Impact of Chronic BDNF Depletion on GABAergic Synaptic Transmission in the Lateral Amygdala

Brain-derived neurotrophic factor (BDNF) has previously been shown to play an important role in glutamatergic synaptic plasticity in the amygdala, correlating with cued fear learning. While glutamatergic neurotransmission is facilitated by BDNF signaling in the amygdala, its mechanism of action at inhibitory synapses in this nucleus is far less understood. We therefore analyzed the impact of chronic BDNF depletion on GABAA-mediated synaptic transmission in BDNF heterozygous knockout mice (BDNF+/−). Analysis of miniature and evoked inhibitory postsynaptic currents (IPSCs) in the lateral amygdala (LA) revealed neither pre- nor postsynaptic differences in BDNF+/− mice compared to wild-type littermates. In addition, long-term potentiation (LTP) of IPSCs was similar in both genotypes. In contrast, facilitation of spontaneous IPSCs (sIPSCs) by norepinephrine (NE) was significantly reduced in BDNF+/− mice. These results argue against a generally impaired efficacy and plasticity at GABAergic synapses due to a chronic BDNF deficit. Importantly, the increase in GABAergic tone mediated by NE is reduced in BDNF+/− mice. As release of NE is elevated during aversive behavioral states in the amygdala, effects of a chronic BDNF deficit on GABAergic inhibition may become evident in response to states of high arousal, leading to amygdala hyper-excitability and impaired amygdala function.


Introduction
Brain-derived neurotrophic factor (BDNF) signaling via its cognate TrkB (tropomyosin-related kinase B) receptor regulates differentiation and survival during neuronal maturation. Furthermore, BDNF plays a pivotal role in synaptic strength and plasticity, and is thereby essential for learning and memory [1][2][3][4][5]. While glutamatergic transmission is, in many brain areas, facilitated by acute or chronic actions of BDNF via TrkB receptors, the role of BDNF signaling at inhibitory synapses seems to be more diverse [1,6,7]. Acute application of BDNF often leads to reduced GABAergic neurotransmission [8][9][10][11], but opposite effects [12][13][14] and biphasic temporal modulation [6,15] have also been described. Chronic BDNF depletion resulted in increased inhibitory synaptic activity in the dentate gyrus and superior colliculus [16,17], while GABAergic inhibitory function was strongly impaired in the visual cortex and in thalamic circuits [18,19].
In the amygdala, inhibitory mechanisms are now recognized to contribute significantly to fear learning and extinction [20]. GABAergic synaptic transmission seems to be involved in rhythmic activity, supporting the interaction of the amygdala with other brain structures during the retrieval of fear memory [21]. Reduction of GABAergic tone may facilitate fear generalization, as well as the development of anxiety disorders [22,23]. In line with this notion, amygdala hyper-excitability is a common phenomenon in disorders like epilepsy, anxiety, and stress-related diseases [24]. In addition, inhibitory control of the amygdala is regulated by different neuromodulatory systems. In the basal amygdala (BA), GABA release is facilitated via presynaptic α1A adrenergic receptors. This mechanism, regulating neuronal excitability in the BA, is severely impaired by stress and attenuated after fear conditioning [25,26].
A critical role for BDNF signaling in amygdala-dependent fear learning, as well as glutamatergic synaptic plasticity, was substantiated by several recent studies (for review and references, see [27,28]). Indeed, long-term potentiation (LTP) at glutamatergic thalamic afferents to the lateral amygdala (LA) was prevented by acute inhibition of BDNF/TrkB signaling [29]. Moreover, learning-induced long-term changes at cortico-LA synapses were absent in heterozygous BDNF +/− mice in parallel with a deficit in fear memory consolidation [30]. As reported previously, BDNF ELISA analysis in the basolateral amygdala in 4-to 5-week-old BDNF +/− mice showed a reduction of BDNF protein levels to around 50% compared to wild-type (WT) littermates [29]. However, information about the influence of BDNF signaling on GABAergic mechanisms in the amygdala is scarce. In amygdala neuronal cell cultures, acute BDNF treatment resulted in rapid internalization of surface GABA A Rα1. This TrkB-dependent internalization of GABA A Rs was hypothesized to partially underlie a transient period of enhanced amygdala activation during fear memory consolidation [31,32]. However, the effect of chronically decreased BDNF levels on GABAergic inhibitory circuits in the amygdala remains elusive. Therefore, in this study, we used patch clamp electrophysiological recordings in an in vitro slice preparation of heterozygous BDNF +/− mice and their wild-type littermates to analyze GABAergic synaptic inputs to LA projection neurons. Our results indicate that chronic BDNF reduction in the amygdala of BDNF +/− mice to about 50% of WT levels [29,33] did neither impair basal synaptic GABAergic transmission nor synaptic plasticity of inhibitory GABAergic inputs (iLTP). However, positive modulation of GABAergic synaptic transmission by norepinephrine (NE) was significantly reduced in BDNF +/− mice, which may lead to reduced GABAergic tone and hyper-excitability in the amygdala during emotionally significant events.
Presynaptic efficacy was further examined by trains of stimulation at a frequency of 40 Hz for 1 s. As expected, we observed synaptic fatigue under these conditions, as calculated for the mean ratios of current amplitudes of the 4th, 10th, 20th, 30th, and 40th IPSC in relation to the first IPSC peak amplitude (factor pulse number in the train: F 4,90 = 13.76, p < 0.0001, Figure 4A,B). Even though the ANOVA revealed a general effect of the factor genotype (F 1,90 = 8.70, p = 0.004), the synaptic fatigue was not influenced by the genotype of the animals (interaction of factors genotype × pulse number in the train: F 4,90 = 0.21, p = 0.93, Figure 4A,B).
In addition, iLTP at GABAergic synapses was independent of NMDA receptor activation, as similar potentiation could be observed when DNQX and the NMDA receptor antagonist AP5 were co-applied (122.4 ± 5.3%, n = 9, p = 0.32).
These results indicate that plasticity of GABAergic IPSCs under our recording conditions may be expressed at the postsynaptic site as suggested by Bauer and LeDoux [34], and that BDNF +/− mice do not show altered plasticity at inhibitory synapses in LA projection neurons.

