Abstract
It is widely accepted that activation of NMDA receptors (NMDAR) is necessary for the formation of fear memories in the basolateral amygdala complex (BLA). This acceptance is based on findings that blockade of NMDAR in the BLA disrupts Pavlovian fear conditioning in rodents when initially innocuous stimuli are paired with aversive and unexpected events (surprising foot shock). The present study challenges this acceptance by showing that the involvement of NMDAR in Pavlovian fear conditioning is determined by prediction errors in relation to aversive events. In the initial experiments, male rats received a BLA infusion of the NMDAR antagonist, D-AP5 and were then exposed to pairings of a novel target stimulus and foot shock. This infusion disrupted acquisition of fear to the target when the shock was surprising (experiments 1a, 1b, 2a, 2b, 3a, and 3b) but spared fear to the target when the shock was expected based on the context, time and other stimuli that were present (experiments 1a and 1b). Under the latter circumstances, fear to the target required activation of calcium-permeable AMPAR (CP-AMPA; experiments 4a, 4b, and 4c), which, using electrophysiology, were shown to regulate the activity of interneurons in the BLA (experiment 5). Thus, NMDAR activation is not required for fear conditioning when danger occurs as expected given the context, time and stimuli present, but is required for fear conditioning when danger occurs unexpectedly. These findings are related to current theories of NMDAR function and ways that prediction errors might influence the substrates of fear memory formation in the BLA.
SIGNIFICANCE STATEMENT It is widely accepted that NMDA receptors (NMDAR) in the basolateral amygdala complex (BLA) are activated by pairings of a conditioned stimulus (CS) and an aversive unconditioned (US) stimulus, leading to the synaptic changes that underlie formation of a CS-US association. The present findings are significant in showing that this theory is incomplete. When the aversive US is unexpected, animals encode all features of the situation (context, time and stimuli present) as a new fear/threat memory, which is regulated by NMDAR in the BLA. However, when the US is expected based on the context, time and stimuli present, the new fear memory is assimilated into networks that represent those features, which occurs independently of NMDAR activation in the BLA.
Introduction
Pavlovian fear conditioning in rats is widely used to study the substrates of learning and memory in the mammalian brain. In a standard protocol, rats are exposed to pairings of an initially neutral conditioned stimulus (CS), such as a tone or a light, and an innately aversive unconditioned stimulus (US), typically foot shock. After these pairings, subsequent presentations of the CS elicit a range of defensive responses, including changes in heart rate and blood pressure, release of stress hormones, potentiated startle and freezing. These responses are taken to indicate formation of a CS-US memory that is retrieved during presentations of the CS and expressed in responses that defend the animal from threat or harm.
Formation of a CS-US memory depends on cellular processes in a network of brain regions, including activation of NMDA receptors (NMDARs) in the basolateral amygdala complex (BLA). This is supported by findings that blocking NMDAR in the BLA impairs formation of a CS-US fear memory; and does so independently of the type of CS (auditory, visual, context), its temporal relation with the US (delay or trace pairings) and the responses that index its formation (freezing, potentiated startle; Gewirtz and Davis, 1997; Rodrigues et al., 2001; Bauer et al., 2002; Parkes and Westbrook, 2010). However, recent studies of context fear conditioning in rats showed that the involvement of NMDAR in fear memory formation is not obligatory (Finnie et al., 2018; see also Sanders and Fanselow, 2003; Hardt et al., 2009; Wang et al., 2012). For example, in the Finnie et al. (2018) study, an infusion of the NMDAR antagonist, D-AP5 into the dorsal hippocampus disrupted context fear conditioning when naive rats were shocked in a distinctive context; but completely spared this conditioning when (1) rats had been previously conditioned to fear a very similar context, and (2) the temporal characteristics of the conditioning were preserved across the shift from context 1 to context 2 (e.g., the placement-to-shock interval). These findings imply that the role of hippocampal NMDAR in fear memory formation is determined by similarity between new and past experiences; and, more generally, that the role of NMDAR in fear memory formation may be determined by the degree to which animals can predict the aversive US based on their prior training. We reasoned, therefore, that activation of NMDAR in the BLA, while required for conditioning when the US is unexpected, may not be necessary when animals expect the US based on their prior training.
