Long-term changes in the CA3 associative network of fear-conditioned mice.

Abstract The CA3 associative network plays a critical role in the generation of network activity patterns related to emotional state and fear memory. We investigated long-term changes in the corticosterone (CORT)-sensitive function of this network following fear conditioning and fear memory reactivation. In acute slice preparations from mice trained in either condition, the ratio of orthodromic population spike (PS) to antidromic PS was reduced compared to unconditioned animals, indicating a decrease in efficacy of neuronal coupling within the associative CA3 network. However, spontaneous sharp wave–ripples (SW-R), which are thought to arise from this network, remained unaltered. Following CORT application, we observed an increase in orthodromic PS and a normalization to control levels of their ratio to antidromic PS, while SW-R increased in slices of fear conditioned and fear reactivated mice, but not in slices of unconditioned controls. Together with our previous observations of altered hippocampal gamma activity under these learning paradigms, these data suggest that fear conditioning and fear reactivation lastingly alters the CORT-sensitive configuration of different network activity patterns generated by the CA3 associational network. Observed changes in the mRNA expression of receptors for glutamate, GABA and cannabinoids in the stratum pyramidale of area CA3 may provide a molecular mechanism for these adaptive changes.


Introduction
The ventral hippocampal CA3 region, in close interaction with the hypothalamic-pituitary-adrenal (HPA) stress axis (Jacobson & Sapolsky, 1991;Maggio & Segal, 2007), plays a key role in anxiety modulation and formation of fear memory (Bannerman et al., 2004;Trivedi & Coover, 2004). Chronic changes in circulating levels of corticosterone (CORT) or deficits in the HPA stress axis response are characteristic for affective and anxiety disorders (Ströhle & Holsboer, 2003). We have previously shown that a single fear reactivation session is sufficient to induce long-term changes in circulating CORT levels as well as in gamma frequency oscillations in area CA3 of the ventral hippocampus (Albrecht et al., 2013). Using slice preparations with and without CORT supplementation, it is possible to address the functional network changes in this region that may mediate the behavioral consequences of conditioned fear stress.
The ventral hippocampal CA3 region is characterized by an associative network built of recurrent axon collaterals that target other pyramidal cells and interneurons in which sequential information can be stored (Le Duigou et al., 2014). In addition to gamma oscillations, this collateral associative network generates sharp wave-ripples (SW-R), positive field potential (FP) transients, which occur spontaneously in ventral hippocampal slices of mice (Maier et al., 2002), and are superimposed by ripples with a frequency of $ 200 Hz (Kranig et al., 2013;Maier et al., 2003Maier et al., , 2009). Gamma oscillatory activity and SW-R are believed to participate in memory formation, but while gamma oscillations occur during more active behaviors, e.g. exploration of a context (Hájos & Paulsen, 2009;Montgomery & Buzsáki, 2007), SW-R occur during quiescent behavior. SW-R are relevant for memory consolidation by replay during sleep as well as for decision-making in a new situation in relation to previous memories in awake animals (Girardeau et al., 2014;Jadhav et al., 2012). Therefore, we determined the effects of fear conditioning and fear reactivation on the strength of collateral associative network activity in area CA3 and its responsiveness to CORT. To address molecular factors potentially involved in the altered network function, we examined the mRNA expression of several receptors (NMDA-R, AMPA-R, GABA A -R, cannabinoid receptor 1 [CB1-R]) that have previously been related to generation and maintenance of SW-R in the ventral hippocampal area CA3. Our data show a differential longterm modification of the CA3 collateral associative network and its sensitivity to CORT following fear learning and fear reactivation.

Materials and methods
Seven-week-old male C57B/6BomTac mice were obtained from M&B Taconic (Berlin, Germany) and acclimatized to our animal facility for one week (12 h light/dark cycle with lights on at 19:00 h and 30 min dawn phases; food and water were provided ad libitum). After another week of single housing, animals were randomly assigned to the different experimental groups. All experiments were conducted in accordance with the European and German regulations for animal experiments and were approved by the Landesverwaltungsamt Sachsonia-Anhalt (AZ 2-618).

