Dual regulation of G proteins and the G-protein–activated K⁺ channels by lithium

Significance

Recent genome-wide association studies suggest a strong linkage between psychiatric disorders, especially bipolar disorder (BPD) and schizophrenia, and ion channels, including G-protein–activated K⁺ channels (GIRK); however, there are no clear functional links. Lithium is a prominent antibipolar treatment. We report a dual regulation of GIRK channels by therapeutic doses of lithium in neurons. We reproduced these regulations in Xenopus oocytes, identified the molecular mechanism (a dual regulation of G proteins via actions on Gα subunit), and verified the action of Li⁺ on Gα–Gβγ interaction by direct biochemical studies. The discovery of a significant regulation of an ion channel by therapeutic doses of lithium may help linking between neuronal excitability and mechanisms of BPD.

Abstract

Lithium (Li⁺) is widely used to treat bipolar disorder (BPD). Cellular targets of Li⁺, such as glycogen synthase kinase 3β (GSK3β) and G proteins, have long been implicated in BPD etiology; however, recent genetic studies link BPD to other proteins, particularly ion channels. Li⁺ affects neuronal excitability, but the underlying mechanisms and the relevance to putative BPD targets are unknown. We discovered a dual regulation of G protein-gated K⁺ (GIRK) channels by Li⁺, and identified the underlying molecular mechanisms. In hippocampal neurons, therapeutic doses of Li⁺ (1–2 mM) increased GIRK basal current (I_basal) but attenuated neurotransmitter-evoked GIRK currents (I_evoked) mediated by G_i/o-coupled G-protein–coupled receptors (GPCRs). Molecular mechanisms of these regulations were studied with heterologously expressed GIRK1/2. In excised membrane patches, Li⁺ increased I_basal but reduced GPCR-induced GIRK currents. Both regulations were membrane-delimited and G protein-dependent, requiring both Gα and Gβγ subunits. Li⁺ did not impair direct activation of GIRK channels by Gβγ, suggesting that inhibition of I_evoked results from an action of Li⁺ on Gα, probably through inhibition of GTP–GDP exchange. In direct binding studies, Li⁺ promoted GPCR-independent dissociation of Gα_i^GDP from Gβγ by a Mg²⁺-independent mechanism. This previously unknown Li⁺ action on G proteins explains the second effect of Li⁺, the enhancement of GIRK's I_basal. The dual effect of Li⁺ on GIRK may profoundly regulate the inhibitory effects of neurotransmitters acting via GIRK channels. Our findings link between Li⁺, neuronal excitability, and both cellular and genetic targets of BPD: GPCRs, G proteins, and ion channels.

Bipolar disorder (BPD) is a common disease of poorly understood molecular mechanisms comprising both neurodevelopmental and genetic factors (1, 2). Lithium (Li⁺) stabilizes the condition in many BPD patients. Historically, the etiology of BPD has been linked to cellular targets of Li⁺, including G proteins, enzymes of phosphoinositide (PI) turnover, and Akt/GSK3β cascade (3–5). However, recent genetic studies have identified a multitude of BPD-associated genes, with a preponderance of proteins related to G-protein–coupled receptors (GPCRs) and ion channels (6), particularly Ca²⁺ channels (2, 7, 8). Linkages with ion channel genes extend to other major psychiatric diseases as well, especially schizophrenia (9, 10). To date, no substantial links between BPD and genes of PI turnover enzymes and GSK3β have been identified.

To understand BPD, it is important to bridge the gap between protein targets of BPD suggested by genetic vs. cellular studies, and to understand how ion channels are involved. Li⁺ may provide a clue. Li⁺ has neuroprotective and neurotrophic effects (11) and affects neuronal excitability (12, 13) by largely unknown mechanisms. Li⁺ acts intracellularly (14), entering neurons through several nonspecific cation channels and voltage-dependent Na⁺ channels (15). Therapeutic doses of Li⁺ are 0.6–1.2 mM in neurons (16); serum concentrations >2 mM are considered toxic. At high doses (EC₅₀ >5 mM), Li⁺ affects AMPA receptor channels (17), voltage-gated Na⁺ channels (18), and Na⁺-activated K⁺ channels (19); however, no clear effects of therapeutic doses of Li⁺ on ion channels have been reported to date.

