Mutations of Nonconserved Residues within the Calcium Channel α₁-interaction Domain Inhibit β-Subunit Potentiation

Influx of calcium from the extracellular medium is mainly mediated by voltage-dependent calcium channels, which are classified according to their threshold into high or low voltage of activation (Ertel et al., 2000). Channels from the high voltage of activation family are composed of at least four nonhomologous subunits: the main pore-forming subunit, Ca_Vα₁, and three auxiliary subunits, Ca_Vα₂/δ, Ca_Vβ, and Ca_Vγ. Ca_Vα₁ encodes all the structural elements of a functional voltage-activated calcium channel (Catterall, 2000). Among the auxiliary subunits, Ca_Vβ stands out as the most potent regulator of channel function and expression (Hidalgo and Neely, 2007; Dolphin, 2003). Four Ca_Vβ isoforms (Ca_Vβ₁ to Ca_Vβ₄) have been cloned from four nonallelic genes, each encoding multiple splice variants. Although all Ca_Vβ isoforms potentiate currents mediated by any of the Ca_Vα₁-encoding high voltage of activation channels, they do so to a different extent (Dzhura and Neely, 2003; Luvisetto et al., 2004). This current potentiation manifests itself in an increase in the ionic-current to charge-movement ratio (I/Q) (Neely et al., 1993; Olcese et al., 1996) and in the channel's open probability (Po) (Wakamori et al., 1993; Neely et al., 1995; Costantin et al., 1998; Wakamori et al., 1999; Dzhura and Neely, 2003). Ca_Vβ may also increase the number of channels present in the plasma membrane by releasing them from the endoplasmic reticulum (Bichet et al., 2000). All Ca_Vβ isoforms are also capable of modulating inactivation, but only Ca_Vβ_2a inhibits voltage-dependent inactivation whereas Ca_Vβ_1b, Ca_Vβ₃, and Ca_Vβ₄ promotes it (Olcese et al., 1994; Qin et al., 1996; Sokolov et al., 2000; Restituito et al., 2000; Hering et al., 2000). A subset of Ca_Vβ isoforms participate as well in the regulation of calcium channel by G-proteins (Sandoz et al., 2004a) or protein phosphorylation (Arikkath and Campbell, 2003). This diversity in modulatory capabilities stands in contrast with the existence of a single well-defined binding site shared by all Ca_Vα₁ and Ca_Vβ subunits. In Ca_Vα₁, this interaction involves a highly conserved consensus sequence, the α₁-subunit interaction domain (AID), which lies within the cytoplasmic loop joining the first and second repeat of Ca_Vα₁ (Pragnell et al., 1994; Fig. 1 A). According to the crystal structure of the AID–Ca_Vβ complex, AID adopts an α-helical structure with the fully conserved tyrosine and tryptophan residues lying on one face of the helix, and becoming buried in a hydrophobic pocket within Ca_Vβ, the so called α-binding pocket (Opatowsky et al., 2004; Van Petegem et al., 2004; Chen et al., 2004). Mutations in the AID sequence change inactivation of Ca_Vα₁ (Leroy et al., 2005; Berrou et al., 2001, 2002; Geib et al., 2002; Dafi et al., 2004), suggesting that this region interacts directly with a yet to be identified component of the channel gating machinery. Residues lying on the opposite face of the AID–Ca_Vβ interaction surface (AID-exposed residues, Fig. 1 A) would be available for this interaction. Interestingly, these residues are poorly conserved among Ca_Vα₁ isoforms and may be partly responsible for isoform-specific variations. Because Ca_Vβ's effect on channel activation is independent of its impact on inactivation (Olcese et al., 1994), we investigated the role of the AID motif in Ca_Vβ-dependent potentiation by either deleting it or replacing conserved and nonconserved amino acids within it. We chose Ca_V1.2 for it shows the largest shift in voltage activation by Ca_Vβ (Olcese et al., 1996). In the absence of Ca_Vβ, mutations of AID residues (E462R, K465N, D469S, and Q473K) did not alter the voltage dependence of activation, I/Q, or single channel kinetics. In contrast, Ca_Vβ-induced increases in I/Q and Po were severely reduced in Ca_V1.2 channels bearing mutations in AID-exposed residues while the apparent binding affinity for Ca_Vβ remained unchanged. These mutant channels visit high Po mode less frequently than do wild-type (WT) channels without causing obvious changes in the kinetic within each gating mode. Deletion of the complete AID sequence (Ca_V1.2 ΔAID) shifts the voltage dependence of activation by +12 mV, suggesting that AID residues by themselves partially display the ability to modulate activation. Collectively, these findings indicate that in Ca_V1.2, AID-exposed residues interact with the gating machinery controlling modal gating once they are properly oriented upon Ca_Vβ binding.

