Correction of amygdalar dysfunction in a rat model of fragile X syndrome

Fragile X syndrome (FXS), a commonly inherited form of autism and intellectual disability, is associated with emotional symptoms that implicate dysfunction of the amygdala. However, current understanding of the pathogenesis of the disease is based primarily on studies in the hippocampus and neocortex, where FXS defects have been corrected by inhibiting group I metabotropic glutamate receptors (mGluRs). Here, we observe that activation, rather than inhibition, of mGluRs in the basolateral amygdala reverses impairments in a rat model of FXS. FXS rats exhibit deficient recall of auditory conditioned fear, which is accompanied by a range of in vitro and in vivo deficits in synaptic transmission and plasticity. We find presynaptic mGluR5 in the amygdala, activation of which reverses deficient synaptic transmission and plasticity, thereby restoring normal fear learning in FXS rats. This highlights the importance of modifying the prevailing mGluR-based framework for therapeutic strategies to include circuit-specific differences in FXS pathophysiology.


In brief
Fernandes et al. investigate the synaptic basis of deficient conditioned fear and its reversal in FXS rats. They find presynaptic mGluR5 in the amygdala, activation of which restores normal synaptic transmission, plasticity, and fear learning. This highlights the importance of circuit-specific differences in FXS pathophysiology and mGluR-based therapeutic strategies.

INTRODUCTION
Fragile X syndrome (FXS), a leading genetic cause of intellectual disability and autism spectrum disorder, is caused by the absence of the fragile X mental retardation protein (FMRP) produced by the fragile X mental retardation1 (FMR1) gene. One key feature of the symptoms of FXS is abnormal emotional behavior, and clinical evidence points to the amygdala's contribution to these affective symptoms. Neuroimaging revealed that men with the FMR1 premutation exhibit reduced activation in the amygdala in response to fearful faces (Hessl et al., 2007). These men also had impaired startle potentiation while viewing fearful faces, implicating amygdalar dysfunction. However, earlier mechanistic analyses using rodent models of FXS focused almost exclusively on the hippocampus and neocortex, not the amygdala. These studies gave rise to the ''mGluR theory'' of FXS, an influential framework for understanding the synaptic and molecular underpinnings of FXS (Bear et al., 2004). This theory proposes that various symptoms of the disorder are the consequences of unchecked activation of mGluR5, a group 1 metabotropic glutamate receptor. A major observation leading to this theory is that mGluR5 is involved in long-term depression (LTD) (Fitzjohn et al., 2001;Palmer et al., 1997), a form of hippocampal synaptic plasticity that is elevated in the Fmr1 knockout (Fmr1 À/y ) mouse (Hou et al., 2006;Huber et al., 2002;Nakamoto et al., 2007;Till et al., 2015). This is significant in light of the contrasting nature of mGluR-dependent synaptic plasticity in the hippocampus versus amygdala. In the lateral amygdala, mGluR5 mediates long-term potentiation (LTP) of synaptic strength (Rodrigues et al., 2002). Although hippocampal mGluR-LTD is enhanced in Fmr1 À/y mice, the opposite effectimpaired LTP-was reported in the amygdala of these mice (Suvrathan et al., 2010). In other words, the direction of synaptic plasticity mediated by mGluRs (i.e., LTP versus LTD) and its dysfunction (impairment versus enhancement) in Fmr1 À/y mice are both different in the amygdala compared to the hippocampus. As predicted by the mGluR theory, downregulating mGluR signaling corrects multiple FXS-induced abnormalities in the hippocampus and neocortex (Dö len et al., 2007). In contrast, an mGluR-antagonist failed to reverse deficient amygdalar LTP in Fmr1 À/y mice (Suvrathan et al., 2010). These observations underscore the need to modify the current mGluR-based framework to better explain FXS-related synaptic dysfunction in the amygdala.
How do these divergent patterns of mGluR-plasticity, and their dysfunction in FXS, affect hippocampal versus amygdalar function in the intact animal? Despite extensive analyses of the molecular and synaptic signaling defects in the hippocampus of FXS mice, very few assessed how these alterations lead to specific behavioral abnormalities. For instance, abnormalities in hippocampal mGluR-LTD, and its reversal, have been studied in considerable detail. LTD and mGluR5 expression levels in the hippocampus also modulate spatial learning performance (Kemp and Manahan-Vaughan, 2004;Braunewell, 1999, 2005) However, the impact of defects in mGluR-LTD on hippocampal circuit function in vivo is not clear and has not yet been linked to specific deficits in learning and memory in mouse models of FXS.
An effective strategy to bridge these gaps comes from rodent models of fear memory, for which the neural circuitry has been characterized extensively across biological scales in the amygdala (Johansen et al., 2011;Maren and Quirk, 2004;Ressler and Maren, 2019;Tovote et al., 2015). Of direct relevance to FXS and mGluR-signaling are findings that mGluR5 in the lateral amygdala plays a pivotal role in fear conditioning and in LTP at synapses involved in fear conditioning (Rodrigues et al., 2002). Further, in vivo activation of mGluR5 receptors in the lateral amygdala enhances cue-specific fear, in addition to facilitating LTP in vitro (Rahman et al., 2017). Moreover, the gap between the synaptic and behavioral levels can be bridged in a fear conditioning-based framework because recordings in freely behaving animals have shown acquisition of conditioned fear responses to be associated with LTP-like physiological changes in vivo in the lateral amygdala (Rogan et al., 1997). An equivalent framework is not available with respect to hippocampal mGluR-LTD. Thus, as a model system, the amygdala offers a way to address several unresolved issues relevant to FXS. For instance, previous findings on synaptic dysfunction caused by FXS were gathered from in vitro measurements in amygdalar slices (Suvrathan et al., 2010), however, the functional consequences of these changes, at the circuit and behavioral levels, remain unexplored in the intact animal. At the other end, earlier studies on fearrelated behavior have yielded mixed results wherein knockout mice showed either impaired fear recall (Paradee et al., 1999) or no difference compared to normal mice (Dobkin et al., 2000;Hamilton et al., 2014;Peier et al., 2000). Further, in vivo electrophysiological analysis of these behaviors has not been attempted in rodent models of FXS. It is also unclear if pharmacological manipulations of mGluRs can correct potential abnormalities in amygdala-dependent conditioned fear and its synaptic correlates. Specifically, would the mGluR theory of FXS also hold in the amygdala despite the divergent patterns of plasticity defects seen in this brain area? The aim of the present study is to address these questions at multiple levels of neural organization in the amygdala of a rat model of FXS. Such analyses in the amygdala are essential for the development of novel strategies to treat the affective symptoms of FXS.

