Orientation of the Calcium Channel β Relative to the α12.2 Subunit Is Critical for Its Regulation of Channel Activity

Background The Cavβ subunits of high voltage-activated Ca2+ channels control the trafficking and biophysical properties of the α1 subunit. The Cavβ-α1 interaction site has been mapped by crystallographic studies. Nevertheless, how this interaction leads to channel regulation has not been determined. One hypothesis is that βs regulate channel gating by modulating movements of IS6. A key requirement for this direct-coupling model is that the linker connecting IS6 to the α-interaction domain (AID) be a rigid structure. Methodology/Principal Findings The present study tests this hypothesis by altering the flexibility and orientation of this region in α12.2, then testing for Cavβ regulation using whole cell patch clamp electrophysiology. Flexibility was induced by replacement of the middle six amino acids of the IS6-AID linker with glycine (PG6). This mutation abolished β2a and β3 subunits ability to shift the voltage dependence of activation and inactivation, and the ability of β2a to produce non-inactivating currents. Orientation of Cavβ with respect to α12.2 was altered by deletion of 1, 2, or 3 amino acids from the IS6-AID linker (Bdel1, Bdel2, Bdel3, respectively). Again, the ability of Cavβ subunits to regulate these biophysical properties were totally abolished in the Bdel1 and Bdel3 mutants. Functional regulation by Cavβ subunits was rescued in the Bdel2 mutant, indicating that this part of the linker forms β-sheet. The orientation of β with respect to α was confirmed by the bimolecular fluorescence complementation assay. Conclusions/Significance These results show that the orientation of the Cavβ subunit relative to the α12.2 subunit is critical, and suggests additional points of contact between these subunits are required for Cavβ to regulate channel activity.


Introduction
Calcium influx via voltage-gated Ca 2+ channels (Ca v ) play vital roles in cell physiology, such as triggering muscle contraction and hormone secretion [1]. Both the amount of Ca 2+ that enters a cell, and where in the cell it enters, are highly regulated. To fulfill these specialized roles, Ca 2+ channels have evolved into multimeric complexes composed of an a 1 , a 2 d, and b, and each of these subunits has evolved such that there are ten a 1 genes, four a 2 d genes, and four b genes. Other mechanisms by which cells can fine tune Ca 2+ channel activity include: alternative splicing of these Ca v genes, regulation by calmodulin and G protein bc subunits, and phosphorylation by protein kinases. One of the first findings from studies with recombinant Ca v channels was the dominant role of Ca v b subunits [2][3][4]. Although the a1 subunit contains the channel pore, the voltage sensors, and most of the drug binding sites, the auxiliary subunits regulate all of these structures to increase channel opening, shift the voltage and time dependence of channel gating, and to increase drug affinity [5,6].
Ca v b subunits are known to bind with high affinity to the I-II loop of HVA a1 subunits [7]. This site has been termed the alphainteracting domain (AID), and is located 22 amino acids (a.a.) away from the C-terminus of the last transmembrane segment of repeat I (IS6). Recently three groups reported the crystal structure of Ca v b, either alone or in complex with a synthetic peptide corresponding to the AID [8][9][10]. These results confirmed the hypothesis that Ca v b subunits were part of the MAGUK protein family [11], and showed how the a-helical AID is embedded in the guanylate kinase (GK) domain of Ca v b. Despite such a clear picture of where it binds to a1, it is unclear how this translates into channel regulation. In fact, splice variants of Ca v b have been found that lack the GK domain, yet are still able to regulate the probability of channel opening, P o [12,13].
