The N-terminal Domain Tethers the Voltage-gated Calcium Channel β2e-subunit to the Plasma Membrane via Electrostatic and Hydrophobic Interactions*

Background: Membrane recruitment of the calcium channel β2e-subunit remains elusive. Results: Charge neutralization of β2e N-terminal residues abolishes its lipid-binding and surface-targeting abilities and its slow inactivation-conferring phenotype. Conclusion: The β2e-subunit anchors to the membrane via electrostatic interactions likely involving an N-terminal in-plane α-helix and a tryptophan side-chain penetration. Significance: A novel mechanism for β-subunit membrane-anchoring supporting slow inactivation is presented. The β-subunit associates with the α1 pore-forming subunit of high voltage-activated calcium channels and modulates several aspects of ion conduction. Four β-subunits are encoded by four different genes with multiple splice variants. Only two members of this family, β2a and β2e, associate with the plasma membrane in the absence of the α1-subunit. Palmitoylation on a di-cysteine motif located at the N terminus of β2a promotes membrane targeting and correlates with the unique ability of this protein to slow down inactivation. In contrast, the mechanism by which β2e anchors to the plasma membrane remains elusive. Here, we identified an N-terminal segment in β2e encompassing a cluster of positively charged residues, which is strictly required for membrane anchoring, and when transferred to the cytoplasmic β1b isoform it confers membrane localization to the latter. In the presence of negatively charged phospholipid vesicles, this segment binds to acidic liposomes dependently on the ionic strength, and the intrinsic fluorescence emission maxima of its single tryptophan blue shifts considerably. Simultaneous substitution of more than two basic residues impairs membrane targeting. Coexpression of the fast inactivating R-type calcium channels with wild-type β2e, but not with a β2e membrane association-deficient mutant, slows down inactivation. We propose that a predicted α-helix within this domain orienting parallel to the membrane tethers the β2e-subunit to the lipid bilayer via electrostatic interactions. Penetration of the tryptophan side chain into the lipidic core stabilizes the membrane-bound conformation. This constitutes a new mechanism for membrane anchoring among the β-subunit family that also sustains slowed inactivation.

The ␣ 1 -subunit (Ca V ␣ 1 ) of HVA 2 calcium channels encompasses the ion conduction pathway, the voltage sensor, and the multiple intracellular domains that provide sites of interaction for regulatory proteins (1). One of these proteins is the ␤-subunit (Ca V ␤) that binds to a highly conserved site among HVA channels and modulates several aspects of channel function, including surface expression and biophysical properties (2,3). Ca V ␤s are special members of the membrane-associated guanylate kinase family of proteins containing only two of the three canonical domains shared by this group as follows: an Src homology 3 (SH3) and a guanylate kinase domain (4 -6). In contrast to these highly conserved domains, the linker region between them as well as the N-and C-terminal flanking segments (HOOK, NT, and CT, respectively, Fig. 1) vary in sequence and length among the four ␤-subunit isoforms (␤ 1 to ␤ 4 ) identified until now. Each ␤ subtype is encoded by a different gene that generates multiple splice variants giving rise to functionally different proteins. An extreme case occurs among the ␤ 2 splice variants (␤ 2a to ␤ 2e ) that, despite their sequence similarity, have different subcellular localizations and even have opposing regulatory effects on calcium currents (7,8). ␤ 2 variants are between 604 and 655 amino acids long and they differ only in their most proximal N-terminal segment, hereafter referred to as NT v (the subscript v stands for variable, Fig. 1). The subtypes 2a and 2e are the only members of the Ca V ␤ family that are targeted to the plasma membrane in the absence of the ␣ 1 pore-forming subunit, and when associated with it, they inhibit the voltage-dependent inactivation of the channel complex (7,9,10). Coexpression of the fast-inactivating R-type calcium channel with ␤ 2a in Xenopus laevis oocytes or mamma-lian cell lines increases considerably the time needed for the current to return to 50% of its peak amplitude and shifts the steady-state inactivation curve toward more positive potentials as compared with channels composed of cytosolic, nonmembrane-associated ␤-isoforms (10 -13). These unique modulatory properties of ␤ 2a rely on two continuous N-terminal cysteine residues that undergo palmitoylation and target the protein to the plasma membrane (9,14,15). Substitution of these two residues disrupts membrane localization as well as the slow inactivation-conferring phenotype of ␤ 2a (16). The prevalent view is that membrane anchoring restricts the free movement of the channel inactivation gate, preventing it from closing. Colecraft and co-workers (7) showed that Ca V 1.2 L-type calcium channels also inactivate more slowly in the presence of the two membrane-anchored ␤ 2 variants as compared with their fully cytosolic distributed counterparts, ␤ 2b , ␤ 2c , and ␤ 2d . In contrast to ␤ 2a , the mechanism by which the ␤ 2e -subunit is targeted to the plasma membrane remains to be elucidated. Here, we exchanged the variable N-terminal region between the ␤ 2e and the ␤ 1b isoform, which is distributed entirely in the cytoplasm, and we showed that this maneuver is sufficient to target the ␤ 1b chimera to the plasma membrane. Moreover, the same N-terminal segment binds in vitro to negatively charged liposomes in an ionic strength-dependent manner. ␤ 2e NT v includes seven positively charged residues and a net charge of ϩ4 at physiological pH. We studied the role of these basic amino acids in membrane association and found that substitution of these residues by alanine inhibits the targeting of the protein to the plasma membrane and the in vitro interaction with anionic lipids. Coexpression of wild-type ␤ 2e with the fast inactivating Ca V ␣ 1 encoding the Ca V 2.3 R-type calcium channel results in channel complexes that inactivate significantly more slowly than Ca V 2.3 expressed with a ␤ 2e mutant in which all positively charged residues within NT v were replaced by alanine. Our results demonstrate that the positively charged N-terminal variable segment of ␤ 2e is responsible for tethering the protein to the plasma membrane via electrostatic interactions and support the model suggesting that membrane anchoring underlies the slow inactivation-conferring phenotype.

