Parvalbumin interneurons gate amygdala excitability and response to chronic stress via kainate receptor-driven tonic GABAB receptor-mediated inhibition

Abstract

Amygdala hyperexcitability is a hallmark of stress-induced anxiety disorders. Stress-associated changes in both principal neurons and interneurons contribute to the increased excitability, but how exactly these mechanisms interact to regulate the function of behaviorally relevant circuits in the amygdala remains unclear. Here, we show that GluK1 subunit-containing kainate receptors in parvalbumin (PV) interneurons maintain high GABA release and control excitability of lateral amygdala (LA) principal neurons via tonic GABA_B-receptor-mediated inhibition. Downregulation of GluK1 expression in PV interneurons after chronic restraint stress (CRS) releases the tonic inhibition and increases excitability of LA principal neurons. Stress-induced LA hyperexcitability was associated with increased glutamatergic transmission to central amygdala PKCδ-expressing neurons, implicated in fear generalization. Consistent with significance in anxiogenesis, absence of GluK1-GABA_B regulation confers resilience against CRS-induced LA hyperexcitability and anxiety-like behavior. Our data reveal a unique novel mechanism involving an interplay between glutamatergic and GABAergic systems in the regulation of amygdala excitability in response to chronic stress.

Fig. 1. Loss of GABA_B R-mediated tonic inhibition contributes to amygdala hyperexcitability after chronic stress.

A A scheme of the experimental protocol (top) and a graph illustrating body weight changes during the chronic restraint stress (CRS, n = 10, control n = 10; RM ANOVA F_{(1, 18)} = 10.54, **p = 0.0045). B Results of the open field (OF) test. The graphs show the total time spent in the center area of the open field (OF), the center time in 5 min bins (control, n = 10, CRS, n = 10, t-test t = 3.232, df = 18, **p = 0.0046; RM ANOVA F_{(1, 18)} = 10.45, **p = 0.0046) and total distance traveled during OF test (t-test, t = 4.546, df = 18, ***p = 0.0003). C Action potential (AP) firing rate of principal neurons (PN) in the lateral amygdala (LA) in response to depolarizing current steps, recorded from brain slices of control and CRS-exposed mice (control, n = 17 (8 mice), CRS, n = 19 (7 mice); RM ANOVA F _{(1, 29)} = 4.877, *p = 0.035). The example traces illustrate the response to 240 pA step current. D Representative traces and pooled data illustrating the effect of CRS on sIPSC frequency and amplitude in LA PNs. Recordings were done using high-Cl containing electrode filling solution and in the presence of antagonists for AMPA and NMDA receptors (control, n = 12 (4 mice), CRS, n = 15 (4 mice), frequency: t-test, t = 1.381, df = 25, p = 0.1796; amplitude: t-test, t = 1.276, df = 25, p = 0.2136). The decay time distribution of sIPSCs for the same data is shown below (multiple t-test, Holm-Sidak, 1 ms: t = 3.070 df = 390.0, *p = 0.027; 2 ms : t = 3.250 df = 390.0, *p = 0.016). E Tonic GABA_B receptor-mediated currents recorded from LA PNs in response to application of GABA_B antagonist CGP55845 (10 μM), in control and CRS-exposed mice as well as in the presence of GDP-β-S in control mice. All recordings were done in the presence of 50 μM of D-AP5, 200 μM picrotoxin, and 50 μM GYKI 53655 to block NMDA, GABA_A, and AMPA receptors, respectively. Pooled data on the maximal amplitude of the GABA_B current under control conditions and in the presence of GDP-β-S (750 µM) in the electrode filling solution (control, n = 8 (3 mice), GDP-β-S n = 5 (3 mice); t-test, t = 4.599, df = 11, ***p = 0.0008). Amplitudes of the tonic GABA_B current in control and CRS-exposed animals (control, n = 17 (7 mice), CRS, n = 14 (3 mice); t-test t = 3.786, df = 29, ***p = 0.0007). F Effect of GABA_B antagonism on firing frequency of LA PNs in control and CRS-exposed mice, as well as in the presence of GDP-β−S in control mice. Action potential frequencies in response to depolarizing current steps were recorded from brain slices under control conditions (in ACSF) and in the presence of CGP55845 (5 µM) (control, n = 10 (4 mice), control+CGP55845, n = 10 (4 mice); RM ANOVA, F_{(2, 26)} = 3.498, *p = 0.045; GDP-β−S n = 12 (3 mice), GDP-β−S+CGP55845 n = 10 (3 mice); RM ANOVA F_(1,20) = 0.8588, p = 0.365; CRS, n = 17 (6 mice), CRS+CGP55845, n = 9 (3 mice); RM ANOVA, F_{(1, 24)} = 1.824, p = 0.1894). Example traces show the response to 240 pA current step. All the data are presented as mean ± SEM.

