The Role of Distal S6 Hydrophobic Residues in the Voltage-dependent Gating of CaV2.3 Channels*

The hydrophobic locus VAVIM is conserved in the S6 transmembrane segment of domain IV (IVS6) in CaV1 and CaV2 families. Herein we show that glycine substitution of the VAVIM motif in CaV2.3 produced whole cell currents with inactivation kinetics that were either slower (A1719G ≈ V1720G), similar (V1718G), or faster (I1721G ≈ M1722G) than the wild-type channel. The fast kinetics of I1721G were observed with a ≈+10 mV shift in its voltage dependence of activation (E0.5,act). In contrast, the slow kinetics of A1719G and V1720G were accompanied by a significant shift of ≈-20 mV in their E0.5,act indicating that the relative stability of the channel closed state was decreased in these mutants. Glycine scan performed with Val 349 in IS6, Ile701 in IIS6, and Leu1420 in IIIS6 at positions predicted to face Val1720 in IVS6 also produced slow inactivating currents with hyperpolarizing shifts in the activation and inactivation potentials, again pointing out a decrease in the stability of the channel closed state. Mutations to other hydrophobic residues at these positions nearly restored the channel gating. Altogether these data indicate that residues at positions equivalent to 1720 exert a critical control upon the relative stability of the channel closed and open states and more specifically, that hydrophobic residues at these positions promote the channel closed state. We discuss a three-dimensional homology model of CaV2.3 based upon Kv1.2 where hydrophobic residues at positions facing Val1720 in IS6, IIS6, and IIIS6 play a critical role in stabilizing the closed state in CaV2.3.

The hydrophobic locus VAVIM is conserved in the S6 transmembrane segment of domain IV (IVS6) in Ca V 1 and Ca V 2 families. Herein we show that glycine substitution of the VAVIM motif in Ca V 2.3 produced whole cell currents with inactivation kinetics that were either slower (A1719G ≈ V1720G), similar (V1718G), or faster (I1721G ≈ M1722G) than the wild-type channel. The fast kinetics of I1721G were observed with a ≈؉10 mV shift in its voltage dependence of activation (E 0.5,act ). In contrast, the slow kinetics of A1719G and V1720G were accompanied by a significant shift of ≈؊20 mV in their E 0.5,act indicating that the relative stability of the channel closed state was decreased in these mutants. Glycine scan performed with Val 349 in IS6, Ile 701 in IIS6, and Leu 1420 in IIIS6 at positions predicted to face Val 1720 in IVS6 also produced slow inactivating currents with hyperpolarizing shifts in the activation and inactivation potentials, again pointing out a decrease in the stability of the channel closed state. Mutations to other hydrophobic residues at these positions nearly restored the channel gating. Altogether these data indicate that residues at positions equivalent to 1720 exert a critical control upon the relative stability of the channel closed and open states and more specifically, that hydrophobic residues at these positions promote the channel closed state. We discuss a three-dimensional homology model of Ca V 2.3 based upon Kv1.2 where hydrophobic residues at positions facing Val 1720 in IS6, IIS6, and IIIS6 play a critical role in stabilizing the closed state in Ca V 2.3.
Voltage-dependent Ca 2ϩ channels (VDCC) 3 are membrane proteins that play a key role in promoting Ca 2ϩ influx in response to membrane depolarization in excitable cells. VDCCs arise from the multimerization of distinct subunits: Ca V ␣1, Ca V ␤, and Ca V ␣2␦, and sometimes Ca V ␥ (1). To this date, molecular cloning has identified the primary structures for 10 distinct calcium channel Ca V ␣ 1 subunits (2)(3)(4)(5)(6)(7)(8) that are classified into three main subfamilies according to their gating properties (Ca V 1, Ca V 2, and Ca V 3). The Ca V ␣1 subunit is the main pore-forming subunit that carries the channel activation gating among other functions. The Ca V ␣1 subunits of VDCCs are evolutionarily related to the ␣ subunit of Kv channels with a single polypeptidic chain carrying four domains of six transmembrane segments (S1-S6) (9). Although the overall identity at the primary sequence level is very low between Ca V and Kv channels, it goes up to 10 -25% when comparing the S6 transmembrane segments. As in Kv channels, the S6 transmembrane segments of Ca V ␣1 are believed to line the channel pore and form the channel inner vestibule. It was inferred from the three-dimensional structures of KcsA, MthK, KvAP, KirBac, and Kv1.2 channels that the M2/S6 transmembrane segments include the activation gate that controls channel opening (10 -14). In the Shaker K ϩ channels, the residue hydrophobicity in this region could alter the channel closed-open equilibrium (15,16).
