Extrasynaptic GABA_A receptors in mediodorsal thalamic nucleus modulate fear extinction learning

To test the effect of GABA_ARs activities in the MD on fear extinction learning, we injected gabazine, a GABA_AR inhibitor, into the MD of wild-type mice (Figure 1A-B). In day 1 (conditioning), we exposed the cannula-implanted mice to three trials of tone (CS), each co-terminated with an electric foot shock (US). In day 2 (extinction learning), 30 min after injection of gabazine or vehicle through the cannula, the fear-conditioned mice were exposed to 20 repeated CS-only trials. In the extinction learning gabazine-injected mice showed a faster decrease in freezing levels compared to the mice in the vehicle-injected group (two-way repeated measures ANOVA, group effect, F1,21 = 71.745, P < 0.001; group × trial interaction, F21,399 = 5.414, P < 0.001; Figure 1C). In comparison with the mice in the vehicle-injected group, the gabazine-injected mice showed lower initial levels of freezing (t test; P = 0.0001; Figure 1C). Therefore, to compare the rates of within-session extinction between the two groups, the freezing levels for trials in each group were normalized to the value for the first trial of each group (Figure 1D) following the convention [25, 26]. There was a difference between groups (two-way repeated measures ANOVA, group effect, F1,21 = 55.024, P < 0.001; group × trial interaction, F21,399 = 5.318, P < 0.001), indicating that gabazine injection into the MD has changed the extinction rate. In the third day (extinction recall), mice were re-exposed to the CS without the US. The gabazine-treated group showed significantly lower freezing levels than the vehicle-injected group (t test, P = 0.0001; Figure 1C), indicating that gabazine facilitated extinction learning in the mice.

To rule out the possible confounding effect of GABA_ARs activity in the MD on the locomotion and anxiety level of the mice, we injected gabazine or vehicle into the MD of the mice 30 min before performing an open field test. The two groups did not differ in total distance moved during a 1 hr period in the open field arena (t test, P = 0.365; Figure 1E), and showed no significant difference in the percentage of time spent in the central area compared to the whole arena in the first 5 min (Mann–Whitney Rank Sum Test, P = 0.758; Figure 1F), indicating that the lower freezing response of gabazine-injected mice was not the result of an increased locomotor activity in the extinction box or decreased anxiety.

The α4β2δ eGABA_ARs are highly expressed in the thalamus including the MD [27, 28]. To test whether eGABA_ARs activity in the MD could affect fear extinction learning, we injected THIP (4,5,6,7-tetrahydroisoxazolo[5,4-c]pyridin-3-ol), a selective GABA_AR agonist with a preference for δ-subunit containing eGABA_ARs [29], into the MD 30 min before fear extinction learning. During the extinction learning, the THIP-injected mice showed less decrease in freezing levels than the vehicle-injected group (two-way repeated measures ANOVA, group effect, F1,16 = 32.315, P < 0.001; group × trial interaction, F16,304 = 0.665, P < 0.852; Figure 2A). Due to higher freezing levels in the first trial in the THIP-injected group compared with the vehicle-injected group (t test; P = 0.00003; Figure 2A), we further examined the rates of within-session extinction between the two groups by normalizing freezing levels to that of the first trial for each group (Figure 2B) following the convention [25, 26]. There was significant difference between the two groups in the rates of within-session extinction (two-way repeated measures ANOVA, group effect, F1,16 = 12.127, P = 0.003; group × trial interaction, F16,304 = 1.065, P < 0.387). This indicates that THIP injection changed within-session extinction rate. In the extinction recall, the THIP-treated group showed significantly higher freezing levels than the vehicle-injected group (t test, P = 0.002; Figure 2A), indicating that THIP attenuated extinction learning in the mice.

To check the possible effect of eGABA_ARs activity in the MD on the locomotion and anxiety level of the mice, we injected THIP or vehicle into the MD of the mice 30 min before performing an open field test. The two groups did not differ in locomotion during a 1 hr period (t test; P = 0.200; Figure 2C), and showed no significant difference in the percentage of the time spent in the central area compared to the whole arena in the first 5 min (t test; P = 0.702; Figure 2D), which means that the higher freezing response of the THIP-injected mice was not due to a change in the locomotor activity or anxiety.