Accordingly, the present study examined whether the role of NMDAR in fear conditioning depends on whether the US was expected or unexpected. To do so, the initial experiments used a serial-order conditioning protocol in which rats were exposed to a sequence of stimuli, e.g., a visual stimulus (labeled S2) whose presentations terminated in onset of an auditory stimulus (labeled S1), which then terminated in foot shock. Immediately before this conditioning session, rats received a BLA infusion of the NMDAR antagonist, D-AP5 or vehicle, and 2 d later were tested for their levels of fear (measured in freezing) to presentations of the S1 and S2 alone. Critically, the experiments differed in whether or not the S1 element of the serial-order compound had been pretrained as a signal for shock; and the question of interest concerned the impact of this pretraining on the NMDAR dependence of fear to S2.
The results of these experiments showed that NMDAR in the BLA encode fear to S2 when it is paired with an unexpected US (i.e., the shock is not signaled by S1), but not when it is paired with an expected US (i.e., the shock is signaled by S1; experiments 1a and 1b). The remaining experiments then used variations of the protocol to examine the conditions under which NMDAR encode fear to S2 (experiments 2a, 2b, 3a, and 3b) and the substrates of this fear when it is NMDAR independent (experiments 4a, 4b, 4c, and 5).
Materials and Methods
Subjects
Subjects were 182 experimentally-naive adult Long–Evans rats, obtained from a colony maintained by the School of Psychology, University of New South Wales. There were 94 males weighing 400–500 g and 65 females weighing 250–350 g. Rats were housed by sex in plastic tubs (22 cm high × 67 cm long × 40 cm wide), with four rats per tub. The tubs were located in a colony room maintained at 20–22°C, with lights on between 7 A.M. and 7 P.M. Rats were handled each day for at least 5 d before behavioral procedures began. Food and water were continuously available in the tubs.
Apparatus
Unless otherwise stated, all procedures were conducted in four identical chambers (30 cm wide × 26 cm high × 30 cm deep), each located in sound-attenuating and light-attenuating wooden cabinets. The rear and front walls of the chambers were made of Plexiglas, the side walls and ceiling of aluminum, and the floors of stainless-steel rods (7 mm in diameter, spaced 1.8 mm apart). Shock could be delivered through each floor via custom-built generators located in another room in the laboratory. A tray below each floor contained bedding material. An LED covered with a reflector was mounted on the rear wall of each cabinet and used to deliver a visual stimulus which flashed at a rate of 3.5 Hz and whose intensity was ∼8 lux at the center of the chamber. A speaker, also mounted on the rear wall of each chamber, was used to deliver an auditory stimulus, a 620-Hz square wave tone at an intensity of 65 dB against a background noise of ∼45 dB. An infrared light, mounted on the rear wall of the chamber, allowed each rat to be recorded via a camera, which was also mounted on the wall of the cabinet, and connected to a DVD recorder located in another room in the laboratory.
Experiment 3a used the chambers described, as well as a second set of four identical chambers (30 cm wide × 26 cm high × 30 cm deep). Each of these chambers was also located in its own sound-attenuating and light-attenuating wooden cabinets but differed from the chambers already described in the size and spacing of the stainless-steel rods (2 mm in diameter, spaced 1.5 cm apart). They were also located in a different room in the laboratory and were scented with drops of liquid peppermint extract, which were placed in plastic weigh boats located inside the wooden cabinets, adjacent to the training chambers. The rear wall of each cabinet contained the same sources of stimuli and recording as described.
Stimuli
Each experiment used the auditory and visual stimuli, counterbalanced in their roles as S1 and S2, such that the auditory stimulus was S1 and the visual stimulus was S2 for half of the subjects in each experiment, while the visual stimulus was S1 and the auditory stimulus was S2 for the remainder. The durations of S2 and S1 were 30 and 10 s, respectively, and were those used in previous studies of second-order (Lay et al., 2018) and serial-order (Leidl et al., 2018; Williams-Spooner et al., 2019) conditioning. The intensity of the foot shock was 0.8 mA, and its duration was 0.5 s.
Scoring and statistical analysis
Freezing was the measure of fear conditioning. A time sampling procedure was used in which each rat was scored for 1 s every 2 s for freezing or not freezing (Fanselow, 1980). The number of occasions on which the rat was scored as freezing was converted to a percentage of the total number of observations.
