The Cacna1h mutation in the GAERS model of absence epilepsy enhances T-type Ca2+ currents by altering calnexin-dependent trafficking of Cav3.2 channels

Low-voltage-activated T-type calcium channels are essential contributors to the functioning of thalamocortical neurons by supporting burst-firing mode of action potentials. Enhanced T-type calcium conductance has been reported in the Genetic Absence Epilepsy Rat from Strasbourg (GAERS) and proposed to be causally related to the overall development of absence seizure activity. Here, we show that calnexin, an endoplasmic reticulum integral membrane protein, interacts with the III-IV linker region of the Cav3.2 channel to modulate the sorting of the channel to the cell surface. We demonstrate that the GAERS missense mutation located in the Cav3.2 III-IV linker alters the Cav3.2/calnexin interaction, resulting in an increased surface expression of the channel and a concomitant elevation in calcium influx. Our study reveals a novel mechanism that controls the expression of T-type channels, and provides a molecular explanation for the enhancement of T-type calcium conductance in GAERS.

by small organic molecules reduces thalamic burst firing and suppresses seizures in rodent models of absence, temporal lobe epilepsies 21,22 , and other types of seizures 23 . Third, T-type calcium channel inhibitors are effective in the treatment of absence seizures in humans 24 . Finally, T-type Ca 2+ conductances are elevated in thalamic neurons of several rodent models of absence epilepsy [25][26][27] , whereas mice lacking the Ca v 3.1 T-type channel exhibit increased resistance to absence seizures 28 . The causal link between a primary elevation of T-type Ca 2+ conductances and the development of absence epilepsy is further supported by the observation that genetic enhancement of Ca v 3.1 channel expression in mice is sufficient to induce an epileptic phenotype 29 .
The GAERS Cacna1h missense mutation results in a splice-variant specific gain-of-function of Ca v 3.2 currents that exhibit significantly faster recovery from channel inactivation and greater charge transference during high-frequency bursts 6 . However, consistent with the polygenic nature of generalized non-motor epilepsies the Cacna1h missense mutation does not by itself entirely account for absence seizure activity in GAERS. Other reported alterations in GAERS include elevated levels of thalamic Ca v 3.2 mRNA expression 30 and whole cell T-type currents 25 , although the genetic and molecular mechanisms by which upregulation of T-type channel activity might occur in GAERS and other rodent models of absence epilepsy remain unknown.
It is well described that calnexin, a type I endoplasmic reticulum integral membrane protein and molecular chaperone, is responsible for the folding, quality control and sorting of newly-synthetized (glyco)proteins 31 .
Although T-type channels undergo asparagine (N)-linked glycosylation [32][33][34][35] , the role of calnexin in the biogenesis and sorting of T-type channels is not understood. Here we show that calnexin binds to and modulates trafficking of Ca v 3.2 channels to the cell surface by altering the retention of the channel in the endoplasmic reticulum (ER). In addition, we show that calnexin-dependent regulation of Ca v 3.2 channels is disrupted by the GAERS mutation, thereby leading to an increased surface expression of T-type channels. These results thus reveal a mechanism for the enhanced T-type Ca 2+ conductance in GAERS, and provide new fundamental knowledge into the biogenesis and molecular trafficking of T-type channels.

