βL–βM loop in the C-terminal domain of G protein-activated inwardly rectifying K⁺ channels is important for G_βγ subunit activation

Abstract

The activity of G protein-activated inwardly rectifying K⁺ channels (GIRK or Kir3) is important for regulating membrane excitability in neuronal, cardiac and endocrine cells. Although G_βγ subunits are known to bind the N- and C-termini of GIRK channels, the mechanism underlying G_βγ activation of GIRK is not well understood. Here, we used chimeras and point mutants constructed from GIRK2 and IRK1, a G protein-insensitive inward rectifier, to determine the region within GIRK2 important for G_βγ binding and activation. An analysis of mutant channels expressed in Xenopus oocytes revealed two amino acid substitutions in the C-terminal domain of GIRK2, GIRK2_L344E and GIRK2_G347H, that exhibited decreased carbachol-activated currents but significantly enhanced basal currents with coexpression of G_βγ subunits. Combining the two mutations (GIRK2_EH) led to a more severe reduction in carbachol-activated and G_βγ-stimulated currents. Ethanol-activated currents were normal, however, suggesting that G protein-independent gating was unaffected by the mutations. Both GIRK2_L344E and GIRK2_EH also showed reduced carbachol activation and normal ethanol activation when expressed in HEK-293T cells. Using epitope-tagged channels expressed in HEK-293T cells, immunocytochemistry showed that G_βγ-impaired mutants were expressed on the plasma membrane, although to varying extents, and could not account completely for the reduced G_βγ activation. In vitro G_βγ binding assays revealed an ∼60% decrease in G_βγ binding to the C-terminal domain of GIRK2_L344E but no statistical change with GIRK2_EH or GIRK2_G347H, though both mutants exhibited G_βγ-impaired activation. Together, these results suggest that L344, and to a lesser extent, G347 play an important functional role in G_βγ activation of GIRK2 channels. Based on the 1.8 Å structure of GIRK1 cytoplasmic domains, L344 and G347 are positioned in the βL–βM loop, which is situated away from the pore and near the N-terminal domain. The results are discussed in terms of a model for activation in which G_βγ alters the interaction between the βL–βM loop and the N-terminal domain.

Many inhibitory neurotransmitters exert their actions by stimulating pertussis toxin (PTX)-sensitive G protein-coupled neurotransmitter receptors (GPCR), which in turn open G protein-activated inwardly rectifying K⁺ channels (GIRK or Kir3) (Hille, 1992). K⁺ ions exit the cell when GIRK channels open, thereby hyperpolarizing the cell's membrane potential and making it more difficult to elicit an action potential. The loss of GIRK channels leads to hyperexcitability and seizures in the brain (Signorini et al. 1997; Slesinger et al. 1997), cardiac abnormalities (Wickman et al. 1998) and hyperactivity and reduced anxiety (Blednov et al. 2001). Four mammalian GIRK channel subunits (GIRK1–4) have been identified, which coassemble to form neuronal GIRK channels (Yamada et al. 1998). Coimmunoprecipitation studies using brain tissues have demonstrated that GIRK channels are composed of heteromultimers of GIRK1/2 and GIRK2/3 (Liao et al. 1996; Jelacic et al. 2000). In some areas of the brain, such as in the substantia nigra, GIRK channels may exist as GIRK2 homomultimers (Liao et al. 1996; Inanobe et al. 1999).

GIRK channels have the canonical features of the inwardly rectifying K⁺ channel family (Kir1–7), which includes cytoplasmic N- and C-terminal domains, two putative transmembrane domains (M1, M2), and a highly conserved pore–loop complex involved in ion selectivity (Doupnik et al. 1995). Of the seven distinct families of Kir channels, however, only GIRK channels are activated by G protein G_βγ subunits (Logothetis et al. 1987; Reuveny et al. 1994; Wickman et al. 1994). In addition, GIRK channel activity is also regulated by G protein-independent signalling molecules. Intracellular Na⁺, MgATP and PIP₂, and extracellular ethanol have all been reported to open GIRK channels in the absence of functional G proteins (Ho & Murrell-Lagnado, 1999; Kobayashi et al. 1999; Lewohl et al. 1999; Petit-Jacques et al. 1999; Zhang et al. 1999).

Since the discovery that G_βγ subunits activate GIRK channels (Logothetis et al. 1987), the molecular mechanism through which G_βγ subunits open the channel has remained elusive. Initially, studies using chimeras and biochemistry demonstrated that G_βγ subunits bind directly to the N-terminal domain and the distal part of the C-terminal domain of GIRK1 (Huang et al. 1995; Inanobe et al. 1995; Kunkel & Peralta, 1995; Slesinger et al. 1995). Krapivinsky et al. (1998) examined the effects of synthetic peptides derived from GIRK1, GIRK4 or IRK1 on the biochemical binding of G_βγ to GIRK1/4 channels and concluded that two regions within the C-terminal domain were important for G_βγ binding and activation. He et al. (1999, 2002), on the other hand, examined the G_βγ sensitivity of GIRK4 and IRK1 chimeras and identified specific amino acids in the N- and C-terminal domains that were important for generating either the agonist-activated or G_βγ-dependent basal current (He et al. 1999, 2002). At present, there is no clear consensus on which region of the GIRK channels is essential for G_βγ activation.