Amino acid substitutions were performed by overlapping PCR using two complementary oligonucleotides bearing the appropriate point mutation and two flanking oligonucleotides to amplify a 388-bp segment flanked by BamHI and SpeI restriction sites. The PCR product was digested with BamHI and SpeI and subcloned into pAGA-2 vector carrying the coding sequence for Ca_V1.2 (GenBank accession no.: X15539). The SpeI silent site was incorporated by standard PCR methods in the Ca_V1.2 sequence at position 497. Constructs were selected by restriction pattern and confirmed by automated DNA sequencing. Ca_V1.2 ΔAID was constructed by removing residues 459 to 475 (459-QLEEDLKGYLDWITQAE-475). In all experiments, a variant of Ca_V1.2 bearing a deletion of 60 amino acids at the N terminus that increases expression was used (Wei et al., 1996).

Capped cRNAs were synthesized from Hind III-linearized templates using the MESSAGE machine (Ambion), as described previously (Hidalgo et al., 2006). Proteins and cRNAs were injected into Xenopus oocytes using a nanoliter injector (Nanoliter 2000; World Precision Instruments). Ca_Vβ_2a protein was purified and handled as described by Hidalgo et al. (2006). Electrophysiological recordings were performed as in Dzhura and Neely (2003). Macroscopic currents were recorded using the cut-open oocyte voltage-clamp technique (Taglialatela et al., 1992) with a CA-1B amplifier (Dagan Corporation) 4–5 d after cRNA injection and 2–5 h after protein injection, as described in Gonzalez-Gutierrez et al. (2007). The external solution contained 10 mM Ba²⁺, 96 mM n-methylglucamine, and 10 mM HEPES, and was adjusted to pH 7.0 with methanesulfonic acid. The internal solution contained 120 mM n-methylglucamine, 10 mM EGTA, and 10 mM HEPES, pH 7.0, adjusted with methanesulfonic acid. Pipettes were filled with 2 M tetramethylammonium-methanesulfonate, 50 mM NaCl, and 10 mM EGTA and showed a resistance from 0.5 to 1.2 MΩ. Current recordings were filtered at 10 kHz, and the linear components were subtracted by a P/−4 prepulse protocol. For patch-clamp recordings of single channel activity, we used an Axopatch-200B (MDS Analytical Technologies). Patch pipettes were pulled from aluminum silicate capillary (Sutter Instrument) coated with Silgard 184 (Dow Corning Corporation) and filled with a solution containing 76 mM Ba²⁺, 10 mM HEPES, and 100 nM S(-)Bay K 8644 (Sigma-Aldrich), adjusted to pH 7.0 with methanesulfonic acid. Pipette resistance ranged from 3 to 8 MΩ. Oocytes were placed in the recording chamber containing 110 mM K⁺ and 10 mM HEPES titrated to pH 7.0 with methanesulfonic acid. Calcium channels were activated by 200-ms pulses to 0 mV at 1 Hz from holding potential of −70 mV, sampled at 20 kHz, and filtered at 2 kHz. Data acquisition and analysis was performed using the pCLAMP system and software (MDS Analytical Technologies). All data are expressed as the mean ± SEM.

