Differential behavioral, stress, and sleep responses in mice with different delays of fear extinction

Abstract

Study Objectives

Sleep, in particular rapid eye movement (REM), has been linked to fear learning and extinction; however, their relationship is poorly understood. We determined how different delays of extinction training (ET) impact fear-conditioned behaviors, changes in sleep, and stress responses.

Methods

EEG activity, movement, and body temperature in mice were monitored via telemetry. Following contextual fear conditioning (shock training [ST]), separate groups of mice were reexposed to the context at 24-hour post-ST (24h ET-1) and at 48-hour post-ST (48h ET-1). Post-ET sleep amount and sleep-associated EEG (delta and theta) activity were compared to baseline and to post-ST sleep. Freezing, locomotion, grooming, and rearing were monitored to determine effects of ET on fear behaviors. Body temperature immediately after ET was monitored to assess stress-induced hyperthermia (SIH).

Results

24h ET-1 and 48h ET-1 produced similar freezing and REM reductions, but dissimilar rearing activity and SIH. 24h ET-1 was followed by periods of suppressed REM-associated theta (REM-θ) activity, immediately after ET and during the subsequent dark period. Suppressed REM-θ was specific to sleep after 24h ET-1, and did not occur after ST, nor after 48h ET-1.

Conclusions

ET-1 at 24 and 48 hours after ST was associated with similar freezing and REM amounts, but with differences in other overt behaviors, in REM-θ, and in SIH. Freezing was not predictive of changes in other fear-associated responses. This study demonstrated that consideration of time delay from fear acquisition to extinction is important when assessing the relationships between extinction and behavior, sleep, and stress responses.

Statement of Significance.

Fear extinction is a learning mechanism essential for coping with stressful memories that can disrupt sleep. Here, we examined fear behaviors, sleep quantity, EEG delta and theta activity, and the stress response following two delays of extinction training (ET) for contextual fear conditioning. ET at 24 and 48 hours after fear conditioning resulted in similar freezing and rapid eye movement (REM) amounts, but differences in other overt behaviors, in REM-theta and in stress-induced hyperthermia. These data demonstrate that expressions of fear memory, and of extinction, can vary with time after the initial fearful events. They demonstrate that effects of fear memory on multiple systems and passage of time need to be considered in modeling pathological fear and assessing its relationship to sleep.

Introduction

Animal studies have demonstrated that, even after consolidation, recalled fear memories can become transiently labile, thereby providing a “window of opportunity” for modifications that can result in lessened fear responses [1, 2]. These modifications underlie fear extinction, a basic mechanism important for regulating negative emotion (reviewed in Pace-Schott et al. [3]). The importance of understanding extinction is underscored by the fact that it provides critical rationale for treatments such as exposure-based psychotherapy used to treat phobias and psychiatric disorders, including post-traumatic stress disorder (PTSD) [4].

Neuroimaging studies [5, 6] have suggested that rapid eye movement (REM) sleep may be important for extinction based on observations that REM is accompanied by selective activation of neural structures that regulate extinction, e.g. the amygdala, the medial prefrontal cortex (mPFC), and the hippocampus. Recently, studies on linkages between extinction and sleep have been increasing; however, there still is minimal information on how REM interacts with extinction and the consolidation of extinction [7–9]. A complexity with examining the role of REM in fear learning and extinction is that alterations in REM can become fear conditioned in much the same way that freezing and other behavioral and physiological responses can become fear conditioned [10–14]. After conditioning, fear memories alone, evoked by cued [15] and contextually [16, 17] conditioned stimuli, produce changes in REM similar to those produced by the initial fearful event. Fear-conditioned alterations in REM are regulated by the amygdala [10–12], a region crucial for fear conditioning [18–21], and can extinguish with subsequent non-reinforced presentation of the fearful stimulus [14]. Thus, many of the processes that regulate fear-conditioned changes in REM are similar to those seen with other fear-conditioned responses. However, REM can be either decreased or increased following virtually identical fear-induced freezing and stress responses [10–13, 16], indicating that fear memory can differentially regulate outputs across functional systems. It is currently unknown whether fear-conditioned responses generated by different systems also are differentially altered by extinction.