Previously we have shown that some aspects of Ca v b regulation could be conferred on a T-type channel a1 subunit (a 1 3.1) by transfer of the AID region from a 1 2.2 [14]. Similar to their regulation of HVA channels, Ca v b shifted the voltage dependence of activation to more hyperpolarized potentials, and increased the amount of current observed at the end of a sustained pulse. These studies provided the first evidence that b regulation required a rigid linker between IS6 and the AID, thereby providing support for the direct coupling hypothesis [15], which postulates that Ca v b alters movements of the IS6 segment that occur during gating. Notably missing from the a 1 3.1-2.2 chimera was Ca v b's regulation of channel P o and closed state inactivation, which has been observed with wild-type N-type channels [16,17]. Due to these limitations, we have now tested the direct coupling hypothesis by mutating a 1 2.2 directly. We show that deletion of a single amino acid in the IS6-AID linker is sufficient to abolish most aspects of Ca v b regulation (except trafficking to the plasma membrane). This result seemingly contradicts the direct-coupling hypothesis, and highlights the importance of b's orientation with respect to a 1 in allowing interaction with its gating machinery.

Results
The direct coupling model for Ca v b regulation predicts that the linker separating the AID from IS6 is a rigid a helix or b sheet. To test this hypothesis, we replaced six consecutive amino acids in the middle of the linker with either glycine (PG6) residues to introduce flexibility, or as a positive control for charge disruptions, with alanines (PA6) to conserve a rigid structure (Fig. 1). Previous circular dichroism studies using peptides designed against wild-type, PG6, and PA6, confirmed the PG6 mutation increased the random coil content from 38.5 to 51%, and the PA6 mutation decreased it to 28% [14]. As a second test of the direct coupling hypothesis, we deleted 1, 2, or 3 amino acids in the middle of the linker in order to alter the orientation of AID-bound b subunit with respect to a 1 . These mutations were introduced into a rat brain a 1 2.2, then studied in HEK-293 cells by whole cell patch clamp electrophysiology.
b regulation of wild-type a 1 2.2 Hallmarks of Ca v b regulation of high voltage-activated a 1 subunits include: their ability to increase the number of functional channels at the plasma membrane (due to both an increase in surface expression of a1 and an increased probability that these channels will open (P o ) in response to a test depolarization); and to shift the voltage dependence of channel activation [18]. This regulation was reliably detected under our experimental conditions (Fig. 2). Notably, Ca v b regulation of inactivation was isoform-specific. Various b2 splice variants, such as b2a, have the ability to dramatically slow inactivation of a 1 2.x currents [19,20], which can be quantitated by measuring the residual current after 350 ms of depolarization and normalizing it to the peak current amplitude (R 350 ; Fig. 2C,D). The ability of b2a to increase R 350 was relatively voltage independent, therefore the value at +20 mV is representative and is the value reported in the Tables. b3 did not have a statistically significant effect on the R 350 of WT channels (Fig. 2D, Table 1). Isoform-specific effects on the steady-state inactivation curve (h ' ) were also observed, with b2a shifting the mid-point (V 50 ) +6 mV, while b3 produced a large 230 mV shift in the V 50 (Fig. 2F). As noted previously, this b3 effect is due to acceleration of closed state inactivation [17]. Preliminary results with b1a and b4 were similar to those obtained with b3 (data not shown), therefore we selected b2a and b3 for further study. In addition, native N-type channels are formed by a 1 2.2 and either b2a or b3 [21,22]. These electrophysiological signatures provide assays of Ca v b regulation that can be used to test a 1 2.2 mutants for loss of regulation (Table 1). Specifically, these bs increase current density over 20-fold and shift the activation curve 210 mV, b2a increase the R 350 current 10-fold, and b3 shifts steady-state inactivation 234 mV (Table 1).
b regulation of poly-glycine and poly-alanine mutants Replacement of six amino acids in the IS6-AID linker with glycines (PG6) had a dramatic impact on channel expression and gating ( Fig. 3A-C). Current density was decreased 10-fold relative to wildtype (WT) channels and all other notable aspects of b regulation were lost in the PG6 mutant, yet, b2a was still capable of increasing the expression of functional channels ( Fig. 3A-D, Table 1). Notably b2a no longer modulated the voltage dependence of activation, b2a no longer increased R 350 , and b3 no longer modulated steady-state inactivation. In contrast to their equipotency at increasing WT currents, b2a increased PG6 currents ,20-fold, while b3 had no significant effect (Table 1). A second notable difference was the ability of b3 to induce ultra-rapid inactivation of open channels (Fig. 3C). This gain-of-function might be explained by b3 interacting with novel regions of the channel, while the decreased current density is likely due to channels inactivating before opening, a phenomenon that has been observed in Na + and T-type Ca 2+ channels [23,24]. In any case, the PG6 mutation largely disrupted normal Ca v b regulation as predicted by the direct coupling model.