LTP at GABAergic Synapses (iLTP)
Experiments were conducted in the presence of DNQX to block non-NMDA glutamatergic synaptic transmission. Pairing of afferent stimulation (100 Hz, 1 s) with postsynaptic depolarization to − 10 mV induced inhibitory long-term potentiation (iLTP) of IPSCs in LA projection neurons. We observed iLTP as an increase in average IPSC amplitude at 30 min after the induction protocol compared to baseline levels in slices from wild-type mice as well as BDNF +/− mice (WT: 128.6 ± 4.8%,
These results indicate that plasticity of GABAergic IPSCs under our recording conditions may be expressed at the postsynaptic site as suggested by Bauer and LeDoux [34], and that BDNF +/− mice do not show altered plasticity at inhibitory synapses in LA projection neurons.

Modulation of sIPSCs by Norepinephrine (NE)
Norepinephrine (NE) is known to facilitate sIPSCs in the basal amygdala of rats and mice [35][36][37][38] In our experiments in the lateral amygdala, application of 10 µM NE also strongly increased the frequency of sIPSCs in BDNF +/− and WT littermate neurons, while sIPSC amplitudes were only affected in WT mice ( Figure 6). Current traces under control conditions and during maximal NE effect ( Figure 6A,B) illustrate sIPSCs in the same representative WT ( Figure 6A) or BDNF +/− ( Figure 6B) projection neuron, respectively. Mean results for both genotypes are depicted in Figure 6C
In the presence of the α1-adrenergic antagonist prazosin, neither amplitudes nor frequencies of sIPSCs recorded in LA projection neurons were altered by NE (mean sIPSC amplitude, prazosin 16.2 ± 1.3 pA; prazosin/NE 15.3 ± 1.9 pA n = 7, p = 0.44; mean sIPSC frequency, prazosin 11.7 ± 1.8 Hz; prazosin/NE 11.3 ± 1.9 Hz n = 7, p = 0.47, Figure 7A,B). Moreover, mIPSCs that were recorded in LA projection neurons in the presence of TTX were also not modified in the presence of NE (mean mIPSC amplitude, control 15.5 ± 0.7 pA; NE 14.9 ± 0.9 pA, n = 8, p = 0.12; mean mIPSC frequency, control 7.6 ± 1.3 Hz; NE 7.6 ± 1.2 Hz n = 8, p = 1, Figure 7C,D). These data indicate that in the LA, NE increases sIPSCs via activation of α1-adrenergic receptors, similar to results for the BA as previously reported [38]. Since addition of TTX abolished the NE effect, excitation of interneurons and subsequent spike propagation seems to be required for the observed NE modulation of sIPSCs. The reduced facilitation of sIPSCs by NE in BDNF +/− mice might lead to increased excitability of LA projection neurons. charge transfer increased significantly upon addition of NE (factor NE treatment: F1,46 = 46.4, p < 0.0001). This effect was significantly enhanced in WT compared to BDNF +/− mice (factor genotype: F1,46 = 11.48, p = 0.002; interaction of factors genotype × NE treatment: F1,46 = 14.03, p = 0.001).
In the presence of the α1-adrenergic antagonist prazosin, neither amplitudes nor frequencies of sIPSCs recorded in LA projection neurons were altered by NE (mean sIPSC amplitude, prazosin 16.2 ± 1.3 pA; prazosin/NE 15.3 ± 1.9 pA n = 7, p = 0.44; mean sIPSC frequency, prazosin 11.7 ± 1.8 Hz; prazosin/NE 11.3 ± 1.9 Hz n = 7, p = 0.47, Figure 7A,B). Moreover, mIPSCs that were recorded in LA projection neurons in the presence of TTX were also not modified in the presence of NE (mean mIPSC amplitude, control 15.5 ± 0.7 pA; NE 14.9 ± 0.9 pA, n = 8, p = 0.12; mean mIPSC frequency, control 7.6 ± 1.3 Hz; NE 7.6 ± 1.2 Hz n = 8, p = 1, Figure 7C, D). These data indicate that in the LA, NE increases sIPSCs via activation of α1-adrenergic receptors, similar to results for the BA as previously reported [38]. Since addition of TTX abolished the NE effect, excitation of interneurons and subsequent spike propagation seems to be required for the observed NE modulation of sIPSCs. The reduced facilitation of sIPSCs by NE in BDNF +/− mice might lead to increased excitability of LA projection neurons.

