A Polybasic Plasma Membrane Binding Motif in the I-II Linker Stabilizes Voltage-gated CaV1.2 Calcium Channel Function

Background: L-type Ca2+ channels (LTCCs) are fine-tuned by different molecular mechanisms. Results: Pore-forming α1-subunits of LTCCs contain a polybasic amino acid sequence within their I-II linkers that binds to the plasma membrane. This polybasic motif is required for normal channel gating and modulation. Conclusion: The polybasic cluster stabilizes normal channel activity. Significance: We discovered a new modulatory domain of LTCCs within their pore-forming α1-subunit.

L-type voltage-gated Ca 2؉ channels (LTCCs) regulate many physiological functions like muscle contraction, hormone secretion, gene expression, and neuronal excitability. Their activity is strictly controlled by various molecular mechanisms. The poreforming ␣ 1 -subunit comprises four repeated domains (I-IV), each connected via an intracellular linker. Here we identified a polybasic plasma membrane binding motif, consisting of four arginines, within the I-II linker of all LTCCs. The primary structure of this motif is similar to polybasic clusters known to interact with polyphosphoinositides identified in other ion channels. We used de novo molecular modeling to predict the conformation of this polybasic motif, immunofluorescence microscopy and live cell imaging to investigate the interaction with the plasma membrane, and electrophysiology to study its role for Ca v 1.2 channel function. According to our models, this polybasic motif of the I-II linker forms a straight ␣-helix, with the positive charges facing the lipid phosphates of the inner leaflet of the plasma membrane. Membrane binding of the I-II linker could be reversed after phospholipase C activation, causing polyphosphoinositide breakdown, and was accelerated by elevated intracellular Ca 2؉ levels. This indicates the involvement of negatively charged phospholipids in the plasma membrane targeting of the linker. Neutralization of four arginine residues eliminated plasma membrane binding. Patch clamp recordings revealed facilitated opening of Ca v 1.2 channels containing these mutations, weaker inhibition by phospholipase C activation, and reduced expression of channels (as quantified by ON-gating charge) at the plasma membrane. Our data provide new evidence for a membrane binding motif within the I-II linker of LTCC ␣ 1 -subunits essential for stabilizing normal Ca 2؉ channel function.
Ca 2ϩ influx through voltage-gated L-type Ca 2ϩ channels (LTCCs) 3 is essential for many cellular events. It causes muscle contraction, initiates hormone secretion and neurotransmitter release, tunes neuronal excitability, and regulates gene expression (1,2). LTCCs are large multiprotein complexes consisting of a pore-forming transmembrane ␣ 1 -subunit and accessory intracellular ␤and extracellular ␣ 2 ␦-subunits. To adapt Ca 2ϩ entry to different cellular needs, functional diversity is achieved by multiple molecular mechanisms. Four different ␣ 1 -subunits (Ca V 1.1-1.4) can confer different biophysical properties (1) that are further adjusted by alternative splicing (3)(4)(5). In addition, the channels' accessory subunits tune their gating behavior and various protein interaction partners further provide LTCC currents with cell-and tissue-specific properties (1). However, rapid regulatory changes in current dynamics cannot be accomplished by changing channel composition. Instead, quick adaptive responses of channel activity require fast regulatory processes, including Ca 2ϩ entry itself (inducing inactivation through channel-bound calmodulin (6)), phosphorylation by various kinases (7)(8)(9), extracellular pH (10), and direct G-protein modulation (11). Next to these mechanisms, membrane lipids are also important regulators of Ca 2ϩ channel activity (12,13). Depletion of phosphatidylinositol 4,5-bisphosphate (PIP 2 ) from the plasma membrane causes a rapid decrease of Ca 2ϩ channel activity. In excised membrane patches, non-LTCC currents (Ca V 2.1 and Ca V 2.2) run down within minutes, which is attenuated by application of PIP 2 or its water-soluble analogue diC8-PIP 2 (14,15). Rapid depletion of membrane PIP 2 content in intact cells by muscarinic 1 (M1) receptor activation, voltage-dependent phosphatases, or rapa-mycin-induced translocation of inositol-lipid phosphatases confirmed the direct dependence of Ca V 1 (Ca V 1.2 and Ca V 1.3) and Ca V 2 (Ca V 2.1 and Ca V 2.2) channels on endogenous membrane PIP 2 (16). Channel inhibition by PIP 2 depletion occurs in a ␤-subunit-dependent manner because the ␤ 2a -subunit, an isoform anchored to the plasma membrane by palmitoylation, attenuates inhibition (13,17). Together with observations that arachidonic acid inhibits channel activity, a model has been proposed in which PIP 2 , the fatty acid side chain of palmitoylated ␤ 2a -subunits of voltage-gated Ca 2ϩ channels (VGCCs), and arachidonic acid have overlapping binding sites, resulting in complex channel regulation by lipid metabolism (12,(17)(18)(19). There is also experimental evidence for more than one PIP 2 regulatory site in VGCCs: a higher affinity stimulatory "S" site supporting channel activity and a lower affinity inhibitory site ("R" site) stabilizing reluctant gating properties (12,14,20). Current working models suggest at least one facilitatory PIP 2 site on LTCCs and at least two on Ca V 2.2 channels (16).
Anionic phospholipids, like phosphatidylserine and phosphoinositides, target proteins with clusters of positive charges to the plasma membrane (21,22). Despite unique insights into the structural basis of PIP 2 modulation within the crystal structure of K ϩ channels (23,24) and the identification of potential polyphosphoinositide binding domains within several ion channel proteins (for a review, see Refs. 13 and 25), the structural basis for fast modulation of VGCC function by plasma membrane lipids is unknown. It occurs at a very fast time scale, indicating that channel-lipid interactions are rapidly reversible and do not require more complex biochemical pathways, such as channel internalization (26). Similar to K ϩ channels (23,24), it appears likely that cytoplasmic domains located close to the plasma membrane participate in lipid interactions that stabilize channel function and allow modulation of channels by activation of phospholipase C (PLC). In addition to fast channel modulation, it is at present also unknown whether anionic lipids participate in the constitutive stabilization of VGCC function, which would be expected for very high affinity interactions not subject to regulation by changes in plasmalemmal phospholipid content (27).