The IS6-AID linker sequence is more highly conserved than the AID itself. For example, in the a 1 2.x family there is only 1 substitution in 20 residues of the linker (a leucine in a 1 2.1 and a 1 2.2 is substituted by methionine in a 1 2.3), but 3 substitutions in the 20 AID residues. This level of conservation, many of which are charged, suggests that both the structure and properties of amino acid side chains in the linker are important. Therefore, as a control for the PG6 mutation, we attempted to maintain a rigid structure by replacing the same 6 residues with alanine (PA6). As predicted, almost all the hallmarks of b regulation were observed with PA6 channels: bs increased peak currents .15-fold, shifted activation ,210 mV, b2a increased R 350 , and b3 shifted the h ' curve ( Fig. 3E-H, Table 1). Three aspects of b2a regulation were diminished by the PA6 mutation: one, its ability to increase current density was diminished 2-fold; two, it increased R 350 to a lesser extent (0.55 in WT vs 0.28 in PA6); and three, its ability to affect steady-state inactivation was lost. In contrast, b3 The residual current at 350 ms of depolarizing pulse was divided by the maximum inward current and plotted against the test potential. (E) Representative current traces obtained during a test pulse to +40 mV after 15 s prepulses to varying potentials from a holding potential of 290 mV. Traces recorded after prepulses to 260 and 220 mV are highlighted to emphasize the b3 induced shift in steady-state inactivation. Scale bar represents 20 ms and 200 pA. (F) Effects of b2a and b3 on steady-state inactivation. The mean normalized amplitude of the currents is expressed as a function of membrane potential and fit with a Boltzmann equation (smooth curves). Data represent mean6SEM, in which the number of cells used to calculate the average is reported in Table 1. doi:10.1371/journal.pone.0003560.g002 regulation of PA6 was similar to its regulation of WT channels in terms of current density and ability to shift the h ' curve. The results with PG6 and PA6 are entirely consistent with the IS6-AID linker being a ordered structure as observed in circular dichroism studies of isolated peptides [14].

b regulation of deletion mutants
Deletion mutants lacking 1, 2, or 3 amino acids (Bdel1, Bdel2, Bdel3, respectively) in the middle of the IS6-AID linker (see Fig. 1) were prepared using PCR-based mutagenesis. Expression of Bdel1 (with a 2 d1) led to the appearance of small barium currents whose current density was similar to WT. Coexpression of Bdel1 with either b2a or b3 led to the appearance of robust barium currents, and the stimulation over Bdel1 alone was 13-fold for b2a and 24fold for b3 ( Fig. 4A-D, Table 2). Other than their ability to increase functional channels, most other aspects of b regulation were lost in the Bdel1 mutant: there was no shift in the activation curve, b2a did not affect R 350 , and b3 had no effect on the h ' curve.
Little or no current could be detected from Bdel2 channels when expressed with only a 2 d1. Measurable currents were    detected in 3 of 17 cells, and these currents were too small (current density 20.660.2, n = 17) for reliable analysis. No currents could be detected with Bdel3 alone. In contrast, over 300 pA of Ba 2+ current could be easily measured when a b was cotransfected with these mutants (Fig. 4E). Other signs of b regulation were: one, that Bdel2+b2a currents inactivated 6-fold slower than with b3, but still much faster than WT currents; and two, that b3 modulated the closed-state inactivation of Bdel2 as observed with WT channels, producing a 220 mV shift in the h ' curve ( Fig. 4F-H. Similar gain-of-function effects observed with PG6 were also present, with b3 inducing rapid inactivation of currents, and stimulating current density to a lesser extent than b2a. In contrast, inactivation of Bdel3 was not regulated by either b2a or b3, as there was no effect on either open-or closed-state inactivation (Fig. 4J-L).

