Generalized extinction of fear memory depends on co-allocation of synaptic plasticity in dendrites

Abstract

Memories can be modified by new experience in a specific or generalized manner. Changes in synaptic connections are crucial for memory storage, but it remains unknown how synaptic changes associated with different memories are distributed within neuronal circuits and how such distributions affect specific or generalized modification by novel experience. Here we show that fear conditioning with two different auditory stimuli (CS) and footshocks (US) induces dendritic spine elimination mainly on different dendritic branches of layer 5 pyramidal neurons in the mouse motor cortex. Subsequent fear extinction causes CS-specific spine formation and extinction of freezing behavior. In contrast, spine elimination induced by fear conditioning with >2 different CS-USs often co-exists on the same dendritic branches. Fear extinction induces CS-nonspecific spine formation and generalized fear extinction. Moreover, activation of somatostatin-expressing interneurons increases the occurrence of spine elimination induced by different CS-USs on the same dendritic branches and facilitates the generalization of fear extinction. These findings suggest that specific or generalized modification of existing memories by new experience depends on whether synaptic changes induced by previous experiences are segregated or co-exist at the level of individual dendritic branches.

Fig. 1. The motor cortex is important for fear extinction generalization.

a Mice (YFP-H line) were subjected to CS1-US and CS2-US pairings followed by CS1 or CS2 extinction (CS1: 1 kHz; CS2: 10 kHz). The freezing response to CS1 after CS2 extinction was comparable to that after no extinction (P = 0.3829, Mann–Whitney U test; n = 7, 6, and 7 mice in no extinction, CS1 extinction, and CS2 extinction groups respectively). No significant difference in the freezing response to CS2 between CS1 extinction and no extinction groups (P = 0.3154, Mann–Whitney U test; n = 10, 8 and 6 mice in no extinction, CS1 extinction and CS2 extinction groups respectively). P values for comparison between groups were calculated using Mann–Whitney U test and shown in the graph. b Mice (YFP-H line) were subjected to CS1-US, CS2-US, and CS3-US pairings followed by CS2 or CS3 extinction (CS1: 1 kHz; CS2: 10 kHz; CS3: 5 kHz). After CS2 extinction, the freezing response to CS3 was significantly lower than that in no extinction group (P = 0.002, unpaired t test; n = 15, 12 and 5 mice in no extinction, CS2 extinction and CS3 extinction groups respectively). After CS3 extinction, the freezing response to CS1 and CS2 was lower than that in no extinction group (CS1: P = 0.0269; n = 15, 12, and 12 mice in no extinction, CS2 extinction and CS3 extinction groups respectively; CS2: P = 0.0016; n = 14, 5, and 12 mice in no extinction, CS2 extinction and CS3 extinction groups respectively; unpaired t test). Statistical analyses used were unpaired t test for mice receiving CS2 extinction and CS1 or CS3 recall or mice receiving CS3 extinction and CS1 or CS2 recall and Mann–Whitney U test for mice receiving CS2 extinction and CS2 recall or mice receiving CS3 extinction and CS3 recall. c Mice (YFP-H line) were subjected to CS1-US, CS2-US, CS3-US, and CS4-US pairings followed by CS4 extinction (CS4: 2.5 kHz). After CS4 extinction, the freezing response to CS1, CS2, or CS3 was lower than that in no extinction group (CS1, P = 0.0026; CS2 and CS3, P < 0.0152 and 0.0488, respectively; unpaired t test; n = 11 mice in each group). P values for comparison between groups were calculated using unpaired t test. d Mice (YFP-H line) were subjected to three different CS-US pairings followed by CS2 extinction (CS1: 1 kHz; CS2: 10 kHz; CS3: 5 kHz). Muscimol or vehicle was infused bilaterally into the auditory or primary motor cortex before CS2 extinction. The freezing response to CS3 after CS2 extinction in mice with muscimol infusion into the motor cortex was significantly higher than that in vehicle-treated group (P < 0.0379, Mann–Whitney U test), and comparable to that in no extinction group (P > 0.9999, Mann–Whitney U test; n = 8 mice in each group). P values for comparison between groups were calculated using Mann–Whitney U test. Error bars, ±S.E.M. All statistical tests were performed two-sided.