To identify regions important for G_βγ activation, we studied chimeras of GIRK2 and a G protein-insensitive inward rectifier (IRK1). We chose GIRK2 because GIRK2, unlike GIRK1, readily forms functional homomultimers in neurones as well as in heterologous expression systems (Slesinger et al. 1996; Inanobe et al. 1999) and because the identity of amino acids involved in G_βγ activation of GIRK2 is unknown. The G_βγ sensitivity of chimeras was evaluated in two different expression systems: Xenopus oocytes and mammalian cells. We also examined the response of each chimeric channel to ethanol, which served as an indicator of GIRK channel gating that is G protein independent (Kobayashi et al. 1999; Lewohl et al. 1999; Zhou et al. 2001). Finally, we measured the biochemical binding of G_βγ to glutathione S-transferase (GST) fusion proteins containing the C-terminal domain of the different mutant channels. We identified two amino acids in the middle of the C-terminal domain of GIRK2 that contribute to G_βγ binding and activation. Some of these results have been published in the form of an abstract (Finley et al. 2003).

Methods

Molecular biology

GIRK1 was in pBSK (Kubo et al. 1993b), GIRK2a cDNA was in pBTG (Lesage et al. 1994) and IRK1 was in pBSK (Kubo et al. 1993a). Chimeras were constructed by identifying transition points between GIRK2 and IRK1, using CLUSTAL alignment analysis, as illustrated in Fig. 1A. The nomenclature, I1G2_xx-xx, refers to IRK1 (I1) and the amino acid sequence in GIRK2 (G2_xx-xx) in the chimeric channel. The following point mutations in GIRK2 were tested: F338Y, T343F, L344E, G347H, F348Y and L344E/G347H (EH). All mutants were constructed using PCR and the mutation confirmed with DNA sequencing. The following chimeras were constructed but did not lead to detectable currents when expressed in oocytes: I1G2_25–190, I1G2_25–414, I1G2_25–414, I1G2_51–414, I1G2_73–414, I1G2_97–271, I1G2_1–271, I1G2_97–310, I1G2_97–335 and I1G2_97–352.

Figure 1. The amino acid segment that differs between chimeras I1G2_1–335 and I1G2_1–352 is important for G_βγ activation.

A, schematic diagram showing design of chimeras. Examples of macroscopic currents recorded from Xenopus oocytes injected with GIRK2 (B), I1G2_1–335 (C), I1G2_1–352 (D) cRNA and either the M2 muscarinic receptor (+ M2R), or G_β1 and G_γ2 subunits (+ G_βγ). GIRK2 shows both carbachol (carb)-activated and G_βγ-stimulated currents as well as ethanol (EtOH)-induced currents. I1G2_1–335 has a large basal current, but is not activated following carbachol stimulation. Currents were recorded with TEVC and perfused with 95mm KCl, 95mm KCl + 3μm carbachol (filled bar), 95mm KCl + 200mm ethanol (hatched bar) or 95mm NaCl (grey bar). The current in the presence of 95mm NaCl represents the leakage current, and was used to measure the agonist-independent basal current. Holding potential was −80 mV. E and F, bar graphs show the carbachol-induced current normalized to the basal current (E) and the agonist-independent current in the absence and presence of G_βγ subunits (F). *Significant difference from wild-type GIRK2 (P < 0.05). **Significant difference between basal and G_βγ subunits (P < 0.05). n= 5–42.

In vitro methyl-capped cRNA was made from linear cDNA and T3 or T7 RNA polymerase (Stratagene). The quality of cRNA was estimated using an ethidium-stained formaldehyde gel and the concentration measured by UV spectrophotometry. Xenopus oocytes were isolated as previously described (Slesinger et al. 1996). Briefly, oocytes were surgically removed from one side of the frog under anaesthesia (0.1% tricaine), the frog was sutured, and then allowed to recover from surgery. The experimental procedures were approved by the Institutional Animal Care and Use Committee (IACUC) at The Salk Institute. Oocytes were injected with a 46nl solution containing cRNA for the G protein G_β1 (∼2–8ng) and G_γ2 (∼2–8ng) subunits or the human muscarinic receptor (0.2–2ng), and the GIRK channels (0.5–5ng). In some experiments, βARK1-ct cRNA (∼6ng) was coinjected with the cRNA for GIRK channels (He et al. 1999). Oocytes were incubated in ND96 (96mm NaCl, 2mm KCl, 1mm CaCl₂, 1mm MgCl₂, 5mm Hepes, pH 7.6 with NaOH) for 1–7 days at 18°C.

For expression in mammalian cells, the channel cDNA was subcloned into pcDNA3 and transfected into human embryonic kidney cells (HEK-293T). HEK-293T cells were cultured in Dulbecco's Modified Eagle's Medium (DMEM) supplemented with fetal bovine serum (10%), glutamine (2mm), penicillin (50 U ml⁻¹) and streptomycin (50μg ml⁻¹; Gibco) in a humidified 37°C incubator with 95% air–5% CO₂. Cells were plated onto 12 mm glass cover slips (Warner Instruments) coated with poly-d-lysine (20μgml⁻¹; Sigma) and collagen (100μgml⁻¹; BD Biosciences) in 24-well plates. HEK-293T cells were transiently transfected with cDNA using the calcium phosphate method. Briefly, cDNA (0.08μgml⁻¹) was mixed in sterile de-ionized water with 0.25m CaCl₂, then combined 1 : 1 with Hepes-buffered saline (280mmNaCl, 10mmKCl, 1.5mmNa₂HPO₄, 12mm glucose, 50mm Hepes (pH 6.9 with ∼1 n NaOH); 50μl of this mixture was added to each well and incubated for 16–32 h.