When scoring freezing responses, we occasionally see bouts of darting and/or flight behavior when animals are exposed to the foot shock US. This is consistent with recent findings reported by Fadok et al. (2017) and can complicate interpretation of freezing data from training sessions. However, we have never seen any such behaviors during test sessions where we assess levels of fear to CSs (S1 and S2) under conditions of extinction. This is also consistent with the findings reported by Fadok and colleagues, and for this reason, we place greater emphasis on the freezing data from test sessions rather than the training sessions in each experiment.
All test data were scored by the experimenter and cross-scored by an experienced observer who was blind to the purposes of the experiment. The Pearson product moment correlation was calculated to assess the reliability of the scores by the two observers. This correlation was >0.9 in each of the experiments. Any discrepancies between the scores were resolved in favor of those by the naive observer. The freezing elicited by S1 and S2 were analyzed separately. Freezing across training and testing were analyzed using a contrast testing procedure with a between-subject factor of group and a within-subject factor of trial. The criterion for rejection of the null hypothesis (α) was set at 0.05. For all significant differences, we calculated standardized 95% confidence intervals (95% CI) as well as Cohen's d (d) as a measure of effect size. By convention, a Cohen's d of 0.2 indicates a small effect, 0.5 indicates a medium-sized effect, and 0.8 or above indicates a large effect.
Methods
The surgical, infusion and histologic procedures described below were used in all subsequent experiments, unless otherwise indicated.
Surgery
All rats were surgically implanted with bilateral cannulas targeting the BLA. Rats were placed into an induction chamber and anesthetized with isoflurane (5%) delivered in oxygen (1.0 l/min). The anesthetized rat was then positioned on a stereotaxic apparatus, and the plane of anesthesia was maintained through continued delivery of isoflurane (1.5–2.5%) in oxygen (0.5–0.7 l/min) via a face mask. Each rat's head was shaved, and subcutaneous injections were administered. These injections included the sodium channel blocker, bupivacaine (0.5%, 0.2 ml), which was administered along the line of the first incision to numb the local area; the antibiotic Benacilin (0.15 ml/kg), which was administered near the neck immediately posterior to the skull; and the nonsteroidal, anti-inflammatory carprofen (1 ml/kg, mixed 1:9 with 0.9% sterile saline), which was also administered near the neck. After the skull was cleaned, two 26-gauge guide cannulas were implanted into the brain via holes drilled in the skull. The tips of the guide cannulas were aimed at the BLA in each hemisphere (2.3–2.5 mm posterior to bregma, 4.8–4.9 mm lateral to the midline and 8.1–8.2 mm ventral to bregma). The guide cannulas were secured in position with dental cement and four jeweler's screws, and a dummy cannula was kept in each guide at all times except during infusions. Following surgery, rats received a fluid-replacing subcutaneous injection of 0.9% sterile saline (3 ml) below the loose skin around the hips and were placed on a heating mat until they recovered from the effects of the anesthetic. They were then returned to their home tubs and given a minimum of 7 d to recover from surgery, during which time they were handled and weighed daily.
Figure 1a–c shows the most ventral point of cannulas within the BLA in the experiments involving the competitive, nonselective NMDAR antagonist, D-AP5 (Fig. 1a; experiments 1a–3b), the competitive, nonselective AMPAR antagonist, NBQX (Fig. 1b; experiment 4a), and the use-dependent CP-AMPAR antagonist, NASPM (Fig. 1c; experiments 4b and 4c).
Cannula placements as verified on Nissl-stained sections. Black dots represent the most ventral point of the cannula track for each rat on coronal sections based on the atlas of Paxinos and Watson (2007). Cannula placements are shown for the experiments that involved: based on the drug that (a) D-AP5 to block NMDAR (experiments 1a, 1b, 2a, 2b, 3a, and 3b, n = 120), (b) NBQX to block AMPAR (experiment 4a, n = 19), and (c) NASPM to block CP-AMPAR (experiments 4b and 4c, n = 43).
Drug infusions
All drug infusions in this and subsequent experiments were administered into the BLA in the same volume (0.5 μl/side) and at the same rate (0.25 μl/min). Rats were infused with: D-AP5 [10 μg/μl, mixed with artificial CSF (ACSF) = 50.7 mm, Sigma-Aldrich; based on Maren et al., 1996] to block NMDAR activity (experiments 1a–3b), NBQX (10 μg/μl, mixed with 0.9% sterile saline = 26.3 mm, Sigma-Aldrich; based on Walker and Davis, 1997; Zimmerman and Maren, 2010) to block AMPAR-mediated synaptic signaling (experiment 4a), NASPM [80 μg/μl, mixed with PBS (pH 7.2, 0.1 M) = 41.7 mm, Tocris Bioscience; based on Conrad et al., 2008; White et al., 2016] to block CP-AMPAR activation (experiments 4b and 4c), or the corresponding infusion of vehicle alone.
