Mechanism of Auxiliary Subunit Modulation of Neuronal α_1E Calcium Channels

Voltage-gated calcium channels are molecular transducers that trigger cellular processes ranging from muscle contraction to neurotransmission. Modulation of these channels thereby constitutes a key potential mechanism for functional adaptation and plasticity. At least three different subunits are believed to comprise native calcium channels: a main, pore-forming α₁ subunit, a cytoplasmic β subunit, and a disulfide-linked α₂δ subunit (for review, see Perez-Reyes and Schneider, 1994; De Waard et al., 1996). So far, seven different genes encoding α_{1A,B,C,D,E,S,G} subunits, and four different genes encoding β_1,2,3,4 subunits have been identified, along with multiple splice variants. Given this heteroligomeric structure, regulation of channel properties by variations in subunit composition have been widely studied as a potential mechanism for tuning channel gating properties to support a given physiologic role.

Despite the potential importance of modulation by subunit combination, fundamental uncertainties remain about the effects of auxiliary subunits (for review see Perez-Reyes and Schneider, 1994; De Waard et al., 1996; Walker and De Waard, 1998). While coexpression studies have demonstrated that the addition of auxiliary subunits (β, α₂δ) can have striking effects on channel gating and/or channel expression, the specific effects observed vary across studies, even using the same α₁ subunit. At least some of the differences in subunit effects may reflect isoform-specific variations in the effects of distinct β subunit isoforms on α₁ gating. Further variability may arise from the use of diverse expression systems, electrophysiological methods, and experimental solutions. These points underscore the need to explore subunit modulation of each α₁ isoform individually, and to undertake comprehensive studies with uniform experimental conditions.

Although most previous work has focused on α_1C, neuronal α_1E channels (Soong et al., 1993) have recently emerged as important channels with which to attempt such comprehensive investigation for several reasons. First, subunit modulation of α_1E has potential physiological relevance, as α_1E (presumed “R-type”) channels have been implicated in neuronal functions including neurotransmitter release (Wu et al., 1998). Second, α_1E demonstrates an exceptional capacity for high-level recombinant expression, which permits well-resolved measurements of both ionic and gating currents, even when the α_1E subunit is expressed alone (which generally lowers overall expression of current). This capability enables examination of changes in both peak open probability and channel density (Olcese et al., 1996), two critical measures for resolving how auxiliary subunits affect the overall level of calcium current. Third, β subunits may affect α_1E expression in a uniquely different manner than observed with other pore-forming α₁ subunits, providing a potentially useful clue as to the underlying mechanism of subunit modulation. Olcese et al. (1994, 1996) provide the most biophysically detailed results in this regard, using Xenopus oocytes. In contrast to other α₁ subunits, β subunits caused little change or even a decrease in overall α_1E current density. This outcome resulted from decreased channel density, as assessed by maximal gating charge, countered by increased channel opening. By contrast, in mammalian expression systems, β subunits increased overall α_1E current density (Williams et al., 1994; Stephens et al., 1997). Here, however, no α_1E gating-current measurements have been made to permit assessment of underlying changes in channel density and open probability. Finally, α_1E is one of the channels in which a second β binding site has been explicitly identified (Tarelius et al., 1997; Walker et al., 1998). Characterization of mutant α_1E constructs lacking this site would allow determination of the functional importance of the secondary site.

Here, we therefore examine subunit modulation of α_1E channels coexpressed with various combinations of auxiliary subunits (β₁–β₄, α₂δ) in mammalian HEK 293 cells. The same recombinant expression system, along with a consistent set of experimental solutions and protocols, is used throughout to facilitate direct comparison of channels with differing molecular composition. Measurements of both ionic and gating currents permits in-depth analysis of subunit modulatory effects. We focus on three key questions. (a) To what degree does modulation of α_1E current density reflect modulation of channel gating and/or number of functional channels? (b) How do different auxiliary subunits compare with regard to modulation of activation gating? (c) What is the functional impact of the secondary β binding site in α_1E? Through addressing these questions, this study helps to establish a more refined picture of auxiliary-subunit modulation of α_1E calcium channels.

HEK 293 cells, obtained from Dr. Jeremy Nathans (Johns Hopkins University; Gorman et al., 1990), were grown at 37°C in Dulbecco's modified Eagles medium (GIBCO BRL, Grand Island, NY), 10% fetal calf serum (GIBCO BRL), 1% L-glutamine (Sigma Chemical Co., St. Louis, MO), 1% penicillin-streptomycin (P0906; Sigma Chemical Co.), in 5% CO₂. Low-passage number cells were used (<P20). cDNAs encoding channel subunits α_1E (Soong et al., 1993), α_1C (Wei et al., 1991), β_1b (Pragnell et al., 1991), β_2a (Perez-Reyes et al., 1992), β₃ (Castellano et al., 1993b), β₄ (Castellano et al., 1993a), and α₂δ (Tomlinson et al., 1993) were subcloned into mammalian expression plasmids (pMT2; Genetics Institute, Cambridge, MA, for β₄, pZEM229R; ZymoGenetics, Inc., Seattle, WA, for α₂δ, pGW1; British Biotechnologies, Cowley, Oxford, UK for all others). α_1EΔ was constructed by replacing the Bst 1107I (α_1E: nucleotide 4299, given start codon at nucleotide 1) and SalI (3′ polylinker) region of α_1E in pGW1 with a shorter polymerase chain reaction fragment, including a premature stop codon after the codon for amino acid 1871. The portion of the channel derived from PCR was verified in its entirety with the use of the fluorescent dideoxy terminator method of thermocycle sequencing on an automated DNA sequencer (Applied Biosystems Division 373a; Perkin-Elmer Cetus Instruments, Emeryville, CA). HEK 293 cells were transiently transfected using a standard, calcium-phosphate precipitation procedure (Brody et al., 1997) with a total of 30 μg of DNA per 10-cm plate. 10 μg of a plasmid containing a pore forming subunit was included (α_1E or α_1C) and mixed with 10 μg of each desired auxiliary subunit (none, a β subunit, and/or the α₂δ subunit). If the amount of DNA totaled <30 μg, pBluescript was added to make up the difference. For certain experiments, both β_2a and β₃ were simultaneously transfected either in a 1:1 ratio (10 μg of each plasmid) or a 5:1 ratio (15 μg of β₃, 3 μg of β_2a). More than 20% of cells transfected with a pore forming subunit exhibited detectable high-threshold calcium currents.