Fear extinction has primarily been examined using freezing to define memory modifications and assess reductions in fear with the general assumption that reduced freezing is indicative of extinguished fear. There is minimal information on how extinction affects other overt fear behaviors, the stress system, and sleep. It is also unknown whether extinction of fear responses in these systems is synchronized with extinction of freezing, or on whether extinction of fear responses varies with time since the initial fearful event. Therefore, our goal in this study was to test the hypothesis that extinction is a unitary process that produces similar changes across fear response systems that are consistent across time. We examined fear extinction using multiple behavioral outputs (freezing, locomotion, grooming, and rearing), stress-induced hyperthermia (SIH) as an index of the stress response [22], and effects on non-rapid eye movement (NREM) and REM. We examined NREM-associated delta (δ) activity (NREM-δ) as a measure of potential changes in sleep intensity [23–25] and REM-associated theta (θ) activity (REM-θ), which primarily reflects θ oscillations in the hippocampus [26], and which has been linked to fear memory consolidation and extinction [27]. We also conducted extinction training (ET) at 24 and 48 hours after fear conditioning to determine whether the extinction of different fear responses varied with time since fear memory consolidation.

Methods

Participants

Fourteen adult male C57BL/6 mice were obtained from Charles River (Wilmington, MA). The mice were 8–9 weeks old and weighed 20–25 g upon arrival. They were individually housed and were kept in a colony room with food and water available ad libitum. The colony room was maintained on a 12:12 light to dark cycle and ambient temperature at 24.0°C ± 1.5°C. At least 1 week was allowed for acclimation before surgery was conducted. Throughout the experimental procedures, measures were taken to minimize unnecessary pain and discomfort of the animals. All procedures were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Experimental Animals and were approved by Eastern Virginia Medical School’s Institutional Animal Care and Use Committee.

Surgical procedures

Surgical procedures were conducted during the light period with the mice under isoflurane anesthesia as inhalant (5% induction, 2% maintenance). All animals received prophylactic potassium penicillin (25 IU/g), gentamicin (0.005 mg/g), and dexamethasone (0.0005 mg/g) subcutaneously. The mice were implanted intraperitoneally with telemetry transmitters (ETA-F10 or ETA-F20; Data Sciences International (DSI), St. Paul, MN) to allow determination of sleep and wakefulness. Electrode leads from the transmitter body were led subcutaneously to the head, and the free ends were placed into holes drilled in the skull to record cortical EEG (approximate coordinates from Bregma: 1.5 and −3.5 mm AP, ± 1.5 mm ML), movements, and body temperature. Ibuprofen (30 mg/kg, oral) to alleviate potential pain was continuously available in each animal’s drinking water for 24–48 hours preoperatively and for a minimum of 72 hours postoperatively. Animals were given at least 2 weeks for post-surgery recovery. During this period, the animals were kept undisturbed except for weekly bedding changes.

Experimental procedures

Methods for contextual fear conditioning and fear memory retrieval have been described in Machida et al. [13], and the current experiment was conducted following the schedule depicted in Figure 1A. Mice were approximately 13–14 weeks old when experiments were performed. First, an undisturbed recording was obtained for each animal and used as a baseline. Mice were divided into two groups (ET24h [n = 8] and ET48h [n = 6] based on delay between shock training [ST] and first reexposure [ET-1] to the shock context [context A]). There were no statistical differences (independent t tests) between groups in baseline body temperature and sleep values prior to beginning experimental procedures (Supplementary Table S1). Subsequently, all animals were habituated to a neutral context (context B; home cage placed within a frame containing photobeam, and outside colony room). Experiments started at the 6th hour after lights on and were conducted in a behavior test room adjacent to the colony/recording room. On the ST day, the mice were placed in a shock chamber (context A; Coulbourn Instruments, Whitehall, PA) and contextually fear conditioned with footshocks administered via grid floors (0.5 mA, 0.5-second duration, 20 trials, 1-minute intervals) by Coulbourn Precision Regulated Animal Shocker (Model E13-14; Coulbourn Instruments, Whitehall, PA). The 30-minute ST procedure consisted of a 5-minute pre-shock period, 20-minute shock presentation period, and 5-minute post-shock period. At 24 hours after ST, the ET24h mice were reexposed to context A for 30 minutes (24h ET-1), whereas the ET48h mice were placed in context B. The latter group experienced delayed ET at 48 hours post-ST (48h ET-1). In order to test retention of fear memory, ET24h animals were again exposed to context A at 48 hours post-ST. Both groups were reexposed to context A at 1 week post-ST to evaluate long-term memory. It should be noted that, at this time point (1 week), ET24h mice were exposed to context A twice (24 and 48 hours), but ET48h mice were exposed only once (48 hours). Thus, due to this difference, the results for the 1-week time point were cautiously interpreted. Figure 1B depicts when the behavioral, stress, and sleep fear responses were evaluated.

Figure 1.