Histology
Following behavioral testing, rats were euthanized with an intraperitoneal injection of a lethal dose of sodium pentobarbital, decapitated, and their brains rapidly removed and frozen. Brains were sliced on a cryostat into coronal sections of 40-µm thickness. Every second section through the BLA was mounted on a glass microscope slide and then stained with cresyl violet. Slides were observed under a microscope to confirm the location of cannula tips using the brain atlas of Paxinos and Watson (2007). Rats with inaccurate cannula placements (one or both cannulas positioned outside the boundaries of the BLA) or with extensive damage to the BLA were excluded from the statistical analysis.
Behavioral procedures
Experiment 1a
Unless otherwise stated, the training and testing procedures in experiment 1a were those used in subsequent experiments.
Context preexposure
On days 1–2, there were two daily sessions of context preexposure, each lasting 20 min, one in the morning and the other in the afternoon. This was done to familiarize the rats with the chambers and minimize any neophobic reactions that might obscure the development of freezing across conditioning.
First-order conditioning
On day 3, rats received four S1-shock pairings. Offset of the 10-s S1 co-terminated with the 0.5-s foot shock. The first S1-shock pairing occurred 5 min after placement in the context, the intertrial interval (ITI) between pairings (S1 offset to S1 onset) was 5 min, and rats remained in the context for 1 min following the final pairing.
Context extinction
On day 4, rats received two 20-min exposures to the conditioning chambers in the absence of any scheduled events, one in the morning and the other in the afternoon. Rats received these exposures to extinguish any context-elicited freezing that might obscure the detection of freezing to the discrete cues across serial-order conditioning.
Serial-order conditioning
On day 5, rats received bilateral intra-BLA infusions of either vehicle (ACSF, n = 8) or the competitive, nonselective NMDAR antagonist, D-AP5 (n = 9) ∼10 min before a session of serial-order conditioning. This session involved four presentations of the sequence S2-S1-shock. On each trial, offset of the 30-s S2 co-occurred with onset of the 10-s S1 which co-terminated in foot shock. The first S2-S1-shock sequence occurred 5 min after placement in the chambers, the interval between the sequences was 5 min (S1 offset to S2 onset), and rats remained in the chambers for 1 min following the final S2-S1-shock sequence.
Context extinction
On day 6, rats received two, 20-min exposures to the chambers in the absence of any scheduled events, one in the morning and the other in the afternoon. Rats received an additional 10-min exposure to the chambers on the morning of day 7. These exposures were intended to extinguish any spontaneously recovered context-elicited freezing that could interfere with detection of freezing elicited by S2 and S1 at test.
Test
On the afternoon of days 7 and 8, rats were tested for freezing to S2 and S1, with the order of testing counterbalanced within each group. Each test consisted in eight presentations of S2 or S1. The interval between presentations was 3 min and the durations of S2 and S1 remained 30 and 10 s, respectively. For a schematic, see Figure 2a.
Extinction of S2 and its reconditioning and testing as a first-order CS
The results of the tests on days 7 and 8 indicated that the BLA infusion of D-AP5 before serial-order conditioning failed to disrupt learning about S2 or any additional learning about S1. Therefore, additional training and testing was conducted to confirm the efficacy of the BLA infusion of D-AP5. This consisted in extinction of S2, its reconditioning via pairings with shock under a BLA infusion of D-AP5 or vehicle, and its subsequent testing.
On each of days 9 and 10, rats received two extinction sessions of S2, one in the morning and the other in the afternoon. Each of the four extinction sessions consisted in eight S2-alone presentations. The first presentation occurred 2 min after placement in the chambers, each presentation lasted for 30 s, the interval between the presentations (S2 offset to S2 onset) was 3 min, and rats remained in the chambers for 1 min following the final presentation.
