Familial hemiplegic migraine type 1 mutations K1336E, W1684R, and V1696I alter Cav2.1 Ca2+ channel gating: evidence for beta-subunit isoform-specific effects.

Mutations in the Cav2.1 alpha1-subunit of P/Q-type Ca2+ channels cause human diseases, including familial hemiplegic migraine type-1 (FHM1). FHM1 mutations alter channel gating and enhanced channel activity at negative potentials appears to be a common pathogenetic mechanism. Different beta-subunit isoforms (primarily beta4 and beta3) participate in the formation of Cav2.1 channel complexes in mammalian brain. Here we investigated not only whether FHM1 mutations K1336E (KE), W1684R (WR), and V1696I (VI) can affect Cav2.1 channel function but focused on the important question whether mutation-induced changes on channel gating depend on the beta-subunit isoform. Mutants were co-expressed in Xenopus oocytes together with beta1, beta3, or beta4 and alpha2delta1 subunits, and channel function was analyzed using the two-electrode voltage-clamp technique. WR shifted the voltage dependence for steady-state inactivation of Ba2+ inward currents (IBa) to more negative voltages with all beta-subunits tested. In contrast, a similar shift was observed for KE only when expressed with beta3. All mutations promoted IBa decay during pulse trains only when expressed with beta1 or beta3 but not with beta4. Enhanced decay could be explained by delayed recovery from inactivation. KE accelerated IBa inactivation only when co-expressed with beta3, and VI slowed inactivation only with beta1 or beta3. KE and WR shifted channel activation of IBa to more negative voltages. As the beta-subunit composition of Cav2.1 channels varies in different brain regions, our data predict that the functional FHM1 phenotype also varies between different neurons or even within different neuronal compartments.

Disease-relevant structural defects in the gene encoding Ca v 2.1 ␣1-subunits have been described in mice and humans (6). In humans Ca v 2.1 ␣1 mutations cause e.g. familial hemiplegic migraine type-1 (FHM1), episodic ataxia type-2 (EA2), and inherited forms of epilepsy (7)(8)(9). FHM1 is a rare form of migraine with aura characterized by often unilateral obligatory motor aura symptoms (motor weakness, paralysis) (8). In some patients EA2, FHM1 (10), or epileptic symptoms (9) co-exist, suggesting that these allelic diseases form part of a broad disease spectrum. These mutations provide us with the opportunity to relate well defined structural changes of a single Ca 2ϩ channel subunit to (paroxysmal) neurological dysfunction and thereby also gain insight into the neurobiology of more common forms of migraine.
Heterologous expression of EA2 mutants demonstrated that EA2 is caused by complete or severe loss of mutated channel function (11)(12)(13). In contrast, a common functional feature of FHM1 mutations is a shift of the Ca v 2.1 activation curve to more hyperpolarized voltages (8,14). This implies the existence of a gain-of-function phenotype allowing Ca 2ϩ influx through mutant channels in response to small depolarizations that are insufficient to open wild-type (WT) channels and an increase of Ca 2ϩ influx through single mutant channels over a large voltage range (8). This has recently been confirmed in cerebellar neurons isolated from mice containing a FHM1 mutation (15). Increased action potential-evoked Ca 2ϩ influx and neurotransmitter release can explain the facilitation of cortical spreading depression (15) and phenomena of enhanced cortical network hyperexcitability associated with common forms of migraine (8).
The validity of this gain-of-function hypothesis is challenged by two important questions: (i) As Ca v 2.1 channels can associate with different ␤-subunits in mammalian brain (mainly ␤ 3 , ␤ 4 , and ␤ 1 (16 -18)) the functional effects of FHM1 mutations may vary depending on the associated ␤-subunit; (ii) so far only seven of at least sixteen known FHM1 mutations have been analyzed functionally. It remains unclear whether all FHM1 mutants can also induce the observed negative shift in activation gating.