Experimental schedule and design. (A) Schematic of the experimental procedure used in the study. Initially, 24-hour undisturbed recordings were obtained and used as a baseline (Base). Mice were divided into two groups (ET24h and ET48h, depending on timing of the first extinction training, ET-1). After habituation in context B, all mice were then contextually fear conditioned (ST) in context A. Twenty-four hours post-ST, ET24h mice were reexposed to context A (n = 8; 24-hours ET-1), whereas ET48h mice were reexposed to context B (n = 6). ET24h and ET48h mice were reexposed to context A at 48 hours post-ST (24h ET-2 and 48h ET-1, respectively). One week post-ST, all mice were reexposed to context A. Recordings were conducted after each procedure. (B) Timing and duration of each measure evaluated in this study. Freezing was scored for 5 minutes, and locomotor activity, grooming, and rearing were evaluated for 30 minutes. Post-ST and post-ET, body temperature and sleep were measured for 4 hours immediately after the animals were returned to their home cage. Sleep quantity and associated EEG spectral parameters were also evaluated during the 12-hour dark period.

Evaluation of freezing, overt active behaviors, and body temperature

Freezing, defined as a rigid posture with the complete absence of visible movement except for respiration, is frequently used to evaluate fear and retention of original fear memory [28]. As freezing induced by fear memory is transient, dissipates, and is replaced by risk assessment behaviors [29], we concentrated our analyses on the first 5 minutes after reexposure to context A when freezing is greatest. The percentage time freezing was calculated for each animal and expressed as mean ± SEM.

We also assessed locomotor activity, grooming, and rearing as additional indices of fear. Fear-induced stress and anxiety are thought to be reflected behaviorally in lower levels of locomotion whereas less anxious animals show greater exploration, which is reflected in higher levels of locomotion and in specific behaviors such as rearing [30–32]. Self-grooming typically occurs spontaneously at low arousal as a maintenance behavior, becomes longer during moderate arousal, and can be inhibited by higher stress states that elicit freezing [33].

To evaluate fear-associated behavioral changes during ET, contexts A and B were placed in an apparatus (dimensions: 42 cm × 42 cm) to enable activity measures to be determined by interruption of infrared photobeam sensors and a computerized analysis system (AccuScan Digiscan Model RXYZCM (16) CCD) using previously described procedures [30, 31]. The following parameters were obtained automatically from the photobeam apparatus during the 30-minute reexposures to context A: (1) locomotor activity (total distance animals travelled with values calculated from the number of beam interruptions detected by the horizontal sensors), (2) grooming activity (total numbers of counts when the animal breaks the same beam [or set of beams] repeatedly as typically happens during grooming), and (3) rearing activity (assessed by total number of beam interruptions detected by the vertical sensors). The apparatus may potentially induce a moderate level of anxiety due to its open field-like setting; thus, following baseline recording but prior to the ST day, all animals were habituated to the photobeam-based apparatus for 30 minutes in context B. After each trial, the ratio of vertical counts and horizontal counts (V/H) was evaluated. In cases where the value was zero, which indicates either the animal was inactive, or a lack of concordance between the two measures, we utilized only horizontal counts because this measure shows less individual variability in the test environment [30]. As a result, rearing data from two ET24h mice were excluded from the analysis.

A stress-induced increase in body temperature (SIH, also called psychogenic fever) that occurs in all mammals including humans in preparation for fight-or-flight reactions [22] was evaluated for 4 hours immediately after exposure to ST and ET. The time course of SIH parallels that of hypothalamic-pituitary-adrenal axis activation and has been used as a measure of acute stress response [22]. SIH was defined as the difference between stress-induced (T_S) and basal (T_B) measurements (ΔT = T_S − T_B) of body temperature [22]. Under basal conditions, the mean core body temperature of the ET24h and ET48h mice did not significantly differ (Supplementary Table S1).

Recording and determination of sleep states

Immediately after each procedure, the animals were returned to the colony/recording room to obtain uninterrupted recordings for 24 hours. Individual cages were placed on a telemetry receiver (RPC-1, DSI), and EEG and transistor–transistor logic (TTL) pulses from the transmitter were processed and collected by DSI software for subsequent offline data analyses. TTL pulses generated by the telemetry system when the mice moved around in their cages were used as a measure of movement. EEG signals were digitized at 256 Hz and spectrally analyzed by SleepSign for Animal (Kissei, Nagano, Japan). Sleep and wakefulness of the animals were visually scored by a trained observer based on EEG and activity in 10-second epochs using the scoring module of the SleepSign program as previously described [13, 34]. Briefly, each epoch was scored as wakefulness with or without movement (AW or QW, respectively), NREM, or REM. During NREM, the EEG was characterized as high-amplitude slow waves, whereas REM was characterized by regularly spaced lower amplitude waves. Nine hours of dark period data for 4 ET24h and 3 ET48h mice were lost during the 48-hour post-ST recording due to a power outage.