To study the functional importance of S6 residues in the gating of Ca V 2.3, we searched for conserved motifs of hydrophobic residues. The VAVIM motif is conserved in the S6 transmembrane segment of domain IV (IVS6) of high voltageactivated Ca V 1 and Ca V 2 families (Fig. 1A). Numerous algorithms align the PVPVIV activation locus in Shaker Kv channels with the FVAVIM (50% identity) (Fig. 1B) suggesting that this locus could play a role in the activation gating of HVA VDCCs. Precious clues regarding the role of the VAVIM motif in channel function came from genetic diseases. Mutation of the conserved Ile to Leu in Ca V 2.1 was identified in patients suffering from familial hemiplegic migraine (17,18). I1811L altered the channel gating properties by shifting the voltage dependence of activation by Ϫ5 to Ϫ12 mV, depending upon the expression system (17, 19 -21
Functional Expression of Ca V 2.3-Oocytes were obtained from female Xenopus laevis clawed frogs as described previously (25)(26)(27). Oocytes were injected with 46 nl of a solution containing cRNA coding for the Ca V ␣1, Ca V ␣2b␦, and Ca V ␤3 subunits in a 3:1:2 weight ratio. Co-expression with the ancillary Ca V ␤3, which is predominant in brain tissues like Ca v 2.3 (28,29), promotes strong functional expression (30,31), and emphasizes closed-state inactivation (32) of Ca V 2.3. Functional expression of mutants was deemed significant with whole cell Ba 2ϩ currents Ͼ300 nA.
Data Acquisition and Analysis-PClamp software 8.2 (Molecular Devices, Sunnyvale, CA) was used for on-line data acquisition and analysis. A series of 450-ms voltage pulses (5 mV steps) were applied at a frequency of 0.2 Hz from a holding potential of Ϫ120 mV to take into account the negative shift of the Ile 1721 mutants in the inactivation curve. Activation parameters were estimated from the peak I-V curves obtained for each channel combination and are reported as the mean of individual measurements Ϯ S.E. Briefly, the I-V relationships were normalized to the maximum amplitude and were fitted to a Boltzmann equation with E 0.5,act being the mid-potential of activation as described elsewhere (25)(26)(27). The estimation of E 0.5,act using non-stationary measurements rests upon the assumption that the transition to the open state is much faster than the transition to the inactivated state. The measure of E 0.5,act yields an estimation of the free energy differences between closed (C) and open (O) states as explained previously (16). Changes in E 0.5,act can thus be interpreted as a modification of the ratio between the open and closed states. Time constants of whole cell current traces were estimated with the predefined Equation 1 in Clampfit 8.2 that uses the Chebyshev routine and a 4-point smoothing filter with n ϭ 1 for deactivation time constants and n ϭ 3 for inactivation time constants. Under the latter, the inactivation time course requires two time constants inact,fast and inact,slow .
As the number of exponential functions needed to account for the inactivation process varied between mutants, inactivation kinetics were quantified using r 50 values. The r 50 ratio is defined as the ratio of peak whole cell currents remaining 50 ms later (I 50 ms /I Peak ). The voltage dependence of inactivation was obtained after a series of 5-s prepulses (h5000 for h measured at 5000 ms) that varied from Ϫ120 to ϩ30 mV at a frequency of 0.02 Hz with E 0.5,inact being the mid-potential of inactivation (25)(26)(27). Most mutations inactivated completely within the 5-s pulse with less than 2% of the peak currents remaining at the end of the inactivating pulse. However, the phrase "isochronal inactivation" is more accurate when estimating the voltage dependence of inactivation of the I701G mutant that displayed 15% residual currents. Sta- In this alignment, the mid-glycine residues in IS6 (Gly 338 ) and IIS6 (Gly 690 ) are not aligned with the universal glycine of K ϩ channels. The alignment of IIS6 was chosen based upon its overall homology with the other S6 segments of Ca V 2.3 (IS6, IIIS6, and IVS6) but is not the best possible alignment of IIS6 with S6 of Kv1.2. Indeed, CLUSTALW predicts the FLAIAVD locus of IIS6 to be aligned with ALPVPVI of Kv1.2. The residues of the putative activation gate in Kv channels (PVPVIV) are underlined and shown to be aligned with FVAVIM in IVS6 of Ca V 2.3.
tistical analyses were performed using the one-way analysis of variance fitting routine for two independent populations included in Origin 7.0. Data were considered statistically significant at p Ͻ 0.01.