Fear conditioning and fear reactivation
Data were obtained from fear conditioned, fear reactivated and control animals described previously (Albrecht et al., 2013). Different sets of animals underwent auditory-cued fear conditioning and its reactivation and were either used for examining oscillatory activity of the ventral hippocampus in slice preparations or gene expression changes in ventral CA3 sublayers via laser capture microdissection and real-time PCR. In brief, all animals were adapted in four sessions (twice a day) to the fear conditioning chamber, a 16 cm Â 32 cm Â 20 cm acrylic glass arena with a grid floor, loudspeaker and ventilation fan in a sound isolation cubicle (background noise 70 dB SPL, light intensity510 lux; TSE, Bad Homburg, Germany). Each adaptation session consisted of 2-min exposure to the conditioning context, followed by six exposures to a neutral tone (CS-: 2.5 kHz for 10 s, 80 dB; 20 s inter stimulus intervals (ISI)). One day later, after 2 min of context exposure, mice were trained to associate a tone (CS+: 10 kHz for 10 s, 80 dB) immediately followed by a footshock (US: 0.4 mA for 1 s) in three pairings, separated by 20 s ISI. Reactivation of fear memory took place 24 h later by re-exposure to the training context (2 min) and to four CS-(10 s each, 20 s ISI) and four CS + (10 s each, 20 s ISI) delivered in the training context. While the fear reactivation group received the full protocol (group R), the reactivation session 24 h posttraining was omitted in the non-reactivation group (NR). An additional control group (CTL) received only tones (3 Â 10 kHz, 10 s, 80 dB) but no foot shocks during the training session. During sessions, the animal's behavior was assessed online via a photo beam detection system. All mice were left undisturbed except for animal care and then either slice electrophysiology or gene expression was studied in ex vivo preparations four weeks after fear conditioning.
Electrophysiological recordings of SW-R activity and population spikes in the ventral CA3 Animals of groups reactivation (R, N ¼ 7), NR (N ¼ 7) and control (CTL, N ¼ 6) underwent the respective fear conditioning and reactivation protocols as described above and were decapitated 30 d later under deep isoflurane anesthesia. Horizontal ventral hippocampal slices (400 mm) were cut at an angle of about 12 in the fronto-occipital direction. The preparation of slices was done in ice-cold, carbogenated (5% CO 2 /95% O 2 ) artificial cerebrospinal fluid (aCSF) containing (in mM) 129 NaCl, 21 NaHCO 3 , 3 KCl, 1.6 CaCl 2 , 1.8 MgCl, 1.25 NaH 2 PO 4 and 10 glucose. Slices were transferred to an interface chamber perfused with aCSF at 34 ± 0.1 C (flow rate: 1.8 ± 0.2 ml/min, pH 7.4, osmolarity $ 300 mosmol/kg). Slices were incubated at least for an hour before starting recordings.
Extracellular field recordings were obtained from stratum pyramidale (SP) of area CA3. FP responses were evoked by constant voltage stimulation of stratum radiatum (SR) in area CA1 using bipolar platinum wire electrodes with exposed tips of 50-80 mm and tip separations of 100-200 mm (electrode resistance in aCSF: $10 KOhm). Drugs were applied via continuous bath perfusion: CORT (1 mM, Sigma-Aldrich, Steinheim, Germany) was diluted freshly prior to the experiment using a stock solution with dimethyl sulfoxide (DMSO; Merck KGaA, Darmstadt, Germany). Microelectrodes were filled with aCSF (resistance: 5-10 M). Signals were pre-amplified using a custom-made amplifier and low-pass filtered at 3 kHz. Signals were sampled at a frequency of 5 kHz and stored on a computer hard disc for off-line analysis.
Statistical data were reported as mean ± standard error of the mean. Before statistical comparison of different groups, normality test (Shapiro-Wilk test) and equal variance test were performed. Group differences were determined by oneway ANOVA. Post hoc Fisher's LSD or Dunn's method was used for pair-wise comparison. To statistically compare the CORT effect, Student's t test or Mann-Whitney U test was used (SigmaPlot for Windows Version 11.0, 2008, Systat Software, Erkrath, Germany).