On day 11, rats received a BLA infusion of vehicle or D-AP5 and, ∼10 min later, four S2-shock pairings. The first S2-shock pairing occurred 5 min after placement in the context, each 30-s S2 presentation co-terminated in the onset of foot shock, the ITI (S2 offset to S2 onset) was 5 min, and rats remained in the context for one additional minute after the final S2-shock pairing. Importantly, the two groups in this part of the experiment, group vehicle (n = 8) and group D-AP5 (n = 9), were reorganized so that each contained equal numbers of rats that had received a BLA infusion of vehicle or drug before the serial-order conditioning session on day 5. One rat had to be immediately euthanized, as its infusion needle broke off inside one of its cannulas on day 11 (group vehicle, n = 8; group D-AP5, n = 8). As its cannulas were correctly positioned inside the BLA, its data from the first part of the experiment were retained.
On day 12, rats were exposed to the chambers for 20 min in the absence of any events and were again exposed for 10 min on the morning of day 13 to extinguish any freezing elicited by the chambers. On the afternoon of day 13, rats were tested with S2 in the manner described previously: the first of the eight 30 s S2 presentations occurred 2 min after placement in the chamber, the interval between the S2 presentations was 3 min, and rats remained in the chamber for 1 min after the final S2 presentation (for a schematic, see Fig. 2a).
Experiment 1b
This experiment had two parts. The first examined the effects of a BLA infusion of D-AP5 on conditioning of the S2 and S1 elements of the serial-order compound but in the absence of the prior conditioning of S1. Briefly, rats were preexposed to the chambers on days 1 and 2, infused with vehicle (n = 11) or D=AP5 (n = 11), and exposed to the S2-S1-shock sequences on day 3. Following extinction of any freezing elicited by the chambers on day 4 and the morning of day 5, rats were tested in a counterbalanced manner with S2 and S1 (for a schematic, see Fig. 3a).
The second part of the experiment was a replication of experiment 1a, it examined the effects of the BLA infusion of D-AP5 on conditioning of the S2 and S1 elements of the serial-order compound when S1 entered the compound already conditioned. To examine these effects, rats were first extinguished to S1 and S2. Extinction of the former occurred across 16 S1-alone presentations, spaced 3 min apart, on each of days 7 and 8; extinction of S2 occurred across 16 S2-alone presentations, spaced 3 min apart, on each of days 9 and 10. Rats then received four S1-shock pairings on day 11 in the manner described in experiment 1a. Following extinction of any freezing elicited by the chambers on day 12, rats received four S2-S1-shock sequences on day 13 in the manner previously described. This occurred under a BLA infusion of vehicle or D-AP5. Half of the rats in each of the groups had been previously infused with D-AP5 and the remainder with vehicle. Any freezing elicited by the chambers was extinguished on day 14 and the morning of day 15. Finally, rats were tested with S2 and S1 on the afternoon of days 15 and 16 in the manner previously described (for a schematic, see Fig. 3a).
Experiment 2a
This experiment conditioned rats to S1 and then conditioned S2 in a trace protocol, i.e., where S1 was omitted. Following preexposure to the chambers (days 1 and 2), conditioning of S1 (day 3) and extinction of freezing to the chambers (day 4), rats received a BLA infusion of vehicle (n = 8) or D-AP5 (n = 8) before trace conditioning of S2 on day 5. This was identical to the serial-order protocol except that the already-conditioned S1 was omitted. That is, termination of each of the four 30-s presentations of S2 was followed 10 s later by foot shock. Freezing elicited by the chambers was then extinguished (day 6 and the morning of day 7) before rats were tested with S2 (day 8) and S1 (day 9; for a schematic, see Fig. 4a).
Experiment 2b
This experiment conditioned S1 and then exposed rats to a second-order protocol, i.e., where S2 preceded S1 as in the serial-order protocol but with the shock US omitted. Across days 1–4 rats received context exposure, S1 conditioning and context extinction in the manner described for days 1–4 in experiment 1a. On day 5, rats received a BLA infusion of vehicle (n = 9) or D-AP5 (n = 11) ∼10 min before exposure to four S2-S1 pairings, each of which consisted in a 30-s S2 presentation that terminated in the onset of the 10-s S1. Rats were exposed to the chambers for 10 min on the morning of day 6 to ensure that there was little or no freezing elicited by the context. They were then tested with S2 on that afternoon, and with S1 on the afternoon of day 7 (for a schematic, see Fig. 5a).