Homology Modeling of S6 Regions in Ca V 2.3-An analysis based on the SAMT02 algorithm confirmed Kv1.2 as a suitable template to model the S5 and S6 segments in each of the 4 domains of Ca V 2.3. The identity of Kv1.2 with Ca V 2.3 at the primary sequence varies from 11% for IIIS6 to up to 25% for IVS6 with values of 17% for IS6 and 22% for IIS6. The primary sequence of each of the four S6 segments of Ca V 2.3 was aligned with the S6 segment of Kv1.2 by using LALIGN without any gap in the structure. Introducing gaps in the alignment could in certain cases (such as in IIS6) improve local alignments while decreasing the overall homology. For each S6, the score for the alignment of the distal S6 was systematically higher than for the beginning of the S6 segment. A Kv1.2-based model of the selectivity filter and pore helix could not, however, be generated as the linkers were generally longer in Ca V than in Kv1.2. Hence, the computer-based molecular model of these regions without the selectivity filter and the pore helix was achieved with Modeler 9v1 (salilab.org/modeler) using the molecular coordinates of Kv1.2 (Protein Data Bank 2A79). 100 models were built and the models with the lowest objective function were further checked for internal consistency using Procheck (biotech. ebi.ac.uk/cgi-bin/sendquery). The precision of the models decreased significantly for the residues located at the C-terminal end in the absence of structural constraints. The global score of the final model was 0.10, whereas values higher than Ϫ0.50 are generally considered to be acceptable. The model was minimized with the DISCOVER module of INSIGHTII (Accelrys, San Diego, CA) with a dielectric constant of 80 to simulate the presence of solvent. The three-dimensional representations were generated with the INSIGHTII software as described elsewhere (34,35).

RESULTS
The Role of IVS6 in the Gating of Ca V 2.3-The I1811L mutation is one of the four original mutations of the Ca V ␣1 subunit of the human Ca V 2.1 that is associated with familial hemiplegic migraine (17,36). As the Ile residue is conserved in HVA VDCCs, we hypothesized that it might be important for the normal gating properties of HVA channels such as Ca V 2.3. Most Ile 1721 mutants yielded high voltage-activated inward Ba 2ϩ currents (Fig. 2, supplemental Fig. S1 and Table S1) with the exception of I1721D, I1721R, and I1721K that failed to express functional channels. The observation that Ile 1721 tolerated a wide range of point mutations suggests a minimum of structural constraints in its local environment. Other than I1721L and I1721A, functional Ile 1721 mutants produced currents with faster inactivation kinetics than the wild-type channel (supplemental Fig. S1). At ϩ10 mV, r50 decreased from wild-type . This is the first report of mutations speeding up the inactivation kinetics of Ca V 2.3 because mutations in the I-II linker decreased the inactivation kinetics (33). Our results with I1721T in Ca V 2.3 contrast with reports that the equivalent mutation I1475T in IVS6 of the L-type Ca V 1.2 did not markedly alter the kinetics of whole cell current inactivation (37).