“Mock-transfected” cells were transfected with 10 μg of β_1b, 10 μg of α₂δ, and 10 μg of pBluescript. In our usual ionic current recording conditions (detailed below), we observed no high threshold, voltage-gated, calcium-channel currents in such cells (n = 32 cells, over two independent rounds of transfection), or in cells transfected with the β_2a subunit alone (n > 40 cells; Patil et al., 1998). In mock-transfected cells, we occasionally (∼10% of cells) observed endogenous, low threshold calcium channel currents of small amplitude (peak ionic current ∼20 pA in 10 mM Ba²⁺), as reported previously by Sun et al. (1994). Although endogenous currents of such small amplitude would contribute negligibly to our results, cells with low threshold activity were nevertheless rejected. At the biochemical level, Western blots performed on total membrane protein (30 μg/lane) from untransfected cells revealed no known high threshold α₁ (A, B, C, D, E) or β (1b, 2e, 3a, 4) subunits, and only low levels of α₂δ (personal communication, Mark Williams, SIBIA Neurosciences Inc., La Jolla, CA). Blots were probed individually with appropriate antibodies, and the lack of subunit proteins was gauged from the absence of bands that were clearly present using cells transfected with corresponding recombinant subunits. The result that coexpression of α₂δ with α_1E potentiated current by approximately threefold suggests that trace expression of endogenous α₂δ did not significantly influence our results.

Whole-cell recordings were obtained at room temperature 48–72 h after transfection using an Axopatch 200A (Axon Instruments, Foster City, CA) and standard patch-clamp techniques. Cell capacitance ranged from 10–40 pF. Series resistance was typically <5 MΩ, and compensated 70–85%, resulting in a typical settling time of ∼80 μs. Voltage pulses were delivered every 15–20 s from a holding potential of −110 mV, except for prepulse inactivation protocols, where voltage pulses were given every minute from a holding potential of −120 mV to allow recovery from inactivation. Data were typically acquired at 50 kHz and filtered at 10 kHz (−3 dB, four-pole Bessel). Displayed traces have generally been additionally processed with a gaussian digital filter at 2 kHz. Leak and capacity currents were subtracted by a P/8 protocol (ionic currents) or P/−8 protocol (gating currents) from the −110-mV holding potential, unless otherwise noted (Armstrong and Bezanilla, 1974). To allow better resolution of small currents, we often subtracted a smooth curve fitted to the leak currents. In some cases, the first 200 μs after a voltage step contains a large leak subtraction artifact, which was zeroed when present before digital filtering.

The base external solution contained (mM) 155 N-methyl-d-glucamine (NMG) aspartate, 10 HEPES, 10 4-aminopyridine, 0.1 EGTA, pH 7.4 with NMG, 280–300 mOsm with no added charge carriers. The internal solution contained (mM) 150 NMG-methanesulfonate (MeSO₃), 1 MgCl₂, 4 MgATP, 10 HEPES, 10 EGTA, pH 7.3, with NMG, typically 280–290 mOsm. The h(∞)-V relations shown in Fig. 11 for α_1C were obtained with an internal solution in which the 150 mM NMG-MeSO₃ was replaced by 150 mM Cesium-MeSO₃. For measurement of ionic currents, either 2 or 10 mM BaCl₂ was added to the external solution. For typical gating current measurements, 0.2 mM LaCl₂/2 mM MgCl₂ was added. External solution flowed continuously at a rate of 1–2 ml/ min during recording. The bath solution was grounded by a 0.5 M KCl agar bridge attached to a Ag-AgCl wire. Measurements were started after >5 min of dialysis with the internal solution. In all cases, the junction potential between external and internal solutions was ∼5 mV (Neher, 1992). To determine the true applied potential, this value should be added to the voltages in the figures and text.

For measurement of α_1E activation curves, 2 mM BaCl₂ was the charge carrier throughout. Test depolarizations were 30 ms long and ranged from −70 to +70 mV (see Fig. 2 A, top) with repolarization to −50 mV to allow good resolution of tail currents. For each cell, plots of peak tail current at −50 mV (I_tail) vs. test pulse voltage (V_test) were normalized by an estimate of maximal peak tail current (I_tail,max). I_tail,max was taken as the saturating value of Boltzmann fits to the I_tail-V_test data. The resulting normalized relations are equivalent to normalized P_o-V relations, and are referred to as G-V curves. G-V curves were then averaged across cells. Such G-V curves were indistinguishable from G-V relations obtained using 15-ms test depolarizations (data not shown). We did not correct tail currents for the contribution of the “OFF” gating current. To assay the magnitude of the error that such OFF gating currents might produce, we corrected G-V relations for six cells transfected with α_1Eβ_2a by subtracting the OFF gating currents measured during repolarization to −50 mV. We found that the average single-Boltzmann fit parameters for the corrected and uncorrected G-V curves were statistically indistinguishable (P < 0.05, Student's t test, uncorrected: z = 3.51 ± 0.4, V_1/2 = −20.1 ± 3.2 mV; corrected: z = 3.57 ± 0.5, V_1/2 = 19.7 ± 3.4 mV), although I_tail,max was reduced by ∼5% (−2,638 ± 568 pA [uncorrected] vs. 2,516 ± 542 pA [corrected]). G_max was calculated according to G_max = I_tail,max/(V − V_rev), where V_rev was +40 mV in 2 mM BaCl₂. Therefore, the small error in I_tail,max will lead to a slight overestimate of the G_max/Q_max ratio, which may vary slightly for the different subunit combinations.