Here we report a previously unknown dual regulation by Li⁺ of G protein-gated K⁺ (GIRK) channels, major mediators of action of inhibitory neurotransmitters. Activated by direct binding of the Gβγ subunit of G_i/o proteins, GIRK channels regulate analgesia, reward-related behaviors, and mood and have been implicated in several neurologic disorders (20, 21). Recently, KCNJ3, the gene encoding the ubiquitous GIRK1 subunit of GIRK channels, has been genetically linked to schizophrenia and BPD in Asian populations (10). We propose that Li⁺ regulation of GIRK channels offers a link between molecular and cellular targets of Li⁺, neuronal excitability, and BPD genetics.

Results

Li⁺ Dually Regulates GIRK Channels in Hippocampal Neurons.

Because Li⁺ inhibits the GPCR-induced GTP–GDP exchange at Gα (5, 22), we hypothesized that Li⁺ should impede activation of GIRK channels by G_i/o-coupled GPCRs. We measured GIRK currents in cultured mouse hippocampal neurons via a whole-cell voltage clamp technique. To mimic conditions occurring in patients treated with Li⁺, we incubated the neurons for 3–6 d with 1 mM Li⁺. This reduced GIRK currents elicited by the selective GABA_B agonist baclofen (I_baclofen) or by purinergic A1 receptor agonists by ∼43–50% (Fig. 1 A and B and Fig. S1A). Acute (∼2 min) intracellular application of 2 mM Li⁺, present only in the patch pipette, also significantly reduced I_baclofen (Fig. 1B and Fig. S1F), indicating that Li⁺ was acting intracellularly. Given the limitations of Li⁺ diffusion from the pipette, we assume that Li⁺ concentration was lower in the dendrites, where GIRK channels are preferentially located (24).

Fig. 1.

Li+ modulates GIRK activity in hippocampal neurons. (A) Incubation of neurons with 1 mM Li⁺ decreased I_baclofen measured in low-K⁺ (4 mM K⁺) solution. (Lower) Boxes represent the 25–75% percentiles and the median, the red line indicates the mean, and whiskers indicate the 10–90% percentiles. (Upper) Representative traces without (control) and with Li⁺ incubation. (B) Normalized effects of Li⁺ on neurotransmitter-evoked currents. Acute effects of 2 mM Li⁺ were measured in 25 mM K⁺ solution (Fig. S1E). Experiments with 1 mM Li⁺ incubation were performed in low-K⁺ solution. In each cell, I_evoked in Li⁺ was normalized to the mean value of I_evoked in neurons of the control group (no Li⁺) of the same experiment. (C) Incubation of neurons with 1 mM Li⁺ increased I_basal in low-K⁺ solution. (Upper) I_basal was measured at −120 mV (voltage step from the holding potential of −60 mV), to enhance the sensitivity of measurement (23). (Lower) Representative traces recorded in the low-K⁺ solution before (black) and after (pink) application of sTPNq (TPN). Subtracting the sTPNq-resistant current from total current gives the net GIRK current (bottom traces). (D) Effect of Li⁺ incubation on I_basal: summary of data. (E) Normalized effects of Li⁺ incubation on I_basal. **P < 0.01; ***P < 0.001.

