this time period (for details, see legend to Fig. 4).
In order to predict the consequences of mutation S652L on Ca 2+ -influx during action potential firing we simultaneously measured I Ca and cytosolic Ca 2+responses upon stimulation of HEK-293 cells transfected with WT L and S652L L with 10-Hz-trains of action potential-like waveforms (APW) (Fig. 5a; for details, see legend) [26,38]. These stimuli resulted in typical I Ca transients (Fig. 5a) with maximal I Ca reached during the repolarization phase of the APW [29,38,39]. Peak I Ca slowly decreased during trains, an effect that was significantly enhanced by the mutation (Fig. 5a, b). However, the mutation enhanced the increase in intracellular Ca 2+ -levels measured simultaneously during this stimulation protocol (Fig. 5c). To explain this discrepancy, we measured the total Ca 2+ -charge during the 30-s train. This was significantly higher in S652L mutated channels even after 15 s of stimulation ( Fig. 5d; for details, see legend). It was not due to higher S652L channel expression, because we normalized data to current density for each cell. As shown in Fig. 5e, I Ca amplitude during APW repolarization was significantly larger for S652L as compared to WT L (peak 1 st AP [pA/pF]: WT L , − 12.43 ± 1.95, n = 19; S652L L , − 33.64 ± 3.13***, n = 21; unpaired Student's t test, ***p < 0.001). This could be explained by the more negative activation voltage range as well as a pronounced slowing of I Ca deactivation. Deactivation of tail currents following repolarizations from + 80 mV to − 60 mV or − 40 mV was significantly slower in S652L L as compared to WT L . This was primarily caused by a decrease in the contribution of the fast component and an increase of the slow component of the bi-exponential deactivation process ( Fig. 5f; for statistics, see Table 3). Thus, higher Ca 2+ -levels and Ca 2+ -charge can be attributed to the slower deactivation kinetics and higher current amplitudes induced by mutation S652L during action potential like firing.

Mutation S652W produces a loss of Cav1.3 channel function
Our data predict that only CACNA1D mutations that are able to enhance Cav1.3 channel activity can confer high risk for neurodevelopmental symptoms. Accordingly, like any of the other previously described pathogenic variants, S652L has not been reported in the genomes of 141,456 control individuals free from pediatric disease (gnomAD database [21];). In contrast, the pathogenic potential should be low or absent from missense mutations causing gating defects favoring reduced function, as outlined above, in mice [40,41] and humans [24,25]. Interestingly, the gnomAD database also reports the rare variant p.Ser672Trp (chr3: 53757881 C>G, human reference genome hg19) variant (S652W), located in the same position as S652L, in three healthy unrelated individuals. PolyPhen2 and SIFT predict a probably damaging (score: 1.00) and deleterious (100%) effect on protein function, respectively, but these algorithms cannot predict gating changes. This provided us with a unique opportunity to further test our above hypothesis by studying the biophysical properties of S652W. As shown in Fig. 6b, this mutation has opposite effects on the voltage-dependence of gating compared to S652L. Both steady-state activation and inactivation were significantly shifted to more positive voltages by 4-5 mV (for statistics see Table 4).
Consequently, window currents were not increased and even tended to be shifted to more positive voltages also compatible with a loss-of-function at threshold voltages (Fig. 6c). S652W also failed to enhance channel function by other mechanisms: it did neither slow inactivation kinetics ( Fig. 6d; for statistics, see Table 5) nor change the fraction of non-inactivating current as evident from the steady-state inactivation analysis (Fig. 6b).
