The α1-β-Subunit Interaction That Modulates Calcium Channel Activity Is Reversible and Requires a Competent α-Interaction Domain*

High voltage-gated calcium channels consist of a pore-forming subunit (α1) and three nonhomologous subunits (α2/δ, β, and γ). Although it is well established that the β-subunit promotes traffic of channels to the plasma membrane and modifies their activity, the reversible nature of the interaction with the α1-subunit remains controversial. Here, we address this issue by examining the effect of purified β2a protein on CaV1.2 and CaV2.3 channels expressed in Xenopus oocytes. The β2a-subunit binds to the α1-interaction domain (AID) in vitro, and when injected into oocytes, it shifts the voltage dependence of activation and increases charge movement to ionic current coupling of CaV1.2 channels. This increase depended on the integrity of AID but was not abolished by bafilomycin, demonstrating that the α1-β interaction through the AID site can take place at the plasma membrane. Furthermore, injection of β2a protein inhibited inactivation of CaV2.3 channels and converted fast inactivating CaV2.3/β1b channels to slow inactivating channels. Inhibition of inactivation required larger concentration of β2a in oocytes expressing CaV2.3/β1b channels than expressing CaV2.3 alone but reached the same maximal level as expected for a competitive interaction through a single binding site. Together, our data show that the α1-β interaction is reversible in intact cells and defines calcium channels β-subunits as regulatory proteins rather than stoichiometric subunits.

extent to which both processes are independent from each other remain elusive. Early studies show that in Xenopus oocytes, coexpression of ␤ 2a with the pore-forming ␣ 1 subunit from cardiac cells (Ca V 1.2) augments ionic currents mostly by increasing ionic current to charge movement ratio (3). Later on, it was shown that the addition of the ␤-subunit as purified protein is capable of modulating channel activity of the ␣ 1 subunit expressed in Xenopus oocytes (4,5) and also on isolated membranes from skeletal muscle (6). These results suggest that modulation of function is separated from the effect on channel expression and predicts that binding sites remain available on the mature channel. However, the ␣ 1 -␤-subunit association depends primarily on the so-called ␣-interaction domain (AID), 2 located within the intracellular loop joining the first and second repeats of the ␣ 1 -subunit. Secondary binding sites have been identified, but they appear to be specific to certain ␣ 1 -␤ pairs, and they modulate particular aspects of channel function (7,8). Recent work indicates that binding of a single ␤-subunit recapitulates function (9). Hence, it seems likely that ␤-subunit-binding sites available at the plasma membrane arise from unbinding of the ␤-subunit to the AID site during channel trafficking. This would imply a reversible interaction rather than a stoichiometric association, as proposed by Tareilus et al. (2). Here, we purified the ␤ 2a -subunit isoform, injected in oocytes expressing ␣ 1 -subunit, and measured ionic and gating currents. Gating currents were correlated to the number of channels in the membrane by immunoassay. Using this approach we demonstrate that the ␣ 1 -␤-subunit interaction is dynamic and can occur at the plasma membrane. Binding to mature channels is reversible, and association of the ␤-subunit to the AID site is an absolute requirement to modulate channel function.

EXPERIMENTAL PROCEDURES
Protein Preparation, Binding Assay, and Mutagenesis-The cDNA encoding the rat ␤ 2a subunit (3) was subcloned by conventional PCR methods into pRSET vector (Stratagene) to add an N-terminal polyhistidine tag (His 6 -␤ 2a ). The His 6 -␤ 2a was expressed in BL-21 (DE-3) Escherichia coli bacteria by 2 h of induction with 0.5 mM isopropyl-␤-D-thiogalactopyranoside at 37°C and purified from the cleared cell lysate by metal affinity chromatography (Talon; BD Biosciences) followed by size exclusion chromatography on a Superdex TM S-200 column (Amersham Biosciences) pre-equilibrated with buffer containing 50 mM Tris buffer, 300 mM NaCl, 1 mM EDTA, pH 8.0. The fractions containing the protein were pooled, concentrated up to 2-4 mg/ml, and stored at Ϫ80°C. Binding to AID was assayed as by Neely et al. (4), using a glutathione S-transferase (GST) fusion protein encoding the ⌱-⌱⌱ loop of the Ca V 1.2 subunit (GST-AID), and as a negative control we used GST alone or fused to a 126-amino acid peptide derived from the C-terminal end of the chloride channel from human skeletal muscle ClC-1 (GST-ClC 126 ).
