A Quartet of Leucine Residues in the Guanylate Kinase Domain of CaVβ Determines the Plasma Membrane Density of the CaV2.3 Channel*

Background: CaVβ subunits stimulate cell surface expression of CaV2.3 channels. Results: Of 33 positions and domains tested, leucine mutants in the guanylate kinase domain of CaVβ3 decreased significantly the surface protein density of CaV2.3. Conclusion: Leucine residues are responsible for the functional modulation by CaVβ. Significance: A quartet of leucine residues forms the hydrophobic pocket surrounding the α-interacting domain of CaV2.3. CaVβ subunits are formed by a Src homology 3 domain and a guanylate kinase-like (GK) domain connected through a variable HOOK domain. Complete deletion of the Src homology 3 domain (75 residues) as well as deletion of the HOOK domain (47 residues) did not alter plasma membrane density of CaV2.3 nor its typical activation gating. In contrast, six-residue deletions in the GK domain disrupted cell surface trafficking and functional expression of CaV2.3. Mutations of residues known to carry nanomolar affinity binding in the GK domain of CaVβ (P175A, P179A, M195A, M196A, K198A, S295A, R302G, R307A, E339G, N340G, and A345G) did not significantly alter cell surface targeting or gating modulation of CaV2.3. Nonetheless, mutations of a quartet of leucine residues (either single or multiple mutants) in the α3, α6, β10, and α9 regions of the GK domain were found to significantly impair cell surface density of CaV2.3 channels. Furthermore, the normalized protein density of CaV2.3 was nearly abolished with the quadruple CaVβ3 Leu mutant L200G/L303G/L337G/L342G. Altogether, our observations suggest that the four leucine residues in CaVβ3 form a hydrophobic pocket surrounding key residues in the α-interacting domain of CaV2.3. This interaction appears to play an essential role in conferring CaVβ-induced modulation of the protein density of CaVα1 subunits in CaV2 channels.

Voltage-dependent Ca 2ϩ channels (Ca V ) are membrane proteins that play a key role in promoting Ca 2ϩ influx in response to membrane depolarization in excitable cells. The total Ca 2ϩ influx through Ca V proteins is controlled by the single-channel conductance, the probability that the channel is open at a given time and voltage, what is also referred to as gating, and the number of proteins expressed at the membrane. Modulating any of these parameters could be used to alter Ca 2ϩ influx and prevent excitable cells from Ca 2ϩ overload.
Voltage-gated Ca 2ϩ channels form oligomeric complexes that are classified according to the structural properties of the pore-forming Ca V ␣1 subunit. The primary structures for 10 distinct Ca V ␣1 subunits (1-7) are classified into three main subfamilies according to their high voltage-activated gating (HVA Ca V 1 and Ca V 2) or low voltage-activated gating (LVA Ca V 3). In HVA Ca V 1 and Ca V 2 channels, auxiliary subunits include a cytoplasmic Ca V ␤ subunit, a mostly extracellular Ca V ␣2␦ subunit, and calmodulin constitutively bound to the C terminus of Ca V ␣1 (7-12) (for review see Ref. 9).
There are four subfamilies of Cav␤s, each with multiple splicing isoforms and unique modulatory functions (14,15). Auxiliary Ca V ␤ subunits modulate activation gating as well as inactivation kinetics of the main Ca V ␣1 subunits of HVA Ca V 1 and Ca V 2 channels (16 -23). They also promote cell surface expression in part by preventing protein ubiquitination and degradation of the protein by the ERAD complex (20,24).
The principal Ca V ␣1-Ca V ␤ interaction site on the poreforming ␣1 subunit is a conserved 18-residue sequence in the I-II loop called the ␣-interacting domain (AID) 3 (25). The nanomolar affinity between the AID helix and the Ca V ␤ protein has been thoroughly investigated (26 -29). It is primarily secured by the projection of the conserved tryptophan and isoleucine (WI) pair of residues onto the Ca V ␤ fold (27)(28)(29). Single point mutations of the C-terminal WI residues on the AID decreased by 1000-fold the affinity of the Ca V ␤2 for Ca V 1.2 (26) and decreased its cell surface density (17).