Calnexin associates with Ca v 3.2 to modulate channel expression and function. To determine
whether T-type channels and calnexin associate at the protein level, we performed co-immunoprecipitation experiments of Ca v 3.2 with calnexin from wild type (WT) versus Ca v 3.2 knock out (KO) brains. As shown in Fig. 1A, a specific anti-calnexin antibody precipitated Ca v 3.2 from WT mouse brain homogenate, suggesting the existence of T-type channel/calnexin complexes in neuronal tissue. Note that the immunoprecipitated reactive species above 250 KDa that corresponds to the Ca v 3.2 channel is not present in the co-immunoprecipitation performed from Ca v 3.2 KO brain, thus demonstrating the specificity of the anti-Ca v 3.2 antibody used in these experiments. Furthermore, western blot analysis of Ca v 3.2 from brain homogenate before and after immunoprecipitation indicated that most of Ca v 3.2 channels are immunoprecipitated by calnexin (Fig. 1B). In contrast, Ca v 3.2 does not specifically interact with calreticulin and binding immunoglobulin protein (BiP), two other endoplasmic reticulum proteins known to interact with newly synthesized proteins (Fig. S1). The Ca v 3.2/calnexin interaction was also observed in tsA-201 cells expressing a recombinant HA-tagged human Ca v 3.2 channel (HA-hCa v 3.2) (Fig. 1C) where the channel is highly colocalized with calnexin (Fig. 1D). To investigate the impact of Ca v 3.2/calnexin interaction on channel function, we performed whole-cell patch-clamp recordings of T-type currents. The co-expression of calnexin with Ca v 3.2 channels in tsA-201 cells revealed a substantial decrease of T-type currents (Fig. 1E). For instance, in response to a depolarizing pulse to −20 mV, the mean peak Ba 2+ current density was decreased by 58% (p < 0.001) in calnexin-overexpressing cells (−17.3 ± 1.7 pA/pF, n = 42) as compared to control cells (−41.1 ± 3.4 pA/pF, n = 36) (Fig. 1F). The maximum slope conductance (Fig. 1F, inset) was reduced by 53% (p < 0.001) in calnexin-overexpressing cells (331 ± 28 pS/pF, n = 42) as compared to control cells (706 ± 55 pS/pF, n = 36). Similar results were obtained with hCa v 3.1 and hCa v 3.3 channels (Fig. S2). Immunoblot analysis of HA-hCa v 3.2 revealed that the decreased T-type conductance in calnexin-overexpressing cells was not accompanied by a diminution of the total expression levels of the channel (Fig. 1G). In contrast, nonstationary noise analysis of T-type currents revealed a significant decrease by 76% (p < 0.001) of the number of functional channels expressed at the cell surface (N) in calnexin-overexpressing cells (74.8 ± 37.6 channel/ pF, n = 11) as compared to control cells (311.8 ± 38.6 channel/pF, n = 12) (Fig. S3A-C). However, the unitary conductance (γ) of hCa v 3.2 channels was enhanced ~2.5 fold (p < 0.01) in cells overexpressing calnexin (4.0 ± 0.5 pS, n = 12) as compared to control cells (1.6 ± 0.2 pS, n = 11), without additional effects on the opening probability (Po) of the channel (Fig. S3C). It is worth noting that the unitary channel conductance was obtained from the noise analysis of whole cell current traces recorded in the presence of 5 mM Ba 2+ and is thus lower than the unitary conductance obtained from direct single channel recordings that are usually performed using a high concentration of charge carrier (routinely 110 mM Ca 2+ or Ba 2+ ). The unitary conductance of 1.6 pS in the absence of overexpressed calnexin is consistent with previous studies using similar recording conditions and analysis 36 . In addition, we observed small alterations in some of the biophysical properties of the channel. For instance, activation and inactivation kinetics of Ca v 3.2 at −20 mV were slightly slower (p = 0.0295 and p = 0.0066, respectively) in calnexin-expressing cells (τ activ = 5.5 ± 0.6 ms; τ inac 22.5 ± 1.2 ms, n = 11) as compared to control cells (τ activ = 4.0 ± 0.3 ms; τ inac = 18.3 ± 0.8 ms, n = 12) and we observed a 5 mV depolarizing shift (p < 0.0001) of the voltage-dependence of activation in calnexin-expressing cells (−35.2 ± 0.6 mV, n = 11) as compared to control cells (−40.0 ± 0.6 mV, n = 12).
Altogether, these data support the notion that calnexin participates in the maturation of the hCa v 3.2 channel and controls its expression at the cell surface.

Molecular determinants of calnexin interactions with hCa v 3.2 channels.