Electrophysiology

Macroscopic currents were recorded from oocytes with a two-electrode voltage-clamp (TEVC) amplifier (Geneclamp 500, Axon Instruments), filtered at 0.05–2kHz, digitized (0.1–2kHz) with a Digidata 1200 A/D interface (Axon Instruments) and stored on a laboratory computer. Electrodes were filled with 3m KCl and had resistances of 0.6–1 MΩ. Oocytes were perfused continuously with an extracellular solution containing 90mm XCl (X = K⁺ or Na⁺), 2mm MgCl₂ and 10mm Hepes (pH 7.5 with ∼5mm XOH). The leakage current was determined using 95mm Na⁺ and subtracted directly from the current measured in 95mm K⁺. For ethanol activation, 100% ethanol was added directly to the 95mm K⁺ solution to give 200mm ethanol (EtOH density = 0.7893 g ml⁻¹). A small chamber (3mm × 15mm) with fast perfusion was used to change the extracellular solution and was connected to earth via a 3m KCl agarose bridge.

The whole-cell patch clamp technique (Hamill et al. 1981) was used to record macroscopic currents from HEK-293T cells. Borosilicate glass (Warner; P6165T) electrodes had resistances of 1–3 MΩ and were coated with Sylgard to reduce capacitance. Membrane currents were recorded with an Axopatch 200B (Axon Instruments) amplifier, adjusted electronically for cell capacitance and series resistance (80–100%), filtered at 2kHz with an 8-pole Bessel filter, digitized at 5kHz with a Digidata 1200 interface (Axon Instruments) and stored on a laboratory computer. Intracellular pipette solution contained (mm) 130 KCl, 20 NaCl, 5 EGTA, 2.56 K₂ATP, 5.46 MgCl₂ and 10 Hepes (pH 7.2 with ∼14mm KOH). With these ion concentrations there was ∼140mm K⁺, 1.5mm free Mg²⁺ and 2mm Mg-ATP in the intracellular solution. Li₃-GTP (300 μm; RBI) was added fresh to the intracellular pipette solution to sustain the activation of GIRK channels. The external bath solution (20mm K⁺) contained (mm) 140 NaCl, 20 KCl, 0.5 CaCl₂, 2 MgCl₂ and 10 Hepes (pH 7.2). The osmolarity was 310–330 mosmol l⁻¹. Current–voltage relations were not corrected for the junction potential of ∼4 mV, estimated using the Junction Potential Calculator (Axon Instruments).

Biochemistry

The C-terminal domains of GIRK2 (beginning with M191) and IRK1 (beginning with V179) were subcloned into pGEX2T (Amersham Pharmacia Biotech) using BamH I and Sma I restriction sites engineered by PCR at the 5′ and 3′ ends of the C-terminal domains. The resulting GST-fused C-terminal domains were purified using standard procedures. G_βγ binding to GST fusion proteins was measured as previously described (Huang et al. 1995) with the following modifications. Binding was performed for 45 min on ice and then each reaction sample was transferred onto a Cytosignal spin column. After three washes with 500μl PBS–0.1% lubrol, the GST-fused proteins–G_βγ complexes were eluted from the columns with 15μl 2 × SDS sample buffer. Anti-G_β antibody (SC-20: Santa Cruz) and anti-GST antibody (Amersham Pharmacia) were used for Western blot analysis. Western blots were quantified using the ‘gel’ module in Image J (NIH software). For each blot, the optical density (OD) of the G_β band was divided by the OD for the GST band and then normalized to the GIRK2 for that experiment.

Immunocytochemistry

HEK-293T cells were cultured in DMEM containing 10% fetal bovine serum, 2.5 i.u. ml⁻¹ penicillin–streptomycin, and 2mm glutamine and transfected with HA-tagged constructs 24h later using the calcium phosphate method. A haemagglutinin (HA) tag was inserted into the extracellular p-loop of mutant channels by subcloning the C-terminal mutation (via the Bst EII site) into HA-GIRK2, which was kindly provided by Chen et al. (2002). Cells were fixed 20–24h after transfection by incubation in 1% paraformaldehyde for 30 min. The cells were washed two times with PBS and half were permeabilized by incubation with 0.25% Triton X-100 in PBS for 10 min. All cells were washed with PBS and incubated in blocking buffer (2% donkey serum and 2% IgG free bovine serum albumin in PBS) for 1h at room temperature. The cells were incubated in the dark in Alexa488-conjugated anti-HA antibody (Covance) in blocking buffer (1 : 400) for 2h at room temperature, washed three times with PBS and mounted onto glass coverslips with 1,4-diazabicyclo(2,2,2)octane (DABCO) in glycerol (Slow Fade Light Antifade Kit; Molecular Probes). Cells were imaged (0.35μm slice thickness) using a Zeiss LSM 5 Pascal laser confocal microscope with a ×63 objective. To compare different mutants, the same gain, pin-hole and exposure time were used for all channels.

Analysis

All values are reported as mean ±s.e.m. Carbachol- (‘c’ in Fig. 1B inset) and ethanol-induced (‘e’) currents were expressed as a ‘fold’ increase over basal current (‘b’), fold increase = c/b or e/b, respectively. One-way ANOVA followed by a post hoc Dunnett's test was used to test for statistical significance (P < 0.05), using GIRK2 as control. A Bonferroni post hoc test was used to evaluate differences among mutants. In experiments examining the effect of expressing G_βγ subunits, we used Student's two-tailed t test on the absolute current levels to compare basal versus coexpression with G_βγ subunits. Distance between amino acids in the GIRK1 structure (PDB:1 N9P, Biological Unit) were measured using a Swiss-PDB viewer and displayed using Accelrys Viewerlite 5.0. The G_βγ domains were defined based on the following studies. Huang et al. (1995, 1997) narrowed the G_βγ binding domains to Q34–I86 in the N-terminal domain and V273–P462 in the C-terminal domain of GIRK1. Kunkel & Peralta (1995) identified T290–Y356 in GIRK1. Ivanina et al. (2003) demonstrated G_βγ binding to F181–G254 and, to a lesser extent, G254–P370 of GIRK1 C-terminal domain. In GIRK2, G_βγ binds to I46–L96 of the N-terminal and L310–E380 of the C-terminal domain of GIRK2 (Ivanina et al. 2003). He et al. (2002) demonstrated G_βγ binding to N253–Y348 of GIRK4. Krapivinsky et al. (1997) identified two peptide sequences, M364–R383 of GIRK1 and S209–R225 of GIRK4, that exhibited potent inhibition of G_βγ binding to native GIRK channels.