Oocyte Injections and Electrophysiological Recordings-Xenopus laevis oocytes were prepared, injected, and maintained as in previous report (4). All capped cRNA were synthesized using the MESSAGE-machine (Ambion, Austin, TX), resuspended in 10 l of water and stored in 2-l aliquots at Ϫ80°C until use. The Ca V 1.2 subunit used in this study bears a deletion of 60 amino acids at the N-terminal end that increase expression (10), and the Ca V 2.3 subunit corresponds to the human form (11). The W470S mutation on the Ca V 1.2 subunit was incorporated by standard PCR methods. Electrophysiological recordings using the cut-open oocyte technique (12) with a CA-1B amplifier (Dagan Corp., Minneapolis, MN) were performed 4 -6 days after cRNA injection as described (4). The external solution contained in mM, 10 Ba 2ϩ , 96 n-methylglucamine, and 10 HEPES, pH 7.0, and the internal solution 120 n-methylglucamine, 10 EGTA, and 10 HEPES, pH 7.0. For recording of oocytes expressing the Ca V 2.3 subunit, EGTA was replaced by BAPTA (1,2-bis(o-aminophenoxy)ethane-N,N,NЈ,NЈ,-tetraacetic acid) in the internal solution for a better control of calcium-activated chloride currents. Data acquisition and analysis were performed using the pCLAMP system and software (Axon Instruments Inc., Foster City, CA). The currents were filtered at 2 kHz and digitized at 10 kHz. Linear components were eliminated by P/Ϫ4 prepulse protocol (3). For experiments with the purified ␤ 2a protein, the oocytes were injected with 50 nl of the protein stock solution 2-7 h before recordings. The final protein concentration was calculated assuming an oocyte volume of 525 nl that corresponds with a diameter of 1 mm. For bafilomycin treatment, the oocytes were exposed for 24 h to 500 nM concentration obtained from a 1 mM stock solution of bafilomycin A 1 (Sigma-Aldrich) in Me 2 SO and compared with control oocytes incubated for the same period in 0.05% Me 2 SO.
Surface Expression Measurements in Xenopus Oocytes-Surface expression of Ca V 1.2 was measured by immunoassay as described (13). The hemagglutinin (HA) epitope was inserted into the extracellular loop S5-H5 of domain II at residue 713 of the Ca V 1.2 subunit (Swiss-Prot P15381) by standard overlapping PCR using complementary oligonucleotides encoding the HA epitope and flanked with extra amino acids to yield Ca V 1.2-HA. 5-6 days after Ca V 1.2 RNA injection, the oocytes were separated into two groups: one for electrophysiological recordings and the other for immunoassay. Unless otherwise stated, all of the incubations for immunoassay were carried out at 4°C. The oocytes were incubated in ND96 solution (96 mM NaCl, 2 mM KCl, 1.8 mM CaCl 2 , 1 mM MgCl 2 , and 5 mM HEPES, pH 7.6) with 1% bovine serum albumin (blocking buffer) for 30 min and then for 60 min in the presence of 1 g/ml rat monoclonal anti-HA antibody (3F10; Roche Applied Science). After washes, the oocytes were incubated for 30 min with horseradish peroxidase-coupled secondary antibody (goat anti-rat FAB fragments; Jackson ImmunoResearch). Thereafter, the oocytes were extensively washed with blocking buffer and rinsed once in ND96 at room temperature. Individual oocytes were then placed in 50 l of SuperSignal enzyme-linked immunosorbent assay femto substrate (Pierce) in 96-well microplates (Optiplate; PerkinElmer Life Sciences), and chemiluminescence was quantified 30 s later with a luminometer (Viktor2; PerkinElmer Life Sciences).