EXPERIMENTAL PROCEDURES
cDNA Constructs-Full-length rat ␤ 2e (UniProtKB accession number Q8VGC3-4) and rat ␤ 1b (UniProtKB accession number P54283) cDNAs were subcloned into pcDNA 3.1 vector and fused at the C terminus to YFP (for confocal imaging) or monomeric red fluorescence protein (for electrophysiology) using standard overlapping PCR methods. ␤ 2e mutants were generated by PCR using a common antisense and specific sense primers bearing the corresponding mutation. The chimeric ␤ 2e NT v / ␤ 1b was also generated using overlapping PCR methods by replacing residues 1-57 of ␤ 1b with residues 1-23 of ␤ 2e . The cDNAs encoding wild-type ␤ 2e NT v (amino acids 1-23) or mutant versions were subcloned into PGEX 6P-1 vector using standard PCR methods to produce GST-␤ 2e NT v domain fusion constructs. YFP was fused to the N terminus of Ca V 2.3-coding region (UniProtKB accession number Q15878) using standard PCR methods and subcloned into pRcCMV vector.
Transfection of tsA201 Cells-For confocal microscopy, cells were grown on plastic -dishes (Ibidi) in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and L-glutamine (2 mM) and incubated in a 5% CO 2 -humidified atmosphere. Cells were transiently transfected using Lipofectamine 2000 TM (Invitrogen) according to the manufacturer's instructions with the corresponding plasmids encoding the different ␤ 2e derivatives. For whole-cell patch clamp recordings, cells were cotransfected with YFP-Ca V 2.3 and the corresponding ␤ 2e fused to monomeric red fluorescence protein-encoding plasmids.
Confocal Microscopy and Colocalization Analysis-Confocal images were acquired 24 -30 h after transfection on living cells using a 63ϫ oil immersion objective on a Leica inverted confocal microscope. For colocalization analysis, prior to imaging the cells were stained with the plasma membrane marker Cell-Mask TM (Invitrogen) according to the manufacturer's instructions. YFP fusion proteins were excited with a 514-nm laser, and emission was monitored at 520 -550 nm, and the plasma membrane stain was excited at 633 nm and monitored at 640 -700 nm. Quantification of the degree of colocalization between fluorescently labeled ␤-subunit and the plasma membrane marker was done by calculating the Manders coefficient (MC) using the JACoP plugin (17) embedded in ImageJ 1.44p software (National Institutes of Health, Bethesda) (18). At least five different fields of view (in total 100 -200 transfected cells) for each ␤ 2e derivative from at least two different transfections were analyzed. The Manders coefficient represents the fraction of pixels in the YFP channel that overlaps with some signal in the CellMask TM (red) detection channel. For comparison of samples, a two-tailed Student's t test was used. Values are expressed as mean Ϯ S.E.
Liposome Preparation-Phospholipid unilamellar vesicles were prepared by extrusion of multilamellar vesicles through 100-nm pore size polycarbonate filters using a handheld extruder (LiposoFast, Avestin, Inc). 1-Palmitoyl-2-oleoyl-snglycerophosphocholine (PC, Avanti 840034) either alone or combined with 1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-Lserine (PS, Avanti 840034) in a 1:3 (PS/PC) molar ratio was dried under a stream of nitrogen. Any residual solvent was removed under vacuum. The resulting lipid film was rehydrated to reach a final concentration of 1 mM with a buffer containing 50 mM HEPES pH 7.5 supplemented with the indicated NaCl concentration and then subjected to three-to five freeze-thaw cycles using liquid nitrogen and a water bath at 37°C. The lipid suspension was either extruded to obtain uniformly sized (about 100 nm diameter) large unilamellar vesicles (LUVs) for protein-liposome cosedimentation assays or sonicated at RT using a bath sonicator to obtain small unilamellar vesicles (SUVs) for tryptophan fluorescence studies (Morrissey Lab Protocols, University of Illinois) (19,20).