EEG spectral analysis

EEG signals were spectrally analyzed by SleepSign using a Fast Fourier transformation algorithm. Signals were sorted by frequency in 0.5 Hz bins from 0 to 20 Hz and values were normalized to the total power within 0–20 Hz. Power values in the spectrum band between 0–4 Hz were summed and defined as delta wave amplitude during NREM to represent delta (δ) activity (NREM-δ), and the power values in the spectrum band between 5–8 Hz were summed and defined as theta wave amplitude during REM to represent theta (θ) activity.

Data analysis

Data were analyzed using between or within measures analysis of variance (ANOVA) procedures (Sigma Plot 12.0, Systat Software, Inc., San Jose, CA). In cases where the equal variance test failed, the data were analyzed using the Kruskal–Wallis ANOVA on ranks. Post hoc pairwise comparisons were conducted using Tukey tests when appropriate. Differences were considered significant at p less than 0.05.

Results

ET-1 at 24 and 48 hours produces similar freezing, but differs in active behaviors and stress responses

Freezing

Freezing in the ET24h and ET48h groups was virtually identical on first reexposure to the shock context (ET-1; Figure 2A). Freezing of the ET24h animals differed across exposures (F_{(2, 14)} = 17.526, p < 0.001); specifically, pairwise comparison revealed significant reductions at 48 hours (p = 0.002) and 1 week (p < 0.001) in comparison to levels at 24 hours, but freezing levels at 48 hours and 1 week were almost identical (Figure 2A, left panel). At 48 hours post-ST, freezing differed between the ET24h (second exposure) and ET48h (first exposure) groups (F_{(1, 5)} = 11.884, p = 0.018), and at 1 week post-ST, the levels of freezing for both groups were almost identical.

Figure 2.

Effects of 24h ET-1 and 48h ET-1 on freezing and overt active behaviors. Effects of extinction training (ET) in context A at 24 hours, 48 hours, and 1 week post-ST on freezing (A), locomotor activity (B), grooming (C), and rearing (D) of ET24h mice (left, n = 6–8) and ET48h mice (right, n = 6). Freezing was scored for 5 minutes after exposure to context A. Locomotor activity (cm), grooming, and rearing (counts) were evaluated for 30 minutes. ET-1: first ET. All data are represented as mean ± SEM. *p < 0.05, comparison to ET-1. #p < 0.05, between 24h ET-1 and 48h ET-1.

Locomotion, grooming, and rearing

Activity levels of each group did not significantly differ during habituation in the first exposure to context B (Supplementary Table S2). The ET24h and ET48h mice did not significantly differ in locomotion or grooming during ET-1, though the ET48h mice had somewhat lower levels (Figure 2, B and C). However, the ET24h mice did show significantly greater rearing activity during ET-1 than did the ET48h mice on their first context reexposure (H = 9.000, p = 0.01; Figure 2D). The rearing activity level was low in the ET48 animals at both 48 hours and 1 week post-ST. These data demonstrate that the ET24 and ET48 mice differed in some behavioral measures of anxiety on their first context reexposure, even though freezing was virtually identical. This did not appear to be due to simple passage of time as overt active behaviors, including rearing, of the ET48h mice in context B at 24 hours post-ST did not differ from the level observed during pre-ST habituation (Supplementary Table S2).

Stress-induced hyperthermia

We evaluated SIH as an index of the conditioned stress response induced by fear memory. Baseline body temperatures (Supplementary Table S1) and SIH responses to ST (Figure 3) did not differ between groups. To compare SIH in ET24h and ET48h mice, we examined temperatures during the 4-hour period immediately after reexposures to context A. There were significant differences in body temperature at H1 among treatments (ET24h: F_{(4, 28)} = 40.256, p < 0.001; ET48h: F_{(4, 20)} = 13.821, p < 0.001), and pairwise comparisons found that all treatment conditions in both groups produced significant increases in body temperature in comparison to baseline (Figure 3A). However, we found that SIH after ET-1 in the ET24h mice was significantly smaller than that in the ET48h mice on their first context reexposure (F_{(1, 12)} = 7.229, p = 0.02; Figure 3B). These data suggest either reduced stress or a more rapid reduction in the stress response after the earlier extinction trials in the ET24h mice.

Figure 3.