RESULTS
As a first step in our analysis of how FXS disrupts amygdaladependent behavior, we subjected Fmr1 À/y rats to auditory fear conditioning, wherein animals rapidly learn to associate a previously neutral tone (conditioned stimulus [CS]) with a coincident aversive stimulus (unconditioned stimulus [US]). Re-exposure to the CS alone evokes a cessation of locomotor activity, or ''freezing,'' which serves as a behavioral measure of the learned association. Accumulating evidence has established that plasticity in the basolateral amygdala (BLA) is essential for encoding fear memories (Blair et al., 2001;Rodrigues et al., 2001;Wilensky et al., 1999). Therefore, we characterized the effects of loss of FMRP on conditioned fear and then examined the underlying in vivo and in vitro mechanisms at lower levels of neural organization.
Impaired recall of conditioned fear in Fmr1 À/y rats Rats chronically implanted with recording electrodes in the BLA first underwent habituation to the context (days 1 and 2) (Figure 1A) and then to the tone that was subsequently used as the CS for repeated pairings with a foot shock (US) (day 3) (Figure 1A). Both wild-type (WT) and Fmr1 À/y rats showed higher levels of freezing at the end of the CS-US pairings compared to tone habituation ( Figure 1B) (unpaired t test, p < 0.001) (i.e., they were both capable of learning the tone-shock association). 24 hours later, Fmr1 À/y animals exhibited significantly lower freezing, compared with WT animals, when presented with only the CS in a different context (testing, day 4) ( Figure 1C). Together, these results point to impaired recall of fear memories in Fmr1 À/y rats. Impaired potentiation of CS-evoked responses in the BLA of fear conditioned Fmr1 À/y rats Next, we examined the neural basis of this memory deficit by recording CS-evoked local field potentials in the BLA of the same freely behaving rats ( Figure 1D). Specifically, amplitudes of auditory evoked potentials (AEP), in response to the tone CS, were measured as the difference between the first maxima and first negative peak of the CS-evoked AEP ( Figure 1E). During fear recall, AEPs in Fmr1 À/y rats exhibited a significant reduction in amplitude (Figures 1E and 1F) as well as slope ( Figure S1A) compared to WT rats. Because the same animals were used to simultaneously monitor changes in freezing behavior and in vivo recordings of AEPs in response to the same presentations of the CS, we also quantified the correlation between the two measures. This analysis revealed a significant positive correlation between the behavioral and electrophysiological responses ( Figures 1G and S1B). Notably, WT animals exhibiting robust potentiation of CS-evoked AEPs in the BLA also responded to the CS with higher freezing, thereby clustering in the upper right quadrant of the correlation plot ( Figure 1G). On the other hand, data points for the Fmr1 À/y rats were clustered in the lower left quadrant, indicating deficits in both behavioral and electrophysiological indices of fear recall. Finally, increase in CS-evoked theta power has been identified as a neural correlate of conditioned fear (Likhtik et al., 2014). Consistent with this, BLA theta power (measured as the power of auditory evoked responses in the 2-12 Hz frequency band) exhibited a significant increase in WT rats but not Fmr1 À/y rats during fear recall (Figures S1C and S1D).
Deficient LTP and excitatory synaptic transmission in the LA of Fmr1 À/y rats There is a significant body of evidence that acquisition of fear memory is associated with LTP of synaptic transmission at thalamic inputs to the LA (Bauer et al., 2001;McKernan and Shinnick-Gallagher, 1997;Rogan et al., 1997). Hence, we reasoned that impaired LTP is likely to underlie the deficits seen in our behavioral and in vivo analyses in the intact animal. To test this, we compared LTP in principal neurons of the LA using whole-cell current-clamp recordings in coronal brain slices prepared from Fmr1 À/y and WT rats ( Figure 2A). We monitored excitatory postsynaptic potentials (EPSPs) elicited by stimulation of thalamic inputs to LA neurons. Consistent with earlier reports, in brain slices from WT rats, two trains of 100 pulses at 30 Hz resulted in robust LTP. However, the same induction protocol failed to elicit any significant LTP in slices prepared from Fmr1 À/y rats ( Figures 2B and 2C). We next examined the overall status of basal synaptic transmission in these LA neurons by comparing the frequency and amplitude of miniature excitatory postsynaptic currents (mEPSCs) in slices from Fmr1 À/y and WT rats. Fmr1 À/y neurons exhibited a significant reduction in mEPSC frequency ( Figures 2D and 2E) but not amplitude ( Figure S2) compared to WT neurons in the LA, indicating a reduction in basal synaptic transmission.
Reduction in surface GluA1 levels in the BLA of Fmr1 À/y rats Next, we probed these impairments in synaptic transmission and plasticity in the BLA at the level of postsynaptic glutamate receptors. Earlier studies have shown that the impact of exaggerated signaling through mGluR5, caused by a loss of the translational repressor FMRP, is manifested in hippocampal area CA1 as enhanced internalization of the AMPAR subunit, GluA1 (Muddashetty et al., 2007;Nakamoto et al., 2007;Snyder et al., 2001). Specifically, this abnormally high GluA1 internalization has been linked to greater mGluR-LTD in the hippocampus. If a similar mechanism is in play in the amygdala, it would disrupt the stabilization of LTP ( Figures 2B and 2C). We investigated Figure 1. Fmr1 À/y rats exhibit impaired recall of auditory fear memory and learning-induced enhancement of CS-evoked responses in the BLA (A) Experimental protocol for auditory fear conditioning. Day 1 and 2, context habituation. Day 3, tone habituation and auditory fear conditioning (CS = 10 s, US = 1 s). Day 4, testing. Auditory evoked potentials (AEPs) were recorded in vivo from the BLA in response to all tones during the tone habituation and testing sessions. (B) There was no difference between the freezing levels of WT and Fmr1 À/y rats during each of the habituation tones. Unpaired t test, p > 0.05. However, although both groups showed an increase in freezing after consecutive CS-US pairings compared to tone habituation (unpaired t test, p < 0.001), Fmr1 À/y rats froze less than WT rats at the end of the pairings. Two-way RM ANOVA, factor: genotype, F (1,14) = 13.69, **p < 0.01. (C) Fmr1 À/y rats showed reduced CS-induced freezing during testing in a different context on day 4. Unpaired t test, ***p < 0.001. (D) Representative photomicrograph (left) and schematic coronal sections (right) showing placement of recording electrodes in the BLA. (E) Representative traces showing AEPs recorded from the BLA in response to the CS during tone habituation (dotted lines) and testing (solid lines). The amplitudes of CS-evoked AEPs were calculated as the difference between the maxima (dot) after the onset of the tone and the negative peak (arrowhead). (F) Percent change in mean amplitude of AEPs normalized to tone habituation (dotted line). Learning-induced increase in the amplitude of AEPs during testing was seen in WT, but not Fmr1 À/y rats. Unpaired t test, **p < 0.01. (G) AEP amplitude was positively correlated with the percentage of freezing levels in the same animal during testing of fear memory. Pearson's correlation, r = 0.64, p < 0.01. WT, N = 8; Fmr1 À/y , N = 8. Data are represented as mean ± SEM. See also Figure S1.