RESULTS
Purified ␤ 2a Protein Binds to GST-AID Fusion Protein and Modulates Ca V 1.2 Channel Activity-The purified ␤ 2a elutes as a monodisperse peak from a size exclusion chromatography (Fig. 1A) and was capable of binding specifically to the Ca V 1.2-AID site because no binding activity was observed with GST alone or GST fused to an unrelated sequence (GST-ClC 126 ; Fig.  1B). As in cRNA coexpression experiments (3,14), the purified ␤ 2a protein increases ionic currents and shifts the voltage dependence of calcium channel activation toward more negative potentials. Maximal barium current (I Ba ) was increased from Ϫ18 Ϯ 4 nA (Ca V 1.2 alone) to Ϫ184 Ϯ 19 nA (Ca V 1.2/␤ 2a  Table 1. The thick line describes the sum of 0.8 times the GV obtained with Ca V 1.2/␤ 2a cRNA and 0.2 times the one obtained without the ␤-subunit (see Table 1 for details).
protein), which compares with Ϫ101 Ϯ 12 nA obtained from oocytes expressing Ca V 1.2/␤ 2a cRNA (Fig. 1C). The reason for this difference is not clear. Channel expression depends on several factors, such as translation efficiency of microinjected cRNA and protein trafficking capacity of the oocyte, which complicates interpretation of peak current data. Nevertheless, our result shows the functional competence of the ␤ 2a protein.
The fraction of channels activated during a 66-ms pulse to increasing potentials was measured as the peak of the tail current during repolarization to Ϫ40 mV. Normalized peak tail currents were plotted with respect to the pulse potential to yield the GV curves (Fig. 1D). The sum of two Boltzmann distributions was adjusted to each GV curve, and the parameters defining these distributions are described in Table 1. The only visible difference when comparing the effect of ␤ 2a as cRNA or as protein was that with the latter the relative contribution of the first Boltzmann distribution (G 1 ) is slightly smaller (42.0 Ϯ 2.5%) than with Ca V 1.2/␤ 2a cRNA (50.9 Ϯ 2.0%). This change impacts macroscopic conductance at positive voltages and may reflect channels not being modulated by the auxiliary subunit when injected as a protein. To illustrate this point we constructed a GV curve by adding 80% of a GV obtained with Ca V 1.2/␤ 2a cRNA and 20% of the one obtained without ␤-subunit (thick line in Fig. 1D). This new plot superimposes almost perfectly with the data obtained from oocytes injected with the Currents recorded in the absence of exogenous ␤-subunit showed a rapid outward transient corresponding to gating currents that was followed by a small noninactivating inward current mediated by the influx of Ba 2ϩ ( Fig. 2A, top panel ). In contrast, oocytes coinjected with Ca V 1.2/␤ 2a cRNA displayed inward ionic currents that were larger than gating currents ( Fig.  2A, middle panel). Likewise, injection of the ␤ 2a protein to Ca V 1.2-expressing oocytes also leads to an increase in ionic current amplitude relative to gating currents ( Fig. 2A, bottom  panel). Here we integrated the first 2 ms of the gating current evoked by a pulse from Ϫ80 to Ϫ30 mV (Q 2ms ; shaded area in Fig. 2A) and compared it with I Ba measured at the end of a pulse to 0 mV as described previously (15). Although in this type of measurements, I Ba /Q 2ms may be overestimated because inward currents are subtracted from outward gating currents, the impact is the same for ␤ 2a cRNA or ␤ 2a protein, because voltage dependence of activation is similar for both subunit combinations. Fig. 2B shows scatter plots of the Q 2ms versus I Ba for the three channel subunit combinations. Over a wide spectrum of Q 2ms amplitudes, I Ba amplitudes are near 0 when Ca V 1.2 was expressed alone, as reflected