A Short Polybasic Segment between the Two Conserved Domains of the β2a-Subunit Modulates the Rate of Inactivation of R-type Calcium Channel*

Background: Membrane anchoring underlies inhibition of voltage-dependent inactivation (VDI) of calcium channels by the β2a-subunit. Results: A polybasic segment of β2a slows VDI without membrane association. This effect is abolished by charge neutralization. Conclusion: VDI inhibition by the β2a-subunit can occur without membrane anchoring by a mechanism that relies on positively charged residues. Significance: A novel mechanism underlying inhibition of VDI in calcium channels is revealed. Besides opening and closing, high voltage-activated calcium channels transit to a nonconducting inactivated state from which they do not re-open unless the plasma membrane is repolarized. Inactivation is critical for temporal regulation of intracellular calcium signaling and prevention of a deleterious rise in calcium concentration. R-type high voltage-activated channels inactivate fully in a few hundred milliseconds when expressed alone. However, when co-expressed with a particular β-subunit isoform, β2a, inactivation is partial and develops in several seconds. Palmitoylation of a unique di-cysteine motif at the N terminus anchors β2a to the plasma membrane. The current view is that membrane-anchored β2a immobilizes the channel inactivation machinery and confers slow inactivation phenotype. β-Subunits contain one Src homology 3 and one guanylate kinase domain, flanked by variable regions with unknown structures. Here, we identified a short polybasic segment at the boundary of the guanylate kinase domain that slows down channel inactivation without relocating a palmitoylation-deficient β2a to the plasma membrane. Substitution of the positively charged residues within this segment by alanine abolishes its slow inactivation-conferring phenotype. The linker upstream from the polybasic segment, but not the N- and C-terminal variable regions, masks the effect of this determinant. These results reveal a novel mechanism for inhibiting voltage-dependent inactivation of R-type calcium channels by the β2a-subunit that might involve electrostatic interactions with an unknown target on the channel's inactivation machinery or its modulatory components. They also suggest that intralinker interactions occlude the action of the polybasic segment and that its functional availability is regulated by the palmitoylated state of the β2a-subunit.

Besides opening and closing, high voltage-activated calcium channels transit to a nonconducting inactivated state from which they do not re-open unless the plasma membrane is repolarized. Inactivation is critical for temporal regulation of intracellular calcium signaling and prevention of a deleterious rise in calcium concentration. R-type high voltage-activated channels inactivate fully in a few hundred milliseconds when expressed alone. However, when co-expressed with a particular ␤-subunit isoform, ␤ 2a , inactivation is partial and develops in several seconds. Palmitoylation of a unique di-cysteine motif at the N terminus anchors ␤ 2a to the plasma membrane. The current view is that membrane-anchored ␤ 2a immobilizes the channel inactivation machinery and confers slow inactivation phenotype. ␤-Subunits contain one Src homology 3 and one guanylate kinase domain, flanked by variable regions with unknown structures. Here, we identified a short polybasic segment at the boundary of the guanylate kinase domain that slows down channel inactivation without relocating a palmitoylation-deficient ␤ 2a to the plasma membrane. Substitution of the positively charged residues within this segment by alanine abolishes its slow inactivation-conferring phenotype. The linker upstream from the polybasic segment, but not the N-and C-terminal variable regions, masks the effect of this determinant. These results reveal a novel mechanism for inhibiting voltage-dependent inactivation of R-type calcium channels by the ␤ 2a -subunit that might involve electrostatic interactions with an unknown target on the channel's inactivation machinery or its modulatory components. They also suggest that intralinker interactions occlude the action of the polybasic segment and that its functional availability is regulated by the palmitoylated state of the ␤ 2a -subunit.
The entry of calcium ions into the cell triggers a multitude of cellular responses that rely on a tight spatio-temporal regulation of the calcium transient spreading within the cell for their coordination (1,2). High voltage-activated (HVA) 4 calcium channels open in response to large membrane depolarization and constitute the major entry pathway for calcium into excitable cells. The largest component of HVA channels, ␣ 1 -subunit (Ca V ␣ 1 ), contains the ion conduction pore, the voltage sensor, and multiple intracellular domains that regulate calcium influx and provide interaction sites for regulatory proteins. Two subfamilies of HVA Ca V ␣ 1 are recognized as follows: Ca V 1.x or L-type channels and Ca V 2.x, also referred as to neuronal channels (Ca V 2.1 encoding P/Q-type; Ca V 2.2 encoding N-type; and Ca V 2.3 encoding R-type) (3). Following strong depolarization and calcium permeation, HVA channels enter into an inactivated nonconducting state that constrains the amount of calcium influx and protects the cells from the cytotoxic effects of an excessive calcium rise. Inactivation of HVA channels is triggered by both calcium increase and prolonged membrane depolarization, referred as to calcium-dependent inactivation and voltage-dependent inactivation (VDI), respectively (4). Although calcium-dependent inactivation depends on the binding of calmodulin to a conserved site among HVA channels (5), VDI is an intrinsic property of Ca V ␣ 1 , but it is strongly modulated by the regulatory ␤-subunit (Ca V ␤) (6 -11) that binds to a site highly conserved among HVA channels, termed the ␣-interaction domain (12).