SIH after shock and ET. (A) Core body temperature of ET24h (left) and ET48h (right) mice presented hourly for the first 4 hours after ST and ET. (B) SIH values during H1 expressed as differences from baseline (Base) after ST and ET at 24 hours, 48 hours, and 1 week. ET24h (left) and ET48h (right). All data are represented as mean ± SEM. ET24h (n = 8). ET48h (n = 6). *p < 0.05, comparison to Base. #p < 0.05, comparisons between 24-hours ET-1 and 48-hours ET-1. SIH, stress-induced hyperthermia; ST, shock training; ET, extinction training.

ST and ET-1 at 24 hours produce similar effects on sleep and different effects on REM-θ

Effects of ST and ET on sleep in ET24h mice

Figure 4 shows that REM and NREM of ET24h animals were reduced immediately after ST and ET-1, followed by robust rebounds primarily during the subsequent dark period. ANOVA revealed treatment effects at H1 (REM: F_{(2, 14)} = 26.529, p < 0.001; NREM: F_{(2, 14)} = 8.150, p = 0.004) and pairwise comparisons found that both ST and ET-1 reduced REM and NREM significantly in comparison to time-matched baseline amounts (Figure 4, A and B). However, at H2, sleep of both stages returned to baseline levels. Although REM loss following ST and ET-1 was small compared to NREM loss, the time required for REM recovery was longer. ANOVA showed treatment effects in the H1–H4 totals of REM (F_{(2, 14)} = 9.014, p = 0.003), and pairwise comparisons revealed that after ST (p = 0.010) and ET-1 (p = 0.005), 4-hour total REM was significantly less than that of baseline (Figure 4A). Significant reductions in REM (H1-H4 total) were also found after the second ET at 48 hours (ET-2; F_{(1, 7)} = 9.450, p = 0.018), but not after the third ET at 1 week (ET-3; Supplementary Figure S1, left).

Figure 4.

Sleep and EEG parameters in ET24h mice during selected time periods. Post-ST and 24h ET-1 results for sleep and EEG analysis were compared to Base. Sleep amount during the immediate 4 hours (H1–H4 and H1–H4 total): REM (A), NREM (B), and 4-hour dark period blocks (D1–D3) and 12-hour dark total: REM (C), NREM (D). Relative EEG power (% to the total of 0–20 Hz) of spectral bands for 0–4 Hz or 5–8 Hz was summed and defined as delta (δ) wave amplitude (DWA), or theta (θ) wave amplitude (TWA), to represent δ activity or θ activity, respectively. REM-associated θ (REM-θ) activity during the immediate 4 hours (E) and 4-hour dark period blocks (G). NREM-associated δ (NREM-δ) activity during the immediate 4 hours (F) and 4-hour dark period blocks (H). ET24h (n = 8). All data are represented as mean ± SEM. *p < 0.05, comparison to Base. ST, shock training; ET, extinction training.

Significant differences in REM during the dark period were found when the data were analyzed in 4-hour blocks (D1: F_{(2, 14)} = 10.278, p = 0.002; D2: F_{(2, 14)} = 6.642, p = 0.009; D3: χ ² = 6.000, p = 0.047) and for the total 12-hour dark period (F_{(2, 14)} = 25.377, p < 0.001). Pairwise comparisons among means revealed increased dark period REM after ST and ET-1 for all analysis periods (Figure 4C). Treatment effects during dark period were also observed in NREM, but only for D1 (F_{(2, 14)} = 20.242, p < 0.001) and for the 12-hour dark total (F_{(2, 14)} = 23.578, p < 0.001) (Figure 4D). Pairwise comparison showed that the increases were similar for ST and ET-1. These results show that ST and ET produced almost identical effects on subsequent REM and NREM in terms of amounts, direction, and duration.

Effects of ST and ET on REM-θ and NREM-δ in ET24h mice

We found that statistically significant changes in REM-θ during H2 and during D2 in the subsequent dark period (H2: F_{(2, 14)} = 8.242, p = 0.004; D2: F_{(2, 14)} = 6.583, p = 0.01), and that, after ET-1, but not after ST, the amplitude of REM-θ was significantly reduced at H2 (p = 0.003) and during D2 (p = 0.007) compared to baseline (Figure 4, E and G, respectively). Directional changes in REM amounts during these periods did not predict the reductions in θ activity. For example, although significant increases in REM were observed in D2 after both ST and ET-1, the reduction in θ activity was observed only after ET-1.

ST and ET-1 produced similar effects on NREM-δ. NREM-δ was significantly increased for the first 3 hours of recording (H1: F_{(2, 14)} = 14.789, p < 0.001; H2: F_{(2, 14)} = 11.300, p = 0.001; H3: F_{(2, 14)} = 11.057, p = 0.001) after both ST and at ET-1 compared to baseline (Figure 4F). No significant alterations were observed in NREM-δ during the dark period (Figure 4H).