Here we report the identification of a polybasic cluster within the cytoplasmic I-II linker of Ca V 1.2 ␣ 1 -subunits, which specifically binds to the plasma membrane. Activation of PLCs and polyphosphoinositide breakdown together with elevated intracellular Ca 2ϩ levels reverse membrane binding. Four positively charged arginines are required for binding, most likely by strongly favoring a straight helix conformation of that region, positioning it entirely at the interface between the hydrocarbon core and lipid headgroups. Their neutralization in the intact channel facilitates channel opening and thus suggests a role of this polybasic cluster in stabilizing normal channel activity.

Experimental Procedures
All chemicals, reagents and antibodies were purchased from Sigma-Aldrich (Vienna, Austria) except where otherwise indicated. diC8-PIP 2 was purchased from Echelon Biosciences Inc. (Salt Lake City, UT).
Cell Culture and Transfection-For immunofluorescence microscopy, live cell imaging (LCI), and electrophysiology, tsA-201 cells were cultured in Dulbecco's modified Eagle's medium (DMEM), 10% fetal calf serum (Gibco, 10500.064), 2 mM glutamine (Sigma, G753) penicillin (10 units/ml), and streptomycin (10 g/ml) and maintained at 37°C in a humidified environment with 5% CO 2 . Cells were grown and split when they reached about 80% confluence using 0.05% trypsin for cell dissociation. Cells were transiently transfected using Ca 2ϩ -phosphate precipitation as described previously (38). For immunofluorescence and LCI, cells were replated 24 h after transfection onto 12 mm (for immunofluorescence) and 18 mm (for LCI) poly-L-lysine-coated coverslips and kept at 37°C for 24 -48 h until further experimentation. For whole-cell patch clamp recordings, tsA-201 cells were transiently transfected with equimolar ratios of cDNA encoding C-terminally long or short wild-type or mutant Ca V 1.2 ␣ 1 -subunits together with auxiliary ␤ 3 -(rat, NM_012828) and ␣ 2 ␦ 1 -(rabbit, NM_001082276) sub-units. To visualize transfected cells, GFP was co-transfected. Cells were then plated onto a 35-mm culture dish coated with poly-L-lysine. The cells were kept at 30°C and 5% CO 2 and subjected to electrophysiological measurements about 48 -72 h after transfection.
Immunofluorescence Microscopy-48 h after plating, tsA-201 cells were fixed in 4% (w/v) paraformaldehyde (Electron Microscopy Sciences, 15710) for 15 min at room temperature. Cells were washed thoroughly with phosphate-buffered saline and blocked for 30 min at room temperature with 5% (w/v) normal goat serum (GibcoBRL, 16210-064) in 0.2% (v/v) Triton/phosphate-buffered saline for cell permeabilization. Cells were incubated overnight at 4°C with primary antibodies diluted in washing buffer containing 0.2% (v/v) Triton X-100 and 0.2% (w/v) BSA (immunoglobulin-free) in phosphate-buffered saline. After washing the cells in washing buffer, they were incubated with the secondary antibody for 1 h at room temperature in the dark. Washing was repeated, and the coverslips were mounted with Vectashield mounting medium (Vector Laboratories, H-1000) and sealed with nail polish on microscope slides. The following antibodies were employed: mouse monoclonal anti-V5, working dilution 1:500 (Invitrogen, R96025); rabbit polyclonal anti-FLAG (Sigma, F7425; 1:500); mouse monoclonal anti-FLAG (Sigma, F3165; 1:5000); Alexa Fluor-488-conjugated goat anti-rabbit and Alexa-594-conjugated goat anti-mouse antibodies (1:4000; Life Technologies, Invitrogen, A-11008 and A-11005). Images were captured with an Axiophot microscope (Carl Zeiss Inc., ϫ63, 1.4 numerical aperture Zeiss plan apochromat oil immersion lens) using a cooled CCD camera and at room temperature and Meta View image processing software (Universal Imaging Corp., West Chester, PA). Images were manually adjusted for brightness and contrast with Adobe Photoshop version 7.0.
Live Cell Imaging-After 24 h of replating, 18-mm coverslips containing live tsA-201 cells were washed once with Tyrode solution (130 mM NaCl, 2.5 mM KCl, 2 mM CaCl 2 , 2 mM MgCl 2 , 10 mM HEPES, 30 mM glucose, pH 7.4, 319 mosmol/liter), subsequently mounted on a Ludin chamber (Life Imaging Services) and kept in Tyrode solution until further processing. The chamber was placed on an ASI stage of an inverted Zeiss Axiovert 200M epifluorescence microscope (Carl Zeiss). Cells were imaged at room temperature. Different drugs (10 M oxotremorine M, 50 M m-3M3FBS (2,4,6-trimethyl-N-(meta-3-trifluoromethyl-phenyl)-benzenesulfonamide), 5 M ionomycin, 20 M wortmannin, 5 M rapamycin (Calbiochem, Merck Millipore)) in different combinations (as indicated under "Results") were diluted to final concentrations in Tyrode solution and either directly added to the imaging chamber or added to a 35-mm culture dish containing the coverslip when preincubation was desired. Twelve-bit grayscale images were recorded at different time intervals using a cooled CCD camera (SPOT, Diagnostic Instruments), Metavue image processing software (Universal Imaging, Corp.), and a ϫ63, 1.4 numerical aperture Zeiss Plan Apochromat oil immersion lens. For presentation, selected images were linearly adjusted using Adobe Photoshop version 7.0 and CS4. For quantification, areas of interest of equal size were selected in the membrane, cytoplasm, and background of cells, and average pixel intensity was recorded and used to calculate the membrane to cytoplasm ratio for every image. All quantitative data are represented as mean Ϯ S.E. for the indicated number of experiments (n), except if stated otherwise. Data were analyzed by paired Student's t test, unpaired Student's t test, or one-way analysis of variance with Bonferroni post-hoc test using GraphPad Prism version 5.01 (GraphPad Software Inc.). Significance level was set to p Ͻ 0.05.