The presence of simultaneous openings during consecutive traces obtained after depolarization to +20 mV was the first criteria used to exclude patches containing multiple channels. From this set, the ones with no coincident openings during the next 1,000 traces using pulses to 0 mV were included in the analysis. To set a higher limit to the probability of not observing overlapping openings in a 1,000 consecutive trace, double channel patches were simulated over 5,000 traces. Channel activity was simulated as in Dzhura and Neely (2003), adjusting kinetic parameters to yield an overall Po < 0.03, as observed with E462R and K465N mutants. About 1% of the simulated traces displayed double openings. Thus, in a 1,000 traces run, the likelihood of not observing multilevel events in a two-channel patch is (1–0.01)^1,000 = 4.3 × 10⁻⁵. Based on these simulations, we assumed that recordings from Ca_V1.2 E462R and Ca_V1.2 K465N channels coexpressed with Ca_Vβ_2a did not include activity from multichannel patches. We make no assumption about the number of channels recorded using channels lacking Ca_Vβ; thus, in those cases we limited the analysis to burst-duration histograms.

The online supplemental material includes four tables and six figures. Table S1 summarizes maximal I/Q values. Table S2 describes the parameters defining the sum of two Boltzmann distributions that best fits the normalized conductance curve (GV) for mutant and WT channels in the absence or presence of Ca_Vβ_2a, Ca_Vβ_1b, or Ca_Vβ_3xo. Table S3 includes parameters defining the exponential distribution that best fits shut, open, and burst-duration histograms for mutants and WT channels with Ca_Vβ. Table S4 contains the parameters defining the exponential that best fits burst histograms from individual patches of WT and E462R and K465N mutants of Ca_V1.2 without Ca_Vβ_2a. Fig. S1 compares ionic currents and I/Q versus voltage plot of Ca_V1.2 WT, Ca_V1.2 W470S, and Ca_V1.2 ΔAID in the presence or absence of Ca_Vβ_2a. Fig. S2 compares I/Q versus voltage plots for CaV1.2 WT, Ca_V1.2 E462R, Ca_V1.2 K465N with Ca_Vβ_1b, or Ca_Vβ_3xo. Fig. S3 shows single channel current versus voltage plot for Ca_V1.2 WT, Ca_V1.2 E462R, Ca_V1.2 D469S, and Ca_V1.2 Q473K coexpressed with Ca_Vβ_2a. Examples of single channel activity from Ca_V1.2 D469S and Ca_V1.2 Q473K channels in the presence of Ca_Vβ_2a are shown in Fig. S4. Dwell-time and burst-duration histograms are plotted in Figs. S5 and S6, respectively.

Mutating the conserved tryptophan residue within the AID (Ca_V1.2 W470S described in Hidalgo et al., 2006) or deleting the AID sequence (as in Gonzalez-Gutierrez et al. [2007]) of Ca_V1.2 does not affect channel expression but abolishes Ca_Vβ-induced increase in I/Q (Fig. S1 and Table S1). Fig. 2 B shows that there is a near perfect overlap in the activation curves of Ca_V1.2 W470S with Ca_V1.2 WT, independently of whether the mutant channel was expressed alone or together with Ca_Vβ_2a. Here, we report that in addition, the voltage dependence of activation in Ca_V1.2 ΔAID channels appears shifted 12 mV toward more positive potentials (Fig. 2 B and Table S2). Half-voltages (V_1/2) for both Boltzmann components describing the conductance versus voltage plot were shifted to the right by ∼10 mV, and this difference was statistically significant (P < 0.01; t test). This suggests that residues within the AID sequence participate in Ca_Vβ-mediated regulation of the coupling between voltage sensor and the channel gate.