Estimation of surface expression and P o of Bdel1
The only typical form of b regulation retained in the Bdel1 mutant was the ability to increase current density. The whole cell current is proportional to the number of channels in the plasma membrane multiplied by their P o (assuming no change in single channel conductance, but see [25]). We hypothesized that trafficking of Bdel1 to the cell surface would be same as for WT [26], since the deletion did not affect the ability of b3 to increase current density. To measure surface expression of Bdel1 and WT channels, we labeled each a 1 subunit at the N-terminus with GFP, expressed them in HEK-293 cells with a 2 d1 and b2a, then used confocal microscopy to quantitate GFP signal at the plasma membrane as described previously [27]. Similar amounts of GFP signal were detected at the plasma membrane with Bdel1 (7867 au/pixels, n = 12) as WT (95616, n = 8, P = 0.3), and the percent of the GFP signal at the surface was also similar (Bdel1, 4161; WT, 3762).
To estimate the effect of the Bdel1 mutation on channel P o , we used the method of Yue and coworkers that relies on the ratio of ionic to gating currents [28]. In this method gating currents are measured at the reversal potential and integrated to estimate Q max , the whole cell current is normalized to driving force to yield the maximal conductance (G max ), and the P o is estimated by G max /Q max . Using the same voltage protocols as Alger et al., (2005) we were able to reliably measure gating currents in cells transfected with b2a, a 2 d1, and either WT or Bdel1 (Fig. 5).
Representative traces from the same cells clearly show that WT channels generate large currents from a small number of channels, while Bdel1 generates smaller currents from a larger number of channels. The slope of the line correlating G max to Q max was 19fold lower for Bdel1 (0.07460.004, n = 7) than for WT channels (1.3260.13, n = 8, P,0.01). This results shows the P o of Bdel1 channels is extremely low, explaining why current density was so low. As noted in the Methods, the transfection protocol was different between WT and Bdel1, thereby precluding direct comparison of G max and Q max in this assay. We conclude that deletion of a single amino acid in the IS6-AID linker abolishes almost all b regulation of the biophysical properties of a 1 2.2, including its ability to increase P o , leaving only the b regulation of trafficking.

Bimolecular Fluorescence Complementation
To address the question of whether the deletions in the IS6-AID linker altered the orientation of b to a 1 2.2 more directly, we utilized bimolecular fluorescence complementation (BiFC) analysis [29]. In this method a fluorescent protein such as CFP is split into two fragments, and then fused to proteins of interest. If the proteins of interest interact, and the two CFP fragments are brought into the proper orientation, then they will reassemble and restore fluorescence. In the present study we relied on the high affinity binding of b to the AID region on a 1 2.2 [30], and fused the fragments of the cyan fluorescent protein (CFP) to the N-terminus of a 1 2.2 and either the N-or C-termini of b3. Preliminary experiments with a full-length b3 tested which combination of tagged proteins could restore the proper orientation (see Methods for all combinations tested), and found the strongest BiFC signal when the big N-terminal fragment of CFP (a.a. 1-158) was fused to the N-terminus of a 1 2.2, and the small C-terminal fragment of CFP was fused to the C-terminus of b3. To restrict the movement of the C-terminal CFP fragment, we truncated the C-terminus of b3 to the same length (b3-core) used in the crystallographic studies [9]. HEK-293 cells were also cotransfected with the RFP, mCherry, which allowed for both selection of transfected cells and calculation of the cyan BiFC to red ratio. As described previously [31], the specificity of the BiFC signal can be calculated from the median ratio of the cyan/red signals (Fig. 6). In our experiments the strongest BiFC signal was observed with tagged Bdel1+b3core constructs, being 3.4-fold higher than WT (Fig. 6). In contrast, the tagged Bdel2+b3core combination produced a lower BiFC signal than Bdel1+b3core, but still higher than WT, a difference that was statistically significant. These results are consistent with the electrophysiology results, where both WT and Bdel2 channels are regulated by b subunits, while Bdel1 is not. We affirm that these results strongly support our hypothesis that the Bdel mutations alter the orientation of the b subunit with respect to the a 1 subunit.

