Complex effects on CaV2.1 channel gating caused by a CACNA1A variant associated with a severe neurodevelopmental disorder

P/Q-type Ca2+ currents mediated by CaV2.1 channels are essential for active neurotransmitter release at neuromuscular junctions and many central synapses. Mutations in CACNA1A, the gene encoding the principal CaV2.1 α1A subunit, cause a broad spectrum of neurological disorders. Typically, gain-of-function (GOF) mutations are associated with migraine and epilepsy while loss-of-function (LOF) mutations are causative for episodic and congenital ataxias. However, a cluster of severe CaV2.1 channelopathies have overlapping presentations which suggests that channel dysfunction in these disorders cannot always be defined bimodally as GOF or LOF. In particular, the R1667P mutation causes focal seizures, generalized hypotonia, dysarthria, congenital ataxia and, in one case, cerebral edema leading ultimately to death. Here, we demonstrate that the R1667P mutation causes both channel GOF (hyperpolarizing voltage-dependence of activation, slowed deactivation) and LOF (slowed activation kinetics) when expressed heterologously in tsA-201 cells. We also observed a substantial reduction in Ca2+ current density in this heterologous system. These changes in channel gating and availability/expression manifested in diminished Ca2+ flux during action potential-like stimuli. However, the integrated Ca2+ fluxes were no different when normalized to tail current amplitude measured upon repolarization from the reversal potential. In summary, our findings indicate a complex functional effect of R1667P and support the idea that pathological missense mutations in CaV2.1 may not represent exclusively GOF or LOF.


Results
The Ca V 2.1 R1667P variant. The Ca V 2.1 R1667P variant was identified earlier in a female child with focal seizures, generalized hypotonia, dysarthria, congenital ataxia and fatal cerebral edema 57 . We now report a second female child carrying the R1667P variant who presented with global developmental delay, microcephaly, pontocerebellar hypoplasia, thinning of the corpus collosum, small cerebellum, brainstem and pons. In both cases, the arginine to proline substitution was a consequence of a de novo guanine to cytosine mutation at bp 5000 of the CACNA1A coding sequence.
Clinical presentation. The proband is a 6-year, 10-month-old female product of a 37-week in-vitro fertilization, twin gestation. The pregnancy was complicated by an intrauterine bleed at 5 weeks and premature contractions at 35 weeks. She was born via scheduled Caesarean section and was discharged with no post-natal concerns. Birth weight was 6 lbs, 8 oz and she was 18.8 inches long. Her fraternal twin sister is developmentally normal. At four months old she was diagnosed with severe hypotonia, a brain MRI found pontocerebellar hypoplasia, a thin appearing corpus callosum and enlargement of the extra-axial subarachnoid spaces; her electroencephalogram (EEG) was normal. Currently, she is unable to crawl or walk but can be propped to sit. A sleep study at 15 months showed severe obstructive sleep apnea and moderate central apnoea.
Upon examination at 3 years and 10 months, she was found to have dysmorphic features including a large face with long palpebral fissures, prominent epicanthal folds, wide nasal bridge, a high and narrow palate and micrognathia. Growth failure was a significant concern with weight 10.3 kg (< 3%), height 82 cm (3-10%), head circumference 46.5 cm (6% -1SD) and Body Mass Index for age (10-25%). She had a well-balanced, varied diet and her nutrition labs were all normal, though she has chronic constipation.
Awake eye exam revealed normal eye structure, post-surgery strabismus, gaze induced nystagmus, and possible mild cortical visual impairment. Dilated sedated eye exam revealed bilateral hyperopia, optical coherence tomography showed normal macular architecture, however, the peripheral retina had mottled pigment changes. Further neurological examination demonstrated generalized low muscle tone, generalized intentional tremors, with truncal instability and truncal ataxia. Extended EEG showed asymmetric posterior dominant rhythm, vertex waves and spindles with right less than left suggestive of right sided focal dysfunction but no epileptiform discharges. The sensory and motor nerve testing did not reveal evidence of neuropathy or a primary muscle disorder. She had a normal hearing for speech and pure tones and normal auditory brainstem response bilaterally.
A developmental evaluation showed varying degrees of delay in all developmental areas. An MRI at this time showed marked atrophy of the cerebellum and pons with progression since her first MRI at 6 months of age. There was also delayed myelination in the cerebellum with gliotic white matter changes. A magnetic resonance spectroscopy demonstrated deficits of N-acetyl aspartate in the pons and superior cerebellar vermis.
