Differential regulation of G protein-gated inwardly rectifying K+ channel kinetics by distinct domains of RGS8


  • Author's present address S.-W. Jeong: Department of Physiology, Wonju College of Medicine, Yonsei University, Wonju, Kangwon-Do, Korea.

Corresponding author S. R. Ikeda: Laboratory of Molecular Physiology, Guthrie Research Institute, One Guthrie Square, Sayre, PA 18840 USA.
Email: sikeda@inet.guthrie.org


  • 1The contribution of endogenous regulators of G protein signalling (RGS) proteins to G protein modulated inwardly rectifying K+ channel (GIRK) activation/deactivation was examined by expressing mutants of GαoA insensitive to both pertussis toxin (PTX) and RGS proteins in rat sympathetic neurons.
  • 2GIRK channel modulation was reconstituted in PTX-treated rat sympathetic neurons following heterologous expression of G protein subunits. Under these conditions, noradrenaline-evoked GIRK channel currents displayed: (1) a prominent lag phase preceding activation, (2) retarded activation and deactivation kinetics, and (3) a lack of acute desensitization.
  • 3Unexpectedly, heterologous expression of RGS8 in neurons expressing PTX-i-RGS-insensitive GαoA shortened the lag phase and restored rapid activation, but retarded the deactivation phase further. These effects were found to arise from the N-terminus, but not the core domain, of RGS8 thus suggesting actions on channel modulation independently of GTPase acceleration.
  • 4These findings indicate that different domains of RGS8 make distinct contributions to the temporal regulation of GIRK channels. The RGS8 core domain accelerates termination of the G-protein cycle presumably by increasing Gα GTPase activity. In contrast, the N-terminal domain of RGS8 appears to promote entry into the G protein cycle, possibly by enhancing coupling of receptors to the G protein heterotrimer. Together, these opposing effects should allow for an increase in temporal fidelity without a dramatic decrease in signal strength.

Regulators of G protein signalling (RGS) proteins comprise a growing family of proteins containing a highly conserved core domain of approximately 120 amino acids flanked by less well-conserved regions (Dohlman & Thorner, 1997; Koelle, 1997; Berman & Gilman, 1998; Hepler, 1999; De Vries & Farquhar, 1999; Ross & Wilkie, 2000). RGS proteins accelerate the intrinsic GTPase activity of Gα subunits thereby functioning as GTPase-activating proteins (GAPs) for heterotrimeric G proteins. In addition, RGS proteins can act as a competitive antagonist by preventing the binding of Gα GTP to effector molecules (Berman & Gilman, 1998). The crystal structure of RGS4 complexed with Gαi1 GDP-AlF4 reveals that the RGS core domain interacts with ‘switch’ regions closely associated with guanine nucleotide binding and GTP hydrolysis (Tesmer et al. 1997). Unlike the RGS core domain, the regions flanking the core domain were disordered and thus were neither resolved nor predicted to be involved in RGS binding to Gα. However, comparison of the primary sequence of multiple RGS protein isoforms reveals that non-core domains contain diverse structural and functional motifs that may subserve novel functions (De Vries & Farquhar, 1999; Hepler, 1999; Zheng et al. 1999).

To date, our understanding of the physiological roles subserved by endogenous RGS proteins is largely indirect having been extrapolated from in vitro biochemical assays and heterologous overexpression experiments. In this study, we examined the contribution of endogenous RGS proteins by reconstituting G protein coupling in sympathetic neurons with a Gα subunit containing two sets of point mutations. The first mutation, a cysteine to glycine switch at position -4 from the C-terminus (C351G), renders the GαoA subunit insensitive to pertussis toxin (PTX) treatment (Jeong & Ikeda, 2000a). The second set of mutations were designed to disrupt the interaction of RGS proteins with Gα subunits (DiBello et al. 1998; Lan et al. 1998; Natochin & Artemyev, 1998). By introducing both sets of mutations into a single Gα subunit, we generated PTX- and RGS-insensitive (PTX-i-RGS-i) GαoA mutants and subsequently used this technique to probe the contribution of endogenous RGS proteins to α2-adrenergic receptor-mediated modulation of N-type Ca2+ channels (Jeong & Ikeda, 2000b). Recently, Chen & Lambert (2000) used a similar technique to determine the role of endogenous RGS proteins in hippocampal neuron presynaptic inhibition. Here, we extend this strategy to the investigation of GIRK-type K+ channel activation. The primary impetus for extending our studies was the higher temporal resolution with which G protein action could be studied using GIRK channels as a readout. As GIRK channels are not natively expressed in sympathetic neurons (Ruiz-Velasco & Ikeda, 1998; Fernandez-Fernandez et al. 1999), GIRK1 (Kir3.1) and GIRK4 (Kir3.4) subunits were expressed along with the PTX-i-RGS-i GαoA mutants and Gβ1γ2 by intranuclear injection of plasmids coding for the desired protein. Under these conditions, G protein coupling via PTX-i-RGS-i GαoA mutants significantly retarded GIRK activation and deactivation kinetics. Unexpectedly, heterologous expression of RGS8 in PTX-i-RGS-i GαoA mutant-expressing neurons accelerated the GIRK activation phase while further retarding the deactivation phase. This activity was found to arise from non-core domains of RGS8. Together, these findings provide evidence that different domains of RGS proteins make distinct contributions to the temporal regulation of GIRK channels.


Dissociation of sympathetic neurons

Adult rats were anaesthetised with carbon dioxide and killed using a laboratory guillotine. All procedures were approved by the Institutional Animal Care and Use Committee (IACUC) and determined to meet NIH guidelines for the care and use of animals. Superior cervical ganglion (SCG) neurons were enzymatically dissociated as described previously (Ikeda, 1997). The dissociated neurons were plated onto polystyrene culture dishes (35 mm) coated with poly-l-lysine and maintained in a humidified atmosphere of 95 % air-5 % CO2 at 37 °C. All neurons were used within 24 h following intranuclear injection of vectors. As appropriate, neurons were incubated overnight (16-20 h) with 500 ng ml−1 pertussis toxin (List Biological Laboratories, Campbell, CA, USA).