Protein-Liposome Cosedimentation Assays-Cosedimentation assays were performed as described previously (21). Briefly, 2.0 g of protein were incubated with LUVs in 50 mM HEPES pH 7.5 buffer supplemented with the indicated NaCl concentration (total volume 100 l) for 15 min at room temperature. Samples were ultracentrifuged at 150,000 ϫ g for 40 min at 4°C using a TLA 120.2 rotor (Beckman Instruments). Pellets were resuspended in the same volume as the supernatant, and equal volumes of each fraction from each sample were resolved by SDS-PAGE (12%). The quantification of proteins present in the different fractions was performed by densitometry of Coomassie Blue-stained gels using ImageJ software (18). The fraction of the protein bound to liposomes (proteinbound) was calculated from the quotient between the fraction of nonaggregated protein in the pellet and the pool of nonaggregated protein as shown in Equation 1,

Protein-boundϭ
Fraction nonaggregated protein in pellet Fraction nonaggregated protein where PP(Lip) is the fraction of protein in the pellet in the presence of lipids, and PP(Ø Lip) is the fraction of protein in the pellet in the absence of liposomes (corresponding to the aggregated protein). Tryptophan Fluorescence Measurements-Measurements were performed on a FluoroLog 3 spectrofluorometer (HORIBA Jobin Yvon) using 1-2 M of a synthetic peptide (Eurogentec) containing the N-terminal region of ␤ 2e ( 1 MKATWIRLLK-RAKGERLKDSD 21 ) dissolved in different concentrations of SUVs in 50 mM HEPES pH 7.5 buffer supplemented with the indicated NaCl concentration. After combining the reagents in a 1-cm cuvette, the sample was mixed by pipetting and allowed to settle for ϳ5 min. Upon excitation of the samples at 280 nm, the emission spectrum was recorded in front face excitation mode from 290 to 440 or 300 to 440 nm. At each wavelength, the signal was integrated over 0.2 s. The bandwidth was 6 -8 nm on the excitation side and 1-2.5 nm on the emission side to optimize the signal levels. Data analysis was performed using Origin (OriginLab) and SigmaPlot (Systat Software). To infer the wavelength of maximal emission, a log-normal distribution was fitted to the spectra (22). For the quantification of binding of vesicles to the peptide, it was assumed that the fluorescence emission results from a two-state equilibrium between unbound and bound peptide. It follows that every measured spectrum is a linear combination of two spectra. Accordingly, the fraction of peptide that is bound to lipids was determined by a deconvolution of the spectral data into the spectra resulting from experiments with 0 and 1.500 M lipid concentration, corresponding to free and fully lipid-bound peptide, respectively (23). The resulting data were fitted by a Hill function.
Protein Purification-Wild-type ␤ 2e and ␤ 2e NT v (7ϩ less), both fused to a hexahistidine tag, and the GST fusion proteins were expressed in bacteria and purified using affinity and sizeexclusion chromatography as described previously (11).
Protein-Protein Binding Assays-The highly conserved ␤binding site among all ␣ 1 pore-forming subunits of high voltage gated calcium channels known the as ␣ 1 interaction domain (AID) was fused to GST (GST-AID). Either 4 g of GST-AID or of GST alone (in control experiments) was pre-coupled to glutathione beads (11). Coupled beads were then incubated with 4 g of either ␤ 2e or ␤ 2e NT v (7ϩ less) for 1 h at room temperature. After several washes, the bound proteins were eluted with SDS-loading buffer and resolved on SDS-PAGE. The gel was stained with Coomassie Blue.

RESULTS
N Terminus of ␤ 2e Is Responsible for Targeting the Protein to the Plasma Membrane-Because ␤ 2 variants differ only in the most proximal amino acid sequence within the N-terminal region, it is likely that the differences in their subcellular loca-tion are caused by signals within this part of the protein. To assess the impact of ␤ 2e NT v on membrane anchoring, we deleted this segment (residues 2-23) from the rat ␤ 2e . Wildtype ␤ 2e , the deletion mutant, and the entire cytosolic ␤ 1b isoform were expressed in tsA201 cells as YFP fusion proteins for confocal imaging analysis. Plasma membrane localization of the YFP fusion proteins was analyzed by staining the transfected cells with the CellMask TM plasma membrane marker and quantifying their degree of colocalization using the MC as described under "Experimental Procedures." For the plasma membrane-associated ␤ 2e -YFP (7), we obtained an MC of 0.48 Ϯ 0.01, for YFP alone, which distributes homogeneously within the cell, the value was 0.10 Ϯ 0.01 (t test; p Ͻ 0.01, Fig. 2). After deletion of the NT v segment, the degree of colocalization decreased significantly (MC ϭ 0.14 Ϯ 0.01, t test; p Ͻ 0.01, Fig.  2). We also exchanged the N-terminal variable regions between ␤ 2e and the entire cytosolic ␤ 1b -subunit to generate a chimeric protein encompassing ␤ 2e NT v in the ␤ 1b background (␤ 2e NT v / ␤ 1b ). This maneuver increases MC from 0.12 Ϯ 0.01 for ␤ 1b to 0.35 Ϯ 0.03 for the chimeric construct (t test; p Ͻ 0.01, Fig. 2B), a magnitude approaching that of wild-type ␤ 2e . This indicates that substitution of the N-terminal segment of ␤ 1b by ␤ 2e NT v transfers to the former the capability to associate with the plasma membrane.