Less is known in regard to the molecular determinants in Ca V ␤ that control protein density at the plasma membrane and modulate the activation gating of the channel. It is understood that the N terminus plays a predominant role in modulating inactivation kinetics, either directly by interacting with the Ca V ␣1 (30,31) and/or indirectly through palmitoylation for Ca V ␤2a (32). High resolution three-dimensional crystal struc-tures have shown that Ca V ␤ subunits consist of five distinct domains: the N terminus, a Src homology 3 (SH3) domain, a HOOK region, a guanylate kinase (GK) domain, and the C terminus (see Fig. 5A) (27)(28)(29). The GK core module appears to be sufficient to confer the activation properties (33)(34)(35). The latter includes the structural determinants of the ␣-binding pocket that are distributed among ␣3, ␣6, ␣7, and ␣9 helices and ␤9 and ␤10 sheets (26 -29).
With some notable exceptions (23,26,36), few studies have systematically attempted to correlate Ca V ␤ high affinity binding with the modulation of channel gating and trafficking. Whereas interaction between Ca V ␤ and Ca V ␣1 is required for the trafficking of Ca V ␣1 to the plasma membrane, it remains to be seen whether nanomolar binding of Ca V ␤ to the AID motif of Ca V ␣1 is required to carry both roles. Furthermore, whereas the binding of Ca V ␤ on the AID of Ca V ␣1 has been well characterized (17,22,26,33,37,38), few concur on the GK residues in Ca V ␤ that determine protein density at the plasma membrane and contribute to channel modulation.
Complications have arisen in functional studies performed with the Xenopus recombinant system (26, 27, 33, 39 -41), including our own (37,38), because endogenous Ca V ␤s (42) are likely to boost expression of noninteracting mutants. The presence of the Ca V ␣2␦ subunit in some experiments could complicate data interpretation. In addition, the Ca V ␤2a subunit used in most studies is an atypical subunit with an N-terminal palmitoylation site (43) that anchors it to the membrane and could offset the disruption of the high affinity interaction site. Indeed, the palmitoylated Ca V ␤2a was still able to modulate the biophysical properties of Ca V 2.2 W391A in Xenopus oocytes, indicating that the plasma membrane anchoring afforded by its palmitoylation can substitute in some cases for high affinity interaction with the I-II linker (22).
In this work we have systematically addressed the role of the structural domains of Ca V ␤3 in regard to protein density and channel function after recombinant expression of HVA Ca V 2.3 in HEKT cells. Selective deletion of the SH3 and HOOK structural domains did not significantly alter the cell surface density or the channel functional modulation by Ca V ␤. More surprisingly, mutations of residues in the GK domain that were previously identified as underlying the affinity to HVA Ca V ␣1 (26) caused little change in the cell surface density of Ca V 2.3 or in the channel activation gating. Nonetheless, point mutations within a quartet of leucine residues in the ␣3, ␣6, ␤10, and ␣9 regions of the GK domain significantly decreased cell surface density. Altogether our results suggest that affinity interaction in the micro-to nanomolar range is sufficient to carry the typical Ca V ␤-induced hyperpolarizing shift in the channel activation gating in the presence of overexpressed Ca V ␤ subunits. Furthermore, we have identified a quartet of leucine residues in the GK domain of Ca V ␤3 that play a critical role in promoting surface plasma density of Ca V 2.3.
The HA epitope tag (YPYDVPDYA) was inserted in the first extracytoplasmic predicted loop in Domain I at position 367 (nucleotides) for Ca V 2.3. The biophysical properties of the HAtagged Ca V ␣1 subunit of Ca V 2.3 expressed in HEKT cells with the auxiliary Ca V ␤3 subunit were found not to be significantly different from the wild-type Ca V 2.3 channel expressed under the same conditions (see Table 1). Ca V ␤3 deletion mutants were produced as described elsewhere (17). The Ca V ␤3 58 -362 fragment was used previously in (17). The boundaries of the five domains/regions of the rat Ca V ␤3 are: N terminus, Met 1 -Pro 59 ; SH3 domain, Val 60 -Ser 123 and Pro 170 -Pro 175 ; HOOK region, Pro 120 -Pro 169 ; GK domain, Ser 176 -Thr 360 ; and C terminus, His 361 -Tyr 484 (see Fig. 5A).