ET-1 at 48 and 24 hours produce similar effects on sleep, but ET-1 at 48 hours did not alter REM-θ

Effects of ST and ET on sleep in ET48h mice.

We assessed the effects of delayed ET (48h ET-1) on sleep and sleep-associated EEG parameters using the ET48h mice (Figure 5). Significant changes were found in both REM and NREM at H1 (REM: F_{(2, 10)} = 13.24, p = 0.002; NREM: F_{(2, 10)} = 12.268, p = 0.002). Significant reductions in both REM and NREM were observed after ST and 48h ET-1 in comparison to time-matched baseline amounts (Figure 5, A and B, respectively). The H1–H4 totals of REM, but not NREM, also were significantly changed (F_{(2, 10)} = 8.074, p = 0.008), and pairwise comparisons revealed both ST and ET produced significant reductions in total REM (Figure 5, A and B). The significant changes in sleep amounts found for the 48h ET-1 session were virtually identical to those found for the 24h ET-1 session (see later).

Figure 5.

Sleep and EEG parameters in ET48h mice during selected time periods. Post-ST and 48h ET-1 results for sleep and EEG analysis were compared to Base. Sleep amount during the immediate 4 hours (H1–H4 and H1–H4 total): REM (A), NREM (B), and 4-hour dark period blocks (D1–D3) and 12-hour dark total: REM (C), NREM (D). Relative EEG power (% to the total of 0–20 Hz) of spectral bands for 0–4 Hz or 5–8 Hz was summed and defined as delta (δ) wave amplitude (DWA), or theta (θ) wave amplitude (TWA), to represent δ activity or θ activity, respectively. REM-θ during the immediate 4 hours (E) and 4-hour dark period blocks (G). NREM-δ during the immediate 4 hours (F) and 4-hour dark period blocks (H). ET48h (n = 6) in A, B, E, and F, and n = 3 in C, D, G, and H. All data are represented as mean ± SEM. *p < 0.05, comparison to Base. ST, shock training; ET, extinction training.

During the dark period, significant differences were found for both REM (D1: F_{(2, 7)} = 6.204, p = 0.028; D2: F_{(2, 7)} = 12.195, p = 0.005; D3: F_{(2, 7)} = 5.948, p = 0.031; 12-hour dark total: F_{(2, 7)} = 18.447, p = 0.002; Figure 5C) and NREM (D1: NS; D2: F_{(1, 5)} = 8.446, p = 0.034; D3: F_{(2, 7)} = 7.882, p = 0.016; 12-hour dark total: F_{(2, 7)} = 9.247, p = 0.011; Figure 5D). Pairwise comparisons revealed that increases were found after ST (REM: D1: p = 0.028, D2: p = 0.013, D3: p = 0.049, dark total: p = 0.002; NREM: D2: p = 0.034, D3: p = 0.016, dark total: p = 0.011), that were similar to those observed post-ST found in ET24h animals (see later). However, potentially due to a nighttime power outage after 48h ET-1 and lost data, post-ET significance was only observed for increased REM during D2 (p = 0.010) and for the total dark period (p = 0.010). Apparent increases in dark period NREM did not reach significance for ET.

Effects of ST and ET on REM-θ and NREM-δ in ET48h mice

Although the immediate effects of ET-1 at 24 and 48 hours on sleep were quantitatively similar, there was no significant decrease in REM-θ activity during H2 or during the dark period ET48h mice (Figure 5, E and G). By comparison, similar to the results observed in the ET24h animals, immediate changes in NREM-δ were observed (H1: F_{(2, 9)} = 10.764, p = 0.004; H2: F_{(2, 10)} = 7.883, p = 0.009), and pairwise analysis revealed a significance difference compared to baseline both post-ST (H1: p = 0.02, H2: p = 0.009) and post-ET-1 (H1: p = 0.005, H2: p = 0.035) in the ET48h mice (Figure 5F). There was no significant difference found in NREM-δ during the dark period (Figure 5H).

Comparisons of REM and REM-θ at 24h ET-1 and 48h ET-1

Figure 6 shows direct comparisons of REM and REM-θ for 24h and 48h ET-1 using values normalized to baseline. There were no significant differences between 24h ET-1 and 48h ET-1 either for immediate alterations in total REM amounts (Figure 6A) or for 12-hour dark total REM (Figure 6C). However, REM-θ values for 24h ET-1 and 48h ET-1 were significantly different at H2 (F_{(1, 12)} = 5.452, p = 0.038; Figure 6B) and during the dark period at D2 (F_{(1, 9)} = 9.606, p = 0.013; Figure 6D). These results suggest that alterations in REM-θ associated with ET-1 only occur within a specific period after initial fear learning, and are not necessarily linked to quantitative changes in REM. By comparison, NREM amounts and NREM-δ did not differ across groups during either time period (data not shown).