in an average I Ba /Q 2ms ratio of 0.3 Ϯ 0.1 nA/pC. When the ␤ 2a -subunit is present in either form, this ratio is severalfold larger ( (16). This treatment is expected to interrupt the incorporation of new Ca V 1.2 subunits into the plasma membrane and cause a net reduction of channel density caused by constitutive endocytosis that removes membrane proteins at a The mean ؎ S.E. of parameters defining the sum of two Boltzmann distributions that best fitted the normalized conductance GV values were obtained by measuring the peak of tail currents recorded during deactivation at Ϫ40 mV following depolarizing pulses of increasing voltages from a holding potential of Ϫ80 mV. The sum of two Boltzmann distributions was adjusted to tail currents amplitudes from individual experiments. %G 1 is the relative contribution of component developing at more negative potential. Each Boltzmann distribution is characterized by slope factors z 1 and z 2 and half activation potential V 1 and V 2 as used by Olcese et al. (24). I max corresponds to the average maximum tail current estimated from the fit. rather constant rate. To assess that bafilomycin was effectively preventing the incorporation of newly synthesized channels, surface expression was assayed concurrently by two independent methods: immunoassay and charge movement (Q on ) during a voltage step to I Ba reversal potential (Fig. 3A). Immunoreactivity and Q on were reduced by more than 70% by bafilomycin, and this reduction was not reverted upon injection of ␤ 2a protein. For protein-injected oocytes Q on was 164 Ϯ 26 pC (n ϭ 8), which compares with 144 Ϯ 21 pC (n ϭ 9) measured in the absence of ␤ 2a . Similarly, chemiluminescence measurements were virtually identical with (0.97 ϫ 10 5 Ϯ 0.24 ϫ 10 5 cps; n ϭ 16) or without (1.02 ϫ 10 5 Ϯ 0.31 ϫ 10 5 cps; n ϭ 20) ␤ 2a protein.

Ca
Thus, in fact the bafilomycin treatment under our conditions prevents channel incorporation into plasma membrane. If ␤ 2a could only interact with channels before they reach the plasma membrane, then injection of ␤ 2a protein should not modify the function of calcium channels in bafilomycin-treated oocytes. However, I Ba /Q 2ms increases from 0.6 Ϯ 0.2 nA/pC to 18.6 Ϯ 5.6 nA/pC by injection of ␤ 2a protein in bafilomycintreated oocytes (Fig. 3, B and C), indicating that ␤ 2a interacts with channels already present in the plasma membrane.
We next asked whether modulation of function of mature channels depends on an intact AID site and generated a point mutation in Ca V 1.2 (W470S) homologous to a mutation that impairs binding of the ␤-subunit in Ca V 2.1 (17). In Ca V 2.3, this mutation abolishes modulation of function as well as binding (18). Because expression may be impaired by alteration of the AID sequence, we documented the ability of HA-tagged Ca V 1.2 carrying the W470S mutation of reaching the plasma membrane by immunoassay and gating current measurements (Fig. 4A). Moreover, Q on measured at I Ba reversal potential correlates with chemiluminescence values over a wide range (Fig. 4B).
We also show that substituting the conserved tryptophan within the AID sequence in Ca V 1.2 abolishes binding of ␤ 2a protein to a peptide carrying this mutation (Fig. 4C, GST-AIDW470S). When injected into Ca V 1.2-W470S-expressing  oocytes, the ␤ 2a protein is no longer capable of increasing ionic currents (Fig. 4, D and E) nor of shifting the voltage dependence (Fig. 4F). For protein-injected oocytes, I Ba /Q 2ms was 1.05 Ϯ 0.07 nA/pC (n ϭ 11), which is virtually identical to 0.95 Ϯ 0.22 nA/pC (n ϭ 13) in oocytes without ␤ 2a , demonstrating that modulation of function by the ␤-subunit requires an intact AID site.