Cell Culture and Transfections-HEK293T or HEKT were grown in high glucose DMEM supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin at 37°C under 5% CO 2 atmosphere as described elsewhere (17). RT-PCR conducted in these cells failed to highlight the presence of Ca V ␤ and Ca V ␣2␦ auxiliary subunits (47). HEKT cells (80% confluence) were transiently transfected with similar amounts of DNA (4 g each or 12 g total): HA-Ca V 2.3, Ca V ␤3, Ca V ␣2␦-1, or empty vector pCMVTag5 in 10 l of Lipofectamine 2000 (Qiagen) using a DNA:lipid ratio of 1:2.5 as described elsewhere (17). Transfection rate of the control peGFP plasmid was estimated to be 66 Ϯ 2% (n ϭ 8) as assessed by flow cytometry from the fluorescence of the GFP. Preliminary tests showed that Ca V 2.3 protein expression peaked 24 -30 h after transfection.
Quantification of Ca V 2.3 Surface Expression with FACS-FACS experiments were conducted and analyzed as described elsewhere. Briefly, 24 -28 h after transfection, the cells were harvested and stained with the anti-HA FITC conjugate (10 g/ml) at room temperature for 45 min. A maximum of 10,000 cells resuspended in standard PBS were counted using a FACScalibur flow cytometer (BD Biosciences) with a FITC filter (530 nm) at the flow cytometry facility located in the Department of Microbiology of the Université de Montréal. The level of background fluorescence was adjusted with cells incubated in the absence of the fluorophore. The parameters of the M1 region were set from Gaussian distribution of the fluorescence intensity obtained in the absence of the FITC antibody. The M2 region was calculated from the Gaussian distribution observed in the higher decades of the fluorescence log scale in the presence of the FITC antibody. The cell fluorescence intensity and, by extension, the cell surface expression of the HA tag were estimated from the M2 peak value (see supplemental Each novel mutant and/or experimental condition was conducted in triplicate and sometimes repeated over 2 weeks. The experiments performed in the 24 -28-h range after transfection yielded M2 values with an experimental variation lower than 10% between samples and between series of experiments. In any case, a maximum of 10% variation in the measure of fluorescence was tolerated. All control conditions were pooled and reported in Tables 3 and 4. Patch Clamp Experiments in HEKT Cells-Whole cell voltage clamp recordings were performed 30 h after transfection using the methods described above in the presence of the peGFP vector (0.2 g) as a control for transfection. Patch clamp experiments were carried out with the Axopatch 200-B amplifier (Molecular Devices, Union City, CA). Electrodes were filled with a solution containing 140 mM CsCl, 0.6 mM NaGTP, 3 mM MgATP, 10 mM EGTA, 10 mM Hepes titrated to pH 7.3 with NaOH. Pipette resistance ranged from 2 to 4 M⍀. The cells were bathed in a modified Earle's saline solution 135 mM NaCl, 20 mM TEACl, 2 mM CaCl 2 , 1 mM MgCl 2 , 10 mM Hepes titrated to pH 7.3 with KOH. PClamp software Clampex 10.2 coupled to a Digidata 1440A acquisition system (Molecular Devices) was used for on-line data acquisition and analysis. Pipette and cell capacitance cancellation and series resistance compensation were applied (up to 80%) using the cancellation feature of the amplifier. Cellular capacitance was estimated by measuring the time constant of current decay evoked by a 10-mV depolarizing pulse applied to the cell from a holding potential of Ϫ100 mV. A series of 150-ms voltage pulses were applied from a holding potential of Ϫ100 mV at a frequency of 0.2 Hz, from Ϫ80 to ϩ60 mV at 5-mV intervals. Unless stated otherwise, the data were sampled at 5 kHz and filtered at 1 kHz. Experiments were performed at room temperature (20 -22°C). Activation parameters were estimated from the peak I-V curves obtained for each channel combination and are reported as the means of individual measurements Ϯ S.E. as described elsewhere (48). Briefly, the I-V relationships were normalized to the maximum amplitude and were fitted to a Boltzmann equation with E 0.5,act being the midpotential of activation. The free energy of activation was calculated using the midactivation potential with Equation 1, where z is the effective charge displacement during activation, and T, F, and R have their usual meaning (49). The r 50 ratio is defined as the ratio of peak whole cell currents remaining 50 ms later (I 50 ms /I Peak ). It has been used elsewhere to estimate inactivation kinetics of Ca V 2.3 (48,50,51).