Preparation and expression of vectors

Vectors encoding human GIRK1 and GIRK4 subunits (D. E. Logothetis, Mount Sinai Medical School, New York, NY, USA), wild-type of mouse GαoA, bovine Gβ1 and Gγ2 (M. I. Simon, California Institute of Technology, Pasadena, CA, USA), and EGFP-fused RGS8 core domain and ΔRGS8 (B. A. Adams, Utah State University, Logan, UT, USA; see Melliti et al. 1999) were generous gifts from the indicated investigators. To generate a PTX-insensitive mutant of GαoA (PTX-i GαoA), the codon specifying the fourth amino acid, cysteine (C), from the carboxyl terminus was mutated to code for a glycine (G) (C351G; Jeong & Ikeda, 2000b). Using GαoA(C351G) as a template, the G184S and S207D mutations were introduced using GeneEditor (Promega, Madison, WI, USA) site-directed mutagenesis kit according to instructions of the manufacturer (for details, see Jeong & Ikeda, 2000b). The cDNA encoding rat RGS8 was cloned using the PCR and a high fidelity polymerase (Pfu, Stratagene, La Jolla, CA, USA) from a commercial rat brain cDNA preparation (Clontech) using the following oligonucleotide primers: sense, 5′-ACTGGAATTCCTGTATGGCTGCCTTACTGATGCCAC-3′; antisense, 5′-CAGTGATATCAAGAGTCCCCATCTGAGGTCTAGCTG-3′. The resulting product was ligated into the mammalian expression vector, pCI (Promega). The cDNA encoding the N-terminus of RGS8 (residues 1-45) was amplified by the PCR from the RGS8 cDNA using the same sense primer as described above and the following antisense primer 5′-AATGGTACCAATCTCTTGAGAGCGCGGTTG-3′. The resulting insert was cloned into pcDNA3.1(-) (Invitrogen, Carlsbad, CA, USA). The sequence of each construct was verified with an ABI 310 DNA sequencer (PE Applied Biosystems, Foster, CA, USA). All vectors were propagated with XL-1 Blue E. coli, and purified using Qiagen (Chatsworth, CA, USA) anion exchange columns. Vectors were expressed in SCG neurons using an intranuclear microinjection technique as described previously (Ikeda, 1996; 1997).The cDNAs encoding Gα (wild-type or mutants), Gβ, and Gγ subunits (10 mm Tris and/or 1 mm EDTA, pH 7.4) were injected at a ratio (weight) of 0.3-0.5:1:1(30-50:100:100 ng μl−1). The injection concentration of the cDNAs encoding all tested RGS proteins except the EGFP-RGS8 core (5-10 fold higher) was 1 ng μl−1.

Electrophysiology and data analysis

GIRK channel currents were recorded using the whole-cell variant of the patch-clamp technique as described previously (Jeong & Ikeda, 1998; Ruiz-Velasco & Ikeda, 1998). To isolate GIRK currents, patch pipettes were filled with a solution containing (mm): 135 KCl, 11 EGTA, 1 CaCl2, 2 MgCl2, 10 Hepes, 4 MgATP, 0.1 Na2GTP (pH 7.2, 296 mosmol (kg H2O)−1). The external solution contained (mm): 130 NaCl, 5.4 KCl, 10 Hepes, 0.8 MgCl2, 10 CaCl2, 15 glucose, 15 sucrose, 0.0001 tetrodotoxin (pH 7.4, 323 mosmol (kg H2O)−1). In experiments quantifying GIRK kinetic parameters, [K+] in the external solution was increased to 20 mm from 5.4 mm by replacing an equal amount of Na+. Noradrenaline was applied to single neurons via either a gravity-fed fused silica capillary tube connected to an array of seven polyethylene tubes or a gravity-fed three barrel square glass tube connected to automated stepper motor-driven perfusion system (Model SF-77B, Warner Instrument Co., Hamden, CT, USA). The outlet of the perfusion system was located within 100 μm of the cell. All experiments were performed at room temperature (21-24 °C).

In initial reconstitution experiments, maximal inward GIRK currents during a voltage-ramp protocol were determined by subtracting basal GIRK currents (Ba2+-sensitive plus leak currents) from those acquired after agonist application. For quantifying alterations in current kinetics, GIRK currents were activated by a 60 s agonist application using the computer-driven automated perfusion system in neurons held at -80 mV. Non-linear curve fitting (Marquardt-Levenberg algorithm) was performed with the IGOR Pro data analysis package (WaveMetrics, Lake Oswego, OR, USA). Data are presented as means ± standard error of the mean (s.e.m). ANOVAs followed by Fisher's PLSD were performed using StatView 5.0 (SAS Institute, Cary, NC, USA). Statistical significance was defined as P < 0.05.