In the absence of agonists, when G protein cycle is not activated (see Fig. 4A), GIRK channels exhibit basal activity (I_basal), which regulates neuronal excitability and plasticity (20, 23, 24). We measured I_basal using a specific GIRK blocker, tertiapin (TPNq) (Fig. 1C). Incubation of neurons with 1 mM Li⁺ for 3–6 d significantly increased I_basal (Fig. 1 C–E). Interestingly, I_basal measured in the low-K⁺ solution was not altered by acute Li⁺ application (2 mM in the pipette) (Fig. S1 B–F). When both I_basal and I_evoked were measured in the same cells and at the same membrane potential in a high-K⁺ solution, acute Li⁺ treatment did not alter I_basal, but did reduce I_evoked (Fig. S1 E and F), indicating that the decrease in I_evoked can occur independently of a change in I_basal. The difference in Li⁺ effects in acute and incubation conditions could result from poor diffusion of Li⁺ to the dendrites in acute application, but a contribution of unknown long-term effects of Li⁺ in the incubation experiments is a distinct possibility. In summary, Li⁺ has two effects on GIRK channels in hippocampal neurons, steadily elevating agonist-independent I_basal but attenuating neurotransmitter-induced inhibitory GIRK signaling.

Fig. 4.

Proposed mechanisms of GIRK-mediated action of Li⁺. (A) Molecular mechanisms of dual Li⁺ action on GIRK channels via G proteins. When the G protein cycle is not activated, basal dissociation of Gα_GDP from Gβγ (reaction 1) (31) is enhanced by the addition of Li⁺ and Na⁺, resulting in increased free Gβγ and activation of GIRK channels (reaction 5). Agonist (A) binds to GPCR (R), and the activated agonist-receptor complex (AR*) binds to Gαβγ heterotrimer (reaction 2), starting the G protein cycle (37). AR* catalyzes the GDP–GTP exchange at Gα (reaction 3) resulting in the departure of A and R and separation of Gα_GTP from Gβγ, which activates GIRK (reaction 5). Activation is terminated by rebinding of Gβγ to Gα_GDP after the hydrolysis of GTP by Gα (reaction 4). Reaction 1 is several orders of magnitude slower than the forward rate of reaction 3, which is rate-limiting in the GPCR-activated cycle; thus, the contribution of reaction 1 to the production of free Gβγ during an active G protein cycle is negligible. Li⁺ inhibits reaction 3, thereby slowing the production of free Gβγ and reducing GIRK activation. (B). Li⁺ reduces the span of inhibitory GIRK-dependent regulations in a single neuron. Li⁺ increases basal inhibitory tonus by increasing GIRK I_basal, but weakens the inhibiting effects of inhibitory G_i/o-coupled neurotransmitters.

Li⁺ Dually Regulates GIRK in Excised Patches via G Protein-Dependent Mechanisms.

We examined the mechanism of action of Li⁺ in excised patches of Xenopus oocytes expressing the GIRK1/2 channel composition, predominant in the hippocampus (20). We first tested whether Li⁺ enhances I_basal in the absence of a GPCR-initiated G protein cycle. The addition of Li⁺ (>0.2 mM) to a guanine nucleotide-free bath solution facing the intracellular side of the excised patch induced GIRK activation, which reached its peak after 4–6 min (Fig. 2 A and B and Fig. S2A). Li⁺ acted by increasing the open probability, producing an increased number and frequency of openings without changing the single-channel current (Fig. S2D). The maximal activation by Li⁺ was ∼3.5-fold; half-maximal activation was achieved at 0.34 mM Li⁺ (K_d,app), within the therapeutic range (Fig. 2C).

Fig. 2.

Li⁺ modulates GIRK in excised patches. (A) Experimental paradigm for studying GPCR-independent basal activity. GIRK channel currents were recorded at −80 mV in cell-attached mode (c.a.), after which the patch was excised, exposing the intracellular side to the bath solution. After 3 min, Li⁺ was applied to the bath solution. (B) Example trace showing enhancement of I_basal by Li⁺. Oocytes expressed GIRK1/2 + Gα_i3. (Insets) Channel activity before and after Li⁺ application. (C) Dose dependence of the effect of Li⁺ on I_basal. Data were fitted to a single-site binding isotherm. B_max, maximal fractional effect. (D) Li⁺ did not activate GIRK when Gα_i3GA was expressed instead of WT Gα_i3. (E) Coexpression of m-cβARK abolished the Li⁺-induced increase in I_basal in the GIRK1/2_D228N channel. (F) Experimental paradigm for studying GPCR-dependent activity. The pipette contained 2 µM ACh. (G) Exemplary traces of the ACh- and GTP-induced GIRK activation from two patches, in control (Left) and 1 mM Li⁺-containing (Right) bath solution. Arrows indicate GTP addition. (Insets) Channel activity before and after GTP application. Oocytes expressed GIRK1/2_D228N, Gα_i3, and m2R. (H) Summary of the effect of Li⁺ on GTP-induced activation at 1–2 min after GTP addition (peak) and 4–5 min after GTP application. *P < 0.05; **P < 0.01.