Cell Reports 37, 109805, October 12, 2021 3 Article ll OPEN ACCESS this possibility by quantifying levels of biotin-labeled surface GluA1 in BLA slices from Fmr1 À/y and WT rats. We found a reduction in GluA1 surface expression in Fmr1 À/y rats ( Figures  2F and 2G), which is in agreement with earlier findings in both the amygdala and hippocampus of Fmr1 À/y mice.

Effects of mGluR-activation on synaptic transmission in the amygdala
The results presented so far identify a range of deficits in amygdalar function in Fmr1 À/y rats, from impaired fear learning to a reduction in synaptic plasticity and transmission. The mGluR theory of FXS predicts that inhibition of mGluR activity can correct FXS-induced aberrations in the hippocampus (Bear et al., 2004;Dö len et al., 2007). Contrary to this prediction, a previous study found that pharmacological inactivation of mGluR5 with 2methyl-6-phenylethynyl-pyridine (MPEP) failed to rescue deficient LTP in the LA of Fmr1 À/y mice (Suvrathan et al., 2010). Perhaps this is not surprising because MPEP is known to block mGluR-dependent LTP (Rodrigues et al., 2002), which was already impaired in the LA of Fmr1 À/y mice. However, a separate study in rats showed that the opposite manipulation-in vitro activation of group 1 mGluRs in the LA using the agonist (RS)-3,5-dihydroxyphenylglycine (DHPG)-caused a robust facilitation of weak LTP, although DHPG by itself did not induce LTP (Rahman et al., 2017). How DHPG achieves its facilitating effects on synaptic transmission in the LA remains unclear.
Studies in the hippocampus reported that DHPG treatment reduces the surface expression of postsynaptic AMPA receptors, a process thought to underlie mGluR-dependent LTD (Fitzjohn et al., 2001;Snyder et al., 2001;Xiao et al., 2001). Hence, we first quantified the impact of bath applied DHPG on the levels of biotin-labeled surface GluA1 in BLA slices ( Figure 3A, top). We found a persistent reduction in GluA1 surface expression in slices treated with 50 mM DHPG ( Figure 3B). Thus, the reduction in postsynaptic GluA1 surface expression in the amygdala is similar to that reported previously in the hippocampus. Therefore, we reasoned that the difference in the electrophysiological effects on synaptic transmission between the two brain areas is likely to lie on the presynaptic side. Hence, we recorded mEPSCs from the same LA principal neurons before, during, and after bath application of DHPG ( Figure 3A, middle). Using this within-cell comparison, we found that the same in vitro application of DHPG, despite reducing postsynaptic surface GluA1, increased the frequency of mEPSCs that persisted even after the drug was washed out ( Figures 3C and 3D). There was no change in the amplitude of mEPSCs ( Figure S3A). Further, DHPG failed to alter the mEPSC frequency in the presence of an mGluR5 antagonist, 3-[(2-methyl-1,3-thiazol-4-yl) ethynyl] pyridine (MTEP) ( Figures S3B-S3D), indicating that the increase in frequency was due to DHPG-induced activation of mGluR5. Because this increase in mEPSC frequency is suggestive of changes in presynaptic release probability, we also analyzed the effects of DHPG on paired-pulse facilitation at thalamic inputs to LA neurons. Specifically, we measured paired-pulse ratios (PPR) of evoked EPSCs before, during, and after bath application of DHPG in the same LA neuron ( Figure 3A, bottom). This within-cell comparison revealed that DHPG also causes a reduction of PPRs, indicating an increase in presynaptic release Figure 2. LA neurons in Fmr1 À/y rats show impaired synaptic transmission and plasticity (A) Placement of stimulating electrode, activating thalamic afferents, and recording electrodes in the lateral amygdala (LA) in a coronal brain slice. (B) 30 Hz tetanus (2 trains of 100 pulses, delivered 10 s apart; arrow) induced LTP in principal neurons of slices obtained from WT (n = 10) but not Fmr1 À/y (n = 9) rats. Two-way RM ANOVA, factor: genotype F (1,17) = 5.61, p < 0.05. Inset: superimposed representative EPSP traces before (gray, WT; pink, Fmr1 À/y ) and after (black, WT; red, Fmr1 À/y ) 30-Hz stimulation of thalamic afferents. (C) Summary of LTP experiments showing normalized EPSP slopes averaged over 25-30 min after tetanic stimulation. LA neurons in Fmr1 À/y rats exhibited impaired LTP. Unpaired t test, *p < 0.05. (D) Significantly higher frequency of mEPSCs recorded in LA neurons of WT rats (n = 13) compared to that in Fmr1 À/y rats (n = 8; unpaired t test, *p < 0.05) as evidenced by a rightward shift in the cumulative probability plot of inter-event intervals (Kolmogorov-Smirnov [KS] test, p < 0.001) and reduced mean frequency (inset, unpaired t test, *p < 0.05) in Fmr1 À/y LA neurons. (E) Representative mEPSC traces. (F) Representative western blots showing levels of surface GluA1 normalized to avidin. FMRP was absent in tissue from Fmr1 À/y rats. (G) Surface expression of GluA1 in the BLA was reduced in Fmr1 À/y rats (N = 7) compared to WT rats (N = 6). Unpaired t test, *p < 0.05. Data are represented as mean ± SEM. See also Figure  Presynaptic mGluR5 in the amygdala Postsynaptic mechanisms of mGluR signaling and plasticity have received greater attention in published analyses of FXSinduced changes in hippocampal area CA1. Analyses using a mouse model of FXS revealed lower presynaptic release as evidenced by a decrease in the frequency