Kinetic modelling

The model illustrated in Fig. 6 is based loosely on the work of Chuang et al. (1998). Rate constant k1 defines a hypothetical transition (S1→S2) that is accelerated by the N-terminus of RGS8. State S1 could represent a ‘preactivated’ state of the G protein that is converted to an ‘activated’ state (S2) by ligand-occupied receptor (R*). Rate constant k2 defines the GDP-GTP exchange transition; the release of GDP is believed to be the rate-limiting step under physiological conditions. Other factors such as Gβγ dissociation (S3) from the heterotrimer were lumped into this rate constant. Rate constant k3 defines the GTP hydrolysis rate which is influenced by the core domain of RGS8. Reactions associated with ligand and/or receptor binding, Gβγ/GIRK bindings, and heterotrimer reformation were considered to be rapid in relationship to the defined transitions. GIRK channel activation kinetics (Fig. 6B) were simulated by numerically solving differential equations describing each of the states (see Chuang et al. 1998) using the ODEIVP solver in HiQ version 2.2.1d4 (National Instruments, Austin, TX, USA). The initial value conditions were set such that the entire G protein population occupied state S1 prior to agonist application. Deactivation kinetics were simulated as a first order decay from state S3 as defined by k3. The fraction of G proteins in S3 following a simulated 60 s application of agonist was used as the initial value. The rate constants used to simulate each condition are listed in Fig. 6A and were arrived at empirically (with some guidance from the literature, e.g. Mukhopadhyay & Ross, 1999). After solving the time course of state S3 (which reflects Gβγ), GIRK current time course was arrived at by assuming (rapid) highly co-operative binding (Hill, coefficient of 4) with an excess of Gβγ (Kd set to 0.001). These assumptions resulted in: (1) the slow sigmoid trajectory of the ‘RGS-i’ activation phase, (2) the non-exponential trajectory of the ‘RGS-i’ and ‘RGS-nt’ deactivation phase, and (3) the dramatic slowing of the ‘RGS-nt’ deactivation phase when compared with the ‘RGS-i’ trace. The latter phenomenon arises from a greater ‘saturation’ of GIRK channels by Gβγ as a consequence of increased entry (i.e. greater k1) into S3. This model is clearly an oversimplification (e.g. see Ross & Wilkie, 2000) and is intended to serve only as a possible explanation for the observed phenomenology and rudimentary framework to guide future studies.

Figure 6.

Possible model for RGS actions

A, kinetic model for a simplified G protein cycle. S1→S2 defines a hypothetical transition (k1) triggered by ligand binding to receptor (R*) that occurs prior to GDP-GTP exchange (S2→S3, k2). Hydrolysis of GTP (k3) defines the transition S3→S1. Ligand-receptor binding and GIRK activation by Gβγ are assumed to be rapid compared with the G protein cycle. GIRK channel current is assumed to be proportional to [Gβγ]4 (see table) with Gβγ in large excess over GIRK channels (see Methods). The effect of RGS8 core domain and N-terminus are modelled as increases in k3 and k2, respectively (see inset). B and C, simulated activation and deactivation phase currents (normalized) designed to approximate signalling by wild-type Gα (wt-Gα), RGS-insensitive Gα (RGSi-Gα), and RGSi-Gα in the presence of the RGS8 N-terminus (RGSi-Gα+ RGS8-nt). Rate constants used to model the various conditions are show in A (table). Traces were generated using numerical approximation as described in Methods.


PTX- and RGS-insensitive GαoA mutants couple α2-adrenergic receptors to GIRK channels

Figure 1 illustrates the coupling of heterologously expressed GIRK channels to natively expressed α2-adrenergic receptors (α2-ARs) in dissociated rat superior cervical ganglion (SCG) neurons. GIRK currents were evoked at 0.1 Hz from a holding potential of -60 mV using a voltage protocol consisting of a 0.2 s voltage ramp from -140 to -40 mV (Fig. 1A, top inset). In the absence of agonist, GIRK currents (basal) were negligible (< 100 pA). In contrast, application of 10 μm noradrenaline (NA) rapidly activated large inward currents that reached maximal amplitude between -135 and -125 mV (Fig. 1A). Following agonist removal, the GIRK currents rapidly returned to the basal level. The NA-induced GIRK activation was completely prevented by PTX pretreatment thus establishing that coupling occurs exclusively via Go/i proteins (Fig. 1B). These results are consistent with previous findings in SCG neurons (Ruiz-Velasco & Ikeda, 1998). Co-expression of GαoA(C351G), a PTX-i mutant of GαoA hereafter denoted GαoA(CG), together with Gβ1 and Gγ2 subunits, reconstituted α2-AR/GIRK coupling in PTX-treated neurons (Fig. 1C). The GIRK currents activated by the mutant Gα subunit were virtually indistinguishable from current activated by native G proteins (cf. Fig. 1A and D). As previously observed for Ca2+ channel modulation (Jeong & Ikeda, 2000a), reconstitution of GIRK activation required a functional stoichiometric match between Gα and Gβγ subunits. For example, if Gα expression greatly exceeded Gβγ expression, NA-induced GIRK activation was attenuated, presumably reflecting sequestration of ‘free’ Gβγ subunits by excess GDP-bound Gα. In contrast, if Gβγ expression greatly exceeded Gα expression, a large basal GIRK activation was observed in the absence of agonist reflecting excess ‘free’ Gβγ subunit interaction with GIRK channels (see also Ikeda, 1996; Jeong & Ikeda, 1999). Successful reconstitution of NA-induced GIRK activation was observed in neurons showing basal currents (Ba2+-sensitive plus leak currents) of 100-500 pA (Fig. 1C and D).

Figure 1.

Expression of mutant Gα reconstitutes GIRK channel modulation in pertussis toxin (PTX)-treated superior cervical ganglion (SCG) neurons

Acutely dissociated SCG neurons were injected with plasmids coding for GIRK1 and GIRK4. Following overnight incubation at 37 °C, GIRK currents were recorded using whole-cell patch-clamp in K+ containing solutions. GIRK currents were evoked every 10 s from a holding potential of -60 mV with the voltage ramp protocol illustrated. Current traces represent a subtraction of currents obtained before (1) and after (2) application of 10 μm noradrenaline (NA, continuous bar). The ordinate represents the maximal inward current amplitude. No significant inward current was observed in the absence of GIRK plasmid injection (see Ruiz-Velasco & Ikeda, 1998) indicating a lack of natively expressed GIRK channels in SCG neurons. A and B, GIRK channel currents from SCG neurons expressing GIRK1/4 following overnight incubation with control or PTX (500 ng ml−1) containing culture media, respectively. C and D, GIRK channel activation in a neuron previously injected with GIRK1/4, Gβ1, Gγ2 and either GαoA(C3351G) or GαoA(G184S:C3351G), respectively. The neurons were treated with PTX as in B. Horizontal calibration was 25 ms; vertical calibration was 0.5, 0.5, 0.25 and 1.0 nA for the traces in A-D, respectively. E, bar graph summarizing mean NA-mediated GIRK channel current density for conditions depicted in A-D. The number of cells tested are indicated in parentheses. The current density was calculated as maximal NA-mediated inward current divided by membrane capacitance (as determined from integration of the capacitance transient). GαoA mutants are denoted as CG (C3351G), GS:CG (G184S:C3351G), and SD:CG (S207D:C351G).

