Behavioral characterization of mice lacking GIRK/Kir3 channel subunits


*K. Wickman, Department of Pharmacology, University of Minnesota, 6-120 Jackson Hall, 321 Church Street S.E., Minneapolis, MN 55455, USA. E-mail:


G protein-gated inwardly rectifying K+ (GIRK/Kir3) channels mediate the postsynaptic inhibitory effects of many neurotransmitters and drugs of abuse. The lack of drugs selective for GIRK channels has hindered our ability to study their contributions to behavior. Here, we assessed the impact of GIRK subunit ablation on several behavioral endpoints. Mice were evaluated with respect to open-field motor activity and habituation, anxiety-related behavior, motor co-ordination and ataxia and operant performance. GIRK3 knockout (−/−) mice behaved indistinguishably from wild-type mice in this panel of tests. GIRK1−/− mice and GIRK2−/− mice, however, showed elevated motor activity and delayed habituation to an open field. GIRK2−/− mice, and to a lesser extent GIRK1−/− mice, also displayed reduced anxiety-related behavior in the elevated plus maze. Both GIRK1−/− mice and GIRK2−/− mice displayed marked resistance to the ataxic effects of the GABAB receptor agonist baclofen in the rotarod test. All GIRK−/− mice were able to learn an operant task using food as the reinforcing agent. Within-session progressive ratio scheduling, however, showed elevated lever press behavior in GIRK2−/− mice and, to a lesser extent, in GIRK1−/− mice. Phenotypic differences between mice lacking GIRK1, GIRK2 and GIRK3 correlate well with the known impact of GIRK subunit ablation on neurotransmitter-gated GIRK currents, arguing that most neuronal GIRK channels contain GIRK1 and/or GIRK2. Altogether, our data suggest that GIRK channels make important contribution to a range of behaviors and may represent points of therapeutic intervention in disorders of anxiety, spasticity and reward.

The slow postsynaptic inhibitory effects of many neurotransmitters are mediated by G protein-gated inwardly rectifying K+ (GIRK/Kir3) channels (North 1989). GIRK channels are activated by Gi/o G proteins and are formed by products of four genes (GIRK1–GIRK4) (Mark & Herlitze 2000). Three channel subunits (GIRK1–GIRK3) exhibit broad distributions in the central nervous system, whereas GIRK4 expression is limited (Karschin et al. 1996). The overlapping distribution of neuronal GIRK messenger RNAs suggests the potential for considerable diversity among neuronal GIRK channels (Karschin et al. 1996).

In the absence of selective drugs, GIRK knockout (GIRK−/−) mice have helped to link channel subtypes to specific neurotransmitters and behaviors. Studies involving neurons from GIRK−/− mice argue that GIRK2 contributes to most, if not all, neuronal GIRK channels (Cruz et al. 2004; Koyrakh et al. 2005; Labouebe et al. 2007; Luscher et al. 1997; Slesinger et al. 1997; Torrecilla et al. 2002). Not surprisingly, GIRK2−/− mice display several deficits including seizures (Signorini et al. 1997), hyperactivity (Blednov et al. 2001b), hyperalgesia (Marker et al. 2002) and blunted analgesia (Marker et al. 2004; Mitrovic et al. 2003). Neuronal GIRK channels are thought to consist of GIRK1 and GIRK2 (Koyrakh et al. 2005; Liao et al. 1996; Marker et al. 2004), and the similar electrophysiological profiles of neurons from GIRK1−/− mice and GIRK2−/− mice support this contention (Koyrakh et al. 2005; Marker et al. 2006). In midbrain dopamine neurons, however, GIRK2 is expressed in the absence of GIRK1 (Cruz et al. 2004; Eulitz et al. 2007; Karschin et al. 1996).

GIRK3 contributes to both channel trafficking (Lunn et al. 2007; Ma et al. 2002) and channel formation (Cruz et al. 2004; Jelacic et al. 2000; Koyrakh et al. 2005; Labouebe et al. 2007; Lunn et al. 2007; Torrecilla et al. 2002). Surprisingly, GIRK3 ablation had little impact on GIRK current in the hippocampus, substantia nigra pars compacta (SNc) or spinal cord (Koyrakh et al. 2005; Marker et al. 2006). In the locus coeruleus, however, GIRK3 ablation did correlate with reduced opioid inhibition (Torrecilla et al. 2002). Furthermore, loss of GIRK3 led to an increase in the potency of the GABAB receptor agonist baclofen in dopamine neurons of the ventral tegmental area (VTA) (Labouebe et al. 2007). These observations suggest a role for GIRK3 in the behavioral effects of drugs of abuse, a possibility supported by recent behavioral studies (Morgan et al. 2003; Smith et al. in press).