We tested whether Li⁺ acts on GIRK similarly to another small monovalent cation, Na⁺. Intracellular Na⁺ activates GIRK1/2 and elevates I_basal in a GPCR-independent manner by two mechanisms: a fast mechanism (within seconds; EC₅₀ = ∼20–40 mM) through a direct interaction with GIRK2 (25, 26) and a slow mechanism (within minutes; EC₅₀ = ∼10 mM) through enhanced dissociation of Gα^GDPβγ heterotrimers, which increases free Gβγ concentration (27). The slow time course of Li⁺ activation was inconsistent with the fast mechanism. In support of this, 1 mM Li⁺ activated the mutant channel GIRK1/2_D228N, which is resistant to direct activation by Na⁺ (26) (Fig. 2E and Fig. S2F). Moreover, activation by Li⁺ was strongly enhanced by coexpression of Gα_i3, without which it was barely detectable (Fig. S2 B and C). Activation was suppressed when, instead of WT Gα_i3, we expressed the “constitutively inactive” Gα_i3 mutant Gα_i3G203A (Gα_i3GA) (Fig. 2D and Fig. S2E), which strongly associates with Gβγ (28). Coexpression of the Gβγ scavenger m-cβARK (28) blocked the Li⁺-induced activation of GIRK1/2_D228N (Fig. 2E and Fig. S2F). Using GIRK1/2_D228N activity in cβARK as a reference, activation of GIRK1/2_D228N by 1 mM Li⁺ was ∼3.6-fold, taking into account the 51 ± 11% rundown revealed by cβARK in this channel (but not in the WT GIRK1/2). Taken together, these results rule out direct activation of GIRK by Li⁺ and support a Gα- and Gβγ-dependent mechanism.

To explore the effect of Li⁺ on GPCR-induced activation of GIRK, we devised a protocol to measure (re)activation of G protein cycle by GTP in excised patches (Fig. 2F). Agonist (acetylcholine, ACh) was present in the pipette to activate the GPCR (muscarinic 2 receptor, m2R), causing high GIRK activity in cell-attached patches (Fig. 2G, c.a.). Excision of the patch into a GTP-free bath solution was followed by a strong reduction in activity, in part because GTP was hydrolyzed by Gα. At 3 min after excision, GTP (100 µM) was added, causing a ∼10-fold activation that slowly subsided, reaching a new steady-state level after 3–5 min (Fig. 2 G and H). After the 3-min exposure, 1 mM Li⁺ in the bath solution did not significantly affect basal activity, but strongly reduced the GTP-evoked response (Fig. 2 G and H and Fig. S3), suggesting that the decrease in GPCR-induced I_evoked is independent of the Li⁺-induced increase in I_basal.

Finally, Li⁺ did not affect the direct activation of the channel by purified Gβγ (Fig. S4), suggesting that Li⁺ does not act directly on GIRK, Gβγ, or their interaction but does act directly on Gα. In summary, our results in the model system show that Li⁺ dually regulates GIRK1/2 in a G protein-dependent manner, decreasing agonist-evoked responses and increasing basal activity, like in neurons. Both of these actions of Li⁺ involve intracellular and membrane-delimited mechanisms.

Li⁺ Attenuates Gα^GDP–Gβγ Interaction.