Expression of a PTX-i-RGS-i GαoA mutant, GαoA(G184S:C351G), hereafter termed GαoA(GS:CG), together with Gβ1γ2, also reconstituted NA-induced GIRK activation in PTX-treated neurons (Fig. 1D). In this case, the activation and deactivation of GIRK currents was noticeably slower when compared with neurons expressing PTX-i GαoA (cf. Fig. 1A and D). Overall, the alterations in GIRK current kinetics were similar those of observed for N-type Ca2+ channel modulation under similar conditions (Jeong & Ikeda, 2000b). Expression of a second PTX-i-RGS-i GαoA mutant, GαoA (S207D:C351G), hereafter termed GαoA(SD:CG), produced similar results (data not shown). A summary of mean NA-induced GIRK current density in neurons co-expressing GIRK with PTX-i GαoA or PTX-i-RGS-i GαoA is shown in Fig. 1E. All the Gα mutants tested were capable of reconstituting NA-mediated GIRK currents following PTX treatment. Together, these data demonstrate the feasibility of confining receptor/GIRK coupling to a molecularly identified Gα using PTX-i Gα mutants.

Expression of PTX-i-RGS-i GαoA alters GIRK current kinetics

In the experiments described above, temporal resolution was limited by a manual drug perfusion system and a low frequency of stimulation. To increase the temporal resolution of the measurements, solution exchange was accomplished by a stepper motor perfusion system (see Methods) under computer control. The speed of solution change was tested by switching from an external solution containing 5.4 mm K+ to one containing 60 mm K+. The increase in basal current (resulting from the increase in driving force) turned on with a biphasic time course and terminated rapidly with a single component time course following reintroduction of 5.4 mm K+ (Fig. 2A, top). The mechanism underlying the slow component of current increase was unclear, but seemed unlikely to arise from the perfusion system since a similar component was not present during washout. Analysis of the rapid phase (Fig. 2B, top) revealed a 0.23 ± 0.01 s (n= 6) lag followed by a relatively linear increase in current. The washout phase (Fig. 2C, top), analysed in greater detail, revealed a similar lag time (0.31 ± 0.03, n= 6) and a normalized slope of 1.42 ± 0.13 s−1 (determined between 20 and 80 %). From these data, solution exchange (under actual experimental conditions) was judged to be complete about 1 s after initiation of the event.

Figure 2.

Effects of PTX-i-RGS-i expression on GIRK channel kinetics

A, representative current traces. The top trace depicts an increase in GIRK current following a switch from 5 to 60 mm external K+ using a computer-controlled perfusion device. The current increase results from an increase in driving force. Subsequent traces represent GIRK current evoked from a holding potential of -60 mV by applying 10 μm NA for 60 s (continuous bar and dashed vertical lines) in solution containing 20 mm external K+. The second trace down was recorded from a neuron expressing GIRK1 and GIRK4. Traces 3, 4 and 5 from top were recorded from PTX-treated neurons expressing GIRK1/4, Gβ1, Gγ2 plus GαoA(CG), GαoA(GS:CG), and GαoA(SD:CG), respectively. Current amplitudes were normalized such that maximal inward current was equal for all traces. B, activation time course. Expanded time scales for normalized currents shown in (A). Currents in the top panel (CG, CON) and bottom panel (GS:CG, SD:CG) were fitted (thin line) to a modified Hodgkin-Huxley equation (see text) using non-linear regression. C, deactivation time course. Expanded time scales for normalized currents shown in (A).

Using this solution exchange system, NA (10 μm) was applied for 60 s to activate the inward GIRK currents in neurons held at -80 mV and bathed in an external solution containing 20 mm K+ (Fig. 2A, traces 2-5). To facilitate the comparison of activation and deactivation kinetics, the current traces illustrated in Fig. 2A were normalized to the maximal inward current obtained during agonist application. In ‘control’ neurons expressing GIRK alone (Fig. 2A, trace 2), the GIRK currents were rapidly activated and deactivated by application and removal of NA, respectively. In addition, substantial (≈60 %) current decay or desensitization occurred during the 60 s agonist application. These features of GIRK current kinetics are similar to those of acetylcholine-induced GIRK currents natively expressed in atrial myocytes (Breitwieser & Szabo, 1988; Yamada et al. 1998). In PTX-treated neurons expressing GαoA(CG), GIRK current activation/deactivation kinetics (Fig. 2A, trace 3) were similar to those of control neurons. Conversely, the activation and deactivation of the GIRK currents were dramatically altered in neurons expressing the PTX-i-RGS-i GαoA mutants, i.e. GαoA(GS:CG) and GαoA(SD:CG) (Fig. 2A, traces 4 and 5, respectively). Three distinct alterations in GIRK current kinetics were apparent in these neurons: (1) activation of current was slower and started after a prolonged delay; (2) desensitization during agonist application was small or absent; and (3) deactivation during agonist washout was prolonged.