While focused studies have been conducted with all three GIRK−/− lines, a comparative behavioral characterization has not been performed. This type of study was delayed until each line had undergone extensive backcrossing against a well-characterized inbred mouse strain (C57BL/6J). Here, we describe the performance of GIRK−/− mice in established behavioral assays, including open-field activity, rotating rod, elevated plus maze (EPM) and operant tests. Our data highlight the contribution of GIRK channels to a range of behaviors and suggest that therapeutic interventions targeting GIRK channels could be useful in a variety of clinical settings, including anxiety and addiction.

Materials and methods

Experimental subjects

All animal use was reviewed and approved by the University of Minnesota Institutional Animal Care and Use Committee and carried out in accordance with National Institutes of Health guidelines. Efforts were made to minimize the pain and discomfort of the animals throughout the study and when possible, to reduce the number of animals used in each test. Mice were housed in groups (two to four) on a 12-h light/dark cycle with food and water available ad libitum. The generation of GIRK−/− mice was described previously (Bettahi et al. 2002; Signorini et al. 1997; Torrecilla et al. 2002). GIRK−/− lines were backcrossed through at least 12 generations against the C57BL/6J mouse strain prior to beginning this study. Studies were conducted between 0800 and 2000 h, and each test involved a different cohort of adult wild-type and GIRK−/− mice (8–10 weeks) naïve to experimental manipulation. Male mice were evaluated in open-field, EPM and operant tests. Male and female mice were evaluated in the rotating rod test; data were pooled given the lack of gender differences apparent in this study. The wild-type control groups for the studies reported below consisted of siblings derived from crosses of mice heterozygous for each null allele. In addition, some of the wild-type mice generated in these crosses were bred to yield entire litters of wild-type mice that were subsequently used in behavioral tests. All GIRK2−/− mice evaluated in these studies were generated by crossing heterozygous mice. GIRK1−/− mice and GIRK3−/− groups consisted of a mix of mice derived from heterozygous and homozygous crosses. No significant differences were detected among same-genotype offspring derived from different cages.

Locomotor activity

Activity studies were performed in Plexiglas open-field environments (ENV-515; 17″ W × 17″ L × 12″ H; Med Associates, Inc., St. Albans, VT, USA) enclosed in melamine sound-attenuating cubicles. Each cubicle was equipped with a ventilated cover and three 16-beam infrared arrays permitting automated measurements of distance traveled and rearing (open field activity software package version 4.2; Med Associates, Inc.). On test day, mice (n = 16–27 per genotype) were habituated to testing room environment for 1 h. Mice were then placed in the open-field chamber, and activity was monitored for 60 min. Each subject was evaluated during three sessions performed in the same open-field environment on three consecutive days. An index of thigmotaxis (wall hugging) was generated for each subject by dividing the distance traveled in the 7.5-cm wide perimeter of the environment by the total distance traveled during the 60-min session (Clement et al. 1995). The number of entries in the central area of the open field (total area minus outer perimeter) was also measured. In order to differentiate between large (ambulatory) and small (stereotypic) movements, analysis parameters were set such that a subject had to cross four beams (7.5 cm) within 500 milliseconds to maintain its ambulatory status. Stereotypic movements were quantified as movements (such as grooming and head weaving) performed in a given area, without triggering ambulatory counts.

Elevated plus maze

Three days before testing, wild-type and GIRK−/− mice (n = 15–32 per genotype) were transferred from group to individual housing. On test day, mice were acclimated to the procedure room for 1 h. Mice were then placed in the EPM (Columbus Instruments, Inc., Columbus, OH, USA) center facing an open arm, and their activity was recorded for 5 min by video camera. Elevated plus maze data were scored during two rounds of videotape playback per subject by an investigator blind to subject genotype, with time spent in open and closed arms measured by stopwatch. In round 1, time spent in the open arms and open-arm entries was assessed. In round 2, time spent in the closed arms and closed-arm entries was counted. Entry to or exit from a region of the maze was counted if the base of the tail crossed the boundary between regions.