Analysis of population spikes in the ventral CA3
Orthodromic population spike (PS) generation is due to activation of recurrent excitatory interactions between CA3 pyramidal cells (Behrens et al., 2005;Fano et al., 2012). Schaffer collateral (SC) stimulation induces an antidromic PS in the CA3 network, which leads to generation of a secondary orthodromic PS (Behrens et al., 2005;Fano et al., 2012). To obtain a stimulus response curve of CA3 PS, constant voltage stimuli ranging from 2 V to 10 V were delivered to SR in area CA1. SR stimulation results in activation of SCs, which induces first an antidromically evoked PS in area CA3, strength of which indicates the number of activated fibers. This is followed by a field EPSP superimposed by a secondary PS (orthodromic) due to activation of recurrent axon collaterals between CA3 pyramidal cells and between CA3 pyramidal cells and interneurons ( Figure 1A). The strength of PS was measured by calculating the area (mV.ms) of the PS. To measure the strength of collateral associative network activation in area CA3, the orthodromic PS was normalized to the antidromic PS for all stimulation intensities. Stimulus intervals of 20 ms, 30 ms, 40 ms, 50 ms and 100 ms were used to calculate paired-pulse ratios by dividing the second orthodromic PS area to the first orthodromic PS area. Paired-pulse ratios were measured using the responses obtained from stimulations with 6-8 V intensities. Data analysis was performed off-line using Spike2, version 6, software (Cambridge Electronic Design, Cambridge, UK).

Analysis of SW-R in the ventral CA3
After one hour of incubation, SW-R appeared in SP of area CA3 of ventral horizontal hippocampal slices (Supplementary Figure 1). For analysis of SW-R, 2-min data files were extracted to be further analyzed using a MATLAB-based code (MathWorks, Natick, MA). Sharp waves (SWs) were detected by low-pass filtering the data at 45 Hz (Butterworth, 8th order). The threshold for event detection was set to 2.5 times the standard deviation (SD) of the lowpass-filtered signal. Minimum interval between two subsequent SW was set to 80 ms. Data stretches of 125 ms centered to the maximum of SW event were stored for further analysis. To analyze the area under the curve of SW, the points crossing the mean of the data were used as the start and the end point of SW. The SW area was measured using trapezoidal numerical integration of low pass-filtered data.
To isolate the ripples, the raw data was band-pass filtered at 120-300 Hz (Butterworth, 8th order). Data stretches of 15 ms before and 10 ms after the maximum of SW event (25 ms) were stored for further analysis. Threshold for ripple detection was set to three times SD of the band-pass filtered signal. To analyze the ripple amplitude, triple-point-minimaxdetermination was used. If the difference between falling and rising component of a ripple was higher than 75%, ripples were discarded from analysis. Frequencies were calculated only from subsequent ripples. Furthermore, to decompose a time-series into time-frequency space, a wavelet analysis using the Morlet wavelet transform was performed (Erchova et al., 2004;Farge, 1992). Figure 1. Collateral excitatory interaction in area CA3 of ventral hippocampal slices is reduced in fear-exposed mice. (A1) A sketch depicting the positioning of the extracellular field electrode at stratum pyramidale (SP) of CA3 and the bipolar stimulation electrode (two black dots) at the stratum radiatum (SR) of area CA1-CA2. (A2) Field response of CA3 pyramidal layer to an antidromic Schaffer collateral (SC) stimulation with an antidromic population spike (PS) followed by an orthodromic PS. (B) Stimulus-response curves of (B1) antidromic PS and (B2) orthodromic PS comparing control, NR (no reactivation) and R (reactivated; n ¼ 5-7 slices per group). Note that the orthodromic PS tends to decrease in group NR, while the antidromic PS is unaltered. (C) Stimulus-response curves of (C1) antidromic PS and (C2) orthodromic PS with or without corticosterone (CORT) in control slices. Note that both of the PS are not altered by CORT. (D) Stimulus-response curves of (D1) antidromic PS and (D2) orthodromic PS with or without CORT in NR slices. Note that CORT augmented the orthodromic responses in group NR. (E) Stimulus-response curves of (E1) antidromic PS and (E2) orthodromic PS with or without CORT in R slices. Note that CORT augmented the orthodromic PS without any change in antidromic response in group R. (F) Summary graph illustrating decrease in normalized orthodromic responses in area CA3 of ventral hippocampal slices of both fearconditioned and fear-reactivated mice. Note that CORT reversed this effect. *indicates significant difference between groups with p50.05, **p50.01 (one-way ANOVA). #indicates significant CORT effect with p50.05, ##p50.01 ###p50.001 (Student's t test or Mann-Whitney U-test).