Results

C-terminal segment of GIRK2 channels implicated in G_βγ activation

To localize the G_βγ activation site(s) in GIRK2, 21 different chimeras of IRK1 and GIRK2 (using chimera I1G2_97–190 as the backbone) were constructed by systematically replacing 15–40 amino acid segments of IRK1 with the homologous amino acids from GIRK2 (Fig. 1A). I1G2_97–190 contained the hydrophobic core domain (m1-P-m2) from GIRK2 and N- and C-terminal domains from IRK1. I1G2_97–190 was shown previously to be K⁺ selective and not gated by G proteins like IRK1 (Slesinger, 2001). In addition, chimera I1G2_97–190 preserved part of the G protein gate that was localized to the m2 transmembrane domain (Sadja et al. 2001; Yi et al. 2001). Each chimera cRNA was injected into Xenopus oocytes with either the cRNA for the M2 muscarinic receptor (M2R) or G_β1 and G_γ2 cRNAs. To evaluate channel function, we examined three parameters of channel activation. First, we measured the activation of current following stimulation of the M2 muscarinic receptor with carbachol (Fig. 1B, ‘c’ in inset); this activation relies on the endogenous G proteins and is G_βγ sensitive. To account for possible changes in expression levels in oocytes, we expressed the carbachol-activated current as a function of the agonist-independent (basal) current. Second, we compared the amplitude of basal current (Fig. 1B, ‘b’ in inset) in oocytes coexpressing the chimera with G_βγ subunits (+G_βγ) with those coexpressing the chimera and M2R receptor. Coexpression of G_βγ subunits in oocytes bypasses the endogenous G proteins, leading to the persistent G_βγ activation of GIRK channels (Reuveny et al. 1994). We chose to express this as absolute current and used statistical analysis to assess whether G_βγ subunits significantly enhanced the basal current. Third, we measured the amplitude of the inward current induced with 200mm ethanol (Fig. 1B, ‘e’ in inset). Ethanol (EtOH) activates GIRK channels but inhibits IRK1 channels (Kobayashi et al. 1999; Lewohl et al. 1999; Zhou et al. 2001). Activation by ethanol does not require functional G proteins and therefore provides an important assessment of channel function for putative G protein-impaired mutants.

The GIRK2/IRK1 chimeras could be classified into three main groups: no current, G_βγ sensitive, and G_βγ impaired. Generally, chimeras containing the N-terminal domain of IRK1 and part or all of the C-terminal domain of GIRK2 failed to generate currents when expressed in oocytes (data not shown); these and other non-functional chimeras were not studied further (see Methods). Two chimeras (I1G2_1–352 and I1G2_1–390) were G_βγ responsive and activated by ethanol, similar to GIRK2 (Fig. 1). The remaining chimeras (e.g. I1G2_1–190–I1G2_1–335) displayed moderate to large basal currents but little or no activation following stimulation of the muscarinic receptor (Fig. 1E). In addition, the basal current was not enhanced by coexpression with G_βγ subunits (Fig. 1F). Thus, these chimeras appear to have impaired G_βγ sensitivity. The change in G_βγ sensitivity occurred between chimeras I1G2_1–335 and I1G2_1–352. To explore whether the agonist-independent current was G_βγ sensitive for these two chimeras, we coinjected the cRNA for the C-terminal domain of βARK1 (βARK1-ct), which sequesters free G_βγ subunits in oocytes (He et al. 1999). A G_βγ-sensitive basal current would be expected to be smaller in the presence of βARK1-ct. βARK-ct reduced the carbachol-activated current for GIRK2 and I1G2_1–352, indicating suppression of G_βγ activity. The basal currents for I1G2_1–335 changed little in the presence of βARK1-ct, from −6.6 ± 0.6μA (n= 5) to −5.8 ± 0.3μA (n= 5) with βARK1-ct. Similarly, the basal current for I1G2_1–352 was not affected by βARK1-ct (−0.45 ± 0.19μA (n= 8) versus−0.46 ± 0.22μA (n= 5) with βARK1-ct. These results suggest the agonist-independent current is relatively insensitive to G_βγ subunits, although we cannot rule out a small component of G protein-dependent activation.

To define the amino acids involved in G_βγ activation better, we constructed point mutations in the region that differed between I1G2_1-335 and I1G2_1-352. We focused on amino acids that are conserved among GIRK channels but differ from IRK1 (Fig. 2A). Five amino acids met this criterion, and each was changed from the amino acid in GIRK2 to the corresponding amino acid in IRK1. GIRK2_F388Y, GIRK2_T343F and GIRK2_F348Y all exhibited carbachol-induced currents that were comparable or larger than those of GIRK2 (Fig. 2B and C). Two mutations in GIRK2, GIRK2_L344E and GIRK2_G347H, however, showed dramatically smaller carbachol-induced currents, 0.13-fold and 0.54-fold increases over basal, respectively (compared to a 7.7-fold increase over basal for GIRK2). Despite the small carbachol response, both mutants showed a stimulated basal current when coexpressed with G_βγ subunits as well as normal ethanol-induced currents (Fig. 2B–E). If L344E and G347H decrease the sensitivity of the mutant channel to G_βγ subunits, then the agonist-activated response may be impaired because there is an insufficient amount of G_βγ subunits generated during carbachol stimulation. In contrast, the high levels of G_βγ subunits that are present when G_βγ subunits are coexpressed may be sufficient to activate the mutant channels.