Alterations in GIRK activation kinetics were quantified by fitting the rising phase of the current to:

display math(1)

where I(t) is current amplitude at time t, Imax is maximum current amplitude, t is time following initiation of agonist application, delay is a constant, and τ is a time constant. Although the function was arrived at empirically, the fits to the data were reasonably good (Fig. 2B, continuous lines) and allowed unbiased comparison of the parameters. The above function was chosen because the sigmoid trajectory of the GIRK current rising phase was reminiscent of the classic Hodgkin-Huxley voltage-gated K+ current rising phase. Other investigators have employed similar functions to describe GIRK activation in native systems (Otis et al. 1993; Sodickson & Bean, 1996). The delay parameter was introduced to allow for the lag in response especially apparent with G protein coupling via the PTX-i-RGS-i GαoA mutants (Fig. 2B, lower traces). In the absence of the delay parameter, adequate fits were not obtained unless the exponent (4 in this case) was allowed to vary. Under this circumstance, large exponent values were required that rendered the non-linear fitting algorithm unstable. Note that the delay parameter has not been corrected for the 0.2-0.3 s delay arising from the perfusion apparatus (see above). The time to reach half-maximal amplitude, t0.5, can be derived from eqn (1) as:

display math(2)

Summary data for mean τ, delay, and t0.5 are depicted in Fig. 3A. For both PTX-i-RGS-i GαoA mutants (Fig. 3A, GS:CG, SD:CG), all three parameters were significantly increased when compared with control neurons (GIRK expressing, non-PTX treated). Of note, the mean t0.5, which is influenced by both the delay and τ, was increased about 10-fold over control values (see also Fig. 2B, lower panel). Conversely, activation kinetic parameters for the PTX-i mutant (CG) were not significantly different from control values suggesting that neither PTX treatment nor the C→G mutation alone were responsible for the altered kinetics.

Figure 3.

GIRK kinetic parameters for PTX-i-RGS-i expressing cells

Mean (±s.e.m.) kinetic parameters for control (GIRK1/4 expressing neurons) and PTX-treated neurons expressing GIRK1/4, Gβ1, and Gγ2 plus GαoA(CG), GαoA(GS:CG), and GαoA(SD:CG), respectively. Numbers of neurons in parentheses. Statistical comparisons with control values determined by ANOVA followed by Fisher's PLSD. **P < 0.01. A, mean activation parameters (delay, τ, t0.5) were determined from non-linear curve fits to currents (see Fig. 2B and text). B, mean desensitization as determined as the fraction of current remaining following a 60 s application of agonist (10 μm NA, see inset). C and D, mean deactivation parameters as determined by time required to reach 50 % of post-agonist current (C, see inset) or fractional area under the current during the 90 s washout period (D, see inset). Statistical comparisons with control values determined by Mann-Whitney U test. **P < 0.01.

Desensitization was calculated as the fraction of current (normalized to the maximum) remaining at the end of the 60 s NA application (Fig. 3B, inset). Summary data for mean desensitization are depicted in Fig. 3B. For both control and PTX-treated neurons expressing GαoA(CG), mean GIRK current decreased significantly (approximately 60 %) during agonist application. In contrast, GIRK current coupled via either PTX-i-RGS-i GαoA mutant displayed little desensitization. Deactivation kinetics were quantified using two different measures: (1) half-decay time (t0.5), or (2) fractional area. The latter parameter represents the area under the decay phase (as determined by numerical integration) normalized to the area assuming no decay of current for 90 s, the duration of sampling following agonist removal (Fig. 3D, inset). This measurement was useful in later experiments in which deactivation kinetics were greatly retarded such that greater than 50 % of the current remained 90 s after agonist termination. Both measurements (Fig. 3C and D) showed that deactivation was significantly retarded by GαoA(GS:CG) and GαoA(SD:CG) while GαoA(CG) produced results similar to control values. As deactivation parameters from PTX-i-RGS-i GαoA mutant expressing neurons varied considerably, a non-parametric statistical test (Mann-Whitney U test) instead of ANOVA was employed for these data.

Expression of RGS proteins modifies GIRK current kinetics in neurons expressing PTX-i-RGS-i GαoA

Biochemical studies have shown that the G184→S (for Gαo) or S202→D (for Gαt) mutations disrupt RGS-mediated acceleration of Gα-catalysed GTP hydrolysis and disrupt RGS binding to Gα-GTP (Lan et al. 1998; Natochin & Artemyev, 1998). These studies, however, do not rule out interactions between RGS proteins and Gα subunits in the presence of GPCRs. Furthermore, it was not clear whether the observations made in vitro were applicable to the neuronal environment. To address these questions, we expressed RGS proteins concurrently with GαoA(GS:CG) and examined the GIRK current kinetics. Among the various RGS protein isoforms, RGS8 and RGS10 were selected because both RGS proteins are: (1) abundant in neuronal tissues (Hunt et al. 1996; Gold et al. 1997; Saitoh et al. 1997), and (2) have been shown to modify the kinetics of N-type Ca2+ and GIRK channel modulation (Jeong & Ikeda, 1998; Melliti et al. 1999; Saitoh et al. 1997, 1999, 2001). Figure 4A depicts representative current traces from PTX-treated neurons expressing GαoA(CG), GαoA(GS:CG), GαoA(GS:CG) + RGS8, and GαoA(GS:CG) + RGS10. When compared with neurons expressing GαoA(GS:CG) alone (Fig. 4A, trace 2), co-expression of either RGS8 (trace 3) or RGS10 (trace 4) produced several distinct effects: (1) the lag phase was dramatically reduced; (2) the activation phase was accelerated; (3) the deactivation phase was made even slower; and (4) desensitization during agonist application appeared unaffected (i.e. was absent). These effects occurred in 4/4 RGS10- and 5/7 RGS8-expressing neurons. In the latter sample, GIRK current from two neurons appeared similar to those neurons expressing only GαoA(GS:CG), for example, the t0.5 of activation was between 11.4-12.9 s as compared with 1.6-6.9 s for the remaining five RGS8 expressing neurons. Thus, the two non-responding neurons were not included in the summary data.