As thalamic expression of the α4-subunit is limited to the extrasynaptic area [16, 27], we tested the behavioral effect of MD-specific KO of GABRA4 gene in the fear extinction paradigm. AAV-mCherry (control virus) or AAV-mCherry-Cre viral particles were injected into the MD of GABRA4 ^flx/flx mice 3 weeks before fear conditioning (Figure 3A). Immunohistochemistry was used to test Cre expression and functionality. Cre expression can be seen in the mCherry-expressing cells in the MD region of the mice injected with AAV-mCherry-Cre (Figure 4A-B). A lack of GABRA4 expression in the injected side of the MD in brain slices from the Cre virus-injected mice is evident when compared with the contralateral non-injected side or slices from the control virus-injected mice (Figure 4C).

There was no difference between the two groups in fear conditioning (two-way repeated measures ANOVA, group effect, F1,15 = 4.109, P = 0.061; group × trial interaction, F15,60 = 0.365, P = 0.832; Figure 3B). There was no significant difference in the initial freezing levels between the two groups (t test; P = 0.186; Figure 3B). However, the mice injected with AAV-mCherry-Cre showed faster extinction learning compared to the control group (two-way repeated measures ANOVA, group effect, F1,15 = 9.947, P = 0.007; group × trial interaction, F15,285 = 1.111, P = 0.338; Figure 3B). In the extinction recall test, the group injected with AAV-mCherry-Cre showed significantly lower freezing levels than the control virus-injected group (t test; P = 0.018; Figure 3B), indicating that GABRA4-deletion facilitated extinction learning in the mice.

The two groups did not differ in locomotion (t test, P = 0.679; Figure 3C) or the percentage of the time spent in the central area compared to the whole arena in the first 5 min (Mann–Whitney Rank Sum Test, P = 0.880; Figure 3D), indicating that the lower freezing response of the Cre virus-injected mice was not the result of increased locomotor activity or decreased anxiety.

Adult (12–16 weeks old) male wild-type F1 (B6 × 129) hybrid mice, used for the drug infusion experiments, were obtained by mating parental strains of C57BL/6J and 129s4/svJ. For the virus injection experiments, adult male (12–16 weeks old) B6.GABRA4 ^flx/flx mice were used. B6.GABRA4 ^flx/flx mice were produced by mating B6.GABRA4 ^flx/+ mice. Mice had ad libitum access to food and water and were housed under a 12:12 hr light/dark cycle (lights on at 8 AM). Animal care and experiments followed the guidelines from the Institutional Animal Care and Use Committee of Institute for Basic Science.

The mice went through fear conditioning, extinction, and extinction recall as described previously [5]. Briefly, the mice were conditioned in conditioning chamber (context A) using three trials of CS + US: tones (3 KHz, 30 s, and 85 dB) co-terminating with foot shocks (0.35 mA, 1 s). For the B6GABRA4 ^flx/flx mice, five trials of tone and foot shock (0.7 mA) were used. The intertrial interval was 120 s. After 24 hr, mice were exposed to twenty CS-only (tone) trials in the extinction box (context B) with an intertrial interval of 5 s. For the extinction recall test, mice were exposed to the tone on the third day (context B).

Mice behavior was recorded with a video camera to score freezing (lack of movement except for respiration) using the FreezeFrame software (Actimetrics, Coulbourn Instruments, PA, USA).

Locomotor activity was tested in an open field box (40×40×50 cm) as reported previously, one week apart from fear conditioning and extinction [5]. The ratio of the time spent in the central 20×20 cm area to the time spent in the whole 40×40 cm arena in the first 5 min was used as a measure of anxiety. Automated measurements were done using EthoVision software (Noldus Information Technology, Netherlands).