Figure 2. Mutation of two conserved amino acids in the C-terminal domain of GIRK2 reduces G_βγ activation.

A, sequence alignment of the amino acids in GIRK1, GIRK2, GIRK3, GIRK4 and IRK1. Amino acids that are conserved in GIRK family but differ from IRK1 are indicated in bold. Asterisk indicates 100% identity and small filled circles indicate conserved substitutions. B–E, oocytes were injected with cRNA for the indicated channel and with either the cRNA for M2R or G_βγ subunits (B). Continuous recordings show the response to 0.3μm carbachol (carb), 200mm ethanol (EtOH) and 95mm NaCl for GIRK2_L344E, GIRK2_G347H and GIRK2_EH. Bar graphs show the average carbachol-induced current normalized to the basal current (C), the average ethanol-induced current normalized to the basal current (D) and the average G_βγ-induced and basal currents (E). Except for F338Y, the carbachol-induced currents for all other mutants are statistically different from GIRK2. None of the EtOH-activated currents is statistically different from GIRK2. All of the +G_βγ currents are statistically larger than basal (P < 0.05). The increase for GIRK2_EH (4-fold), however, is smaller than that for GIRK2 (∼25-fold). n= 5–49.

We next tested for additivity of L344E and G347H effects on GIRK currents by introducing both mutations into GIRK2 (GIRK2_EH). Like the individual mutations, GIRK2_EH exhibited a small basal current with a small increase (∼0.02-fold) during carbachol stimulation. In contrast to the single point mutations, the G_βγ-stimulated currents for GIRK2_EH were significantly reduced; the basal current for GIRK2_EH was enhanced only ∼4-fold with coexpressed G_βγ subunits, as compared to the ∼25-fold increase for GIRK2. Interestingly, the ethanol-induced currents for GIRK2_EH were comparable to those of GIRK2 (Fig. 2D), suggesting that GIRK2_EH retains ethanol sensitivity but has impaired G_βγ activation. Thus, G_βγ activation appears to be affected more severely in GIRK2_EH than with either mutation alone.

Functional studies with G_βγ-impaired mutants expressed in HEK-293T cells

In Xenopus oocytes, GIRK channels can coassemble with an endogenous GIRK subunit (XIR) to produce heteromeric channels (Hedin et al. 1996). To eliminate any possible influence of XIR on the G protein sensitivity of the mutant channels studied in oocytes, we examined the G_βγ-impaired mutants expressed in HEK-293T cells using whole-cell patch clamp technique. Similar to oocytes, chimera I1G2_1-335 exhibited a large, agonist-independent current that was not increased further with carbachol (Figs 3B and 4A–C). By contrast, I1G2_1-352 showed both carbachol-induced and ethanol-induced currents, like GIRK2 (Fig. 4B and C). Thus, the G_βγ sensitivity of chimeras I1G2_1-335 and I1G2_1-352 expressed in HEK-293T paralleled those observed in oocytes. Interestingly, we also observed a change in ethanol activation between chimeras I1G2_1-335 and I1G2_1-352 (Fig. 4C) in HEK-293T cells, which is similar to that reported by Lewohl et al. (1999).

Figure 3. GIRK2_EH expressed in HEK-293T cells has reduced carbachol activation but normal ethanol activation.

HEK-293T cells were transiently transfected with cDNA for the M2R and either GIRK2 (A), I1G2_1-335 (B), GIRK2_EH (C), or GIRK1 plus GIRK2_EH (D) cDNA. Whole-cell patch clamp currents were elicited by voltage steps from −120 to 30 mV. Bath solutions contained 20mm KCl (20 K) solution, 20 K + 0.3μm carb, 20 K + 200mm EtOH or 160mm NaCl.

Figure 4. Summary of G protein sensitivity for mutants channels expressed in HEK-293T cells.

Bar graphs show the agonist-independent (A), the carbachol-induced (B), and the ethanol-induced (C) current density for each mutant. Using the ethanol-activated current as a measure of channel expression, the carbachol-activated current was normalized to the amplitude of ethanol-activated current (D). *Significant difference from GIRK2 (P < 0.05). n= 5–18.

We next examined the G protein sensitivity of GIRK2 point mutants expressed in HEK-293T cells. As in the oocytes, the GIRK2_L344E, GIRK2_G347H and GIRK2_EH all displayed small agonist-independent currents and markedly reduced carbachol-activated currents (Figs 3C and 4A and B). Ethanol (200mm), on the other hand, activated the mutant GIRK2 channels, although the ethanol-activated current for G347H was dramatically smaller than control (Fig. 4C). The GIRK2 point mutants expressed in HEK-293T showed defects in G_βγ activation similar to those observed in oocytes, indicating that XIR did not contribute to the G protein phenotype. Because GIRK1 cannot express on the membrane surface in the absence of other GIRK subunits (Kennedy et al. 1996), we could now examine unequivocally whether GIRK2_EH coassembles with GIRK1. Carbachol-activated currents were restored when GIRK1 was cotransfected with GIRK2_EH (Figs 3D and 4B) and displayed the slow voltage-dependent activation kinetics typical of heteromultimers containing GIRK1 (Slesinger et al. 1996). Thus, GIRK2_EH forms homotetramers as well as heterotetramers, like native GIRK2, indicating that the double (EH) mutation does not impair subunit assembly. Furthermore, the presence of one or more GIRK1 subunits in the tetramer appears to restore G_βγ sensitivity.