Figure 4.

Effects of RGS8 or RGS 10 on PTX-i-RGS-i expressing cells

A, representative current traces. Currents were recorded as in Fig. 2A. Traces 1-4 from top were recorded from PTX-treated neurons expressing GIRK1/4, Gβ1, and Gγ2 plus GαoA(CG), GαoA(GS:CG) and GαoA(GS:CG) + RGS8, and GαoA(GS:CG) + RGS10, respectively. Current amplitudes were normalized such that maximal inward current was equal for all traces. B-D, mean (±s.e.m.) activation and deactivation parameters for the conditions shown in A. Means were compared with neurons expressing GαoA(GS:CG). GαoA(CG) condition shown for comparison.

Figure 4 B depicts the mean activation phase parameters for the experimental groups shown in Fig. 4A. Expression of RGS8 or RGS10 significantly decreased the mean delay, τ, and t0.5 of activation when compared with neurons expressing GαoA(GS:CG) alone (Fig. 4B). Although the activation phase was greatly accelerated by RGS expression, the activation kinetics were not as rapid as those recorded from GαoA(CG) expressing neurons. For example, the mean t0.5 of activation was 1.36 ± 0.09 (n= 5), 13.4 ± 1.4 (n= 14), 3.55 ± 0.98 (n= 5) and 5.03 ± 0.41 s (n= 4) for neurons expressing GαoA(CG), GαoA(GS:CG), GαoA(GS:CG) + RGS8, and GαoA(GS:CG) + RGS10, respectively. In contrast, expression of RGS8 or RGS10 dramatically slowed the deactivation phase in neurons expressing GαoA(GS:CG). Mean rate of deactivation, as determined by either t0.5 (Fig. 4C) or fractional area (Fig. 4D), was decreased severalfold when compared with neurons expressing GαoA(GS:CG) alone. In 3/9 neurons examined (pooled RGS8 and RGS10 data), the current was greater than 50 % of the initial amplitude at the end of the 90 s washout sampling period (e.g. Fig. 4A, 3rd tracedown). In these cases, t0.5 was set to 90 s which leads to an underestimation of the mean value.

These data indicate, somewhat unexpectedly, that heterologously expressed RGS proteins interact functionally with a mutant Gα, GαoA(GS:CG), believed to be immune to the GTPase accelerating properties of RGS proteins. Furthermore, similar results were observed in neurons co-expressing GαoA(SD:CG) (data not shown, n= 5) thus confirming the generality of these effects. Finally, similar phenomena were not observed in neurons co-expressing GαoA(CG) and RGS8 (data not shown, n= 2). Thus, the results do not support the re-establishment of RGS-mediated GAP activity as might occur if the G184S or S207D mutations simply lowered the affinity of GαoA for RGS proteins. If this were the case, one would expect a ‘normalization’ of all kinetic parameters. Instead, the deactivation time course was dramatically prolonged and acute desensitization was not restored. Together, these results suggest that PTX-i-RGS-i GαoA mutants interact with RGS proteins to control the GIRK current kinetics by mechanism(s) independently of the GAP action.

RGS8 domains underlying the modification of GIRK current kinetics in neurons expressing PTX-i-RGS-i GαoA

To identify the region(s) of RGS8 responsible for the kinetic modifications observed in GαoA(GS:CG) expressing neurons, we co-expressed several RGS8 truncation/ deletion mutants (Fig. 5B). The RGS-core construct consisted of the RGS8 core domain fused to EGFP to confirm expression (Melliti et al. 1999). The ΔRGS8 and RGS8-nt constructs coded for N- and C-terminus with the core domain deleted and the RGS8 N-terminus alone, respectively. As illustrated in Fig. 5A, co-expression of RGS8 core did not accelerate the activation kinetics or slow the deactivation kinetics of GIRK currents as expression of RGS8 did (cf. Fig. 5A, traces 1 and 2 from top). Conversely, co-expression of either ΔRGS8 (Fig. 5A, trace 3) or RGS8-nt (Fig. 5A, trace 4) mimicked the effects of RGS8 expression by restoring the fast GIRK activation and slowing the GIRK current deactivation. While the effects of ΔRGS8 were relatively consistent (seen in 5/6 cells), the effects of RGS8-nt exhibited more variability (11/15 cells showed these effects). As with the effects of RGS8 (see above), the negatively responding neurons were easily identified by a greatly prolonged t0.5 of activation (range 10-15 vs. 2-6 s for responsive neurons). These neurons were excluded from the summary analyses shown in Fig. 5C and D. Figure 5C summarizes the effects of different RGS8 domains on GIRK deactivation kinetics in neurons co-expressing GαoA(GS:CG). The mean fractional area was between 0.7 and 0.8 for neurons expressing RGS8, ΔRGS8 and RGS8-nt but, in comparison, significantly decreased to 0.38 by RGS8-core expression. A similar pattern was observed for the mean activation kinetic parameters (Fig. 5D). Expression of RGS8-core domain failed to accelerate the mean delay or τ when compared with RGS8 expression. Conversely, expression of ΔRGS8 or RGS8-nt produced rapid GIRK activation similar to that seen for RGS expression. Analogous results were obtained when the different RGS8 domains were expressed along with a different PTX-i-RGS-i Gα mutant, GαoA(SD:CG) (data not shown, n ≈ 3-5). To determine whether ΔRGS8 expression affected natively expressed Gα subunits, ΔRGS8 was expressed in SCG neurons along with GIRK1/4. Under these conditions, ΔRGS8 expression produced no apparent alteration of GIRK current activation/deactivation parameters (n= 4; data not shown) when compared with cells expressing only GIRK1/4. Taken together, these findings suggest that the N-terminus of RGS proteins contributes to the GIRK current activation.