Here we directly addressed these questions by studying the functional consequences of three FHM1 mutants, K1336E (KE), W1684R (WR), and V1696I (VI) (19), expressed in Xenopus laevis oocytes together with ␣ 2 ␦ 1 and different ␤-subunit isoforms (␤ 1 , ␤ 3 , or ␤ 4 ). We not only demonstrate that KE, WR, and VI cause significant changes in Ca v 2.1 gating, but our systematic analysis of ␤-subunits also revealed isoform-selective effects on all three FHM1 mutants. This suggests that the functional FHM1 phenotype varies between different neurons or even within different neuronal compartments.
Electrophysiological Recordings in X. laevis Oocytes-1-2 days after cRNA injection I Ba was measured at 19 -23°C using the two-microelectrode voltage clamp technique with the two-electrode voltage-clamp Turbo TEC 01C amplifier (NPI Electronics, Germany) (11). Data analysis and acquisition was performed by using the pClamp software package version 9.0 (Axon Instruments) after adjusting current traces by a conversion factor calculated from the difference between the leak at Ϫ80 and Ϫ90 mV, respectively. Microelectrodes were filled with 2.8 M CsCl, 0.2 M CsOH, 10 mM HEPES, and 10 mM EGTA (pH 7.4 with HCl), were pulled as described above and had pipette resistances between 0.3 and 1.1 M⍀. The extracellular solution contained 10 mM Ba(OH) 2 , 50 mM NaOH, 2 mM CsOH, and 5 mM HEPES (pH 7.4 with methanesulfonic acid).
Voltage-dependent inactivation during depolarization was estimated during 3-s pulses from a holding potential of Ϫ80 mV to a test potential 10 mV positive to the peak potential (V max ) of the I-V relations of the respective cell. Decrease of maximal I Ba during pulse trains was determined by applying 1-Hz trains of 15 100-ms pulses to the V max from a holding potential of Ϫ60 mV as described (22,23).
The voltage dependence of activation was determined from currentvoltage (I-V) curves at a holding potential of Ϫ80 mV which were fitted according to Equation 1, where V rev is the extrapolated reversal potential of I Ba , V is the membrane potential, I is the peak current, G max is the maximum conductance of the cell, V 0.5,act is the voltage for half-maximal activation, and k act is the slope factor of the Boltzmann term. Recovery of I Ba from inactivation was studied using a double pulse protocol. After a 3-s depolarizing prepulse from holding potentials of Ϫ80 mV to V max , the time course of I Ba recovery was determined at Ϫ60 mV by applying 300-ms test pulses to V max at various time intervals (between 0.05 and 20 s) after the prepulse. Peak I Ba was normalized to the peak current amplitude measured during the prepulse. The doublepulse protocol was repeated for each recovery time interval in the same

FIG. 2. Expression of WT and FHM1 mutants in tsA-201 cells.
Cells were transfected with equal amounts of Ca v 2.1 WT and KE, WR or VI mutants together with ␤ 3 -and ␣ 2 ␦ 1 -subunit cDNA and expressed ␣1-subunit protein was quantified by immunoblotting as described under "Materials and Methods." 15 g of membrane protein was loaded per lane and separated on a 7% SDS-polyacrylamide gel. Immunoreactivity was visualized using anti-Ca v 2.1␣1 1141-1156 antibody (11). No immunoreactivity was present in mock-transfected cells on the same gel (data not shown). Antibody staining intensity was quantified by digital image analysis of antibody stained Ca v 2.1 bands. When staining intensity was corrected for protein load (Coomassie Blue staining of immunoblots) and normalized to WT signal intensity, the following relative expression densities were obtained: KE, 1.07 Ϯ 0.21; WR, 1.21 Ϯ 0.13; VI, 0.95 Ϯ 0.10 (means Ϯ S.E., n ϭ 4; no statistically significant difference between mutants and WT).