For gating currents, ionic currents were blocked by the external solution containing 0.2 mM LaCl₃ (Bean and Rios, 1989). The effective free La³⁺ concentration was 0.1 mM due to the presence of 0.1 mM EGTA in all external solutions. The voltage protocol was the same as for ionic currents, except that the test pulse duration was decreased to 15 ms, and repolarization to −110 mV (see Fig. 5 A, top). Total charge moved during test depolarization (Q_on) was obtained by integrating over the entire depolarizing epoch, taking as the zero baseline the average current over the last 3 ms of the test pulse. Total charge moved during repolarization (Q_off) was calculated similarly. For each cell, Q_on-V and Q_off-V curves were normalized by an estimate of maximal mobile charge (Q_max), taken as the saturating value of the Boltzmann fit (detailed below) to the Q_on-V or Q_off-V curves, as indicated in the text. Such normalized Q_on-V and Q_off-V curves were averaged across cells.

To ensure that La³⁺ does not alter activation gating, we obtained Q-V relations both in the presence and absence of La³⁺ blockade. Fig. 1 A shows the results of the analysis, in which we compared Q_off-V curves acquired in 2 mM MgCl₂ (•) and 2 mM MgCl₂/0.2 mM LaCl₃ (○). The identity of the two curves, absent the expected surface-potential shift, provides additional strong support that La³⁺ does not perturb activation gating.

To determine explicitly the surface-charge shift between solutions used for ionic and gating currents, we exploited the property that isolated gating currents can actually be measured in the solution for ionic current, so long as the voltage range is negative to the threshold (∼−65 mV, Fig. 1,B, inset) for ionic-current activation. We could then calculate the surface-charge shift by direct comparison of the rising “foot” of Q-V curves obtained in ionic and gating current solutions. Fig. 1,B shows the results of this approach. Before averaging across cells, Q-V data for a single cell was normalized by the value of Q_on at −65 mV in 2 mM Ba²⁺. In the ionic-current solution containing 2 mM Ba²⁺, the Q_on-V (Fig. 1,B, inset, ○) and Q_off-V (inset, •) curves matched at potentials negative to −65 mV, indicating that gating currents were isolated below this potential. The main graph in Fig. 1 B demonstrates that, over this range of voltages, Q_on-V relations obtained in 2 mM Ba²⁺ (○) and 2 mM MgCl₂/0.2 LaCl₃ (⋄) are essentially indistinguishable, indicating that there is little if any surface-charge shift between solutions. To quantitate the value of the shift, for each cell the voltage shift required to fit the same dual-Boltzmann to both sets of Q_on-V data was taken to be the surface potential difference. Averaging this value across cells gave a value of 3 ± 1 mV (n = 9). These results excluded the need for surface-charge correction between ionic and gating current measurements.

Steady state inactivation curves were approximated using a protocol in which a 20-s prepulse was followed by a step to peak of the current–voltage (I-V)¹ curve (typically −5 mV) to measure the fraction of inactivated current. In some cases, a 10-ms normalizing prepulse at the test pulse potential was included before the 20-s prepulse to assay for the presence of cumulative inactivation or rundown. Steady state inactivation (h(∞)-V) curves were derived by normalizing test pulse currents by either the current during the normalizing test pulse, or by the value of the test pulse with no prepulse. All steady state inactivation curves were measured with 10 mM Ba²⁺ as charge carrier. Voltage commands were given every minute from a holding potential of −120 mV. Typically, prepulse voltages ranged from −120 to −20 mV in 10-mV increments. Normalized h(∞)-V relations were averaged across cells. For cells transfected with two β subunits, the h(∞)-V relation was fit with a dual-Boltzmann function to obtain parameters for the low and high threshold components, in addition to the relative contribution of each component.

Boltzmann fits to either G-V or Q-V relations were performed with functions of the form B(V) = B_max/{1 + exp[−zF(V − V_1/2)/ RT]}, where B_max is the saturating value, z is the effective charge, and V_1/2 is the midpoint of activation. Q_on-V data above +40 mV were sometimes unreliable and were therefore excluded.

For dual-Boltzmann fits to h(∞)-V relations, we used a function of the form B(V) = f_low{1 + exp[z _low F(V − V_1/2,low)/RT]}⁻¹ + f_high{1 + exp[z _high F(V − V_1/2,high)/RT]}⁻¹, where V_1/2,low and V_1/2,high are midpoints of activation, z_low and z_high are the effective valences, and f_low and f_high are amplitudes of low and high threshold components. Fits were obtained using nonlinear, least-squares minimization. All reported values are mean ± SEM.

Transfection of HEK 293 cells with the α_1E subunit alone, or in combination with various auxiliary subunits, led to the expression of well-resolved inward barium currents carried by recombinant calcium channels (Fig. 2,A). The relative magnitudes of the various sets of traces illustrate that addition of auxiliary subunits caused striking increases in the level of expressed current. To quantify the relative increase in current density, we calculated the maximum tail current upon repolarization to 50 mV [G_max = nP_o,max g(−50 mV)h], where g is the unitary conductance, h is the fraction of noninactivated channels at the end of the test pulse, n is the number of channels, and P_o,max is the maximum open probability. Fig. 2,B compares the average values of G_max for all different subunit combinations examined. The largest effect was the ∼12-fold enhancement of expressed current with the coexpression of β subunits. All β subunits were approximately equipotent in this regard, although the average β₃ effect was slightly smaller (approximately sevenfold). Addition of α₂δ to α_1E produced a weaker increase in current (about threefold), and the combination of α₂δ and β subunits yielded no appreciable current enhancement over the coexpression of β subunits alone. Since modulation of G_max values may reflect not only changes in nP_o,max, but also differences in the number of noninactivated channels (h) with different subunits, we examined another measure of current density (Fig. 2 C), I_peak = nP_o[V_peak] i[V_peak], where P_o[V_peak] and i[V_peak] are the open probability and unitary current at the voltage (V_peak) yielding the maximum test-pulse current. This measure (I_peak), which is less sensitive to test pulse inactivation, gave similar results. Therefore, we are confident that G_max can henceforth be used as a quantitative indicator of relative changes in current (nP_o,max).