Crystallographic studies from three of the four known Ca V ␤ isoforms (Ca V ␤ 1 to Ca V ␤ 4 ) revealed the molecular aspects of this interaction. Ca V ␤ shares a common structural arrangement with members of the membrane-associated guanylate kinase family (MAGUK), a highly conserved Src homology 3 (SH3) and guanylate kinase (GK) domain flanked and joined by variable segments (13)(14)(15). The binding motif in the pore-forming subunit, ␣-interaction domain, forms an ␣-helix that penetrates into a hydrophobic groove within the GK domain. In contrast to the structural core of the protein, namely the SH3-GK unit, the structural data for the variable regions and their interactions with the rest of the protein are as yet unknown. Functionally, all ␤-subunits shift the voltage dependence of activation toward more negative voltages, but they differentially modulate inactivation of HVA Ca V ␣ 1 subtypes (6,8).
The native R-type Ca V 2.3 channel, originally cloned from human and mouse brain (16), inactivates very fast upon depolarization, and in heterologous expression systems, most Ca V ␤s speed up inactivation even further. Ca V ␤ 2a , hereby referred as to ␤ 2a , is the only isoform that slows inactivation and is also the only isoform that undergoes palmitoylation (6,17). This splice variant encompasses two adjacent cysteine residues at positions 3 and 4 within its N-terminal variable region. These are dynamically palmitoylated and target the protein to the plasma membrane (17)(18)(19)(20). In oocytes expressing Ca V 2.3 channels, ␤ 2a induces a 13-fold increase in the time for the current to return to 50% of its peak amplitude evoked by depolarizing pulse to 0 mV (hereby referred as to T 0.5 ). Even after 15 s, inactivation is still incomplete; leaving around 10% of residual current (6). Substituting the two cysteines by alanines impairs membrane localization and also abolishes this slow inactivation phenotype (17,18). The prevalent view is that anchoring of ␤ 2a to the membrane slows down channel inactivation by immobilizing the inactivation machinery and therefore preventing it from occluding the conduction pathway (19,21). However, it was shown later that the slow inactivation-conferring phenotype can be manifested in nonmembrane-anchored Ca V ␤, suggesting that this is not the only mechanism that contributes to slowing down VDI (7,22,23). The same studies also implicate the linker joining the SH3 and GK domains in VDI regulation. Among the possible mechanisms that come to mind is that the linker contributes to anchoring the ␤-subunit to the plasma membrane, so that it interacts directly with the inactivation machinery or modifies the interaction of the SH3-GK core with the channel protein. Here, we used Xenopus oocytes to study the effect of different Ca V ␤ derivatives on VDI and tsA-201 cells to investigate the intracellular localization of these protein constructs. The rationale is that although mammalian cells are well suited for confocal microscopy, they are too small for protein injection. The opposite is true for Xenopus oocytes. Moreover, expression of Ca V ␣ 1 in mammalian cells is sensitive to the co-expression of the ␤-subunit. In contrast, in Xenopus oocytes association with the Ca V ␤ does not seem to be a requirement for the pore-forming subunit to reach the plasma membrane. In fact, Ca V 2.3 bearing a mutation in the primary ␤-subunit anchoring site expresses as well as wild-type channels (24). This is also true for Ca V 1.2 channels (25)(26)(27). Using this approach, we identified within the linker region of the ␤-subunit a positively charged segment, 12 amino acids long, that slows down inactivation without conferring membrane anchoring to the protein.
The effect of this linker segment was abolished by neutralizing the charged residues and is masked by the upstream linker region but not by the other variable (N and C terminus) segments of the protein. We propose the existence of electrostatic interactions between the polybasic linker segment and the channel inactivation machinery itself or its regulatory components that contribute to the inhibition of VDI by the ␤ 2a -subunit. The exposition of this segment appears to be regulated by the palmitoylated state of the N terminus of ␤ 2a .