FHM1 Mutations in Ca v 2.1 Ca 2ϩ Channels
oocyte. Between protocols oocytes were held at Ϫ100 mV for 1 min. The time course of recovery was fit to a biexponential decay yielding time constants for the fast ( fast ) and slow ( slow ) component and the contribution of the fast component (% fast ). The voltage dependence of inactivation (steady-state inactivation, holding potential of Ϫ80 mV) was estimated from normalized inward currents elicited during steps to V max after 10-s steps to various holding potentials (conditioning pulses) between Ϫ90 and ϩ30 mV. The half-maximal voltage (V 0.5,inact ) and the slope factor for steady-state inactivation (k inact ) were obtained by fitting the data to the following Boltzmann equation (Equation 2), where I SS is the non-inactivating current component.
Immunoblotting and Preparation of Affinity-purified Sequence-directed Antibodies-Experiments were carried out as described using a sequence directed antibody (anti-Ca v 2.1␣1 1141-1156 ) against a synthetic peptide corresponding to residues 1141-1156 of the rabbit Ca v 2.1 ␣1 sequence (GenBank TM accession number X57477 (11)). Antibody binding was visualized using the ECL-system (Pierce) employing a horseradish peroxidase labeled anti-rabbit antibody.
Statistics-All data are presented as mean Ϯ S.E. for the indicated number of experiments. Statistical significance was determined by oneway ANOVA followed by the Bonferroni post test using Sigma Plot 2001 (SPSS Inc.) or GraphPad Prism 4 (GraphPad software Inc.). Fig. 1 illustrates the positions of the mutations within the ␣1 subunit. KE converts a positive to a negative charge in the S3b-S4 linker of repeat III, which forms the putative voltagesensor paddle of the channel (26). WR introduces an additional positive charge in the S4 -S5 linker of repeat IV, which is believed to confer movements of the paddle to the pore structure (26). VI in helix IVS5 is the first FHM1 mutation in a S5 helix subjected to electrophysiological analysis.

RESULTS
We introduced these mutations into the human Ca v 2.1 ␣1subunit and expressed them together with ␣ 2 ␦ 1 and ␤ 1 , ␤ 3 , or ␤ 4 ␤-subunits in X. laevis oocytes. This expression system has proven to be very reliable to rapidly screen for FHM1-induced changes of whole cell Ca v 2.1 Ca 2ϩ channel currents using the two-electrode voltage clamp technique (11,22,23). ␤ 3 and ␤ 4 subunits were selected for co-expression, because Ca v 2.1 channels predominantly associate with these isoforms in mammalian brain (16,17). ␤ 1 -subunit containing channels are a minor component of the Ca v 2.1 channel population (16 -18) but were The half-maximal voltage for activation (V 0.5,act ) and the steepness of the curve at V 0.5,act (k act ) were obtained by fitting the data to the Boltzmann equation. Data are means Ϯ S.E. for the indicated number of experiments. Only currents with amplitudes between 0.2 and 1.4 A were included in the analysis. To rule out that differences in the activation parameters were due to voltage errors introduced by different current amplitudes, comparisons with WT were also performed with currents not differing more than 0.3 A. All significant differences persisted in this analysis (not shown). The statistically significant differences to WT are expressed with the same ␤-subunit determined by one-way ANOVA followed by the Bonferroni post test.   Tables I and II. A, I-V relationship of WT and mutant Ca v 2.1 ␣1-subunits co-expressed with ␣ 2 ␦ 1 and ␤ 4 (10 mM Ba 2ϩ as charge carrier). Currents were elicited by depolarizing pulses from a holding potential of Ϫ90 mV to test potentials shown between Ϫ40 mV and ϩ50 mV. For complete statistics of all mutants and ␤-subunits see Table I  All three mutant ␣1-subunits yielded robust I Ba 1-2 days after cRNA injection. In Western blots of tsA-201 cell-expressed wild-type (WT) and mutant ␣1-subunits (together with ␤ 3 and ␣ 2 ␦ 1 ), ␣1 immunoreactivities migrated with the expected molecular mass and showed indistinguishable protein expression densities (see also legend to Fig. 2).