Discussion
Since the seminal experiments of Ringer on cardiac muscle contraction [32], it has been recognized that Ca 2+ entry into cells provides a crucial trigger for life and death processes. Key pathways for Ca 2+ entry are voltage-gated Ca 2+ channels, and to fulfill specialized roles these channels have evolved into ten a 1 subunit genes that are extensively spliced to generate unique channels. Biochemical purification of high voltage-activated channels revealed their multi-subunit structure, being composed of a 1 , a 2 d, and b subunits [33,34]. Studies with the cloned subunits clearly established the critical roles of the a 2 d and b subunits in trafficking and regulating the biophysical properties of the a 1 subunit. For these reasons, the mechanism of action of these subunits has been extensively investigated. Arguably the greatest progress has been made in understanding the roles and mechanism of action of b subunits [18,35,36]. Campbell and coworkers provided a major breakthrough by mapping the site on a 1 that binds b, termed the AID [7]. The precise details of this interaction were elucidated by X-ray crystallographic studies of AID peptides bound to b2, b3, and b4 [8][9][10]. The AID anchors the b subunit 22 amino acids away from the end of IS6, a distance that is invariant in all HVA a 1 subunits. S6 segments of Ca 2+ channels are thought to form an inner gate that opens during channel gating [37], as so clearly observed in K + channel crystal structures [38]. S6 segments are also involved in drug binding, and this binding can be regulated by b subunits [15,39]. Taken together, these observations led to the direct coupling hypothesis, whereby b modulates a 1 gating by direct modulation of IS6 movement [8,15].
In sharp contrast to HVA channels, expression of cloned LVA Ca v 3 a 1 subunits produces robust currents with properties that are nearly identical to native T-type currents [40]. In a previous study we exploited this difference to make chimeras that would test the direct coupling hypothesis, moving the I-II loop of a 1 2.2 into a 1 3.1 [14]. Four key results from this study were: one, that some aspects of b regulation could be conferred (b induced a shift in activation curve and slowed open channel inactivation); two, that this regulation was completely dependent on the IS6-AID linker; 3) circular dichroism of peptides corresponding to WT, PA6, and PG6 confirmed the ability of these mutations to disrupt structure; and 4) that this linker was likely to form a rigid structure [14]. Two other interesting conclusions from this previous study were: 1) that the distal portion of the S6 segment is part of an inner gating ring that controls channel inactivation; and 2) that the post-IS6 linker acts as a gating brake in Ca v 3 channels. The structure of this gating brake is likely an antiparellel helix-loop-helix [41], and is conserved in all three Ca v 3 channels [42]. A limitation to this previous study was that bs had no effect on P o , and b isoform specific regulation of inactivation from open and closed states were not conferred to the chimera. Therefore, in the present study we have explored the importance of the IS6-AID structure by mutating this region in a 1 2.2.
We show that mutations that induce flexibility and destabilize ahelices and b-sheets, as with PG6, totally abolish b subunit regulation. Specifically, b subunits no longer shift the activation curve, b3 no longer shifts the h ' curve, and b2a no longer increases plateau currents. As a control, we show that replacement of these same amino acids with alanines largely retain b regulation of gating. Interestingly, b3 accelerated inactivation kinetics of PA6, and b2a's ability to affect inactivation was altered. Two possible explanations for the loss of b regulation of PA6 are: one, particular amino acid residues may play a role (e.g. charged residues stabilize interactions with other channel regions) or two, the structure of PA6 is not identical to WT.