mRNA expression in sublayers of ventral hippocampal CA3 region
Mice from groups CTL, NR and R (N ¼ 6 each) were killed by cervical dislocation 30 d after fear conditioning, at a day time with low internal CORT plasma levels (2.00-3.00 pm at an inverse light-dark cycle; Albrecht et al., 2013). Brains were quickly removed, embedded in Tissue Tek freezing compound and snap-frozen in methylbutane cooled by liquid nitrogen. For laser capture microdissection, 8-10 horizontal cryosections per animal (20 mm thick) at the level of the ventral hippocampus were mounted on 0.05% poly-L-lysinecoated RNase-free PEN membrane slides (Carl Zeiss, Jena, Germany), fixated in -20 C cold 70% ethanol and briefly stained with 1% cresyl violet acetate (Sigma-Aldrich) under nuclease-minimized conditions. Using a laser capture microdissection system (Carl Zeiss), the SR, SP and stratum oriens (SO) of the ventral CA3 were microdissected and collected in an adhesive cap capture device (Carl Zeiss). Sample lysis, removal of genomic DNA and isolation of total RNA was done with the RNeasy Micro Plus kit (Qiagen, Hilden, Germany) according to manufacturer's instructions. Firststrand synthesis of cDNA was performed with the Sensiscript Reverse Transcription kit (Qiagen) for low amounts of RNA, in the presence of 2.5 mM dNTPs, Oligo (dT)18/random decamer first-strand primer mix (50 mM each; Life Technologies, Darmstadt, Germany) and RNase Inhibitor (SuperaseIN; 20 U/ml; Life Technologies) for 60 min at 37 C. Determination of relative gene expression levels of different glutamate receptor subunits was done in a 1:5 dilution of cDNA via quantitative multiplex PCR (ABI Prism Step One real-time PCR; Life Technologies). TaqMan Õ reagents and predesigned assays were used with different fluorescent dyes labeling glutamatergic and GABAergic targets (TaqMan gene expression assays, Life Technologies; assay IDs: GluA1 (Gria1): Mm00433753_m1; GluN1 (Grin1): Mm00433790_m1; GluN2A (Grin2a): Mm00433802_m1; GluN2B (Grin2b): Mm00433820_m1; GABA A a2 (Gabra2): Mm00433435_m1; GABA A a3 (Gabra3): Mm01294271_m1; CB1 (Cnr1): Mm01212171_s1) or the housekeeping gene glycerinaldehyd-3-phosphat-dehydrogenase (GAPDH; endogenous control, Life Technologies). All samples were run in triplicates in 50 cycles of 15 s at 95 C and 1 min at 60 C, preceded by a 2-min decontamination step at 50 C with uracil-N-glycosidase and initial denaturation at 95 C for 10 min.
For data analysis, the cycle thresholds (CTs) in each triplicate assay and each target gene was determined and relative quantification (RQ) was conducted with the ddCT method (Livak & Schmittgen, 2001) by normalizing each sample to the overall content of cDNA using GAPDH as an internal control (dCT; dCT ¼ dCT (target gene) ¼ (CT (target gene)) -(CT (GAPDH)). All ddCT values were normalized to the mean of control group for each target gene and area with ddCT ¼ dCT (sample) -mean dCT (control group). Transformation to RQ values was done according to RQ ¼ 2-ddCT Â 100 (%) with RQ (control) ¼ 100%. Statistical analysis was performed based on RQ% values by one-way ANOVA for the factor training group followed by LSD tests for post hoc comparison in each layer and for each target gene.
