Abstract
We investigated which subtypes of G-protein β subunits participate in voltage-dependent modulation of N-type calcium channels. Calcium currents were recorded from cultured rat superior cervical ganglion neurons injected intranuclearly with DNA encoding five different G-protein β subunits. Gβ1 and Gβ2 strongly mimicked the fast voltage-dependent inhibition of calcium channels produced by many G-protein-coupled receptors. The Gβ5 subunit produced much weaker effects than Gβ1 and Gβ2, whereas Gβ3 and Gβ4 were nearly inactive in these electrophysiological studies. The specificity implied by these results was confirmed and extended using the yeast two-hybrid system to test for protein–protein interactions. Here, Gβ1 or Gβ2 coupled to the GAL4-activation domain interacted strongly with a channel sequence corresponding to the intracellular loop connecting domains I and II of a α1subunit of the class B calcium channel fused to the GAL4 DNA-binding domain. In this assay, the Gβ5 subunit interacted weakly, and Gβ3 and Gβ4 failed to interact. Together, these results suggest that Gβ1 and/or Gβ2 subunits account for most of the voltage-dependent inhibition of N-type calcium channels and that the linker between domains I and II of the calcium channel α1 subunit is a principal receptor for this inhibition.
Inhibition of neuronal calcium currents by G-protein-coupled receptors probably occurs in every type of neuron. It is a ubiquitous mode of modulation of electrical activity and synaptic function by neighboring neurons. A common form uses a signaling pathway whose components appear to be restricted to the cell membrane (membrane-delimited), develops in <1 sec, and is voltage-dependent, i.e., the inhibition can be partially relieved by a strong depolarizing pulse (for review, see Hille, 1994). A well studied example is the inhibition of N-type calcium channels by norepinephrine (NE) acting on α2 adrenergic receptors in superior cervical ganglion (SCG) neurons. This type of inhibition is mediated by βγ-subunits of G-proteins, apparently acting directly on the channel (Herlitze et al., 1996; Ikeda, 1996).
The specificity for Gβ and Gγ subunits has not been investigated, and the target site for the interaction of G-protein βγ subunits with calcium channels is the subject of some debate. By analogy with other Gβγ-effector proteins, several investigators have proposed that the target site on voltage-gated calcium channels includes a QXXER motif in the intracellular loop connecting domains I and II (LI-II) of the α1 subunits of classes A, B, and E (the P/Q-, N-, and R-type) calcium channels (De Waard et al., 1997; Herlitze et al., 1997; Page et al., 1997;Zamponi et al., 1997). This QXXER sequence overlaps with the region required for interaction with the calcium channel β subunit. In contrast, some investigators argue that an interaction of Gβγ subunits with the C terminus of calcium channel α1subunit is the functionally important one (Zhang et al., 1996a; Qin et al., 1997). Here, we inject DNA for various G-protein β subunits into the nucleus of adult rat SCG neurons and find that only certain Gβ subunits will induce inhibition of the endogenous N-type Ca2+ currents. Then, we used the yeast two-hybrid assay to look for interactions between Gβ subunits and a domain of the channel. We find that exactly the same Gβ subunits that inhibit current also interact well with the linker between domains I and II of the α1 subunit of the class B (N-type) calcium channel. Together, these results identify the Gβ subunits that can mediate fast, membrane-delimited, and voltage-dependent inhibition of N-type calcium channels, and they identify a target domain of the channel that has the appropriate binding specificity.
MATERIALS AND METHODS
Cell culture and intranuclear microinjection. Single SCG neurons were enzymatically dissociated from 5-week-old male rats (Sprague Dawley rats were used for the experiments in Figs. 1-3 in Seattle, and Wistar rats were used in the experiments in Figs. 4 and 5in Mexico) as described previously (Beech et al., 1991; Bernheim et al., 1991). After a 4 hr wait for attachment to the substrate, the neurons were intranuclearly microinjected using an Eppendorf 5242 pressure microinjector and 5171 micromanipulator system (Eppendorf, Madison, WI). The injection solution contained varying amounts of G-protein expression plasmids mixed with two injection markers, 1 mg/ml 10,000 kDa dextran–fluorescein (Molecular Probes, Eugene, OR), and green-fluorescent protein (GFP) plasmid as an expression reporter. Injection at pressures of 10–20 kPa for 0.5–0.8 sec resulted in no obvious increase in cell volume. After 12–16 hr, successfully injected neurons were identified by their characteristic greenish-blue GFP fluorescence using an inverted microscope equipped with epifluorescence and fluorescein optics.