For each novel mutant tested, the biophysical parameters of Ca V 2.3 WT ϩ Ca V ␤3 WT were measured the same day under the same experimental conditions from the transfection protocol up to the bath solutions. Experiments performed under the same conditions yielded peak current densities Ϯ 15% between samples and between series of experiments. All of the experiments were pooled and are reported in Tables 1 and 2. Ca V 2.3 wild-type or HA-Ca V 2.3 channels were expressed in HEKT cells with Ca V ␤3 WT or constructs. Biophysical parameters were measured in the presence of a physiological saline containing 2 mM Ca 2ϩ as described elsewhere (17). Ca V ␣2␦-1 was omitted. Activation properties (E 0.5,act and ⌬G act ) were estimated from the mean I-V relationships and fitted to a Boltzmann equation. The data are shown as the means Ϯ S.E. of the individual experiments, and the number of experiments appears in parentheses.

TABLE 2
Biophysical properties of Ca V 2.3 channels in the presence of Ca V ␣2␦-1 Ca V 2.3 wild-type channels were expressed in HEKT cells with Ca V ␤3 WT or constructs with Ca V ␣2␦-1. Biophysical parameters were measured in the presence of a physiological saline containing 2 mM Ca 2ϩ as described elsewhere (17). Activation properties (E 0.5,act and ⌬G act ) were estimated from the mean I-V relationships and fitted to a Boltzmann equation. The data are shown as the means Ϯ S.E. of the individual experiments, and the number of experiments appears in parentheses.

RESULTS
␣-Interacting Domain and Cell Surface Expression of Ca V 2.3-Co-expression of Ca V 2.3 with Ca V ␤3 increased whole cell currents in HEKT cells from Ϫ6 Ϯ 1 pA/pF (n ϭ 18) (Table 1) for the wild-type Ca V 2.3 channel expressed alone to a current density of Ϫ18 Ϯ 2 pA/pF (n ϭ 49) for Ca V 2.3 ϩ Ca V ␤3. Similar results were obtained for the HA-tagged Ca V 2.3 channels (Table 1). Co-expression with Ca V ␣2␦-1 subunit further increased whole cell currents to Ϫ65 Ϯ 5 pA/pF (n ϭ 95) (Fig. 1, A and B, and Table 2), confirming that Ca V ␣2␦-1 increases whole cell current density of HVA Ca V 1 and Ca V 2 channels (16,17). The increase in cell current density was accompanied by an apparent acceleration of the inactivation kinetics with r 50 values decreasing from 0.38 Ϯ 0.01 (n ϭ 49) to 0.25 Ϯ 0.01 (n ϭ 95) at ϩ10 mV in the presence of Ca V ␣2␦-1 (Fig. 1C) but failed to significantly shift the midpotential of activation (E 0.5,act ) toward negative potentials (Tables 1 and 2), an observation that was also reported elsewhere (47).
Flow cytometry assays carried out with the HA-tagged Ca V 2.3 in the stable Ca V ␤3 confirmed that Ca V ␤ is the critical auxiliary subunit in stimulating the plasma membrane density of Ca V 2.3 (Fig. 1D). Roughly 40% of the cells transfected with HA-Ca V 2.3 and Ca V ␤3 were fluorescent, suggesting that a substantial fraction of Ca V 2.3 proteins remained in cytoplasmic compartments as noted for Ca V 1.2 (17,20). Ca V ␤1a, Ca V ␤1b, and Ca V ␤4 isoforms were as proficient as Ca V ␤3 in chaperoning Ca V 2.3 to the membrane, but the palmitoylated Ca V ␤2a did

Fluorescence in HEKT
not appear to be as powerful as the other Ca V ␤ subunits ( Fig. 2 and Tables 1 and 3). There was no significant increase in cell fluorescence when overexpressing Ca V ␣2␦-1 or calmodulin wild type alone with Ca V 2.3 (Table 3). Because the whole cell current density results from the NP o ⌬ i product, where N is the number of proteins at the membrane, P o is the open channel probability, and ⌬ i is the single channel conductance, these data suggest that Ca V ␣2␦-1 improves the open channel probability rather than increasing the number of channels at the membrane as shown for Ca V 1.2 (17,20). The strong functional modulation by Ca V ␣2␦-1 contrasts with its milder effects on the total Ca V 2.3 protein density, especially when compared with the robust increase conferred by Ca V ␤3 (Fig. 3).