Figure 6.

Comparisons of sleep and EEG parameters between ET24h and ET48h mice. Effects of the first extinction training (ET-1) session on sleep and EEG activity during selected time periods were directly compared in the ET24h and ET48 mice. (A) Total REM amount for immediate 4 hours. (B) REM-θ during immediate 4 hours (H1–H4). (C) Total REM amount for 12-hour dark period. (D) REM-θ during 4-hour dark period blocks (D1–D3). All values were normalized to baseline (Base) for each group. All data represent mean ± SEM. #p < 0.05, comparisons between 24h ET-1 and 48h ET-1 (24h ET-1, n = 8; 48h ET-1, n = 6 in A and B, n = 3 in C and D).

Discussion

Data in this study demonstrate complex outcomes for fear conditioning and extinction when multiple outcome measures and time points are considered. Critically, freezing alone was not predictive of changes in other fear-associated responses. Freezing is a conventionally used index of fear memory with greater freezing being interpreted as an indication of greater fear and of stronger fear memory retention [35]. Concomitantly, reduced freezing is typically interpreted as indicating extinguished fear and/or attenuated fear memory. However, the current results that used multiple measures to assess fear-conditioned responses demonstrate that a change in freezing is not reflective of all aspects of the fear response. For example, freezing was not predictive of behaviors that varied with ET delay. We found that 24h ET-1 and 48h ET-1 induced different effects on overt behavior with 24h ET-1 resulting in significantly increased rearing and a nominal increase in locomotion indicating a putatively lower level of anxiety compared to 48h ET-1. This suggests potential differences in the anxiolytic effect of extinction at different post-conditioning times on subsequent behaviors, even though the responses in freezing after 24h ET-1 and 48h ET-1 were very similar. This suggestion is in line with previous finding that the age of the memory is an important variable that regulates the temporal dynamics of memory reconsolidation, such that younger memories are more easily reconsolidated than older memories [28].

In addition to the dissociation of freezing and overt behavior, freezing was not predictive of fear-conditioned changes in temperature (SIH). This finding is consistent with work in rats that has demonstrated that freezing can extinguish at high levels of SIH [10–12]. The difference in SIH between the ET24h and ET48h mice after ET-1 was remarkable because it suggests that early extinction may actually be associated with a reduction in the stress response. We have never observed an SIH response this small (~0.5°C over baseline) in any study in which we conducted ET at longer delays after ST [10–12].

The lack of a relationship between REM amounts and fear learning and memory consolidation (as indicated by freezing) is consistent with our prior work where we have found increases and decreases in REM after virtually identical freezing amounts [16]. As noted earlier, changes in sleep can be fear-conditioned responses that show extinction with repeated reexposures to fearful stimuli [14]. Interestingly, work examining the effect of REM deprivation found that both immediate [7–9] and delayed [8] deprivation had minimal effects on extinction of contextual freezing. By comparison, REM deprivation performed for 0–6 hours immediately after ET has been reported to impair extinction of cued freezing (measured 24 hours post-ST), whereas delayed (6–12 hours after ET) does not produce impairment [8]. Thus, there is the possibility that the association between REM and freezing differs with cued and contextual fear.

A major finding of this study was that extinction determined by multiple outcome measures in the ET24h mice was associated with two distinct “time windows” when REM-θ activity was significantly attenuated, one immediately post-ET (H2) and one during the following dark period (D2). Reduced REM-θ activity was found only post-24h ET-1, and not post-ST, nor post-48h ET-1, and it was not associated with any apparent alteration in REM quantity. Unfortunately, a power outage interrupted recording of the 48-hour time point for several animals and prevented full analysis of REM-θ activity during the dark period. Thus, it is possible that a reduction in REM-θ activity occurred at some later time point; however, the early response after ET was significantly different depending on whether ET was conducted at 24 or 48 hours after ST. It is also possible that the intervening exposure to context B was a factor; however, this manipulation was essentially carried out with the animal remaining in its home cage. There also was no indication that fear generalized to context B after ST as there were no significant differences in pre- and post-ST behaviors, sleep, or body temperature (SIH) of ET48h animals (Supplementary Table S2).