Modelling the R1667P mutation. The R1667P mutation resides in the "R4" position in the voltagesensing S4 α-helix of Ca V 2.1 Repeat IV 6,10 (Fig. 1a). This R4 residue is highly-conserved amongst vertebrate Ca V channels including the closely-related N-type Ca V 2.2 channel 13 (Fig. 1b). Initially, Alphafold2 modelling was used to establish the orientation of R1667 within the Ca V 2.1 Repeat IV voltage-sensing module (Fig. 1c) 58 . In general, the AlphaFold2 model is consistent with the recent cryo-EM structure of Ca V 2.2 in which the Repeat IV voltage-sensor is shown in the "up" position 10 . The comparison with the Ca V 2.2 structure demonstrated that R1667 is positioned just extracellular to a highly-conserved phenylalanine (F1609) in the S2 helix in the open state. F1609 putatively constitutes the isoelectric border between the intra-and extra-cellular compartments of the Repeat IV voltage-sensing module by dividing the aqueous conduit formed by the S1-S3 helices around the S4 helix 59 (Fig. 1c). AlphaFold2 analysis also inferred that R1667 likely makes hydrogen bonds with N1579 in S1, T1606 in S2, and S1641 in S3 (Fig. 1d). To predict how introduction of a proline at position 1667 might affect www.nature.com/scientificreports/ these interactions, a homology model based on the Ca V 2.2 cryo-EM structure was generated using Phyre2. The arginine to proline substitution at position 1667 was introduced using the software Missense 3D (Fig. 1e). This second modelling strategy indicated that the proline: (1) causes a 65º bend in the S4 α-helix, and (2) lacks the positively-charged side chain which extends into the aqueous conduit. The former observation implies a steric   58 . The S1-S4 helices are viewed from the lateral aspect (c) and from an extracellular (d) vantage points. Potential hydrogen bonds between R1667 and N1579 in S1, T1606 in S2, and S1641 in S3 are indicated by the blue dashed lines. F1609 (i.e., the gating charge transfer center) is also labelled. Panels (c) and (d) were published with permission https:// creat iveco mmons. org/ licen ses/ by/4. 0/ (e) Missense 3D model showing the impact of the arginine to proline substitution at position 1667. The mutant stick structure (blue) is overlaid on the wild-type stick structure (white) with R1667 shown in green and the R1667P substitution shown in red. In both cases, F1609 is shown; this residue is colored green and red in the in the wild-type and mutant structures, respectively. www.nature.com/scientificreports/ "kink" in the voltage-sensor, while the latter suggests the loss of electrostatic contacts with N1579, T1606 and S1641.
R1667P reduces peak current density but shifts activation to more hyperpolarizing potentials. To investigate the impact of the R1667P mutation on the biophysical properties of Ca V 2.1, whole-cell Ca 2+ currents were recorded from tsA-201 cells co-expressing either GFP-fused wild-type Ca V 2.1 or GFP-fused Ca V 2.1 carrying the R1667P mutation (GFP-Ca V 2.1 and GFP-Ca V 2.1 R1667P, respectively) with auxiliary β 4 and α 2 δ-1 subunits. In this set of experiments, GFP-Ca V 2.1 and GFP-Ca V 2.1 R1667P both supported inward currents when near physiological 2 mM Ca 2+ was used as the charge carrier (Fig. 2a). A comparison of the Ca 2+ current-voltage (I-V) relationships for GFP-Ca V 2.1 and GFP-Ca V 2.1 R1667P revealed a considerable decrease in peak current density for the mutant channel (I peak = − 34.9 ± 2.8 pA/pF at + 10 mV, n = 17 vs. − 9.6 ± 1.2 pA/pF at 0 mV, n = 19, respectively; P = 5.5 × 10 -10 ) (Fig. 2b). Tail current amplitudes evoked by repolarization from + 70 to − 40 mV were substantially reduced in cells expressing GFP-Ca V 2.1 R1667P compared to cells expressing GFP-Ca V 2.1 (I tail = − 15.5 ± 1.8 pA/pF vs. − 67.8 ± 5.6 pA/pF, respectively; P = 2.1 × 10 -10 ) consistent with the idea that fewer mutant channels are present in the plasma membrane in our heterologous system. Notably, a ~ 10 mV hyperpolarizing shift in activation was evident in both the I-V and the normalized conductance-voltage (G-V) relationships (V G = 0.9 ± 0.7 mV vs. − 10.0 ± 0.9 mV for GFP-Ca v 2.1 and GFP-Ca V 2.1 R1667P, respectively; P = 1.4 × 10 -10 ) (Fig. 2c).