To further investigate the role of AID-exposed residues in channel modulation, we replaced them by the corresponding amino acids present in other Ca_Vα₁ subunits (Fig. 1) to yield Ca_V1.2 E462R, Ca_V1.2 K465N, Ca_V1.2 D469S, and Ca_V1.2 Q473K. We also tested the effect of a charge conservative mutation by replacing glutamate at position 462 by aspartate (Ca_V1.2 E462D). Fig. 3 A shows voltage-clamp recordings from Ca_V1.2 E462D, Ca_V1.2 E462R, and Ca_V1.2 K465N. In the absence of Ca_Vβ_2a, all mutants yielded I/Q versus voltage curves (Fig. 3 B) and peak values (Table S1) that were indistinguishable from Ca_V1.2 WT. Except for E462D, there were no apparent differences in the kinetic or voltage dependence of macroscopic currents recorded from oocytes expressing these variants in the presence of Ca_Vβ_2a. However, closer inspection revealed that inward ionic currents appeared reduced relative to the transient outward current (gating currents) for both mutants carrying a charge modification. Peak I/Q, when Ca_Vβ_2a was coexpressed, for Ca_V1.2 E462R and Ca_V1.2 K465N were 23 and 31% of Ca_V1.2 WT, respectively, whereas Ca_V1.2 E462D in the presence or absence of Ca_Vβ was nearly identical to Ca_V1.2 WT (Fig. 3, B and C, and Table S1). Thus, Ca_Vβ_2a increases peak I/Q of Ca_V1.2 WT over 20-fold compared with sixfold and threefold for Ca_V1.2 E462R and Ca_V1.2 K465N channels, respectively. The other two mutations examined (D469S and Q473K), when coexpressed with Ca_Vβ_2a, also yield a peak I/Q smaller than Ca_V1.2 WT (Table S1).

To test whether these mutations hindered potentiation by other Ca_Vβ isoforms, we coexpressed the isoform endogenous to Xenopus oocytes, Ca_Vβ_3xo (Tareilus et al., 1997), and a rat neuronal isoform (Ca_Vβ_1b). Ca_Vβ_3xo increased peak I/Q of WT Ca_V1.2 channels by ∼12-fold, whereas Ca_Vβ_1b did so by 10-fold. More importantly, Ca_Vβ_3xo and Ca_Vβ_1b potentiated Ca_V1.2 E462R and K465N channels to a lesser extend than did WT channels (Table S1 and Fig. S2). These reductions in I/Q were not accompanied by changes in the voltage dependence of activation nor in the ability of Ca_Vβ_2a to shift the GV curves toward negative potentials (Fig. 4, A and B, and Table S2). However, the maximal conductance (G_max) normalized by charge movement (Q_on) was greatly reduced for E462R and K465N mutants in the presence of Ca_Vβ_2a (Fig. 4 C).

We examined four factors that may account for this reduction in I/Q and G_max: (1) a shift in the equilibrium toward inactivated states, (2) a decrease in the fraction of channels bound to Ca_Vβ, (3) a decrease in unitary conductance, and (4) a reduction in the channel probability of being open.

The participation of the I-II loop of Ca_Vα₁ in the regulation of voltage-dependent inactivation is well documented (Stotz et al., 2004a,b; Sandoz et al., 2004b; Berrou et al., 2001). Moreover, WT Ca_V1.2 coexpressed with Ca_Vβ_2a and Ca_Vα₂ undergoes an ultraslow inactivation that develops over several seconds (Ferreira et al., 2003). Here, in the absence of Ca_Vα₂, the fraction of channels that inactivated after a 10-s depolarizing pulse did not exceed 30% over a wide range of voltages. Fig. 5 A shows the time courses of inward currents during a single 10-s pulse to 0 mV. The percentage of residual current remaining at the end of the pulse was virtually identical for WT, E462R, and K465N Ca_V1.2 channels (Fig. 5 B). To evaluate whether the reduction in I/Q was due to an increase in the fraction of mutant channels remaining inactive, we compared currents recorded at 0 mV after a 10-s prepulse to either −120 or −60 mV (Fig. 5 C). We were unable to detect any differences; thus, the reduction in I/Q observed in the presence of Ca_Vβ_2a cannot be attributed to an enhancement of a slow voltage-dependent inactivation.