The foregoing functional data strongly suggest that Li⁺ regulates G proteins and subsequent GIRK channel activity by acting on G proteins through more than one mechanism. Previous biochemical studies established the inhibitory effects of Li⁺ on GDP–GTP exchange and GTP binding (5, 22, 30), which may account for Li⁺’s effect on I_evoked; however, whether Li⁺ affects the agonist-independent basal state of the G protein heterotrimer remains unknown.

We used a biochemical approach to test our hypothesis that Li⁺ directly affects the dissociation of Gα^GDP from Gβγ. We monitored the interaction between Gβ₁γ₂ and GST-fused Gα_i3^GDP in a high-K⁺ buffer containing 5 mM Mg²⁺. We used GST-Gα_i3^GDP to pull-down purified Gβ₁γ₂ and found that 1 mM Li⁺ reduced binding (Fig. 3A). The decrease was significant at low Gβγ concentrations (2.5 and 10 nM), but blunted at high Gβγ concentrations. No significant effect of Li⁺ on the binding of Gβγ was observed at a high concentration of GST-Gα_i3^GDP (120 nM) (Fig. S5 A and B). This finding is compatible with a Li⁺-induced reduction in Gα^GDP– Gβγ binding affinity, given that the sensitivity of the pull-down method to changes in affinity is reduced when concentrations of interactors greatly exceed their dissociation constant, with a K_d for Gα_i1^GDP and Gβγ of <8 nM (31).

Fig. 3.

Li⁺ attenuates the binding between Gα_i3^GDP and Gβγ. (A) Application of 1 mM Li⁺ decreased the binding of purified Gβγ to purified GST-Gα_i3^GDP at low Gβγ concentrations, but not at high Gβγ concentrations. (Left) Representative Western blot (divided into two parts taken at different film exposures). (Right) Data summary. (B) Representative autoradiogram showing the effect of 1 mM Li⁺ on the binding of [³⁵S]methionine-labeled IVT-Gβγ or IVT-G1NC to GST-Gα_i3^GDP. (C) Effect of 1 or 20 mM Li⁺ or 20 mM Na⁺ on binding between GST-Gα_i3 and IVT-Gβγ in the absence or presence of 5 mM Mg²⁺. (D) Summary of the experiments shown in B and C. (E and F) Effect of Li⁺ on binding of IVT-Gα_i3, IVT-Gα_i3GA, and IVT-G1NC to purified His₆-tagged Gβγ in the absence or presence of GDP, as indicated. (G) Example gel showing dose dependency of the effect of Li⁺ on binding of GST-Gα_i3 (2.4 nM) and IVT-Gβγ. (H) Dose dependence of the effect of Li⁺. The solid line represents the fit of the data to a single-site binding isotherm. *P < 0.05; **P < 0.01; ***P < 0.001.

To better quantitate changes in Gα-Gβγ binding, we used autoradiography with in vitro translated (IVT) Gβγ or Gα_i3 labeled with [³⁵S]-methionine, which allows a wide dynamic range of quantitative protein measurements. The amount of IVT Gβγ was calibrated to obtain Gβγ concentrations in the low nM range (Fig. S5C). Pull-down of IVT Gβγ with 24 nM GST-Gα_i3 confirmed that 1 mM Li⁺ reduced the Gα^GDP–Gβγ binding by 32 ± 5%. In contrast, 1 mM Li⁺ did not affect the binding of the full cytosolic domain of GIRK1, G1NC (28), to GST-Gα_i3 (Fig. 3 B–D). 20 mM Na⁺ reduced the Gα_i3-Gβγ binding by ∼45%, as shown before (27) (Fig. 3 C, D, and G). Interestingly, 20 mM Li⁺ did not reduce but slightly increased Gα_i3-Gβγ binding (by 18 ± 7.5%), indicating an opposing effect of high concentrations of Li⁺ (Fig. 3 C and D). The attenuation of Gα_i3^GDP-Gβγ binding by 1 mM Li⁺ was further supported by a reverse paradigm, pull-down of IVT Gα_i3 by purified His-Gβγ. 1 mM Li⁺ reduced the Gα_i3^GDP-Gβγ binding by ∼32%, whereas 20 mM Li⁺ was without effect. Importantly, Li⁺ did not affect the interaction between His-Gβγ and IVT G1NC (Fig. 3 E and F). The Li⁺-induced decrease in binding persisted in the absence of added nucleotides (Fig. 3F and Fig. S5E), a condition that mimics the patch clamp experiments shown in Fig. 2 A–D. In this case, most of Gα is presumed to still contain bound GDP. Notably, 1 mM Li⁺ only slightly (<12%) reduced the binding of the constitutively inactive Gα_i3GA mutant to Gβγ both in the presence and absence of GDP (Fig. 3F and Fig. S5E). The impaired dissociation of Gα_i3GA from Gβγ correlates with, and may explain, the suppression of Li⁺-induced enhancement of basal GIRK1/2 activity by this Gα mutant (Fig. 2D and Fig. S2E).