The R1667P mutation slows activation and deactivation kinetics. A closer examination of GFP-
Ca V 2.1 and GFP-Ca V 2.1 R1667P current families indicated that GFP-Ca V 2.1 R1667P activated much more slowly than GFP-Ca V 2.1 (τ act at + 20 mV = 0.7 ± 0.0 ms, n = 17 vs. 1.8 ± 0.1 ms, n = 15, respectively, P = 4.0 × 10 -15 ) ( Fig. 3a-c). Similarly, the R1667P mutation also impaired deactivation (Fig. 3d,e); analysis of tail currents evoked by repolarization steps from + 30 mV to a range of increasingly more positive potentials beginning at -50 mV  R1667P has little effect on closed-state inactivation. Five second conditioning steps were employed to assess channel closed-state inactivation (illustrated in Fig. 4a). The conditioning step was followed immediately by a 5 ms return to -80 mV before a 25 ms depolarizing test step to + 20 mV. This protocol yielded no However, the window current range was shifted to more hyperpolarizing potentials (Fig. 4e) as a consequence of the ~ 10 mV hyperpolarizing shift in activation for GFP-Ca V 2.1 R1667P (Fig. 2c).
R1667P reduces Ca 2+ flux in response to action potential-like stimuli. The R1667P substitution shifts activation and, consequently, the window current to more hyperpolarizing potentials (Figs. 2, 4). The mutation also retards deactivation (Fig. 3). These effects imply that the R1667P manifests in channel GOF under physiological circumstances. By contrast, the reduced current density and slower activation kinetics suggest that R1667P may result in overall channel LOF (Figs. 2, 3). To investigate these effects of the mutation in a somewhat more physiological context, we evoked Ca 2+ flux via an action potential-like voltage-clamp protocol similar to that used by Bahamonde et al. 54 . Robust Ca 2+ influx via GFP-Ca V 2.1 was observed using this protocol (Q = 12.0 ± 1.6 fC/pF; n = 13) (Fig. 5a) but the total integrated charge flux for GFP-Ca V 2.1 R1667P was found to be much less (3.4 ± 0.3 fC/pF, n = 12; P = 4.6 × 10 -5 ) (Fig. 5b).
The disparity in current density between cells expressing GFP-Ca V 2.1 and GFP-Ca V 2.1 R1667P (Fig. 2) raised the possibility that the observed reduction in Ca 2+ flux (Fig. 5c) was a consequence of fewer functional channels resident in the plasma membrane. To assess this possibility, the total integrated charge flux in each cell (i.e., Q) was normalized to maximal whole-cell conductance (i.e., the amplitude of the tail current upon repolarization from the reversal potential to − 40 mV) to reasonably estimate flux via individual channels. Per this metric, the Ca 2+ fluxes supported by individual GFP-Ca V 2.1 and GFP-Ca V 2.1 R1667P channels were not significantly different (Q/I tail = 0.17 ± 0.02 fC/pA, n = 13 vs. 0.25 ± 0.04 fC/pA, n = 12, respectively; P = 0.063) (Fig. 5d). However, flux could vary between the wild-type and mutant channels during repetitive use.

Discussion
In this study, we identified a second patient carrying the pathological CACNA1A R1667P variant and characterized the biophysical impact of the mutation on Ca V 2.1 channel function. In tsA-201 cells expressing GFP-fused Ca V 2.1 R1667P, we observed a substantial reduction in current density (Fig. 2) and multiple alterations in gating including a ~ 10 mV hyperpolarizing shift in activation (Fig. 2), slowed activation kinetics (Fig. 3a-c) and slowed deactivation kinetics (Fig. 3d-f). Intuitively, one would expect alterations in gating considering that R1667 is a basic residue in the "R4" position of one of the channel's four S4 voltage-sensing α-helices 13,60 . Our models showed that, in the putative open-state, R1667 is located just extracellular to the Repeat IV S2 phenylalanine (F1609) which represents the isoelectric point, or gating charge transfer center, of the channel 59 . The introduction of a proline at this position was predicted to partially disrupt the helical nature of the S4 helix (Fig. 1c,d) and to preclude key hydrogen bonds of R1667 with N1579, T1606, and S1641 in S1, S2 and S3, respectively (Fig. 1e). The ablation of these molecular contacts provides a plausible explanation for the hyperpolarizing shift in activation whereby stabilization of the Repeat IV S4 helix with the R4 position extracellular to the gating charge transfer center (i.e., F1609) favours a "primed" conformation. By the same token, we posit that the inability of R1667P to efficiently form hydrogen bonds with N1579, T1606 and/or S1641 slightly impedes voltage-sensor return and subsequent channel closure upon repolarization (Fig. 3d-f). On the surface, the increased activation of the current at more hyperpolarizing membrane potentials (Fig. 2) and slower activation kinetics (Fig. 3) of the R1667P mutant seem counterintuitive. A potential explanation for the incongruence is that introduction of the proline has two independent effects on gating. First, the "priming" of the Repeat IV voltage-sensor shifts activation to more hyperpolarizing potentials (please see above). Second, the proline-induced 65° twist in the S4 α-helix may impose a steric impediment to later, voltage-independent transitions critical for opening of the pore, as has been observed previously for a proline substitution at the R6 position in Repeat IV 61 . Thus, the R1667P mutation seems to precipitate both GOF and LOF in the same channel.