Fig. 2. Fear conditioning and extinction-induced dendritic spine elimination and formation of layer 5 pyramidal neurons in the motor cortex likely contribute to neuronal activity changes.

a Schematic of experimental design. Spine structure and Ca²⁺ activity were examined with or without CS (1 kHz) presentation before and after CS-US pairing and CS extinction. b Left and middle panels: Representative images of spine activity on apical dendrites of GCaMP6 and tdTomato co-expressing layer 5 pyramidal neurons with or without CS presentation before fear conditioning. Right panel: GCaMP6 fluorescence traces of seven spines on the left panel without and with CS presentation. Yellow bar denotes the period of CS presentation. Experiments were repeated independently on 178 spines in 6 mice with similar results. c Percentage of spines showing increased or decreased activity or no activity in response to CS before fear conditioning. Spines were considered showing increased or decreased responses to CS if total ΔF/F₀ of spines during CS/pre-CS presentation period was ≥1 or <1, respectively (during CS vs. pre-CS, P < 0.0001 respectively, Wilcoxon matched-pairs signed rank test). Total 178 spines were analyzed. d Spine elimination and formation after fear conditioning and extinction. The hollow triangle indicates the spine (S4) eliminated on a dendrite (b) after fear conditioning. The solid triangle indicates the newly-formed spine (NS) after fear extinction. Experiments were repeated independently on 178 spines in 6 mice with similar results. e After CS-US pairing, spines with increased activity in response to CS had a higher elimination rate than spines with decreased activity to CS (~14.0% vs. 1.5%, P < 0.0062, chi-square test; n = 85 and 66 spines respectively from 6 mice). f Images and Ca²⁺ fluorescence traces of spines on the same dendrite (in b and d) before and during CS presentation after fear extinction. The solid triangle indicates the newly-formed spine (NS) after fear extinction. Experiments were repeated independently on 13 spines in 6 mice with similar results. g After CS extinction, newly-formed spines, but not existing spines, showed a higher activity level during CS presentation when compared to pre-CS presentation period (new spines: P = 0.0061; existing spines: P = 0.1167; Wilcoxon matched-pairs signed rank test; n = 13 and 53 spines respectively). A larger percentage of newly-formed spines than existing spines also showed increased activity in response to CS (~92% vs. 51%, P = 0.0066, chi-square test). Error bars, ±S.E.M. All statistical tests were performed two-sided. ** P < 0.01; **** P < 0.0001.

Fig. 3. Distribution of spine elimination induced by different CS-US pairings on individual neurons.

a Experimental design to examine spine elimination induced by different CS-USs in the motor cortex of YFP-H line mice (CS1: 1 kHz; CS2: 10 kHz; CS3: 5 kHz). b Representative images of spine elimination induced by three different CS-US pairings on apical dendritic branches from same neuron in YFP-H line mice. The hollow triangles indicate the spines eliminated in the succeeding image. The solid triangles indicate the newly-formed spines when compared to the preceding image. The asterisks indicate filopodia. Experiments were repeated independently on 36 branches in 14 mice with similar results. c Comparison of spine elimination rates of individual neurons (circles) between mice subjected to different CS-US pairings and untrained or unpaired control mice (CS1-US pairing vs. untrained or unpaired, P < 0.0001 or P = 0.0003, respectively; n = 28, 14 and 12 neurons respectively. CS2-US pairing vs. untrained or unpaired, P < 0.0001; n = 28, 8 and 11 neurons respectively. CS3-US pairing vs. untrained or unpaired, P = 0.0001 and P = 0.0036 respectively; n = 14, 14 and 11 neurons respectively; Mann–Whitney U test). d After each CS-US pairing, the majority of Layer 5 pyramidal neurons showed the spine elimination rate higher than the mean plus two times the standard deviation of that in untrained control mice. e Percentage of overlap between neurons with spine elimination induced by two different CS-US pairings. The percentage in the non-overlapping area indicates the proportion of non-overlapping neurons relative to neurons with spine elimination induced by each of the different CS-US pairings. Error bars, ±S.E.M. All statistical tests were performed two-sided. ** P < 0.01; *** P < 0.001; **** P < 0.0001.