EXPERIMENTAL PROCEDURES
Construction of cDNA and Protein Expression-cDNA encoding all Ca V ␤ 2a (Swiss-Prot accession number Q8VGC3-2, described elsewhere (27)) and Ca V ␤ 3xo (Swiss-Prot accession number Q91630) derivatives were subcloned by PCR methods into pRSET vector (Invitrogen) to include the N-terminal histidine tag and Xpress epitope tag. The proteins were overexpressed in bacteria and purified by metal-affinity chromatography, followed by size-exclusion chromatography onto a Superdex S-200 column (GE Healthcare), as described previously (27). Proteins were concentrated to 0.7-5.0 mg/ml by ultrafiltration, fast frozen, and stored at Ϫ80°C until use. For confocal fluorescence microscopy, the same cDNAs encoding for the different Ca V ␤ 2a derivatives used for protein expression were subcloned into pcDNA 3.1 vector and fused to YFP. All constructs were verified by DNA sequencing. The human form of the Ca V 2.3 subunit (Swiss-Prot accession number Q15878) used in this study has been previously described (28,29).
Western Blot Analysis-For Western blot analysis of ␤ 2a -derivative protein constructs, five noninjected control oocytes and five oocytes were injected with purified His-␤ 2a protein constructs 5 days before recording. Oocytes were transferred to a 1.5-ml tube and homogenized by pipetting up and down with 50 l of lysis buffer (100 mM sodium phosphate, pH 7.2), supplemented with 0.1% Triton X-100 and protease inhibitor mixture (Sigma). After incubation on ice for 15 min (15 s vortexing every 5 min), the samples were centrifuged twice at 12,000 rpm in a microcentrifuge for 10 min at 4°C. The supernatant was diluted with 5ϫ SDS loading buffer and loaded on a 12% acrylamide gel. After SDS-PAGE, the proteins were electrically transferred to ECL Plex membranes (GE Healthcare) for 1 h. Membranes were blocked using 3% BSA in TBS buffer (10 mM Tris, 150 mM NaCl, pH 7.5) and incubated with primary antibody diluted according to the manufacturer's instructions in TBS supplemented with 1% BSA. Three primary antibodies were used depending on the ␤-construct to be detected as follows: anti-calcium channel ␤ 2 -subunit antibody (Sigma) when the C terminus was present (␤ 2a C3S,C4S, ␤ 2a SH3-PBLK-GK ϩ CT, and ␤ 2a ⌬PBLK), anti-CACNB2 antibody (Abcam) only when the full linker sequence was present (␤ 2a SH3-LK-GK), and anti-Xpress antibody (Invitrogen) for the rest of the constructs (␤ 2a SH3-PBLK-GK, ␤ 2a SH3-PBLK-GK ϩ ⌵⌻ C3S,C4S , ␤ 2a SH3-PBLK Ala -GK, and ␤ 2a SH3-GK). Immunodetection was carried out using HRP-conjugated secondary antibody (Pierce Antibodies, Thermo Scientific) diluted 1:40,000 to 1:60,000 and a chemiluminescent detection kit (SuperSignal West Femto Chemiluminescent Substrate, Thermo Scientific). Membranes were exposed between 15 and 60 s into a Gene-Gnome chemiluminescence imaging system (SYNGENE).
Oocytes Injection and Electrophysiological Recordings-The cRNA encoding for Ca V 2.3 ␣ 1 -subunit was synthesized using a capping RNA transcription kit (mMESSAGE mMACHINE TM , Ambion) and injected into Xenopus laevis oocytes, as reported previously (30). To minimize cRNA degradation, the protein was injected 1 h after cRNA injection, using the same method as for cRNA injection. Electrophysiological recordings were performed using the cut-open oocyte technique (31) with a CA-1B amplifier (Dagan Corp., Minneapolis, MN), as described previously (27). Data acquisition and analysis were performed using the pCLAMP system and software (Axon Instruments Inc., Foster City, CA). Linear components were eliminated by P/Ϫ4 pre-pulse protocol. All experiments were carried out in at least two batches of X. laevis oocytes from two different frogs. The external solution contained (in mM), 10 Ba 2ϩ , 96 n-methylglucamine, and 10 HEPES, and the internal solution contained 120 n-methylglucamine, 10 EGTA, and 10 HEPES. Both solutions were adjusted to pH 7.0 with methanesulfonic acid.