Fig. 2. GABA_B R-dependent tonic inhibition is driven by GluK1 kainate receptors in PV interneurons.

A Tonic GABA_A currents recorded as a change in the holding current in response to the application of 25 μM bicuculline in LA PNs, in slices from mice with floxed GluK1 gene (Grik1^fl/fl) and mice lacking GluK1 expression selectively in PV interneurons (PV-Cre::Grik1^fl/fl). All recordings were done at −90 mV holding potential in the presence of 50 μM of D-AP5 and 50 μM GYKI 53655 to block NMDA and AMPA receptors, respectively. The graph illustrates averaged data on the amplitude of the tonic GABA_A current (Grik1^fl/fl: n = 10 (4 mice); PV-Cre::Grik1^fl/fl: n = 9 (3 mice); t-test, t = 1.681, df = 17, p = 0.11). B Tonic GABA_B currents recorded from LA PNs in slices from mice with floxed GluK1 gene (Grik1^fl/fl) and mice lacking GluK1 expression selectively in PV interneurons (PV-Cre::Grik1^fl/fl), before and after ACET application (200 nM). All recordings were done at −50 mV holding potential in the presence of 50 μM of D-AP5, 200 μM picrotoxin, and 50 μM GYKI 53655 to block NMDA, GABA_A, and AMPA receptors, respectively. The amplitude of the tonic current was measured as a change in the holding current in response to the application of 10 μM CGP55845. The graph illustrates averaged data on the amplitude of the tonic GABA_B currents under various conditions (Grik1^fl/fl: control, n = 17 (5 mice), ACET, n = 11 (4 mice); PV-Cre::Grik1^fl/fl: control, n = 16 (4 mice), ACET, n = 7 (3 mice); Dunnett test, ***p < 0.001, **p = 0.0095). C Action potential firing rate of LA PNs in response to depolarizing current steps under control conditions and in the presence of ACET (200 nM). Recordings were done in brain slices from Grik1^fl/fl and PV-Cre::Grik1^fl/fl mice (Grik1^fl/fl: control, n = 17 (5 mice), ACET, n = 17 (5 mice), RM ANOVA, F_{(1, 32)} = 5.967, *p = 0.02; PV-Cre::Grik1^fl/fl: control, n = 24 (8 mice), ACET, n = 22 RM ANOVA, F_{(1, 44)} = 1.534, p = 0.22). D Action potential (AP) firing rate of LA PNs in response to depolarizing current steps under control conditions and in the presence of CGP55845 (5 µM), in PV-Cre::Grik1^fl/fl mice (control, n = 15 (3 mice), CGP55845, n = 13 (3 mice), RM ANOVA, F_{(1, 44)} = 0.1779, p = 0.6766). The example traces in C and D illustrate the response to the 240 pA current step. All the data is presented as mean ± SEM. See Supplementary Data 2 for the membrane properties of LA principal neurons in Grik1^fl/fl and PV-Cre::Grik1^fl/fl mice.

Fig. 3. GluK1 kainate receptors regulate action potential-induced and asynchronous GABA release in PV interneurons.