Wild-type and GIRK−/− mice (n = 28–69 per genotype) were habituated for 1 h to the testing room prior to training or testing. On day 1, mice were trained during two sessions separated by 2 h to endure for 300 seconds on a rod rotating at 12 rpm (Ugo Basile, Comerio, Italy); falls during training sessions were recorded. On day 2, the performance of all subjects was evaluated. Subjects achieving an endurance score of 300 seconds received a subcutaneous (s.c.) injection of either saline or (R+)-baclofen (1, 3, 6, 9 or 12 mg/kg; Sigma, St Louis, MO, USA). Baclofen solution was prepared fresh each day with sterile 0.9% saline. Endurance score (time spent on the rod in seconds) was recorded at 1, 2, 3 and 4 h postinjection. If mice spent more than 80% of the total time clinging to the rod, their endurance score was considered equal to zero (Jacobson & Cryan 2005).

Operant behavior

Conditioned operant behavioral studies were conducted with mouse operant chambers (Med Associates, Inc.) as described (Morgan et al. 2003). Wild-type and GIRK−/− mice (n = 7–18 per genotype) were housed individually for 60 h, and food restricted to 3 g/day for 36 h, prior to the first operant session. Food pellets (20 mg, PJA/100020 dust-free pellets, Noyes Precision Pellets; Research Diets, Inc., New Brunswick, NJ, USA) used in the study were presented to mice 1 day before testing to avoid food neophobia phenomena (Hayward & Low 2007). Subjects were trained to lever press for food pellets under a fixed ratio (FR) 1 schedule of reinforcement, during daily 3-h sessions (phase 1). Correct responding on the active lever resulted in stimulus light illumination for 5 seconds and delivery of one food pellet to the food hopper. Responses on the inactive lever resulted in stimulus light illumination but did not lead to food pellet delivery. The criteria for the acquisition of food-reinforced behavior were discrimination (≥3:1 ratio) between the active and the inactive levers and delivery of ≥30 pellets for three consecutive sessions. Active and inactive lever responding as well as pellets earned during phase 1 were averaged over the three sessions. Animals that acquired these criteria within 9 days proceeded to phase 2 (Morgan et al. 2003). In phase 2, mice were evaluated in three sessions under FR 5 scheduling. In phase 3, performance was evaluated using a within-session progressive ratio (PR) schedule to test the reinforcing effect of food (phase 3). During the PR test, the response requirement for each successive food pellet increased as follows: 1, 2, 4, 6, 9, 12, 15, 20, 25, 32, 40, 50, 62, etc. (Richardson & Roberts 1996). Progressive ratio performance was assessed during 3-h sessions on three consecutive days under both restricted and ad libitum feeding conditions. Active and inactive lever responding as well as pellets earned during the final two sessions for phases 2 and 3 were averaged.

Statistical analysis

Data are expressed throughout as the mean ± SEM. Data were analyzed by spss, version 11.0 (SPSS©, Chicago, IL, USA) software. Most data from each study were analyzed by one-way analysis of variance (anova), with genotype as grouping factor, followed by Dunnett’s post hoc test to compare individual GIRK−/− lines with wild type and Tukey’s Honestly Significant Differences (HSD) test to probe for differences between GIRK−/− lines. To evaluate habituation in the motor activity test, one-way repeated measure anova was performed for each genotype. The effect of genotype on falls from the rotarod during training sessions and on the number of subjects failing to master the task was analyzed by Kruskal–Wallis one-way anova on ranks because of failure of normality in assumption testing followed by Dunn’s method for post hoc comparisons (Jacobson & Cryan 2005). No difference in the performance of males and females in the rotarod test was detected by independent t-test; data from both genotypes were combined to increase the power of the study. The effect of genotype on the number of subjects failing to acquire operant responding (phase 1) was analyzed in the same fashion.


The goal of this work was to compare the performance of wild-type and congenic GIRK−/− mice in behavioral paradigms that can show significant motor, cognitive or emotional impairments. Wild-type and GIRK−/− mice were assessed in open-field activity, EPM, rotating rod and food-reinforced operant conditioning tests. Different cohorts of experimentally naïve mice were evaluated in each study in order to avoid interference and cross-sensitization between different tests. GIRK−/− mice used in this study were generated following extensive backcrossing (≥12 rounds) against the C57BL/6 mouse strain, mitigating concerns about inter-subject differences in genetic background.