Electrophysiological recording. Currents were recorded using the whole-cell voltage-clamp technique (Hamill et al., 1981) with a patch-clamp amplifier at room temperature. Pipettes (0.85–2 MΩ) were pulled from microhematocrit glass and fire polished. During recording, neurons were constantly perfused locally (1–2 ml/min) with the appropriate external solution. Solution reservoirs were selected by means of solenoid valves, and solution changes were accomplished in <10 sec. Voltage protocols were generated, and data were digitized, recorded, and analyzed using BASIC-FASTLAB (Indec Systems, Capitola, CA).
To measure calcium currents, we used two sets of solutions. In the experiments of Figures 1-3, the pipette solution contained (in mm): 125 N-methyl-d-glucamine, 20 TEA-Cl, 10 HEPES, 0.1 tetracesium-BAPTA, 4 MgCl2, 0.1 leupeptin, 4 Na2ATP, and 0.3 Na2GTP, pH adjusted to 7.2 with methanesulfonic acid. The external solution contained (in mm): 140 TEA-OH, 10 HEPES, 2 CaCl2, 15 glucose, 0.0001 TTX, and 0.002 nifedipine, pH adjusted to 7.3 with methanesulfonic acid. In the experiments of Figures 4 and 5, the pipette solution contained (in mm): 125 methanesulfonic acid, 20 TEA-Cl, 10 HEPES, 0.1 tetracesium-BAPTA, 4 MgCl2, 5 MgATP, 0.3 Na2GTP, and 0.1 leupeptin, pH adjusted to 7.2 with CsOH. The external solution contained (in mm): 162.5 TEA-Cl, 10 HEPES, 8 glucose, 2 CaCl2, 1 MgCl2, 0.0002 TTX, and 0.002 nifedipine, pH adjusted to 7.3 with TEA-OH.
The Ca2+ current of rat SCG neurons is carried ∼85–90% in N-type channels and the remainder in L-type channels, with no detectable P/Q component (Mintz et al., 1992). Therefore, the N-type Ca2+ current could be defined as the component of the current sensitive to 100 μmCd2+ in the presence of 2 μmnifedipine. Currents were sampled at 10 kHz. To emphasize the effects of kinetic slowing of Ca2+ current activation, current amplitudes were always taken as the mean value of recorded points between 5 and 6 msec after the start of the depolarizing test pulse. This time is before the current reaches a peak. Because the magnitude of the Ca2+ current was dependent on cell size, aggregate current data are presented as current densities normalized to cell capacitance. To avoid one source of systematic bias, experimental and control measurements were alternated whenever possible, and concurrent controls were always performed. Where appropriate, data are expressed as mean ± SE.
Plasmids and materials. The DNAs encoding Gβ2, Gβ4, and Gβ5 were cloned in pcDNA I plasmid, Gβ1 was cloned in pCDM8, Gβ3 was cloned in pCIS, Gγ3 was cloned in pCI (all from M. Simon, Caltech, Pasadena, CA), and GFP was cloned in pEGFP-N1 (Clontech, Palo Alto, CA). Although not in identical vectors, all G-protein-containing plasmids were driven by the cytomegalovirus promoter. DNA encoding the LI-II loop of rat brain N-type calcium channel (rbB-I) α1B was subcloned into the pAS2–1 vector (Clontech) as an EcoRI–BamHI fragment (nt 1069–1449; aa 357–483) to express a GAL4 DNA-binding domain (GBD) hybrid protein. The bovine G-protein β1 subunit and β2 subunit genes were subcloned into the pACT vector (Clontech) to express GAL4-activation domain (GAD) hybrid proteins. The plasmids pACT-Gβ3, pACT-Gβ4, and pACT-Gβ5 were described previously (Yan et al., 1996). Plasmids were purified using commercial kits (Qiagen, Valencia, CA). The yeast strain was PJ69–4A (MAT a trp1–901 leu2–3,112 ura 3–52 his3–200 gal4Δ gal 80Δ GAL2-ADE2 LYS2:: GAL-HIS3 met2:: GAL7-lacZ) (a gift of Dr. Philip James, University of Wisconsin) (James et al., 1996). This strain contains three reporters (ADE2, HIS3, andLacZ) that can be activated as a result of protein–protein interaction by reconstituted GAL4 transactivator function.
Tetracesium-BAPTA was obtained from Molecular Probes, and all other salts were obtained from Sigma (St. Louis, MO).