It has been proposed that some of the effects of Li⁺ on G proteins arise from competition between Li⁺ and Mg²⁺ (22). In the absence of Mg²⁺, the ∼30–35% decrease in Gα-Gβγ binding induced by 1 mM Li⁺ or 20 mM Na⁺ persisted, but the increase caused by 20 mM Li⁺ was not observed (Fig. 3 C and D). This suggests that part of the increase in Gα-Gβγ binding by 20 mM Li⁺ could represent displacement of Mg²⁺ from Gα, because Mg²⁺ reduces the binding between Gα^GDP and Gβγ (31). We did not investigate the mechanisms of the effects of high Li⁺ concentrations any further in this study. Importantly, these experiments demonstrate that the effect of therapeutic doses of Li⁺ is Mg²⁺-independent.

We further explored the dose dependency of therapeutic doses of Li⁺ (up to 2 mM) on the binding of IVT-Gβγ to GST-Gα_i3^GDP at low concentration (2.4 nM) to accentuate the Li⁺-induced changes in binding (Fig. 3 G and H). The decrease in binding was already apparent at 0.2–0.5 mM Li⁺, and the reduction was ∼40% with 1 mM Li⁺ or 20 mM Na⁺ (Fig. 3H and Fig. S5D). The calculated apparent K_d for Li⁺ was ∼0.34 mM, comparable to the patch clamp data (compare with Fig. 2C). At 2 mM, the effect of Li⁺ became weaker, and 20 mM Li⁺ produced no change in binding (Fig. S5D), corroborating an opposing action (increase in Gα-Gβγ binding) of high-Li⁺. In summary, our findings indicate that therapeutic doses of Li⁺ significantly decrease the binding between Gα_i3^GDP and Gβγ in a Mg²⁺-independent manner, but do not interfere with the binding between GIRK1 and either Gα_i3 or Gβγ.

Discussion

We report a unique dual regulation of GIRK channels by Li⁺ in neurons. Importantly, both regulations take place at therapeutic doses of Li⁺. We reproduced these regulations in a heterologous system (Xenopus oocytes). We identified the underlying molecular mechanisms, which involve a complex regulation of G proteins via two distinct actions on the Gα subunit, and verified a novel mechanism of action of Li⁺ on Gα–Gβγ interaction by direct biochemical studies. We propose that these findings can help illuminate the etiology of BPD.

Molecular Mechanism of Li⁺ Action.

Biochemical and electrophysiological experiments suggest that Li⁺ regulates the basal and GPCR-evoked activity of GIRK channels by acting on Gα subunit of G_i/o proteins rather than on the channels themselves. Both effects occur through intracellular actions of Li⁺ on G proteins in the absence of cytosol, in a membrane-delimited manner. The proposed molecular mechanism for the action of Li⁺ is shown in Fig. 4A.