Open-field activity

To determine whether GIRK subunit ablation impacts activity of mice in a novel environment, we evaluated total distance traveled, thigmotaxis, stereotypy and the number of entries made into the central zone of a novel open field during 60-min sessions on three consecutive days. Over the three sessions, the behavior of GIRK3−/− mice was indistinguishable from wild-type controls. Several differences between wild-type, GIRK1−/− and GIRK2−/− mice, however, were evident. Consistent with published observations (Blednov et al. 2001b, 2002), we found that GIRK2−/− mice were hyperactive in the open field. On day 1, total distance traveled by GIRK2−/− mice as well as GIRK1−/− mice was significantly greater than that seen for wild-type counterparts (Fig. 1a, left). Over the next two sessions, GIRK1−/− mice and GIRK2−/− mice continued to display higher levels of horizontal motor activity compared with wild-type mice. Furthermore, GIRK2−/− mice exhibited delayed and/or incomplete between-session habituation with respect to total distance traveled. While there were only minor differences between groups with regard to thigmotaxis (Fig. 1a, right) and total stereotypic counts (Fig. 2a, left), genotype-dependent differences in the number of central zone entries mirrored those seen with total distance traveled (Fig. 2b, right). GIRK1−/− mice and GIRK2−/− mice exhibited elevated numbers of central zone entries on all 3 days; significant differences were measured on day 1 (GIRK1−/− mice, P < 0.05) and day 2 (GIRK1−/− mice and GIRK2−/− mice, P < 0.05).

Figure 1.

Open-field activity of wild-type and GIRK−/− mice. (a) Left panel – Total distance traveled (m) per 60-min session by wild-type (white), GIRK1−/− (light gray), GIRK2−/− (dark gray) and GIRK3−/− (black) mice. GIRK1−/− (139 ± 6 m, n = 27; P < 0.05) and GIRK2−/− (124 ± 7 m, n = 16; P < 0.05) mice displayed elevated horizontal motor activity in the open field on day 1 relative to wild-type controls (93 ± 5 m, n = 22) as well as on subsequent test days. Right panel – Thigmotaxis is expressed as a ratio of distance traveled in the periphery to total distance traveled. On day 3, GIRK2−/− mice displayed slightly reduced thigmotaxis (0.65 ± 0.03, P < 0.05) compared with wild-type controls (0.73 ± 0.02). (b) Left panel – Stereotypic counts per session. GIRK1−/− mice showed elevated stereotypic counts on day 2 (5758 ± 293, n = 27; P < 0.05) relative to wild-type controls (4637 ± 249, n = 22). Right panel – Total number of entries into the central area. On day 1, GIRK1−/− mice displayed more entries (500 ± 31, P < 0.05) than wild-type mice (375 ± 25). On day 2, both GIRK1−/− mice (401 ± 35, P < 0.05) and GIRK2−/− mice (368 ± 48, P < 0.05) made more entries into the center compared with wild-type mice (199 ± 24). *P < 0.05 vs. wild-type mice; P < 0.05 vs. GIRK1−/− mice.

Figure 2.

Elevated plus maze performance of wild-type (wt) and GIRK−/− mice. (a) The time spent in open and closed arms during the 5-min (300-second) test. GIRK1−/− mice (59 ± 6 seconds, n = 31; P < 0.05) and GIRK2−/− mice (136 ± 16 seconds, n = 14; P < 0.05) showed significantly elevated time spent in open arms compared with wild-type controls (34 ± 5 seconds, n = 32). Correspondingly, time spent in closed arms by GIRK1−/− and GIRK2−/− mice was significantly reduced. (b) The number of open- and closed-arm entries made during the test. GIRK1−/− mice (7 ± 1, P < 0.05) and GIRK2−/− mice (11 ± 2, P < 0.05) also made significantly more open-arm entries than wild-type mice (3 ± 0.4). The number of closed-arm entries did not differ between genotypes. *P < 0.05 vs. wild-type mice; P < 0.05 vs. GIRK1−/− mice.

Anxiety-related behavior

GIRK−/− mice were evaluated next using the EPM, a test of anxiety-related behavior in rodents (Belzung & Griebel 2001). As with the open-field test, GIRK3−/− mice did not differ from wild-type mice with respect to performance in the EPM test. Consistent with published literature (Blednov et al. 2001b), GIRK2−/− mice exhibited reduced anxiety-related behavior, spending more time in the open arms of the maze than wild-type controls, at the expense of time spent in the closed arms (Fig. 2a). GIRK2−/− mice also made more open-arm entries than wild-type mice (Fig. 2b, left). Closed-arm entries did not differ between GIRK2−/− mice and wild-type mice (Fig. 2b, right), indicating that the difference seen with respect to open-arm entries was not a reflection of general hyperactivity in GIRK2−/− mice. GIRK1−/− mice behaved in a qualitatively similar manner to GIRK2−/− mice, spending more time in the open arms and exhibiting more open-arm entries than wild-type mice, while not differing in the number of closed-arm entries. Despite the qualitative similarity in their performance, GIRK2−/− mice displayed significantly less anxiety-related behavior than GIRK1−/− mice.