To determine the molecular determinants of calnexin/Ca v 3.2 interaction, we developed mCherry-tagged full-length calnexin (CNX AD ) and deletion mutants of calnexin corresponding to the luminal (CNX AC ) and cytosolic regions of the protein (CNX BD ) ( Fig. 2A). The transmembrane domain of calnexin was conserved in all constructs in order to preserve ER localization. Confocal images of live tsA-201 cells expressing GFP-tagged calreticulin together with mCherry-tagged calnexin-deletion mutants confirmed that the fusion proteins are effectively expressed in the ER compartment (Fig. 2B). A similar colocalization of mCherry-tagged CNX constructs was also observed with GFP-tagged BiP and with an ER-targeted GFP construct (data not shown). We then investigated by co-immunoprecipitation which region of calnexin is required for the interaction with hCa v 3.2 channels. As shown in Fig. 2C, the mCherry-tagged luminal (CNX AC ) and cytosolic regions of calnexin (CNX BD ) both immunoprecipitated HA-hCa v 3.2 from tsA-201 cells. Conversely, HA-tagged hCa v 3.2 immunoprecipitated mCherry-tagged CNX AC and CNX BD (Fig. 2D). To examine the impact of Ca v 3.2/calnexin interactions on the trafficking of T-type channels at the plasma membrane, we performed surface immunostaining of HA-hCa v 3.2 expressed in tsA-201 cells along with calnexin-deletion mutants. Confocal images of surface HA-hCa v 3.2 are  shown in Fig. 3A. Surface expression of HA-hCa v 3.2 was reduced by 48% (p < 0.05) and 47% (p < 0.05) in cells expressing CNX AC and CNX BD , respectively (Fig. 3B). In contrast, surface expression of the membrane-targeted form of GFP (Lck-GFP), which does not interact with calnexin, was unaffected by the co-expression of calnexin, indicating that the effect of calnexin on surface expression of the channel is not mediated by unspecific saturation of the cellular machinery (Fig. S4). Consistent with a decreased surface expression of hCa v 3.2 channels in the presence of calnexin constructs, T-type currents were significantly reduced (Fig. 3C). For instance, the mean peak Ba 2+ current recorded in response to a depolarizing pulse to −20 mV was decreased by 49% (p < 0.001) and 38% (p < 0.001) in cells expressing CNX AC (−21.1 ± 1.9 pA/pF, n = 41) and CNX BD (−25.7 ± 2.3 pA/pF, n = 42), respectively, as compared to control cells (−41.2 ± 2.0 pA/pF, n = 95) (Fig. 3D). The maximum slope conductance was reduced by 43% (p < 0.001) and 31% (p < 0.001) in cells expressing CNX AC (415 ± 35 pS/pF, n = 41) and CNX BD (489 ± 39 pS/pF, n = 42), respectively, as compared to control cells (732 ± 34 pS/pF, n = 95) (Fig. 3D). In contrast, expression of an ER-targeted construct that does not have biological function (ER-dsRed) had no significant effect on T-type currents, indicating that the decreased conductance observed in the presence of calnexin-deletion mutants is not mediated by a non-specific stress of the ER that could have resulted from overexpression, but rather from the specific binding of calnexin regions with hCa v 3.2 and the retention of the channel in the ER. This notion is further supported by our observation that uncoupling of hCa v 3.2 channels from Note that co-expression of the ER-dsRed as no significant effect on hCa v 3.2-HA currents (dark red bar, n = 26). Data are presented as mean ± SEM and were analyzed by Student's unpaired t test; NS not significant, ***p < 0.001. calnexin by co-expression of the mCherry-calnexin C-tail (CNX CD ) fusion protein produced a robust increase of T-type currents (Fig. 4A,B). For instance, in response to a depolarizing pulse to −20 mV, the mean peak Ba 2+ current density was increased by 76% (p < 0.001) in cells expressing CNX CD (−58.4 ± 3.9 pA/pF, n = 48) as compared to control cells (−33.2 ± 2.4 pA/pF, n = 47) (Fig. 4C). The maximum slope conductance (Fig. 4C, inset) was increased by 70% (p < 0.001) in CNX CD -expressing cells (1028 ± 61 pS/pF, n = 48) as compared to control cells (607 ± 41 pS/pF, n = 47). Consistent with the notion that the increased T-type conductance in the presence of the calnexin C-tail arose from the uncoupling of Ca v 3.2 from calnexin and thus increased trafficking of the channel to the plasma membrane, the surface expression of HA-tagged Ca v 3.2 channels assessed by immunostaining was increased by 30% (p < 0.001) in cells expressing the calnexin C-tail ( Fig. 4D-E). In contrast, calnexin C-tail had no effect on the total expression of the channel protein. Because the elevated T-type conductance in the presence of calnexin C-tail could also have resulted from a decrease of channel internalization rather than from an increase of the trafficking of the channel to the plasma membrane 37 , we assessed internalization kinetics of HA-hCa v 3.2 channels. The time constant of internalization of HA-hCa v 3.2 channels at 37 °C was unaffected in cells expressing calnexin C-tail (75.3 ± 2.2 min, n = 3) as compared to control cells (80.8 ± 3.3 min, n = 3) (Fig. 4F).