mEPSCs, increased PPR, and slower use-dependent block of NMDA receptor currents in the LA (Suvrathan et al., 2010). Further, in vitro mGluRactivation using DHPG enhances presynaptic release (Figure 3), which is also consistent with previous findings (Rahman et al., 2017). However, an earlier study on mGluR-dependent plasticity and fear learning, using light and electron microscopy, found that mGluR5 is predominantly located in postsynaptic structures in the LA (Rodrigues et al., 2002). To reconcile the different loci of mGluR action, we examined the localization of mGluR5 in the amygdala and hippocampus. To assess the distribution of mGluR5 relative to the pre-and postsynaptic compartment with high spatial resolution, we used super-resolution protein retention expansion microscopy together with immunocytochemistry (Tillberg et al., 2016). With an increased resolution (of up to $43), expansion microscopy allowed us to better measure and quantify mGluR5 localization to either the pre-or postsynaptic compartment (Chien et al., 2015;Hafner et al., 2019). We examined the localization of mGluR5 relative to pre-and postsynaptic terminal markers, RIM1 (Kaeser Top: the levels of surface GluA1 receptors were measured in BLA slices from control rats after a 10-min bath application of 50 mM DHPG followed by a 30-min washout. Middle: voltage-clamp recordings of mEPSCs were made before (5-min baseline), the last 5 min of 10-min bath application of 100 mM DHPG, and during the last 5 min of a 30-min washout each in the same LA neuron. ( ) Indicates the time points at which mEPSCs were recorded. Bottom: paired-pulse ratios (PPRs) in response to stimulation of thalamic afferents were measured in the same LA neuron before, during, and 15 min after bath application of 100 mM DHPG. (Y) Indicates the time points at which PPRs were monitored. (B) DHPG induced a reduction in surface GluA1 levels in the BLA. Top: representative western blots showing surface levels of GluA1 normalized to avidin in control and DHPG treated slices. Unpaired t test, **p < 0.01. Control, N = 9; DHPG, N = 9. (C) DHPG increased the frequency of mEPSCs in LA neurons, which remained elevated 30 min after washout. One-way RM ANOVA, F (1,18) = 7.59, Bonferroni's test **p < 0.01. There was a significant leftward shift in the cumulative probability histogram of inter-event intervals during DHPG application, as well as washout, compared to baseline. KS test, **p < 0.01, n = 12. (D) Representative mEPSC traces before (baseline), during (DHPG) and after washout of DHPG. (E) Bath application of 100 mM DHPG decreases PPRs recorded in LA neurons. One-way RM ANOVA, F (1,9) = 38, Bonferroni's test ***p < 0.001, n = 8. (F) Sample of averaged EPSC traces at 25-ms inter-stimulus intervals before, during, and after DHPG application. Data are represented as mean ± SEM. See also Figure (Ehrengruber et al., 2004), in the BLA ( Figures 4A and 4C, top) and CA1 region of the hippocampus ( Figures 4A and 4C, bottom). We assessed the signal distribution of mGluR5 relative to the pre-and postsynapse, by calculating the peak-signal to peak-signal distance of the immunolabeled puncta for mGluR5 and RIM1 or Homer1 in both the BLA and CA1 (see STAR Methods). We found that mGluR5 was positioned closer to the postsynaptic protein Homer1 in the CA1 area compared to the BLA ( Figure 4B). In contrast, mGluR5 was localized closer to the presynaptic protein RIM1 in the BLA than in CA1 ( Figure 4D). Taken together, these data indicate a predominantly presynaptic localization for mGluR5 in the BLA.
The analysis in the BLA using expansion microscopy reveals the presence of presynaptic mGluR5 in the BLA. If this is indeed adequate for DHPG to enhance mEPSC frequency, then disrupting the signaling cascade triggered by postsynaptic mGluR5 in (B) Homer1 and mGluR5 had a greater peak-topeak distance in the BLA compared to the CA1. Unpaired t test, *p < 0.05. N = 3 experiments. (C) Representative images of the BLA (top) and the CA1 region of the hippocampus (bottom) showing localization and overlap of mGluR5 (green) and RIM1 (red, presynaptic marker). Scale bar, 2 mm. (D) RIM1 and mGluR5 had a shorter peak-to-peak distance in the BLA compared to the CA1. Unpaired t test, *p < 0.05. N = 3 experiments. (E) DHPG increased mEPSC frequency in LA neurons even after chelating intracellular Ca 2+ using 10 mM BAPTA. One-way RM ANOVA, F (1,12) = 9.95, Bonferroni's test **p < 0.01. n = 8. Bath application of DHPG causes a significant leftward shift in in the cumulative probability histogram of inter-event intervals compared to baseline, and this persists after washout. KS test, p < 0.05. (F) Representative mEPSC traces. Data are represented as mean ± SEM. See also Figure S4. the recorded neuron should not diminish the presynaptic effects of DHPG manifested as electrophysiological changes in mEPSC frequency. To test this, we added the Ca 2+ -chelator, 1,2-bis(2-aminophenoxy) ethane-N,N,N 0 ,N 0 -tetraacetic acid (10 mM BAPTA) in the pipette solution while recording the effects of bathapplied DHPG on LA mEPSCs (using the same protocol depicted in Figure 3A, middle). Strikingly, this manipulation did not prevent the DHPG-induced enhancement in mEPSC frequency in the LA ( Figures  4E, 4F, and S4). This argues strongly for a presynaptic mechanism of action for DHPG in enhancing synaptic transmission in the LA.