Electrophysiological Recordings-Whole-cell patch clamp experiments were performed in transiently transfected tsA-201 cells using an Axopatch 200B amplifier (Axon Instruments). Electrodes with a final resistance of 2-4 megaohms were pulled from borosilicate glass capillaries using a P-97 micropipette puller (Sutter Instruments) and subsequently fire-polished (MF-830 microforge, Narishinge). Ca 2ϩ currents were measured using the following solutions: pipette solution, 135 mM CsCl, 10 mM Cs-EGTA, 1 mM MgCl 2 , 10 mM HEPES, 4 mM Na 2 -ATP, adjusted to pH 7.3 with CsOH; bath solution, 15 mM CaCl 2 , 150 mM choline chloride, 1 mM MgCl 2 , 10 mM HEPES adjusted to pH 7.3 with CsOH. All voltages were corrected for a junction potential of Ϫ9.3 mV. Recordings were digitized (Digidata 1322A digitizer, Axon Instruments) at 50 kHz, low pass-filtered at 5 kHz, and analyzed using pClamp version 10.2 software (Axon instruments). Currentvoltage (I-V) relationships were obtained by applying a 20-ms square pulse protocol to various test potentials starting from a holding potential of Ϫ90 mV. Resulting I-V curves were fitted to Equation 1, where I is the peak current amplitude, G max is the maximum slope conductance, V is the test potential, V rev is the extrapolated reversal potential, V 0.5 is the half-maximal activation voltage, and k act is the activation slope. Channel inactivation was measured using a 300-ms pulse from a holding potential of Ϫ90 mV to V max . For the estimation of the open probability (P O ), the area of the ON-gating current (Q ON ) was integrated and compared with the amplitude of the ionic tail current (I Tail ) at V rev . Data were analyzed using Clampfit version 10.2 (Axon Instruments) and Sigma Plot 12 (Systat Software Inc.). To assess the effects of diC8-PIP 2 or the PLC activator m-3M3FBS, cells were depolarized from a holding potential of Ϫ90 mV to V max for 20 ms at 0.1 Hz. As a control, Ca V 1.2S-or Ca V 1.2S 4E -expressing cells were perfused (flow rate: 250 l/min) with bath solution only. To test whether diC8-PIP 2 can stabilize I Ca decline, the same protocol was performed in the presence of 100 M diC8-PIP 2 in the internal solution. DiC8-PIP 2 was prepared according to the manufacturer's instructions as a 1 mM stock in intracellular recording solution. Aliquots were kept frozen and diluted before use. To analyze the effects of phosphoinositide depletion on Ca V 1.2 currents, cells were perfused with 20 M wortmannin and 50 M m-3M3FBS. (flow rate: 250 l/min). All quantitative data are represented as mean Ϯ S.E. Statistical significance was determined by one-or two-way analysis of variance followed by Bonferroni post-hoc test or unpaired Student's t test as indicated using GraphPad Prism version 5.01 (GraphPad Software Inc.). Significance level was set to p Ͻ 0.05.

Rosetta Membrane Modeling of the Domain I-II Linker Region
and Voltage-sensing Domain II of Ca V 1.2 Channel-Homology, de novo, and full-atom modeling of the voltage-sensing domain (VSD) of native and mutant Ca V 1.2 channels was performed using the Rosetta membrane method (39 -41) and the x-ray structure of the bacterial voltage-gated Na ϩ channel (Na V Ab) VSD (42) as a template. Sequence alignment between native Ca V 1.2 and Na V Ab VSDs shown in Fig. 8A was generated using the HHpred server (43,44). The backbone structure of the transmembrane regions of Ca V 1.2 was built based on Na V Ab VSD template. The 19-residue N-terminal region and S1-S2, S2-S3, and S3-S4 loops of Ca V 1.2 VSD were built de novo using the Rosetta cyclic coordinate descent loop modeling method (45) guided by membrane environment-specific energy function (39,46). 10,000 models were generated for each Ca V 1.2 channel construct, and the top 10% of models ranked by total score were clustered (47) using root mean square deviation threshold that generates at least 150 -200 models in the largest cluster. Models representing centers of the top five clusters and the best 10 models by total score were chosen for visual analysis. The top cluster and all 10 lowest energy models of native Ca V 1.2 showed very similar conformation of the domain I-II linker region (see Fig. 8). None of the top five clusters and 10 lowest energy models of alanine or glutamate mutants of Ca V 1.2 showed similar conformations of the domain I-II linker region (see Fig. 8). All structural figures were generated using the UCSF Chimera package (48).

Results
We (49) and others (50) have recently reported that the I-II linker of LTCC Ca V 1.3 and Ca V 1.2 ␣ 1 -subunits is targeted to the plasma membrane when expressed alone as a soluble protein in tsA-201 cells. Because the I-II linker harbors the high affinity interaction site for Ca 2ϩ channel ␤-subunits (51), it is also capable of targeting co-expressed ␤-subunits to the plasma membrane, even in complex with ␤ 1 -subunit-bound Rab3-interacting molecule (49). Plasma membrane association of the linker is independent of ␤-subunits. Introduction of mutation W441A, which disrupts ␤-subunit binding to the linker (31, 52), prevented ␤-subunit targeting without affecting the plasma membrane localization of the linker (49). This suggested a specific, ␤-subunit-independent interaction of the I-II linker with the plasma membrane either through a membrane-associated protein or through direct interaction with membrane lipids.