Confocal Fluorescence Microscopy-Live cell confocal imaging was carried out with a ϫ40 oil immersion objective on a Leica inverted confocal microscope equipped with an argon ion laser using transiently transfected tsA201 cells. Cells were cultivated on a glass coverslip and imaged 18 -48 h after transfection using Lipofectamine TM (Invitrogen) according to the manufacturer's instructions. For excitation of YFP, the 514-nm laser was used, and the emitted light was monitored between 520 and 560 nm. ␤2a-YFP has been described elsewhere (32).

Co-injection of Full-length ␤-Protein with ␣ 1 -cRNA Fully Reconstitutes ␤-Subunit Modulation of Activation and Inactivation of Ca V 2.3 R-type Calcium Channel Expressed in Xenopus
Oocytes-Instead of injecting the cRNA encoding for Ca V ␣ 1 and Ca V ␤, i.e. the widespread strategy to study ␤-modulation of calcium channels, we injected ␤-derived constructs as purified proteins together with the cRNA encoding the ␣ 1 -subunit (27,29,30). With this strategy we were able to study isolated domains and short constructs of ␤ 2a that are not functional when injected as cRNA (29) but that regulate calcium channel function when injected as purified proteins. To confirm the integrity and folding of the recombinant proteins, all constructs used in this study were subjected to SDS-reducing PAGE and size-exclusion chromatography (supplemental Fig. S3). Fig. 1 illustrates the strategy followed in this study to investigate the effect of different ␤-subunit derivatives on VDI of R-type Ca V 2.3 channels. Protein constructs were injected into Xenopus oocytes 1 h later than the cRNA encoding the ␣ 1 -subunit, so as to avoid RNA degradation. Electrophysiological recordings were performed 5-6 days after cRNA/protein injection, corresponding to the time necessary for the ␣ 1 -subunit to be assembled in the plasma membrane (Fig. 1A).
As observed from earlier co-injection experiments of cRNA encoding ␣ 1 and ␤-subunits, ␤ 2a inhibits VDI of Ca V 2.3 channels expressed in Xenopus oocytes, and ␤ 3xo stimulates it (8,(33)(34)(35). This effect is reflected in much greater values of T 0.5 for Ca V 2.3/␤ 2a channel complexes than Ca V 2.3/␤ 3xo and in steadystate inactivation curves, exhibiting opposing voltage shifts with respect to the Ca V 2.3 ␣ 1 -subunit alone (Fig. 1, B and C, and Table 1). see Equation 1, where I max is the current at peak; I res is the noninactivating current; F is the Faraday's constant; R the universal constant of gases; t is temperature (298 K), and V m is the membrane voltage. V 1 , V 2 , z 1 , and z 2 are the parameters defining each Boltzmann component. V1 ⁄ 2 is then obtained by solving numerically I(V) ϭ I max /2. I res corresponds to the fraction of channels that Values are expressed as mean Ϯ S.E. The dashed line corresponds to T 0.5 for Ca V 2.3 channels expressed alone (in the absence of ␤-subunit). C, voltage-dependent inactivation curves from oocytes injected with Ca V 2.3 encoding cRNA and the indicated ␤-protein. The degree of inactivation was determined after a 10-s pre-pulse ranging from Ϫ120 to ϩ30 mV in 15-mV increments from a holding potential of Ϫ90 mV followed by short deactivation pulse to Ϫ90 mV and a 0.5-s test pulse at 0 mV. Continuous lines correspond to double Boltzmann distributions that best fit the experimental data for the indicated subunit combination. For comparison, the fit to the inactivation curve obtained with Ca V 2.3 expressed alone is shown by a dotted line. D, voltagedependent activation curves from oocytes injected as in C obtained from tail currents measured at Ϫ40 mV after a 70-ms pulse ranging from Ϫ50 to ϩ105 mV in 5-mV increments from a holding potential of Ϫ80 mV. Continuous lines correspond to the sum of two Boltzmann distributions that best fit the experimental data. The curve obtained with Ca V 2.3 expressed alone is shown by a dotted line.
do not inactivate at extreme positive voltages. Moreover, by shifting the steady-state inactivation curve to more depolarizing potentials, ␤ 2a also induces a residual noninactivating current component that is unique to this isoform (Fig. 1C). In contrast to their differences in the channel inactivation curve, ␤ 2a and ␤ 3xo shift the activation curve to the same extent and direction ( Fig. 1D and supplemental Table S1). This change in the midpoint of the activation curve toward more hyperpolarizing potentials is the signature of ␤-subunit modulation of HVA calcium channels (33,36), and we used it as a parameter to confirm the functional state of the protein construct within the oocytes. We conclude that the effect of ␤-subunit protein on Ca V 2.3 channels is undistinguishable from one of the cRNA sources when injected with Ca V ␣ 1 cRNA. This, however, is not the case for ␤-subunit protein injected at a later stage of Ca V ␣ 1 biogenesis (29).