To exclude the possibility that the alterations in protein sorting were due to expression artifacts or degradation, crude homogenates from cells expressing the different fluorescent proteins were run on SDS-PAGE and visualized by fluorescence scanning (Fig. 2A, bottom panels). The results show that all YFP fusion proteins were expressed in their full length indicating that the changes observed in their subcellular location are associated with the absence or presence of ␤ 2e NT v . Together, these results demonstrate that ␤ 2e NT v is the molecular determinant for targeting the protein to the plasma membrane.
The Variable N-terminal Segment of ␤ 2e Associates in an Ionic Strength-dependent Manner with Phospholipid Vesicles-Peripheral membrane proteins commonly associate with the plasma membrane via covalent lipid modification of specific residues and/or via electrostatic interactions between positively charged amino acid residues and negatively charged phospholipids (24). Substitution of the single cysteine residue in ␤ 2e NT v by alanine (␤ 2e C23A) does not impair the relocation FIGURE 2. The Variable N-terminal segment of the ␤ 2e -subunit is responsible for targeting the protein to the plasma membrane. A, confocal fluorescence images of tsA201 cells expressing the indicated protein constructs fused to YFP as follows: full-length ␤ 2e (␤ 2e ); ␤ 2e lacking NT V (␤ 2e ⌬NT V ); full-length ␤ 1b (␤ 1b ); chimeric protein encompassing NT V of ␤ 2e in a ␤ 1b background (␤ 2e NT V /␤ 1b ). YFP moiety alone was used as control protein. Prior to visualization, cells were stained with the red plasma membrane marker CellMask TM . Images from ␤ 2e fusion proteins, green channel (panels a-e); plasma membranes, red channel (panels f-j); and merge (panels k-o) are shown. Scale bar, 15 m and is valid for all images. The bottom panels show crude lysates from cells expressing the indicated construct resolved by SDS-PAGE and visualized by fluorescence scanning. Numbers denote the molecular mass of standard proteins. The migration of the fusion constructs is consistent with their predicted molecular mass. B, bar plot of the colocalization analysis between ␤ 2e derivatives and the plasma membrane marker according to Manders coefficient. Values are expressed as mean Ϯ S.E.
of the protein to the plasma membrane (Fig. 3). The degree of colocalization between this mutant and the plasma membrane marker is slightly higher than the one estimated for the wildtype protein (MC ϭ 0.62 Ϯ 0.06, Fig. 3B). We conclude that palmitoylation is not responsible for the membrane association of this subunit, as is the case for ␤ 2a (14). ␤ 2e NT v contains no predictable additional protein modification sites such as prenylation or myristoylation but is enriched in positively charged residues. Therefore, we investigated the role of electrostatic interactions in the membrane association of ␤ 2e using proteinliposome cosedimentation assays. ␤ 2e NT v was fused to GST (GST-␤ 2e NT v ), expressed in bacteria, purified, and incubated at different ionic strengths with LUVs containing either negatively charged (PS/PC) or neutral (PC) phospholipids. After centrifugation, the supernatant and pellet fractions containing lipid-free and lipid-associated protein, respectively, were analyzed using SDS-PAGE (Fig. 4A). The presence of lipids in the supernatant and in the pellet fractions was confirmed in control experiments where a fluorescently labeled PS analog (7-nitro-2-1,3-benzoxadiazol-4-yl-PS) was added during the preparation of liposomes. The majority of the lipids was found in the pellet fraction at the different NaCl concentrations tested (99 Ϯ 1, 96 Ϯ 2, and 91 Ϯ 4% in 50, 300, and 400 mM NaCl, respectively).
To discard the possibility that the protein observed in the pellet arises from an interaction between GST and the lipids, parallel experiments were performed in which GST-␤ 2e NT v was replaced by GST moiety alone. GST is virtually excluded from the pellet in all conditions after centrifugation (Fig. 4A).
A lower molecular band copurifies with GST-␤ 2e NT v and distributes almost evenly in the supernatant fraction (Fig. 4A). In Western blot analysis, this protein was recognized by an anti-GST antibody (Fig. 4A) indicating that it is a degradation product that may lack completely or partially the 23-amino acid residues of ␤ 2e NT v . Although unplanned, this protein provided an extra internal control for the assay.
A fraction of the GST-␤ 2e NT v is found in the pellet in the absence of phospholipids, indicating a degree of aggregation of the fusion protein arising from ␤ 2e NT v moiety. Despite this, quantification of the protein bound to the liposomes demonstrates that at relatively low ionic strength (50 mM NaCl) a significantly higher amount of protein cosediments with negatively charged LUVs (PS/PC) than with neutral PC liposomes   (Fig. 4B). Moreover, association of GST-␤ 2e NT v with PS/PC liposomes shows more acute dependence on the salt concentration (Fig. 4B). Increasing the ionic strength up 400 mM NaCl results in a persistent decrease in the quantity of GST-␤ 2e NT v found in the pellet with PS/PC liposomes that is not observed in the presence of PC liposomes. Altogether, these results indicate that electrostatic interactions play an important role in the binding of ␤ 2e NT v to anionic liposomes.