Charged residues at position 1721 produced non-functional mutants save for I1721E that showed little difference in activation and inactivation potentials despite faster inactivation kinetics. I1721T displayed faster inactivation kinetics without any significant changes in E 0.5,act or E 0.5,inact . Both the activation and inactivation gating of I1721L and I1721A were significantly shifted by ХϪ5 to Ϫ10 mV as compared from the wild-type channel (supplemental Table  S1) but inactivated with kinetics similar to the wild-type  Fig. 3. B, the r 50 values (the fraction of peak whole cell currents remaining after a 50 ms pulse) are shown Ϯ S.E. from 0 to ϩ30 mV for Ca V 2.3 wild-type (32), I1721L (10), I1721A (9), I1721T (12), and I1721G (26) (from left to right on the bar graph). At ϩ10 mV, the values of I1721T and I1721G were significantly different at p Ͻ 0.0001, whereas r 50 of I1721L and I1721A were not significantly different (p ϭ 0.9) as compared with the wild-type. C, the differences in the mid-points of inactivation and activation of mutants were obtained by subtracting the E 0.5 values measured for the wild-type from the values of the mutants measured under the same conditions such that a ⌬ value ϭ 0 indicates no difference from the wild-type channel. All mutants were modulated by Ca V ␤3 in a typical fashion (data not shown). The dotted lines drawn at Ϯ5 mV around 0 mV indicates a value significantly different from the wild-type channel by p Ͻ 0.01. The mid-points of inactivation E 0.5,inact (striped columns) of all mutants were significantly shifted at p Ͻ 0.01 to negative potentials as compared with the wild-type channel. As compared with the wild-type channel, the midpoints of activation E 0.5,act (plain gray columns) of I1721L and I1721G were significantly shifted at p Ͻ 0.001 in the negative and positive directions, respectively.
channel. In the case of I1721V, I1721P, and I1721G, the activation potentials were significantly shifted toward positive voltages, whereas inactivation potentials were shifted in the negative direction. These observations coupled with the faster inactivation kinetics of these mutants suggest that the fraction of channels in the open state at any given voltage was decreased in I1721V, I1721P, and I1721G. This conclusion holds true even when taking into account the changes in the slope factor Z. As estimated from the ⌬G app aa scale calculated by the group of Stephen White (38), hydrophobicity decreases in the following order: Ile Х Leu Ͼ Val Ͼ Ala Ͼ Thr Ͼ Gly Ͼ His Ͼ Pro Ͼ Glu, whereas volume decreases from Ile Х Leu Ͼ His Ͼ Val Х Glu Ͼ Thr Ͼ Pro Ͼ Ala Ͼ Gly. As a result, no single parameter such as hydrophobicity, charge, or volume can account for the altered channel gating of Ile 1721 mutants.
V1720G and I1721G Yield Clues Regarding the Position of the Channel Activation Gate in IVS6-With I1721V and I1721P, I1721G produced the strongest changes in channel gating. To minimize steric and structural constraints, we chose to identify residues that are critical for channel activation gating by performing a glycine scan of the conserved VAVIM locus in the distal portion of IVS6 (Fig. 3A). Glycine mutants decreased the inactivation kinetics of neighboring A1719G and V1720G residues but sped up the kinetics of I1721G and M1722G as compared with the wild-type and V1718G channels. The r 50 values thus decreased from A1719G Х V1720G Ͼ wild-type Х V1718G Ͼ I1721G Х M1722G (Fig. 3B). The fast inactivation kinetics of the latter were accompanied by a faster recovery from inactivation, suggesting that the inactivated state was not more stable in I1721G and M1722G (supplemental Fig. S2A). The activation potentials of the latter were either not significantly different (M1722G) or more positive (I1721G, p Ͻ 0.001) than the wildtype channel. Coupled with the fact that these 2 mutants inactivated at potentials ϷϪ30 mV more negative than the wild-type channel (Fig. 3C), these observations altogether indicate that glycine mutations introduced at positions 1721 and 1722 result in a decreased channel open state. In all these mutants, activation and inactivation potentials were typically shifted toward negative voltages by Ca V ␤3 (supplemental Fig.  S2B). For I1721G, this means that E 0.5,act ϭ 6.9 Ϯ 0.4 mV (n ϭ 6) and E 0.5,inact ϭ Ϫ64 Ϯ 3 mV (n ϭ 3) in the absence of Ca V ␤3, values that remain significantly different from E 0.5,act ϭ Ϫ1.4 Ϯ 0.4 mV (n ϭ 29) and E 0.5,inact ϭ Ϫ33.7 Ϯ 0.8 mV (n ϭ 36) for the wild-type channel under the same conditions.