To examine which AID residues (Fig. 4A) are required to promote cell surface expression of Ca V 2.3, conserved and nonconserved residues in the AID were mutated and functionally expressed in HEKT cells. Cell surface expression were maintained with Ca V 2.3 G382A, Y383A, and Y383F (Fig. 4, B and C), although the cell surface density decreased significantly for Y383G. The Ca V ␤3 modulation of activation gating and inactivation kinetics of Y383F and Y383A mutants was not significantly different from Ca V 2.3 WT (Fig. 4D and Table 1). Nonetheless, W386A failed to migrate to the plasma membrane. Hence, only the most severe reduction in the binding affinity as seen with W386A (26) was found to disrupt plasma membrane density.
Isothermal titration calorimetry assays (26) showed that the mutation of the nonconserved arginine residue (Arg 384 ) located between the GY and WI in the AID helix increased the binding affinity of the Ca V 2.3 peptide for Ca V ␤2a with a K d decreasing from 54 to 8.6 nM. Ca V 2.3 R384M and R384L mutations, however, did not significantly influence the number of Ca V 2.3 proteins at the membrane, nor did they alter the Ca V ␤3 modulation of channel gating (Fig. 4 and Table 1).

The SH3 and HOOK Domains of Ca V ␤ Are Not Essential for
Ca V ␤-mediated Modulation-We next addressed the importance of the different structural domains of Ca V ␤ in the Ca V ␤induced modulation of cell surface density and channel gating of Ca V 2.3. Co-expression with the SH3-HOOK-GK fragment of Ca V ␤3 (in amino acids 58 -362) boosted cell surface expression and generated whole cell currents with biophysical parameters not significantly different from the wild-type version of Ca V ␤3 ( Fig. 5 and Table 1), suggesting that the C-terminal does not contribute significantly to functional modulation in contrast to results obtained with Ca V ␤2a (19).
The SH3 domain was shown to support channel endocytosis (53) and to promote calpain-mediated Ca V ␤3 proteolysis through a PEST-like motif (54). Deletion of the N-terminal of SH3 domain (⌬57-123), deletion of the six-residue ␤5 strand (⌬170 -175 or ⌬PYDVVP, formerly known as the ␤-interaction domain or BID (55)), or deletion of both regions (⌬SH3) yielded channels with robust peak current density and typical voltagedependent activation gating in the absence and in the presence of Ca V ␣2␦-1 as a background subunit (Fig. 7 and Tables 1 and 2). Deletion of the HOOK domain (⌬122-169 and ⌬122-175) in Ca V ␤3 did not significantly alter any of the biophysical parameters (activation potential, peak current density, and inactivation kinetics) ( Fig. 7 and Table 1). In contrast, small six-residue deletions in the GK domain with Ca V ␤3 ⌬␤6 or    Table 1 ⌬175-180 and Ca V ␤3 ⌬195-200 failed to promote cell surface density of Ca V 2.3 channels (Fig. 7B).
Mutations of Leucine Residues in the GK Domain Decrease Plasma Membrane Density-A mutational analysis was thus carried to seek out single residues in the GK domain that are responsible for modulation of gating and cell surface density of Ca V 2.3. We focused on residues of Ca V ␤3 that were highly likely to interact with Trp 386 and Ile 387 residues in Ca V 2.3 (38) based upon the changes in the energetics of interaction between the AID peptide from Ca V 1.2 and Ca V ␤2a (26), as well   (Table 4).  Table 1.
as from the predictions of the homology model (Fig. 6). Note that the GK domains of Ca V ␤2a and Ca V ␤3 are well conserved with more than 87% identity in this stretch of 188 amino acids. Residues in Ca V ␤3 were either substituted by glycine to minimize side chain interactions or substituted with alanine to preserve the ␣-helicoidal structure.
Mutations at position Arg 307 (R307A, R307G, and R307K) in the ␣6 helix significantly increased peak current densities as compared with Ca V ␤3 WT (Tables 1 and 2). In addition, the Ca V ␤3 Arg 307 mutants decreased the inactivation kinetics, suggesting that the open state was stabilized by these mutants.
These effects were observed despite a documented decrease in the affinity of the equivalent R356A mutation in Ca V ␤2a (equivalent to Arg 307 in Ca V ␤3) with a K d increasing from 5 to 345 nM (26).