Mice were anesthetized using intraperitoneal injection of 2% Avertin solution (tribromoethyl alcohol/tertiary amyl alcohol, vol/vol, Sigma-Aldrich), and framed in stereotaxic device (Kopf instruments, CA, USA). Mouse surgery for unilateral implantation of guide cannula was done according to the protocols [5], targeting the MD (AP: −1.5 mm, ML: 0.3 mm, and DV: −3.3 mm from the brain surface). Gabazine or THIP (from Sigma-Aldrich, St. Louis, MO, USA) were dissolved in normal saline, and a volume of 0.3 μl (30 and 50 μM, respectively) was infused (0.1 μlmin⁻¹) through a 33-gauge injection cannula inserted in the guide cannula. The injector was kept in place for 3 min after the end of the injection. Histologic confirmation of the infusion position was carried out and only the data from properly injected mice were used for statistical analysis.

For AAV2 production, we used the pAAV-MCS plasmid (Stratagene, La Jolla, CA) carrying the cDNA for mCherry-Cre (NLS-Cre was kindly provided by Dr. C. Justin Lee, KIST) or mCherry downstream of the CMV promoter. For recombinant virus generation, AAV-293 cells were co-transfected with pAAV-RC (Stratagene) encoding the AAV genes rep and cap and the helper plasmid (Stratagene) encoding E24, E4 and VA. Virus particles were purified and concentrated according to the procedure [38]. The AAV viral titer was determined using a QuickTiterTM AAV Quantitation Kit (Cell Biolabs, San Diego, CA) and a minimum titer of 10¹¹/ml was obtained. The AAV-mCherry control or AAV-mCherry-Cre virus was injected in the MD region (AP: −1.5 mm, ML: 0.3 mm, and DV: −3.3 mm from the brain surface) of B6GABRA4 ^flx/flx mice unilaterally, using a Hamilton syringe connected to a microinjection pump (sp100i instrument, WPI, USA) with a volume of 0.6 μl at a rate of 0.1 μlmin⁻¹. The needle was held in place for 10 min after completion of the injection to permit the diffusion of the viral particles into the brain tissue.

To check Cre and α4-subunit expression in the MD, we performed histology as described previously [39]. Mice were perfused transcardially with cold saline, followed by 4% paraformaldehyde in phosphate-buffered saline (PBS) (10 mM, pH 7.4) under Avertin anesthesia. Brains were postfixed in 4% paraformaldehyde overnight at 4°C. Coronal sections (50 μm thickness) of the MD were prepared using a vibratome (VT1200S, Leica, Germany).

For diaminobenzidine (DAB) immunostaining, free-floating sections were initially washed (3×) in PBS, and then cell membranes were permeabilized using PBS supplemented with 0.1% Triton X-100 (PBST) for 10 min and then incubated in 0.3% H₂O₂ in PBS (20 min). Next, tissue was blocked (1 h) in 5% goat serum⁄PBS and incubated (overnight, 4°C) in a rabbit polyclonal anti α4-subunit antibody (1:1000 final dilution, ab4120 abcam, UK). After 3× washes in PBS, sections were incubated (2 h) at room temperature (25°C) in biotinylated anti-rabbit IgG (1:500, Vector Laboratories, Burlingame, CA, USA) and then placed in an avidin⁄biotin horseradish peroxidase complex for 1 h (prepared according to the manufacturer’s instructions, Vector Laboratories). The signal was visualized by the addition of diaminobenzidine-nickel-intensified substrate (Vector Laboratories, USA), mounted on gelatin-coated slides and coverslipped with Permount media (Fisher Scientific, Houston, TX, USA).

For fluorescent immunolabeling against Cre, tissue sections were incubated (overnight, 4°C) with mouse monoclonal anti-Cre antibody (1:1000, MAB3120, EMD Millipore, Germany). The sections were washed with PBS and then incubated (2 h at room temperature) with an Alexa 488-conjugated anti-mouse IgG antibody (1:500, Molecular Probes, Eugene, OR, USA). Sections were washed in PBS, and then incubated (30 min) with the DNA stain 4′,6-diamidino-2-phenylindole (DAPI) (Thermo Scientific, USA), washed in PBS again, and then mounted with VECTASHIELD (Vector Laboratories, Inc. Burlinggame, CA, USA).

Figures were prepared using Adobe Photoshop CS6 (Adobe Systems Inc., USA) and Microsoft Office 2010 (Microsoft Corporation, Redmond, USA). Image manipulation was restricted to threshold and brightness adjustments applied to the entire image.