The faster kinetics of I1721G and M1722G contrast with the slower kinetics produced by G352A in IS6 and R378E mutations in the I-II linker (33,39). It was then proposed that the slow kinetics of these mutants resulted from either a disruption of the fast inactivated state or from a stabilization of the channel open state. To investigate whether the increased inactivation gating of I1721G and M1722G could antagonize the effects of the slow mutants, the double mutants G352A/M1722G and R378E/I1721G were produced. Unfortunately, neither double mutant expressed functional whole cell currents, leaving this question unanswered.
In contrast to I1721G and M1722G, A1719G and V1720G displayed slow inactivating kinetics accompanied with negative shifts of ϷϪ20 mV in their activation potentials. This change in gating departs from our observations with G352A in IS6 (25) and the charged mutations of Arg 378 in the I-II linker (33,39) where the decreased inactivation kinetics occurred without any change in the channel activation potential of Ca V 2.3. The significant shifts in the E 0.5,act  Table S1. All mutants were modulated by Ca V ␤3 in a typical fashion (supplemental Fig. S2B). B, the r 50 values in the presence of Ca V ␤3 are shown Ϯ S.E. from 0 to ϩ30 mV for Ca V 2.3 wild-type (32), V1718G (18), A1719G (20), V1720G (25), I1721G (26), and M1722G (17) (from left to right on the bar graph). C, the differences in the mid-points of inactivation and activation of mutants were obtained by subtracting the E 0.5 values measured for the wild-type from the values of the mutants such that a ⌬ value ϭ 0 indicates no difference from the wild-type channel (see supplemental Table S1 for actual values). The dotted lines drawn at Ϯ5 mV around 0 mV indicates a value significantly different form the wild-type channel by p Ͻ 0.01. The midpoints of inactivation E 0.5,inact (striped columns) of most mutants but A1719G (p Ͻ 0.01) were significantly shifted at p Ͻ 0.001 to negative potentials as compared with the wild-type channel. As compared with the wild-type channel, the mid-point of activation E 0.5,act (gray columns) of I1721G was significantly shifted at p Ͻ 0.001 toward positive potentials by ϩ10 mV, whereas E 0.5,act of A1719G and V1720G was significantly shifted toward negative voltages by ХϪ17 mV (p Ͻ 0.001). In contrast, E 0.5,act of V1718G was only slightly shifted at p Ͻ 0.1 by Ϫ3 mV.
toward negative voltages were accompanied by similar increases in the slope factor Z. Hence, introducing glycine residues at positions 1719 and 1720 appears to shift the open/closed equilibrium toward the open state. Mutations of Val 1720 to Ala and Ile nearly restored the normal channel gating with smaller shifts in the mid-potentials of activation and inactivation than measured with V1720G ( Fig. 4; supplemental Table S1) suggesting that hydrophobic residues at this position could shift the closed/open state equilibrium toward the channel closed state.
Glycine Mutations in Positions Equivalent to Val 1720 Promote the Channel Open State-As the primary structure of VDCCs identifies 4 distinct S6 domains, domain-specific effects were investigated at positions equivalent to Val 1720 (Fig. 5) and Ile 1721 (L350 in IS6, Ala 702 in IIS6, and Ile 1421 in IIIS6) (supplemental Fig. S3). As seen for I1721G (IVS6), I1421G (IIIS6) displayed inactivation kinetics significantly faster than the wild-type channel at all voltages (supplemental Fig.  S3). Unlike I1721G, however, the activation potential of I1421G was not significantly altered, which suggests that the mutation did not affect the relative stability of the open and closed states. L350G (IS6) behaves mostly like the wild-type channel, whereas the slower inactivation kinetics of A702G (IIS6) were observed in the absence of significant shifts in its activation and inactivation potentials (supplemental Fig. S3).