I-II Linkers of All LTCC Isoforms Bind to the Plasma Membrane-To test whether plasma membrane binding is a property of all LTCC ␣ 1 -subunits, we transfected FLAG-labeled I-II linkers of Ca V 1.1, Ca V 1.2, Ca V 1.3, and Ca V 1.4 ␣ 1 -subunits ( Fig. 1, A-D, left) into tsA-201 cells. Immunoblot analysis (not shown; n ϭ 3) confirmed their expression as intact polypeptides. All LTCC I-II linkers were localized at the plasma membrane ( Fig. 1, A-D). This localization pattern was independent of expression levels and indistinguishably observed in cells with weak and strong expression of the respective linkers (not shown). More than 85% of transfected cells (three independent experiments; 300 cells/experiment analyzed) showed this typical plasma membrane binding. In contrast, a FLAG-labeled control fragment derived from the Ca V 1.3 C terminus (FLAG-C158) clearly revealed a cytoplasmic distribution (Fig. 1E, first panel), as did the non-palmitoylated ␤ 2a mutant C3S/C4S ␤ 2a (Fig.  1E, middle). In contrast, normal palmitoylated ␤ 2a revealed the expected plasma membrane targeting (Fig. 1E, last panel) and thus served as a positive control. Consistent with our previous findings, C3S/C4S ␤ 2a (n ϭ 3; Fig. 1, A-D, middle panel) and ␤ 3 (not shown, n ϭ 3) (49) also were localized at the plasma membrane after co-expression with one of the four LTCC I-II linkers. This shows that all LTCC I-II linkers support ␤-subunit plasma membrane targeting.
Secondary structure analysis of the Ca V 1.2 I-II linker using PSIPRED (53) (not shown) predicted the region between amino acids 531 and 550 to form a polybasic amphipathic ␣-helix with 8 of the 9 positive charges (Fig. 3B) located on one side of the helix as shown in a helical wheel plot (Fig. 3B). This included four arginines (Arg-537, Arg-538, Arg-541, and Arg-544) located within the region 536 -544 required for membrane translocation (Fig. 3C, yellow), all of which are conserved among the I-II linkers of LTCCs (Fig. 3D). To test whether these positive charges are essential for plasma membrane binding, we neutralized them by mutations to alanines (mutant I-II-4R4A) or converted them to negatively charged glutamates (mutant I-II-4R4E; Fig. 3C). As expected, both mutations prevented the plasma membrane binding of the Ca V 1.2 I-II linker and of the co-expressed C3S/C4S ␤ 2a subunits (Fig. 3E).
To assess whether this polybasic motif is sufficient to induce plasma membrane binding when attached to an otherwise cytoplasmic protein, we fused residues 526 -554 to the C terminus of GFP ( GFP 526 -554; Fig. 3, A and C). This construct localized to the plasma membrane (Fig. 3A). It also accumulated in the nucleus of the vast majority of transfected cells. Notably, a distinct feature of lipid-interacting polybasic membrane targeting motifs is their similarity to nuclear localization sequences (22), a property that may also account for the nuclear targeting of GFP 526 -554. However, nuclear staining did not obscure plasma membrane binding (see also Figs. 4A and 5B), and this was also confirmed in dividing cells in which nuclear staining was completely absent (Fig. 3A, GFP 526 -554, right panel; see also Fig. 5A). Taken together, these findings clearly demonstrate that positive charges at the C-terminal end of the Ca V 1.2 I-II linker form a plasma membrane binding motif sufficient for translocating the cytoplasmic I-II linker and fused GFP to the plasma membrane.
Involvement of Membrane Phosphoinositides in Plasma Membrane Targeting-The polybasic motif at the distal end of the I-II linker closely resembles clusters of positive charges found in other proteins, such as Rit and K-Ras (22) (Fig. 3D), and polybasic domains in other ion channels mediating protein-lipid interactions at the plasma membrane (for a review, see Ref. 13). We therefore hypothesized that this cluster of positive charges could serve a similar function. This notion was further supported by current folding models of Ca 2ϩ channel ␣ 1 -subunits, which position these charges near the cytoplasmic end of the transmembrane helix IIS1 (42), which, in turn, places them in close proximity to the anionic lipids of the inner leaflet of the plasma membrane.
Many modulatory lipid interactions of ion channels involve polyphosphoinositides (13). Ca V 1 and Ca V 2 VGCCs are modulated by metabotropic receptors through activation of PLC, also independently of direct G-protein modulation (16). Activation of M1 receptors induces a partial inhibition of Ca 2ϩ inward currents mostly by membrane depletion of PIP 2 in tsA-201 cells (16). This effect is mediated via activation of PLC␤ (54). If binding of LTCC I-II linkers is dynamically regulated by interaction with PLC-metabolized polyphosphoinositides (primarily PIP 2 ), then it should be reversed in a PLC-sensitive manner.