Previous studies revealed that FHM1 mutations increase channel activity at negative membrane potentials, an effect that was also evident as a shift of the half-maximal activation voltage (V 0.5,act ) in six of seven mutations studied by us previously in Xenopus oocytes (22). A statistically significant negative shift of the half-maximal activation voltage (V 0.5,act ) was indeed found for KE and WR independent of the co-expressed ␤-subunit (Table I). For VI a small (Ϫ4 mV) negative shift was only seen upon ␤ 1 co-expression, but this trend did not reach statistical significance ( Fig. 3A; but see also legend to Table I). The differences in V 0.5,act could not be attributed to voltage errors due to different current amplitudes, because the statistically significant difference remained when subgroups of oocytes in each group with current amplitudes within a 0.3-A range (0.4 -0.7 A) were compared (not shown). The apparent reversal potentials for WT and all mutants ranged between 51.6 and 55.8 mV indicating that the permeability for Ba 2ϩ was not affected by the mutations.
Next we investigated the inactivation kinetics of the mutant channels in comparison to WT for all three ␤-subunits. The half-maximal voltage for steady-state inactivation (V 0.5,inact ) induced by 3-s conditioning prepulses to V max (holding potential of Ϫ80 mV) was significantly altered by the WR and KE mutations (Table II and Fig. 3B). Mutation WR induced a significant shift of V 0.5,inact to more negative potentials with all three ␤-subunits, which was most pronounced with ␤ 3 (Ϫ8.1

TABLE II Effects of FHM1 mutations on I Ba inactivation properties
The half maximal voltage for inactivation (V 0.5,inact ) and the steepness of the curve at V 0.5,inact (k inact ) were obtained by fitting the data to the Boltzmann equation. Data are means Ϯ S.E. for the indicated number of experiments. The statistically significant differences to WT are expressed with the same ␤-subunit determined by one-way ANOVA followed by the Bonferroni post test.

FIG. 4. Effects of FHM1 mutations on voltage-dependent inactivation during test pulses.
A and B, representative current traces illustrating voltagedependent inactivation of KE and VI mutants compared with WT co-expressed with ␤ 3 -and ␣ 2 ␦ 1 -subunits. I Ba was elicited by 3-s pulses from a holding potential of Ϫ80 mV to a depolarizing test potential 10 mV positive to V max . Traces were normalized to the peak current amplitude. C-E, for statistical analysis I Ba inactivation of normalized current traces of mutants expressed with ␣ 2 ␦ 1 and ␤ 1 (C), ␤ 3 (D), or ␤ 4 (E) was calculated for the indicated time points and expressed as means Ϯ S.E. 1, p Ͻ 0.05; 2, p Ͻ 0.01; 3, p Ͻ 0.001, statistically significant differences to WT determined by one-way ANOVA followed by the Bonferroni post test for n ϭ 9 -15. FHM1 Mutations in Ca v 2.1 Ca 2ϩ Channels mV). Interestingly, for mutation KE this occurred in a ␤-subunitdependent manner as a significant shift (Ϫ4.5 mV) was only observed with co-expressed ␤ 3 -subunits, but not with ␤ 1 and ␤ 4 . Mutation VI did not change V 0.5,inact .