As a second test of the direct coupling hypothesis, we deleted 1, 2, or 3 amino acids from the middle of the IS6-AID linker. The key results were loss of b regulation in Bdel1 and Bdel3, and retention of regulation in Bdel2. A limitation to these studies is that little or no currents could be recorded from Bdel2 or Bdel3 when expressed alone (+a 2 d1), so the presence or absence of b regulation could only be inferred by comparisons between b2a and b3. Nevertheless, the ability of bs to increase functional channels at the plasma membrane was retained in all 3 Bdel mutants. We conclude that the central region of the IS6-AID adopts a b-sheet structure. Deletion of 1 or 3 amino acids in this b-sheet would alter the orientation of Ca v b with respect to a1 by 180u, while deletion of two would return the orientation back to WT (Fig. 7). Although b subunits were crystallized with peptides corresponding to the AID sequence, these peptides did not include the 20 amino acids of the IS6-AID linker, consequently the exact structure of this region was not determined. Opatowsky et al., (2004) proposed an interesting hypothesis whereby b binds to a disordered AID and induces its folding into an a helix that extends all the way to IS6 [10]. Secondary structure prediction programs [43] suggest that this linker has a high tendency to form an a helix, and near the middle of the linker an equal tendency to form a b sheet. Deletions in this region may preferentially stabilize the b-sheet structure. Regardless of the structure, the orientation of b subunit is critical for its ability to regulate the biophysical properties of a 1 2.2 channels. We hypothesize that b is precisely anchored at the AID to orient other parts of b towards regions of a 1 that control channel opening (thereby modulating P o and the activation curve) and channel inactivation.
The palmitic moieties of b2a may alter this orientation, thereby altering its interaction with inactivation gates [19]. This difference in orientation is supported by the finding that the ability of b2a and b3 to increase current density and their effect on inactivation kinetics varied between the mutants.
Mutations in the AID that weaken its interaction with b subunits may allow a new orientation, which explains why the W391A mutation in a 1 2.2 abolishes b1 but not b2a regulation [30]. Alternative splicing of either b subunits [44][45][46] or a 1 subunits may provide additional isoform specific regulation [47]. The crystal structures of bs provided clues for additional binding sites: one, the SH3 domain, which is a common structural motif at protein interaction sites, and two, the large groove between the GK and SH3 domain [8][9][10]. Notably, a rigid IS6-AID linker would orient this groove directly below the cytoplasmic face of a 1 . The other interaction sites on a 1 are likely to be of much lower affinity, allowing regulatory proteins such as G protein bc to disrupt these interactions. In this scenario, Gbc could shift channels into reluctant states without complete dissociation of Ca v b from the AID, which reconciles conflicting observations [6,8,48]. Mutational studies support the hypothesis that all voltage-gated channels have a similar inverted teepee structure where the intracellular gate is formed by a bundle crossing of S6 segments as observed in the crystal structure of Shaker K + channels [15,37,38]. The precise positioning of Ca v b near this bundle crossing would allow it to interact with any of the post-S6 segments. Inappropriate interactions appear to severely accelerate channel inactivation, as observed with the needle-like kinetics of PG6 and Bdel2 induced by b3, which may be due to novel interactions with other post-S6 segments. Accordingly, De Waard and colleagues have provided evidence that post-IIIS6, as well as the carboxyl terminus, play important roles in Ca v b regulation of a 1 2.1 [49,50]. An alternative version of the direct coupling hypothesis is that the other a 1 -b interactions serve mainly as a fulcrum, allowing b to restrain movement of IS6. The gating current studies show that deletion of one amino acid from IS6-AID linker largely disrupts bs ability to increase P o , therefore, we conclude that orientation of b is critical for its regulation of the biophysical properties of a 1 2.2, and confirm the importance of additional points of contact between the two subunits.