Yeast two-hybrid assay. Plasmids were made expressing hybrid proteins consisting either of a specific Gβ subunit coupled to a GAD or an intracellular loop of the α1 N-type calcium channel fused to a GBD. A plasmid expressing one of the GBD hybrids and the plasmid expressing the GAD hybrid were cotransformed into yeast strain PJ69–4A (usually without introducing any mammalian Gγ subunit). Cells were first plated on a tryptophan-free leucine-free medium to select Trp+ Leu+ transformed cells containing the GBD hybrid- and GAD hybrid-expressing plasmids. To detect protein–protein interactions, these transformants were then transferred onto medium that also lacked adenine and histidine.
Immunoblotting. To check the expression of G-protein subunit hybrids in yeast, transformed yeast cells were grown to saturation phase. Cultures (1.5 ml) were harvested and resuspended in 150 μl of Z buffer (in mm: 30 Na2HPO4, 35 NaHPO4, 10 KCl, and 0.4 MgSO4, pH 7.0) with 5% glycerol, 0.5 mm dithiothreitol, 1 μg/ml aprotinin, and 2 μg/ml leupeptin, followed by addition of 0.1 gm of 425–600 μm glass beads (Sigma). The cells were vortexed for 4 min and centrifuged for 10 min at 4°C. The supernatants were saved, and 20 μl cell lysates were fractionated on a SDS gel. The separated proteins were electrophoretically transferred to a nitrocellulose membrane, and the blot was probed with anti-GAD antibody (0.2 μg/ml; Clontech), followed by horseradish peroxidase-conjugated anti-mouse IgG (1:2000; Amersham, Piscataway, NJ). The GAD hybrid proteins were detected by the enhanced chemiluminescence method (Amersham).
RESULTS
We begin with the typical NE-induced modulation of N-type calcium currents. In control SCG neurons, a 20 msec depolarizing test pulse elicited a rapidly activating inward Ca2+ current (Fig. 1A,left), which was primarily carried in N-type channels because the medium contained 2 μm nifedipine to block L-type channels. Perfusion with 10 μm NE reduced the inward current amplitude by ∼60% and slowed its rate of activation. To monitor current facilitation, as well as amplitude, we clamped the membrane potential using the voltage protocol in Figure1A. Inward Ca2+ currents were evoked every 10 sec with a pair of 20 msec depolarizing test pulses to +10 mV from a holding potential of −80 mV, one before (test pulse 1) and the other after (test pulse 2) a 25 msec prepulse to +125 mV. The depolarizing prepulse transiently relieves much of the NE-induced inhibition, and the resulting facilitation can be seen by comparing currents in test pulse 2 with those in test pulse 1. The “facilitation ratio,” defined as the current during test pulse 2 (at 5–6 msec) divided by the current during test pulse 1, is a convenient measure of voltage-dependent membrane-delimited inhibition by certain G-proteins. In standard Ringer’s solution, this facilitation ratio is near 1.1, and after treatment with NE, it rises to 1.6.
Injecting Gβ2 DNA increases calcium current facilitation. A–E, Superimposed calcium current (ICa) tracesduring a 20 msec depolarization (test pulse) to +10 mV from a holding potential of −80 mV, in the absence (bottom trace) or presence (top trace) of NE (10 μm), before (left) or after (right) a 25 msec +125 mV prepulse. Successive panels show uninjected cells (A) and cells injected with increasing concentrations of Gβ2 DNA (B–E). Gβ2 DNA was coinjected with 100 ng/μl Gγ3DNA in all cases. F, Summary of facilitation ratios for injections of several concentrations of Gβ2 and Gβ3 DNA.
Dose-dependent suppression of ICa by overexpression of Gβ2
Previously, we had found that injection of RNA for Gβ2 subunits, with and without RNA for Gγ3, into the cytoplasm of SCG neurons inhibited Ca2+ currents, increased the facilitation ratio, and partially occluded the actions of NE (Herlitze et al., 1996). Ikeda (1996) reported even stronger effects with intranuclear injection of DNA. To optimize conditions for intranuclear injection here, we first determined the dose–response relationship for the action of Gβ2 by varying the concentration of Gβ2 DNA injected, which was always coinjected with 100 ng/μl Gγ3 and 100 ng/μl GFP DNA. Injection of Gβ2 plasmid at 10, 20, and 100 ng/μl induced progressively larger facilitation and kinetic slowing of activation in standard Ringer’s solution (Fig. 1B–D). Injection at 600 ng/μl (Fig. 1E) gave about the same facilitation as 100 ng/μl. Figure 1F shows the dose–response relationship for development of facilitation after these injections of DNA for Gβ2 compared with that after injection of Gβ3. Quite clearly, Gβ2 has a much greater effect than Gβ3. Injection with 100 or 500 ng/μl Gβ3 DNA solution did not induce facilitation (Fig. 1F); indeed, injection of Gβ3slightly reduced facilitation below control levels (facilitation ratio, 1.03 ± 0.02; n = 10; p < 0.05).