In contrast, glycine mutations at positions Val 349 (IS6), Ile 701 (IIS6), and Leu 1420 (IIIS6) equivalent to Val 1720 (IVS6) significantly altered the biophysical properties of the channel. In particular, all mutations slowed down the channel inactivation kinetics (Fig. 5B). At 0 mV, the r 50 values decreased from wild-type Х V349G (IS6) Ͻ L1420G (IIIS6) Х V1720G (IVS6) Ͻ I701G (IIS6). The decrease in channel kinetics in these mutants was achieved through the progressive elimination of the fast inactivation time constant up to the point where the contribution of the fast time constant becomes negligible in I701G. The mutations also significantly hyperpolarized the activation potentials by Ϫ5 to Ϫ35 mV and the inactivation potentials by Ϫ18 to Ϫ24 mV (supplemental Tables S1 and S2 and Fig. S4). The most spectacular alterations in channel properties were, however, observed for I701G in IIS6. Indeed, the hyperpolarization of its activation curve was stronger than the negative shift in its mid-potential of inactivation (Fig.  5C). Finally, channel activation kinetics were also affected as the activation kinetics of I701G were remarkably voltage-dependent with slow act ϭ 5.8 Ϯ 0.4 ms (n ϭ 9) at the peak voltage of Ϫ35 mV to reach act ϭ 2.5 Ϯ 0.1 ms (n ϭ 9) at Ϫ5 mV, whereas the wild-type channel activated with reach act ϭ 2.1 Ϯ 0.3 ms (n ϭ 17) at Ϫ20 Ͻ V m Ͻ 25 mV. I701G significantly slowed deactivation kinetics (Fig. 6). Altogether, these changes in channel gating suggest that introducing a glycine residue at this position either destabilized the channel closed state or else stabilized its open state. Similar changes in channel gating were reported after substitutions of Ile 781 , the Ca V 1.2 residue corresponding to Ile 701 in Ca V 2.3, suggesting a congruence in the gating mechanism between Ca V 1 and Ca V 2 channel types (40).
Although the altered gating was more spectacular with I701G, glycine mutations of the LAIAV motif in IIS6, save for A702G, produced slowly inactivating channels that activated at more negative potentials than the wild-type channel (supplemental Fig. S5 and Table S2). In particular, V703G imparted significant hyperpolarizing shifts in the activation and inactivation potentials of Ca V 2.3. These data indicate that IIS6 plays a unique role in the channel equilibrium between the closed and open state(s).
Altogether, introducing glycine mutations in the distal portion of S6 segments in Ca V 2.3 produced position specific alterations in activation and inactivation gating within IIIS6 and IVS6 that were absent in IS6 and IIS6. Conservative mutations of the same residues to hydrophobic amino acids nearly restored the channel gating (supplemental Table S2). These observations confirm that S6 transmembrane segments of Ca V channels are not functionally equivalent, an idea that was first brought up by the mutational analysis of . C, the ⌬ of the mid-points of inactivation E 0.5,inact (striped columns) and activation E 0.5,act (gray columns) were calculated as described before in the presence of Ca V ␤3 such that a ⌬ value ϭ 0 indicates no difference from the wild-type channel (see supplementary Table S2 for actual values). The dotted lines drawn at Ϯ 5 mV around 0 mV indicates a value significantly different form the wild-type channel by p Ͻ 0.01. the selectivity filter of Ca V 1.2 (41,42) and recently substantiated by the SCAM study of Ca V 2.1 (43,44).

DISCUSSION
In Kv channels, mutations to residues less hydrophobic than wild-type residues "at the closed bundle" was shown to shift the closed-open equilibrium in favor of the open state (15,16) prompting the suggestion that the closed conformation is stabilized by hydrophobic interactions. Thus hydrophobic interactions or "hydrophobic seals" may prevent ion flow in the closed state (45). Three hydrophobic seals have been postulated in KcsA (Thr 107 , Ala 111 , and Val 115 ) (45), whereas only Val 474 and Val 478 are present in Shaker K ϩ channels (46). Three-dimensional structures of the KcsA and Kv1.2 channels show that the residues postulated to form the hydrophobic seal project their side chains toward the channel pore. Of these two positions, Val 478 was clearly identified as a key candidate for occlusion of the pore in the closed state (15).
