Regulating anxiety with extrasynaptic inhibition

Abstract

Aversive experiences can lead to complex behavioral adaptations including increased levels of anxiety and fear generalization. The neuronal mechanisms underlying such maladaptive behavioral changes, however, are poorly understood. Here, using a combination of behavioral, physiological and optogenetic approaches in mouse, we identify a specific subpopulation of central amygdala neurons expressing protein kinase C δ (PKCδ) as key elements of the neuronal circuitry controlling anxiety. Moreover, we show that aversive experiences induce anxiety and fear generalization by regulating the activity of PKCδ(+) neurons via extrasynaptic inhibition mediated by α5 subunit-containing GABAA receptors. Our findings reveal that the neuronal circuits that mediate fear and anxiety overlap at the level of defined subpopulations of central amygdala neurons and demonstrate that persistent changes in the excitability of a single cell type can orchestrate complex behavioral changes.

Figure 1

Fear conditioning enhances anxiety. a, Schematic illustrating the experimental paradigm. Animals were exposed to five CSs or five CS-US pairings. Twenty four hrs later, anxiety behavior was assessed on the EPM, and fear generalization was quantified in a retrieval test carried out in a novel context. b, Freezing values (percent of time) of the two experimental groups (CS only, n =11 mice, and CS-US, n =12 mice) during CS⁺ and CS^- exposure. ***P < 0.001, paired t-test for the CS-US group, t = 8.599, 11 degree of freedom. c, Left, Example EPM trajectories of animals previously exposed to CS only or to CS-US pairings. Right, Animals subjected to CS-US pairings (n = 12 mice) exhibit enhanced anxiety behavior on the EPM as compared to animals exposed to CS only (n = 11 mice; ***P < 0.001, Mann-Whitney rank sum unpaired t-test). d, Overall EPM track length is the same for CS only and CS-US groups (P = 0.865, unpaired t-test, t = 0.144, 21 degrees of freedom). e, Regression analysis reveals a significant correlation between EPM anxiety behavior and fear generalization for individual animals. White symbols represent values from individual animals. Black symbols represent binned averages of animals exhibiting different levels of fear generalization (0-0.2, 0.2-0.4, and > 0.4). Linear regression: R = 0.6, P = 0.003, n = 12 mice (CS-US group). f, There was no correlation between EPM anxiety behavior and freezing to the CS⁺ and CS^- for the CS only group (n = 11 mice, CS only group; freezing to the CS^- vs. time spent in open arms: R = 0.298, P = 0.260; freezing to the CS⁺ vs. time spent in open arms: R = 0.287, P = 0.335. White symbols represent values from individual animals. All error bars indicate mean ± s.e.m.