To examine the possibility that other biophysical properties are also affected in a ␤-subunit isoform-specific manner, we determined the effects of the mutants on the kinetics of voltagedependent inactivation. We measured the decay of I Ba during 3-s depolarizations to voltages 10 mV positive to V max (holding potential Ϫ80 mV). Mutation WR did not affect inactivation kinetics with any of the co-expressed ␤-subunits (for statistics see Fig. 4). KE significantly accelerated inactivation when coexpressed with ␤ 3 (Fig. 4, A and D) but not with ␤ 4 or ␤ 1 (Fig.  4, C and E). Although VI did not change activation and inactivation voltage, it significantly slowed the inactivation time course with ␤ 1 and ␤ 3 (Fig. 4, B-D) but not with ␤ 4 (Fig. 4E). All three mutations also affected the recovery of Ca v 2.1 channels from inactivation. This was investigated using a double-pulse protocol (Fig. 5A) in which time-dependent recovery of I Ba was measured at increasing time intervals after a conditioning 3-s prepulse to V max (holding potential Ϫ80 mV). All three mutations significantly slowed the bi-exponential recovery from inactivation, but this effect was mainly observed upon co-expression with ␤ 1 and ␤ 3 (Fig. 5, C and D, and Table III). With ␤ 4 (Fig. 5E) a small but significant inhibition of the late phase of I Ba recovery (after 10, 15, and 20 s) was evident only for mutation WR but not for KE and VI. Non-linear fits of the bi-exponential recovery data (Table III) revealed that this slowed recovery could be explained by a corresponding increase of the time constants and/or a smaller contribution of the fast recovering component and/or an increase of the non-recovering component.
Changes of inactivation parameters should also affect the availability of Ca v 2.1 channels during trains of frequent pulses. Delayed recovery from inactivation should allow less I Ba to recover between pulses and promote accumulation of channels in inactivated states resulting in a cumulative decrease of I Ba during the train. Fig. 6 illustrates that WT maximal current amplitude decreased during 15 1-Hz trains of 100-ms depolarizations by 26.8% (␤ 1 ) to 32.5% (␤ 4 ). All three mutants significantly enhanced the current decrease with co-expressed ␤ 1 and ␤ 3 , but not with ␤ 4 . Our data demonstrate that all three FHM1 mutants can potentially reduce Ca 2ϩ influx during frequent depolarizations, at least if ␤ 1 and ␤ 3 form part of the channel complex. DISCUSSION The major novel finding of our study is that ␤-subunits critically determine the disturbances of channel gating found for all three FHM1 missense mutations, which we selected randomly from nine not yet functionally characterized ones. This implies that mutant phenotypes could vary between dif- ferent neurons expressing different ␤-subunit isoforms. Another important finding is that two mutations, KE and WR, allow Ca v 2.1 channel activation at more negative voltages. This is in accordance with the hypothesis that enhanced activity of Ca v 2.1 channels at negative voltages serves as a pathogenetic mechanism in FHM1 (8). Our study emphasizes that further analysis of FHM1 mutations on channel function requires the co-expression with those ␤-subunit isoforms most frequently associated Ca v 2.1 Ca 2ϩ channels in mammalian brain.
As shown in biochemical studies, brain Ca v 2.1 channel complexes predominantly contain either ␤ 3 (36%) or ␤ 4 (48%) and to a much smaller extent ␤ 1 (8.4%) and ␤ 2 (7.2%) subunits (16). Similar findings were obtained for brain Ca v 2.2 (27) and Ca v 1 (L-type) (17) channels. Analysis of the ␤-subunit composition of Ca v 1, Ca v 2.2, and Ca v 2.1 channels in defined brain regions revealed that equally large fractions of channels associate with ␤ 3 or ␤ 4 in cerebral cortex, whereas ␤ 4 -subunits predominate in channel complexes immunoprecipitated from the cerebellum (17). In the cerebellum, ␤ 4 and ␤ 3 mRNAs are both expressed in granule cells, whereas ␤ 4 predominates in the Purkinje cells (18). ␤ 4 , ␤ 3 , and ␤ 1 mRNA was detected in cerebral cortex (18). In the human hippocampus ␤ 1 -, ␤ 2 -, and ␤ 3 -subunit immunoreactivity is mainly localized to somata, whereas ␤ 4 staining is intense at dendritic locations (28,29). Based on the differential localization of these subunits and our findings FHM1 mutations may induce different patterns of Ca v 2.1 dysfunction in different neurons and even between different subcellular compartments of a single neuron. For example, in the presence of ␤ 3 -subunits, which appear to contribute more to presynaptic channels than ␤ 4 , at least in hippocampal neurons (29), all three mutations increase channel accumulation in inactivated states during frequent channel opening. This effect is mainly due to slowed recovery from voltage-dependent inactivation (Table III and Fig. 5). This is not seen with channels containing the ␤ 4 subunit, which appears to be the predominant form in dendritic compartments of hippocampal neurons (28,29). Ca v 2.1-current decrease during high neuronal firing rates may therefore be affected differently by the mutations at pre-and postsynaptic sites and/or in different neurons.