The small basal and ethanol-activated currents for GIRK2_G347H and GIRK2_EH suggested that these mutants might express less efficiently than the other mutants in HEK-293T cells. To examine the surface expression of GIRK2 mutants, it was necessary to engineer an extracellular haemagglutinin (HA) epitope. The presence of the HA tag does not appear to alter the function of GIRK2 but may have some effect on trafficking (Chen et al. 2002; Ma et al. 2002). HEK-293T cells transfected with HA-GIRK2_L344E showed intense membrane staining, which was the same or slightly more intense than that for wild-type HA-GIRK2 (Fig. 5B and C). By contrast, both HA-GIRK2_G347H and HA-GIRK2_EH expressed at lower levels on the membrane surface, though they were clearly detectable above the background staining of untransfected cells (Fig. 5A, D and E). In the presence of the L344E mutation, the G347H mutation appears to reduce the surface expression of the double mutant HA-GIRK2_EH. GIRK2_EH expression, however, was recovered by coexpression with GIRK1 (Fig. 5F), consistent with the whole-cell patch clamp recordings (Fig. 4). If we normalize the carbachol-activated current to the amplitude of ethanol-activated current, which can serve as a measure of functional channels on the membrane surface, then I1G2_1-335, GIRK2_L344E and GIRK2_EH clearly show reduced carbachol activation (Fig. 4D). Taken together, these results suggest that some of the reduction in carbachol-activated current in HEK-293T cells may be due to the lower expression levels for GIRK2_G347H but not for GIRK2_L344E and GIRK2_EH.

Figure 5. Immunostaining of HA-tagged GIRK2 channels expressed in HEK-293T cells.

HEK-293T cells were transfected with the channel cDNA indicated in the figure. A–F, surface localization was revealed by immunostaining non-permeabilized cells using an Alexa488 conjugated anti-HA antibody. Images were obtained with a laser confocal microscope, using the same gain, slice thickness (0.35μm), pin-hole and exposure time.

G_βγ binding to C-terminal domains of GIRK2/IRK1 chimeras

We next examined whether L344 or G347 is important for the biochemical binding of G_βγ to the channel. We used a coaffinity precipitation assay to measure the binding of G_βγ subunits to fusion proteins containing the C-terminal domain of the channel (Fig. 6A). In contrast to previous studies examining G_βγ binding with increasingly smaller fragments of the C-terminal domain of GIRK channels (Huang et al. 1995; He et al. 1999; Ivanina et al. 2003), we constructed GST fusion proteins that contained the entire cytoplasmic C-terminal domain of the chimera or the GIRK2 point mutant (Fig. 6A). As shown previously, the C-terminal domain of GIRK2 but not IRK1 binds G_βγ subunits (Huang et al. 1997). Surprisingly, fusion proteins containing the C-terminal domain from the different chimeras all exhibited G_βγ binding similar to GIRK2 (Fig. 6B). Chimeras containing either the proximal (GST-I1G2_181–230) or distal (GST-I1G2_353–414) region of GIRK2 exhibited similar G_βγ binding that was clearly greater than the G_βγ binding to the GST alone or the C-terminal domain of IRK1. By contrast, the G_βγ binding to the C-terminal domain of GIRK2_EH appeared slightly reduced (Fig. 6C). To quantify these possible differences in G_βγ binding, we measured the optical density (OD) of the G_β band, divided by the OD for the GST band and then normalized to the G_βγ binding for GIRK2 (Fig. 6D). Compared to the full-length C-terminal domain of GIRK2 (GST-G2_191–414), only GST-G2_L344E and GST-I1_173–428 showed significantly less G_βγ binding (Fig. 6D). However, G_βγ binding to GST-G2_L344E was not statistically different from GST-G2_G347H or GST-G2_EH. Taken together, the electrophysiological and biochemical experiments suggest that L344 plays a major role in G_βγ activation and, to a lesser extent, in G_βγ binding.

Figure 6. Mutations in βL–βM loop of GIRK2 show variable changes in G_βγ binding.

In vitro coaffinity precipitation (AP) assays were used to measure the G_βγ binding to GST fusion proteins containing the C-terminal domain of the wild-type or mutant channel. A, schematic diagram of GST fusion proteins. B, GST fusion proteins (200nm) were mixed with purified bovine brain G_βγ subunits (40nm). Immuno-blot (IB) using antibody against G_β (upper panel) and, after stripping the blot, against GST (lower panel). All chimeras showed some G_βγ binding. C, G_βγ binding assay for GST, GST-I1_173–428, GST-G2_191–414, GST-G2_EH, GST-G2_336–414 and GST-G2_353–414. D, quantification of G_βγ binding. The optical density (OD) of the G_β band was divided by the OD of the GST band, and then normalized to the GIRK2 for each blot (n indicated in parentheses). *Statistical difference from GST-G2_191–414 (P < 0.05, ANOVA followed by post hoc Bonferroni test). GST-G2_L344E is not statistically different from GST-G2_G347H or GST-G2_EH.