Figure 2

Bi-directional regulation of fear generalization and anxiety through optogenetic control of CEA PKCδ⁺ neurons. a, Stainings for Cyan Fluorescence Protein (CFP, green) and endogenous PKCδ(red) confirm overlap of CFP with endogenous PKCδ (98% CFP⁺ cells were PKCδ⁺; 100% PKCδ⁺ cells were CFP⁺). White dashed lines denote the contour of CEA (CEl and CEm). b, Left, Representative cartoon illustrating the injection of conditional AAVs expressing ChR2-YFP or ARCH-GFP into CEA. Middle, Overlap of endogenous PKCδ (red) and ChR2-YFP (green). Right, Overlap of endogenous PKCδ (red) and ARCH-GFP (green). Arrows indicate cells with overlap, red triangles point to cells stained only for PKCδ. Scale bar: 20 μm. c, Top, Experimental configuration illustrating in vivo optrode implants; Bottom, Light-responsive units were identified using 300 ms light pulses followed by 3 min continued test stimulations as used during behavioral experiments. d, Top, Raster plot illustrating firing of an identified ChR2-expressing PKCδ⁺ neuron before, during and after 300 ms blue light stimulation; Bottom, PSTH of the z-score shows a marked increase in firing during the 300 ms light pulse. Inset, z-score of the short-latency light-induced responses. e, Rate histogram average before, during and after 3 min light stimulation on identified ChR2-expressing PKCδ⁺ neurons (n = 5 cells/2 mice). f, z-score histogram illustrating increased firing of ChR2-expressing PKCδ⁺ neurons during 3 min light stimulation (n = 5 cells/2 mice). g, Top, Raster plot illustrating firing of an identified ARCH-expressing PKCδ⁺ neuron before, during and after 300 ms yellow light stimulation; Bottom, PSTH of the z-score shows a marked decrease in firing during the 300 ms light pulse. h, Rate histogram average before, during and after 3 min light stimulation on identified ARCH-expressing PKCδ⁺ neurons (n = 6 cells/3 mice). i, z-score histogram illustrating increased firing of ARCH-expressing PKCδ⁺ neurons during 3 min light stimulation (n = 6 cells/3 mice). All error bars indicate mean ± s.e.m. j, Schematic representation of a freely moving mouse bilaterally implanted with optical fibers after injection of DIO-AAVs expressing either ChR2 or ARCH. k, Coronal sections of a mouse brain indicating the location of CEA and the expression of a conditional rAAV expressing ChR2 and mCherry. White dashed lines indicate anatomical boundaries of BLA, CEl and CEm. The blue dashed line indicates the insertion of the optical fiber. Scale bar: 500 μm. l, Top, Schematic of the experimental paradigm. Bottom, Bar graph illustrating increased fear generalization upon stimulation with blue light in PKCδ-Cre animals infected with DIO-AAV-ChR2 (n = 7 mice), but not in control animals (n = 6 mice; Two way ANOVA: F_(1,11) = 11.83, P= 0.005 for the interaction injection × light stimulation; post-hoc Sidak's multiple comparison test: P = 0.01 for light_on vs. light_off in DIO-AAV-ChR2 injected mice). **P = 0.008, paired t-test between light_off and light_on in ChR group, t = -3.952, 6 degrees of freedom. = 0.946, paired t-test between light_off and light_on in control group, t = 0.0707, 5 degrees of freedom. m, Top, Schematic representation of the experimental protocol used for analyzing optogenetic manipulations of PKCδ⁺ neurons during EPM behavior. EPM and open field behaviour was analyzed prior to subjecting animals to fear conditioning. Bottom, Example EPM trajectories of PKCδ-Cre animals injected with rAAV 2/7 EF1α∷DIO-ChR2(H134R)-2A-NpHR-2A-Venus (top) or rAAV 2/5 CBA∷DIO-ARCH-GFP (bottom) under light_on or light_off conditions. n, Enhancing the activity of PKCδ⁺ neurons decreases the percentage of time animals spend on open arms (n = 7 mice, **P = 0.006, paired t-test, t = -3.952, 6 degrees of freedom between light_off and light_on for ChR2 group.), whereas decreasing the activity of PKCδ⁺ neurons has the opposite effect (n = 6 mice, **P = 0.005, paired t-test, t = -4.702, 5 degrees of freedom between light_off and light_on for ARCH group.). The behavior of control animals was not altered by light stimulation (n = 9 mice, P = 0.214, aired t-test, t = 1.349, 8 degrees of freedom between light_off and light_on for control group.). Two-way ANOVA: F_{(2, 19)} = 26.72, P < 0.001 for the interaction injection × light stimulation; post-hoc pairwise Sidak's tests revealed significant differences between light_on in the ARCH group, light_on in the ChR2 and control group. o, Top, Schematic representation of the experimental protocol used for analyzing optogenetic manipulations of PKCδ⁺ neurons during open field behavior. Bottom, Example open field trajectories of PKCδ-Cre animals injected with rAAV 2/7 EF1α∷DIO-ChR2(H134R)-2A-NpHR-2A-Venus (top) or rAAV 2/5 CBA∷DIO-ARCH-GFP (bottom) under light_on or light_off conditions. p, Enhancing the activity of PKCδ⁺ neurons decreases the number of center crossings (red square) per unit track length (n = 7 mice, **P = 0.006, paired t-test between light_off and light_on for ChR group, t = 4.124, 6 degrees of freedom), whereas decreasing the activity of PKCδ⁺ neurons has the opposite effect (n = 8 mice, **P = 0.007, aired t-test between light_off and light_on for ARCH group, t = -3.956, 6 degrees of freedom.). The behavior of control animals was not altered by light stimulation (n = 7 mice, P = 0.408, paired t-test between light_off and light_on for control group, t = 0.880, 7 degrees of freedom.). Two-way ANOVA: Two-way ANOVA: F_{(2, 19)} = 8.587, P = 0.002 for the interaction injection × light stimulation; post-hoc pairwise Sidak's tests revealed significant differences between light_on in the ARCH group, light_on in the ChR2 and control group. All error bars indicate mean ± s.e.m.