A A scheme of the experimental protocol (left). AAV viral vectors encoding for Cre-dependent ChR2 were injected to BLA of PV-Cre mice for PV neuron-specific expression of ChR2. GABAergic responses were recorded from LA PNs in response to light stimulation. B Examples of IPSCs, evoked by paired-pulse stimulation with 470 nm blue light before and after 200 nM ACET application. Pooled data on the effect of ACET on the 1st IPSC amplitude and paired-pulse ratio (PPR) (n = 9 (5 mice), PPR, t = 4.567, df = 8, **p = 0.0018; IPSC, t = 3.246, df = 8, *p = 0.0118 paired t-test). C Asynchronous barrage of IPSCs recorded from LA PNs in response to 750 ms opto-stimulation of PV interneurons, before and after application of 200 nM of ACET. Pooled data illustrating the frequency of IPSCs, normalized to the level before opto-stimulation and analysed in 10 min bins under control conditions and in the presence of ACET (n = 8 (6 animals), RM ANOVA, F _{(1, 7)} = 11.49, *p = 0.012). D The decay time distribution of sIPSCs at baseline, and during 10 s period after opto-stimulation of PV interneurons. Optogenetic activation of the PV interneurons increased the relative occurrence of sIPSCs with fast decay time (multiple t-test, Holm-Sidak; 1 ms : t = 3.486, df = 130.0, **p = 0.008). E The decay time distribution of sIPSCs at baseline and in the presence of ACET, during periods when opto-stimulation was not applied. In the presence of ACET, there was a lower percentage of 1 ms decay time sIPSC events (multiple t-test, Holm-Sidak, t = 3.566, df = 299.0, **p = 0.005466). All the data are presented as mean ± SEM.

Fig. 4. Grik1 expression and function in PV interneurons is downregulated after chronic stress.

A Representative images illustrating triple in situ hybridization staining for Grik1 (magenta), Gad1 (white), and Pvalb (green) in the LA in control and CRS-treated mice. The image with merged channels also shows DAPI staining (blue). Yellow arrows in the merged image point to cells co-expressing Grik1, Gad1 and Pvalb. B Bar charts summarizing the density of cells expressing Grik1 and Pvalb in the LA in control and CRS-treated mice. The Grik1 expression level is expressed as a percentage of DAPI stained nuclei, and Pvalb as a percentage of all Gad1 positive GABAergic cells (control, n = 20 sections (3 mice), CRS, n = 21 sections (3 mice), Grik1: t-test, t = 0.04723, df = 39, p = 0.9626; Pvalb: t = 0.05947, df = 42 p = 0.9529). The values represent mean ± SEM. C Bar charts summarizing the percentage of PV neurons (Gad1+Pvalb+) and other subtypes of GABAergic interneurons (Gad1+Pvalb-) co-expressing Grik1 mRNA in LA of control and CRS-exposed mice (Pvalb+: control, n = 14 section (3 mice), CRS, n = 15 (3 mice), Mann-Whitney U = 38.50, **p = 0.001; Gad1+Pvalb+: control, n = 13 section (3 mice), CRS, n = 15 (3 mice), GAD: t = 1.594, df = 26, p = 0.1230). The values represent mean ± SEM. D High-magnification images illustrating the expression of Grik1 mRNA (magenta) in individual Pvalb positive LA neurons (left), and in Pvalb negative, Gad1 expressing neurons (Pvalb, green, Gad1, white). The image with merged channels also shows DAPI staining (blue). Pooled data summarizing the average intensity of Grik1 mRNA staining (# dots) per cell, in PV neurons (Pvalb+) and in other subtypes (Gad1+Pvalb-) of GABAergic interneurons in LA of control and CRS animals (Pvalb+: control, n = 22 sections from 3 animals, CRS, n = 19 sections from 3 animals, Mann-Whitney test, U = 128, *p = 0.034; Pvalb-Gad1+: control, n = 14 sections from 3 animals, CRS, n = 15 sections from 3 animals, Mann-Whitney test, U = 100.5, p = 0.8554). Data are presented as median and quartile. Histogram demonstrating the distribution of Grik1 mRNA staining intensity in individual PV neurons in control and CRS mice (control, n = 60 cells, CRS, n = 46 cells, 3 animals in both groups, Kolmogorov-Smirnov test, D = 0.4841, ****p < 0.0001). E Current clamp recordings from PV interneurons in acute slices from control and CRS-exposed PV-TdTomato mice. Example traces illustrate the response of the PV neurons to depolarizing currents steps (50 pA and 200 pA), for control and CRS groups. Pooled data on the action potential frequencies in response to depolarizing current steps (control, n = 17 (7 mice), CRS, n = 18 (6 mice), RM ANOVA F_{(1, 32)} = 4.370, *p = 0.0446). F Example traces illustrate the response of the PV neurons before and after ACET application (50 pA current step), in control and CRS-exposed PV-TdTomato mice. Pooled data on the action potential frequencies in response to depolarizing current steps (control: n = 13 (6 mice), F_{(1, 12)} = 11.36, **p = 0.0056; CRS: n = 11 (6 mice), RM ANOVA F_{(1, 10)} = 4.112, p = 0.0701). All the data presented as mean ± SEM. Data on resting membrane potential (Vm) and properties of the action potentials (AP) (rheobase, threshold and half-width) for LA PV interneurons in control and CRS-exposed mice is shown in Supplementary Data 2.