The current traces after injection of Gβ2 show that the progressive increase of facilitation was accompanied by a decreasing effectiveness of NE and a decrease in basal calcium current density. The increasing amounts of expressed Gβ2 subunits presumably occlude the actions of NE and block N-type calcium channels. The facilitation ratio is consistently a more sensitive indicator of Gβ2 actions than the occlusion of NE action or the basal current density. From the dose–response curves for these three actions (Fig. 2), we elected to use 100 ng/μl Gβ plasmids in the next experiments.
Concentration–response relationships for facilitation ratio, inhibition by NE, and ICa density. Gβ2 DNA concentrations ([β2]) are plotted on a logarithmic scale. Data were fitted to a Hill function, y = [(A1-A2)/{1+([β2]/[β2]0)n}] + A2, with best fitting values of the midpoints, [β2]0 = 20.0, 156, and 103 ng/μl (vertical solid line); Hill coefficients,n = 2.0, 1.2, and 1.51; low-concentration asymptotes, A1 = 0.94, 57, and 12.7; and high-concentration asymptotes, A2 = 2.4, 1.9, and 1.4 for A, B, andC, respectively. The horizontal dashed line in A indicates the facilitation ratio for control cells (1.1), and the vertical dashed lineindicates a theoretical injected Gβ2 DNA concentration (7.0 ng/μl) that would be equivalent to this basal facilitation ratio.
The dashed arrow in Figure 2Arepresents an extrapolated Gβ2 DNA concentration (7 ng/μl) that would be equivalent to the control basal facilitation levels, assuming that the facilitation ratio in the total absence of endogenous Gβ would be the fitted value 0.94.
Identification of the Gβ subtypes that modulate Ca2+ currents
The apparent inability of Gβ3 to produce facilitation prompted us to undertake a more systematic study of the effectiveness of the various G-protein β subunits in calcium channel inhibition. As before, we measured facilitation, tonic inhibition, and occlusion of further inhibition by NE. Again in control cells, application of 10 μm NE strongly reduced the Ca2+ current magnitude (Fig.3A, left). Figure3, B and D, illustrates Ca2+currents recorded from neurons that had been injected with 100 ng/μl DNA for Gβ1, Gβ2, or Gβ5, with GFP DNA but without coinjection of a Gγ DNA. In the absence of agonist, the Ca2+currents in neurons injected with DNAs for Gβ1 or Gβ2 displayed kinetic slowing (Fig. 3, left traces of each pair) and facilitation (right traces) virtually identical to that produced by application of NE in control cells. Coinjection with Gγ3 DNA did not change the results for Gβ1 (data not shown) or Gβ2 (Fig. 1). Only a small effect was observed in cells injected with Gβ5 DNA, whether alone (Fig. 3D) or in combination with Gγ3 DNA (data not shown). Finally, in neurons injected with DNAs for Gβ3γ3, Gβ4γ3, or CB1λ (a truncated, nonmembrane spanning form of the rat cannabinoid receptor in the pcDNA3 expression vector), the Ca2+ current activation was similar to that of uninjected neurons (data not shown).
Identification of the Gβ subtypes that cause voltage-dependent modulation of Ca currents. Cells were injected with DNA for G-protein β and γ subunits, as indicated, and currents were recorded by the double-pulse protocol. A–D, Current records for uninjected control cell (A) and neurons injected with 100 ng/μl Gβ1, Gβ2, or Gβ5 DNA (B–D). E, Summary data for facilitation ratios; numbers above individualcolumns represent the number of tested cells, and the same numbers apply to F and G. Data plotted as mean ± SEM. *p < 0.05; two-tailedt test versus control. F, Percent ofICa inhibited by NE (during test pulse 1).G, ICa density during test pulse 1 in the absence of NE.