Site-directed mutagenesis
The starting material (wild-type, WT) for these studies was the cDNA encoding the ''a'' isoform of rat brain a 1 2.2 cDNA (GenBank entry #AF055477) cloned into the plasmid vector The b3 core structure was modeled from PDB code 1VYT [9]. The CFP, Cirulean, was modeled from PDB code 2QYT [57]. The fragments of CFP were generated using PyMOLWin (Delano Scientific), where CFP-N corresponds to residues 1-158, and CFP-C corresponds to residues 159-238. The approximate size of the a 1 2.2 domains and linkers were estimated using the method of Helton and Horne, where the volume occupied by each segment is calculated from the number of amino acids in each segment [45]. The b3 subunit was scaled using the same method. doi:10.1371/journal.pone.0003560.g007 pcDNA6 [51]. A 1.5 kb fragment was subcloned into pCR2.1-TOPO (Invitrogen, Carlsbad, CA), then mutated using the QuikchangeH protocol and Pfu Ultra DNA polymerase (Stratagene, La Jolla, CA). Oligonucleotide primers were obtained from Invitrogen and used without purification. All restriction enzymes were purchased from New England Biolabs (Ipswich, MA). The full-length cDNA was reassembled in the original plasmid vector that was cut with AscI and BsiWI by ligating the following fragments: AscI(32)/BlpI(355), BlpI/SacI (1407), and SacI/BsiWI (2991). The mutations were contained in the BlpI/SacI fragment, and the sequence of this fragment was verified for each mutant by automated sequencing at the University of Virginia Biomolecular Research Facility.
Transfection of a 1 2.2, and any b under these conditions led to the appearance of robust currents that could be reliably clamped. In contrast, transfection of a1 alone produced little or no currents. Addition of a 2 d1 increased the amplitude of the currents 2-to 5fold, and increased the percentage of cells with current. With some of the mutants (e.g. Bdel2, Bdel3), there was no detectable current under these conditions. To boost expression, transfections were modified as follows: 1) increasing the amount of mutant a 1 2.2 and a 2 d1 plasmids to 3 mg, 2) adding 0.5 mg of plasmid containing the SV40 T antigen, and 3) incubating the cells 48 hours before plating onto chips. Under these conditions currents from all the mutants could be reliably measured, however, wild-type currents were so large that in many cases they saturated the amplifier (.20 nA). As a consequence, the ability of b subunits to stimulate expression of functional channels is underestimated. Similar experimental conditions have been used in previous studies to record recombinant Ca v 2.2 currents [28,54].

Electrophysiology
Electrophysiological experiments were carried out using the whole cell configuration of the patch clamp technique. Recordings were obtained using an Axopatch 200A amplifier equipped with a CV201A headstage. The amplifier was connected to a computer (Dell, Round Rock, TX) through a Digidata 1200 A/D converter, and controlled using pCLAMP 9.2 software (Axon Instruments, Union City, CA). Currents were recorded using the following external solution (in mM): 10 BaCl 2 , 156 tetraethyl ammonium (TEA) chloride, and 10 HEPES, pH adjusted to 7.4 with TEA-OH. The internal pipette solution contained the following (in mM): 125 CsCl, 10 BAPTA, 2 CaCl 2 , 1 MgCl 2 , 4 Mg-ATP, 0.3 Li-GDPbS, and 10 HEPES, pH adjusted to 7.2 with CsOH.
Pipettes were made from TW-150-3 capillary tubing (World Precision Instruments, Inc., Sarasota, FL). Initial pipette resistance was typically 2-3 MV. Access resistance and cell capacitance were measured using on-line exponential fits to a capacitance transient (Membrane Test, Clampex). Cell capacitance ranged between 8-30 pF. Access resistance averaged 4.2 MV. Data from cells where the access resistance exceeded 5.5 MV were discarded. Series resistance was compensated between protocols to 70% (prediction and correction; 10 ms lag), resulting in maximal residual voltage error below 1.6 mV during measurement of the current-voltage relationship. Data were collected at room temperature.