We therefore co-transfected tsA-201 cells with M1 receptors and examined effects on membrane binding of our constructs in response to receptor activation with oxotremorine M (Oxo-M) using live cell imaging (Fig. 4, A and B). The PIP 2specific mRFP-tagged pleckstrin homology domain of PLC␦ (PH-PLC␦) served as positive control for monitoring PIP 2 breakdown (55). Oxo-M induced the relocation of PH-PLC␦ but did not reverse the plasma membrane association of I-II linker-bound ␤ 3 -GFP (n ϭ 3 independent experiments; Fig. 4A) or of GFP 526 -554 (n ϭ 3; Fig. 4B). Translocation of membranebound ␤ 3 -subunits was also absent in cells responding to co- Previous studies have shown that plasma membrane targeting of cytoplasmic proteins through cationic motifs resembling our polybasic sequence may also involve negatively charged lipids other than PIP 2 , in particular PIP 3 , PIP (22), or phosphatidylserine (21). We therefore used a rapamycin-inducible system (37) causing PIP and PIP 2 depletion. Briefly, RFP-tagged pseudojanin (PJ), an enzymatic chimera that converts PIP 2 to PIP and dephosphorylates PIP, was translocated from the cytosol to the plasma membrane upon rapamycin-induced Lyn 11 -FRB interaction. Translocation of PJ to the plasma membrane started 15.6 Ϯ 3.12 s after rapamycin application and was completed after 30.4 Ϯ 3.61 s in the control experiments (Fig. 4, C  and D). This was followed by membrane dissociation of cotransfected GFP-PH-Osh2x2 (starting 29.6 Ϯ 4.03 s after drug administration, completed after 57.1 Ϯ 3.28 s; mean Ϯ S.E., n ϭ 24). Notably, in 13 of 24 cells, some residual membrane staining of GFP-PH-Osh2x2 still remained (Fig. 4C). When we co-transfected the rapamycin-inducible system and the I-II linker together with ␤ 3 -GFP, PJ translocation started at 13.1 Ϯ 1.43 s and was completed within 25.3 Ϯ 1.07 s (mean Ϯ S.E., n ϭ 16) after rapamycin application. However, even after a prolonged recording time (20 min), we never observed a I-II linker-mediated translocation of ␤ 3 -GFP to the cytoplasm (n ϭ 16 cells; three independent experiments; Fig. 4E). We also combined the rapamycin-inducible system with wortmannin treatment. Preincubation with 20 M wortmannin for 45-90 min before rapamycin application did not affect targeting of controls (PJ translocation started only upon rapamycin application after 20 Ϯ 3.33 s, completion after 37.0 Ϯ 5.17 s, n ϭ 10) or linker-bound ␤ 3 -GFP (n ϭ 10, three independent experiments; Fig. 4F).
Earlier studies have shown very strong plasma membrane binding of similar polybasic targeting motifs of small GTPases, such as Rit or K-Ras (21,22) (Fig. 3D) or the plasma membrane targeting domain of RasGRP1, a Ras-specific exchange factor (56). Their dissociation requires, in addition to phosphoinositide hydrolysis, either depletion of PI3K products (22,56) or prolonged elevation of intracellular Ca 2ϩ (21). Because neither the rapamycin-inducible system nor the Oxo-M response appears to induce complete phosphoinositide depletion and induces no or, in the case of Oxo-M, only a transient intracellular Ca 2ϩ signal (55), we employed the PLC activator m-3M3FBS together with wortmannin to induce PIP, PIP 2 , and PIP 3 depletion and inhibit PI3K. m-3M3FBS activates PLC␤, -␥, and -␦ isoforms and also induces a prolonged intracellular Ca 2ϩ release (55,57). Application of 50 M m-3M3FBS alone was not sufficient to induce translocation within our observation period. However, co-application of m-3M3FBS together with wortmannin caused a delayed translocation of PH-PLC␦ (n ϭ 4; Fig. 5A, left) as well as of GFP 526 -554 (n ϭ 11 independent experiments; Fig. 5A, right) from the plasma membrane to the cytosol starting within a time frame of 20 -30 min and completion in the following 7-10 min. This suggested a combined effect of intracellular Ca 2ϩ (which acts as a PLC co-activator (55)) and lipid hydrolysis for slow translocation. Accordingly, the addition of 5 M ionomycin together with m-3M3FBS (58) induced an even faster relocation of PH-PLC␦ to the cytoplasm (Fig. 5B, left). Translocation started 1.9 Ϯ 0.2 min (mean Ϯ S.D., n ϭ 15) after m-3M3FBS/ionomycin stimulation and was complete within 45-55 s. When we treated cells co-transfected with the I-II linker together with ␤ 3 -subunits, PLC activation reversed the linker-mediated ␤ 3 -GFP association with the plasma membrane (Fig. 5B, middle). Although cell to cell differences for the translocation time course after adding m-3M3FBS/ionomycin were observed, the translocation of ␤ 3 -GFP and PH-PLC␦ always occurred in parallel (n ϭ 5 independent experiments; not shown). The GFP-tagged cationic peptide ( GFP 526 -554) also redistributed to the cytoplasm under these experimental conditions (n ϭ 3; Fig. 5B, right). In contrast, no translocation was observed in control experiments with ␤ 2aGFP subunits, which are anchored to the membrane through palmitoylation, known to be insensitive to polyphosphoinositide breakdown (n ϭ 6; Fig. 5C). Taken together, our data demonstrate that polyphosphoinositide breakdown as well as increased intracellular Ca 2ϩ levels are required for unbinding of the distal portion of the Ca V 1.2 I-II linker from the plasma membrane.

Role of the Polybasic Motif for Ca V 1.2 Channel Function and
Modulation-Due to the close proximity of the polybasic motif to the membrane and its lipid-dependent interaction with the plasma membrane, our data strongly suggest that this interaction also occurs in the pore-forming ␣ 1 -subunit of intact Ca V 1.2 channel complexes. We therefore introduced the mutations found to prevent membrane binding into the I-II linker constructs I-II-4R/4A and I-II-4R/4E (Fig. 3D) into the corresponding positions of the intact Ca V 1.2 ␣ 1 -subunits and expressed these mutants (together with ␤ 3 and ␣ 2 ␦ 1 ) in tsA-201 cells. Mutation-induced changes of Ca 2ϩ inward current (I Ca ) properties were then analyzed using whole-cell patch clamp experiments with Ca 2ϩ as charge carrier. Ca V 1.2 ␣ 1 -subunit channels undergo partial proteolytic processing, giving rise to a C-terminally long (Ca V 1.2L) and short (Ca V 1.2S) variants. Both variants exist in the heart and brain and exhibit different current properties (59,60). In particular, Ca V 1.2S exhibits a higher open probability (59,60). We therefore tested the effects of the I-II linker mutations in both variants (long, mutants Ca V 1.2L 4A and Ca V 1.2L 4E ; short, mutants Ca V 1.2S 4A and Ca V 1.2S 4E ).