The hyperpolarized activation and slower rate of deactivation would together serve to increase Ca 2+ entry into synaptic terminals 54,62 , whereas the slower activation kinetics would almost certainly promote a reduction in neurotransmission 63 . However, the opposing effects of the mutation on channel gating appeared to effectively cancel (Fig. 5d), leaving the reduction in current density as the primary determinant in reducing total Ca 2+ flux during single action potential-like stimuli (Fig. 5c). If the GOF and LOF effects largely cancel in situ and the net effect of R1667P on unitary Ca 2+ flux is effectively minimal, the explanation that mutation simply impairs axon terminal Ca 2+ influx through haploinsufficiency becomes a tempting conclusion. That said, a more sophisticated experimental model (e.g., iPSCs, knock-in animals) is required to rigorously test the hypothesis that The complex presentation of the two patients carrying the R1667P mutation may be related to a perturbation of the delicate balance of GOF and LOF effects on the channel 57 . Two other patients, a mother and daughter, have been identified with an arginine to tryptophan substitution at the same position (R1667/8W) 39 . The mother presented with progressive cerebellar ataxia, ocular deficiencies and cerebellar atrophy. The daughter shared these characteristics but also displayed migraine and symptoms more consistent with EA2. Some of these attributes were common with the two non-related R1667P patients. However, neither the seizure symptoms nor cerebral edema associated with the first case of the R1667P mutation were present within the R1667/8W family suggesting that overall effects of the proline introduction are more disruptive of normal channel function than introduction of a tryptophan.
The generalization that GOF mutations in Ca V 2.1 are associated with migraine while LOF mutations are causative for episodic ataxia is usually correct 31-37 , though exceptions have been documented 30,64,65 . Moreover, the phenotypes of Ca V 2.1 channelopathies can be spectral and differ in degree of severity [38][39][40][41][42][43][44][45][46][47][48][49][50][51] . In some of these rapidly presenting cases, there is an almost stroke-like presentation [66][67][68] . For many of the subset of disorders which include these overlapping characteristics, both GOF and LOF effects on Ca V 2.1 function have been reported for the same mutation. Early on, Pietrobon and colleagues demonstrated that the primarily FHM1-linked mutations increased channel P o but also produced lower current amplitude owing to a lower number of functional channels in the membrane 52,53 . In particular, the widely-studied S218L variant displayed paradoxical increased voltage-sensitivity and decreased current density 69 . Likewise, a pathogenic Ca V 2.1 variant missing a phenylalanine at position 1502 had a substantial LOF (reduced peak current) which was overridden by strong GOF (~ 20 mV hyperpolarizing shift in activation and slowed deactivation) during action potential-mimicking stimuli 54 . More recently, Gandini et al. 56 demonstrated that the Y1384C mutation causes both channel GOF (increased window current) and LOF (reduced current density) in tsA-201 cells. Thus, our current study adds to a growing body of work that supports the idea that increased voltage-sensitivity and decreased current density are an electrophysiological signature of a subset of severe Ca V 2.1 channelopathies which feature both ataxia and migraine.

Compliance with ethics of experimentation. The current study was approved by Institutional Review
Board of The National Institutes of Health, National Human Genome Research Institute (15-HG-0130). Informed consent for publication was obtained and is on record. The study was performed in accordance with the principles of the Declaration of Helsinki. No animals were used in this study.
Health and safety. All laboratory procedures were approved by University of Maryland Baltimore Environmental Health and Safety (00005440).