Discussion
In the present study, we focused on GABAergic synapses on LA projection neurons. Therefore, glutamatergic inputs were blocked with DNQX and AP5, and the stimulation electrode was placed within the LA to directly activate local interneurons. Chronic BDNF reduction in the LA of BDNF +/− mice to about 50% of WT values [29] did neither impair basal synaptic GABAergic transmission nor inhibitory synaptic plasticity. However, positive modulation of interneuron activity by noradrenaline was significantly reduced by BDNF haplo-insufficiency, suggesting a role of BDNF signaling in regulating neuromodulatory transmitter effects in inhibitory synaptic circuits of the LA.

Basal GABAergic Transmission in BDNF +/− Mice
Chronic BDNF reduction in BDNF +/− mice has been shown previously to yield opposite effects on GABAergic synaptic transmission in different brain areas. In the dentate gyrus of the hippocampus, BDNF +/− mice showed increased inhibitory synaptic activity accompanied by decreased excitability of granule cells. As the frequency of mIPSCs and paired-pulse depression of eIPSCs were both enhanced, the increased inhibition most likely resulted from enhanced presynaptic release probability of GABA [16]. Moreover, a rise of sIPSC frequencies recorded in dentate gyrus granule cells of BDNF +/− mice was suggested to be due to an increase in interneuron firing rates [39]. Furthermore, enhancement of GABAergic inhibition, as seen in the superior colliculus of homozygous BDNF knockout mice (BDNF −/− mice), was suggested to result from postsynaptic increase in GABA A receptor expression [17].
In contrast, GABAergic inhibitory function was strongly impaired by chronically reduced levels of BDNF in the visual cortex [18]. BDNF +/− mice showed decreased frequency and amplitude of mIPSCs, as well as diminished paired-pulse depression. In line with altered presynaptic GABA release in BDNF +/− mice, release probability, steady-state release, and synchronous release of GABA were decreased [18]. Consistently, frequency and amplitude of mIPSCs were also reduced in the thalamic ventrobasal nucleus of BDNF +/− mice, while the decay time constant was prolonged. Evoked IPSCs showed no alteration in paired-pulse depression or synaptic fatigue compared to WT littermates. This reduced GABAergic transmission under conditions of chronic BDNF deficit suggests reduced presynaptic function and/or reduced number of functional GABAergic synaptic boutons [19]. Likewise, disruption of activity-dependent BDNF transcription was reported to impair inhibitory synaptic transmission also in prefrontal cortex pyramidal neurons [40].
In the present study, we observed neither increased nor decreased efficacy of GABAergic synaptic inputs on LA projection neurons in BDNF +/− mice. These results resemble glutamatergic synaptic transmission in the amygdala, where neither intrinsic membrane properties of LA projection neurons, nor presynaptic glutamate release, nor postsynaptic membrane properties were affected in BDNF +/− mice [29]. Different brain areas may therefore show differential susceptibility to BDNF depletion for the development and/or preservation of proper GABAergic synaptic function.