As for REM-θ, the general consensus is that the large-scale θ oscillation detected in the EEG, as recorded and analyzed in our study, is primarily produced in the hippocampus [26], and that changes in power amplitude arise from macroscopic changes in synchronization within local neural ensembles [36]. Accordingly, the phasic changes in θ activity that we observed likely reflected neural activity in the hippocampus [37]. Recently, a functional role of θ oscillations has been hypothesized as a means to enable network-level cooperation for the collective actions of single neuron computations underlying cognitive functions [38, 39], including memory consolidation [27] and fear extinction [40]. For example, Boyce et al. suggested that in vivo hippocampal θ oscillations during REM are necessary for consolidation of contextual fear memory. In addition, while the hippocampus is central to the regulation of θ oscillations, studies using local field potential and unit recordings have found phase-locked discharge of neurons responding to hippocampal θ waves in the amygdala and the mPFC [27, 41–43]. These structures exhibit “synchronized θ activity” reflecting selective involvement of the hippocampus [44], for example, θ coherence increases during contextual fear conditioning [45], but declines during extinction learning [40]. Thus, synchronized θ activity has been hypothesized to provide means for connecting neural ensembles temporally and functionally [38, 40, 46]. Accordingly, the reduced REM-θ activity shortly after ET and during the following dark period in mice receiving ET at 24 hours after ST may reflect changes in synchronization within hippocampal-associated neural ensembles acting to regulate the consolidation of fear extinction. However, if this is the case, the lack of a similar reduction in REM-θ activity in mice receiving ET at 48 hours after ST, at least in the early post-ET period, is difficult to explain. Differences may relate to the timing of ET and also to the intensity of ST. Many fear-conditioning studies carry out ET at 24 hours post-ST [40, 45] and use fewer shock presentations [40, 45] than we typically use in our studies.

Regardless of the potential differences between paradigms, the presence of two periods with decreased REM-θ activity in the ET24h group, and lack of altered REM-θ activity in the ET48h group, is intriguing. Even at 1 week post-ST, REM-θ of the ET24h animals decreased during the dark period (D2), whereas REM-θ of the ET48h mice at 1 week did not differ from baseline, and the difference between the two groups was significant (p < 0.013, F(1, 5) = 14.408; Supplementary Figure S2). It should be noted that, at this time point, the ET24h animals had received three ET sessions, whereas the ET48h animals had received two ET sessions. Therefore, it is possible that frequency of memory retrieval could also be a factor in the suppression of REM-θ.

Similar to the association we found between extinction and REM-θ activity at 24 hours after ST, Datta et al.[47] group found that the rat P-wave, an REM-associated phasic potential originating in the pons, is linked to successful extinction learning as indicated by reduced freezing. In the study, among rats that underwent contextual fear conditioning followed by ET, and only those that displayed an increase in P-waves in post-ET sleep continued to show reduced freezing [47]. They also found that changes in REM quantity after ET did not guarantee consolidation of extinction for the freezing response, a finding consistent with the current results. It should also be noted that this study conducted ET at 24 hours post-ST, a time that our results suggest may have a different relationship to REM-related phenomena from that of ET conducted at longer post-consolidation periods.

In summary, the current work demonstrates that the relationship between behavioral indices of fear, the stress response, and subsequent sleep is complex and is not fully explainable by current concepts guiding research on fear memory and extinction. Freezing is the generally accepted standard of conditioned fear memory, and its presence in response to a fearful context or cue does indicate the evocation of fear memory and some level of emotional fear. However, the presence of freezing is not predictive of subsequent fear-induced alterations in sleep, and as the current results and other studies demonstrate [10–12], freezing may also show different patterns of extinction than do fear-conditioned changes in sleep and the stress response as well as fear-related behaviors. These differences may be related to the presence or absence of alterations in REM-θ activity, as extinction in the ET24h mice, but not ET48h mice, was associated with two distinct “time windows” when REM-θ activity was significantly attenuated. Thus, although freezing is attractive as a simple index of fear, its predictive limitations suggest that multiple fear outputs will likely need to be considered before true insight is attained into how fear memory regulates the relationships between overt fear, stress responses, and subsequent sleep. This study also demonstrates that considering the time delay between fear acquisition and extinction will be important for assessing these relationships at the mechanistic level. Considering multiple outputs and time passed since a stressful experience will be particularly important for studies seeking to understand the neural circuits and mechanisms underlying failed extinction in fear and stress-related disorders, such as PTSD, which have, thus far, primarily relied solely on overt indices of fear.

Funding

This work was supported by National Institutes of Health research grant MH64827.

Conflict of interest statement. This was not an industry-supported study. The authors have indicated no financial conflicts of interest. There was no off-label or investigational use of drugs.