Fig. 5. Chronic stress and loss of the GluK1-GABA_B-dependent tonic inhibition regulates LA output to CeL.

A Voltage clamp recordings of sEPSCs from PKCδ+ and PKCδ- neurons in CeL, in acute slices from control and CRS-exposed PKCδ-TdTom mice. Pooled data on the sEPSC frequency in the two cell types (PKCδ+: control, n = 19 (7 mice), CRS, n = 16 (5 mice), t-test, t = 3.302, df = 33, **p = 0.0023; PKCδ-: control, n = 9 (4 mice), CRS, n = 9 (5 mice), t-test, t = 1.848, df = 6, p = 0.1140). B The experimental approach for chemogenetic inhibition of the BLA PNs in PKCδ-Cre mice. AAV viral vectors encoding for inhibitory DREADD receptor hM4Di under the CaMKII promoter were injected into the BLA to target PNs. PKCδ+ neurons in the CeL were visualized by injection of AAV viral vectors encoding Cre-dependent EGFP. C The effect of CNO (10 μM) on resting membrane potential (RMP, recorded under whole cell current clamp) and spontaneous AP firing (cell-attached recording) in hM4Di expressing BLA PNs. CNO application resulted in 6.4 ± 1.8 mV hyperpolarizing shift in the membrane potential and 28 ± 10% reduction in the frequency of spontaneous AP firing (RMP: n = 6 (3 mice), paired t-test, t = 3.468, df = 5, *p = 0.0179; AP : n = 13 (3 mice), one sample t-test, t = 2.655, df = 12, *p = 0.021). D Examples of sEPSC recordings from PKCδ+ neurons in slices from control and CRS-exposed mice, expressing the inhibitory DREADD receptor hM4Di in the BLA PNs, at control conditions and in the presence of CNO (10 µM). Pooled data on the effect of CNO on sEPSC frequency (control: baseline, n = 11, CNO, n = 10 (4 mice), paired t-test, t = 2.746, df = 19, *p = 0.0128; CRS: baseline, n = 10, CNO, n = 8 (4 mice), paired t-test, t = 4.493, df = 16, ***p = 0.0004). Comparisons between control vs. CRS: unpaired t-test, t = 3.836, df = 19, **p = 0.0011; control + CNO vs. CRS+CNO: unpaired t-test, t = 1.204, df = 16, p = 0.2463. E Example traces and pooled data on sEPSC recordings from PKCδ+ neurons before and after application of ACET (200 nM) in control and CRS-exposed animals (control: n = 12 (5 mice), paired t-test, t = 4.452, df = 11, ***p = 0.001; CRS: n = 7 (5 mice), paired t-test, t = 0.7174, df = 6, p = 0.50). F Examples of sEPSC recordings from PKCδ+ neurons before and after application of ACET (200 nM) in mice expressing hM4Di in the BLA, in absence or continuous presence of CNO. Pooled data on the effect of ACET on sEPSC frequency (hM4Di: n = 6 (4 mice), paired t-test, t = 4.881, df = 5, **p = 0.0045; hM4Di + CNO: n = 6 (4 mice), paired t-test, t = 1.447, df = 5, p = 0.2074). All the data presented as mean ± SEM.