Collectively, these results indicate that various regions of calnexin are engaged in the interaction with hCa v 3.2 channels. Of particular interest is the demonstration that interactions with the cytosolic calnexin C-tail are involved in the retention of hCa v 3.2 in the ER, thus restricting surface expression of the channel. The calnexin C-tail associates with the III-IV linker region of hCa v 3.2 channels. To identify the Ca v 3.2 molecular determinants of calnexin interactions, we assessed the ability of different intracellular regions of hCa v 3.2 to interact with calnexin C-tail. The main intracellular regions of hCa v 3.2 channels fused to the GFP (Fig. 5A) were expressed in tsA-201 cells together with the mCherry-tagged calnexin C-tail (mCherry-CNX BD ). As shown in Fig. 5B, the calnexin C-tail was specifically immunoprecipitated with the III-IV linker region of hCa v 3.2. A similar observation was made with the calnexin C-tail deleted of the ER transmembrane domain (mCherry-CNX CD ) indicating that the transmembrane domain is not required for this interaction (Fig. S5). Interestingly, the GAERS mutation correlating with absence seizures in GAERS is located within exon 24 of Cacna1h that encodes the proximal region of the III-IV linker. This missense mutation produces an arginine to proline switch (R1584P) that could potentially affect the electrostatic potential and/or tertiary structure of the III-IV linker and alter protein-protein interactions. To investigate whether the binding of calnexin C-tail is altered by the GAERS mutation, we introduced the homologous mutation (R1573P) into the human III-IV linker. In addition, we generated a III-IV linker construct containing a human mutation (T1606M) previously identified in patients with generalized non-motor epilepsies 2° (Fig. 5C, inset). Co-immunoprecipitations from tsA-201 cells expressing the GFP-tagged III-IV variants together with mCherry-tagged calnexin C-tail (mCherry-CNX BD ) revealed a substantially decreased interaction of the calnexin C-tail with the III-IV variant containing the GAERS mutation (Fig. 5C). Indeed, the ability of III-IV R1573P to immunoprecipitate calnexin C-tail was decreased by 70% (p < 0.001) as compared to the wild-type III-IV linker (Fig. 5D). In contrast, the T1606M mutation did not have a significant effect, although there did appear to be a trend towards a weakening interaction. To further support the notion that Ca v 3.2/calnexin interaction is altered by the GAERS mutation, we performed co-immunoprecipitation experiments of Ca v 3.2 with calnexin from GAERS brain compared to its non-epileptic control (NEC) strain. Consistent with co-immunoprecipitation studies using calnexin C-tail and Ca v 3.2 III-IV linker, co-immunoprecipitations from GAERS brain homogenates revealed a substantial decrease of the interaction of Ca v 3.2 with calnexin ( Fig. 5E) in GAERS brain by 42% (p < 0.01) compared to NEC brain (Fig. 5F).
Altogether, these data identified the III-IV linker region of Ca v 3.2 as the channel region engaged in the interaction with calnexin C-tail, and indicate that this interaction is disrupted by the GAERS Ca v 3.2 missense mutation.

The GAERS mutation alters surface expression of Ca v 3.2 channels by modulating calnexin interactions.