To determine the origin of the increased current density (nP_o,max), we wished to measure the maximum amount of mobile gating charge (Q_max = nq, q is the charge per channel), which provides a convenient assay for the relative number of functional channels (n). Measuring Q_max involves good resolution of the currents arising from gating charge movement (gating currents; Armstrong and Bezanilla 1977; Sigworth, 1994), which in turn requires a blocker that eliminates ionic currents without significantly perturbing channel gating behavior. Previous work indicates that the highly potent block by La³⁺ can be used to isolate gating currents of calcium channels containing the α_1B subunit (Jones et al., 1997a), but not the α_1C subunit (Kamp et al., 1996). To determine the feasibility of La³⁺ blockade of calcium channels containing α_1E, we examined gating currents with either 2 mM Ba²⁺ or 2 mM Mg²⁺/0.2 mM La³⁺ added to the bath solution (Fig. 3,A). Although 2 mM Mg²⁺/0.2 mM La³⁺ (solid traces) completely blocked ionic currents, the early outward transients that are dominated by gating current were unchanged, arguing strongly that La³⁺ does not alter the voltage sensor movement that underlies activation gating. Furthermore, the block of ionic current was completely reversible (Fig. 3,B, •), and did not alter Q_rev during repetitive stimulation in the presence of La³⁺ (○, obtained by integrating outward transients at the reversal potential). The lack of change of Q_rev argues that La³⁺ does not promote channel inactivation, which would be apparent as a reduction in Q_rev (Jones et al., 1997b) known as gating-charge immobilization (Armstrong and Bezanilla, 1977; Bezanilla et al., 1991). Similar results to those for α_1E + β_2a (Fig. 3, A and B) were obtained with the other subunit combinations (data not shown). Further experiments (see materials and methods) demonstrated that La³⁺ did not affect the voltage dependence of charge movement, and that the surface charge shift between solutions used for ionic and gating current measurements was ∼3 mV.

With assurance that La³⁺ does not detectably alter gating or surface-charge properties, we turned to analysis of extensive sets of currents recorded during La³⁺ block for α_1Eβ_2a (Fig. 4,A). These traces represent genuine calcium-channel gating currents for several reasons. First, no such currents are observed in mock-transfected cells (Fig. 4,B). Second, no nonlinear charge movement is present in the range of our leak pulses (Fig. 4,C). Third, the measured charge movement is not affected by the choice of the leak subtraction protocols (data not shown). Finally, these “nonlinear displacement” currents have the standard properties typically associated with gating currents (Fig. 4,D): time integrals of outward (Q_on) and inward (Q_off) displacement currents saturated with increasing test depolarization; Q_off ∼ Q_on in the absence of inactivation (Fig. 4,D); charge movement (Q_on-V or Q_off-V curves) occurs before, and then parallels, ionic-current activation (G-V curve); and finally, the maximal amount of gating-charge (Q_max) is linearly correlated with maximal current density (G_max) (Fig. 4 E).

With the ability to isolate gating currents, we could now compare auxiliary subunits with regard to their mechanism for current potentiation. Fig. 5,A displays representative gating-current records for most of the different subunit combinations. These traces illustrate that all auxiliary subunits boost the maximum amount of gating charge (Q_max), which is taken as the saturating value of the Boltzmann fit to the Q_on-V relation. Fig. 5,B compares the average values of Q_max for all different subunit combinations examined. Q_max for channels expressed from α_1E alone was characteristically small, with a mean of 0.8 ± 0.1 fC/pF (n = 9). β Subunits induced the strongest enhancement of Q_max, ranging from fourfold for β₃ to sevenfold for β₄. Coexpression of α₂δ also produced clear augmentation of Q_max, though the effect was less potent than for β subunits. Table I summarizes the complete details of the analysis. Given that gating-charge per channel (q) does not appear to be affected by auxiliary subunits (Noceti et al., 1996), the rise in Q_max likely reflects an increase in the number of functional channels (n). Hence, our results indicate that the enhancement of current density by auxiliary subunits arises, at least in part, from an increase in the number of functional channels. Such an increase in the number of functional channels may reflect either improved processing and trafficking of α_1E channels (increasing total amount of α_1E protein), or an increase in the fraction of functional α_1E protein in the membrane (with no increase in total amount of α_1E protein) by the α₂δ and β subunits.

To determine whether an increase in the maximal open probability (P_o,max) also contributes to higher ionic-current densities, we calculated the ratio G_max/ Q_max, which is directly proportional to P_o,max, so long as auxiliary subunits do not alter permeation properties of the channel (as in Fig. 6,D and Noceti et al., 1996). Fig. 5,C shows that all β subunits approximately doubled G_max/Q_max, but α₂δ left the ratio unchanged. Table I reports further details of the calculations. The data in Fig. 5 suggest that β subunits enhance α_1E current density by jointly increasing the number of functional channels (as reported by Q_max) and the maximal open probability (as reflected by G_max/Q_max). The enhancement of current by the α₂δ subunit appears to be fundamentally different: there may be a pure increase in the number of functional channels, without change in P_o,max.

A second goal of this study was to compare auxiliary subunit effects on the kinetics and voltage dependence of channel activation. To qualitatively compare activation kinetics for channels with different β subunits, we normalized the rising phases of exemplar ionic-current records (Fig. 2,A) evoked by voltage steps to −30, 10, and +10 mV (Fig. 6,A). The identical trajectories of traces from all four β subunits suggest that β subunits produce channels with similar activation kinetics. Fig. 6 B shows the identical analysis for channels expressed from α_1E alone (solid trace) or from α_1E + α₂δ (dashed trace). The records for α_1E + β_1b (gray traces) are reproduced for comparison. Here again, the close correspondence between traces suggests that auxiliary subunits do not significantly modulate activation kinetics.