LTP at GABAergic Synapses in the LA of BDNF +/− Mice
In the present study, we focused on iLTP of GABAergic synapses on projection neurons. All glutamatergic inputs in the LA were inhibited by addition of DNQX and AP5 and local interneurons were directly activated by focal electrical stimulation. This experimental paradigm was chosen to directly investigate synaptic plasticity at GABAergic synapses in the absence of interfering effects from glutamatergic network activity [34,[41][42][43][44]. We observed stable LTP induced by pairing postsynaptic depolarization with high-frequency presynaptic stimulation (HFS) at 100 Hz. Evoked IPSCs showed no change in PPR 30 min after LTP induction, resembling results in the rat LA [34].
In amygdala cultures kept 2-3 weeks in vitro, application of exogenous BDNF led to a reduction in surface GABA A Rα1 within 5 min. This effect was TrkB receptor and PKC dependent [32]. We observed previously that LTP at the glutamatergic thalamic input to LA projection neurons is supported by acute endogenous BDNF signaling, in which BDNF may arise from thalamic or intra-amygdala sources [29].
It seems reasonable to assume that the same stimulation paradigm applied directly in the LA as in the present study may elicit BDNF release from the same afferents/PN somata. Therefore, we hypothesized that BDNF +/− mice might show a substantial reduction in acutely-released BDNF, less internalization of GABA A receptors, and hence larger GABA LTP. Nevertheless, no differences were found between WT and BDNF +/− mice. Therefore, a chronic reduction of endogenous BDNF to around 50% does not seem to compromise GABA LTP in LA projection neurons.

Norepinephrine
In our experiments, application of NE strongly increased the frequency and amplitude of sIPSCs in WT projection neurons via activation of α1-adrenergic receptors ( Figure 6). Interestingly, NE application reduced the frequency of sIPSCs recorded in rat LA neurons [45]. This opposing finding to our result in the murine LA may be related to the different species studied. Alternatively, it is feasible that α1-adrenoceptors are particularly involved in the facilitation of inhibitory synaptic transmission in the LA. In line with this notion, α1-adrenergic receptor activity was found to enhance feed-forward inhibition and constrain plasticity related to fear conditioning in the rat LA [46]. Similar to our results reported here, NE was shown previously to facilitate inhibitory synaptic transmission in rat and mouse projection neurons in the BA [35][36][37][38]. Interestingly, Kaneko and coworkers identified a specific subpopulation of GABAergic neurons in the BA of mice which were selectively excited by NE. This effect was mediated via α1-adrenoreceptors and caused increased spontaneous IPSCs in projection neurons [36]. In line with a similar mechanism of NE action in the LA, application of NE did not change sIPSCs in the presence of a α1-adrenoreceptors antagonist (Figure 7). In addition, NE effects were absent after addition of TTX. We conclude that similar to the BA, excitation of LA interneurons by NE leads to increased spiking, thereby causing enhanced frequencies and amplitudes of spontaneous IPSCs in postsynaptic projection neurons.
Importantly, modulation of sIPSCs by NE was strongly reduced in BDNF +/− mice compared to WT littermates. While frequency of sIPSCs was augmented by NE application, amplitudes were enhanced in WT mice only. This variation between genotypes caused a substantial reduction in charge transfer upon NE addition in BDNF +/− mice. Most probably, a specific NE-responsive interneuron subpopulation in the LA is altered in BDNF +/− mice with respect to excitability and/or quantal size or content. In addition, a decrease in the expression of α1-adrenergic receptors [47] could account for our observations. Indeed, BDNF deficiency seems to especially affect parvalbumin and somatostatin/NPY-expressing interneurons in the cortex [40,[48][49][50][51], and somatostatin/NPY-positive interneurons in the amygdala of female mice [52]. Interestingly, different subtypes of LA interneurons were reported to impose unitary IPSCs with unique features onto projection neurons, with stutter-firing interneurons showing the largest amplitudes of unitary IPSCs [53]. In mice, LA interneurons with stutter-firing properties were described to express somatostatin [54] and may represent a GABA interneuron subtype vulnerable to low BDNF function [52]. A detailed characterization of the NE-sensitive interneuron subtype of the LA and its regulation by BDNF remains an important topic of future studies.