Fig. 4. Distribution of spine elimination induced by different CS-US pairings on individual dendrites.

a Comparisons of spine elimination rates on individual dendritic branches between mice subjected to each CS-US pairing and untrained or unpaired control mice (CS1: 1 kHz; CS2: 10 kHz; CS3: 5 kHz; CS1-US pairing vs. untrained or unpaired, P < 0.0001; n = 78, 28, and 34 branches, respectively. CS2-US pairing vs. untrained or unpaired, P = 0.0014 or P = 0.0075; n = 60, 14, and 24 branches respectively. CS3-US pairing vs. untrained or unpaired, P = 0.0024 and P = 0.0233, respectively; n = 58, 32, and 28 branches, respectively; Mann–Whitney U test). b After each CS-US pairing, ~40% of individual dendritic branches showed spine elimination rate higher than the mean plus two times the standard deviation of that in untrained control mice. c Percentage of overlap between branches with spine elimination induced by any two different CS-US pairing. The percentage in the non-overlapping area indicates the proportion of non-overlapping branches versus branches with spine elimination induced by each of different CS-US pairings. d The rate of spine elimination rate on individual branches after CS2-US pairing was inversely correlated with that after CS1-US pairing (P = 0.0006, Pearson’s correlation; n = 60 branches). e The rate of spine elimination on individual branches after CS3-US pairing was not correlated with that after CS1-US pairing (P = 0.1717, Pearson’s correlation; n = 36 branches). f The rate of spine elimination on individual branches after CS3-US pairing was positively correlated with that after CS2-US pairing (P = 0.0098, Pearson’s correlation). Error bars, ±S.E.M. All statistical tests were performed two-sided. * P < 0.05; ** P < 0.01; **** P < 0.0001.

Fig. 5. Fear extinction induces CS-specific spine formation and neuronal activity increase in response to CS when two different CS-US pairings induce spine elimination on separate dendritic branches.

a Experimental design to examine spine remodeling and somatic Ca²⁺ activity in mice subjected to two different CS-USs followed by CS1 extinction (CS1: 1 kHz; CS2: 10 kHz). b Representative images of spine remodeling induced by two different CS-US pairings followed by CS1 extinction on individual dendritic branches in YFP-H mice. The hollow triangles indicate spines eliminated in the succeeding image. The solid triangles indicate newly-formed spines when compared to the preceding image. The asterisks indicate filopodia. Experiments were repeated independently on 36 branches in 8 mice with similar results. The rate of spine formation on individual branches after CS1 extinction was positively correlated with the rate of spine elimination after CS1-US pairing (c) and inversely correlated with the rate of spine elimination after CS2-US pairing (d) (P < 0.0001 and P = 0.0487, respectively, Pearson’s correlation; n = 36 branches). e Majority of newly-formed spines after CS1 extinction were located within 2 µm distance to spines eliminated after CS1-US pairing, but not after CS2-US pairing (P = 0.0002, chi-square test; n = 91 spines from 8 mice). f The survival rate of newly-formed spines induced by CS1 extinction was lower after reconditioning by CS1-US pairing than after reconditioning by CS2-US pairing (P = 0.0002, chi-square test; n = 36 and 40 spines from 6 mice, respectively). g Left panel: Representative images of somatic activity of layer 5 pyramidal neurons in GCaMP6S line 3 mice after CS1-US and CS2-US pairings; The blue triangle indicates the soma with reduced activity to both CS1 and CS2 as compared to the pre-CS period after two different CS-US pairings. Somatic Ca²⁺ changes in this cell were measured by ΔF/F₀. Right panel: Representative images of somatic activity in response to CS1 or CS2 after CS1 extinction. The labeled neuron with reduced response to both CSs in the (left panel) showed increased activity to CS1 but not CS2 after CS1 extinction. Experiments were repeated independently in 189 somas in 10 mice with similar results. h After CS1-US or CS2-US pairings, the majority of neurons exhibited reduced somatic Ca²⁺ activity in response to CS1 or CS2 as compared to pre-CS period (n = 189 somas from 10 mice). i Percentage of overlap between neurons with reduced activity to two different CSs. The percentage in the non-overlapping area indicates the proportion of non-overlapping neurons relative to neurons with reduced activity to each CS. j After CS1 extinction, a larger percentage of overlapping neurons with reduced activity to both CSs showed increased activity in response to CS1 when compared to without extinction (P < 0.0001, chi-square test; n = 66 and 25 somas in extinction and no extinction group, respectively). The percentage of overlapping neurons exhibiting increased activity in response to CS2 after CS1 extinction was comparable to that after no extinction (P = 0.2081, chi-square test). All statistical tests were performed two-sided. *** P < 0.001; **** P < 0.0001.