It is possible that the decrease in calnexin C-tail binding to the human Ca v 3.2 III-IV linker region carrying the equivalent GAERS mutation may be channel splice variant specific. We thus examined the effect of the GAERS mutation on the III-IV linker of the two major Ca v 3.2 channel splice variants expressed in the rat thalamus, Ca v 3.2 (+25) and Ca v 3.2 (−25), which differ in the presence or absence of the exon 25. This exon encodes a short stretch of six amino acid residues and is located 14 residues downstream the GAERS mutation (Fig. 6A). Consistent with our previous observation, the binding of calnexin C-tail to the rIII-IV linker (+25) containing the GAERS mutation was significantly decreased by 82% (p < 0.001) compared to the wild-type rIII-IV (+25) (Fig. 6B and C). In addition, the binding of calnexin C-tail to the short wild-type rIII-IV linker variant (−25) was also significantly decreased by 75% (p < 0.001). Introducing the GAERS mutation in the rIII-IV linker (−25) variant produced an additional 69% decrease (p < 0.01) in binding to the calnexin C-tail compared to wild-type rIII-IV (+25).
To establish whether the decrease in the calnexin C-tail interaction with the rCa v 3.2 III-IV linker also alters surface expression of full-length rCa v 3.2 channels, we examined T-type currents in cells expressing various channel variants together with overexpressed full-length calnexin. T-type currents recorded from tsA-201 cells expressing the rCa v 3.2 (+25) channel isoform were significantly increased when the GAERS mutation was introduced ( Fig. 6D and E). The maximal conductance was increased by 67% (p < 0.01) in rCa v 3.2 GAERS (+25)-expressing cells (283 ± 30 pS/pF, n = 21) as compared to cells expressing the wild-type rCa v 3.2 (+25) channel (169 ± 20 pS/ pF, n = 20) (Fig. 6F). In addition, and consistent with our observation that binding of calnexin C-tail is enhanced in the presence of exon 25, T-type currents recorded from cells expressing the rCa v 3.2 (−25) channel variant were elevated ( Fig. 6D and E). For instance, the maximal conductance was increased by 60% (p < 0.01) in rCa v 3.2 (−25)-expressing cells (270 ± 20 pS/pF, n = 23), and by 133% (p < 0.01) in rCa v 3.2 GAERS (−25)-expressing cells (393 ± 65 pS/pF, n = 21) (Fig. 6F). Remarkably, the electrophysiological data in Fig. 6F appear to be inversely correlated with the biochemical measurements in Fig. 6C, suggesting the possibility that the differences in cell surface expression of the various constructs is mediated by their differential interactions with calnexin. To test this hypothesis, we co-expressed the different rCa v 3.2 channel variants with the calnexin C-tail peptide to disrupt calnexin regulation of Ca v 3.2 channels. Preventing calnexin interactions with the channel not only eliminated the differences in current densities observed with the two exon 25 splice isoforms, but importantly also ablated the effects of the GAERS mutation. (Fig. 6G-I). Altogether, these data indicate that calnexin mediated ER retention of Ca v 3.2 channels is a key determinant of the differential cell surface expression of Ca v 3.2 III-IV linker splice isoforms and are essential for the functional effects of the GAERS mutation.

Discussion
Here, we describe a new regulatory pathway for low-voltage-activated T-type Ca 2+ channels, and provide insights into molecular basis that may underlie the enhancement of T-type Ca 2+ conductance of thalamic neurons in the Genetic Absence Epilepsy Rat from Strasbourg (GAERS) and other rodent models of absence epilepsy. Our data show that the III-IV linker region of Ca v 3.2 channels interacts with calnexin to increase ER retention of the channel. Of particular importance, our data revealed that the GAERS mutation critically alters the interaction of calnexin with Ca v 3.2, resulting in a significant increase in surface expression of the channel, which may account for the elevated T-type Ca 2+ conductance in GAERS reticular thalamic neurons.