Because Ca_Vβ_2a shifts normalized GV curves of mutant and WT channels to the same degree, a similar fraction of channels is expected to be complexed with the auxiliary subunit and, therefore, substitutions of AID-exposed residues should not influence the affinity of Ca_Vβ_2a for the anchoring domain. Following the same strategy used previously (Hidalgo et al. 2006), we performed dose-response experiments for each variant of Ca_V1.2 with purified Ca_Vβ_2a protein and estimated the fraction of bound channels by modeling I/Q versus voltage plots (Fig. 6, A and B). In the presence of Ca_Vβ_2a, these plots differ in size and shape from the one recorded in oocytes expressing Ca_V1.2 variants alone. In oocytes exposed to intermediate concentrations of purified Ca_Vβ_2a, I/Q plots could be described by the weighted sum of template I/Q curves from Ca_V1.2 alone and from Ca_V1.2-expressing oocytes injected with saturating concentration of Ca_Vβ. The relative weight of Ca_V1.2-β_2a template was taken as the fraction of Ca_Vβ_2a-bound channels (β2a-like). The apparent dissociation constants (K_d) were then calculated from the fit of β2a-like coefficients versus concentrations of purified Ca_Vβ_2a plot to a standard Hill's equation (Fig. 6 C). The resulting values for K_d's were similar for all Ca_V1.2 variants (0.20 μM for WT Ca_V1.2, 0.22 μM for Ca_V1.2 E462R, and 0.25 μM Ca_V1.2 K465N). As in Hidalgo et al. (2006), the Hill coefficient was allowed to vary freely, and best fits were obtained with values between 1.4 and 1.6. As discussed in a previous paper (Hidalgo et al., 2006), this type of experiment does not rule out a second binding site, but this seems unlikely in light of experiments showing that channel modulation is fully recapitulated by covalently linking a single Ca_Vβ_2b to Ca_V1.2 (Dalton et al., 2005). On the other hand, it is possible that full equilibrium was not reached when the experiments were performed. Nevertheless, the fact that K_d and Hill coefficient were virtually identical for all Ca_V1.2 variants strongly indicates that the fraction of Ca_Vβ_2a-bound channels was similar in all cases, and that the substituted residues indeed did not influence AID–Ca_Vβ interaction. This result agrees with a recent report that K_d between Ca_Vβ and synthetic AID peptides derived from different Ca_Vα₁ subtypes is nearly identical (Van Petegem et al., 2008).

To determine whether changes in unitary conductance contribute to the reduced I/Q in mutant channels, we measured single channel conductance in the presence of Ca_Vβ_2a for all four mutants of AID-exposed residues (Ca_V1.2 E462R, Ca_V1.2 K465N., Ca_V1.2 D469S, and Ca_V1.2 Q473K) and found it to be ∼17 pS for all them (Fig. S3). We then compared single channel activity in 1,000 traces that recorded the responses to 200-ms pulses to 0 mV repeated once a second as in Dzhura and Neely (2003). Fig. 7 A shows 20 consecutive traces with this protocol for Ca_V1.2 E462R and Ca_V1.2 K465N (data for Ca_V1.2 D469S and Ca_V1.2 Q473K is shown in Fig. S4). A decrease in the overall Po in all mutants is evident by simple inspection because of the prevalence of sweeps lacking channel openings. Channel activity in traces from K465N and E462R mutants (Fig. 7), and also from D469S and Q473K channels (Fig. S4), display long bursts. These were similar to the ones that dominated activity in Ca_V1.2 WT in the presence of Ca_Vβ. In contrast, in Ca_V1.2 W470S, channel openings are rather brief, as reported for WT calcium channel expressed in the absence of Ca_Vβ (Dzhura and Neely, 2003). The Po for each trace, estimated as the fraction of time that the channel spends in the open state, was plotted with respect to trace number to yield a typical diary plot (Fig. 7 B). As reported for WT Ca_V1.2 (Costantin et al., 1998; Dzhura and Neely, 2003), these plots show that channel activity is clustered and all variants retain the ability to gate in the different modes. Each mode has its own set of transition rates that determine the channel Po within a trace, which is also revealed in Po histograms using log binning (Fig. 7 C). These histograms are clearly multimodal and demonstrate that the Po among active sweeps for Ca_V1.2 E462R and Ca_V1.2 WT were distributed similarly, whereas with Ca_V1.2 K465N, low Po sweeps were more frequent. However, low and high Po modes appear to peak at the same values for both mutants. This stands in clear contrast with Ca_V1.2 W470S, which is near mono-modal; it lacks sweeps with Po > 0.1, and it is dominated by sweeps with Po around 0.05. Although the chance of including traces with more than one active channel is not negligible for this mutant, the observed changes resemble those of WT Ca_V1.2 channels lacking Ca_Vβ (Dzhura and Neely, 2003).