Molecular modeling of Cav1.3 WT, S652L, and S652W α 1subunits On the structural level, the loss of a newly discovered inter-domain hydrogen bond connecting the S4-S5 linkers of repeats II and I could explain the gating differences between the two variants (Fig. 7). Our Cav1.3 channel homology model, based on the cryo-EM structure of the Cav1.1 α 1 -subunit [30], localizes S652 at the C-terminal end of the S4-S5 linker in channel repeat II (Fig. 7, upper, left). The S4-S5 ; WT L , − 11.93 ± 1.46; S652L L , − 8.51 ± 1.04) determined in a ramp protocol before the start of the train. S652L L Cav1.3 channels showed higher levels of [Ca 2+ ] than WT L after 30 s of stimulation. Statistics: unpaired Student's t test, **p < 0.01. d Ca 2+charge of WT L and S652 L L obtained by integrating the area of the I Ca transient normalized to maximum I Ca determined in a ramp protocol before the start of the train. Statistics: two-way ANOVA of selected time points (every 5 s), *p < 0.05, **p < 0.01, ***p < 0.001. e Overlay of 1 st AP of WT L and S652L L I Ca transients normalized to current density (pA/pF) determined in a ramp protocol before the start of the train to demonstrate larger APs induced by mutation S652L. f Normalized representative I Ca transients of WT L and S652L L obtained from repolarisations from + 80 mV to − 60 mV (left) or − 40 mV (right); scale bars correspond to the traces of the same color; for parameters and statistics see Table 3. The AP-like command voltage also triggered an outward current component occurring at the peak of the AP spike. We and others (see references in Ortner et al. [29]) have observed this previously. The outward component is likely composed of Q ON and a passive non-LTCC component (also found in non-transfected cells, [29]) Table 3 Tail current parameters of mutation S652L Normalized tail currents were fitted to a bi-exponential decay (τ fast, τ slow: time constants of slow and fast component; A fast , A slow : amplitudes of slow and fast components; C: non-inactivating component). Statistics: Mann-Whitney U test, *p < 0.05, **p < 0.01, ***p < 0.001. All values are presented as mean ± SEM (> 3 independent transfections). n, number of recordings linkers in each repeat are known to form contacts with the cytoplasmic end of their corresponding S6 helices, which together form the inner mouth of the channel (activation gate [30];). This allows the S4-S5 linkers to transmit voltage-sensor movements to the activation gate. Our model predicts that S652 forms a hydrogen bond with S256 in the S4-S5 linkers of the neighboring repeat I (Fig. 7a). Therefore, this hydrogen bond connects the S4-S5 linkers in two different repeats and could be crucial for controlling the voltage-sensitivity of channel gating. This is supported by our finding that in the S652L mutant, this inter-domain interaction is not stabilized by any hydrogen bonds. Instead, leucine forms much weaker hydrophobic contacts with the residues V259, V260, and L261 located in the S4-S5 linkers of repeat I (Fig. 7b). In contrast, the aromatic side chain of the tryptophan in the S652 W mutant is capable of forming an interdomain pi-H interaction with S256 in IS4-S5 as well as an intra-domain hydrogen bond with its own IIS4-S5 backbone (K648) (Fig. 7c), which should also allow a rigidifying effect similar as in the WT channel. Together with our functional data, this reveals the importance of an inter-domain hydrogen bond for normal  Table 4. c Window currents of WT L vs S652W L were calculated as in Fig. 1e, f. Data are presented as means for the indicated number of experiments. d Inactivation kinetics of WT L vs. S652W L during a 5-s depolarization to V max with Ca 2+ as a charge carrier showing no difference in inactivation kinetics between WT L and S652W L . Inset shows the first 2000 ms. Data are presented as mean ± SEM; for statistics and number of experiments, see Table 5. The significant shift of activation and inactivation half-maximal voltage to positive voltages and the absence of a change in inactivation time course was independently confirmed in an independent set of experiments using HEK-293 cells stably expressing β 3 and α2δ-1 subunits with essentially identical results. Statistics: one-way ANOVA followed by Bonferroni post hoc test, *p < 0.05. Data are represented as mean ± SEM for the n-numbers indicated. Data were collected from > 3 independent transfections

Mutation S652L increases the sensitivity of Cav1.3 channels for inhibition by the dihydropyridine LTCC blocker isradipine
The fact that pathogenicity is associated with enhanced channel function but that reduced Cav1.3 channel function in the brain is not associated with CNS symptoms (see above), make available LTCC blockers a potential therapeutic option for the symptomatic treatment of individuals affected by gain-of-function mutations.