All four mutant constructs conducted inward Ca 2ϩ currents. Current-voltage relationships revealed a significant 6 -10 mV shift of V 0.5 in the hyperpolarizing direction for both mutants (Table 1 and Fig. 6A). This was due to a significant increase in the slope without changes in activation threshold (Table 1), suggesting more efficient coupling of pore opening to membrane depolarization. The mutations had little effect on the kinetics of I Ca inactivation during 300-ms depolarizing pulses to the voltage corresponding to the peak of the I-V relationship ( Table 2). The apparent reversal potential was also unchanged ( Table 1). Differences between the mutants and wild-type channels were observed when we studied the relationship between the size of ON-gating charges (as a measure of active channels on the cell surface; enlarged ON-gating currents are shown in the insets of Fig. 6B) and the size of ionic tail currents (Fig. 6B). As shown by us and others previously, this ratio provides an estimate for the channel's open probability (61)(62)(63). The two wild-type constructs served as an internal control because a higher open probability was reported earlier for Ca V 1.2S (59,61). This is evident as a statistically significant difference of the I Tail /Q ON ratio in our experiments ( Fig. 6B; see Table 1 for statistics). This is also evident from the steeper slopes of the regression lines of plots of I Tail versus Q ON (Fig.  6C). In Ca V 1.2L, the 4R/4A mutation caused a 64% increase in the I Tail /Q ON ratio. A similar increase was also observed in the Ca V 1.2S mutants despite higher basal open probability. An even larger effect was seen for the Ca V 1.2L 4E and Cav1.2S 4E mutants, which more than doubled the open probability. We also found that all four arginine mutant constructs significantly reduced Q ON in both Ca V 1.2L and Ca V 1.2S, with the reduction again more pronounced for the two 4R/4E mutations. Our data demonstrate that positive charges located in proximity to the transmembrane segment IIS1 within the I-II linker are impor-tantdeterminantsofnormalCa V 1.2Ca 2ϩ channelfunctionindependent of the length of their C-terminal tails and basal open probabilities. The negative shift in V 0.5 and higher open probability both indicate a tighter coupling between the voltage sensor and the pore. Based on our membrane targeting analysis, we propose that we have identified a site that is involved in the membrane lipid-dependent stabilization of channel function as well as the regulation of the expression of functional channels at the plasma membrane.
We also tested whether the arginine mutations in Ca V 1.2S 4E (versus Ca V 1.2S as wild-type control) affect the time-dependent decline of whole-cell I Ca and the modulation by added PIP 2 (by intracellular application of a 100 M concentration of the water-soluble PIP 2 analogue diC8-PIP 2 , (15, 64)) or phosphoinositide hydrolysis induced by wortmannin (20 M) plus   's t test). Significance was defined as p Ͻ 0.05 (a, b, and c), p Ͻ 0.01 (aa, bb, and cc), and p Ͻ 0.001 (aaa, bbb, and ccc). For reliable calculation of Q ON , we omitted data from recordings in which Q ON was not larger than 4-fold the signal of noise (5 of a total of 79 recordings).  5A). Perfusion of cells with control solution induced a slow decrease in activity with time ("run-down"), as expected for Ca v 1.2 channels (65). I Ca decline during perfusion with control solution was similar in wild-type and the mutant channel and was also not affected by intracellular application of diC8-PIP 2 (Fig. 7). However, extracellular perfusion with wortmannin/m-3M3FBS significantly inhibited I Ca of wild-type and Ca v 1.2 4E channels, but this current inhibition was significantly attenuated in the mutant channel (Fig. 7). This indicates that membrane association of the distal I-II linker through its positive charges not only stabilizes a more reluctant channel state ( Fig.  6) but also weakens inhibition of channel activity by phosphoinositide depletion.

Ca
Structural Modeling of the Domain I-II Linker Region of Ca V 1.2 Channel-To further explore the conformation of this distal domain I-II linker region of the Ca V 1.2 channel (corresponding to residues 536 -554 containing the polybasic cluster), its interaction with the inner leaflet of the membrane bilayer, and potential changes introduced by charge neutralizations, we applied the Rosetta membrane method (39 -41) and the x-ray structure of the voltage-sensing domain of a recently crystallized bacterial voltage-gated Na ϩ channel, Na V Ab (Fig.  8A) as described under "Experimental Procedures." The best models of the native Ca V 1.2 channel revealed convergence between the top cluster model (the most frequently sampled conformation) and the 10 lowest energy models (Fig. 8B). In those models, the distal I-II linker forms a straight helix in  Cells were depolarized for 300 ms from a holding potential of Ϫ90 mV to V max . R values represent the remaining fraction of I Ca (15 mM) after 50, 100, or 250 ms. Parameters are indicated as means Ϯ S.E. for a given number of experiments (n). Statistical significances are indicated for comparisons of Ca V 1.2L versus Ca V 1.2L 4A and Ca V 1.2L 4E (a) (one-way analysis of variance with Bonferroni post-hoc test as indicated in the table) and Ca V 1.2L versus Ca V 1.2S (b and bb, unpaired Student's t test). Significance was defined as p Ͻ 0.05 (a and b) and p Ͻ 0.01 (bb). which arginine residues are oriented away from the hydrophobic layer of the membrane (lines in Fig. 8) and are therefore in a favorable position for possible interactions with lipid phosphate groups (Fig. 8C). Large hydrophobic residues in the domain I-II linker region are oriented toward the hydrophobic layer of the membrane, which is in good agreement with their favorable environment based on analysis of available membrane protein structures (Fig. 8D). The best models of the domain I-II linker arginines mutated to alanines or glutamates are shown in Fig. 8, E and F. Unlike in wild-type models, there is no convergence among any of the top cluster models and the best scoring models. Our model suggests that native arginines in the distal I-II linker strongly favor straight helix conformation, positioning it entirely at the interface between the hydrocarbon core and lipid head groups. Based on Rosetta predictions, the alanine mutations clearly disturb this conformation, making the helical region structurally unstable. We therefore suggest that alanine and glutamate mutations in the domain I-II linker not only affect its plasma membrane binding but also disturb secondary structure, leading to altered positioning of the domain I-II linker helix near the membrane-spanning helices of repeat II.