Functional Implications
The amygdala receives dense noradrenergic afferents, primarily originating from the locus coeruleus (LC, [55]), which mostly form non-junctional appositions in the LA [56] that typically give rise to volume transmission. During aversive stimuli such as foot shocks, NE release in the amygdala is strongly enhanced [57]. Indeed, release of NE in the amygdala seems to be essential for encoding and retention of memories for emotionally significant events (for recent review, see [58]). The diminished facilitation of sIPSCs by NE in BDNF +/− mice may result in enhanced activation of projection neurons in the amygdala during states of high arousal, and may thereby impair amygdala function [24]. Interestingly, noradrenergic facilitation of GABAergic inhibition was disrupted by chronic stress [35]. Diminished GABAergic tone upon NE action in the amygdala of BDNF +/− mice might therefore enhance stress susceptibility. Importantly, major depressive disorder (MDD), as well as diverse neurodegenerative diseases, are associated with reduced levels of BDNF paralleled by diminished GABAergic neurotransmission [7,59]. The interaction of BDNF and GABA neurotransmission in the amygdala may preferentially involve control of distinct interneuron subtypes by modulatory neurotransmitters, while GABAergic synapses in the LA are not directly modified by chronic BDNF depletion.

Animals
In the present study, male C57BL/6J mice (Charles River, Sulzfeld, Germany) were analyzed. The animals were kept in groups of three to four animals per cage, had free access to food and water, and were maintained at a 12-12 h light-dark-cycle (lights on at 7:00 a.m.). All experiments were carried out in accordance with the European Committees Council Directive (2010/63/EU).

Drugs
Drugs were added to the external ACSF. NE was bath-applied for 5 min and reached its maximal effect around 3 min after application. Slices were pre-incubated with prazosin for 15 to 20 min. All substances were obtained from Sigma-Aldrich (Diesenhofen, Germany), except for DNQX, AP5 and CGP55845 hydrochloride (Tocris, Bristol, UK).

Data Analysis
Data were analyzed with Origin 8.0 (OriginLab Corporation, Northampton, MA, USA). Miniature postsynaptic currents were detected using the program Mini-Analysis (Jaejin software, Leonia, NJ, USA). The rise time of IPSCs was calculated between 10 and 90% of the peak amplitude onset, and the time course of decay was fitted to a mono-exponential function. Cumulative histograms without bins were calculated within time periods of 3 min duration containing exactly 300 events. Release probability (Pr) and synaptic release sites activated by the stimulus (Nsyn) were calculated as described in detail by [18]. In short, the readily releasable pool (RRP) was quantified by back-extrapolating the linear phase of the cumulative amplitude plot of eIPSCs to the y-axis. Pr was calculated as first eIPSCs amplitude divided by the RRP, and Nsyn was given as RRP divided by quantal size. The latter was estimated as the median amplitude of spontaneous IPSCs immediately following 40 Hz stimulation [63].
Statistical analysis was performed using Kolmogorov-Smirnoff (Mini-Analysis) and nonparametric tests by Graph Pad Prism software (San Diego, CA, USA; Wilcoxon signed-rank test for paired observations, Mann-Whitney test for non-paired observations) or by JMP (SAS Institute Inc., Cary, NC, USA, Version 8; analysis of variance (ANOVA) tests, followed by post-hoc Tukey comparisons).

Conclusions
GABAergic synapses in the LA are neither impaired in pre-nor postsynaptic properties, nor in synaptic plasticity in 4-to 5-week-old BDNF +/− mice, which show around 50% reduced BDNF protein levels in the amygdala. However, facilitation of GABAergic synaptic transmission by NE was significantly decreased in BDNF +/− mice. These findings suggest that BDNF regulates neuromodulation of inhibitory synaptic circuits in the LA, which may become evident during states of high arousal. Chronic BDNF depletion might therefore lead to amygdala dysfunction due to diminished GABAergic tone during emotionally significant events.