To test whether extinction depends on increased levels of feed-forward inhibition in CEm, we first compared the responses of CEm neurons to BLA inputs () in coronal slices of the amygdala obtained from rats that were previously subjected to fear conditioning only (n = 16) vs. rats that were fear conditioned and trained on extinction the next day (n = 14; ). This data was contrasted to that obtained in naïve rats (n = 11) and rats presented with the CSt and USt in an unpaired fashion (n = 10; ). We first describe the behavior of these rats () and then analyze how the training procedures affected the responsiveness of CEm neurons to BLA inputs in vitro (). In this and subsequent experiments, the individuals carrying out the in vitro experiments and scoring the rats’ behavior were blind to group identity.
Fig. 1 Increased inhibition of CEm neurons in extinction and conditioned inhibition. (a) Experimental set-up. (b) Control and experimental groups. (c) Proportion of time spent freezing (average ± s.e.m.) during the various phases of the behavioral protocol (more ...)
Analysis of percent time spent freezing () confirmed that rats from the fear conditioning only (, black) vs. fear conditioning plus extinction (, red) groups exhibited nearly identical levels of conditioned freezing by the end of the fear conditioning session (Day 2). Although rats from the unpaired groups (, blue) did not receive paired CSt-USt presentations on Day 2, they did express significant freezing levels (paired t-test, habituation vs. last CSt, P = 0.002), which presumably represents contextual freezing. On day 3 in a different context, only rats from the extinction group (, red) received CSt presentations. We measured freezing in the unpaired or fear conditioned rats during corresponding 30 s periods, revealing significantly higher freezing levels in the extinction group than other groups (ANOVA F (2,59) = 37.83, P = 0.0001; Bonferroni-corrected post-hoc t-tests, P ≤ 0.0001).
Twenty-four hours after exposure to context B, we anesthetized the rats and prepared coronal sections of their amygdala. We obtained patch recordings from samples of 10–16 CEm cells per group and compared their responsiveness to electrical stimuli delivered at a standard position in BLA (). We carried out these tests from a membrane potential of −45 mV with the lidocaine derivative QX-314 in the pipette solution to facilitate the measurement of IPSP amplitudes without contamination from spike afterhyperpolizations. We observed significant inter-group differences in the amplitude of BLA-evoked IPSPs (; ANOVA F
(3,40) = 4.823, P
= 0.006). Bonferroni-corrected post-hoc t-tests
revealed that CEm neurons from naïve and fear conditioned rats exhibited IPSPs of significantly lower amplitude than in the extinguished and unpaired groups (; 400 µA; P
≤ 0.028). However, IPSPs from extinguished rats were not significantly different from those of the unpaired group (P
= 0.9) and both were abolished by picrotoxin (Fig. S1
). Group differences in IPSP amplitudes were not attributable to variations in the passive properties or GABA-A reversal potentials of CEm neurons since we observed negligible differences along these dimensions (Suppl. Tables 1–2 and Fig. S2
Although EPSP slopes tended to be higher in CEm neurons from the fear conditioned and extinction groups compared to naïve and unpaired animals, these differences did not reach significance (ANOVA, F(3,38) = 2.31, P = 0.92; , inset). However, because we studied these CEm cells at a depolarized level (−45 mV) to facilitate IPSP measurements, the testing conditions were not optimal to study BLA-evoked EPSPs. We therefore repeated these tests in separate samples of 13–18 CEm neurons (≥ 5 rats/group) from a membrane potential of −70 mV (). In these conditions, we detected significant inter-group differences in EPSP slopes (ANOVA F(3,54) = 4.443, P = 0.006). Bonferroni-corrected post-hoc t-tests revealed that EPSP slopes in the fear conditioned group were significantly higher than in the naïve (P = 0.001) and unpaired (P = 0.0014) animals but did not significantly differ from extinguished animals (P = 0.099).
Fig. 2 Group-related differences in CEm EPSP slopes and orthodromic spiking in response to BLA stimulation. (a) Slope of BLA-evoked EPSPs (initial 2 ms; from −70 mV; average ± s.e.m.) and (b) percent BLA stimuli (400 µA) eliciting orthodromic (more ...)