Fig. 6. Fear extinction induces CS-nonspecific spine formation and neuronal activity increase in response to CS when spine elimination induced by three different CS-US pairings occurs on the same dendritic branches.

a Experimental design to examine spine remodeling and somatic Ca²⁺ activity in mice subjected to three different CS-USs followed by CS2 extinction respectively (CS1: 1 kHz; CS2: 10 kHz; CS3: 5 kHz). b Representative images of spine remodeling induced by CS2-US and CS3-US pairings followed by CS2 extinction on individual dendritic branches in YFP-H line mice. The hollow triangles indicate spines eliminated in the succeeding image. The solid triangles indicate newly-formed spines when compared to the preceding image. The asterisks indicate filopodia. Experiments were repeated independently on 27 branches in 7 mice with similar results. c Comparisons of spine formation rates after CS2 extinction on branches with spine elimination induced by one or more CS-US pairings (spine formation rate after CS2 extinction on branches with spine elimination induced by only CS2-US pairing vs. by only CS3-US pairing or by both CS2-US and CS3-US pairing, P < 0.0001 or P < 0.0108, respectively; spine formation rate after CS2 extinction vs. no extinction on branches with spine elimination induced by only CS2-US pairing or by both CS2-US and CS3-US pairing, P = 0.0002 or P = 0.0004, respectively; Mann–Whitney U test; n = 10, 8, and 9 branches with spine elimination induced by only CS2-US or by only CS3-US or by both CS2-US and CS3-US in CS2 extinction group; n = 6, 5, and 6 branches with spine elimination induced by only CS2-US pairing or by only CS3-US pairing or by both CS2-US and CS3-US pairings in no extinction group). d On branches with spine elimination induced by both CS2-US and CS3-US pairings, a larger percentage of new spines induced by CS2 extinction were located within 2 µm to spines eliminated by CS2-US pairing or by CS3-US pairing as compared to no extinction condition (P = 0.0234 or 0.0252, respectively, chi-square test; n = 58 and 19 newly-formed spines from 7 and 5 mice in extinction and no extinction groups respectively). e On branches with spine elimination induced by both CS2-US pairing and CS3-US pairings, CS3-US reconditioning reduces the survival rate of new spines that were formed after CS2 extinction and located within 2 µm to spines eliminated prior CS3-US pairing (P = 0.0019, compared to no reconditioning group; chi square test; n = 30 and 23 newly-formed spines from 8 and 4 mice in reconditioning and no reconditioning groups respectively). f Left panel: Representative images of somatic activity of layer 5 pyramidal neurons in GCaMP6S line 3 mice after three different CS-US pairings. The blue triangle indicates the soma with reduced activity to all CSs as compared to the pre-CS period after three CS-US pairings. Right panel: Representative images of somatic activity in response to CS1, CS2, or CS3 after CS2 extinction. The labeled neuron with reduced response to all CSs in (left panel) showed increased activity to CS2 and CS3, but not CS1, after CS2 extinction. Experiments were repeated independently in 130 somas in 7 mice with similar results. g After three different CS-US pairings, the majority of layer 5 pyramidal neurons exhibited reduced somatic Ca²⁺ activity in response to CS1, CS2, or CS3 as compared to the pre-CS period (n = 130 somas from 7 mice). h. Overlap among neurons with reduced activity to different CSs. The number in the overlapping or non-overlapping area indicates the number of neurons with reduced activity to different CSs. i After CS2 extinction, a larger percentage of overlapping neurons with reduced activity to all CSs showed increased activity to CS2 and CS3 when compared to that without no extinction (CS2, P < 0.0001, chi-square test; CS3, P < 0.0053; n = 30 and 23 somas in extinction and no extinction groups respectively). The percentage of these overlapping neurons exhibiting increased activity to CS1 after CS2 extinction was comparable to that without extinction (P = 0.4401, chi-square test). Error bars, ±S.E.M. All statistical tests were performed two-sided. * P < 0.05; ** P < 0.01; *** P < 0.001; **** P < 0.0001.