Positively Charged Residues within ␤ 2e NT v Are Required for Targeting the Protein to the Plasma Membrane-The binding of ␤ 2e NT v to negatively charged phospholipid vesicles implies that positively charged residues within this segment are key players. To test this, we generated several ␤ 2e constructs bearing simultaneous substitutions of lysine and arginine residues within ␤ 2e NT v , in which alanine replaced either all seven (␤ 2e NT v (7ϩ less)), four (␤ 2e NT v (4ϩ less)), or two basic residues at once (␤ 2e R7A,K10A and ␤ 2e K2A,R16A). Quantification of the plasma membrane localization of these mutant proteins was done as in Fig. 2. Substitution of all seven positively charged residues decreased dramatically the degree of colocalization as compared with the wild-type protein (MC ϭ 0.13 Ϯ 0.01 for ␤ 2e NT v (7ϩ less), t test; p Ͻ 0.01, Fig. 5B). The MC value obtained for ␤ 2e NT v (7ϩ less) is comparable with the one estimated for YFP alone indicating a loss of surface membrane targeting. Neutralization of four or two basic residues also leads to a significant decrease in the plasma membrane localization of the protein (MC ϭ 0.19 Ϯ 0.01 for ␤ 2e NT v (4ϩ less), 0.22 Ϯ 0.02 for ␤ 2e R7A,K10A and 0.20 Ϯ 0.02 for ␤ 2e K2A,R16A, t test; p Ͻ FIGURE 5. Charge-neutralizing mutations of the basic residues within ␤ 2e NT v impair the association of the protein with the plasma membrane. A, confocal fluorescence images of tsA201 cells expressing ␤ 2e mutants bearing the indicated simultaneous amino acid substitutions of lysine and arginine residues within NT v as follows: NT v (7ϩ less), seven lysine and arginine residues at NT v were substituted by alanine; NT v (4ϩ less), four basic residues were replaced with alanine (Arg 11 , Lys 13 , Arg 16 , and Lys 18 ), and R7A,K10A, K2A,R16A, R7K,K10R, and K2R,R16K. ␤ 2e mutant proteins (panels a-f), plasma membranes (panels g-l), and merge images (panels m-r) visualized as in Fig Fig. 5, A and B). However, the charge-conserving mutations of the double mutant proteins at the same amino acid positions (␤ 2e R7K,K10R and ␤ 2e K2R,R16K, Fig. 5, A and B) preserve plasma membrane localization, demonstrating that the charge of the residue is the determinant factor in targeting the protein to the surface membrane.
We introduced the double mutations R7A,K10A and R7K, K10R into the GST-␤ 2e NT v fusion proteins (GST-␤ 2e NT v R7A,K10A and GST-␤ 2e NT v R7K,K10R, respectively) to investigate the correlation between surface membrane targeting and liposome binding. The charge neutralizing mutations significantly reduced the fraction of protein associated with negatively charged liposomes, whereas the ability of the charge conserving mutant to bind to PS/PC liposomes is preserved to a high extent as compared with wild-type ␤ 2e NT v (Fig. 5, C and  D). These results suggest that the loss of plasma membrane localization upon charge neutralization is mainly due to the impaired ability of the protein to establish electrostatic interactions with anionic phospholipids.
Single point mutations of all arginine or lysine residues within ␤ 2e NT v to alanine moderately decrease the degree of colocalization with the plasma membrane marker compared with the WT protein (Fig. 6, A and B). The values of MC ranged between 0.36 Ϯ 0.01 and 0.42 Ϯ 0.02 for all proteins bearing a single residue mutation except for K2A mutant with an MC equal to 0.28 Ϯ 0.02 (Fig.  6B). This suggests that Lys 2 is involved in a more intimate proteinmembrane interaction.

Fluorescence Emission Maxima of the Trp 5 Blue Shifts in the Presence of Anionic Liposomes Indicating Penetration of This
Residue into the Lipid Bilayer-Tryptophan residues in peripheral membrane proteins are commonly found in association with the membrane's hydrophobic core (25). We took advantage of the presence of a single tryptophan residue at position 5 (Trp 5 ) within ␤ 2e NT v to study the association of this segment with lipids using intrinsic fluorescence measurements. At first, we substituted this residue by alanine in ␤ 2e -YFP and found that this mutation results in a significant decrease in the degree of colocalization with the plasma membrane marker (MC ϭ 0.16 Ϯ 0.01) compared with WT (Fig. 7, A and B). However, the binding of ␤ 2e NT v W5A to negatively charged liposomes under low salt conditions (50 mM NaCl) is only slightly diminished as compared with wild-type ␤ 2e NT v , although association with neutral liposomes is significantly impaired (Fig. 7, C and D). This indicates that Trp 5 may indeed contribute directly to lipid binding through hydrophobic interactions, which will be more relevant at the relatively higher ionic strength of the cellular environment.