To balance the effects of inactivation, slow recovery, and rundown, we selected 350 ms pulses for the current-voltage protocol, and an inter-sweep interval of 20 s. Peak currents at each voltage step were used to calculate the voltage dependence of activation (V 0.5 , and k), and the conductance (G) as described previously [27]. The current at the end of the depolarizing pulse was also measured, and divided by the peak current in that pulse to derive the R 350 value. The voltage dependence of steady-state inactivation was estimated using 15 s prepulses to varying potentials followed by a test pulse to +40 mV to measure channel availability (h). The current elicited during each test pulse was normalized to that observed when the holding potential was 290 mV (I/Imax), and the data from each cell were fit with a Boltzmann equation using PrismH software (version 5, Graphpad Software, San Diego, CA). Results are presented as mean6SEM. Significant differences in the average data were analyzed using one-way ANOVA followed by Bonferroni's multiple comparison test (GraphPad Prism).

Confocal microscopy
Images of live cells were collected using a Cooke Sensicam QE CCD camera (Romulus, MI) mounted on an Olympus BX61WI microscope equipped with an Olympus confocal spinning disk unit (Melville, NY). Channel localization was visualized by measuring the green fluorescent signal from GFP fused to the N-termini of either WT or Bdel1. Data were acquired under identical conditions, and then analyzed using IPLab 4.0 (Scanalytics, Fairfax, VA) as described previously [27]. Plasma membranes were labeled with FM 4-64 (Invitrogen) following the supplier's recommendations. Live cells were treated for at least 5 min at 4uC, and then imaged at 4uC. The FM-4-64 signal was used to localize the plasma membrane, and the amount of green fluorescent signal (arbitrary units, au) was measured and normalized to the number of pixels in this area. For both channels the GFP signal was evenly distributed in the plasma membrane.

Bimolecular Fluorescence Complementation (BiFC)
In preliminary experiments we fused either the CFP N-terminal fragment (a.a. 1-158, abbreviated B) or C-terminal fragment (a.a. 159-238, abbreviated S) to the amino terminus of WT a 1 2.2 (a1B-B and a1B-S); and fused both CFP fragments to either the N-or C-termini of rat b3 (SNB3, BNB3, SCB3, BCB3). The CFP fragments and b3 coding region were PCR amplified with primers that added unique restriction sites in the correct reading frame. The CFP fragments were ligated to full-length a 1 2.2 cDNA using KpnI (polylinker) and AscI (263), thereby creating a flexible 21 a.a. linker (5 glycines) from the 59 untranslated region. The CFP-b3 linker included an SbfI site that encoded PAGAT, while the b3-CFP linker included a BspEI site that encoded SGAT. Of the four combinations that could produce a BiFC signal, the largest signal was detected with a1B-B+SCB3. The crystal structure of b3 was obtained with a construct called b3core (B3c). In agreement with previous reports [9,55], we found that B3c remains functional despite having the non-conserved amino and carboxy termini removed (results not shown). Since the 121 a.a. of the C-terminus might allow flexibility that would obscure the orientation of b, we prepared a C-terminal CFP (a.a. 159-238; SCB3c) fused to the b3core. The Bdel1-B and Bdel2-B were prepared using the same KpnI/AscI cloning strategy as for WT. These constructs (88 ng each) were transiently transfected into HEK-293 cells along with a 2 d1 and the red fluorescent protein (RFP), mCherry [56]. Similar results were obtained using the RFP, DsRed-Monomer (Clontech), so the results were pooled. After 18 hrs the cells were plated onto polylysine-treated glass bottom dishes (Fluorodish, World Precision Instruments, Saratoga, FL). Transfected cells were identified by their red fluorescence, and their red and cyan fluorescence signals were collected with IPLab software and the Olympus microscope described above (406objective, 262 binning, confocal off). Data were background subtracted using a region devoid of cells, and the ratio of green to red signal for each cell was calculated. Following the method of Shyu et al. [31], BiFC specificity was determined using the median of the CFP/RFP signal for each condition. Statistical analysis was performed using one-way ANOVA with the Kruskal-Wallis post test using GraphPad Prism.