Co-ordination and ataxia

Muscle co-ordination was evaluated using the rotating rod (rotarod) test using methods described previously for C57BL/6 mice (Jacobson & Cryan 2005). Genotype-dependent differences in the number of falls observed per subject during training sessions were not observed (data not shown). Similarly, the number of subjects failing to complete the task on test day did not differ across genotypes (one to three animals per genotype). Thus, none of the GIRK−/− lines displayed a co-ordination impairment that would preclude their evaluation using the rotarod test.

Next, subjects were challenged with the GABAB receptor agonist baclofen, which induces a well-characterized ataxic response in rodents (Jacobson & Cryan 2005; Jacobson et al. 2006). Each subject received only one dose of baclofen (0, 1, 3, 6, 9, 12 mg/kg s.c.). No animal, regardless of genotype, showed any ataxic effect of baclofen in the rotarod test at 1 or 3 mg/kg. Conversely, all subjects showed some degree of ataxia at the 12-mg/kg dose. Overall, GIRK3−/− mice performed indistinguishably from wild-type animals at all baclofen doses tested (Fig. 3a–c). GIRK1−/− mice and GIRK2−/− mice, however, displayed blunted ataxic responses to baclofen that were most evident at the 9-mg/kg dose (Fig. 3a–c). Reduced sensitivity to baclofen in GIRK1−/− mice and GIRK2−/− mice was suggested by the accelerated time–course of recovery from the ataxia (Fig. 3b). At 9 mg/kg, endurance scores for GIRK1−/− mice and GIRK2−/− mice returned to normal by 3 h postinjection. Cumulative endurance scores, obtained by summing the endurance values for each dose at each time-point, summarize the impact of GIRK subunit ablation on the ataxic effect of baclofen (Fig. 3c).

Figure 3.

Co-ordination and ataxia in wild-type and GIRK−/− mice. Endurance scores represent the time (in seconds) spent on the rod at a given time-point. Group sizes ranged from six to 20 per genotype and baclofen dose. (a) Dose–response curve displaying group endurance scores 1 h after baclofen injection. Note that GIRK1−/− mice and GIRK2−/− mice exhibited blunted sensitivity to the ataxic effect of baclofen at 9 and 12 mg/kg. (b) Time–course of recovery from baclofen-induced (9 mg/kg) ataxia. Endurance scores at each time-point (1, 2, 3 and 4 h) are plotted for each genotype. Note the accelerated recovery exhibited by GIRK1−/− and GIRK2−/− mice. (c) Cumulative ataxia scores for wild-type and GIRK−/− mice. Cumulative scores for each genotype and baclofen dose were determined by summing the endurance scores measured at each time-point. The maximum possible cumulative score in this assay was 1200 seconds (4 × 300 seconds). *P < 0.05 vs. wild-type mice.

Operant test

To evaluate the impact of GIRK subunit ablation on the ability of mice to perform a conditioned instrumental task, wild-type and GIRK−/− mice were trained in a three-phase, food-reinforced operant test involving lever pressing for palatable food pellets (Hayward & Low 2007). Mice were trained in this task using an FR 1 schedule of reinforcement (phase 1). Once acquisition criteria were satisfied, animals were switched to an FR 5 schedule of reinforcement (phase 2), prior to testing using a within-session PR schedule of reinforcement under both restricted and ad libitum feeding conditions (phase 3). Subjects’ body weights were held at ∼85% of free-feeding values during phases 1 and 2 in an effort to stimulate robust responding, as described (Carroll 1994; Carroll & Lac 1997, 1998; Morgan et al. 2003).