At a low salt concentration, incubation of a synthetic peptide corresponding to ␤ 2e NT v (residues 1-21) with increasing concentrations of negatively charged PS/PC SUVs induced a pronounced blue shift in Trp 5 fluorescence emission maximum with a concomitant increase in the fluorescence intensity (Fig.  7E). The observed shift in emission maximum from 346 nm in the absence of vesicles to 323 nm in the presence of saturating PS/PC concentrations (Fig. 7F) indicates that this residue changes from an aqueous into an essentially hydrophobic environment (e.g. the hydrocarbon core of the membrane) (see also "Discussion") (26). The fit of a Hill function to the lipid concentration dependence of the fraction of bound peptide at low ionic strength yields a K D of 182 Ϯ 34 M (Fig. 7G).
At higher ionic strength (400 mM NaCl), full binding was not reached with the employed lipid concentrations (Fig. 7F) indi- FIGURE 6. Individual substitutions of positively charged residues within ␤ 2e NT v differentially alter plasma membrane targeting. A, confocal fluorescence images of tsA201 cells expressing the indicated ␤ 2e mutations at single lysine or arginine residues within ␤ 2e NT v . ␤ 2e single-point mutants (panels a-f), plasma membranes (panels g-l), and merged images (panels m-r) are visualized as in Fig. 2A. Scale bar represents 15 m and is valid for all images. B, colocalization analysis and dashed lines as in Fig. 5B. cating a decrease of the binding affinity. This finding is consis 2e NT v to PS/PC liposomes observed in cosedimentation assays (see Fig. 4). In the presence of neutral PC vesicles, virtually no changes in the intrinsic Trp 5 fluorescence were monitored at all salt concentrations tested (Fig. 7F). In contrast, we observed a small fraction of the GST-␤ 2e NT v cosedimenting with PC vesicles (Fig. 4). Although the cause of this difference is unclear, it is worth noting that fluorescence measurements were performed with SUVs, rather than LUVs, to minimize differential light-scattering effects (19). Altogether these results indicate that Trp 5 penetrates into the plasma membrane when electrostatic interactions between the positively charged residues and the anionic lipids are allowed.
Plasma Membrane Anchoring of ␤ 2e Regulates the Rate of Inactivation of R-type Calcium Channels-Coexpression of Ca v 2.3 with the membrane-anchored ␤ 2a in tsA201 cells and Xenopus oocytes induces a dramatic increase in the time needed for the current to return to 50% of its peak amplitude (hereafter referred to as t 0.5 ) and shifts the steady-state inactivation curve toward more depolarizing voltages with respect to the Ca V 2.3 coexpressed with cytosolic ␤-isoforms (␤ 1b , ␤ 3 , and ␤ 4 ) (10 -13). Here, we compared the effects of the wild-type ␤ 2e -subunit and the membrane association-deficient mutant containing no basic residues within ␤ 2e NT v (␤ 2e NT v (7ϩ less), see Fig. 5, A and B) on Ca V 2.3 channel complexes expressed in tsA201 cells.
We tested whether substitution of all positively charged residues in ␤ 2e NT v interferes with the ability of the protein to bind to the ␣ 1 -subunit. For this purpose, the highly conserved binding site in all ␣ 1 -subunits of HVA, referred as to AID (27), was fused to GST (GST-AID) and used as bait protein in a standard in vitro binding assay. Either wild-type ␤ 2e or the ␤ 2e NT v (7ϩ less) were added as prey. Control binding reactions were performed in parallel using GST alone as bait. Both wildtype ␤ 2e and the mutant protein bind to the GST-AID to a comparable extent (Fig. 8A).
Coexpression of Ca V 2.3 with the nonmembrane-associated ␤ 2e NT v (7ϩ less) mutant leads to an 8-fold decrease in t 0.5 at 0 mV as compared with Ca V 2.3 coexpressed with wild-type ␤ 2e (Fig. 8, B and C). Moreover, the steady-state inactivation curve from cells expressing Ca V 2.3 and ␤ 2e NT v (7ϩ less) complexes is shifted to more negative potentials with respect to cells transfected with Ca V 2.3 and ␤ 2e (Fig. 8D). Half-inactivation voltages (V 0.5 ) obtained from the fit to a Boltzmann distribution for Ca V 2.3 coexpressed with ␤ 2e WT and mutant were similar to those previously reported for membrane-anchored ␤ 2a and the cytosolic ␤ 1b , respectively (12).
These results imply that different mechanisms of ␤-subunit membrane association lead to a functional analogy in the inhibition of the voltage dependence of inactivation.