Figure 5.

Effects of RGS8 domains on PTX-i-RGS-i expressing cells

A, representative current traces. Currents were recorded as in Fig. 2A. Traces 1-4 from top were recorded from PTX-treated neurons expressing GIRK1/4, Gβ1 and Gγ2 plus GαoA(GS:CG) + RGS8, GαoA(GS:CG) + RGS8-core, GαoA(GS:CG) +ΔRGS8-core, and GαoA(GS:CG) + RGS8-nt, respectively. Current amplitudes were normalized such that maximal inward current was equal for all traces. B, schematic representation of RGS constructs. C and D, mean (±s.e.m.) deactivation and activation parameters, respectively, for the conditions shown in A. Mean values were compared with neurons expressing GαoA(GS:CG) + RGS8.


Reconstitution of GIRK channel modulation in PTX-treated neurons expressing PTX-i Gα subunits

In initial experiments (Fig. 1 and Fig. 2), we established that GIRK channel activation was rescued in PTX-treated neurons by expression of both PTX-i and PTX-i-RGS-i mutants of GαoA. Comparison of GIRK channel kinetics recorded from dissociated hippocampal CA3 neurons (Sodickson & Bean, 1996; 1998) with those of this study revealed several similarities. Activation and deactivation kinetics were comparable although signalling in hippocampal neurons (a completely native system) was more rapid (around 2-fold) than that of heterologously expressed GIRK channels coupled to native (Fig. 2, CON) or PTX-i Gα (Fig. 2, CG) in SCG neurons. In addition, acute desensitization of the GIRK response during prolonged agonist application was observed in both preparations. Finally, the activation and decay phases were characterized by a lag phase and sigmoid trajectory. It should be noted that, amongst other factors, the receptors (e.g. GABAbR vs.α2-AR) and channels (e.g. GIRK1/2 vs. GIRK1/4) in these preparations were different thus precluding emphasis on direct comparisons. Nonetheless, the resemblance of the reconstituted GIRK currents to native GIRK channel responses lends confidence that the C-terminus mutation of Gα (i.e. C351→G) does not severely alter functional aspects of Gα function and that heterologous expression of GIRK channels in SCG neurons represents a physiologically relevant system.

Alteration of GIRK channel kinetics in neurons expressing PTX-i-RGS-i Gα subunits

Numerous laboratories have examined the effects of heterologous expression of RGS proteins on GIRK channel kinetics in Xenopus oocytes (Doupnik et al. 1997; Saitoh et al. 1997, 1999, 2001; Chuang et al. 1998; Herlitze et al. 1999; Fujita et al. 2000; Koovor et al. 2000). Although experimental conditions varied in these studies, three consistent effects of RGS protein expression emerged. (1) GIRK deactivation was accelerated, (2) GIRK activation was accelerated, and (3) acute desensitization was enhanced (studied in detail by Chuang et al. 1998; see also Kobrinsky et al. 2000). Since GIRK channel kinetics in Xenopus oocytes are slower (in the absence of RGS protein expression) than those of native systems, one can rationalize either that Xenopus oocytes express low levels of native RGS proteins or that Xenopus RGS isoforms are suboptimal in regard to GIRK channel modulation. These assumptions lead to the prediction that GIRK channel kinetics resulting from activation of a Gα subunit resistant to the effects of RGS proteins (in cells natively expressing RGS proteins), as in the current experiments, would behave in an opposite manner. As shown in Fig. 2 and Fig. 3, this prediction was borne out. GIRK channel kinetics in neurons expressing either GαoA(GS:CG) or GαoA(SD:CG) displayed: (1) slower activation upon agonist application, (2) lack of desensitization during agonist application, and (3) slower deactivation following agonist removal when compared with neurons expressing GαoA(CG) or untreated neurons (i.e. coupling via native G proteins).

The retardation of GIRK channel deactivation is the easiest phenomenon to explain as the electrophysiological results were in concert with the biochemical data for the RGS-i Gα mutations (Dibello et al. 1998; Lan et al. 1998; Natochin & Artemyev, 1998). A decrease in GAP activity resulting from impaired interaction with endogenous RGS proteins would lead to prolongation of Gα-GTP lifetime and thus delayed Gαβγ heterotrimer reformation. As a consequence, Gβγ lifetime and GIRK channel activation following agonist removal would be prolonged. The most dramatic kinetic alterations produced by GαoA(GS:CG) or GαoA(SD:CG) expression were in the GIRK current rising phase. In addition to an increase in the τ of activation, a prolonged lag phase lasting several seconds was evident prior to discernable current activation (Fig. 2B, lower panel). The slowing of activation was consistent with Xenopus oocyte studies demonstrating an acceleration of GIRK current activation following RGS protein expression (Doupnik et al. 1997; Saitoh et al. 1997, 1999; Chuang et al. 1998; Herlitze et al. 1999; Fujita et al. 2000; Koovor et al. 2000). At present, the mechanism underlying this phenomenon is unknown. As steady-state levels of agonist-induced current were unaltered by RGS protein expression in these studies (Doupnik et al. 1997; Saitoh et al. 1997; Fujita et al. 2000), acceleration of the activation phase was not believed to arise as a consequence of RGS-mediated GAP activity. The increase in the lag phase in GαoA(GS:CG) or GαoA(SD:CG) expressing neurons was both dramatic and unexpected. As noted above, lag phases for GIRK current activation and deactivation have been documented in native systems (Sodickson & Bean, 1996, 1998) and thus represent physiologically relevant phenomena.