Discussion

Important role for βL–βM loop in G_βγ activation

We first evaluated in Xenopus oocytes the ability of mutant GIRK channels to be activated through stimulation of the m2 muscarinic receptor as well as by coexpression with G_βγ subunits. In the latter case, we assume that the concentration of free G_βγ is significantly higher as compared to the G_βγ liberated during stimulation of the m2 muscarinic receptor. This supposition seems justified since the carbachol-activated currents were on average significantly smaller than the basal currents recorded in oocytes coexpressing G_βγ subunits. We found that GIRK2 channels containing either the L344E or G347H mutation exhibited dramatically smaller carbachol-activated currents but were still activated by coexpressed G_βγ subunits. Combining the two mutations (GIRK2_EH) produced channels that were unresponsive to muscarinic receptor stimulation and now showed little enhancement with coexpressed G_βγ subunits. Electrophysiological recordings obtained in HEK-293T cells also supported the conclusion that L344E and, to a lesser extent, G347H, were important for G_βγ activation. Interestingly, L344 and G347 are both conserved among the different types of GIRK channels but differ in G protein-insensitive inward rectifiers (see Supplementary Material available online), suggesting an important functional role for these amino acids. Assuming the three-dimensional structure of the GIRK2 cytoplasmic domains is the same as that of GIRK1 (Nishida & MacKinnon, 2002), L344 and G347 would be located in an exposed loop (βL–βM) facing the intracellular milieu (see Fig. 7B).

Figure 7. Multiple sites for G_βγ interaction with GIRK channels: a role for the βL–βM loop in G_βγ activation.

A, alignment of the G_βγ binding domains in the N- and C-terminal domains of GIRK channel identified in the following studies: ^aHuang et al. 1995, 1997; ^bKunkel & Peralta, 1995; ^cHe et al. 1999, 2002; ^dKrapivinsky et al. 1998; ^eIvanina et al. 2003; and ^fthis study (see Methods for details). Filled bars (red, green, black): regions binding G_βγ subunits. Shaded red bar: regions showing less G_βγ binding. Stippled bars: peptide competing for G_βγ binding. Arrow marks position of the leucine in the βL–βM loop. The GIRK1 sequence included in 3-D structure is indicated by the black line. Three G_βγ binding domains are drawn to indicate regions of most overlap for G_βγ binding, along with a region unique to GIRK1 (see text for details). B, three putative G_βγ segments are highlighted in the GIRK1 structure (Nishida & MacKinnon, 2002). Model shows four subunits arranged around a pore (circle), with two subunits highlighting G_βγ domains: region 1 (R43–E63, turquoise), region 2 (R190–R219, yellow) and region 3 (E300–P370, red). The amino acid sequence between regions 2 and 3 (green) has also been implicated in G_βγ binding. Region 4 is unique to GIRK1 and is not present in the structure. Note proximity of the N-terminal domain to the βL–βM loop. Two amino acids, L262 (yellow, He et al. 2002) and L333E (red) are highlighted in the structure. C, side view of the same structure. Dashed line indicates approximate position of the cytoplasmic side of the plasma membrane.

We examined the surface labelling of HA-tagged channels expressed in HEK-293T cells to determine whether the smaller G_βγ-activated currents could be explained by decreased surface expression. The expression of GIRK2_L344E was the same as, or slightly higher than, that of wild-type GIRK2 when expressed in HEK-293T cells. Consistent with this, the functional response to ethanol for GIRK2_L344E, which activates GIRK channels through a G protein-independent mechanism (Kobayashi et al. 1999; Lewohl et al. 1999; Zhou et al. 2001), was comparable with that of wild-type GIRK2 channels. Cells transfected with either HA-GIRK2_G347H or HA-GIRK2_EH showed weak plasma membrane staining relative to HA-GIRK2. However, normalizing the carbachol-activated currents to the amplitude of ethanol-activated currents, which can serve as an indicator of surface expression, suggests that G_βγ activation is impaired in both GIRK2_L344E and the double mutant, GIRK2_EH. Perhaps G347H amplifies the G_βγ deficiency of L344E in the double mutant. Additional mutagenesis studies are required to determine the precise role of G347H in surface expression and/or G_βγ activation.

The L344 in the βL–βM loop of GIRK2 has been implicated previously in other GIRK channels. In a study of GIRK4, He et al. (1999) discovered that L339E in GIRK4 (homologous to GIRK2_L344E) impaired agonist-induced activation, similar to GIRK2_L344E and GIRK2_EH. In addition to a loss of agonist-activated current, GIRK4*_L339E exhibited a large, agonist-independent current that was not increased by coexpression with G_βγ subunits (He et al. 1999). The basal current was suppressed by coexpression of βARK1-ct or G_α G protein, however, leading He et al. (1999) to postulate that the agonist-induced current and G_βγ-dependent basal current are generated by two different G_βγ binding sites on the channel; a high affinity site that produces the agonist-independent current and is saturated under basal conditions, and a low affinity G_βγ binding site that is occupied following agonist activation. For generating large, agonist-independent currents for homomeric GIRK4 channels, He et al. (1999) introduced a mutation in the pore of GIRK4 (GIRK4*_S143T) to promote the expression of the homomultimer (Chan et al. 1996; He et al. 1999). Although the S143T mutation has no effect on single-channel kinetics, it is unclear how the mutation leads to larger currents. Interestingly, Peleg et al. (2002) found that GIRK channels expressed at high levels in oocytes leads to large, agonist-independent currents with small receptor-activated currents, due to a limiting supply of G_α subunits. In our study, the expression of GIRK2 in oocytes or HEK-293T cells did not generate an agonist-independent current that was large enough to reliably test the effect of βARK1-ct. In a recent study by Ivanina et al. (2003), GIRK1_L333E/GIRK2_L344E heteromultimers expressed in oocytes displayed small, agonist-independent and agonist-activated currents, similar to our results. Collectively, these studies provide convincing evidence that the leucine in the βL–βM loop plays an important functional role in G_βγ activation.