DISCUSSION
Proteins use diverse mechanisms to associate with lipid membranes, including electrostatic interactions between protein segments enriched in positively charged amino acid residues and anionic lipids as well as hydrophobic interactions between the membrane lipids and nonpolar regions of the pro-tein and/or covalently attached lipid anchors (24). Here, we identified a short domain located at the N-terminal region of ␤ 2e that is responsible for tethering the protein to the plasma membrane. This domain consists of several positively charged residues distributed along 23 amino acids and hydrophobic residues intercalated between them. A consensus secondary structure prediction program that integrates several methods based on different principles (28) calculated a helical structure within residues 2-13. Most notably, within this putative helix a short in-plane membrane anchor segment was predicted (Fig. 9A) (29). This region is highly conserved across highly divergent species suggesting a conserved structure-function (Fig. 9A). Alanine scanning mutagenesis revealed that Lys 2 has a stronger effect on membrane affinity than the other residues (see Fig. 6). Lys 2 maps to the same side of a helical wheel plot together with the nonpolar Trp 5 and Leu 9 residues (Fig. 9B). The significant blue shift of the fluorescence emission of Trp 5 upon addition of negatively charged lipids indicates that this residue moves to a highly hydrophobic environment that is not likely to be provided by structural rearrangement of this short peptide bearing a net charge of ϩ4. The direction and magnitude of the changes in the emission maxima of Trp 5 in the presence of anionic phospholipids are consistent with a relatively deep membrane penetration of this residue. Fluorescence studies on model peptides FIGURE 8. Membrane association of the ␤ 2e -subunit modulates the voltage-dependent inactivation of the R-type calcium channels. A, binding assay of wild-type ␤ 2e and ␤ 2e NT v (7ϩ less) membrane association-deficient mutant to the highly conserved ␣ 1 interaction domain (AID). Either GST moiety alone or AID site fused to GST (GST-AID) were pre-coupled to glutathione-agarose beads and incubated with either wild-type ␤ 2e or ␤ 2e NT v (7ϩ less) proteins. After several washes, the bound proteins were eluted and resolved on SDS-PAGE. Lane 1, molecular mass in kDa; lane 2, binding of wild-type ␤ 2e to GST-AID; lane 3, binding of wild-type ␤ 2e to GST; lane 4, binding of ␤ 2e NT v (7ϩ less) to GST-AID; lane 5, binding of ␤ 2e NT v (7ϩ less) to GST; lanes 6-9, inputs proteins. B, representative current traces from tsA201 cells coexpressing the R-type Ca V 2.3 ␣ 1 -subunit of voltage-activated calcium channels either with wild-type ␤ 2e or ␤ 2 NT v (7ϩ less). Currents were elicited during a 10-s pulse to 0 mV from a holding potential of Ϫ90 mV. C, average t 0.5 for Ca V 2.3/␤ 2e and Ca V 2.3/␤ 2e NT v (7ϩ less) channel complexes. t 0.5 significantly decreases from 1.26 Ϯ 0.14 s (n ϭ 6) for Ca V 2.3/␤ 2e to 0.16 Ϯ 0.01 s (n ϭ 6) for Ca V 2.3/␤ 2e NT v (7ϩ less) channel complexes (t test; p Ͻ 0.01). D, average steady-state inactivation curves from cells expressing the indicated channel subunit combinations. 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 a short deactivation pulse to Ϫ90 mV and a 5-s test pulse at 0 mV. The interval between sweeps was of 60 s. Continuous lines correspond to the Boltzmann distribution that best fit the experimental data for the indicated channel complex. Values are expressed as mean Ϯ S.E. V 0.5 significantly varies from Ϫ49.0 Ϯ 3.1 mV (n ϭ 6) for Ca V 2.3-␤ 2e to Ϫ67.0 Ϯ 0.8 mV (n ϭ 6) for Ca V 2.3-␤ 2e NT v (7ϩ less) channel complexes (t test; p Ͻ 0.01).
Plasma Membrane Tethering of the Calcium Channel ␤ 2e -subunit APRIL 11, 2014 • VOLUME 289 • NUMBER 15 assign an average distance of 7 Å from the lipid bilayer center to a Trp emission fluorescence wavelength maximum comparable with the one we observed with the ␤ 2e NT v peptide at the highest lipid concentration (323 nm, Fig. 7, E and F) (30). Considering the width of a monolayer as 15 Å, this would imply that the Trp 5 penetrates about 8 Å into the plasma membrane beyond the lipid head groups and, consequently, that the putative inplane helix is in close contact or interweaved with the headgroups (Fig. 9C). Aside from strengthening the association with the plasma membrane, the Trp 5 may be important for the intracellular trafficking of the protein, for example by promoting binding to intracellular membrane vesicles. This would explain why the protein bearing an alanine at that position binds to anionic lipids but is not found at the cell surface. Mutation of Leu 9 into alanine also results in a loss of plasma membrane targeting (data not shown), reinforcing the possibility that hydrophobic interactions may be important under physiological conditions for associating the protein with the plasma membrane. Our data are consistent with a mechanism by which ␤ 2e anchors to the plasma membrane via an N-terminal helix that orients parallel to the plane of the bilayer (in-plane membrane association) and inserts relatively deeply into the headgroups (Fig. 9C). Association with the plasma membrane would take place in two steps as follows: first, the protein is recruited to the cell surface via electrostatic interactions with its positively charged residues, and second, the membrane-bound confor-mation is stabilized by the penetration of hydrophobic residues (i.e. Trp 5 and Leu 9 ) into the lipidic core. Although we cannot evaluate the precise contribution of electrostatic and hydrophobic interactions to the binding of ␤ 2e to the plasma membrane in vivo, our experiments show that hydrophobic interactions become more relevant in the relatively higher ionic strength of the cellular environment.