The quantitative results of these DNA injections for the various Gβ subunits are summarized in Figure 3, E–G. The Gβ1 or Gβ2 injections with and without Gγ3 give indistinguishable, strong, and highly significant effects when compared with cells injected with the control plasmid CB1λ or to uninjected cells. In these Gβ1- or β2-injected neurons, the facilitation ratio is raised from a mean of 1.1 to 2.3, the inhibition by NE is lowered from 60 to 22%, and the basal calcium current density is lowered from 22 to 9 pA/pF. These actions are significant at the p < 0.005 level. Interestingly, both the depression of current and the increase in facilitation ratio are greater than those which occur with 10 μm NE on control cells. By comparison, the effects of injecting DNA for Gβ3, Gβ4, or Gβ5 are weak or negligible, even in the presence of Gγ3. Of these, Gβ5 appears the most active. Our results suggest that the different Gβ subtypes have different efficacies in voltage-dependent inhibition of Ca2+channels, with a rank order of effectiveness: Gβ1 = Gβ2 > Gβ5 ≫ Gβ4 = Gβ3.
Characterization of the weakly active Gβ subtypes
We wanted to explore more fully the reasons for the weak effects of Gβ3, β4, and β5 DNA injections. Did the plasmids fail to induce proper synthesis of the corresponding proteins? Were the proteins really inactive on Ca2+ channels? Because convenient antibodies for verifying expression of specific Gβ subunits in single injected cells are lacking, we chose less direct approaches. We looked with increased sensitivity for electrophysiological effects on Ca2+ channels.
We made two changes in the electrophysiological assay. One was in the concentration of GFP DNA coinjected into the cells. In some systems, it has been found that the overexpression of GFP or other proteins from one plasmid leads to less expression from a second plasmid. Thus, the most brightly fluorescing cells express lower levels of protein from the second plasmid (N. Davidson, personal communication). Therefore, we lowered the concentration of GFP DNA in the injection solution from 100 to 10 ng/μl in the next experiments. A second factor that may have limited our sensitivity to detect occlusion of the NE response is the high concentration of NE we had used (10 μm) in Figures 1, 2, and 3. This concentration is sufficiently above that needed to have a maximal NE action that a small shift of the sensitivity to NE could have been obscured. Therefore, we reduced the NE concentration to 2 μm, which is submaximal.
Figure 4, A–D, illustrates Ca2+ currents recorded from an uninjected neuron and from neurons injected with Gβ3, Gβ4, and Gβ5 DNAs. In uninjected neurons, 2 μm NE inhibited the Ca2+current by 41 ± 4% (SEM; n = 6), compared with the 55% inhibition obtained with saturating NE (10 μm) (Figs. 1-3). There were statistically significant changes in at least one of the experimental parameters for each injected neuron, suggesting expression of each Gβ subunit. Facilitation was increased in the neurons injected with Gβ5; the inhibition by 2 μm NE was reduced by injection of Gβ4 or Gβ5, and the calcium current density was almost doubled by injection of Gβ3. Further evidence that injection of Gβ3 DNA affects cellular responses is seen in Figure 5, in which neurons were injected with Gβ1 alone, Gβ3 alone, or a mixture of Gβ1 and Gβ3 DNAs. As we have seen before, the facilitation ratio is strongly increased by Gβ1 and not increased by Gβ3; however, coinjection of Gβ3 with the Gβ1gives significantly less facilitation than with Gβ1alone, consistent, for example, with competition between the Gβ subunits for formation of active Gβγ complexes. These results show that Gβ3 and Gβ4 DNA injections do have specific electrophysiological effects, even if they do not mimic actions of NE on Ca2+ currents. Therefore, these Gβ subunits are expressed at functionally significant levels in our injected neurons.
Three weakly active Gβ subtypes.A–D, Current records for uninjected control cell (A) and neurons injected with 100 ng/μl Gβ3, Gβ4, or Gβ5 DNA (B–D). To accentuate possible weak actions by Gβ3, Gβ4, and Gβ5, the concentration of the reporter plasmid (pEGFP-N1) was decreased tenfold to 10 ng/μl, and NE was decreased to 2 μm.E–G, Summary data for facilitation ratio, inhibition by NE, and current density in these cells. Note that the axes have been expanded to reveal smaller differences.
Coinjection of Gβ1 and Gβ3 DNAs. A–C, Currenttraces for cells were injected with 100 ng/μl Gβ1 (A) or Gβ3(B) or with 100 ng/μl Gβ1 plus 100 ng/μl Gβ3 DNAs (C). Same experimental conditions as in Figure 3. D, Mean facilitation ratios for these cells. * indicates data significantly different from control. ** indicates data for Gβ1 plus Gβ3 significantly different from control, Gβ1, or Gβ3.