Distal Linker Segment of Ca V ␤ 2a Comprising the Amino Acid Sequence HSKEKRMPFFKK Slows Down VDI of Ca V 2.3 Channels-Previous results from Qin et al. and Richards et al. (7,22) suggested that the linker joining SH3 and GK domains of the ␤-subunit plays a role in the regulation of VDI. However the variability in length and sequence of this region makes it difficult to define further the structural determinant of VDI inhibition. Eye inspection of the amino acid sequence of the linker region of ␤ 2a reveals a short polybasic segment at the boundary of the GK domain, comprising residues 202-213, hereby referred to as polybasic linker segment (PBLK, Fig. 2A) that is absent in ␤ 3xo . To isolate the potential impact of the polybasic linker segment on VDI regulation, we removed all other variable regions of ␤ 2a , including the N and C termini and the segment upstream PBLK that encompasses a long serine-rich sequence (PSLK, Fig. 2A), containing several predicted phosphorylation sites. The result is a construct that consists of SH3 and GK domains joined only by PBLK (␤ 2a SH3-PBLK-GK, Fig.  2B). We found that this construct confers the slow inactivation phenotype, despite lacking the palmitoylable NT segment that is commonly recognized as responsible for this effect.
To test if indeed the polybasic segment is responsible for this conferring phenotype, we generated a linker-less ␤-subunit construct, consisting of only SH3 and GK domains (␤ 2a SH3-GK, Fig. 2B) that has been proposed earlier to be the functional unit of the ␤-subunit (13,37). In contrast to ␤ 2a SH3-PBLK-GK, ␤ 2a SH3-GK confers a fast inactivating phenotype to Ca V 2.3 channel complexes, even faster than the channel alone ( Fig. 2C and Table 1). As performed for the full-length proteins (see Fig.  1D), the functional state of the ␤-derivative constructs 5 days following injection into oocytes was tested by measuring the activation curves of the different Ca V 2.3 channel complexes. Both ␤ 2a SH3-PBLK-GK and ␤ 2a SH3-GK produced a left shift in the activation curve of around 10 mV accompanied by an increase in the slope, and relative contribution of the first component of the sum of two Boltzmann distributions that fit the activation curves (supplemental Fig. S1 and supplemental Table S1). We also verified the integrity of the ␤-derivatives by Western blot analysis (Fig. 2B, inset). Although no signal was detected in lysates from noninjected oocytes, a strong band that migrated at the same position as the purified protein was visible in lysates from injected Xenopus oocytes showing that ␤-derivatives constructs were stable for at least 5 days within the oocyte (Fig. 2B, inset).
Although only ␤ 2a SH3-PBLK-GK slowed inactivation, both constructs shifted the voltage dependence of VDI toward more negative voltages, compared with Ca V 2.3 expressed alone. The voltage for 50% inactivation (V1 ⁄ 2 ) was shifted by about 8 mV to the left by the presence of ␤ 2a SH3-PBLK-GK and 20 mV by ␤ 2a SH3-GK ( Fig. 2D and Table 1). Nevertheless, as with wild-type ␤ 2a , a residual current can still be detected at the end of a 10-s pulse in Ca V 2.3/␤ 2a SH3-PBLK-GK expressing oocytes. These results identify the distal polybasic linker segment of Ca V ␤ 2a as a structural determinant for slowing down inactivation of Ca V 2.3 channels and reveal divergences in the modulation of kinetic and steady-state parameters.