To examine whether auxiliary subunits affect the steady state voltage dependence of activation (G-V), we tested for subunit-dependent changes in G-V curves derived from peak tail currents (see materials and methods) (Fig. 6,C). Coexpression of the α₂δ subunit had little effect on the G-V (Fig. 6,C) or I-V (Fig. 6,D) relations. The lack of effect of α₂δ on the kinetics and voltage dependence of activation, as well as on G_max/ Q_max (Fig. 5,C), suggests that this subunit is functionally uncoupled from any aspect of activation in α_1E. In contrast, single-Boltzmann function analysis (Fig. 6,C, solid curves, and Table II) clearly demonstrates that coexpression of β subunits produces an ∼7-mV hyperpolarizing shift and a modest increase in the steepness of G-V relations (e.g., the Boltzmann valence (z) increases from 2.4 for α_1E to 3.6 for α_1Eβ_2a). As expected from the shift in the G-V relation, coexpression of β subunits shifted the peak of the I-V relation leftward (Fig. 6 D) without altering the reversal potential.

The results in Fig. 6, C and D, are compatible with earlier work on β subunit effects in Xenopus oocytes (Olcese et al., 1994), in which G-V relations were fitted with dual-Boltzmann functions. In agreement with the earlier report, application of dual-Boltzmann analysis to our G-V data (Fig. 6,C, dashed curves) suggests that the apparent hyperpolarization and steepening of activation by β subunits could arise from an increase in the proportion of the low threshold Boltzmann component from ∼30 to ∼70%, without change in valence or midpoint parameters of individual Boltzmann functions. More in-depth interpretation of the data, like that introduced by dual-Boltzmann analysis, is deferred to the discussion, where explicit fits of a multistate kinetic model will be employed. For simplicity, in the remainder of the results, we retain single-Boltzmann analysis for first-order characterization of experimentally resolvable changes in activation. Regardless of the particular analytical functions used to describe the data, the results thus far (Figs. 5 and 6, A–D) clearly indicate that β subunits increase ionic current by simultaneously modulating the G-V relation and doubling the G_max/ Q_max ratio, in agreement with the findings of Olcese et al. (1994, 1996).

To explore the mechanistic basis of the β subunit effects on activation, we investigated how auxiliary subunits influenced Q-V curves derived from gating currents (Fig. 6,E). The rising phase of Q-V curves is very sensitive to modulation of the early events in the activation pathway, and the interrelation of Q-V and G-V curves lends insight into steps that couple voltage sensor movement to channel openings (Jones et al., 1997a). Fig. 6,E illustrates that all β subunits produced essentially identical effects on the Q-V relation: a small hyperpolarizing shift in the midpoint (∼5 mV) with little change in the steepness (Boltzmann valence [z] ranges from 2.5 to 2.8, Table II). The effects of β subunits on Q-V curves are smaller than on G-V curves, thereby narrowing the gap between Q-V and G-V relations along the voltage axis. As expected from previous null results, α₂δ had no effect on the Q-V relation. All the β subunit effects on gating (Figs. 5,C and 6, C–E), particularly the contraction between Q-V and G-V curves, fit nicely with the idea that all β subunits act primarily to modulate a single locus of weakly voltage-dependent steps late in the activation pathway (see discussion).

Previous reports in Xenopus oocytes indicate that α_1E channels containing different β subunits have strikingly different inactivation characteristics, despite very similar activation gating (Olcese et al., 1994). Here, we sought to confirm this effect in HEK 293 cells so that we could exploit this property to test whether multiple β subunits can simultaneously define the functional behavior of a calcium channel. To assay inactivation properties, we used a 20-s prepulse followed by a test pulse to the peak of I-V relations (Fig. 7,A). Typical currents for α_1Eβ_2a and α_1Eβ₃ channels illustrate the extremes of inactivation behavior observed with the different subunit combinations. β_2a dramatically slowed inactivation, while β₃ accelerated inactivation. β_1b and β₄ subunits also accelerated inactivation during the test pulse (not shown), though not as strongly as β₃. To provide a robust indication of the differences in inactivation properties, we used such records to calculate steady state inactivation curves (h(∞)V curves; Fig. 7,B). While addition of α₂δ did not affect the h(∞)V relation, coexpression of β subunits induced striking modulation of steady state inactivation: β_1b, β₃, and β₄ all left-shifted h(∞)V curves by ∼10, 15, and 10 mV, respectively; β_2a imparted a right shift of ∼15 mV. The profound distinction between the effects of β_2a and the other β subunits has been reported in previous studies of α_1E (Olcese et al., 1994), and of other neuronal calcium channels, including α_1A (Stea et al., 1994) and α_1B (Patil et al., 1998). Table III summarizes the Boltzmann analysis of h(∞)V data.

To investigate whether multiple β subunits can concomitantly specify the functional properties of a single calcium channel, we took advantage of the vast difference between the h(∞)V relations for α_1Eβ₃ and α_1Eβ_2a channels. If there are multiple β subunit sites on α_1E that specify inactivation properties, then cotransfection of both β_2a and β₃ subunits should result in mixed-composition channels (e.g., α_1Eβ_2aβ₃) whose inactivation behavior should be distinct from that of pure α_1Eβ_2a- or α_1Eβ₃-like channels. However, if there is only one functionally active β subunit site per channel, the aggregate h(∞)V relation should possess only two components. Fig. 8, A and B, shows the results for one such experiment in which β₃ and β_2a were cotransfected in a 1:1 weight ratio. This example demonstrates that inactivation is clearly biphasic, with a low threshold, readily inactivating component, as well as a high threshold, inactivation-resistant component. Only two Boltzmanns are required to produce an excellent fit of the data since the average residual for the dual-Boltzmann fit is close to zero (Fig. 8,C). Furthermore, the average fit parameters to the h(∞)V data correspond closely to the steady state inactivation properties of pure α_1Eβ_2a and α_1Eβ₃ channels (Fig. 7,B and Tables III and IV). Cotransfection of β₃ and β_2a in a 5:1 weight ratio merely decreased the relative amplitude of the low threshold component (Fig. 8,D), while preserving the intrinsic properties of the two components (Table IV). The only apparent deviation from parameters obtained with pure-composition channels is a small 7–9-mV increase in the V_1/2 for the high threshold component (compare Tables III and IV). Although this increase could reflect a minor contribution of mixed-composition channels, the overall results are consistent with the functional dominance of pure α_1Eβ_2a and α_1Eβ₃ channels.