In Fig. 8, channel activities from several patches were compiled by normalizing the number of sweeps within each Po bin by the total number of traces in each recording. Ca_V1.2 W470S was excluded because of uncertainties about the number of channels in each patch. In all cases, Po histograms are bi-modals, with one mode peaking around 0.5. The low Po component is dominated by Po's between 0.001 and 0.01 in WT and E462R channels, whereas it is shifted toward slightly higher Po for K465N, D469S, and Q473K mutants (Po ≈ 0.02). However, the relative weight of each mode changed more dramatically and was quantified by calculating the fraction of null traces, non-null traces with Po ≤ 0.1, and traces with Po > 0.1 (Fig. 8 D). WT channels dwell a comparable amount of time in each mode of activity (34.3 ± 8.8, 25.6 ± 6.9, and 40.1 ± 9.2% for nulls, 0 < Po ≤ 0.1, and Po > 0.1, respectively). In contrast, all mutants rarely visit the high Po mode (6.5 ± 1.5% for Ca_V1.2 E462R, 4.3 ± 2.2% for Ca_V1.2 K465N, 8.9 ± 8.0% for Ca_V1.2 D469S, and 11.3 ± 6.0% for Ca_V1.2 Q473K). The rest of the time is spent between nulls and low Po mode, with Ca_V1.2 E462R visiting the low Po less frequently (16.6 ± 2.5%) compared with 40.3 ± 12.0% for Ca_V1.2 K465N, 37.8 ± 13.8% for Ca_V1.2 D469S, and 41.9 ± 10.1% for Ca_V1.2 Q473K. It should be noted though, that K465N, D469S, and Q473K mutants display higher patch-to-patch variability in the modal gating behavior, with low Po traces ranging from 10 to 83%. All these changes contribute to a reduction in the overall Po, which, when measured as the average fraction of time spent in the open state, ranges from 0.21 ± 0.08 for Ca_V1.2 WT to values ranging from 0.05 to 0.03 for mutant channels.

To visualize the impact of the decrease in Po on the macroscopic currents, mean current traces from single channel recordings were built by averaging traces during 0 mV jumps, from multiple patches for several channel variants (Fig. 9). Ca_V1.2 WT yields 210 fA, which compares to 29 fA for E462R and 41 fA for K465N (Fig. 9 A). This corresponds to an 86 and 81% reduction for E462R and K465N, respectively, and nearly matches the reduction in I/Q and Po described above. Single channel activity was recorded in the presence of S(-)Bay K 8644. Because the modal gating behavior may be influenced by the presence of this agonist (Lacerda and Brown, 1989), we compared I/Q versus voltage curves in the presence of 0.1 μM S(-)Bay K 8644 and 76 mM external Ba²⁺ to mimic single channel recording conditions (Fig. 9 B). I/Q's for Ca_V1.2 E462R and Ca_V1.2 K465N were 77 and 63% smaller than for Ca_V1.2 WT, respectively, under the same recording conditions. This compares to the 87 and 79% difference in I/Q's observed in 10 mM Ba²⁺ and in the absence of the calcium channel agonist, indicating that differences in Po cannot be attributed to a differential effect of S(-)Bay K 8644. Collectively, these data show that the reduction in current capacity in channels bearing a mutation in AID-exposed residues and in the presence of Ca_Vβ_2a results from a decrease in Po associated with a decrease in the time spent in high Po mode.