Polybasic Linker Segment That Slows Down Inactivation Does Not Confer Membrane Anchoring-The interaction of membrane-associated proteins with lipid membranes can be driven by electrostatic interactions between positively charged residues and negatively charged lipid membranes (38). The basic nature of PBLK might provide attachment of the ␤-subunit to the plasma membrane, mimicking the mechanism proposed for palmitoyl-membrane-anchored ␤ 2a (17). To investigate the cellular location of ␤ 2a SH3-PBLK-GK, we fused it to YFP (␤ 2a SH3-PBLK-GK-YFP), expressed it in mammalian cells, and

.3 channels expressed alone or with the indicated ␤-protein
The different constructs are described in the text and figures. Parameters from the Ca V 2.3 alone topping the list is followed by the different subunit combinations in decreasing order for T 0.5 , which corresponds to the time for the current to return to 50% of its peak amplitude, evoked by a depolarizing pulse of 10 s to 0 mV. V1 ⁄ 2 corresponds to the voltage at which 50% of the channels are inactivated, estimated from the fit to a double Boltzmann distribution according to the Equation 1, which also defines I res . visualized it by confocal microscopy (Fig. 3). In clear contrast to wild-type ␤ 2a -YFP that is located at the plasma membrane, fluorescently labeled ␤ 2a SH3-PBLK-GK is distributed throughout the cytoplasm, closely resembling the distribution of nonpalmitoylable ␤ 2a C3S,C4S double mutant (Fig.  3). To rule out that the His tag used to purify the recombinant proteins may interfere with palmitoylation of ␤ 2a , we fused the His-tagged ␤ 2a sequence to YFP and expressed it in tsA201 cells. Confocal images of tsA201 cells expressing His-␤ 2a -YFP show that this protein is indeed targeted to the plasma membrane (supplemental Fig. S4). Thus, the polybasic linker confers slow inactivation phenotype by other means, not only providing a membrane anchoring domain to the protein.

Positively Charged Amino Acids within the Polybasic Linker Segment Are Required for Slowing Down Inactivation of Ca V 2.3
Channels-Because the most striking property of this segment is a relatively high density of positively charged residues, we rationalized that they might still be functionally relevant although not for surface membrane association. To test this idea, we substituted all positively charged residues within this segment by alanine (␤ 2a SH3-PBLK Ala -GK, see Fig. 4A). This ␤-derivative remains intact after 5 days of injection into Xenopus oocytes, as judged by Western blot analysis (Fig. 4A) and left shifts of the channel activation curve (supplemental Fig.  S1 and supplemental Table S1).
Channel complexes bearing ␤ 2a SH3-PBLK Ala -GK inactivates nearly as fast as the construct lacking this segment (Fig. 4, B and C, and Table 1). These results demonstrate that the basic residues within the PBLK segment are required for conferring a slow inactivation phenotype to nonmembrane-anchored ␤-subunit and suggest the involvement of electrostatic interactions in VDI modulation.
Linker Region Upstream from the Polybasic Linker Segment Masks Its Effect on VDI-It remains to be explained why the nonpalmitoylated ␤ 2a C3S,C4S double mutant (␤ 2a -C3S,C4S) containing the polybasic linker segment confers fast inactivation to Ca V 2.3 channels (Fig. 5, A and E) (17,29). This observation implies that other parts of the nonmembrane-anchored ␤-proteins counteract the effect of the polybasic linker segment. We followed a simple strategy to identify variable segments of the protein that may occlude the effect of the polybasic linker segment. This involved adding to ␤ 2a SH3-PBLK-GK to the rest of the variable regions, one at a time ( Fig. 5 and see also Adding NT or CT to ␤ 2a SH3-PBLK-GK did not interfere with the ability of the PBLK segment to confer a slow inactivation phenotype, and in the presence of either construct, the channel voltage dependence of inactivation is closer to the parent protein than to ␤ 2a C3S,C4S (Fig. 5, A-C and E and F, and  Table 1). In contrast, ␤-subunit carrying a full-length linker while lacking the N and C termini gives rise to relatively fast inactivating Ca V 2.3 channels, and the inactivation curve is shifted even further to the left than with ␤ 2a C3S,C4S, indicat-ing that the sequence upstream from the polybasic segment occludes its effect on VDI (Fig. 5, D-F, and Table 1).

Determinant for Modulation of Inactivation in ␤ 2a -Subunit
The idea that a post-translational modification may be involved in occluding a slow inactivation phenotype is appealing because it might explain why a nonmembrane-anchored ␤-subunit can slow inactivation when acutely applied to already expressed Ca V 2.3 channels (29). Because the linker region upstream from the polybasic segment contains several predicted phosphorylation sites, there is a possibility that the addition of phosphate groups would hinder potential electrostatic interactions that might, in turn, be responsible for VDI inhibition by the polybasic segment. We substituted all serine and threonine residues by alanine within PSLK in the ␤ 2a -C3S,C4S background, and no effect on T 0.5 or the inactivation curve was observed (supplemental Fig. S2 and supplemental Table S1). This indicates that phosphorylation events within the region upstream from the PBLK do not participate in VDI modulation.
Although the molecular mechanism, by which the upstream linker region hinders the effect of the distal polybasic segment, remains elusive, the above results provide a clear explanation why nonmembrane-anchored full-length proteins containing the polybasic segment do not manifest their phenotype.