The number of days required to complete phase 1 and the number of subjects not reaching acquisition criteria (one to two per genotype) were indistinguishable across the genotypes, indicating that GIRK subunit ablation did not significantly impair associative learning or the ability of the subjects to perform an operant task. Consistent with the results of the other tests, the performance of GIRK3−/− mice did not differ from wild-type mice during phase 1, 2 or 3. GIRK2−/− mice, and to a lesser extent GIRK1−/− mice, showed increased active lever responding and food pellets earned at FR 1 and FR 5 during phases 1 and 2 in comparison to wild-type mice (Table 1). Responding on the inactive lever during phases 1 and 2 did not differ for wild-type, GIRK1−/− and GIRK2−/− mice, indicating that differences seen in active lever responding did not reflect general differences in activity levels. During phase 3, GIRK2−/− mice displayed elevated active lever responding and food pellets earned, particularly under food-restricted conditions (Fig. 4). While significant differences between wild-type and GIRK1−/− mice were not observed during phase 3, GIRK1−/− mice tended to exhibit elevated responding (P = 0.062) under ad libitum food access conditions (Fig. 4b).

Table 1.  Operant responding for food
GroupPhase 1Phase 2
nAge (days)Weight (g)Rate (days)Lever pressesPelletsLever pressesPellets
  • Ages at the start of study were not significantly different across genotypes, ranging from 12 to 16 weeks. Note that the free-feeding body weights of GIRK1−/− and GIRK2−/− mice were lower than mice of other genotypes prior to testing. During phases 1 and 2, mice were restricted to 3 g/day of food plus the earned food during sessions. There were no significant genotype-dependent differences observed with respect to the extent of weight loss associated with food restriction in this study; all subjects were maintained at ∼85% of free-feeding body weight during phases 1 and 2 of this study (not shown).

  • *

    P < 0.05 vs. wild-type mice.

Wild type1896 ± 328.3 ± 0.66 ± 185 ± 75 ± 258 ± 4238 ± 136 ± 244 ± 2
GIRK1−/−1895 ± 325.5 ± 0.7*6 ± 1119 ± 12*7 ± 169 ± 3319 ± 18*11 ± 257 ± 3*
GIRK2−/−7100 ± 326.3 ± 0.77 ± 2134 ± 16*6 ± 286 ± 10*395 ± 46*6 ± 267 ± 9*
GIRK3−/−1496 ± 329.4 ± 0.86 ± 167 ± 66 ± 148 ± 3204 ± 136 ± 139 ± 3
Figure 4.

Operant responding for food: phase 3. Performance of wild-type (wt) and GIRK−/− mice during phase 3 (PR) of the operant task involving food as the reinforcing agent. Performance in phase 3 was evaluated under both food-restricted (a) and ad libitum (b) food access conditions. Lever responding and pellets earned were tabulated during 3-h sessions; values measured on the final 2 days of testing were averaged for each subject. (a) GIRK2−/− mice responded more on the active lever (893 ± 186, n = 7; P < 0.05) and earned more pellets (16 ± 1, n = 7; P < 0.05) than wild-type mice (active lever: 447 ± 61; pellets: 13 ± 1) under mild food restriction. No differences were seen with regard to the number of inactive lever presses (not shown). (b) GIRK2−/− mice responded more on the active lever (263 ± 62, P < 0.05) than wild-type mice (active lever: 142 ± 24) under ad libitum feeding conditions. GIRK2−/− mice also responded more on the inactive lever (28 ± 15, P < 0.05) during phase 3 compared with wild-type mice (9 ± 2) (not shown). *P < 0.05 vs. wild-type mice.


We measured the impact of genetic ablation of GIRK1, GIRK2 and GIRK3 in mice using established behavioral paradigms. Assays were chosen that would provide insight into the contribution of GIRK channels to activity, anxiety, muscle co-ordination, ataxia and reward-related behavior. We found that GIRK1−/− mice and GIRK2−/− mice often displayed robust and similar phenotypes, including elevated open-field activity, decreased anxiety-like behavior, decreased baclofen ataxia and increased operant responding for food.

Overall, GIRK2−/− mice exhibited the most pronounced phenotypes in this study. These observations are consistent with the view that GIRK2 contributes to channel formation in most neuron populations that exhibit a GIRK conductance (Cruz et al. 2004; Koyrakh et al. 2005; Luscher et al. 1997; Slesinger et al. 1997; Torrecilla et al. 2002). GIRK2−/− mice have displayed phenotypes in many behavioral tests. For example, GIRK2−/− mice displayed blunted behavioral responses to ethanol (Blednov et al. 2001a), blunted hypothermic responses to agonists for several G protein-coupled receptors (Costa et al. 2005), thermal and chemical hyperalgesia (Marker et al. 2002, 2004; Mitrovic et al. 2003), blunted antinociceptive responses to several analgesic compounds (Blednov et al. 2003; Marker et al. 2004), reduced cocaine self-administration (Morgan et al. 2003) and enhanced seizure susceptibility (Signorini et al. 1997).