The first action of Li⁺ is to increase the basal activity of GIRK channels by elevating free Gβγ in the absence of agonist activation of the G protein cycle. Here Li⁺ acts through a unique Mg²⁺-independent mechanism, reducing the affinity of the Gα^GDP–Gβγ interaction (31) (reaction 1), similar to Na⁺ (27). This mechanism is supported by the following findings: (i) Li⁺ activates GIRK in excised patches in the absence of GPCR agonists and GTP, ruling out an effect on the G protein cycle; (ii) Li⁺ reduces the interaction between purified Gα^GDP and Gβγ in direct binding experiments; (iii) effective doses of Li⁺ for activation of GIRK in excised patches and for the reduction of Gα^GDP–Gβγ binding are similar, reinforcing the similarity of these mechanisms; and (iv) GIRK activation by Li⁺ is supported by Gα_i3, but not by the constitutively inactive mutant of Gα, Gα_i3G203A, which is deficient in Li⁺-induced dissociation from Gβγ (Fig. 3F).

The addition of Gα would be expected to increase levels of undissociated Gαβγ heterotrimer by shifting reaction 1 (Fig. 4A). This would reduce free Gβγ levels but enrich the substrate for Li⁺ action (Gαβγ), thereby enhancing the relative effect Li⁺ in promoting Gαβγ dissociation (27, 29). A modulatory role of Gα_i itself (28, 32) is also plausible, although the mechanisms of Gα regulations of GIRK remain unclear.

A second action of Li⁺ is inhibition of the agonist-evoked GIRK response, I_evoked. We propose that Li⁺ exerts this effect by inhibiting the GDP–GTP exchange at Gα, thereby slowing the agonist- and GPCR-activated G protein cycle (5, 22). This classical mechanism has not been previously examined in the GIRK signaling cascade. In line with this mechanism, we found that the GPCR- and GTP-induced activation of GIRK was reduced by Li⁺, whereas basal activity was not (Fig. 2), and that Li⁺ did not affect Gβγ activation of GIRK (Fig. S4) or the binding of GIRK to Gβγ (Fig. 3), suggesting an effect on Gα.

The two actions of Li⁺ provide an insight into the nature of GIRK's basal activity in cultured hippocampal neurons and Xenopus oocytes. Under the conditions of our experiments, I_basal was largely independent of an active G protein cycle (e.g., initiated by the presence of an ambient G_i/o-activating neurotransmitter); otherwise Li⁺ would inhibit I_basal instead of enhancing it. Thus, I_basal largely relies on the inherent equilibrium between Gα^GDP and Gβγ (reaction 1 in Fig. 4A). During the activity of the G protein cycle, most of the Gαβγ complexes are dissociated into Gα^GTP and Gβγ (33), and reaction 3 becomes the main source of free Gβγ for I_evoked (Fig. 4A). Thus, the mechanisms of I_basal and I_evoked are distinct, and Li⁺ affects them in opposite ways. Nevertheless, if the total Gβγ available for activation of GIRK is limited, as occurs in Xenopus oocytes (34), then the Li⁺-induced increase in I_basal could contribute to the attenuation of I_evoked. However, Li⁺ still caused a robust decrease in I_evoked under experimental conditions that did not change I_basal, suggesting a separate Li⁺ regulation of I_evoked. Taken together, our findings strongly suggest that dual regulation of GIRK by Li⁺ is related to two separate effects of Li⁺ on G proteins.

Li⁺ Regulates GIRK in Neurons: Possible Relevance to BPD.