To test how these differences in the character of BLA-evoked responses affect the ability of BLA inputs to fire CEm neurons, we next compared the probability that 10 BLA stimuli of fixed intensity (400 µA) would trigger action potentials in CEm cells from the various groups (). We carried out these tests in samples of 10–15 CEm neurons (≥ 5 rats/group) recorded at rest with a control intracellular solution. We observed large differences in spiking probability between the various groups (ANOVA F(3,46) = 4.033, P = 0.014). In keeping with the analysis of EPSP slopes, Bonferroni-corrected post-hoc t-tests revealed that CEm cells from the fear conditioning group had a higher orthodromic responsiveness than in all other groups (P ≤ 0.042) with no differences among the latter. It should be noted that BLA-evoked orthodromic spikes had a relatively long latency in CEm cells from the fear conditioning group (9.8 ± 0.7 ms), longer than the latency of the IPSP onset in CEm cells from the extinction group (7.2 ± 0.49 ms; , red arrows) studied at −45 mV ().
Thus, overall the above indicates that fear conditioning is associated with an enhancement of BLA-evoked EPSPs that is partially reversed following extinction. In parallel, extinction training causes a marked increase in BLA-evoked inhibition, a property also seen in the unpaired group.
The increased amplitude of BLA-evoked IPSPs in the unpaired group was unexpected because this paradigm is commonly used as a control for non-associative influences in fear conditioning. However, it was previously reported that presenting the CSt and USt in an unpaired fashion transforms the CSt in a conditioned inhibitor 19
or in other words, a safety signal, an observation we confirmed in our experiments (Fig. S3
). This implies that treatments such as extinction and conditioned inhibition causing a reduction of fear responsiveness are associated with persistently increased BLA-evoked inhibition in CEm neurons. What is the origin of this enhanced inhibition? In brain slices, there are only two extrinsic sources of GABAergic inputs to CEm: CEl and ITC cells 10
. Thus, we next tested whether treatments that induce an increased inhibition of CEm neurons alter the responsiveness of CEl and ITC cells to BLA inputs.
We first compared BLA-evoked EPSPs in CEl cells from the various groups (). There were significant inter-group differences in EPSP amplitudes (ANOVA F
(3,43) = 3.073, P
= 0.038, ) and slopes (ANOVA F
(3,35) = 9.78, P
= 0.0001; ) with CEl neurons from the unpaired group being significantly more responsive to BLA inputs than in all other groups (Bonferroni-corrected post-hoc t-tests
≤ 0.001; Suppl. Table 3
). Surprisingly however, CEl neurons from rats of the extinction, home cage, and fear conditioned groups displayed similarly lower BLA-evoked EPSP amplitudes and slopes (P
≥ 0.3; ).
Fig. 3 Increased recruitment of CEl neurons by BLA inputs in conditioned inhibition. (a) Representative examples of BLA-evoked responses in CEl cells in control aCSF. Superimposition of four responses elicited by BLA stimuli of 200–500 µA increasing (more ...)
To test whether the stronger EPSPs seen in the unpaired group translated into an enhanced ability of BLA inputs to fire CEl neurons, we next compared the orthodromic responsiveness of CEl cells from the various groups using the same approach as for CEm neurons (samples of 8–11 CEl neurons in ≥ 4 rats/group). Paralleling the EPSP analysis, CEl neurons from the unpaired group had a significantly greater orthodromic responsiveness than all other groups (; ANOVA F(3,33) = 3.98, P = 0.016, t-tests, P ≤ 0.042).
The above suggests that the stronger inhibition seen in CEm neurons from the unpaired and extinction groups depend on different mechanisms. In unpaired rats, an increased recruitment of CEl neurons by BLA inputs appears to be involved. However, this was not the case in extinguished rats. To test whether the larger inhibition seen in CEm neurons from extinguished rats resulted from an increased recruitment of ITC cells by BLA inputs, we compared BLA-evoked EPSPs in ITC cells from the various groups ( and Fig. S4
). Here, we added an additional control group (termed “Unpaired + CS”; n = 12) to determine whether repetitive presentations of a previously unpaired CSt would also modify the responsiveness of ITC cells. These animals were treated like those of the Unpaired group on days 1–2 () but received 20 unpaired presentations in context B on day 3.