In contrast, point mutations of four leucine residues (L200G, L303G, L337G, and L342G) significantly decreased cell surface density of Ca V 2.3 by Ϸ50 -60% (Fig. 8). The three-dimensional homology model suggests that the four leucine residues in Ca V ␤3 could form a hydrophobic pocket surrounding Ca V 2.3 Trp 386 and Ca V 2.3 Ile 387 (Fig. 9). Note that the strongest effects were observed with Ca V ␤3 L303G, which is predicted to interact with Ile 387 in Ca V 2.3. Multiple point mutations were constructed to test the hypothesis that the four leucine residues form a single interaction site. The double, triple, and quadruple Ca V ␤3 mutants L200G/L303G, L200G/L303G/L347G, and L200G/L303G/L337G/L342G Ca V ␤3 mutants all ablated cell surface density of Ca V 2.3 (Table 4). Furthermore, the peak current densities measured with these Ca V ␤3 mutants were not statistically different (p Ͼ 0.1) from the densities measured in the absence of Ca V ␤3 (Tables 1 and 2). Finally, the normalized protein density of Ca V 2.3 measured in membrane lysates was significantly decreased with Ca V ␤3 L303G as compared with Ca V ␤3 WT but was nearly abolished with the quadruple Ca V ␤3 leucine mutant L200G/L303G/L337G/L342G (Fig 10). In the context where the fluorescence level measured in our FACS assay results from the net balance between anterograde and retrograde trafficking, this observation suggests that the decreased protein density at the plasma membrane could result from a decrease in the Ca V 2.3 total protein density.

DISCUSSION
There is considerable interest in identifying molecules that modulate protein-protein interactions in vivo. In this regard, specifically modulating the interaction of Ca V ␤ with HVA Ca V channels could be a strategy to design new HVA Ca V agonists and antagonists. Residues of Ca V ␤ carrying nanomolar affinity binding onto the Ca V ␣1 of Ca V 1.2 have been identified from high resolution three-dimensional crystal structures and isothermal calorimetry assays (26 -29). Although mutations of residues in Ca V ␣1 abolishing the protein-protein interaction are incompatible with Ca V 1.2 function (17,26), there has been some divergence regarding the functional importance of the domains in Ca V ␤ subunits (26, 27, 33, 39 -41). The range of approaches and recombinant systems could explain in part the diverse conclusions. We have thus systematically addressed the role of Ca V ␤ in channel function based upon the structural data currently available. Our results show that protein-protein affinity measured in vitro may not be the only predictor of channel modulation especially in the context where Ca V ␤ subunits control surface density and prevent degradation of Ca V ␣1 subunits from Ca V 1 and Ca V 2 channels by the ERAD complex.
Site Occupancy Is Sufficient to Modulate Protein Density and Gating-It is well understood that high affinity binding of Ca V ␤ onto the AID motif appears to be a prerequisite for both Ca V ␤induced modulation of gating and Ca V ␤-stimulated plasma membrane trafficking of Ca V ␣1 (17,22,56,57). In our hands, mutations within the AID of Ca V 2.3 that were shown to moderately increase (R384M and R384L) or decrease (G382A, given that the reverse observation was reported in the Xenopus expression system. AID-deficient Ca V 2.3 channels migrated to the plasma membrane but were not functionally modulated by Ca V ␤ when expressed in Xenopus oocytes (37,38). It certainly raises questions about the transposition of data from Xenopus oocytes to mammalian cells.
Partial Deletions of the GK Domain of Ca V ␤3 Disrupt Plasma Membrane Density of Ca V 2.3-The deletion of the HOOK region and thus the absence of intramolecular coupling between the SH3 and GK in Ca V ␤ did not impact significantly the voltage-dependent activation gating or the inactivation kinetics of Ca V 2.3 despite the report that the variable HOOK domain plays a role in the inactivation kinetics of Ca V ␤2a (40). Our observation also seems to contradict a previous report that strong SH3-GK intramolecular coupling conferred by short linkers (Ͻ3 amino acids) confers fast inactivation kinetics (19,  (Table 4). C, normalized peak current densities are plotted as a function of applied voltage. All of the Ca V ␤3 mutants produced Ca V 2.3 currents with similar voltage-dependent activation (Table 1). D, the r 50 values are shown Ϯ S.E. from Ϫ10 to ϩ20 mV for Ca V 2.3 WT in the presence of Ca V ␤3 WT, Ca V ␤3 L200G, Ca V ␤3 L303G, Ca V ␤3 L337G, and Ca V ␤3 L342G (from left to right on the bar graph). Inactivation kinetics were significantly slower for Ca V ␤3 L342G at all voltages (p Ͻ 0.01).