The W470S channels displayed a higher I Ba /Q 2ms ratio than wild type channels, suggesting an alteration on the ability of the channel to open. Changes in channel function associated with mutations on residues within the AID domain in the absence of ␤-subunit have been reported (19).
␤ 2a Protein Reverts Inactivation of Ca V 2.3/␤ 1b -mediated Currents-Here, to ascertain whether the ␣ 1 -␤ interaction is reversible, we tested the ability of ␤ 2a to inhibit inactivation in oocytes coexpressing Ca V 2.3/␤ 1b cRNA and compare it with its effects on oocytes expressing Ca V 2.3 alone. Typically, I Ba from oocytes expressing Ca V 2.3 cRNA by itself fully inactivates in a few hundred milliseconds (Fig. 5A). In contrast, when ␤ 2a cRNA is coinjected with Ca V 2.3 cRNA, I Ba inactivation develops over several seconds and reaches a plateau of about 25% of the peak current (20). Injection of ␤ 2a protein on Ca V 2.3-expressing oocytes also slows inactivation (Fig. 5, A and B). T1 ⁄ 2 increases from 0.26 Ϯ 0.03 s (n ϭ 16) for Ca V 2.3 alone to 4.0 Ϯ 0.63 s (n ϭ 12). This is even slower than T1 ⁄ 2 values obtained from oocytes injected with both subunits as cRNAs (2.4 Ϯ 0.42 s; n ϭ 13). More notably, the ␤ 2a protein was also capable of increasing T1 ⁄ 2 in oocytes coexpressing the Ca V 2.3/␤ 1b combination, from 0.15 Ϯ 0.01 s (n ϭ 18) to 1.5 Ϯ 0.3 s (n ϭ 23). The steady state inactivation (SS IN ) is also affected by ␤ 2a -subunit (20). While SS IN is complete for Ca V 2.3 alone or coexpressed with ␤ 1b , a residual component emerges (I RES ) with ␤ 22 . When we injected ␤ 2a protein in oocytes expressing Ca V 2.3 alone or the Ca V 2.3/␤ 1b combination, voltage for half-inactivation was shifted to the right, and a significant component of I RES developed in both cases (Fig. 5, C and D; see also Table 2 for details).
The inhibition of inactivation by ␤ 2a protein in Ca V 2.3/␤ 1b channel complexes may arise either by competitive inhibition or by allosteric modulation through a second binding site. To discriminate among both models, we studied the inhibition of inactivation by measuring T1 ⁄ 2 at different concentrations of ␤ 2a (Fig. 6). In Ca V 2.3/␤ 1b channel complexes, increase in T1 ⁄ 2 requires higher concentration of ␤ 2a than Ca V 2.3 alone (1.2 M versus 0.24 M to increase T1 ⁄ 2 near half-maximal, respectively). With higher ␤ 2a concentration (Ͼ3 M) channels inactivate similarly regardless of the presence of ␤ 1b (T1 ⁄ 2 ϭ 6.4 Ϯ 1.5 s (n ϭ 4) and 6.8 Ϯ 2.0 s (n ϭ 5) for Ca V 2.3 alone or coexpressed with ␤ 1b , respectively). These results suggest a simple competitive replacement of ␤ 1b by ␤ 2a .
To estimate the fraction of newly associated Ca V 2.3/␤ 2a channel complexes, we modeled the voltage dependence of steady state inactivation of Ca V 2.3 and Ca V 2.3/␤ 1b channels exposed to different concentrations of ␤ 2a protein (Fig. 7). Each subunit combination is expected to give rise to a SS IN of a particular shape. In oocytes expressing a mixture of subunit combinations, SS IN Table 2

E. of parameters defining the Boltzmann distribution and the percentage of residual current (I RES ) that best fitted steady state inactivation for the different subunit combinations
Peak current (I PEAK ) during the test pulse (Fig. 5, C and D) were plotted against the prepulse potential (V ) and adjusted to the following equation: I PEAK ϭ I RES ϩ (I MAX Ϫ I RES )/(1 ϩ exp͓(V1 ⁄ 2 Ϫ V )⅐(zF/RT)͔), where I RES corresponds to the residual noninactivating current expressed as a percentage of I MAX in the table.  (Fig. 7B). Note also that in oocytes lacking the ␤ 1b -subunit, this component of inactivation was not detected. Together these results confirm the notion of a competitive reaction between ␤ 1b and ␤ 2a .