Using a combination of biochemical and electrophysiological approaches, we revealed that binding of calnexin to Ca v 3.2 has two important consequences. First, it increases the retention of the channel in the ER, resulting in  7). The CNX BD -mCherry signal was quantified by densitometry using ImageJ software and normalized to the expression level of the corresponding III-IV-GFP used to immunoprecipitate calnexin. (E) Co-immunoprecipitation of Ca v 3.2 from NEC versus GAERS brain homogenates with specific anti-calnexin antibody. Corresponding normalized mean binding values of Ca v 3.2 on calnexin in NEC (black bar) and GAERS brains (red bar). The Ca v 3.2 signal was normalized to the corresponding immunoprecipitated calnexin signal. Data are presented as mean ± SEM and were normalized to total C-tail expression levels and analyzed by one-way ANOVA with Tukey's post tests; NS not significant, **p < 0.01, ***p < 0.001. decreased expression of T-type channels at the cell surface without alteration of the total protein levels. Second, it participates in the maturation/folding of Ca v 3.2 channels as evidenced by an increase of the single channel conductance in the presence of overexpressed calnexin. The increased channel conductance may result from an improved folding of the channel consistent with the chaperone role of calnexin in the maturation of newly synthetized proteins. It is also conceivable that increased ER retention of Ca v 3.2 alters the glycosylation status of the channel, which in turn may affect ion permeability. Consistent with this idea, we have recently reported that N-linked glycosylation of hCa v 3.2 at specific loci influences channel permeability 38 . Nonetheless, the net effect of calnexin interactions with the channel is a decrease of Ca v 3.2 surface expression, consistent with a reduction of the whole cell Ca 2+ conductance. These results support the existence of a Ca v 3.2/calnexin signaling complex that is essential for the maturation and sorting of T-type channels to the cell surface, similar to that for Na v 1.8 channels, CFTR channels, Shaker potassium channels, and other G-protein coupled receptors including acetycholine and dopamine D1 and D2 receptors [39][40][41][42][43][44] . Detailed analysis of the molecular determinants of the Ca v 3.2/calnexin interaction revealed that Ca v 3.2 interacts with two distinct calnexin regions. In part, Ca v 3.2 interacts with the ER luminal domain of calnexin. It is well known that this domain is responsible for lectin-like activity of calnexin and glycosylation-dependent interaction with newly synthetized glycoproteins 45 . Although this interaction has not been investigated in more detail in the current study, it likely involves one or more extracellular regions of the channel (i.e. facing the lumen side when the channel is embedded in the membrane of the ER) located in domains I, II and IV that are known to be N-glycosylated 32,33,35 . In another part, Ca v 3.2 interacts with the cytosolic region of calnexin (calnexin C-tail) via its domain III-IV cytosolic linker. We showed that this interaction is sufficient to mediate the retention of Ca v 3.2 in the ER and to decrease the sorting of channels to the cell surface. Conversely, uncoupling calnexin from its binding site on the channel via coexpression of the calnexin C-tail interfering peptides resulted in a significant increase in the surface expression of the channel. While little is known about the cellular functions of the calnexin C-tail, a role in the control of clathrin-mediated endocytosis of membrane proteins has been proposed 37 . However, our observation that internalization properties of Ca v 3.2 channels were not affected by the presence of the calnexin C-tail strongly supports the notion that the increased channel surface expression is the direct consequence of a decreased ER retention of the channel upon uncoupling of Ca v 3.2 from calnexin.
The steady-state expression of T-type channels at the cell surface is governed by a balance between the number of channels arriving at and being removed from the plasma membrane. Hence, our observation that calnexin affects the sorting of T-type channels to the cell surface may have important consequences on neuronal excitability. Indeed, we have revealed a previously unrecognized effect of the GAERS Ca v 3.2 missense mutation on the trafficking of Ca v 3.2 channels. Of particular interest is the observation that the GAERS mutation alters the binding of calnexin C-tail to the III-IV linker region of Ca v 3.2, resulting in a potent increase of the expression of channels at the cell surface. The observation that the GAERS mutation enhances Ca v 3.2 surface expression by altering calnexin-dependent ER retention of the channel provides a further explanation to the hyperexcitability observed in a subset of GAERS thalamic neurons and believed to underlie absence seizures 25 . A modest increase of Ca v 3.2 mRNA levels has also been reported in GAERS, but whether this correlates with an increased expression levels of the channel and/or is a result of further underlying genetic alterations remains to be investigated 30 .