Finally, we analyzed the paired-pulse ratio of orthodromic PS (n ¼ 5-7 slices for each group) for intervals ranging from 20 ms to 100 ms (Figure 2). Below 30 ms, second orthodromic PS values tended to be lower than those for the first orthodromic PS values without any significant change between the groups (stimulation interval: 20 ms; one-way ANOVA on ranks: H(2) ¼ 1.830, p ¼ 0.401) indicating that fast inhibition was unaffected. Similarly, CORT did not change the paired-pulse facilitation in area CA3 of ventral hippocampal slices in all groups for all intervals (Student's ttest or Mann-Whitney U test: p40.05).
Combined, these data suggest that fear conditioning and its reactivation decrease the strength of collateral coupling within the associative network in CA3 in ventral hippocampal slices and that this can be normalized by CORT.

mRNA expression of glutamatergic, GABAergic and cannabinoid receptor subunits
Assessment of mRNA levels with quantitative real-time PCR revealed a long-lasting reduction of various receptor subunits in the SP of the ventral CA3 region after fear conditioning, but not after its reactivation (Figure 4; for expression in SO and SR, see Figure S2-3). Specifically, the GluA1 subunit of the AMPA receptor was reduced in NR (F(2,15) ¼ 4.239, p ¼ 0.035; LSD post hoc: p ¼ 0.011, NR versus CTL). While the expression of the GluN2A subunit of the NMDA receptor showed no training effect (F(2,15) ¼ 0.652, p ¼ 0.535), significant long-term effects of fear conditioning were observed on the expression of the GluN1 (F(2,15) ¼ 7.370, p ¼ 0.006) and GluN2b subunit (F(2,15) ¼ 6.563, p ¼ 0.009). Post hoc analysis revealed a reduction of both subunits after fear conditioning alone (GluN1: p ¼ 0.005, NR versus CTL; NR2B: p ¼ 0.003, NR versus CTL) but not after reactivation of the conditioned fear (NR1: p ¼ 0.004, NR versus R; NR2B: p ¼ 0.034, NR versus R). However, the expression of selected targets related to GABAergic signaling was affected in a different way. After both, fear conditioning and its reactivation, expression levels of the GABA A receptor a2 subunit were reduced (F(2,15) ¼ 6.892, p ¼ 0.008; p ¼ 0.002, NR versus CTL; p ¼ 0.037, R versus CTL). A similar expression pattern was observed for the CB1-R (F(2,15) ¼ 7.124, p ¼ 0.007; p ¼ 0.002, NR versus CTL; p ¼ 0.023, R versus CTL), while the expression of the a3 subunit was not changed (F(2,14) ¼ 0.729, p ¼ 0.585).
Together, while fear conditioning induces a long-lasting reduced expression of both, glutamatergic and GABAergic modulators of SW-R in the SP of the ventral CA3, fear memory reactivation only restores the expression of glutamatergic receptor subunits, indicating differential experienceinduced mechanism of SW-R modulation.

Discussion
In this study, we report long-term changes in the function of the mouse CA3 collateral associative network following fear conditioning and fear memory reactivation. In acute slice preparations, network patterns with similar characteristics observed in vivo can be obtained. They thus are well suited to study the mechanisms underlying different behaviorally relevant network activity patterns and their change following behavioral stimulation (Albrecht et al., 2013;Fisahn et al., 1998;Maier et al., 2009;Lu et al., 2011). In addition to analyzing slices from differently trained animals, we simulated states of high and low CORT levels during recording. Without supplementation of exogenous CORT, we observed reduced orthodromic/antidromic PS ratio indicating reduced excitatory synaptic interaction within the CA3 network ( Figure 1) as well as decreased gamma oscillations (Albrecht et al., 2013), but normal spontaneous SW-R activity. In contrast, under high CORT levels, PS ratios and gamma activity were both comparable to those in slices from control animals, whereas SW-R were increased in fear conditioned and fear reactivated groups. Thus, fear learning appears to lastingly alter the CORT-sensitive configuration of different network activity patterns generated by the CA3 associational network. This is in line with the different functions attributed to gamma oscillations and SW-R in the behaving animal. For example, we recently showed that in naïve animals, the effects of the stress-related neuromodulators including CORT, corticotropin-releasing factor and tetrahydrodeoxycorticosterone on spontaneous SW-R are rather mild compared to their effects on gamma oscillations (Calışkan et al., 2015).