Three-dimensional structures of the ␣1 subunits of Na V and Ca V channels are not known. Hence, models developed for symmetrical 6-transmembrane ion channels remain widely invoked with asymmetrical 24-transmembrane voltage-gated Na ϩ (47)(48)(49)(50) and Ca 2ϩ channels (51). To explore structural and functional similarities between the Kv and the Ca V channels, we have built a threedimensional model of the S6 segments of Ca V 2.3 using the atomic coordinates for Kv1.2 (Fig. 7). Structural predictions based on the threading SAMT02 algorithm suggested Kv1.2 as a suitable template to model the S5 and S6 segments in each of the 4 domains of Ca V 2.3 but ruled out Kv1.2 as a model for the selectivity filter, as well as for the S4 -S5 and S5-S6 linkers. The absence of the selectivity filter, however, limits the precision in the spatial orientation of the four S6 transmembrane segments. Within the limits of the three-dimensional model, the sequence alignment predicts Val 1720 and Ile 1721 to be present in the bent section of the S6 helix. Indeed there is a Ͼ45% local identity in a 17-amino acid span in the distal region such that LALIGN, CLUSTALW, and TCOFFEE yielded the same residue alignment between the S6 of Kv1.2 and IVS6 of Ca V 2.3 (Fig. 1B).
The three-dimensional model in Fig. 7 is supported by a SCAM study of Ca V 2.1 where the equivalent residues in IVS6, save for the middle Val, were accessible from the intracellular medium (43) suggesting the presence of a wide intracellular vestibule in the open state. The SCAM study of the Ca V 2.1 pore also showed the VAVIM locus to be modified by MTSET in a state-dependent manner (43,44) indicating that it is located close to or at the activation gate. This contrasts with inwardly rectifying GIRK4 or Kir3.4 channels where rigidity below the transmembrane 2 helix bundle crossing appears to be a requirement for channel gating (52).
Transposing the results from mutational analyses in Shaker channels to Ca V 2.3 hence predicts that large hydrophobic residues at positions equivalent to Val 478 should stabilize the closed conformation by strengthening the second hydrophobic seal (46). Conversely, small and/or hydrophilic residues should promote channel activation at more negative membrane potentials. Our mutational analysis of S6 segments in Ca V (19) for I701G (p Ͻ 0.001), 0.59 Ϯ 0.05 (16) for L1420G (p Ͻ 0.001), and 0.70 Ϯ 0.01 (31) for V1720G (p Ͻ 0.001) (from left to right on the bar graph). IS6 already contains 3 endogenous glycine residues that could account for the smaller impact of the glycine mutation at this position (25). C, the differences in the mid-points of inactivation and activation of V349G (Domain I), I701G (Domain II), L1420G (Domain III), and V1720G (Domain IV) were obtained from the difference between E 0.5,wt and E 0.5,mutant in the presence of Ca V ␤3 such that a ⌬ value ϭ 0 indicates no difference from the wild-type channel (see supplemental Table S2 for actual values). The dotted lines drawn at Ϯ 5 mV around 0 mV indicates a value significantly different form the wild-type channel by p Ͻ 0.01. The mid-points of activation E 0.5,act (gray columns) and inactivation E 0.5,inact (striped columns) of mutants were significantly shifted at p Ͻ 0.001 to more negative potentials as compared with the wild-type channel. positive potentials. In contrast, Gly mutations at positions 1720 (IVS6), 1420 (IIIS6), 701 (IIS6), and 349 (IS6) promoted the channel open state with channel activation significantly shifted toward more negative voltages. At positions 1720 and 1420, mutations to Ala, a residue that is more hydrophobic than Gly but less hydrophobic than the respective wildtype valine and leucine residues, tended to restore the channel gating. It is hence proposed that hydrophobic residues at positions 1720, 1420, 701, and 349 contribute to stabilize the closed state in Ca V 2.3. Val 1720 in our three-dimensional model is aligned with Val 476 (supplemental Fig. S6) and hence does not appear to project its side chain toward the pore region unlike Val 478 in Kv1.2. Nonetheless, the V476A mutation in the Shaker Kv channels imparted a Ϫ20 mV shift in its E 0.5,act value (15). The very good identity at the primary sequence level between the distal IVS6 of Ca V 2.3 and S6 of Kv1.2 supports our three-dimensional model in this region.
The position of key residues in IIS6 remains, however, open to discussion. Alignment predictions vary whether or not gaps are introduced in the structure and in some alignments, Val 703 would be facing with Val 476 in Kv1.2 and Val 1720 in IVS6. To note, mutations at position Val 703 were also seen to significantly promote the channel open state. The orientation of the side chains facing directly or not the channel permeation pathway may or may not influence channel activation depending upon helix packing in the region. In this regard, it was concluded from the MTSET modification of 5 consecutive positions in IIS6 of Ca V 2.1, that the S6 segments of Ca V channels are more loosely packed than S6 of Kv channels (43).
