CACNA1D De Novo Mutations in Autism Spectrum Disorders Activate Cav1.3 L-Type Calcium Channels

Background Cav1.3 voltage-gated L-type calcium channels (LTCCs) are part of postsynaptic neuronal signaling networks. They play a key role in brain function, including fear memory and emotional and drug-taking behaviors. A whole-exome sequencing study identified a de novo mutation, p.A749G, in Cav1.3 α1-subunits (CACNA1D), the second main LTCC in the brain, as 1 of 62 high risk–conferring mutations in a cohort of patients with autism and intellectual disability. We screened all published genetic information available from whole-exome sequencing studies and identified a second de novo CACNA1D mutation, p.G407R. Both mutations are present only in the probands and not in their unaffected parents or siblings. Methods We functionally expressed both mutations in tsA-201 cells to study their functional consequences using whole-cell patch-clamp. Results The mutations p.A749G and p.G407R caused dramatic changes in channel gating by shifting (~15 mV) the voltage dependence for steady-state activation and inactivation to more negative voltages (p.A749G) or by pronounced slowing of current inactivation during depolarizing stimuli (p.G407R). In both cases, these changes are compatible with a gain-of-function phenotype. Conclusions Our data, together with the discovery that Cav1.3 gain-of-function causes primary aldosteronism with seizures, neurologic abnormalities, and intellectual disability, suggest that Cav1.3 gain-of-function mutations confer a major part of the risk for autism in the two probands and may even cause the disease. Our findings have immediate clinical relevance because blockers of LTCCs are available for therapeutic attempts in affected individuals. Patients should also be explored for other symptoms likely resulting from Cav1.3 hyperactivity, in particular, primary aldosteronism.

. Position of the Cav1. 3  A. Simplified scheme of Cav1.3 a1-topology with its four homologous repeats (segments S1 -S6). A749G and G407R (in alternative exon 8A) are shown together with two missense mutations discovered in aldosterone producing adenomas (somatic mutations G403R and I750M) for which a gain of function phenotype has been confirmed in previous functional studies (1,2). Mutations G403D (in alternative exon 8B) and I750M were also reported as germline mutations in individuals exhibiting congenital neurodevelopmental defects, intellectual disability and epilepsy in addition to hyperaldosteronism (PASNA) (2). The higher flexibility of the distal IIS6 helix is illustrated by color coding (low B-factor in red, high B-factor in green). This also results in a higher probability of IIS6 interactions with the adjacent S6 helices as part of the activation gate (not illustrated). Methodological details on modeling and MD simulations are given in the Supplemental Methods.
C. Prediction of structural changes imposed by mutation G407R: Location of residue 407 in the wild-type (C1) and mutant (C2) Cav1.3 a1-subunit. In the wild-type channel, it makes close contacts to the S4-S5 linker of repeat I (in particular residues S263, K266, A267). This is illustrated by the Van der Waals surface (grey) in C1. Panel C2 clearly shows that the IS4-IS5 interactions predicted in the wild-type are completely lost in the mutant channel. Moreover, the introduction of a positive charge enables new electrostatic interactions with adjacent IIS6 and IVS6 helices as part of the activation gate (not illustrated). The model also predicts that R407 can form ionic interactions with charged heads of lipids which additionally indicates the loss of interactions with the S4-S5-linker in the mutant. This is also demonstrated by the plot in C3, showing the distance between the IS6 activation gate and the IS4S5 linker in the wild-type (black line) and the mutant (red line).
In wild-type the distance remains stable at about 10 Å and even decreases during the simulation. In contrast, the mutant G407R shows an increase of the distance up to 20 Å.
Thus, the activation gate of domain I is moving away from the IS4S5 linker and is expected to lose key interactions for coupling with its voltage sensor. This is also confirmed by the root mean square deviation (RMSD) of the mutant with respect to the wild-type channel as shown in C4. The RMSD plot of G407R indicates that the structural model is stable after equilibration (indicated by the black line) and differs from the wild-type conformation (RMSD increases up to 7 Å, indicated by the red line).
Although the predictions of our molecular model still need to be confirmed by further mutational analysis, they clearly demonstrate that the mutations occur within regions that are crucial for the electromechanical coupling of the voltage-sensing domains to the channel pore and normal function of the activation gate (for recent reviews see 4,5).
The slowing of the inactivation in G407R can be explained by a loss of Ca 2+dependent inactivation and a slowing of voltage-dependent inactivation, because inactivation is much slower than observed with barium currents through Cav1.3 wild-type channels (cf. 6). This highlights the importance of the IS6 helix also for voltage-dependent inactivation of Cav1.3 channels. The mobility of the IS6 segment can also be affected by Ca 2+ channel bsubunits through binding to the I-II-linker of the channel (7-9) and also allows control of voltage-dependent inactivation (10). It will therefore be important to analyze Cav1.3 gain of function mutations also with b-subunit isoforms other than b3 in future studies.  To confirm that the smaller maximal current size measured for G407R can indeed affect excitability in an electrically excitable cell, we expressed this mutant in skeletal muscle GLTmyotubes. These lack functional a1-subunits, express native auxiliary b-and a2d-subunits and thus allow current recordings that are not contaminated by other Ca 2+ channels. . These changes involved a longer duration of the single events due to a slower Δf/f decay (example trace b) and/or to a longer plateau at each event peak (example trace c). In different myotubes these changes contributed to varying degrees to the shape of the larger single events. Taken together, these data further support a gain-of-function phenotype for the G407R mutation despite its smaller maximal current amplitudes in transfected cells.

Immunoblot Analysis
tsA 201-cells were transfected and cultured as described previously (17 Signals were detected with the image acquisition system Fx7 (Peqlab). Quantitation of band intensity was performed using ImageJ 1.46 (National Institute of Health). After background subtraction integrated densities of mutant and wild-type signals were normalized to the loading control.

Molecular Modeling
Since 3D-structural information from crystallography is not yet available for Cav1. were aligned with ClustalW, as already published by Zhang and coworkers for NavAb and Cav1.3 (19). However, sequences are not conserved in loops. Therefore ab initio modeling has to be used to generate stable conformations taking into account only the sequence information. In particular, we used the Rosetta method to build models for loops (20,21).
Moreover, we used the ab initio Rosetta method to generate structures for the activation gates that were missing in the NavAb channel template. In the case of mutations, the structure was derived from the wild-type model by replacing the mutated residue and carrying out a local energy minimization of its atoms using dedicated tools of MOE.
Moreover, in each case the C-terminal and N-terminal parts of each domain were capped to avoid perturbations by free charged functional groups. Once the protein model was data, such as the relative flexibility and distances between specific residues. Furthermore, the quality of the models has been evaluated using experimental knowledge (also from other Ca 2+ channel isoforms) and molecular dynamics simulations on our in-house GPU cluster that permitted to perform fast calculations (26). Parameters like the root mean square deviation were investigated to determine the structural stability of the models after minimization in their natural lipid environment.

Electrophysiological Recordings and Ca 2+ Imaging in Dysgenic Myotubes
Myotubes of the homozygous dysgenic (mdg/mdg) cell line GLT lack functional a1-subunits and express native auxiliary b-and a2d-subunits and thus allow current recordings that are not contaminated by other Ca 2+ channels. Myotubes were cultured and transfected as previously described (27)