35)
. It also disagrees with the observation that the ␤5 sheet in the SH3 domain of Ca V ␤2a plays an important role in the modulation of Ca V 2.1 channels (35). Discrepancies can be attributed to the idiosyncrasy of the Ca V ␤2a subunit and/or the Xenopus expression system used in those studies. In our hands, complete deletion of the SH3 or the HOOK domains in Ca V ␤3 failed to abrogate either its chaperone function or its modulation of channel gating.
Leucine Residues in the GK Domain of Ca V ␤3 Determine Cell Surface Density of Ca V 2.3-Close to 20 point mutations were performed in the GK domain of Ca V ␤3. The majority of the mutations did not alter the Ca V ␤-induced stimulation of cell surface density and hyperpolarization of gating despite a documented decrease in the binding affinity (26). Among the Ca V ␤ mutations herein tested, four mutations (M196A, L303G, R307A, and L342G) were expected to decrease significantly the binding affinity of Ca V ␤3 to Ca V 2.3 (26). Ca V ␤2a M245A (equivalent to Ca V ␤3 M196A) yielded a 200-fold decrease in the affinity for Ca V 1.2 (26), with the K d increasing from 5 nM to 1.1 M. Nonetheless, whole cell currents of Ca V 2.3 obtained in the presence of Ca V ␤3 M196A were not significantly different from those obtained with Ca V ␤3 WT both in terms of peak current density, voltage dependence of activation, and inactivation kinetics. In contrast, mutations of leucine residues at positions 200, 303, 337, and 342 (either individually or in combination) significantly reduced the modulation of Ca V 2.3 function and its cell surface density when compared with Ca V ␤3 WT. Remarkably, the affinities measured between the AID peptide of Ca V 1.2 and the Ca V ␤2 mutants L352A and L392A (equivalent to Ca V ␤3 L303 and L342) were Ϸ20-fold lower than for Ca V ␤2 WT but still 100-fold higher than with Ca V ␤2 M245A (26). Our functional screen thus discriminated residues in Ca V ␤3 that were initially believed to behave similarly based upon their binding energies.
It remains possible that the binding affinities of the Ca V 2.3 ϩ Ca V ␤3 complex cannot be simply extrapolated from the in vitro binding energies measured with Ca V ␤2a bound onto the AID peptides from Ca V 1.2 channels (26). A recent study performed with larger peptides (Ϸ30 -40 residues) points to important structural differences between the highly conserved I-II linkers of Ca V 1.2 and Ca V 2.2 (13).
With these reservations in mind, our results altogether suggest that nanomolar protein-protein affinity may not be the sole determinant of the modulation of Ca V ␣1 channel function by Ca V ␤. Given that Ca V 2.3 W386A was not measured at the membrane but that Ca V ␤3 M196A sustained surface density of Ca V 2.3 proteins, it can be concluded that functional modulation occurs in the nano-to micromolar range. Even with a moderate affinity, "occupancy of the AID site" could be sufficient to carry the Ca V ␤-induced modulation of channel function in HVA Ca V 1 and Ca V 2 channels (23). In this scheme, occupancy of the AID site by Ca V ␤ could either unmask retention signals for protein targeting or mask ubiquitination sites on the Ca V ␣1 subunit and prevents its degradation by the proteasome.
Conclusion-Mutations of four leucine residues (Leu 200 , Leu 303 , Leu 337 , and Leu 342 ) were each expected to moderately decrease binding affinity, significantly decreasing the cell surface density of Ca V 2.3 protein. Simultaneous mutation of the four leucine residues completely abolished surface and total protein density of Ca V 2.3. We propose that these four leucine residues form a hydrophobic pocket that is required to promote van der Waals interactions with Trp 386 and Ile 387 (WI pair) in the AID region of Ca V 2.3. Hence, the four leucine residues in the GK domain of Ca V ␤3 and the WI pair in the AID of Ca V 2.3 appear to be essential determinants in the modulation of high voltage-activated calcium channel function by Ca V ␤ auxiliary subunits.