Fig. 7. SST interneuron activity regulates dendritic distribution of spine elimination induced by different CS-US pairings and subsequent changes induced by fear extinction.

a Experimental design to examine the regulation of fear conditioning-induced spine elimination by SST INs (CS1: 1 kHz). b Representative images of spine remodeling on apical dendrites induced by fear conditioning in YFP-H line mice crossed with SST Cre mice which were infected with hM3D(Gq) and injected with either CNO or saline. Experiments were repeated independently on 16 branches in 5 hM3D(Gq)-CNO mice and 12 branches in 5 hM3D(Gq)-saline mice with similar results. The hollow triangles indicate spines eliminated in the succeeding image. The solid triangles indicate newly-formed spines when compared to preceding image. The asterisks indicate filopodia. c After fear conditioning, the spine elimination rate on individual branches was significantly higher in the hM3D(Gq)-CNO group than the hM3D(Gq)-saline control group (P = 0.016, unpaired t test; n = 12 and 16 branches in hM3D(Gq)-saline and hM3D(Gq)-CNO groups respectively). d The freezing response was significantly higher in the hM3D(Gq)-CNO group than the hM3D(Gq)-saline control group (P = 0.0349, unpaired t-test; n = 9 and 10 mice in hM3D(Gq)-saline and hM3D(Gq)-CNO groups, respectively). e Representative images of spine remodeling on apical dendrites induced by fear conditioning in YFP-H-line mice crossed with SST Cre mice which were infected with hM4D(Gq) and injected with CNO or saline. Experiments were repeated independently on 12 branches in 4 hM4D(Gi)-saline mice and 14 branches in 4 hM4D(Gi)-saline mice with similar results. f After fear conditioning, the spine elimination rate of individual branches was slightly but not significantly lower in the hM4D(Gi)-CNO group than the hM4D(Gi)-saline group (P = 0.2502, unpaired t test; n = 14 and 12 branches in hM4D(Gi)-saline and hM4D(Gi)-CNO groups respectively). g The freezing response was slightly lower in the hM4D(Gi)-CNO group than the hM4D(Gi)-saline group (P = 0.0952, Mann–Whitney U test; n = 5 mice for each group). h Experimental design (CS1: 1 kHz; CS2: 10 kHz). i Representative images of spine remodeling induced by two different CS-US pairings followed by CS1 extinction on individual dendritic branches in hM3D(Gq)-CNO mice. Experiments were repeated independently on 11 branches in 7 mice with similar results. j Percentage of overlap between branches with spine elimination induced by two different CS-US pairing in hM3D(Gq)-saline and hM3D(Gq)-CNO mice. The percentage in the non-overlapping area indicates the proportion of non-overlapping branches relative to branches with spine elimination induced by each of different CS-US pairings. k After CS1 extinction, the spine formation rate was significantly higher on branches with spine elimination induced by both CS1-US and CS2-US pairings when compared to that on branches with spine elimination induced by only CS1-US pairing or by only CS2-US pairing in both hM3D(Gq)-CNO and hM3D(Gq)-saline groups (hM3D(Gq)-CNO group: both CS-US pairings vs. CS1-US or CS2-US pairing, P = 0.005 or P = 0.0015, respectively; CS1-US vs. CS2-US pairing, P = 0.017; Mann–Whitney U test; n = 11, 10, and 4 branches with spine elimination induced by both CS1 and CS2-US pairings or by only CS1-US pairing or by only CS2-US pairing; hM3D(Gq)-saline group: both CS-US pairings vs. CS1-US or CS2-US pairing, P = 0.0167 or P = 0.0303; CS1-US vs. CS2-US pairing, P = 0.0025; Mann–Whitney U test; n = 2, 14 and 10 branches with spine elimination induced by both CS1-US and CS2-US pairings or by only CS1-US pairing or by only CS2-US pairing in hM3D(Gq)-saline group). l On branches with spine elimination induced by both CS1-US and CS2-US pairings, a large percentage of newly-formed spines after CS1 extinction were located within 2 µm to spines eliminated by CS1-US pairing or by CS2-US pairing respectively in hM3D(Gq)-CNO group (n = 54 newly-formed spines from 7 mice). m After CS1 extinction, the freezing response to CS2 in the hM3D(Gq)-CNO group was significantly lower when compared to that without extinction (P = 0.0003, Mann–Whitney U test; n = 7-8 mice for each group). Error bars, ±S.E.M. All statistical tests were performed two-sided. *P < 0.05; **P < 0.01; ***P < 0.001.