It is established that alternative splicing represents an important mechanism that controls the expression and regulation of ion channels including T-type channels [46][47][48][49][50][51] . Ca v 3.2 splice variants with distinct biophysical properties have previously been identified in the rat thalamus and that differ by the presence or absence of exon 25 encoding for a small region of the III-IV linker region 6 . Our study revealed that binding of calnexin C-tail on the III-IV linker lacking the exon 25 is significantly decreased, which correlated with an increased expression of the Ca v 3.2 (−25) channel variant at the cell surface. This observation further supports the notion that Ca v 3.2/calnexin interaction plays an essential role in the control of surface trafficking of the T-type channel. It also uncovers a new mechanism by which alternative splicing of Ca v 3.2 can affect channel surface expression. In addition, the surface expression of the Ca v 3.2 (−25) variant was further augmented in the presence of the GAERS Ca v 3.2 missense mutation. Of note, the observed splice-variant specific effects of calnexin interactions on Ca v 3.2 current conductance are quantitatively analogous to that for the previously described splice-variant effects of the GAERS mutation on Ca v 3.2 recovery from inactivation 6 . Together, these data suggest that GAERS neuronal hyperexcitability and seizure activity may result from a combination of gain-of-functions derived from domain III-IV linker splice variant Ca v 3.2/calnexin interactions and splice variant specific effects of the GAERS Ca v 3.2 missense mutation on recovery from inactivation. It remains to be determined whether the GAERS genome possess any further underlying genetic alterations associated with the Ca v 3.2/calnexin interactions and that could further affect T-type calcium channel functioning or that of other ion channels.
Alteration of T-type channel expression has been linked to various neurological disorders including absence epilepsy. Interestingly, the III-IV linker region of Ca v 3.2 channels has previously been implicated in the control of channel activation and recovery from inactivation 6,49,50 , as well as in the control of channel expression by ubiquitination 52 . Here, we present evidence that the III-IV linker also mediates calnexin-dependent control of channel surface expression. Hence, the III-IV linker region of Ca v 3.2 channels, albeit relatively short compared to the other intracellular region of the channel, appears to be an essential hub for regulating key aspects of calcium channel trafficking and biophysical properties. The physiological importance of this regulatory hotspot is underscored by its role in chronic pain 52 and as we show here, seizure disorders.

Materials and Methods
Plasmids cDNA constructs. The human wild-type HA-Ca v 3.2 construct (HA-hCa v 3.2 WT ) was previously described 53 . This plasmid was used as a template to amplify by PCR the main cytoplasmic regions of the channel (amino-and carboxy-terminal regions, and I-II, II-III, and III-IV linkers), and the PCR products were inserted into the XhoI/HindIII sites of pEGFP-C1 vector. The GFP-tagged III-IV linker construct was used as a template to introduce by site-directed mutagenesis the R1573P and T1606M mutations using overlap extension PCR method, and the PCR products were inserted into the XhoI/HindIII sites of pEGFP-C1 vector. The rat wildtype and GAERS Ca v 3.2 ± exon 25 variants in pcDNA3.1 were previously described 6 . Sequences encoding the various rat III-IV linkers used in this study were synthetized in vitro (GenScript) and subcloned into the XhoI/ HindIII sites of pEGFP-C1 vector. To generate mCherry-tagged calnexin constructs, plasmids encoding for the full-length mouse calnexin (CNX AD ) and calnexin fragments (CNX AC and CNX BD ) (a generous gift of Dr. Marek Michalak) were used as template for PCR. PCR products encoding for the different domains of calnexin and mCherry were inserted into KpnI/AgeI sites and AgeI/PmeI sites of pcDNA3.1(+) vector, respectively, to generate C-terminal mCherry-tagged calnexin constructs. The PCR primers used to generate those constructed are shown in Supplementary Information. All final constructs were verified by sequencing of the full-length cDNAs. The Lck-GFP construct encoding for a membrane-targeted form of GFP was previously described 54 .