The interaction with lipids can alter the subcellular localization, conformation, and activity of proteins. An example is ␤ 2a , which switches from a fast to a slow inactivation-conferring phenotype upon palmitoylation and membrane association. Here, we demonstrated that ␤ 2e is not an exception; when associated with the plasma membrane, ␤ 2e displays very different channel modulatory activity compared with the membrane anchor-deficient mutant. The fact that both ␤ 2a and ␤ 2e inhibit voltage-dependent inactivation of Ca V 2.3 channels, although they associate differently with lipids, supports the model suggesting that membrane anchoring is sufficient to slow down channel inactivation by restricting the movement of the inactivation gate (31).
The dynamic nature of ␤ 2a palmitoylation provides a further control mechanism regulating in vivo calcium influx through voltage-gated calcium channels (15). However, it remains unknown whether or not the association of ␤ 2e with the cell membrane is subject to regulation in vivo. In fluorescence confocal images of cells expressing ␤ 2e , the protein distributes mainly in the surface membrane, but a slight dif- FIGURE 9. Model depicting the association of the ␤ 2e -subunit with the plasma membrane. A, amino acid sequence and predicted secondary structure for the ␤ 2e N-terminal variable region (␤ 2e NT V ). The positively charged and the nonpolar residues within ␤ 2e NT v are highlighted in blue and yellow, respectively. The green cylinder below the sequence delimits the predicted ␣-helix (residues 2-13) according to the consensus secondary prediction results (NPS@Web server) (28). Within the helix an in-plane membrane anchor segment (residues 4 -7) is predicted (black line) (29). The lower panel shows the amino acid alignment of ␤ 2e NT v segment from (in parenthesis, UniProtKB accession numbers) rat (Q8VGC3), human (Q08289), mouse (Q8CC27), dog (E2RCN6), chicken (E1BZF6), Xenopus (F7EDB7), and zebrafish (B2XY77). B, helical wheel diagram of the predicted consensus helix within ␤ 2e NT v extending from residues 2 to 13 (Helical Wheel Applet, University of Virginia). The potential location of Arg 16 considering a longer helix is also shown (37). The relatively critical residues deduced from alanine scanning mutagenesis, Lys 2 , Trp 5 together with Leu 9 (data not shown) lie on one side of the helix. C, proposed model to explain membrane anchoring of ␤ 2e . The ␤ 2e -subunit includes a helical membrane anchor site in its N-terminal region that associates in plane with the membrane via electrostatic interactions between its positively charged residues and the phospholipids headgroups. After the electrostatic recruitment of the protein to the plasma membrane, the bound conformation is stabilized by a relatively deep penetration of hydrophobic residues, including the aromatic side chain of Trp 5 and the aliphatic chain of Leu 9 into the lipid core. We suggest that the predicted phosphorylation sites (36), represented as lines, in the flexible segment joining the NT v helix and the core of the protein (labeled ␤) may induce dissociation of the protein from the plasma membrane, providing additional regulatory input for calcium entry through voltage-gated calcium channels. fuse fluorescence is visible within the cell, arguing in favor of a regulated membrane targeting (32). Although additional experiments will be required to test this hypothesis, we speculate that the phosphorylation state of the ␤ 2e N-terminal domain may be particularly important for controlling the association of this subunit with the plasma membrane as has been suggested for other proteins (32,33). Phosphorylation of Ca V ␤ subunits that play a role in channel regulation has been described (34,35). The N-terminal region of ␤ 2e contains several potential phosphorylation sites downstream of the membrane anchor segment identified in this study. Prediction programs reveal a phosphorylation hot spot within a very short protein segment encompassing residues 29 -39 (36). Phosphorylation of these sites would decrease the positive charge density of ␤ 2e NT v and, consequently, weaken its interaction with membrane lipids.
In summary, in addition to providing a mechanism by which ␤ 2e anchors to the plasma membrane, our study suggests a structure for the N-terminal region of the protein for which no data are available. The high resolution structure of three ␤ isoforms (␤ 2 , ␤ 3 , and ␤ 4 ) has been solved, but these do not provide structural data for the variable regions outside the SH3-guanylate kinase domains. This is the first ␤-subunit described that associates with the plasma membrane via electrostatic interactions. The fact that single point mutations within this domain appear to alter the protein membrane affinity makes it feasible to use this protein as a model to correlate membrane binding affinities to the degree of inhibition of voltage-dependent inactivation conferred by membrane-anchored ␤-subunits.