Regulation of GIRKs by Li⁺ may profoundly affect excitability in neuronal networks (35) and neuronal plasticity (20). In hippocampal neurons, Li⁺ elevated the steady-state GIRK current (I_basal) in the absence of neurotransmitter-induced GPCR activation. This would be expected to have an inhibitory effect on resting neuronal excitability. However, Li⁺ attenuated inhibitory GIRK signaling by G_i/o-coupled GPCRs when the latter were activated by inhibitory neurotransmitters. Such dual regulation would limit the total span of the inhibitory effects of neurotransmitters acting via G_i/o-coupled GPCRs and potentially dampen “swings” in the inhibitory control of individual neurons via GIRK channels (Fig. 4B). On a larger scale, this effect may be related to the overall mood-stabilizing effect of Li⁺. Unlike previously observed regulation of other ion channels, both activation of I_basal and inhibition of I_evoked occurred at therapeutic doses of Li⁺, indicating (patho)physiological relevance. Our findings suggest links among BPD genetics, Li⁺ actions, and neuronal excitability. Indeed, G proteins present an obvious connection between BPD genetics [i.e, ion channels such as GIRK channels and possibly neuronal voltage-gated Ca²⁺ channels regulated by Gβγ (36)] and the putative cellular targets of Li⁺ and BPD, including PI turnover enzymes (through G_q) and the β-arrestin/Akt/GSK3β pathway, which has multiple reciprocal links to GPCRs and G proteins (4).

Materials and Methods

More detailed information on the materials and methods used in this study is provided in SI Materials and Methods. All experiments were approved by Tel Aviv University’s Committee for Animal Use and Care.

Electrophysiology.

Xenopus oocytes were prepared, injected with RNA, and incubated for 2–4 d before use in experiments. The patch-clamp electrode solution contained 146 mM KCl, 1 mM MgCl₂, 1 mM CaCl₂, 2 mM NaCl, 10 mM Hepes/KOH, and 1 mM GdCl₃ (pH 7.5). The bath solution contained 130 mM KCl (146 mM in the experiments shown in Fig. 2 A–D), 2 mM MgCl₂, 10 mM Hepes/KOH, 1 mM EGTA, and 2 mM Mg-ATP (pH 7.5). The treatment-induced changes in activity were calculated as fold change in total open probability (NPo) relative to basal NPo measured during the last minute of recording in the excised mode before the addition of Li⁺ (GIRK1/2_D228N) or to the first minute after the addition of Li⁺ (WT GIRK1/2).

Primary cultures of mouse hippocampal neurons from neonatal mice were kept in culture for 14–21 d before use in experiments. Whole-cell voltage-clamp recordings were performed in low-K⁺ bath solution (145 mM NaCl, 4 mM KCl, 1.8 mM CaCl₂, 1 mM MgCl₂, 5.5 mM d-glucose, and 5 mM Hepes/NaOH; pH 7.4) supplemented with 0.5 μM tetrodotoxin, 0.5 mM kynurenic acid, and sometimes 30 µM picrotoxin. The patch pipette solution contained 130 mM K-gluconate, 6 mM NaCl, 1 mM EGTA, 1 mM MgCl₂, 10 mM Hepes, 2 mM MgATP, 0.3 mM Tris-GTP, and 0.01 mM Tris-GDP (pH 7.3). In some experiments, a high-K⁺ bath solution (25 mM KCl) was used. Reagent concentrations were baclofen, 100 µM; adenosine, 100 µM; CCPA (Sigma-Aldrich), 10 µM; and sTPN-q and rTPN-q (Alomone Laboratories), 100–120 nM.

Biochemistry.

Interaction of GST-Gα_i3 or His-Gβγ with purified proteins or the IVT [³⁵S]-methionine–labeled proteins was studied by pull-down on glutathion-affinity or Ni-affinity beads, respectively, in a high-K⁺ buffer (150 mM KCl, 50 mM Tris, 5 mM MgCl₂, and 1 mM EDTA, pH 7.0, supplemented with 0.1% Lubrol and 90 µM GDP) unless indicated otherwise (27). The addition of more than 2 mM LiCl or NaCl was compensated for by a reduction in K⁺.

Statistics.

Data are presented as mean ± SEM or, when the data did not pass the normal distribution test, as boxplots including raw data. Multiple-group comparisons were done using one-way ANOVA followed by the Tukey or Dunnet test. Pairwise comparisons were done using the nonpaired one-tailed t test (comparing two groups of cells) or paired t test (comparing data before and after treatment in same cells). When data did not pass the normal distribution test, multiple comparisons were performed using the Kruskal–Wallis test, and pairwise comparisons were performed using the Mann–Whitney or paired t test on ranks.