As a further test for the possible functional role of a second β subunit site (Tareilus et al., 1997), we examined how auxiliary subunits modulated the properties of a COOH-terminal truncation of the α_1E construct (α_1EΔ, amino acids 1–1871 of α_1E [1–2251]) that lacks the secondary binding site. Fig. 9,A displays ionic currents for channels composed of α_1EΔ+ α₂δ or α_1E + β_2a + α₂δ subunits. Coexpression of β_2a with α_1EΔ increased G_max from −306 ± 115 pS/pF (n = 11) to −1,708 ± 251 pS/pF (n = 4), a 5.6-fold increase similar to the 4.1-fold increase in G_max seen for wild-type α_1E (Fig. 2,B). Similarly, modulation of activation by β_2a is unchanged by the COOH-terminal deletion, as demonstrated in Fig. 9,B by the identical subunit modulation of G-V relations for α_1EΔ (circles) and wild-type α_1E (squares). Finally, we compared β subunit modulation of the steady state inactivation properties of α_1EΔ (Fig. 9, C, traces, and D, symbols; Table III) with data obtained with wild-type α_1E (Fig. 9,D, lines). The α_1E data have been shifted uniformly by −7 mV in the Fig. 9,D overlay to account for a difference in inactivation that is present even without β subunit coexpression (e.g., α_1EΔ + α₂δ in Fig. 9,D); this small shift likely reflects a difference in the intrinsic inactivation behavior of the α₁ backbone (Soldatov et al., 1997), rather than a change in the modulatory action of β subunits. The close correspondence between h(∞)-V relations for α_1E (lines) and α_1EΔ (symbols) in the Fig. 9,D overlay illustrates that β subunit modulation of inactivation is similar for the two constructs. Although the small difference between modulation of α_1EΔ and α_1E inactivation (most apparent for α_1EΔβ₃α₂δ, Fig. 9,D, ▿) could reflect a minor contribution of a second β subunit binding site, all the results in Figs. 8 and 9 support the view that a single β subunit binding site predominates in specifying inactivation properties. If present, the potential contribution of a second site appears to be small by comparison.

Although auxiliary subunits clearly have a role in defining channel properties, specific modulatory effects vary widely across studies, underscoring the need to examine comprehensively the modulation of each α₁ subunit under the same experimental conditions. Here, we have performed a systematic evaluation of auxiliary subunit regulation of expression and gating of α_1E calcium channels in HEK 293 cells. The experiments lead to three main conclusions. (a) The α₂δ and β auxiliary subunits differ fundamentally in the manner by which they induce an overall increase in current density. Coexpression of α₂δ with the pore-forming α_1E moiety produced a clear-cut enhancement of current, arising purely from an increase in the number of functional channels (n), without significantly affecting channel gating behavior. By contrast, coexpression of β subunits induced stronger potentiation of current by joint elevation of channel number (n) and maximal open probability (P_o,max), suggesting effects on both channel assembly and gating. (b) While α₂δ had no appreciable effect on activation gating, β subunits produced significant hyperpolarizing shifts in the voltage dependence of ionic-current activation and gating-charge movement, all without discernible change in activation kinetics. Importantly, different β isoforms produced nearly indistinguishable effects in regard to both current potentiation and activation gating. (c) Little functional evidence for a secondary β subunit binding site was found, fitting with earlier biochemical evidence for a 1:1 stoichiometry of α₁ and β subunits for skeletal (De Waard and Campbell, 1995) and neuronal N-type (Witcher et al., 1993) channels. Together, these findings represent an important contribution to clarifying both the mechanism and structural determinants of auxiliary-subunit modulation of calcium channels.

In the sections to follow, we will first relate each of the conclusions to previous studies of α_1E and, where relevant, other calcium channels. For clarity, we will discuss α₂δ and β subunit effects sequentially, as independent parts. A kinetic mechanism is then developed to explain how β subunits can produce all the observed changes in gating, simply by alteration of the equilibrium between a single, weakly voltage-dependent transition near the open state. Finally, we consider the generality of our conclusions to other α₁ isoforms.

The α₂δ subunit produced an approximately threefold increase in α_1E current, which arose almost exclusively from elevated channel expression (Q_max). The α₂δ subunit had no other clear modulatory effects, except to slightly antagonize the effect of β₃ on inactivation (Fig. 7,B). All measures of activation gating, including the maximal open probability (G_max/Q_max), the voltage dependence of charge movement (Q-V), and the voltage dependence of ionic activation (G-V) were similar to α_1E alone. Similar effects on inactivation and expression were reported for doe1 (marine ray analog of α_1E) expressed in Xenopus oocytes (Ellinor et al., 1993). However, in contrast to our results, in studies of rat α_1E in COS-7 cells (Stephens et al., 1997) and human α_1E in Xenopus oocytes (Wakamori et al., 1994), coexpression of α₂δ was found to produce a depolarizing shift in the G-V without modifying expressed current levels; however, these studies agree with our findings concerning the slight antagonism of β₃ effects on inactivation (Fig. 7 B).