Discussion
We describe the identification of a polybasic motif within the I-II linker of the pore-forming ␣ 1 -subunit of voltage-gated Ca V 1.2 L-type Ca 2ϩ channels with biochemical features that allow its attachment to the inner leaflet of the plasma membrane via interaction with negatively charged phospholipids. First, we demonstrate that the I-II linkers of all four LTCC isoforms bind to the plasma membrane in living cells (Fig. 1). This enables this polypeptide to translocate large complexes of cytoplasmic proteins to the plasma membrane (different ␤-subunits, Rab3-interacting molecule in complex with ␤-subunits (49)). Second, this interaction is reversible upon pharmacological activation of PLC together with either Ca 2ϩ as an additional PLC activator (66) or with wortmannin (Fig. 5), which prevents resynthesis of polyphosphoinositides (22). Under both activating conditions, the time course of membrane unbinding of the I-II linker closely follows the time course of unbinding of the PIP 2 -selective sensor PH-PLC␦. No such translocation was found for ␤ 2a -subunits, which are membrane-anchored through their palmitoic acid side chain and thus insensitive to PLCs. Together, these findings provide strong experimental evidence that binding of the I-II linker through this polybasic motif to the plasma membrane also involves interaction with polyphosphoinositides. Third, the primary structure of our lipid-binding motif has the characteristics of previously reported polybasic motifs implicated in polyphosphoinositide-interactions of other ion channels or of small GTPases, such as Rit and K-Ras (Fig. 3C) (22). The polybasic motifs of these proteins are sufficient for their plasma membrane targeting. Similar to our motif, the removal of only a few residues in Rit was sufficient to completely prevent translocation (3 in Rit, 4 in the I-II linker) (22). Moreover, positively charged clusters required for interaction with membrane lipids have recently also been identified in the auxiliary ␤ 2e subunit of VGCC (67).
Using de novo molecular modeling, we predict this membrane binding motif to form a straight ␣-helix positioned entirely at the interface between the hydrocarbon core and lipid headgroups, with the positive charges facing the lipid phosphates. Moreover, according to the model, the four arginines not only support plasma membrane binding of the I-II linker but are also necessary to stabilize this secondary structure in the context of the membrane environment. This stabilization may also explain the functional changes we observed. We demonstrate that the four basic residues are by themselves determinants of Ca V 1.2 Ca 2ϩ channel function. In two different Ca V 1.2 ␣ 1 -subunit variants with distinct intrinsic open probabilities (61), neutralization of these charges enhanced Ca V 1.2-mediated inward currents resulting from increased tail Ca 2ϩ currents relative to total ON-gating charge. In addition, these mutations also shifted half-maximal activation to more negative potentials by increasing the slope of the current-voltage relationship. Both observations indicate that charge neutralization (or conversion) enhances the coupling between voltage sensing and pore opening. This is in accordance with the previous finding that the I-II linker encodes self-reliant molecular motifs for channel activation independent of ␤-subunits (68,69). Our data show that residues remote from the ␤-subunit interaction site in the I-II linker are required for stabilization of normal Ca V 1.2 channel function. Although only crystallization studies such as those recently described for Kir2.2 and GIRK2 K ϩ channels (23, 24) can ultimately confirm the direct interaction of channel domains with membrane lipids, our data pro- Transmembrane segments S1-S4 are underlined by black bars and labeled. Amino acids were colored using the Zappo color scheme in Jalview. B-D, transmembrane view of the ribbon representation of the top cluster models of the VSD of Ca V 1.2 with the 10 lowest energy Rosetta models superimposed in B and space-filling representations of arginine side chains in the domain I-II linker helix in C and of large hydrophobic side chains in the I-II linker helix in D. E and F, transmembrane view of ribbon representation of the top five clusters and 10 lowest energy Rosetta models of the Ca V 1.2 VSD of alanine mutants superimposed in E and of glutamate mutants superimposed in F. Models are colored in a rainbow scheme from blue (N-terminal region before S1 segment) to red (S4 segment). Transmembrane segments S1-S4 are labeled accordingly. Black bars, approximate location of the extracellular and intracellular edges of the hydrophobic layer of the membrane. vide strong evidence that the polybasic motif described here constitutively stabilizes channel function through membrane lipid binding. In addition, we also found that charge neutralization and conversion also reduced Q ON , revealing a reduction of functional channels at the plasma membrane. This strongly indicates that the interaction of this basic motif with negatively charged membrane lipids is also important for regulating the expression of functional Ca v 1.2 channels at the membrane surface.