These drugs, such as the dihydropyridines (DHPs) nifedipine, felodipine, or isradipine, are safely used since decades for the treatment of arterial hypertension and angina. Since DHPs preferentially bind to channels in an inactivated state [42,43], mutations affecting the gating properties, such as S652L may change the sensitivity of the channel for inhibition by DHPs. In order to quantify mutation-induced changes on DHP sensitivity, we assessed inhibition of the Cterminally long WT and S652L mutant channels by the DHP isradipine using a standard square pulse protocol (100 ms to V max , 0.1 Hz, HP: − 89 mV). As illustrated in Fig. 8, S652L-mutated channels required significantly lower isradipine concentrations for channel inhibition with a 3-4-fold decrease of their half maximal inhibitory concentration (IC 50 , mean (95% coincidence interval); WT L : 60.3 (52.0 -70.0) nM, S652L L : 18.1 (15.3 -21.5) nM; Fig. 8). This finding suggests that DHPs with good brain penetrance, such as isradipine [29,44] may preferentially inhibit S652L-mutated Cav1.3 channels. Based on their good clinical safety profile this encourages therapeutic trials with DHPs in the affected individuals.

Discussion
In this study, we provide compelling evidence for the CACNA1D S652L variant as a high-risk and likely disease-causing mutation in two individuals of the Deciphering Developmental Disorders cohort (decipher.sanger.ac.uk) of children with severe, undiagnosed developmental disorders [23]. This evidence builds on a detailed biophysical characterization that demonstrates gating changes able to also induce enhanced channel activity as is typical for six other de novo germline missense mutations in patients with ASD with and without other neurodevelopmental symptoms [14][15][16][17][18]20]. Our data adds the CACNA1D gene to the other 12 developmental-disorder-linked genes identified in the Deciphering Developmental Disorders study and thus further increases its diagnostic yield. Moreover, since S652L has also been found in APAs as a somatic mutation we also confirm S652L as a disease-causing mutation in patients with primary aldosteronism [33]. We also demonstrate that although heterozygous missense mutations in the same position can be damaging, they can have opposite effects on channel function and, in the case of Cav1.3 α 1 -subunits, may comprise low or no pathogenic risk when the resulting gating changes does not support a gain-of-function. This complicates predictions of the disease-causing role of rare de novo CACNA1D variants in genetic studies and emphasizes the need for functional analysis as described in this report. We clearly show this for S652W. Whereas neither S652L nor one of the other pathogenic CACNA1D mutations is reported in the gnomAD database, the heterozygous S652W variant is reported in three neurologically apparently healthy individuals in this database [21]. In contrast to S652L, this variant tended to shift the window current to more positive rather than negative voltages, which is due to the positive shift of the voltagedependence of activation and inactivation. This nicely fits our prediction that only CACNA1D mutations which can also support channel gain-of-function confer high risk for neurodevelopmental disorders. It is also in line with previous studies in mice (for a review, see [3]) and humans [24,25]. These found that heterozygous loss-offunction of Cav1.3 is unlikely to cause symptomatic neurodevelopmental disorders and even homozygous loss of Cav1.3 function has not been reported to cause neuropsychiatric behavioral changes [3,24,25,41]. Therefore, a unifying feature of all pathogenic mutations described so far is the potential to induce gating changes that can enhance Cav1.3 function during neuronal activity. The potential for enhanced channel function may, however, vary between different neurons. For example, in neurons firing from more depolarized membrane potentials, the negative shift in steady-state inactivation may also reduce the availability of Cav1.3 channels. S652L adds to another six de novo missense mutations functionally characterized so far by us [14][15][16] and others [17,18,20,38] in a total of seven patients with neurodevelopmental disorders. Functionally they fall into two major classes. Channel gain-of-function is either predominantly induced by stabilizing a large noninactivating current component (type 1: in particular, G403D, G407R) or by inducing a strong shift of activation voltage to more negative potentials (type 2: A749G, I750M and V401L) [14][15][16][17], as also observed for S652L. Both types of mutations enhance intracellular Ca 2+ -load when expressed in HEK-293 (A749G, [38]; S652L, this paper) or GLT muscle cells (G407R, [14]). Pronounced negative shifts of activation and/or pronounced slowing of channel inactivation can therefore be taken as "diagnostic" feature for the pathogenicity of CACNA1D de novo mutations in patients with neurodevelopmental disease.