Patch-clamp electrophysiology. Patch-clamp recordings were performed 72 h after transfection in the whole-cell configuration of the patch-clamp technique at room temperature (22-24 °C) as previously described 55 . Briefly, the bath solution contained (in millimolar): 5 BaCl 2 , 5 KCl, 1 MgCl 2 , 128 NaCl, 10 TEA-Cl, 10 D-glucose, 10 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) (pH7.2 with NaOH). Patch pipettes had a resistance of 2-4 MΩ when filled with a solution containing (in millimolar): 110 CsCl, 3 Mg-ATP, 0.5 Na-GTP, 2.5 MgCl 2 , 5 D-glucose, 10 EGTA, and 10 HEPES (pH7.4 with CsOH). Whole-cell patch-clamp recordings were performed using an Axopatch 200B amplifier (Axon Instruments). Acquisition and analysis were performed using pClamp 9 and Clampfit 9 software, respectively (Axon Instruments). All traces were corrected online for leak currents, digitized at 10 kHz, and filtered at 2 kHz. The voltage dependence of the peak Ba 2+ current density was fitted with the following modified Boltzmann equation (1): with I(V) being the peak current amplitude at the command potential V, G max the maximum conductance, V rev the reversal potential, V 0.5 the half-activation potential, and k the slope factor. The voltage dependence of the whole-cell Ba 2+ conductance was calculated using the following modified Boltzmann equation (2): with G(V) being the Ba 2+ conductance at the command potential V.

SDS-PAGE and immunoblot analysis.
Immunoprecipitation samples, or total lysates (25 μg), were separated on 12% SDS-PAGE and transferred onto PVDF membrane (Millipore). For detection of the HA-hCa v 3.2 channel, the membrane was incubated with a primary rat monoclonal anti-HA antibody (Roche) diluted at 1:1000; GFP-tagged hCa v 3.2 loops were detected with a primary rat monoclonal anti-GFP antibody (Abcam) diluted at 1:10.000; mCherry-tagged calnexin constructed were detected with a primary mouse monoclonal anti-mCherry antibody (Abcam) diluted at 1:1:1000; native calnexin was detected using a primary rabbit polyclonal anti-calnexin antibody (Abcam) diluted at 1:5000. For detection of native proteins, the following antibodies were used and incubated overnight at 4 °C with the membrane: rabbit polyclonal anti-Ca v 3.2 antibody (Santa Cruz) diluted at 1:5000; mouse monoclonal anti-BiP antibody diluted at 1:1000 (Santa Cruz); mouse monoclonal anti-Calreticulin antibody (Santa Cruz) diluted at 1:1000; and rabbit polyclonal anti-calnexin antibody (Abcam) diluted at 1:10.000, incubated overnight at 4 °C. Membranes were then washed in PBS/Tween-20 buffer, and incubated with the corresponding secondary HRP-conjugated antibody (Jackson ImmunoResearch) diluted at 1:20.000. For immunoprecipitation controls, membranes were stripped to remove antibodies in a stripping buffer (0.2 M glycine, 1% SDS; pH 2.0) and then reblotted as described above. Immunoreactive bands were detected by enhanced chemiluminescence.
Surface immunostaining. Twenty-four hours before the experiment, cells expressing HA-Ca v 3.2 channels were seeded on poly-L-lysine-coated glass coverslips. Cells were incubated for 30 min at 37 °C with a primary monoclonal mouse anti-HA antibody (Abcam) diluted in DMEM at 1:1000, washed with PBS, fixed for 7 min in 4% paraformaldehyde, and blocked for 45 min in blocking buffer (5% FBS in PBS). Cells were then incubated for 1 h at room temperature with a secondary goat polyclonal anti-mouse Alexa488-conjugated antibody (Jackson ImmunoResearch) diluted in blocking buffer at 1:1000, washed, and mounted on microscope glass slides with ProLong Gold mounting medium (Life Technologies). Confocal images were acquired with a Zeiss LSM780 microscope and the field fluorescence intensity was analyzed using ImageJ software.
For internalization studies, cells were incubated with a primary anti-HA antibody as described above, washed, kept at 37 °C for 30, 60, 120, or 180 min to allow internalization of the channel, fixed, and stained with a secondary Alexa488-conjugated antibody to assess the time-dependence of surface expression of the channel. In order to visualize internalized channels, cells were permeabilized with 0.2% Triton X-100 for 10 min before incubation with the secondary antibody.
Statistical analysis. Data values are presented as mean ± S.E.M. for n experiments. Statistical significance was determined using Student's unpaired t test or one-way ANOVA with Tukey's post tests as indicated in the figure legends: *p < 0.05, **p < 0.01, ***p < 0.001, and NS, statistically not different.