Previous studies have identified other potential PIP 2 interaction sites within VGCC ␣ 1 -subunits. Mutation of Ile-1520 in segment IIIS6 of the Ca V 2.1 ␣ 1 -subunit reduced current rundown attributed to PIP 2 depletion (70), thus proposing a channel-stabilizing PIP 2 interaction with a region close to the pore. The C terminus of the Ca V 2.1 ␣ 1 -subunit has also been implicated in a PIP 2 -dependent regulation of the direct (and voltagedependent) inhibition of Ca V 2.1 channels by receptor-mediated G-protein activation. Evidence has been provided that this could be due to PIP 2 -mediated interaction with the plasma membrane because the C-terminal channel fragment showed binding to polyphosphoinositide strips in vitro (71). However, the polybasic cluster implicated in this interaction is not present in the C termini of LTCCs.
So far, we could not obtain evidence that the polybasic binding motif identified here is also important for the dynamic regulation of channel function by physiological modulatory pathways inducing phospholipid breakdown. Although we can show breakdown of PIP 2 in our transfected cells by activation of M1 receptors robust enough to induce cytoplasmic relocation of the PIP 2 -selective probe PH-PLC␦, this did not result in relocation of our I-II linker constructs. In electrophysiological studies, Oxo-M stimulation of tsA-201 cells transfected with M1 receptors and Ca V 1.2 induces a 33-60% inhibition of Ba 2ϩ currents through Ca V 1.2 (16), and more than half of this modulation is dependent on PIP 2 . This is compatible with models predicting that under these experimental conditions, 1-7% of the total PIP 2 and 12-40% of PIP remain non-hydrolyzed in these cells (55). This may be sufficient for maintaining the binding of the I-II linker during stimulation with Oxo-M. Indeed, it is known from studies with proteins containing polybasic targeting motifs, similar to the one described here in the I-II linker, that efficient plasma membrane unbinding requires more than just PIP 2 hydrolysis. The GTPases Rit and K-Ras require complete depletion of both PIP 2 and PIP or PI3K products. Binding still persists when only one of them is depleted (22,37). This is also in accordance with a previous study that observed that complete loss of the binding of the RasGRP1 plasma membrane targeting domain only occurred after combined treatment with the PLC activator m-3M3FBS and wortmannin and not with m-3M3FBS alone (56). In our experiments, dissociation of the Ca V 1.2 I-II linker from the plasma membrane was also achieved by combining m-3M3FBS with ionomycin-induced Ca 2ϩ increase. This suggests that Ca 2ϩ also plays a role in this mechanism. Thus, it requires a second mechanism that either further enhances polyphosphoinositide breakdown (wortmannin) or a strong intracellular Ca 2ϩ signal.
The requirement of Ca 2ϩ could also indicate that PLC␦ or phosphatidylserine plays a major role in reversing the plasma membrane targeting of the Ca v 1.2 I-II linker. PLC␦ is mainly activated by an increase in intracellular Ca 2ϩ levels (10 -100 nM), usually due to prior activation of other PLCs (72)(73)(74). Moreover, it has been shown that sustained Ca 2ϩ increase can cause phosphatidylserine externalization (75). The observed translocation of the I-II linker upon m-3M3FBS and ionomycin application could therefore also be due to dissociation from the negatively charged phosphatidylserine. Hence, we propose that the I-II linker can interact with phosphoinositides as well as with phosphatidylserine, which has also been reported for Ras-GRP1 plasma membrane targeting (56). Independent of the exact mechanism, our biochemical studies provide evidence for a strong and constitutive binding of the polybasic membrane binding motif identified here, which may involve different negatively charged phospholipids.
The observation of facilitated gating and enhanced Ca 2ϩ influx in our channel mutants is unexpected because recent work (76) has demonstrated inhibition of channel activity by M1 receptor-activated PIP 2 breakdown. We were unable to test the effects of our mutations on M1 receptor-induced channel inhibition because the co-transfection was toxic for the cells and precluded their use in patch clamp recordings. Instead, we show that inducing phosphoinositide breakdown by treatment of cells with wortmannin and the PLC activator m-3M3FBS causes the expected inhibition of current amplitude but that this response is significantly attenuated by disruptive mutations in the polybasic motif as shown for Ca v 1.2S 4E (Fig. 7). This further supports our interpretation that this membrane binding motif not only stabilizes a more reluctant gating mode but also supports moderation of channel activity by lipid breakdown once targeted to the plasma membrane.
The lipid-binding motif described here is positioned on the C-terminal end of the I-II linker. In contrast to the N-terminal half of the linker, this region has not previously been implicated in the control of Ca 2ϩ channel function. The ␤-subunit tightly binds to a conserved motif within the N-terminal half of the I-II linker. X-ray crystallography and mutational studies (77)(78)(79) revealed that an ␣-helix between the AID (and its bound ␤-subunit) and IS6 provides a rigid connection to the channel pore and thereby crucially determines the voltage-and Ca 2ϩ -dependent gating properties of the channel. In Ca V 2 channels, direct G-protein modulation is also mediated by G␤␥ interaction in this region (77,80). In contrast to this proximal region, no x-ray structure could be obtained C-terminal to the AID in a recent x-ray crystallographic study of the I-II linker in complex with ␤-subunit (77). Our data suggest that also this portion of the I-II linker forms an ␣-helix and that its interaction with membrane lipids plays a crucial role in stabilizing the channel conformation. This conformational stabilization may be transmitted either upstream through the I-II linker to the pore (IS6) or downstream through IIS1 to the voltage sensor of the repeat II, or both, and thereby control channel gating.
Based on recent genetic findings that already small changes in LTCC functions underlie human diseases (81), this polybasic motif may also be a target for human disease-causing mutations. Further studies must therefore investigate the consequences of single charge neutralizations within this motif as well as the role of adjacent positively charged residues for membrane binding and channel function.
Author Contributions-G. K. and J. S. conceived the study. G. K. designed, performed, and analyzed the experiments shown in Figs. 1, 2, 3, and 5. A. P. designed, performed, and analyzed the experiments shown in Fig. 4