Figure 3

Extrasynaptic inhibition mediated by α₅-GABA_ARs controls the spontaneous firing of PKCδ⁺ neurons. a, Representative whole-cell current trace (V_h = -75.5 mV) recorded from a PKCδ⁺ neuron in vitro illustrating the successive blockade of phasic (synaptic) and tonic (extrasynaptic) GABAergic currents (scale bar: 5 pA, 2 min.). Application of 1 μM SR-95531 and 100 μM PTX is indicated by the pink and brown areas, respectively. Lower dashed line indicates baseline holding current that fits with the average mean of two all-point histograms for baseline (gray) and SR-95531 (pink). PTX caused a shift of the holding current indicated by the upper dashed line that fits the average of a Gaussian distribution (brown). b, Charge transfer of synaptic and extrasynaptic GABAergic currents (n = 6 cells/6 slices/3 mice each; **P = 0.005, two-tailed paired t-test, t = -4.691, 5 degrees of freedom). c, Top, Traces show spontaneous firing of PKCδ⁺ neurons recorded in cell-attached configuration (scale bar: 100 pA, 5 s). Bottom, Spontaneous firing of PKCδ⁺ neurons is enhanced by PTX but not by SR-95531 (n = 4 cells/4 slices/2 mice; *P = 0.028, One-Way ANOVA followed by Dunn's multiple comparisons test between baseline and PTX, P = 1.000 between baseline and SR-95531). d, Top, Representative recordings illustrating the shift in the holding current induced by PWZ-029 (1 μM) in wild-type (α₅^+/+), but not in mice lacking α₅ receptors in PKCδ⁺ neurons (α₅^-/-: PKCδ-Cre × Gabra5^fl/fl)(scale bar:- 10 pA, 10 s). Bottom, α5GABA_AR mediated extrasynaptic inhibition is completely absent in PKCδ⁺ neurons recorded in slices from PKCδ-Cre × Gabra5^fl/fl mice (n = 5 cells/5 slices/2 KO mice and n = 4 cells/4 slices/2 wild type mice; *P = 0.003, two-tailed unpaired t-test, t = 4.172, 6 degrees of freedom). e, Top, Spontaneous firing of a representative PKCδ⁺ neuron recorded in cell-attached configuration before and during the application of PWZ-029 (scale bar: 100 pA, 5 s). Bottom, Spontaneous firing of PKCδ⁺ neurons is enhanced by application of PWZ-029 (1 μM; n = 5 cells/5 slices/3 mice; **P = 0.015, two-tailed paired t-test, t = -4.049, 4 degrees of freedom). f, Pre-embedding double-labeling electron microscopy reveals extrasynaptic α5GABA_AR expression (gold/silver particles) in dendrites of PKCδ⁺ neurons (shown in orange; HRP-DAB reaction) of the CEl. Immunometal particles were observed both at the plasma membrane (indicated by arrows) and at cytoplasmic locations. Synapses are indicated by arrowheads. The axon terminal at1 forms a symmetric (Type II) synapse whereas axon terminals at2 and at3 form asymmetric (Type I) synapses. Scale bar: 1 μm. The lower panel shows a reconstruction of segments (from 5 sections) of the two PKCδ⁺ dendrites and of the axon terminals at1 (blue) and at2 (green). Surface immunometal particle density on PKCδ⁺ dendrites was 0.17 ± 0.02 particles/μm, n = 70 segments. It can be appreciated that the immunometal particles, shown as red dots, are not associated with synaptic specializations. mc: mitochondrion. All error bars indicate mean ± s.e.m.

Figure 4

Experience-dependent reduction of extrasynaptic inhibition predicts fear generalisation. a, Freezing levels before (baseline) and during the presentation of the CS^- and the CS⁺ 24 hrs after training for animals exposed to CS only (n = 8 mice) and animals subjected to CS⁺-US pairings (n = 8 mice; ***P < 0.001 for CS⁺ in CS-US group, H = 81.715, 2 degrees of freedom; P = 0.412 for CS only group, H = 1.771, 2 degrees of freedom; Kruskal-Wallis One-Way ANOVA followed by Dunn's pairwise multiple comparison test). b, Representative current traces recorded in vitro from PKCδ⁺ neurons in slices obtained from control (CS only) and fear conditioned (CS-US) animals. Traces illustrate the sequential application of L-655 (50 nM) and PTX (100 μM). Scale bar: 50 pA, 10 s. c, Left, Total extrasynaptic inhibition is reduced in PKCδ⁺ neurons of fear conditioned animals (n = 26 cells/26 slices/8 mice) compared to control animals (CS only, n = 14 cells/14 slices/8 mice). ***P < 0.001 by Mann-Whitney Rank Sum unpaired t-test. Right, α₅-mediated extrasynaptic inhibition is reduced in PKCδ⁺ neurons of fear conditioned animals (n = 28 cells/28 slices/8 mice) compared to control animals (CS only, n = 14/14 slices/8 mice). ***P < 0.001, Mann-Whitney Rank Sum unpaired t-test. d, Total extrasynaptic inhibition inversely correlates with fear generalization. Each dot represents the average of the extrasynaptic inhibition recorded from several PKCδ⁺ neurons for an individual animal (n = 2-4 cells per animal; n = 8 animals). Linear regression: R = 0.7; P = 0.041. All error bars indicate mean ± s.e.m.