In vivo, gamma oscillations occur during exploratory behavior while the hippocampus receives a strong cholinergic input (Hironaka et al., 2001;Lee et al., 1994;Montgomery & Buzsáki, 2007). Similarly, gamma oscillations can be induced in hippocampal slice preparations by challenging the network via cholinergic or kainate receptor activation (Fano et al., 2012;Fisahn et al., 1998;Wójtowicz et al., 2009). This type of perturbation depolarizes the cells in area CA3, where the gamma network activity emerges due to extensive axon collaterals interacting with pyramidal cells and inhibitory interneurons (Hájos & Paulsen, 2009). Contrary to gamma oscillations, in vivo, SW-R mostly occur during quiescent behavior such as grooming and slow-wave sleep when the level of ACh in the hippocampus is low. In vitro, SW-R are observed in hippocampal slice preparations of mice as spontaneous events, which often originate from area CA3 (Maier et al., 2003).
In fact, different local circuit mechanisms are thought to underlie the generation of SW-R and gamma oscillations in the CA3, whereas gamma oscillations require cycle-by-cycle reciprocal interactions (PING model) between pyramidal cell and inhibitory interneurons (parvalbumin containing basket cells, etc.); SW-R are more dependent on reciprocal inhibitory interactions (FINO model) (Schlingloff et al., 2014). In our experiments, as paired-pulse behavior of orthodromic PS was affected neither by fear conditioning nor by its reactivation, the effects observed appear to be rather of postsynaptic than of presynaptic nature (Zucker & Regehr, 2002).
We found that CORT pre-application augmented SW-R activity in slices from mice after fear conditioning and its reactivation, but not in control slices. A previous study described the concentration-dependent effects of CORT on the SW-R activity in naive slices obtained from male C57Bl6 mice (Weiss et al., 2008). Weiss et al. recorded the SW-R activity in area CA1, whereas in our study, we recorded SW-R in area CA3, where they are generated (Kranig et al., 2013;Maier et al., 2002). However, similar to our study, after 1 mM CORT, relatively mild effects on the occurrence of SW-R were reported (Weiss et al., 2008). Detailed comparison of training groups reveals a somewhat attenuated impact of CORT after fear memory reactivation, as compared to simple fear conditioning, whereas in NR slices, CORT increased the incidence and the size of SW-R as well as the ripple frequency; the additional reactivation of fear memory increased only SW-R incidence but had no effect on SW-R size or ripple frequency. A differential response to prior training experience has also been observed concerning gamma oscillations and their rescue by CORT supplementation (Albrecht et al., 2013). Thus, fear conditioning and fear reactivation appear to lastingly alter the CORT sensitivity of network activity patterns generated in the CA3 associative network. The increase in SW-R occurrence in the fearexposed mice might be due to augmenting effects of CORT on excitatory neurotransmission within the 30-min CORT preexposure interval. This suggests an involvement of fast nongenomic mechanisms rather than CORT-induced changes in gene expression that would just onset in this phase (Joëls et al., 2012).
Training-specific long-term changes were also observed on the molecular level: fear conditioning without reactivation induced a long-lasting reduction of GluA1, GluN1 and GluN2B, as well as GABA A -R a2 subunit and the CB1-R in the SP of area CA3. Following additional fear reactivation, in contrast, only the latter, inhibitory factors, remained reduced in expression. This suggests an allostatic regulation in the mRNA expression of receptors known to control the activity of the CA3 collateral associative network.
Strikingly, mRNA expression changes were exclusively observed in the SP, but not SO or radiatum of the ventral CA3 region. A recent study by Hájos et al. (2013) investigated the contribution of anatomically identified CA3 neurons to SW-R and found that the majority of cells firing during SW-R are interneurons, specifically basket cells and axo-axonic cells located within the SP. Moreover, specific firing properties were observed for different types of basket cells that express either parvalbumin or CB1. Our data thus raise the exciting possibility that fear conditioning alters CA3 network activity by lastingly and differentially modulating the activity and molecular composition of particular types of interneurons. Future studies using cell-type specific analysis methods are required to investigate this hypothesis further.