Palmitoylated ␤ 2a Lacking the Polybasic Linker Segment Inhibits Voltage-dependent Inactivation of Ca V 2.3 Channel Complex to a Lesser Extent than Does the Wild-type Protein-It
is well accepted that membrane anchoring by itself suffices to confer slow inactivation to the Ca V 2.3 subunit. This idea is based on results showing that removal of the N-terminal double cysteine motif in ␤ 2a abolishes membrane localization and the slow inactivation-conferring phenotype that is rescued by artificially anchoring ␤ 2a C3S,C4S to the membrane (see also Figs. 3 and 5) (17,19). Within this framework, the functional relevance of PBLK on the membrane-anchored ␤ 2a -subunit comes into question. To provide an answer to this, we removed the PBLK from wild-type ␤ 2a containing the double cysteine motif (␤ 2a ⌬PBLK). This protein was also stable within the oocyte, as judged by Western blot analysis and its effect on the channel's voltage dependence of activation ( Fig. 6A and supplemental Table S1). Ca V 2.3/␤ 2a ⌬PBLK channels inactivated slightly faster than the Ca V 2.3-␤ 2a channel complex, and when fused to YFP, ␤ 2a ⌬PBLK still localized at the plasma membrane (Fig. 6,  A-C). Thus, dissociation of the membrane does not explain this  effect. Although the reduction in T 0.5 is relatively modest, it is also accompanied by a left shift in the voltage to reach 50% inactivation and a reduction in the noninactivating current ( Fig. 6D and Table 1). This result indicates that in the membrane-anchored ␤-subunit, the polybasic linker segment contributes to the inhibition of VDI.

DISCUSSION
The physiological relevance of the precise regulation of calcium influx appears reflected in the redundancy of elements that regulate inactivation of voltage-gated calcium channels. This added complexity barred a full mechanistic account of voltage-dependent inactivation in Ca V 2.x channels. A simple ball-and-chain model in which the membrane anchoring of the auxiliary subunit hinders the movement of the ball is rather simplistic and cannot explain why binding of a large nonmembrane-anchored subunit may facilitate the process. Although our present study does not provide a definitive answer to this question, it does however contribute to a better understanding of the modulation of inactivation by the regulatory ␤-subunit.
Four conclusions can be drawn from this work as follows: (i) there is an additional structural determinant in ␤ 2a isoform, the distal polybasic linker segment, besides the N-terminal palmitoylable region, that contributes to slowing down inactivation; (ii) the polybasic segment inhibits inactivation in a membraneanchoring independent manner and the basic residues within are required for this effect; (iii) the linker sequence upstream from the polybasic segment masks the contribution of the latter, and (iv) the effect of the palmitoylated N terminus of the  ␤-subunit predominates over the PBLK-mediated inhibition, but both act synergistically to inhibit VDI.
Work by others has also shown that altering the length and sequence of the linker impacts VDI (7,22,23). Qin et al. (7) were the first to identify the linker region as a secondary determinant of VDI. Richards et al. (22) reported that in the palmitoylation-deficient mutant, removal of the linker speeds inactivation, which is expected for a linker encompassing a slow inactivation conferring sequence, such as PBLK. Stotz and coworkers (23) transferred the linker and GK domain of ␤ 3 into ␤ 4 and found that VDI of Ca V 2.2 N-type calcium channel is accelerated. They also found that the N terminus of the ␤ 3 -subunit confers fast inactivation to Ca V 2.2 channels and that transferring this region to ␤ 4 is sufficient to accelerate VDI. Interestingly, this 14-amino acid region contains five negatively charged residues but no positively charged ones. Here, we propose that an electrostatic interaction between the polybasic linker and a putative acidic region within either the channel's inactivation machinery or some of its regulatory determinants slows VDI. An interesting concept is that acceleration of inactivation might involve also electrostatic interactions but of opposite signs, involving a positively charged receptor site in the pore-forming subunit.
We envisage that membrane anchoring not only contributes to immobilize the inactivation particle, as commonly accepted, but also exposes the polybasic distal linker segment that adds an additional component to VDI inhibition through an electrostatic repulsion of a putative inactivation particle (Fig. 7). In the absence of palmitoylation, as the protein is no longer anchored to the plasma membrane, the linker may fold over, occluding the positive charges of the PBLK and masking their functional association with a potential target within Ca V ␣ 1 . Membrane anchoring would give rise to an extended conformation of the linker that exposes the positively charged residues of the PBLK (Fig. 7).