Gβ1 and Gβ2 interact strongly with the α1 subunit of N-type calcium channels
It has been demonstrated that LI-II of the α1 subunits of classes A, B, and E (P/Q-, N-, and R-type) calcium channels interact directly with the Gβ1 subunit (De Waard et al., 1997; Zamponi et al., 1997) and that this loop may mediate the inhibitory modulation by the activated G-protein (Herlitze et al., 1997; Page et al., 1997). Our results presented above suggest that this interaction is not the same for all of the Gβ subunits. We sought to investigate the mechanism underlying the differential effects of the five different Gβ subunit subtypes by testing the possibility that they interact with the channel with different affinities. We examined the interaction of LI-IIof α1B with the Gβ subunits using the two-hybrid assay, an assay that can detect protein–protein interactions in vivo in yeast (Fields and Song, 1989).
Plasmids encoding the GBD-LI-II hybrid and the various GAD-Gβ hybrids were introduced into a yeast reporter strain in which interactions between two hybrid proteins result in transcriptional activation of two reporter genes, HIS3 and ADE2. These two reporter gene products are involved in histidine and adenine synthesis, respectively, and their activation allows the yeast cells to grow on medium lacking histidine and adenine. As shown in Figure6A, coexpression of LI-II with Gβ1 or Gβ2 results in robust growth of the yeast cells on the selection medium, whereas with Gβ5, the cells grow at a much slower rate. In the case of Gβ3 and Gβ4, no transformed yeast cells were able to grow in the absence of histidine and adenine. The same results were obtained in five experiments. Because transcription signals resulting from the protein–protein interactions in the two-hybrid assay generally correlate with the binding affinity observed from in vitro experiments (Li and Fields, 1993; Yan et al., 1996), our data suggest that the order of affinity for LI-II of α1B is Gβ1 = Gβ2 > Gβ5 ≫ Gβ3, Gβ4.
Gβ1 and Gβ2 interact strongly with the α1 subunit of N-type calcium channels.A, Yeast reporter strain PJ69–4A was cotransformed with plasmids encoding GBD-LI-II and the indicated GAD-Gβ subunits. Cells were first plated onto synthetic sucrose medium SS/−Trp/−Leu to select cotransformants. The cotransformants were then replica-plated onto medium SS/−Trp/−Leu/−Ade/−His to select for colonies having successful GBD–GAD interactions.B, Expression of GAD-Gβ hybrid proteins in yeast. Yeast lysates were subjected to SDS–PAGE and immunoblotting. The blot was probed with anti-GAD monoclonal antibody and horseradish peroxidase-conjugated to anti-mouse IgG. The proteins were detected by ECL. Yeast lysate without GAD hybrid proteins was loaded in lane 6 as a negative control. Total protein amounts in the 20 μl lysates quantified by the Bradford assay (Bio-Rad, Hercules, CA) were 1.94, 1.30, 0.72, 0.80, and 1.70 μg in lanes 1–5, respectively. Intensity of individual GAD-Gβ bands on the immunoblot was determined by densitometry (Molecular Dynamics, Sunnyvale, CA), normalized to the protein amount in their respective lysates, and plotted as a percent of GAD-Gβ3 to show the relative levels of these hybrid proteins.
To exclude the possibility that the tighter binding between Gβ1, Gβ2, and LI-II is attributable to their more efficient recruitment of endogenous yeast Gγ subunits, we also performed the assay in the presence of overexpressed rat Gγ3 subunit. The results obtained were the same (data not shown), suggesting that Gγ subunits are not a determinant of the interaction between Gβ and LI-II in our assay. Hence, activation of theHIS3 and ADE2 genes in the assay might be the result of direct interactions between Gβ1, Gβ2, and Gβ5 with the calcium channel LI-II loop region.
Because an apparent differential affinity of Gβ subtypes for LI-II of α1B observed in the two-hybrid experiments could also be attributable to differences of their protein level in the yeast cell, we examined their expression by immunoblotting. This method measures the level of expressed protein but does not test correct folding or association with Gγ subunits. Immunoblots of the yeast cell lysates with anti-GAD antibody showed these GAD-hybrid proteins are expressed at similar levels (Fig.6B, lanes 1–5). Densitometric scanning and normalization to the amount of total cellular protein in each extract showed that expression of Gβ3, Gβ4, and Gβ5 was higher than expression of Gβ1 and Gβ2. A second experiment gave the same result. These results support our conclusion that, in the yeast two-hybrid assay, Gβ1 and Gβ2 are more effective because they bind with higher affinity to LI-II.