To evaluate if AID-exposed residues also modulate gating within each mode, we performed a standard dwell-time analysis for the different mutants. Fig. 10 shows open and shut-time histograms from the same patches shown in Fig. 7, but displayed in Sine-Sigworth coordinates. The maximum likelihood algorithm (Sigworth and Sine, 1987) was used to optimize multi-exponential probability density functions. Histograms for Ca_V1.2 WT and mutant channels are better described by the sum of two exponential distributions. The mean lifetime of long- and short-lived open states are indistinguishable between Ca_V1.2 WT and channels bearing mutations of AID-exposed residues (Table S3). In contrast, single channel activity from Ca_V1.2 W470S mimicked the activity of WT channels recorded in the absence of Ca_Vβ (Dzhura and Neely, 2003). The relative contribution of short-lived openings, typical of low Po activity, was significantly larger in mutants with 25% of the active traces being low Po (Ca_V1.2 K465N, Ca_V1.2 D469S, and Ca_V1.2 Q473K). This should not come as a surprise because the apparent mean open-time in low Po sweeps is shorter (Dzhura and Neely, 2003), and their contribution should be augmented in channels that visit this mode of gating more often. Differences in shut-interval histograms, meaning the lifetime and relative contribution of short and long events, were not correlated with the prevalence of the different gating mode in any obvious manner and did not help in detecting changes in the kinetic within modes.

We next compared burst-duration histograms containing all traces and described them by the sum of two exponential distributions (Fig. 11 A). In all cases, the fast component had a mean lifetime of a fraction of a millisecond, whereas longer-lived burst of openings had a mean duration around 10 ms for all variants, except Ca_V1.2 W470S. In this mutant, long-lived bursts averaged 4.0 ± 0.8 ms, similar to what was previously reported for WT channels lacking Ca_Vβ (Dzhura and Neely, 2003). To separate bursts originating from different gating modes, histograms were built using traces with Po > 0.1 or 0 < Po ≤ 0.1. From this analysis, we can see that burst-duration histograms for Po ≤ 0.1 still required the sum of two exponential distributions for their description (Fig. 11 B), and that the time constants and relative contributions of both components were similar in all cases (Table S3). In contrast, burst-duration histograms from Po > 0.1 traces was described by a single exponential distribution with an estimated mean burst duration that ranges from 8.7 to 17.5 ms (Fig. 10 C, and Fig. S6 and Table S3). On average, mean burst duration for Ca_V1.2 WT is 11.9 ± 1.2 ms (n = 6), which compares to 10.9 ± 1.9 ms (n = 7) for Ca_V1.2 E462R. Although only one half of the patches displayed a reasonable number of traces with Po > 0.1, a similar value was obtained for K465N (10.9 ± 3.6 ms; n = 3). Similarly, mean burst durations obtained for Ca_V1.2 D469S and Ca_V1.2 Q473K were 9.1 ± 1.1 ms (n = 4) and 9.7 ± 3.2 ms (n = 4), respectively. Collectively, these results indicate that the kinetic within modes is rather insensitive to mutations of AID-exposed residues in contrast to the impact in modal gating.

In the absence of Ca_Vβ, the different Ca_V1.2 mutants yielded similar I/Q's (Fig. 3 B), suggesting that these AID mutations do not alter intrinsic gating of the calcium channel. As shown in Fig. 12, single channel activity appeared strikingly similar for all mutants. Unfortunately, channel activity underlying the expression of Ca_Vα₁ by itself is insufficient to rule out the presence of multiple channels; thus, long-shut intervals do not reflect a state lifetime. Consequently, only burst-duration histograms from all sweeps were compared, and two exponential distributions were required to describe them. All parameters for Ca_V1.2 E463R and Ca_V1.2 K465N turned out to be similar to those for WT channels (Table S4), indicating that the mutant phenotype is only expressed in the presence of Ca_Vβ. This suggests that AID-exposed residues come into place to interact with the gating machinery only when they are bound to this auxiliary subunit.

The main conclusion of this work is that substitutions of residues on the opposite face of the AID–Ca_Vβ interaction surface, the so-called AID-exposed residues, reduce channel current capacity only when Ca_Vβ is present. The change in the ionic-current versus voltage relationship, when normalized by Q_on, demonstrates that channel function rather than trafficking or expression is altered by these mutations. In single channel recordings, there is dramatic reduction in the overall Po, whereas unitary conductance remains constant. As described for WT Ca_V1.2 (Dzhura and Neely, 2003), Ca_V1.2 E462R, Ca_V1.2 K465N., Ca_V1.2 D469S, and Ca_V1.2 Q473K present clusters of openings in at least two nonsilent modes that are manifested in Po histograms using logarithmic binning. However, the relative contribution of each nonsilent mode differs among these mutants. Whereas in Ca_V1.2 E462R the low Po mode is seldom visited, K465N, D469S, and Q473K mutant channels spend most of their time between low Po and silent modes. On the other hand, burst-duration histograms derived from sweeps with Po > 0.1 were similar for all mutants and WT channels, suggesting that in the presence of Ca_Vβ, AID-exposed residues modulate channel Po changes in modal gating rather than in the kinetic within mode.