A still unresolved, yet important question concerning the FHM1 genotype-phenotype relationship is why about half of the known FHM1 mutations (black-filled symbols in Fig. 1 (19)) cause permanent cerebellar signs (PCS), such as ataxia, nystagmus, or dysarthria. PCS likely reflect permanent neuronal damage of cerebellar neurons, which may present clinically as cerebellar atrophy (30). Due to the predominant expression of ␤ 4 subunits in cerebellar neurons, it is possible that only mutation-induced abnormalities seen with ␤ 4 -subunits underlie PCS. From the mutations investigated here, only WR is associated with PCS (8). Interestingly, only WR caused a shift of V 0.5,inact to more negative potentials when co-expressed with ␤ 4 . This could decrease channel availability in cerebellar neurons, which may result in neuronal damage. Note that a loss of Ca v 2.1-channel function also occurs in EA2 (11)(12)(13), which leads to PCS and cerebellar atrophy of variable intensity (6). Our results therefore prompt further studies investigating the correlation of ␤ 4 -induced biophysical changes with PCS.
Our data support the hypothesis that FHM1 mutants can activate at more hyperpolarized potentials. This was unequivocally demonstrated for WR and KE. For VI a smaller shift was observed when co-expressed with ␤ 1 , which, however, did not reach statistical significance. Note that this does not rule the possibility of mutation-induced changes in channel activity: FHM1 mutation R192Q also caused only a minor shift of V 0.5,act in whole cell experiments after expression in Xenopus oocytes or tsA-201 cells, but an increased open probability over a broad voltage range was revealed in single-channel recordings (22,31). Further detailed single channel analysis, clearly beyond the scope of the present study, employing these three FHM1 mutants co-expressed with different ␤-subunit isoforms, must address this question. Moreover, we cannot rule out the possibility that other factors not accounted for in our heterologous expression system (such as the association of Ca v 2.1 channels with SNARE proteins (2)) also affect the functional consequences of FHM1 mutations. Thus it will be interesting to see whether the mutant phenotypes of Ca v 2.1 channel currents vary in cultured neurons isolated from different brain regions of FHM1 knock-in mice (15).
In this study we have not systematically addressed the question of mutant-induced changes in I Ba density and/or effects on ␣1-subunit plasma membrane targeting, because such data would be difficult to interpret based on the results of previous studies. Although heterologous expression nicely revealed changes in mutant channel gating that were confirmed in mutant mice in vivo (cf. Refs. 15 and 31), their role for determining effects on current density remain controversial, because increased (31), decreased (14), and unaltered (15) Ca v 2.1 current densities were for example found for R192Q in tsA-201 cells, transfected Ca v 2.1-deficient neurons, and mutant mice, respectively.
Taken together, our data clearly indicate that FHM1 mutation-induced changes of Ca v 2.1 channel function critically depends on the ␤-subunit composition of the channel. Obviously, our studies also prompt an even more detailed analysis of the expression of different Ca 2ϩ channel ␤-subunit isoforms in neuronal circuits relevant to migraine pathophysiology.