Fig. 6. Mice lacking Grik1 expression in PV interneurons are resistant to stress-induced alterations in amygdala excitability and behavior.

A Body weight changes during chronic restraint stress (CRS) protocol (Grik1^fl/fl: control n = 17, CRS n = 19; RM ANOVA, F _{(1, 34)} = 21.32, ****p < 0.0001; PV-Cre::Grik1^fl/fl: control n = 19, CRS n = 17; RM ANOVA; F _{(1, 34)} = 23.89, ****p < 0.0001). B Action potential frequencies in response to depolarizing current steps recorded from brain slices of control and CRS-treated, Grik1^fl/fl or PV-Cre::Grik1^fl/fl mice (Grik1^fl/fl: control, n = 25 (7 mice), CRS, n = 17 (6 mice), RM ANOVA, F _{(1, 40)} = 4.775, *p = 0.035; PV-Cre::Grik1^fl/fl: control, n = 19 (6 mice), CRS, n = 14 (4 mice), RM ANOVA, F _{(1, 31)} = 0.0001180, p = 0.99). C Tonic GABA_B currents recorded from LA PNs in slices from control and CRS-treated, Grik1^fl/fl or PV-Cre::Grik1^fl/fl mice. All recordings were done at −50 mV holding potential in the presence of 50 μM of D-AP5, 200 μM picrotoxin, and 50 μM GYKI 53655 to block NMDA, GABA_A, and AMPA receptors, respectively. The amplitude of the tonic current was measured as a change in the holding current in response to the application of 10 μM CGP55845. The graph illustrates averaged data on the amplitude of the tonic GABA_B currents under various conditions (Grik1^fl/fl: control, n = 15 (5 mice), CRS, n = 11 (3 mice); PV-Cre::Grik1^fl/fl: control, n = 15 (4 mice), CRS, n = 18 (3 mice); two-way ANOVA for genotype and condition, genotype effect F_(1,54) = 19.93_, ****p < 0.0001; post-hoc comparison *** p = 0.0003, Bonferroni). D Current clamp recordings from PV interneurons in acute slices from control and CRS-exposed PV-Cre::Grik1^fl/fl mice. Example traces illustrate the response of the PV neurons to depolarizing current steps (75 and 200 pA), for control and CRS groups. Pooled data on the action potential frequencies in response to depolarizing current steps (control, n = 12 (3 mice), CRS, n = 12 (3 mice), Mixed-effects model (REML) F_{(1, 22)} = 1.157, p = 0.2938). E Results of open field (OF) test. The graphs show the total time spent in the center area of the open field (OF) and the center time in 5 min bins, for control and CRS-exposed mice for the two genotypes, Grik1^fl/fl and PV-Cre::Grik1^fl/fl (Grik1^fl/fl: control, n = 16, CRS, n = 19; PV-Cre::Grik1^fl/fl: control, n = 19, CRS, n = 17). Total time in the center field: Two-way ANOVA: treatment F_{(1, 67)} = 3.086, p = 0.0835, genotype F_{(1, 67)} = 0.7517, p = 0.389, interaction F_{(1, 67)} = 4.133, p = 0.046. * p = 0.019, Holm-Sidak. Time in center zone in 5 min bins: RM ANOVA, Grik1^fl/fl: F_{(1, 34)} = 8.564, **p = 0.0062; PV-Cre::Grik1^fl/fl : F _{(1, 34)} = 0.03320, p = 0.8566. Total distance traveled: 2-way ANOVA, genotype F_(1,67) = 31.74, ****p < 0.0001, treatment F_{(1, 67)} = 1.828, p = 0.1780, interaction F_{(1, 67)} = 1.495, p = 0.2442. All the data are presented as mean ± SEM.