Limited information is available concerning the neurobehavioral profile of GIRK1−/− mice. Like the GIRK2−/− line, GIRK1−/− mice exhibited thermal hyperalgesia and decreased analgesic responses to high doses of intrathecal opioids (Marker et al. 2004). Here, we found that GIRK1−/− mice and GIRK2−/− mice exhibited similar phenotypes with regard to open-field activity and habituation, anxiety and baclofen-induced ataxia. Thus, our work supports the contention that GIRK1 interacts with GIRK2 to form functional channels in most neurons (Karschin et al. 1996; Koyrakh et al. 2005; Liao et al. 1996). Nevertheless, two phenotypes evident in GIRK2−/− mice were attenuated in GIRK1−/− mice. While both lines showed reduced anxiety-related behavior, this phenotype was more pronounced in GIRK2−/− mice. Similarly, GIRK1−/− mice exhibited elevated operant responding for food, but their performance was less pronounced than that of GIRK2−/− mice. There are two likely explanations for these observations. First, some neurons express GIRK2 but not GIRK1, such as the dopamine neurons of the VTA and SNc (Cruz et al. 2004; Eulitz et al. 2007; Inanobe et al. 1999; Karschin et al. 1996). Ventral tegmental area dopamine neurons are vital to reward-related behaviors, including self-administration of food and drugs of abuse (Kalivas & McFarland 2003; Koob 1992; Marinelli et al. 2006; Self 2004; Self & Nestler 1995). Second, although neurotransmitter-induced currents are reduced in neurons from GIRK1−/− mice, the reduction is not always as dramatic as that seen in GIRK2−/− mice (Marker et al. 2006). This is most likely because of the presence of residual current carried by channels formed by GIRK2 and/or GIRK3. Indeed, GIRK2 contains a forward trafficking signal that promotes membrane targeting, while GIRK1 contains an endoplasmic reticulum retention signal and requires coexpression with another GIRK subunit to achieve membrane localization (Hedin et al. 1996; Kennedy et al. 1999; Lunn et al. 2007; Ma et al. 2002).

Behavioral abnormalities were not detected in GIRK3−/− mice in this study. While this may be because of compensatory adaptations that minimize the impact of GIRK3 ablation, some phenotypes have been reported in this mouse line. For example, GIRK3−/− mice exhibited reduced cocaine self-administration (Morgan et al. 2003). In addition, GIRK3−/− mice exhibited thermal hyperalgesia in the hot plate but not the tail flick test (Marker et al. 2002, 2004). Furthermore, the analgesic potency of systemic morphine was reduced in GIRK3−/− mice (Smith et al. in press). Because GIRK3−/− mice showed normal tail flick behavior and responses to intrathecal morphine (Marker et al. 2004), it seems likely that supraspinal GIRK3-containing channels mediate in part the analgesic effect of systemic morphine. Given the otherwise normal behavior of GIRK3−/− mice, these phenotypes suggest the intriguing possibility that GIRK3-containing channels make a selective contribution to the behavioral effects of drugs of abuse.

The mutant mouse lines evaluated in this study were generated using embryonic stem cells from 129-based substrains and were backcrossed against the C57BL/6 strain for 12–22 generations prior to testing. The goal of backcrossing is to minimize inter-subject differences in genetic content that could influence behavioral outcomes. Despite extensive backcrossing, it is not possible to completely eliminate genetic content from the stem cell donor strain using this approach. This is important as differences in the expression level or polymorphisms in ‘hitchhiking’ genes could explain the phenotypic differences between the parent 129 and the C57BL/6 strains and the differences between GIRK−/− mice and wild-type mice. With respect to the paradigms used in this study, behavioral differences have been noted for 129 and C57BL/6 strains. For example, 129 strain shows reduced locomotor activity and heightened anxiety relative to the C57BL/6 strain and a diminished capacity for associative learning (Bothe et al. 2004, 2005; Hagenbuch et al. 2006; Hengemihle et al. 1999; Kelly et al. 2003; Paulus et al. 1999; Rodgers et al. 2002; Tarantino et al. 2000). Thus, contributions of trait-relevant hitchhiking gene(s) in our study are predicted to manifest as hypoactivity, heightened anxiety and diminished operant performance. In contrast, the phenotypes displayed by GIRK−/− mice included heightened locomotor activity, reduced anxiety-related behavior and elevated operant responding for food. Thus, the most parsimonious interpretation of the data is that the lack of the GIRK subunit in question is the dominant source of the phenotypes.