mGluR-activation corrects deficient synaptic transmission and LTP in the LA Taken together, these results suggest that activation of mGluRs using DHPG may be well suited to reverse the range of amygdalar deficits we have identified in the Fmr1 À/y rats. We tested this at three different levels of neural organization (Figures 5 and 6). First, because DHPG enhances mEPSC frequency and the same measure of basal synaptic transmission is suppressed in LA neurons of Fmr1 À/y rats ( Figures 2D and 2E), we examined the effects of mGluR-activation in LA slices prepared from Fmr1 À/y rats. To this end, we recorded mEPSCs from LA principal neurons before, during, and after bath application of 100 mM DHPG (using the same protocol depicted in Figure 3A, middle). This within-cell comparison revealed that 100 mM DHPG causes a significant increase in mEPSC frequency in Fmr1 À/y neurons that reached levels comparable to baseline transmission in WT neurons ( Figure 5A). This enhancement persisted even 30 min after DHPG was washed out (Figures 5A and 5B). As in the case of WT neurons, DHPG did not change the amplitude of mEPSCs in LA neurons from Fmr1 À/y rats (Figure S5). Next, we examined the effects of DHPG bath application on LTP in LA slices applying the same induction protocol used earlier (Figures 2A and 2B). After incubation with 100 mM DHPG, tetanic stimulation of thalamic inputs to LA principal neurons triggered robust LTP in Fmr1 À/y slices, which was significantly greater than the impaired LTP seen in Fmr1 À/y LA neurons in the absence of DHPG (Figures 5C-5E). Thus, mGluR-activation also reversed the deficient LTP in amygdalar slices from Fmr1 À/y rats. mGluR-activation in the amygdala also restores normal fear learning in Fmr1 À/y rats What are the behavioral consequences of these synaptic changes induced by mGluR-activation? Specifically, will the reversal of the deficits in basal synaptic transmission and LTP in the LA also restore recall of conditioned fear in the intact Fmr1 À/y rat? To address this question, we combined the same fear conditioning procedure used earlier ( Figure 1A) with targeted in vivo infusion of saline or DHPG directly into the BLA of freely behaving Fmr1 À/y and WT rats ( Figure 6A, top). Following context habituation, animals were subjected to bilateral in vivo infusions of saline into the BLA and 30 min later underwent habituation to the CS that was subsequently used for auditory conditioning ( Figure 6A, bottom). 24 h later, this conditioning led to a significant increase in the freezing response to the CS ( Figure 6B) in WT rats (i.e., robust recall of conditioned fear). However, the same conditioning was unable to produce any detectable change in CS-induced freezing in the saline-infused Fmr1 À/y rats 24 h later ( Figure 6B), thereby demonstrating an impairment in fear recall similar to that presented earlier in Figure 1. Next, a separate group of Fmr1 À/y and WT animals received in vivo infusions of DHPG into the BLA, followed by the same sequence of tone habituation and auditory conditioning ( Figure 6A, top). In contrast to saline, infusions of DHPG caused higher freezing in both control and Fmr1 À/y animals before the onset of the first tone (pre-CS) (day 1) (Figure 6A), as well as during subsequent tone habituation (Figure 6A). A day later, however, these animals did not exhibit enhanced pre-CS freezing in the testing context ( Figure S6A) but only a selective increase in freezing to the CS (testing, day 2) ( Figure 6B). Notably, the levels of CS-induced freezing during the testing session were indistinguishable between the DHPGtreated Fmr1 À/y and WT rats ( Figure 6B). Thus, auditory conditioning, in conjunction with simultaneous in vivo activation of mGluR in the BLA, restored normal recall of fear memory in Fmr1 À/y rats. Figure 5. In vitro mGluR-activation corrects deficits in synaptic transmission and plasticity in the LA of Fmr1 À/y rats (A) Bath application of 100 mM DHPG reversed the decrease in mEPSC frequency in LA neurons of Fmr1 À/y rats. DHPG application increased mEPSC frequency in Fmr1 À/y neurons back to WT levels (dotted line), which persisted until the end of the 30-min washout. One-way RM ANOVA, F (1,12) = 14, Bonferroni's test *p < 0.05, **p < 0.01. The cumulative frequency histogram of inter-event intervals showed a leftward shift during DHPG application and washout, overlapping with baseline levels of WT neurons. KS test, **p < 0.01. WT, n = 12; Fmr1 À/y , n = 8. (B) Representative mEPSC traces. (C) Representative EPSP traces before (gray, WT; pink, Fmr1 À/y ) the 30-Hz tetanus and up to 30 min after (black, WT; red, Fmr1 À/y ). (D) High-frequency 30 Hz stimulation (arrow) did not induce LTP in LA neurons from Fmr1 À/y rats. However, after pre-incubation with DHPG, the same 30-Hz tetanus induced robust LTP in Fmr1 À/y neurons. Two-way RM ANOVA, factor: DHPG F (1,15) = 6.12, *p < 0.05. (E) Mean values of LTP from minutes 25 to 30, normalized to the 5-min pre-tetanus baseline. DHPG incubation reversed impaired LTP in Fmr1 À/y LA neurons, causing no further increase in WT LTP. Two-way ANOVA factor: interaction F (1,33) = 4.92, *p < 0.05. Tukey's test, *p < 0.05. WT, n = 10; Fmr1 À/y , n = 9; WT DHPG, n = 8; Fmr1 À/y DHPG, n = 10. Data are represented as mean ± SEM. See also Figure  In vivo activation of mGluRs in the BLA restores learning-induced increase in behavioral and in vivo electrophysiological responses to the CS in Fmr1 À/y rats These results demonstrating reversal of amygdalar deficits in FXS rats, using in vitro and in vivo DHPG treatment, were gathered from separate measurements across levels of neural organization. The DHPG-induced corrections achieved at the level of synaptic plasticity and fear memory recall are consistent with each other, suggesting a link between them. To strengthen this link further, we attempted to integrate these different measures with DHPG infusion into a single experiment. To this end, we monitored if DHPG infusions into the BLA simultaneously corrected deficient recall of fear memory and its underlying in vivo correlate -impaired learning-induced potentiation of CS-responses in the BLA -in the same animal. Here too, we used the same fear conditioning protocol ( Figure 6D) with targeted bilateral in vivo infusion of saline or DHPG directly into the BLA of freely behaving Fmr1 À/y and WT rats, while carrying out in vivo recordings of the changes in CS-evoked AEPs unilaterally from the BLA (Figures 6C, S6B, and S6C). First, consistent with results shown in Figure 6B, in vivo mGluR activation in the BLA once again restored normal recall of conditioned fear in Fmr1 À/y rats (data not shown). Next, in these same rats, simultaneous recordings of CS-evoked responses in the BLA ( Figure 6E) allowed us to compare baseline AEPs (dotted lines) recorded on day 1 before infusion of saline/DHPG with those recorded during testing of fear recall (solid lines) on day 2 in both WT (black) and Fmr1 À/y (red) rats. In rats receiving saline infusion into the BLA, auditory fear conditioning caused an increase in AEP amplitude in WT, but not Fmr1 À/y , rats ( Figure 6F). In striking contrast, in conditioned Fmr1 À/y rats receiving DHPG infusions in the BLA, learning-induced potentiation of CS-evoked AEPs was restored ( Figure 6F). Thus, in conditioned Fmr1 À/y rats receiving DHPG infusions in the BLA, the deficient potentiation of CS-evoked responses in the BLA was prevented, allowing normal recall of fear memory. The deficit in learning-induced increase in theta power, seen in Fmr1 À/y rats, was not fully restored by DHPG infusion into the BLA (Figures S6D and S6E). Finally, if learning-induced potentiation of CS responses in the BLA is indeed the substrate underlying normal conditioned fear seen, a comparison of the AEP amplitudes measured in individual animals with the freezing levels exhibited by them should reflect a correlation. Indeed, behavioral and in vivo electrophysiological data obtained from rats in all four experimental groups displayed a positive correlation between the mean percentage change in CS-evoked AEPs and percentage freezing (Figure 6G), supporting a role for learning-induced enhancement in CS responses in the BLA and normal fear recall. Specifically, the groups that exhibited strong potentiation of BLA responses also showed higher freezing to CS (WT-saline, WT-DHPG), thereby clustering in the upper right quadrant of the correlation plot. In contrast, data points for the Fmr1 À/y -saline group were clustered toward the lower left quadrant, indicating deficits in both measures. However, the data points from Fmr1 À/y rats that received DHPG infusions into the BLA overlapped with those from the WT rats in the upper right quadrant. Thus, the same in vivo mGluR-activation that restored normal recall of conditioned fear also reversed its deficient in vivo encoding in the BLA in the same animal.