It can be assumed that altered mRNA levels four weeks after the inducing experience reflect changes in their steadystate expression. Therefore, the observed differential expression changes may be directly relevant for adaptive and maladaptive cellular function in the ventral CA3 region. Indeed, the generation and maintenance of hippocampal SW-R are critically dependent on both, glutamatergic and GABAergic receptor activation (Ellender et al., 2010;Hájos et al., 2013;Papatheodoropoulos, 2007), while activation of CB1-R disrupts CA3 network activity by suppression of excitatory transmission (Holderith et al., 2011;Maier et al., 2012;Sun et al., 2012). Thus under low CORT conditions, reduced NMDA and AMPA receptor expression could prevent over-excitation of area CA3 and thus maintain SW-R within a normal range in fear conditioned animals. A single fear reactivation session, however, is sufficient to disrupt the balanced regulation of excitatory and inhibitory factors, normalizing NDMA and AMPA receptor expression to control levels while GABA A receptor alpha2 subunit and CB1R remain low.
We propose that the differential change in expression of glutamate, GABA A and cannabinoid receptors may not only underlie CORT-sensitive changes in the CA3 associative network but could also be involved in modulation CORTresponsive behavior in these animals. SW-R reflect a temporally condensed replay of learning-activated neuronal ensemble activity during sleep. Recent studies furthermore suggest a role for SW-R in decision-making based on previously acquired memory (Jadhav et al., 2012). Their increase under CORT supplementation indicates a sensitization of the CA3 network in animals with previous fear conditioning experience that may alter the behavioral response to challenging situations with increased levels of circulating CORT. In fact, we have previously shown an increased generalization to the background context of mice that had undergone memory reactivation, and thus presented with increased levels of endogenous CORT ( Figure 5; Albrecht et al., 2013).  expression is also reduced in both groups NR and R. Values are mean RQ % ± SEM (normalized to average expression in CTL group). *Indicates significant difference between groups with p50.05, **p50.01 (n ¼ 6 per group, one-way ANOVA). hippocampus modulates anxiety and memory dependent on previous stress experience (Akirav, 2011). Cannabinoids are further believed to modulate adaptation to stress, ameliorating the impact of severe and chronic stress pre-exposure on plasticity, cognitive tasks, fear memory extinction and expression of GR (Ganon-Elazar & Akirav, 2013;Segev et al., 2014). Furthermore, NMDA receptors in the hippocampus are critical for the formation of context fear memories (McHugh & Tonegawa, 2009;Stiedl et al., 2000;Zhang et al., 2001). A shift toward glutamatergic receptor expression thus also is in line with the generalization to background context in fear reactivated mice (see also Figure 5).
We observed that fear reactivation attenuates the sensitization of CA3 associative network to CORT and recovers normal glutamatergic receptor expression. After reactivation, a fear memory becomes labile and can be updated with nonfearful information (Alberini, 2011). Reorganization of CA3 network activities following fear learning and reactivation may provide a powerful mechanism to support such adaptive Figure 5. Summary of results from this and our previous (Albrecht et al., 2013) study. Thirty days after auditory cued fear conditioning alone (NR), anxiety-like behavior in the elevated plus maze was decreased. At the same time, mRNA expression levels of glucocorticoid (GR) and mineralocorticoid receptors (MR) as well as of AMPA and NMDA receptor subunits were decreased in ventral CA3 sublayers in only NR while the GABA A -R a2 subunit (GABRA2) and the cannabinoid receptor 1 (CB1) were decreased in both NR and R suggesting a specific recovery of excitatory subunit expression by fear memory reactivation. Moreover, fear conditioning reduced lastingly network activity in the ventral CA3 region, reflected by gamma oscillations and associative excitatory interactions as indicated by the ratio of orthodromic to antidromic population spike (PS). These changes were reversible by acute application of CORT in vitro. In addition, CORT augmented SW-R incidence and magnitude. Reactivation of fear memory (R) induced a differential pattern of changes. While anxiety-like behavior was reduced as well, the remote fear memory toward the background context was enhanced and CORT plasma levels were lastingly increased. The mRNA expression levels of MR, GR and glutamatergic receptors were comparable to control levels while gamma oscillations and associative excitatory interactions within area CA3 were reduced as well. However, the impact of CORT on network activity appears reduced after reactivation. changes; it will be interesting to see how these behave during other modifications of fear memory such as extinction learning.