Two aspects of the operant study merit discussion. First, clear genotype-dependent differences in body weights were observed prior to the study. GIRK1−/− and GIRK2−/− mice weighed less than wild-type and GIRK3−/− mice. Thus, the impact of underweight and caloric imbalance on operant performance needs to be considered. Group PR performance was qualitatively similar under both food-restricted and ad libitum conditions, however, arguing that the caloric imbalance cannot fully account for the observed phenotypes. Nevertheless, persistent caloric imbalance (hunger) or elevated metabolism in one or more GIRK−/− lines could impact performance in this test. Certainly, the contribution of GIRK channels to energy homeostasis warrants further examination. Second, the performance of GIRK3−/− mice was consistently lower, although not significantly, than that of wild-type mice in this test. Interestingly, Lüscher and colleagues showed recently that the GIRK channel in VTA dopamine neurons (normally formed by GIRK2 and GIRK3 subunits) from GIRK3−/− mice was more sensitive to GABAB receptor stimulation than GIRK channels in wild-type VTA dopamine neurons (Labouebe et al. 2007). Given that GIRK2 ablation leads to a complete loss of GIRK current in VTA dopamine neurons, it is tempting to speculate that the extent of operant behavior directed at earning food may be inversely proportional to strength of receptor–GIRK channel coupling that exists in VTA dopamine neurons.

Both GIRK1−/− mice and GIRK2−/− mice showed decreased anxiety-like behavior. Given the constitutive and global nature of the GIRK−/− lines, it is not possible to identify the circuit(s) or neurotransmitter system(s) that explain the anxiolytic phenotype. Nevertheless, the elevated anxiety-related behaviors seen in GABAB(1) receptor knockout mice argue that the loss of GABAB–GIRK signaling is an unlikely explanation for the anxiolytic phenotype in GIRK−/− mice (Mombereau et al. 2004). Although other explanations are tenable, our observations are more consistent with those of a recent study showing that chronic administration of the selective serotonin reuptake inhibitor fluoxetine exerts a beneficial influence on a rodent model of depression through suppression of GIRK-dependent signaling in the dorsal raphé nucleus (Cornelisse et al. 2007). Acute inhibition of GIRK channels by fluoxetine was also reported in other systems (Takahashi et al. 2006). Thus, the potential contribution of GIRK channels to emotional disorders and the actions of a common class of antidepressant compounds warrant additional attention.

In clinical settings, baclofen is used to treat spasticity and muscular rigidity associated with cerebral palsy, muscular sclerosis, amyotrophic lateral sclerosis and spinal cord injury (Brogden et al. 1974; Vacher & Bettler 2003). Baclofen also exerts potent antinociceptive effects at the spinal level (Slonimski et al. 2004). Side-effects of baclofen are pronounced, however, and include sedation, hypothermia, amnesia, dizziness, headache, insomnia and sexual dysfunction (Denys et al. 1998; Kofler et al. 2002). These and other unfavorable properties limit its use in clinical medicine and behavioral pharmacology (Vacher & Bettler 2003). Here, we show that GIRK channels, likely formed by GIRK1 and GIRK2, make a significant contribution to baclofen ataxia. Thus, it is worth considering the possibility that drugs that directly activate GIRK channels may evoke the beneficial effects of baclofen (e.g. analgesia and muscle relaxation) while showing diminished off-target effects.

In summary, our findings implicate GIRK channels in motor activity, anxiety, reward and baclofen ataxia. The channels involved in these behaviors are likely formed by GIRK1 and/or GIRK2. Future studies will strive for increased resolution with regard to GIRK modulation. The site-specific injection of drugs and/or genetic reagents that perturb GIRK-dependent signaling with molecular and spatial precision in wild-type mice will be quite informative and will obviate concerns linked to constitutive gene ablation.


The authors gratefully acknowledge the assistance of Dr Cheryl Marker and Cydne Perry with the EPM and operant tests, respectively. The authors thank Dr Paul Slesinger for providing helpful comments on the manuscript. This work was supported by National Institutes of Health grants MH61933 and DA011806 (K.W.) and a predoctoral fellowship from the Epilepsy Foundation (M.P.).