DISCUSSION
Our electrophysiological and protein–protein interaction experiments give concordant results for the β subunits of G-proteins. All Gβ subunits are not equal. Mimicry and occlusion of membrane-delimited voltage-dependent inhibition of N-type Ca2+ currents and interaction with the LI-II loop of αlB (N-type) calcium channels follow the sequence Gβ1 = Gβ2 > Gβ5 ≫ Gβ3, Gβ4. The agreement of the two approaches adds weight to the proposal that the functional target of voltage-dependent Gβ interaction includes the LI-II loop (De Waard et al., 1997; Herlitze et al., 1997; Page et al., 1997; Zamponi et al., 1997). The activity of overexpressed Gβ1 and Gβ2 has been documented before in SCG cells (Herlitze et al., 1996; Ikeda, 1996;Delmas et al., 1998). Unlike our results (Fig. 1), however, Ikeda (1996) and Delmas et al. (1998), who studied only Gβ1, found that a Gγ subunit had to be coexpressed with Gβ1 to have an effect on calcium channels. We do not know any reason for this clear difference from our work. They tested only Gγ2, and we tested only Gγ3. If Gβγ dimers are needed for mimicking action of NE, the Gβ subunits must have combined with endogenous Gγ subunits when we did not inject Gγ DNA. The endogenous complement of Gγ subunits is not known for SCG neurons.
Although the vectors for Gβ3 and Gβ4 did not mimic voltage-dependent Ca2+ current inhibition by NE, they did alter electrophysiological properties of the cells. Thus, they were expressed. Unexpectedly, expression of Gβ3 led to a near doubling of the Ca2+current density (Fig. 4). Because the percent of inhibition by NE was unchanged in these cells, the increased current density was primarily caused by an increase in N-type Ca2+ current. On the other hand, expression of Gβ4 led to a reduction in modulation by NE, without an increase in facilitation or a decrease in current density. These effects of Gβ3 and Gβ4 might have been caused by interactions with signaling systems in the cell that alter gene expression or act in other ways on the channel. They could also represent direct interactions of Gβ3 and Gβ4 with parts of the channel that do not have the binding specificity we found for the LI-IIloop. It is not likely that Gβ3 or Gβ4failed to pair with coinjected Gγ3 subunit, because it has been found previously that this subunit pairs efficiently with the Gβ3 and Gβ4 subunits (Yan et al., 1996). Indeed, in comparison with Gβ2γ3, the reporter activity in a yeast-two-hybrid assay was 55% for Gβ3γ3, 72% for Gβ4γ3, and 83% for Gβ5γ3 (Yan et al., 1996).
The literature does not contain much information on specificity of the actions of Gβ subunits. From our results, the strong membrane-delimited voltage-dependent modulation of N-type calcium channels seen in many neurons is almost certainly mediated by Gβ1 and/or Gβ2 subunits. A similar conclusion may be drawn for the activation of G-protein-coupled inward rectifier K+ (GIRK) channels. A two-hybrid screen with residues 1–83 of the GIRK1 channel showed interaction with Gβ1 and Gβ2 but not with Gβ3, Gβ4, or Gβ5 (Yan and Gautam, 1996), suggesting that Gβ1 and Gβ2 may be generally involved in membrane-delimited modulation of ion channels. Functional tests of Gβ3, Gβ4, or Gβ5 on GIRK channels have not been published.
Specificity has been investigated in a few other systems. Kleuss et al. (1991, 1992, 1993) used injection of antisense oligonucleotides directed against specific members of the G-protein heterotrimer to study agonist-induced inhibition of L-type calcium channels in GH3 cells. They established that Gαo1/β1/γ4mediates inhibition acting via m4 muscarinic receptors, whereas Gαo2/β3/γ3mediates inhibition by somatostatin receptors. The molecular mechanisms of these signaling pathways have not been worked out, and it is possible that the very high specificity lies at the level of the receptors rather than the (unknown) effector(s). This contrasts with work on adenylyl cyclase II in the two-hybrid system in which all β subunits interact to some degree, and the rank order of interaction was β1 > β2 > β3 = β5 > β4 (Yan and Gautam, 1996). Interestingly, β1 stimulates, whereas β5inhibits adenylyl cyclase II (Bayewitch, 1998). Both interactions were strong in the two-hybrid system.