DISCUSSION
Analyses of various abnormalities in synaptic function and molecular signaling in rodent models of FXS have given rise to a powerful framework, including the ''mGluR theory,'' for identifying molecular targets for therapeutic strategies against FXS. However, the focus of most of these foundational studies was on the hippocampus. Here, we present findings that add a new dimension to our understanding of FXS-induced dysfunction in the amygdala, which remains less explored despite its central role in the emotional symptoms of the disorder. Motivated by differences between the amygdala and hippocampus in mGluRdependent modulation of synaptic transmission, we addressed two key questions. First, behavioral characterization of a rat model of FXS revealed deficient recall of conditioned fear. This, in turn, led us to examine both in vivo and in vitro measures of plasticity underlying fear learning in a specific circuit-the Figure 6. Targeted in vivo mGluR-activation in the BLA restores normal fear learning and its neural correlates in Fmr1 À/y rats (A) Inset top: WT and Fmr1 À/y rats received bilateral in vivo infusions (1.0 mL per side) of either saline (0.9% NaCl) or DHPG (50 mM) into the BLA 30 min before tone habituation and were then subjected to auditory fear conditioning (day 1); recall of conditioned fear was tested on day 2. DHPG infusion increased mean freezing levels in both WT and Fmr1 À/y rats during the pre-CS and tone habituation session. Two-way RM ANOVA factor: DHPG F (5,230) = 11.04, ***p < 0.001. All 4 groups showed an increase in freezing during subsequent conditioning (7 CS-US pairings: CS = 5 kHz tone, 30 s; US = 0.5 mA foot shock, 1 s). (B) Although Fmr1 À/y rats with saline infusion into the BLA showed impaired recall of conditioned fear during testing, DHPG infusion restored normal recall of fear memory. Two-way ANOVA factor: interaction F (1,44) = 8.77, **p < 0.01. All groups N = 12. (C) The same saline/ DHPG in vivo infusion protocol, using bilateral cannula (dotted lines), was combined with unilateral AEP recordings in the BLA (arrow head) and simultaneous behavioral measurements of freezing. (D) Experimental protocol for auditory fear conditioning (bottom), combined with in vivo infusion into, and recording from, the BLA (top). Day 1, baseline CSevoked AEPs were measured before saline/DHPG infusion, followed 30 min later by tone habituation and conditioning (CS = 10 s, US = 1 s, 0.5 mA). Day 2, CSevoked AEPs were measured again during recall of fear memory (testing). (E) Representative AEP traces recorded from the BLA during baseline (dotted lines) and during testing (solid lines) on for WT (black) and Fmr1 À/y (red) rats. The AEP amplitude was calculated as the difference between the maxima (dot) after the onset of the tone and the negative peak (arrowhead). (F) Percent change in mean amplitude of AEPs normalized to tone habituation (dotted line). A learning-induced increase in the amplitude of AEPs during testing was seen in both WT groups but not Fmr1 À/y rats receiving saline infusions. In contrast, Fmr1 À/y rats receiving DHPG infusions show a significant learning-induced enhancement in AEP amplitude during testing. Two-way ANOVA factor: interaction F (1,18) = 12.64, p < 0.01. Tukey's test **p < 0.01. WT saline, N = 7; Fmr1 À/y saline, N = 9; WT DHPG, N = 7; Fmr1 À/y DHPG, N = 8. (G) AEP amplitude (normalized to tone habituation) was positively correlated with the percentage of freezing levels in the same animal during testing of fear memory. Pearson's correlation, r = 0.45, p < 0.05. Data are represented as mean ± SEM. See also Figure S6.
Cell Reports 37, 109805, October 12, 2021 9 Article ll OPEN ACCESS lateral amygdala (LA)-that is known to mediate this behavior. Enhanced freezing behavior elicited by the tone CS was accompanied by increases in CS-evoked auditory-evoked field potentials in the intact LA of conditioned WT, but not Fmr1 À/y rats. This deficit was also evident at the synaptic level as impaired LTP in LA principal neurons. Consistent with this LTP deficit, postsynaptic surface expression of the AMPA receptor subunit, GluA1, was reduced in the LA. We also found evidence for reduced presynaptic release in the Fmr1 À/y rats manifested as a reduction in mEPSC frequency. This presynaptic effect, in turn, served as the starting point for a second line of inquiry-how these deficits associated with FXS may be reversed by targeting mGluRs. Strikingly, in contrast to earlier strategies involving inactivation of mGluR-signaling to correct abnormalities in the hippocampus (Dö len et al., 2007), pharmacological activation of mGluRs reversed the impairment in presynaptic release, as well as LTP, in the LA. mGluR-activation caused a reduction in postsynaptic surface expression of GluA1, suggesting a presynaptic mechanism through which DHPG strengthens synaptic transmission. Consistent with this, our analysis demonstrates the presence of mGluR5 in the presynaptic compartment of LA synapses. We confirmed the functional consequences of activating these presynaptic receptors by showing that DHPG was capable of enhancing mEPSC frequency even when signaling triggered by postsynaptic mGluRs was blocked by chelating Ca 2+ in the postsynaptic neuron. Upon probing the benefits of this intervention at the behavioral level, we found that in vivo mGluR-activation in the BLA of freely behaving Fmr1 À/y rats reversed deficient fear recall. Finally, this behavioral rescue was accompanied by a simultaneous reversal of the circuit level deficit in the same animalthe impairment of learning-induced enhancement of CS responses was also corrected by in vivo mGluR-activation in the amygdala. Together, these results demonstrate the contrasting nature of FXS-induced defects in the amygdala compared to previous findings in the hippocampus. Importantly, this led us to adopt an opposite pharmacological strategy to correct amygdalar aberrations-from synapses through circuit to behaviorin Fmr1 À/y rats. Thus, group 1 mGluR signaling may still be an effective target for therapeutic interventions against amygdalar defects caused by FXS but in a manner that takes into account circuit-specific differences and presynaptic mechanisms.