Non-core domains of RGS proteins contribute to GIRK current activation kinetics

A clue to the mechanism underlying the rising phase alterations seen in GαoA(GS:CG) or GαoA(SD:CG) expressing neurons arose from ‘control’ studies in which RGS proteins were co-expressed to determine the ‘RGS insensitivity’ of the Gα mutants. Surprisingly, expression of RGS8 or RGS10, under these conditions, greatly decreased the lag phase and time constant of GIRK current activation (Fig. 4). The trivial explanation for these findings was that the RGS-i Gα mutants possessed a decreased affinity for RGS proteins that was overcome by RGS protein overexpression. This hypothesis was untenable, however, for the following reasons. First, the deactivation phase was slowed rather than accelerated as predicted by an increase in GAP activity. Second, acute desensitization was absent again, an indication that GAP activity was not increased. Third, subsequent experiments (Fig. 5) demonstrated that expression of either the N-terminus or a construct lacking the core domain, but not the core domain alone, of RGS8 mimicked the effects of RGS8 expression. From these results, we conclude that RGS-mediated changes in Gα GAP activity do not underlie the alterations in GIRK current rising phase. Furthermore, the GIRK current rising phase alterations appear to arise from a non-core domain of the RGS protein.

At this time, little is known about non-core RGS domains and thus the mechanism underlying the actions of the RGS8 N-terminus on the GIRK current activation phase remains speculation. One possibility is that RGS8 non-core domains influence the Gα GDP-GTP exchange rate. Several observations, however, argue against this mechanism. First, non-core domains of RGS4 were not visualized in the RGS4-Gαi1 crystal (Tesmer et al. 1997) suggesting flexibility of these regions and thus little interaction with Gα under these conditions (e.g. absence of receptor). Second, in vitro experiments indicate that basal GDP-GTP exchange rates were not significantly altered by the Gα SD mutation (Lan et al. 1998). Third, receptor-catalysed GDP-GTP exchange rate was not altered by RGS4 or another GAP, PLC-β1 (Mukhopadhyay & Ross, 1999). A second possible mechanism is that non-core RGS domains interact with G protein-coupled receptors (GPCR), perhaps acting as scaffolds to ‘pre-couple’ the GPCR and G protein heterotrimer complex (Neubig, 1994). In this regard, RGS proteins have been shown to influence GPCR signalling specificity (Xu et al. 1999) and, conversely, receptor-Gα fusion constructs have been shown to influence RGS specificity for Gα (Cavalli et al. 2000). Moreover, Zeng et al. (1998) have suggested that the N-terminus of RGS4 interacts with GPCRs. Based on these observations and the results presented here, we currently favour a mechanism whereby the N-terminus of RGS8 interacts with the α2-AR to ‘pre-couple’ or facilitate the interaction of GPCR with the G protein (Fig. 6). When interaction between Gα and RGS proteins is disrupted by mutation to specific residues, slowing of activation and deactivation occurs through separate mechanisms mediated by discrete RGS domains. To explain the actions of RGS8 and RGS8-nt on GαoA(GS:CG) or GαoA(SD:CG) expressing neurons, it must be assumed that the functions of the RGS N-terminus can be reinstated by increasing RGS concentration whereas the GAP activity of the core domain cannot. Within the framework of this assumption, the Gα mutations that confer RGS core domain insensitivity are believed to directly or indirectly weaken the actions and/or binding of the RGS N-terminus. Therefore, in the presence of GαoA(GS:CG) or GαoA(SD:CG), the RGS N-terminus can exert influence under conditions of heterologous overexpression but not at levels of natively expressed RGS proteins. An apparent contradiction to this scheme is the paradoxical further slowing of the deactivation phase produced by expression of RGS8 and RGS10 (Fig. 4) or non-core domains of RGS8 (Fig. 5). Since RGS8 enhances activation but does not, in the setting of the GαoA(GS:CG) mutant, enhance deactivation, there may be a large amount of free Gβγ generated. If this is combined with saturation of the GIRK response, then it may take an exceptionally long time for the free Gβγ level to drop sufficiently to permit deactivation of the GIRK channels. Preliminary kinetic modelling (Fig. 6) indicates that saturation of GIRK channels with Gβγ, a situation that might arise with increased coupling efficiency and decreased GTPase activity, could produce such an effect. In this circumstance, GIRK channel deactivation kinetics would not be directly correlated with free [Gβγ] or GTPase activity. It should be emphasized, however, that such a scenario remains highly speculative. Recent findings, such as the putative association of RGS proteins with GIRK (Fujita et al. 2000), the ability of RGS8 to localise in the nucleus under certain conditions (Saitoh et al. 2001), and the influence of receptor/channel stoichiometry on RGS effects (Herlitze et al. 1999) indicate an unexpected level of complexity. Moreover, it must be emphasized that our interpretations are predicated on the assumption that the point mutations conferring insensitivity to PTX and RGS core domains do not significantly alter other aspects of Gα function. Given this complexity, further experiments are required to define the molecular mechanisms underlying the phenomenology described here.

Regardless of the precise mechanisms involved, however, the above results indicate that endogenous RGS proteins greatly influence the kinetics of GIRK channel activation and deactivation. Moreover, it appears that different domains of RGS proteins contribute distinct aspects to the temporal regulation of GIRK channels. The RGS core domain contributes to increased Gα GTPase activity thus accelerating termination of the G protein cycle. In contrast, the N-terminal domain of RGS8 appears to promote entry into the G protein cycle, possibly by enhancing GPCR/heterotrimer coupling. Together, these opposing effects should allow for an increase in temporal fidelity without a dramatic decrease in signal strength. Since GIRK channels underlie postsynaptic responses in both the central and peripheral nervous system, the function of RGS proteins will contribute to the strength and duration of slow synaptic potentials.


The authors are grateful to Ms Marina King for outstanding technical assistance, Dr R. Neubig for reading a preliminary version of the manuscript and providing critical comments, and Drs M. Simon, B. Adams, and D. Logothetis for supplying clones used in the study. This work was supported in part by National Institutes of Health with grants GM56108 and NS37615 to SRI.