Other systems in which G-protein subtype specificity has been examined include inhibition of adenylyl cyclase I (no difference between β1 and β5; Bayewitch, 1998), activation of MAP kinase (β1 more efficient than β5; Zhang et al., 1996b), activation of phospholipase β2 (no difference between β1and β5; Zhang et al., 1996b), and binding to G-protein receptor kinases (GRKs) (β1 and β2, but not β3, bind to GRK2, and β1, β2, and β3 all bind to GRK3; Daaka, 1997). Thus, in the case of classical G-protein effectors (adenylyl cyclase, phospholipase, and ion conducting channels) specificity of interactions with G-protein β subunit types has previously been noted only in β5compared with β1. However, this is not entirely surprising considering that β5 is a unique β subunit type that is only 53% identical to the rest and is primarily expressed in brain (Watson et al., 1994). The differential activity of β1 and β2 subunits compared with the β3 and β4 subunits seen here on Ca2+ channels is of interest, because these β subunit types are very similar. In fact, β1 and β2are more similar to β4 (∼90% identity) than β3 to β4 (∼80% identity) (Yan et al., 1996). The functional differences must, however, be caused by structural difference(s) between the two pairs of proteins. A scan of the sequences to identify residues common to Gβ1 and β2, but different in β3 and β4, yields seven residues: R19, S31, N35, P39, A193, R197, and A305, using Gβ1 numbering. Most of these residues are near regions thought to interact with the Gγ subunit. The molecular basis for the functional difference between β1/β2 versus β3/β4 may not be restricted to these residues, because β3 and β4 diverge individually with respect to the each other, as well as β1 and β2 subtypes at many positions in their amino acid sequences. This divergence is also consistent with the differential effects of β3 compared with β4on Ca2+ channel properties (Fig. 4). A recent major study by Ford et al. (1998) has looked for residues in Gβ1 that affect interaction with five different effectors, including N-type calcium channels containing α1B subunits. They mutated 15 residues within the interaction domain for Gβ subunits with Gα subunits and found eight that affected the calcium channel facilitation. However, because all of the 15 residues tested are identical among Gβ1, Gβ2, Gβ3, and Gβ4, these results do not help to explain why two of the subunits are so efficacious and two of them are ineffective on N-type calcium channels.
As mentioned before, the Gβ5 subunit stands out among Gβ subunits. Unlike the others, it is reported to couple almost exclusively to the Gαq subunit in a manner that can prevent the others from binding (Fletcher et al., 1998). In our experiments, injection of the vector encoding Gβ5produced a small but reliable voltage-dependent inhibition of the Ca2+ current and partial occlusion of NE-induced inhibition (Fig. 4). Using a Gβ5 antibody (provided by M. Simon, Caltech) (Watson et al., 1994), we found that SCG neurons express Gβ5 (data not shown), as anticipated from its enrichment in brain, and therefore stimulation of Gαq via m1 muscarinic or angiotensin II receptors in SCG cells should release Gβ5 subunits. Together, these three results suggest that the small but fast component of calcium channel inhibition seen with receptors linked to Gαq in SCG neurons (such as m1 muscarinic receptors) (Zhou et al., 1997) could be attributable to Gβ5.
The putative sites of interaction of the G-protein β subunit with the calcium channel have been controversial (e.g., Dolphin, 1998). Typically, experiments examining this interaction have relied on overexpression of appropriately engineered calcium channels in cell lines or Xenopus oocytes or on the demonstration of protein–protein interactions with recombinant proteins in vitro. We have taken a complementary approach with the two-hybrid system, and our results strongly support the hypothesis that the interaction of Gβ subunits with the linker between homologous domains I and II of the α1 subunit determines the subunit specificity for voltage-dependent inhibition of N-type calcium channels by G-protein βγ subunits. Although it is likely that G-protein β subunits interact with other regions of the calcium channel, further experimentation will be necessary to determine whether these additional interactions contribute to the voltage-dependent inhibition.
Footnotes
This work was supported by the W. M. Keck Foundation, an Alexander von Humboldt Stiftung Fellowship, DGAPA Universidad Nacional Autonoma de Mexico and Consejo Nacional de Ciencia y Tecnologia, and National Institutes of Health Grants NS01588, DA08934, DA00286, NS08174, NS22625, and GM46963. N.G. is an Established Investigator of the American Heart Association. We thank D. Anderson, S. Brown, and L. Miller for technical help.
Correspondence should be addressed to Dr. Bertil Hille, Department of Physiology and Biophysics, Box 357290, University of Washington, Seattle, WA 98195-7290.
Drs. Li, Garcı́a-Ferreiro, and Hernández-Ochoa contributed equally to this work.