Deficient excitatory synaptic transmission and plasticity in the amygdala, and its behavioral manifestation as impaired recall of cue-specific fear, are consistent with accumulating clinical evidence suggesting disruption of appropriate encoding of fearrelated information seen in FXS individuals. For example, functional MRI revealed reduced amygdalar activity in affected individuals while viewing fearful facial expressions compared with other stimuli (Kim et al., 2014). Notably, the decrease in amygdala activation in these individuals was fear-specific because this was not seen when participants were viewing happy compared with other types of facial expressions. In men with the FMR1 premutation, attenuated amygdalar activation has also been seen during an emotion-matching task, as well as impaired startle potentiation while viewing fear faces (Hessl et al., 2011). Consistent with these neuroimaging studies, emotion-potentiated startle, a probe of amygdala activation, was found to be reduced in children and adolescents with FXS compared to a typically developing control group (Ballinger et al., 2014).
In the broader context of earlier findings, our analyses reveal differences, as well as common endpoints, underlying FXSrelated synaptic defects in the amygdala and hippocampus. First, we found lower surface GluA1 in BLA slices, similar to that reported in cultured hippocampal neurons (Nakamoto et al., 2007). Second, DHPG application caused a postsynaptic reduction of surface GluA1 in the BLA, which is also similar to what has been seen in hippocampal neurons (Fitzjohn et al., 2001;Snyder et al., 2001;Xiao et al., 2001). Thus, the postsynaptic effects of both FXS and DHPG-reduction of surface GluA1-are similar in the amygdala and hippocampus. However, where the two brain areas appear to differ is on the presynaptic side. Consistent with earlier reports of lower presynaptic release in the LA of a mouse model of FXS (Suvrathan et al., 2010), we report reduced mEPSC frequency in the present study. However, no such reduction in presynaptic release has been observed in hippocampal area CA1 of adult Fmr1 À/y mice (Braun and Segal, 2000;Pfeiffer and Huber, 2007). In fact, the opposite effect-higher spontaneous EPSC frequencies and lower paired-pulse ratios-have been reported in the hippocampus of younger Fmr1 À/y mice (Contractor et al., 2015;Tyzio et al., 2014;Klemmer et al., 2011). The most notable difference emerges from the divergent effects of the mGluR-agonist, DHPG, in the two structures. Although DHPG is known to induce LTD in hippocampal area CA1, we show here that it reverses the impairment of LTP in the LA of Fmr1 À/y rats, despite the baseline reduction in postsynaptic surface AMPARs. Moreover, although an earlier study (Rodrigues et al., 2002) reported postsynaptic localization of mGluR5 in the LA, here we identify presynaptic mGluR5 in the LA, providing a basis for the presynaptic effects of DHPG in the amygdala. The synaptic effects of DHPG in the hippocampus and amygdala are particularly interesting in light of our observations using expansion microscopy, which revealed the existence of both pre-and postsynaptic mGluR5 in these two brain areas. This analysis suggests a greater abundance of presynaptic mGluR5 in the BLA relative to the CA1 area, and more postsynaptic mGluR5 in area CA1 compared to the BLA. Hence, future studies will have to take into account both pre-and postsynaptic effects of mGluR-activation, as well as their brain region-specific variations. This is also interesting in light of accumulating evidence for the presence of fragile X mental retardation protein (FMRP) in presynaptic or axonal FMRP-containing granules (Akins et al., 2012(Akins et al., , 2017Christie et al., 2009). Specifically, FMRP, which regulates mRNA localization and translation, exhibits distinct brain region-specific patterns of expression in presynaptic compartments and its dysregulation may also contribute to the neurological symptoms of FXS . This warrants exploring if the levels of presynaptically localized FMRP are different between the LA and the hippocampus. Further, it was recently shown that fear conditioning elicits changes in the translatome in cortical axons that project to the LA (Ostroff et al., 2019). Notably, pathway analysis of the fear learninginduced changes in ribosome-bound mRNAs indicates that FMRP is one of the key upstream regulators of axonal protein synthesis in the LA. In the hippocampus, it is well known that DHPG stimulates postsynaptic protein synthesis (Raymond et al., 2000;Weiler and Greenough, 1993), which is disrupted in Fmr1 À/y mice (Bowling et al., 2019;Darnell and Klann, 2013;Hou et al., 2006;Osterweil et al., 2010;Waung and Huber, 2009). Thus, it will be important to determine whether fear conditioning-induced presynaptic protein synthesis in the LA requires mGluR5 and whether it is disrupted in Fmr1 À/y rats.
Finally, these differences in mGluR-dependent synaptic transmission, and its aberration in FXS, add to growing evidence for brain region-specific, even cell-type-specific, changes induced by the loss of FMRP (Contractor et al., 2015;Wang et al., 2014). This also poses a significant therapeutic challenge because a pharmacological strategy that is effective in one brain area, such as an mGluR5-antagonist in the hippocampus, may need to be modified for another area like the amygdala. Recent failures of clinical trials using mGluR5-antagonists may be indicative of some of these issues related to circuit-specific differences. For example, a randomized, double-blind, placebo-controlled trial using a selective mGluR5-antagonist did not reveal any significant effect on emotional function in FXS patients (Berry-Kravis et al., 2016). Thus, an essential step toward overcoming these challenges requires us to focus on specific behavioral symptoms of FXS and then probe the underlying functional changes at the synaptic and molecular levels within specific circuits underlying those particular behavioral deficits. In the present study, we attempted to achieve this by leveraging the advantages offered by auditory fear conditioning, a well-established model of fear learning for which the underlying neural circuitry has been characterized extensively. This enabled us to systematically probe if perturbations caused by FXS at one level of neural organization are consistent with predicted alterations at another. Findings gathered from such an approach offer a new framework, spanning biological scales, for understanding and modulating activity in a welldefined fear circuit and thereby suggesting strategies for treating affective symptoms associated with FXS.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

Lead contact
Further information and requests for resources and reagents should be directed to and will be fulfilled by the corresponding author, Dr. Sumantra Chattarji (shona@ncbs.res.in).

Materials availability
This study did not generate new unique reagents or biological samples.

Data and code availability
All data reported in this paper is available from the Lead Contact upon request.
All original code has been deposited at Zenodo and is publicly available as of the date of publication. The DOI is listed in the Key resources table.