The applicability of our results to other channel types is unclear. However, it is interesting to note that the reported effects of the α₂δ subunit on other α₁ subunits also varies widely, sometimes agreeing with our findings, other times not. For example, with regard to modulation of α_1C channel density, the α₂δ subunit was found to increase ligand binding sites (Welling et al., 1993), protein levels (Shistik et al., 1995), and gating currents (Bangalore et al., 1996). By contrast, in other studies (Wei et al., 1995; Gurnett et al., 1997) of α_1C, maximal dihydropyridine binding is not increased by α₂δ coexpression. Similarly, activation kinetics of α_1C accelerate in some studies (Singer et al., 1991; Bangalore et al., 1996), but not others (Mikami et al., 1989; Welling et al., 1993). The sources of this variability have yet to be determined.

By contrast to the α₂δ subunit, coexpression of β subunits (β₁–β₄) enhanced current density (G_max) by increasing not only the number of functional channels (Q_max) but also the maximum open probability (G_max/ Q_max). Similar effects on current density have been reported in COS 7 (Stephens et al., 1997) and HEK 293 cells (Williams et al., 1994), but not in Xenopus oocytes (Soong et al., 1993; Olcese et al., 1994, 1996). Furthermore, in a study of α_1E gating currents in Xenopus oocytes (Olcese et al., 1996), coexpression of β_2a with α_1E actually decreased the number of functional channels (Q_max), although surprisingly they found a twofold increase of G_max/Q_max that is qualitatively similar to our result. Additional support for the role of the β subunit in modulating P_o,max comes from a separate study that used fluctuation analysis to determine the effects of β_2a and β_1a on α_1E open probability (Noceti et al., 1996).

Fitting with the doubling of the G_max/Q_max ratio, the β subunit also induced hyperpolarizing shifts of both G-V and the Q-V relations and slightly reduced the gap between the two, all while producing little effect on activation kinetics. Here, the action of β subunits is also somewhat controversial. Although most studies of α_1E report effects on G-V relations and activation kinetics that are similar to ours (Witcher et al., 1993; Olcese et al., 1994; Stephens et al., 1997), in one case (Wakamori et al., 1994), β_1b coexpression with the human α_1E in Xenopus oocytes was found to slow activation kinetics substantially. With respect to gating currents, the only other study of α_1E charge movement (Olcese et al., 1996) also found that the β subunit reduced the gap between the G-V and Q-V. However, in contrast to the small but statistically significant (P < 0.01, Student's t test, comparing cells with and without a β) shift in the Q-V reported here, Olcese et al. (1996) found that the β subunit produced no significant change in the Q-V. Again, these discrepancies may reflect differences between clones (human versus rat α_1E) or expression systems (Xenopus oocytes versus HEK 293 cells). Despite these minor differences, all the results indicate a role for β subunits in modulating activation gating.

In other calcium channels, the reported effects of the β subunits vary even more widely than for the α₂δ subunit. However, at least in some respects, β subunit modulation of other α₁ subunits appears similar to what we find for α_1E. For example, there are reported shifts in the voltage dependence of ionic activation for α_1A (Stea et al., 1994; De Waard and Campbell, 1995) and α_1C (Wei et al., 1991; Neely et al., 1993). Increased current density has also been observed for many of the α₁ subunits including α_1A (Mori et al., 1991), α_1B (Williams et al., 1992a), α_1C (Perez-Reyes et al., 1992), α_1D (Williams et al., 1992b), and α_1S (Ren and Hall, 1997). Furthermore, for α_1C, studies of gating currents in both HEK 293 cells (Kamp et al., 1996; Josephson and Varadi, 1996) and Xenopus oocytes (Neely et al., 1993) find that β subunit modulation of ionic current activation is not associated with much shift in the Q-V, similar to what we find for α_1E. On the other hand, even these few gating current studies disagree in other regards. While coexpression of β_1a (Kamp et al., 1996) or β₃ (Josephson and Varadi, 1996) with α_1C increased both current density and Q_max in HEK 293 cells similar to our results for α_1E, β_2a increased current without changing Q_max in Xenopus oocytes (Neely et al., 1993). Therefore, as with α_1E, the specific effects observed with β coexpression appear to depend on as yet unknown distinctions between expression systems, perhaps the endogenous expression of β_XO subunits in Xenopus oocytes (Tareilus et al., 1997).

There was little isoform dependence to the modulation of all the above measures of activation gating, suggesting that different β subunits act by a similar mechanism to modulate activation and expression of α_1E, despite very different effects on inactivation. While no other study has compared gating currents of α_1E channels containing different β subunits, measurements of α_1E ionic-current G-V curves in Xenopus oocytes support this finding (Olcese et al., 1994). However, for other α₁ subunits, modulation of expression and activation may differ across β subunits. For example, there clearly is isoform specificity in the β subunit modulation of current potentiation in α_1A (Stea et al., 1994; De Waard et al., 1994) and α_1S (Ren and Hall, 1997). This fits with the binding affinity differences in vitro of various β subunits to α_1A (De Waard et al., 1995). However, binding of various β subunits to the I-II linker of α_1B occurs with the same affinity (Scott et al., 1996). Whether differences in in vitro binding affinities translates into discernible gradations of functional effects remains to be established.

Most previous studies have implicitly assumed that only one β subunit is involved in modulating channel properties, consistent with biochemical evidence for a 1:1 stoichiometry of α₁ and β subunits for skeletal (De Waard et al., 1996) and N-type (Witcher et al., 1993) channels. However, a recent report by Tareilus et al. (1997) identified a second β subunit binding site on the COOH terminus of α_1E, raising the possibility that two or more β subunits might collectively determine channel gating properties. Here, we found little evidence that two β subunits modulate the properties of the α_1E channel, either in regard to expression or gating.