Figure 5

Brain area-specific conditional genetic deletion of α₅-GABA_AR expression induces fear generalization and anxiety behavior. a, Top, Scheme showing the experimental protocol. Anxiety tests were performed before subjecting animals to fear conditioning. Bottom, AAV-mediated expression of Cre recombinase and GFP (green staining) in CEA neurons after injection of α₅-floxed animals (α₅^(fl/fl)) with AAV-Cre. Scale bar: 50 μm. b, Top, Representative current traces illustrating Cre-induced loss of extrasynapticα₅-mediated inhibition in infected CEA neurons of α₅-floxed animals (right,α₅^(fl/fl), green) compared to wild-type animals injected with the same virus (left, α₅^(+/+), black). Scale bar: 20 pA, 15 s. Bottom, PTX- and PWZ-sensitive extrasynaptic inhibition in CEA neurons is strongly reduced by expression of Cre-GFP (n = 6 cells/6 slices/3 mice, **P = 0.006, One-Way ANOVA followed by Dunnett's multiple comparisons test vs. PTX in α₅^(+/+)) compared to control (n = 5 cells/5 slices/3 mice,α₅^(+/+)) and non-infected cells in α₅^(fl/fl)animals (n = 5 cells/5 slices/2 mice). c, Top, Schematic representation of the behavioral protocol. Bottom, Brain area-specific genetic deletion of α₅-GABA_ARs (n = 13 mice, α₅^(fl/fl)) in CEA reduces the time spent in the open arms on the EPM compared to control animals injected with the same virus (n = 9 mice, α₅^(+/+)). **P = 0.008, Mann-Whitney rank sum two-tailed unpaired t-test. d, Top, Schematic representation of the behavioral protocol. Bottom, Brain area-specific genetic deletion of α₅-GABA_ARs (n = 8 mice, α₅^(fl/fl)) in CEA enhances fear generalization compared to controls (n = 9 mice, α₅^(+/+)). *P = 0.04, two-tailed unpaired t-test, t = -2.269, 14 degrees of freedom. All error bars indicate mean ± s.e.m.

Figure 6

Brain area- and cell type-specific knock-down of α₅-GABA_AR expression induces anxiety behaviour and fear generalization. a, Conditional expression of Tdtomato and shRNA in PKCδ⁺ neurons (CFP staining, green) after infection of PKCδ-Cre-CFP mice with an AAV containingDIO-α₅-shRNA. Scale bar: 50 μm. b, Representative current traces illustrating shRNA-induced knock-down of extrasynaptic inhibition in PKCδ⁺ neurons (α₅-shRNA) compared to control animals injected with a virus expressing scrambled shRNA. Scale bar: 20 pA, 15 s. c, PTX-sensitive total extrasynaptic inhibition in PKCδ⁺ neurons is strongly reduced by expression of α₅-shRNA (n = 8) compared to cells expressing scrambled shRNA (n = 7). ***P < 0.001, Mann-Whitney Rank Sum unpaired t-test. d, PWZ-029-sensitive α₅-GABA_AR-mediated extrasynaptic inhibition in PKCδ⁺ neurons is strongly reduced by expression of α₅-shRNA (n = 8 cells/8 slices/3 mice) compared to cells expressing scrambled shRNA (n = 7 cells/7 slices/3 mice). ***P < 0.001 by Mann-Whitney Rank Sum unpaired t-test. e, Top, Schematic representation of the experimental protocol. Bottom, Cell-type specific expression of α₅-shRNA (n = 7 mice) in PKCδ⁺ neurons reduces the percentage of time spent in the open arms of the EPM compared to animals infected with an AVV expressing scrambled shRNA (n = 7 mice). ***P < 0.001, two-tailed unpaired t-test, t = 5.514, 12 degrees of freedom. f, Top, Schematic representation of the experimental protocol. Bottom, Cell-type specific expression of α₅-shRNA (n = 7 mice) in PKCδ⁺ neurons reduces the normalized number of center crossings in the open field paradigm compared to animals injected with an AAV expressing scrambled shRNA (n = 7 mice). ***P < 0.001, two-tailed unpaired t-test, 4.611, 12 degrees of freedom. g, Top, Schematic representation of the experimental protocol. Bottom, Cell-type specific expression of α5-shRNA (n = 5 mice) in PKCδ⁺ neurons enhances fear generalization compared to animals infected with an AAV expressing scrambled shRNA (n = 5 mice). **P = 0.015, two-tailed unpaired t-test, t = -3.088, 8 degrees of freedom. All error bars indicate mean ± s.e.m.