Another explanation is that the polybasic distal linker segment has no direct action on the inactivation machinery but confers a permissive orientation to SH3/GK functional unit. This idea appears unlikely in light of our data, showing that several ␤-constructs bearing no linker or linker segments that differ in length and sequence composition (␤ 3xo , ␤ 2a SH3-GK, ␤ 2a SH3-LK-GK, ␤ 2a SH3-PBLK Ala -GK) confer fast inactivation phenotype. Moreover, crystallographic studies did not reveal any structural differences between the ␤-subunit alone and complexes to the peptide corresponding to the binding site in the ␣ 1 -subunit. Unless the relative orientation of SH3/GK domain is different in vivo, one should conclude that there are no major conformational changes associated with the binding of the ␤-subunit to the channel and that relative orientation of SH3 and GK is irrelevant to inactivation.
There is, to date, no direct structural data on the variable regions of ␤-subunit. Here, we show that N-and C-terminal variable segments do not interfere with the effect of PBLK on inactivation, and we conclude that such regions do not physically interact with the linker.
An unexpected observation from this study is that constructs capable of inducing a large increase in T 0.5 , such as ␤ 2a SH3-PBLK-GK, produced a modest change in V1 ⁄ 2 . Thus, slowing the kinetics of inactivation is not always accompanied by a shift in the voltage dependence of inactivation toward positive voltages, as one would expect if VDI inhibition occurs by disfavoring inactivated states. Moreover, ␤ 2a SH3-PBLK-GK that confers a rather slow inactivation phenotype shifts the steady-state inactivation curve to the left when compared with Ca V 2.3 alone, as if the equilibrium for the occupation of inactivated states would be favored by the presence of this construct. This suggests that the ␤ 2a -subunit and its derivatives influence not only the stability of the inactivated state but also the energy barriers that the channel must overcome in its way to inactivation.
All the available data show that shifting the inactivation curve toward more depolarizing potentials and provoking noninactivating components at positive voltages require a membrane-anchored ␤-subunit. Consistent with this view, our studies also show that only palmitoylable ␤ 2a displays this behavior. In contrast, the ability to modulate T 0.5 does not require palmitoylation, leading us to conclude that although membrane anchoring destabilizes the channel inactive state, the linker is geared at regulating the movement of the inactivation gate. This suggests that while the channel transits to the inactive conformation, PBLK comes transiently into close contact with the inactivation machinery.
In a previous work, we showed that acute exposure of Ca V 2.3 channels to the GK domain of ␤ 2a alone was sufficient to slow down inactivation when applied to the ␣ 1 -subunit already assembled into the plasma membrane (29). Unfortunately, the GK domain turned out to be unstable, and we could not test whether a long term exposure of this domain, i.e. the cRNA/protein co-injection strategy, would be sufficient to confer fast inactivation. We reported that ␤-subunits accelerating VDI acquired this phenotype only when co-injecting the protein with ␣ 1 -cRNA. In light of our present results that short constructs also accelerate VDI when co-injected, we rationalized that this could be the case for GK. In favor of this possibility is that the SH3 domain does not have an impact on channel function but on surface expression (25,29). In this scenario, GK would become competent to accelerate FIGURE 7. Model for modulation of voltage-dependent inactivation by ␤ 2a -subunit. Two molecular determinants of ␤-subunit control voltage-dependent inactivation as follows: palmitoylated N terminus and the polybasic linker segment (shown in "red"). The ␤-subunit binds through the GK domain to the conserved ␣-interaction domain in the Ca V ␣ 1 subunit. According to current models, anchoring of ␤ 2a to the membrane restricts the movement of the inactivation particle, and based on our present data, it exposes the polybasic segment that further inhibits VDI (left panel). In the absence of a palmitoylated N terminus, the region upstream from the polybasic linker segment occludes its functional association with the inactivation machinery (right panel).
inactivation following a post-translational modification, yet to be established.
In summary, we were able to identify a new determinant for the inhibition of inactivation within the linker joining the GK and SH3 domains present in all ␤ 2 -subunits and conserved across different species. This linker has already been pointed out as a potential regulator for inactivation (7,22,23). Here, we defined a short segment located in this region that includes positively charged residues. These basic amino acids are required for modulation of VDI through a mechanism that remains to be elucidated and that may involve electrostatic interactions. The functional capability of this segment in nonmembrane-anchored ␤ 2 -subunits remains elusive. One possibility is that the accessibility of PBLK positively charged residues may indeed come about through post-translational modifications of the protein.