Fig. 4 Enhanced efficacy of BLA synapses onto ITC cells in extinction. (a) Intensity-dependence of BLA-evoked EPSPs in ITC neurons (average ± s.e.m.) in control aCSF. Inset shows representative ITC cells from the extinction (red) and fear conditioning (more ...)
We observed significant inter-group differences in EPSP amplitudes (ANOVA F
(4,66) = 5.3, P
= 0.001; ) and slopes (ANOVA F
(4,62) = 3.559, P
= 0.011; ). Bonferroni-corrected post-hoc t-tests
revealed that ITC cells were significantly more responsive to BLA inputs in the extinction group as compared to all other groups (P
≤ 0.018). Yet, the passive membrane properties of ITC cells did not vary between groups (Suppl. Table 4
To test whether the stronger EPSPs seen in the extinction group translated into an enhanced ability of BLA inputs to fire ITC neurons, we next compared the orthodromic responsiveness of ITC cells from the various groups using the same approach as for CEA neurons (samples of 11–19 ITC neurons with ≥ 5 rats/group). Paralleling the EPSP analysis, ITC neurons from the extinction group had a significantly greater orthodromic responsiveness than all other groups combined (; ANOVA F(4,66) = 5.19, P = 0.0011) with no differences between the latter (t-tests, P ≥0.09).
We considered two possible mechanisms for the enhanced responsiveness of ITC cells to BLA inputs in the extinguished group: (1) an enhanced transmitter release probability by BLA axon terminals contacting ITC cells; (2) an altered expression of ionotropic glutamate receptors at BLA inputs to ITC cells. To test the first possibility, in voltage-clamp mode, we compared the effect of paired BLA stimuli (50 ms inter-stimulus interval) and looked for differences in paired-pulse ratio () between cells of the extinction group (n = 9) vs. cells from the various control groups (n = 34). The paired-pulse ratio was significantly lower in the extinction group (n = 9) compared to the control cells (t-test
= 0.036; ), suggesting that extinction training produces a modest increase in transmitter release probability at BLA synapses onto ITC cells. See Figure S5
for individual groups.
Fig. 5 Mechanisms underlying increased BLA responsiveness of ITC cells in extinction. (a) Histogram on left shows paired pulse ratio (average ± s.e.m.) in ITC cells from the control (n = 34) and extinction (n = 9) groups. Traces on right show representative (more ...)
To test whether extinction training is associated with an altered function of ionotropic glutamate receptors at BLA inputs to ITC cells, in the presence picrotoxin (100 µM), we measured the amplitude of BLA-evoked EPSCs at membrane potentials of –80 and 55 mV and expressed the data in the form of a non-NMDA to NMDA ratio (). This analysis revealed that the non-NMDA to NMDA ratio was drastically enhanced in ITC cells from the extinction group (n = 8, t-test
< 0.0001) compared to cells from the control groups (n = 27; see Figure S6
for individual groups). Overall, these results suggest that the enhanced responsiveness of ITC cells in extinguished rats is due to a modest enhancement in transmitter release probability at BLA synapses on ITC cells and an altered expression profile or phosphorylation level of ionotropic glutamate receptors in ITC cells, in favor of non-NMDA receptors.
Finally, to test whether infralimbic inputs are required for the extinction-related facilitation of BLA inputs onto ITC cells, we compared the amplitude of BLA-evoked EPSPs in rats that received infusions of either vehicle or the GABA-A receptor agonist muscimol in the infralimbic cortex 10 min before extinction training ( and Fig. S7
). ITC cells from the muscimol group had a significantly lower responsiveness to BLA inputs than in the vehicle group (; t-test
= 0.011). ITC cells from the muscimol group were indistinguishable from the unpaired control group described above (t-test
= 0.51). Similarly, the non-NMDA to NMDA ratio of ITC cells in the muscimol group was significantly lower than in the vehicle group (, P
= 0.0012), but identical to that of the unpaired group (P
= 0.88). Overall, these results suggest that the extinction-related changes in the efficacy of BLA-to-ITC synapses are critically dependent on infralimbic activity during and/or shortly after extinction training.
Fig. 6 Infralimbic (IL) inactivation blocks extinction-related changes in the efficacy of BLA synapses onto ITC cells. (a) Experimental paradigm. (b) Intensity-dependence of BLA-evoked responses in ITC cells from the vehicle (n = 15) and muscimol (n = 11) groups (more ...)