To account for β subunit effects on G-V and Q-V curves, previous studies have proposed that the β subunit acts mainly on the weakly voltage-dependent steps that “couple” channel opening to voltage sensor movement (Neely et al., 1993; Olcese et al., 1996). Here, we demonstrate that this mechanism may explain not only the modulation of G-V and Q-V curves, but also the doubling of maximum open probability. Fig. 10 depicts a channel gating model that closely resembles those previously used in the study of potassium channel gating (Zagotta and Aldrich, 1990; Schoppa et al., 1992). There are three independent, voltage-dependent transitions between the closed states (C₀, C₁, C₂, and C₃), each associated with an appropriately scaled equilibrium constant K₀. These transitions are followed by a weakly voltage-dependent transition (C₂–C₃) with equilibrium constant K₁ and a final voltage-independent step with equilibrium constant K₂. K₀ and K₁ are voltage dependent according to a Boltzmann distribution, K_i = exp{[z_iF(V − V_i)]/(RT)}. To obtain baseline model parameters (z₀, z₁, V₀, V₁, K₂), we fit α_1E alone Q-V and G-V data (Fig. 10,B). Then, to simulate the observed twofold change in the maximum open probability, we modified only K₂, the equilibrium constant for the last voltage- independent transition leading to channel opening. This simple change reproduced well both the shift in the G-V relationship and the shift in the Q-V relationship (Fig. 10 C). In fact, such simulations indicate that modifying the coupling of charge movement to channel opening (K₂) usually also perturbs the Q-V relation and, therefore, charge movement. Yet in several studies (Kamp et al., 1996; Josephson and Varadi, 1996; Neely et al., 1993; Olcese et al., 1996), shifts in ionic activation have been seen with little or no modification of charge movement. In these studies, it may be that the shift in charge movement is too small to be well resolved.

To determine the effect of modifying K₂ on the time course of activation, we modeled the kinetics of activation. The choice of rate constants is constrained by the equilibrium constants, according to K_i(V) = α_i(V)/ β_i(V), where α_i(V) is the forward rate constant and β_i(V) is the backward rate constant (seconds⁻¹). We chose α_i(V) = f_i exp{[d_i z_iF(V − V_i)]/(RT)}, and β_i(V) = α_i(V)/K_i(V), giving us free parameters f₀, d₀, f₁, d₁, and f₂, which we could vary to fit the activation kinetics. Fig. 10,D shows a model fit (thin solid line) consistent with α_1E alone steady state model parameters (z₀, z₁, V₀, V₁, K₂; Fig. 10,B). Representative ionic current data (Fig. 10,B, thick gray line) are derived by subtracting the gating currents (Fig. 5,A) from the α_1Eβ_1b whole cell records in Fig. 6,A. Because of the subunit invariance of activation kinetics, these ionic currents also represent the expected time course of α_1E alone. Changing K₂ to accord with the twofold increase in maximal open probability produced by β subunits can modify the activation kinetics (Fig. 10 E, dashed line), but this may be compensated for by altering only f₁ and f₂. This amounts to subtle changes in the absolute rate constants of the last two transitions, but only an alteration of the equilibrium constant of the last transition. Although no change in parameters corresponding to the more voltage-dependent steps (z₀ and f₀) is necessary, we found that we could not well reproduce the invariance of activation kinetics by just modifying f₂, corresponding to the last voltage-independent step. Therefore, this simulation argues that all the effects of β subunit modulation of α_1E (increased open probability, hyperpolarization of G-V and Q-V curves, and invariant activation kinetics) can be attributed to actions on one or a few weakly voltage-dependent steps before opening.

From the previous discussion, it is clear that generalization across α₁ subtypes is a difficult proposition. Nevertheless, we wondered whether some of the most robust properties of α_1E modulation by subunits would translate to a different α₁ subunit. In particular, there was striking adherence to two “rules” for α_1E modulation by subunits: (a) all β subunits produce no change in the kinetics of activation, but induce identical but relatively small hyperpolarizing shifts of the G-V curve, and (b) distinct β subunits impart vastly different steady state inactivation curves. Do these features of subunit modulation hold true as general tenets for other α₁ subunits?

Fig. 11 tests this proposition for the α_1C calcium channel. The results indicate a complete reversal of the behavior found with α_1E. Now the kinetics of activation are clearly different for β₃ and β_2a subunits. In addition, the different β isoforms led to large differences in G-V curves. On the other hand, steady state inactivation curves show only small isoform-dependent distinctions.

The diametrically opposite behaviors exhibited by α_1E and α_1C subunits have interesting implications for the structure–function relations underlying α₁-β modulation. The leading candidates for structural interaction between these two subunits are a small motif on the I-II linker of α₁ subunits known as the “alpha interaction domain” or AID (Pragnell et al., 1994), and another small motif in the middle of β subunits known as the “β interaction domain” or BID (De Waard et al., 1995). AID and BID peptides bind with high affinity (tens of nanomolar), and the BID region has documented importance for modulation of α₁ subunits (De Waard et al., 1995). The key points with regard to our findings are that the BID is highly (∼70%) homologous across β subunits, and the AID is also highly conserved across different α₁ subunits (Pragnell et al., 1994; De Waard et al., 1995). The profound differences in β isoform selectivity for entirely different gating properties, depending on the α₁ subtype, suggest that either the AID–BID interaction is exquisitely sensitive to small sequence variations in the AID (Scott et al., 1996), or there are other features that contribute to β subunit modulation of α₁ (Chien et al., 1996). Distinguishing between these two possibilities and identifying any secondary interaction sites will be important challenges for the future.

We thank K.P. Campbell for the β_1b clone, T.P. Snutch for the α_1E and α₂δ clones, E. Perez-Reyes for the α_1C, β_2a, β₃, and β₄ clones, M. deLeon for construction of α_1EΔ, J.G. Mulle for technical assistance, and David Brody and Carla DeMaria for discussion and comments.

This work was supported by the National Institutes of Health (NIH) to D.T. Yue, the National Science Foundation Presidential Faculty Fellowship (D.T. Yue), a Maryland American Heart Association Postdoctoral Fellowship (S.K. Wei), and an NIH Medical Scientist Training Program Award (L.P. Jones).

I-V

current–voltage