Substitutions of the conserved tryptophan residue by serine (Ca_V1.2 W470S) yielded channels that behave as WT channels in the absence of Ca_Vβ (Dzhura and Neely, 2003). This finding appears in discrepancy with previous results showing that Ca_Vβ_2a modulation is spared when this conserved tryptophan residue is substituted by alanine in Ca_V2.2 channels (Leroy et al., 2005). Nevertheless, this tryptophan is buried in the α-binding pocket of the β-subunit (Opatowsky et al., 2004; Van Petegem et al., 2004; Chen et al., 2004) and, according to a recent comprehensive survey on the energetic of Ca_Vα₁–Ca_Vβ interaction, this residue is critical for binding and modulation of function by Ca_Vβ (Van Petegem et al., 2008).

There are several reports indicating that AID residues or more proximal elements of the I-II loop participate in channel inactivation (Sandoz et al., 2004b; Raybaud et al., 2007; Berrou et al., 2005; Cens et al., 2006). Furthermore, it has been reported that E462R mutant of Ca_V1.2 channels display an acceleration of voltage-dependent inactivation when combined to Ca_Vα₂/δ and Ca_Vβ₃ (Berrou et al., 2001; Dafi et al., 2004). If the reduced Po that we observed on E462R plus Ca_Vβ_2a stemmed from an increase in the inactivation rate, it would be visible in prolonged pulses, which we did not observe. Moreover, the fact that this mutation appears to impair Ca_Vβ_2a-mediated increase in I/Q to a larger extend than Ca_Vβ_1b, which promotes inactivation, reinforces the idea that the effect of Ca_Vβ on channel activation is independent of its impact on the rate of channel inactivation (Olcese et al., 1994).

Our data confirm that calcium channels coexpressed with Ca_Vβ alternate between low Po and high Po modes (Shistik et al., 1995, Dzhura and Neely, 2003, Luvisetto et al., 2004). Because Ca_Vα₁–Ca_Vβ interaction is reversible (Hidalgo et al., 2006), it appears plausible that channels switch from low to high Po upon binding to Ca_Vβ. In such a case, a reduction in the fraction of time spent in the high Po mode should correlate with an increase in the proportion of channels lacking Ca_Vβ. Here, we show that Ca_V1.2 mutants that seldom visit the high Po mode maintain their apparent affinities constant for Ca_Vβ. This is entirely in line with experiments showing that AID derived from different calcium channel subtypes share the same affinity for Ca_Vβ (Van Petegem et al., 2008). To explain changes in the prevalence among different gating modes, we envision instead that Ca_Vβ remains attached to the channel and positions AID-exposed residues to interact with a still unknown region of Ca_Vα₁, and that this interaction may take two possible configurations to account for low and high Po modes. In the absence of this putative interaction, channel would remain in a silent mode. Changing the side chain of some of these residues, as in K465N, D469S, and Q473K mutants, destabilizes the configuration that supports high Po, and shifts the equilibrium toward low and silent mode, whereas channel-bearing E462R mutations spend most of their time in the silent mode because the stability of both configurations is reduced. Within this new framework, β-subunit potentiation of calcium channel would arise from an increase in the proportion of channels gating in the high Po mode, and the coupling efficiency between the voltage sensor and the channel can be regulated by changing the distribution of the different gating modes.

We thank Vivian Gonzalez and John Ewer for critical review and valuable comments on this manuscript.

This work was part of G.G-G. Doctoral dissertation under a CONICYT graduate fellowship and was funded by grants from Proyecto Anillo de Ciencia y Tecnología (ACT-46) to A. Neely and the Deutsche Forschung Gemeinschaft (Grant FOR 450, TP1) to P. Hidalgo.

Olaf S. Andersen served as editor.