The observation, that residues Ile 701 and Val 703 located within a half-helicoidal turn of each other in IIS6 affects channel activation, FIGURE 6. Deactivation time constants for the S6 residues facing Val 1720 in Ca V 2.3. A, representative tail currents for wild-type, V349G, L1420G, and V1720G channels. Currents were activated during a 8-ms conditioning depolarization to 0 mV for Ca V 2.3 wild-type, V349G, and L1420G; and Ϫ10 mV for V1720G. Deactivation was recorded during subsequent repolarizations with 10-mV increments starting from Ϫ120 mV (test potentials). Time constants were estimated by fitting current deactivation to a monoexponential function. B, representative tail currents for I701G channels. Currents were activated during a 15-ms conditioning depolarization to Ϫ40 mV. Deactivation was recorded during subsequent repolarizations with 10-mV increments starting from Ϫ120 to Ϫ40 mV (test potentials). Time constants were estimated by fitting current deactivation to a mono-or a biexponential function. Monoexponential functions were found to fit reasonably well the tail currents of all channels and allow for a better comparison between the wild-type and other mutants. C, mean time constants of channel deactivation (monoexponential functions) for Ca V 2.3 wild-type, V349G, I701G, L1420G, and V1720G are plotted versus test potential. At Ϫ40 mV, the time constants deact were: 1.7 Ϯ 0.1 ms (n ϭ 8) for wild-type; 2.0 Ϯ 0.2 ms (n ϭ 7) for V349G; 3.3 Ϯ 0.06 ms (n ϭ 6) for L1420G; 5 Ϯ 1 ms (n ϭ 6) for V1720G; and 130 Ϯ 20 ms (n ϭ 6) for I701G such that deactivation kinetics decreased from wild-type ХV349G Ͻ L1420G Ͻ V1720G Ͻ Ͻ I701G. The pulse protocol is shown in the inset. Computer-based molecular model of the pore region of Ca V 2.3 (IS6, IIS6, IIIS6, and IVS6 transmembrane segments without the selectivity filter and the pore helix) is shown as a ribbon representation. Kv1.2 (PDB code 2A79) is a suitable template for the S5 and S6 segments but not for the selectivity filter or the cytoplasmic S4 -S5 linker. Modeling was achieved with Modeler 9v1. Although the S6 remains helical at the site of the PVP motif (putative activation gate in Kv1.2), the distal residues are located in the relaxed portion of the S6 helix. Color codes indicate IS6 (red), IIS6 (violet), IIIS6 (white), and IVS6 (blue). The residues in IS6 (orange), IIS6 (pink), and IIIS6 (gray) equivalent to the VAVIM locus in IVS6 (turquoise) are shown in the same color with a different shade. The representation shows the channel pore in a transversal cut with the top of the S6 helices facing the extracellular medium. raises the possibility that cytoplasmic linkers might play an active role in controlling transition(s) from the closed to the open state. Indeed, the role of the S4 -S5 intracellular linker in the activation of HVA Ca V channels remains to be addressed. Numerous studies have already shown that mutations in the cytoplasmic S4 -S5 alter the activation kinetics of Kv2.1 (53) and HERG channels (54) among others, suggesting that interactions between the S4 -S5 linker and the S6 helix seem to be crucial for the coupling of the voltage sensor movements to the activation gate. The group of MacKinnon and co-workers (55) has proposed in 2005 that the S4 -S5 linker functions as a lever that exerts a mechanical force on the activation gate. Inward movement of the voltage sensor with hyperpolarization is postulated to compress the S4 -S5 linker against the S6 helices, resulting in the tightening of the permeation pathway (55). It is thus possible that the distal S6 mutations in Ca V 2.3 influence the angle between the proximal and distal S6 such that it disrupts the coupling between the voltage sensor and the activation gate. As mentioned earlier, the S4 -S5 and the S5-S6 linkers were not modeled in our study due to the limited identity between Kv1.2 and Ca V 2.3 in these regions. It thus remains to be seen whether the molecular mechanisms underlying the transition between the closed and open state are transposable from Kv1.2 to Ca V 2.3. Questions regarding the role of the S4 -S5 linker go beyond the scope of this current investigation and await further structural studies.