Any additional information required to reanalyze the data reported in this work is available from the Lead Contact upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS
Male Fmr1 -/y Sprague Dawley rats and their wild-type (WT) littermate controls (8-10 weeks old, 300-350 g) were obtained from Sigma Advanced Genetic Engineering (SAGE) Labs (St. Louis, MO, USA) and maintained on a 14-h/10-h light/dark cycle with food and water provided ad libitum. All experimental subjects were group housed (2-5 animals/cage) to avoid the effects of isolation and randomly assigned to experimental groups. Experiments were done blind to genotype. Rats were handled for 2-3 days to habituate them to the experimenter before the start of each experiment. All experiments were conducted in accordance with the guidelines of the CPCSEA, Government of India and approved by the Institutional Animal Ethics Committees of the National Centre for Biological Sciences and the Institute for Stem Cell and Regenerative Medicine. inches deep x 12 inches high, no odor), while testing was done in a modified home cage (Context B: 14 inches wide x 8 inches deep x 16 inches high, mint odor). Lighting conditions, floors, and walls were different for each context. All chambers were cleaned with 70% alcohol before each experiment. Infrared LED cues, positioned on the walls of the chambers, were triggered in coincidence with auditory stimuli to monitor the tone-evoked freezing response offline. During context habituation (Days 1&2) ( Figure 1A), the rats were allowed to explore context A for 25 minutes each, on 2 consecutive days. During tone habituation (Day 3), the rats received five presentations of an auditory tone (5 kHz, 30 s tone consisting of 30 pips of 100 ms duration at a frequency of 1 Hz; 5 ms rise and fall, 70 ± 5 dB sound pressure level) in context A. This was immediately followed by the fear conditioning protocol, where the tone (CS) was paired (7 pairings, average inter-trial interval < ITI > = 120 s, with a range of 80-160 s) with a co-terminating 0.5 s scrambled foot shock (US; 0.5 mA). In the testing session (Day 4), the rats were introduced into context B and presented with the same tone (CS) to test recall of fear memory. Behavioral recordings were made using a video camera fixed to the wall of the sound isolation box and a frame grabber (sampling at 30 Hz). The videos were stored offline for further manual quantification of freezing behavior by a blind experimenter. Freezing was defined as the absence of movement except due to respiration. In addition to the time spent freezing to the tones, freezing levels were measured for a 10 s period (pre-CS) immediately before the start of the tone trials for every session to assess context-dependent fear. In experiments with targeted bilateral infusions (Figure 6), saline or DHPG was infused into the BLA 30 min before tone habituation. In experiments with both infusions and AEP recordings ( Figures 6C-6G), a baseline was included immediately before the infusion of saline or DHPG.

Surgical procedures and in vivo extracellular recordings
Rats were subjected to anesthesia with 5% isoflurane and then sustained in anesthesia with 1.5%-2% isoflurane. The level of anesthesia was frequently monitored throughout the procedure using the pedal withdrawal reflex to toe pinch and body temperature maintained with a heating pad. Burr holes were drilled at the stereotaxic co-ordinates of the BLA (3.3 mm posterior to bregma and ± 5.3 mm lateral to midline) and a bundle of 4 formavar-insulated stainless steel electrodes (50 mm diameter; AM Systems, Carlsborg, WA, USA) were implanted using the stereotaxic frame (8.3 mm ventral from the brain surface). The implant was secured using anchor screws and dental acrylic cement. One of the anchor screws was connected to the ground electrode. Similar surgical procedures were used during bilateral implantation of stainless-steel cannulae (Figures 6 and S6) for targeted infusion of DHPG into the BLA. Guide cannulae were implanted using the stereotaxic frame (7.0 mm ventral from the brain surface) and dummy cannulae (28 gauge, with a 0.5 mm projection) were inserted into them to prevent clogging. In experiments with both infusion and AEP recordings (Figures 6 and S6), bundled electrodes (as described above) were attached to one of the guide cannulae and implanted unilaterally. Rats were permitted to recover for 7-10 days following surgery. In the post-surgery period, the rats were singly housed in separate cages. AEPs were recorded using a unit gain buffer head stage (HS-36-Flex; Neuralynx, Bozeman, MT, USA) and the Digilynx data acquisition system (Neuralynx). Signals were amplified (1000X) and acquired at a sampling rate of 1 kHz followed by a band-pass filter (1-500 Hz) using Cheetah data acquisition software (Neuralynx).

AEPs -
The recorded AEPs were averaged over all the tone pips for the specified trial blocks. Averaged AEPs were quantified by measuring the amplitude and slope as the difference between the maxima after the onset of the response and the negative peak. AEP amplitudes and slopes were calculated during tone habituation and testing of fear memory. All AEP amplitudes and slopes were normalized as a percentage of the value during tone habituation ( Figure 1) or pre-infusion baseline ( Figures 6C-6G) for each animal.
Time-frequency analysis -Event related variations in spectral power were calculated by time-frequency analysis executed using continuous wavelet transformation (MATLAB, MathWorks Inc., Natick, MA, USA) on the averaged AEPs. Complex Morlet wavelets were used to compute the phase and amplitude of evoked responses within a frequency range from 2 to 100 Hz in steps of 0.1 Hz. The bandwidth parameter and center frequency of the mother wavelet were 2 and 1 Hz respectively. Subsequently, the wavelet power of the time series was calculated and expressed in decibels. Baseline average power for the duration of 0 to À200 ms was subtracted across all time points for each frequency bands. Tone evoked theta power was computed over the duration of 0 to 250 ms from tone onset for frequencies from 2 to 12 Hz. Targeted pharmacological infusion of DHPG into the BLA Bilateral infusion of DHPG in the BLA was done using standard pressure injection methods (Rahman et al., 2017). During the infusion procedure, the rats were retained in their home cages and an injection cannula with 1 mm projection (28 gauge, Plastic One, Roanoke, VA, USA) were inserted through the guide-cannula. Using polyethylene tubing, the injection cannula was connected to a Hamilton syringe (10 ml), mounted on an infusion pump (Harvard Apparatus, Holliston, MA, USA). Either vehicle (0.9% (vol/vol) NaCl (1.0 ml per side) or DHPG (1.0 ml per side, 50 mM in saline; Ab120007 Abcam, Cambridge, UK) was infused at a rate of 0.2 ml min -1 . The injection cannula was held in place for 5 min after the end of infusion, to permit the drug to diffuse into the tissue.

Histology
After the experiment was completed, rats were deeply anesthetized (ketamine/xylazine, 100/20 mg per kg) and electrolytic lesions (20 mA, 20 s) were made through the implanted cannulae and electrodes to mark the recording and infusion sites. The animals were