Modeling. The AlphaFold2 model of human Ca V 2.1 can be accessed at the following web address: https:// alpha fold. ebi. ac. uk/ entry/ O00555 58 . Homology models of full length Ca V 2.1 α 1A subunits were generated using Phyre2 employing Ca V 2.2 α 1B as a template 70 . The sequence of Ca V 2.1 α 1A was downloaded from the universal protein resource (Uniprot; entry: O00555). The optimal template for homology modelling was identified using the BLASTp program 71 . The 3D structure of Ca V 2.2 α 1B (Uniprot ID: Q00975) was downloaded from PDB (PDB ID:7miy) and served as the template structure. The secondary structure of Ca V 2.1 α 1A was predicted using Phyre2. Missense mutation analysis was performed using Missense 3D 72,73 . Structures were visualized in PyMol.
Cell culture and transfection. tsA-201 cells (American Type Culture Collection) were cultured and transfected using Lipofectamine 2000 (Invitrogen) as described previously 61 . The transfection mixture contained expression plasmids encoding GFP-Ca V 2.1 or GFP-Ca V 2.1 R1667P with rat β 4 and rabbit α 2 δ-1 auxiliary subunits (1 µg of each cDNA per 35 mm well). The day following transfection, cells were trypsinized, re-plated onto 35 mm Primaria-treated plastic culture dishes (BD Falcon) and transferred to a 30 °C humified incubator. Successfully transfected (i.e., GFP-positive) cells were used in experiments ~ 24 h later.
Fluorescence imaging. Images of live tsA-201 cells were acquired using a Nikon W1 Spinning Disk microscope. Briefly, GFP was excited with a 488 nm line. The emitted GFP fluorescence was directed to the detector via a 500-550 nm band-pass filter. Similar detector gains were used for both sets of images.
Whole-cell patch clamp electrophysiology. All  www.nature.com/scientificreports/ contained (mM): 145 tetraethylammonium-Cl, 2 CaCl 2 , 10 HEPES, 10 glucose and pH 7.4 with tetraethylammonium-OH. Currents were recorded with an Axon 200B patch amplifier and digitized via a Digidata 1550 analogto-digital converter (both Molecular Devices). Electronic compensation was used to reduce the effective series resistance. Linear components of leak and capacitive currents were corrected with − P/4 online subtraction protocols. Filtering was at 5-10 kHz and digitization was at 50-100 kHz. Cell capacitance (C m ) was determined by integration of a transient from − 80 to − 70 mV using Clampex 10.6 (Molecular Devices) without compensation (23.0 ± 0.8 pF, n = 36) and was used to normalize current amplitudes (pA/pF). The time constant for decay of the whole-cell capacity transient (τ m ) was reduced as much as possible using the analog compensation circuit of the amplifier. The values of τ m and access resistance (R a ) were 108.8 ± 8.9 µs and 4.9 ± 0.2 MΩ respectively. Conductance-voltage (G-V) relationships were obtained from the I tail -V data where individual tail current amplitudes evoked from a given test potential were normalized by the maximal tail current amplitude produced by repolarization to -40 mV. Normalized I tail values were subsequently fit with the equation: where G is the tail current amplitude evoked by repolarization from a given test potential back to − 40 mV, G max is the maximal tail current amplitude obtained, V G is the half-maximal activation potential, and k is the slope factor. Activation and deactivation recordings were fit with a single exponential function: Normalized closed-state inactivation curves were fit by the equation: where I is the step current amplitude evoked by a 25 ms test depolarization to + 20 mV which was evoked following a 5 s conditioning step to potentials ranging from − 100 mV to + 30 mV and 5 ms return to − 80 mV, I 0 is the maximal step current amplitude obtained during the protocol, V 1/2inact is the half-inactivation potential, and k is the slope factor. Window currents were quantified via integration of the area under the conductance and inactivation curves.
Two-phase ramp waveforms intended to mimic neuronal action potentials were generated using a protocol similar to those used by Bahamonde and colleagues 54 . Cells were depolarized by a 1 ms ramp from − 80 to + 30 mV followed immediately by a 1 ms repolarization ramp to − 80 mV. The total charge flux was calculated as the integral of the area under the current evoked by the ramp protocol. Charge flux in each cell was then divided by the amplitude of the tail current following a repolarization from a 25 ms step to + 70 mV (i.e., near the reversal potential) back to − 40 mV to normalize Ca 2+ flux to channel expression.