Importantly, our data also strongly suggest that pathogenicity can also be assumed if the same variant has also been reported as a somatic mutation in APAs. As shown here, this is the case for S652L [33] and likewise for G403D, I750M and V401L [17,45].
We have also observed a small but significant shift of V rev (Fig. 1, Table 1) to more negative voltages. This could indicate a potential change in ion selectivity by the mutation. Interestingly, we have also detected similar shifts by other Cav1.3 α 1 -subunit gain-of-function mutations [16]. If confirmed by single-channel recordings, this may also contribute to the mutation-induced pathological signaling changes.
Very recently identified de novo mutation V259A has been reported in another severely affected individual, a 1-year-old male with seizures, global developmental delay and primary aldosteronism [19]. So far this mutation has not been functionally characterized. Although its pathogenic potential appears to be high based on two different APA mutations reported in the same position (V259D, V259G) [46]), final proof requires functional analysis as described here.
Finally, another important and clinically highly relevant result of our study was the observation that Cav1.3 channels harboring the S652L-mutation require lower concentrations of the DHP isradipine for inhibition. This can be explained by the known voltage-dependence of DHP action due to their preferred binding to inactivated channel states [29,42]. S652L induces a pronounced negative shift of the voltage-dependence of inactivation thus increasing the availability of inactivated channels. This preclinical finding is a strong motivation to test if repurposing of already available DHPs could ameliorate symptoms in affected individuals. Since also other mutations inducing strong shifts in steady-state inactivation are likely to increase DHP sensitivity (e.g. A749G, [14]), this treatment approach, if successful, could also be offered to individuals with other CACNA1D mutations.

Limitations
Although our data strongly support CACNA1D as a high-risk gene for neurodevelopmental disorders and emphasize the need of functional analysis to distinguish likely pathogenic (able to increase Cav1.3 activity) from non-pathogenic de novo mutations (unable to increase Cav1.3 activity), our studies do not provide insight into altered signaling cascades downstream of Cav1.3 channels. This will require introduction of one or more of these human mutations into the mouse Cacna1d gene for electrophysiological and biochemical studies in native cells. Such animal models will also allow to address the important question, if currently available LTCC blocker, such as isradipine, felodipine of nimodipine, can normalize cellular function and, perhaps, even behavioral phenotypes in these mice. Although LTCC blockers may normalize the mutation-induced increase in channel function after diagnosis, it is possible that the mutation may have already caused permanent developmental deficiencies resistant to drug treatment. Therefore the clinical potential of this therapy needs to be tested in small clinical trials in affected individuals.

Conclusions
Taken together our data have important implications for genetic diagnostics. We provide evidence that CACN A1D is a neurodevelopmental disorder-linked gene. Although initially considered to cause high risk only for ASD with or without intellectual disability [14], the increasing number of affected individuals, including S652L, now strongly indicates that the majority presents with a more severe phenotype. This can involve seizures, intellectual disability and, due to the role of Cav1.3 for aldosterone and insulin secretion [40,46]), also with (often transient, [17,18]) endocrine symptoms. Our findings with S652W (the loss-of-function mutation) emphasize that, in the case of CACNA1D, the amino acid position itself does not allow predictions about the disease risk of a variant, even if bioinformatics prediction tools provide high scores for protein damage. A high probability for pathogenicity can also be assumed if a variant identical to the germline mutation has also been found in at least two different individuals as a somatic mutation in an APA or an aldosterone-producing cellcluster [34,47,48]. Our report should raise awareness for the pathogenic potential of CACNA1D mutations, especially in patients without additional congenital endocrine symptoms as diagnostic features. At present, de