DISCUSSION
The data presented here show that the ␣ 1 -␤-subunit interaction leading to modulation of channel activity can take place at the plasma membrane, is reversible, and depends on an intact AID site. The reversibility was shown by a novel approach that took advantage of the differential effect of two ␤-subunit isoforms, ␤ 1b and ␤ 2a , on the inactivation of Ca V 2.3 channels. Through a simple modeling of the voltage dependence of the steady state inactivation of the different subunit combinations, we estimated the K D in intact cells. Ca V 2.3/␤ 1b channels were converted to channels with a behavior typical of Ca V 2.3/␤ 2a channels upon the addition of ␤ 2a protein in a concentrationdependent manner that was consistent with the exchange of ␤-subunits governed by mass action law; i.e. the presence of ␤ 1b shifts the apparent K D without a change in the maximal effect when compared with Ca V 2.3 alone. Thus, our data are consistent with a competitive inhibition of inactivation of Ca V 2.3/␤ 1b channel complexes by ␤ 2a protein. Although our analysis does not conclusively exclude a second binding site, evidence for a single site come from experiments showing that a covalently linked ␤ 2b to Ca V 1.2 recapitulates the effect of the auxiliary subunit on expression and function and no longer responds to coexpression of the ␤-subunit (9). In our case, with both subunit combinations, dose-response curves yield Hill coefficients between one and two (Fig. 7). Several factors may increase the apparent Hill coefficient, such as not attaining full equilibrium in the 5 h that elapsed between protein injection and electrophysiological recordings. According to our experience, longer incubation time reduces the activity of the purified ␤-subunit protein (4).
The K D values we measured in intact cells are hundred-fold larger than reported for in vitro binding to GST-AID fusion proteins. Affinity for the ␤-subunit may be lower for channels in the cell surface than in early stage of protein synthesis and folding as initially proposed by Birnbuamer and co-workers (2). A K D value in the micromolar range is also consistent with a rapidly reversible binding of the ␤-subunit that stands in contrast to the traditional view of an invariant stoichiometric subunit. This reversibility is important to other intracellular regulatory pathways involving the AID site. For example, G protein binds to a site near the AID sequence, displaces the ␤-subunit, and inhibits N and P/Q type calcium channel (21,22).
An issue that remains rather puzzling is our observation that Ca V 1.2-W470S, a mutant with impaired ␤-subunit binding and functional modulation, expresses well in the plasma membrane, as if binding of the ␤-subunit to the AID site were not mandatory for surface expression. It may be that the lack of the conserved tryptophan weakens binding to the AID to the point that is no longer detectable in vitro and on functional modulation, but it is sufficient to allow the ␤-subunit to occlude the putative endoplasmic reticulum retention signal encoded in the I-II loop. In light of recent experiments (23) showing that Ca V 2.1 can also reach the plasma membrane without the AID sequence, retention appears not to be as tight as initially thought, and perhaps the ␤-subunit may help traffic through the endoplasmic reticulum by binding to other sites of the ␣ 1 -subunit. Alternatively, the Ca V 1.2 N-terminal truncation used in this study may help override the retention signal from the AID conferring independence of channel expression on the ␤-subunit. We have reported similar results for other ␤-subunit isoforms (4). This uncoupling would also explain why N-terminal deletion mutant overexpresses (10).
In summary, we demonstrate that the interaction between the ␤-subunit and the pore-forming subunit of voltage-gated calcium channels is dynamic. The dynamic nature of this interaction allows for post-targeting modulation of channel function. Moreover, this interaction requires the AID site, proposed to control targeting of the channel to the plasma membrane.