How does mutating the leucine in the βL–βM loop account for the change in G_βγ activation? One possibility is that the binding of G_βγ subunits to the channel is altered. The effect of the leucine-to-glutamate mutation on the biochemical binding of G_βγ, however, is more equivocal. He et al. (1999) reported an ∼60% decrease in G_βγ binding to the C-terminal domain of GIRK4_L339E. However, G_βγ binding was also reduced for GIRK4_L268I, which had a defective G_βγ-dependent basal current (He et al. 2002). Using in vitro translated ³⁵S-labelled G_βγ to measure binding to GIRK channels, Ivanina et al. (2003) observed a 30–40% decrease in binding to the C-terminal domain of GIRK1_L333E but no detectable change in G_βγ binding to the C-terminal domain of GIRK2_L344E. In our experiments with purified G_βγ subunits, L344E exhibited ∼60% less G_βγ binding but the double mutant (GIRK2_EH) and the chimeric channel (I1G2_1-335), which both exhibited defective G_βγ activation, were not statistically different from control. One complication to the binding studies is that G_βγ subunits bind to multiple regions in the C-terminal domain (see below), thereby potentially masking changes in G_βγ binding.

Even with a twofold decrease in G_βγ binding, it may appear difficult to reconcile this small change with a major loss of agonist-activated and G_βγ-stimulated currents. A limitation to the G_βγ binding assay, however, is that G_βγ binding is measured in vitro with a fusion protein that lacks the transmembrane domains. If we consider GIRK channels to be allosteric proteins, then the affinity for G_βγ subunits may change depending on the state of the channel (Changeux & Edelstein, 1998). Thus, it remains possible that G_βγ binding is altered more dramatically in the context of the intact mutant channel. On the other hand, G_βγ activation is highly cooperative, requiring the binding of multiple G_βγ dimers to the channel (Corey & Clapham, 2001). Thus, a subtle mutation in all four subunits may alter the cooperativity of G_βγ activation and lead to reduced G_βγ activation. Finally, mutations in the βL–βM loop may interfere with the coupling of G_βγ binding to the channel's activation gate. Future studies will clarify the link between G_βγ binding and channel activation.

G_βγ binding sites mapped on the GIRK1 structure

In addition to the βL–βM loop, several other regions in the C-terminal domain of GIRK channels have been implicated in G_βγ binding. We compared the G_βγ binding domains implicated in this and previous studies, and searched for regions of greatest overlap. Based on this criterion, we identified three general regions in GIRK channels important for G_βγ binding (Fig. 7A, see Methods for details). Region 1 contains part of the N-terminal domain. Defining the regions in the C-terminal domain was more difficult and somewhat subjective, since G_βγ sites appear to be distributed over the entire C-terminal domain. We suggest that two general regions, a proximal (region 2) and middle (region 3) C-terminal segment of GIRK1–4, can account for most of the overlap in G_βγ binding sites. Region 3 encompasses the βL–βM loop and surrounding amino acids. A fourth region (region 4) may exist that is unique to the distal end of GIRK1. Mapping these putative G_βγ binding segments onto the three-dimensional structure of GIRK1 reveals that they are clustered on the outer edge of the tetrameric channel, away from the central pore, and well positioned to interact with G_βγ subunits (Fig. 7B). Further refinement of the G_βγ sites on GIRK channels will require the structural determination of a complex of G_βγ subunits and GIRK cytoplasmic domains, as well as incoporation of the G_α binding site (Huang et al. 1995) and PIP₂ sites (Huang et al. 1998; Sui et al. 1998).

The propinquity of the βL–βM loop to the N-terminal domain suggests a testable model for G_βγ activation. The βL–βM loop is situated within 4 Å (Q44–G336 distance is 3.6 Å) of the N-terminal domain. Thus, G_βγ binding may alter the interaction between the βL–βM loop of one subunit with the N-terminal domain of the neighbouring subunit. Strengthening the bonds within the loop and the N-terminal domain might interfere with G_βγ activation. Using the Swiss-Pdb Viewer program to model mutations in GIRK1, the side-chain of the glutamate (L333E) could form hydrogen bonds with the backbone of E334, E335 and F337, within the βL–βM loop. The side-chain of the histidine (G336H) could form a hydrogen bond with the Q44 located in the N-terminal domain. Based on this model, it may be possible to create a mutation that will rescue the G_βγ defect of L344E/G347H. Several lines of evidence support an important role for coupling between the N- and C-terminal domains in other types of inward rectifiers during channel activation. First, the N-terminal domain of GIRK binds G_βγ subunits and interacts cooperatively with the C-terminal domain (Huang et al. 1997). Second, Schulte et al. (1998) found that cysteines in the N- and C-terminal domains of ROMK1 (Kir1.1) were modified only in the closed state, indicating a conformational change in the cytoplasmic domains. Finally, mutations in the N-terminal domain of K_ATP channel subunit (Kir6.2) disrupt the binding of the N-terminal domain to the C-terminal domain (Tucker & Ashcroft, 1999).

Supplementary material

The online version of this paper can be found at:

DOI: 10.1113/jphysiol.2003.056101 and contains supplementary material consisting of a figure entitled:

Clustal alignment of rodent Kir family of channels.

This can also be accessed athttp://www.blackwellpublishing.com/products/journals/suppmat/tjp/tjp131sm.htm