Experiment 3a
The design was similar to that of experiment 1a. Rats were exposed to S1-shock pairings (day 3), S2-S1-shock sequences under a BLA infusion of vehicle or D-AP5 (day 5), and then tested with S2 and S1 (days 7 and 8). The difference was that the S1-shock pairings occurred in a different set of chambers (located in a different room in the laboratory) to that which was used for S2-S1-shock sequences in this and the other experiments. On the morning and afternoons of days 1 and 2, rats were preexposed to the chambers where S1-shock pairings would occur (labeled A) and to those where serial-order conditioning would take place (labeled B). The order of preexposure was completely counterbalanced within and between groups of rats, with half the rats in each group being exposed to the contexts in the order ABBA, while remaining rats were exposed in the order BAAB. On day 3, rats received four S1-shock pairings in the A chambers in the manner previously described. On day 4, rats were exposed to both sets of chambers: a 20-min exposure to the A chamber or the other chamber in the morning, and vice versa in the afternoon. This was done to extinguish any freezing elicited by the A chambers and to assess whether this chamber was discriminated from the chambers (B) where serial-order conditioning would take place. On day 5, rats received a BLA infusion of vehicle (n = 8) or D-AP5 (n = 8) and were then exposed to S2-S1-shock sequences in the B chambers. On day 6 and the morning of day 7, rats were exposed to the B chambers in the manner previously described. They were then tested in those chambers with S2 and S1 on days 7 and 8. Tests were conducted in the manner previously described, and the order of testing with S2 and S1 was counterbalanced (for a schematic, see Fig. 6a).
Experiment 3b
As in experiment 1a, rats received S1-shock pairings in stage one, followed by exposure to S2-S1-shock sequences in stage two. However, instead of the 2-d interval between stages one and two in that experiment (and the others), the interval here was two weeks. Briefly, rats were preexposed to the context (days 1–2), conditioned to S1 via its pairings with shock (day 3), extinguished to the context (day 16), exposed to serial-order conditioning (day 17) under a BLA infusion of D-AP5 (n = 15) or vehicle (n = 14), followed by context extinction (day 18 and the morning of day 19) and counterbalanced tests of freezing to S2 and S1 (days 19–20; for a schematic, see Fig. 7a).
Experiment 4a
Rats were conditioned to S1 and then exposed to S2-S1-shock sequences in the manner described for experiment 1a. The difference here was that rats received a BLA infusion of the competitive, nonselective AMPAR antagonist, NBQX (n = 9) or vehicle (0.9% sterile saline, n = 10) before exposure to the S2-S1-shock sequences. Briefly, rats were preexposed to the chambers (days 1–2), conditioned to S1 (day 3), exposed to the chambers to extinguish any freezing (day 4), subjected to serial-order conditioning under a BLA infusion of NBQX or vehicle (day 5), exposed to the chambers to extinguish any context freezing (day 6 and the morning of day 7), and finally, tested in a counterbalanced manner with S2 and S1 (days 7 and 8; for a schematic, see Fig. 8a).
Experiment 4b
Rats were conditioned to S1 and then exposed to the S2-S1-shock sequences in the manner described for experiment 1a. The difference here was that rats received a BLA infusion of the calcium-permeable AMPAR (CP-AMPAR) antagonist, NASPM (n = 9) or the vehicle control (PBS: pH 7.2, 0.1 m; n = 11) before exposure to the S2-S1-shock sequences. Briefly, rats were preexposed to the chambers (days 1–2), conditioned to S1 (day 3), and re-exposed to the chambers to extinguish any freezing (day 4). They next received the S2-S1-shock sequences under a BLA infusion of NASPM or vehicle (day 5), extinction exposures to the chambers (day 6 and the morning of day 7), and counterbalanced tests of S2 and S1 (afternoons of days 7 and 8; for a schematic, see Fig. 9a).
Experiment 4c
This experiment differed from experiment 4b only in that the S1-shock training in stage one was omitted. Briefly, rats were preexposed to the chambers (days 1–2), and the next day received a BLA infusion of NASPM (n = 10) or the vehicle control (PBS: pH 7.2, 0.1 m; n = 13) before exposure to the S2-S1-shock sequences (day 3). They next received extinction exposures to the chambers (day 4 and the morning of day 5), followed by counterbalanced tests of S2 and S1 (afternoons of days 5 and 6; for a schematic, see Fig. 10a).
Experiment 5, electrophysiology
Coronal brain sections were prepared from 20- to 30-d-old Long–Evans rats. Rats were anaesthetized (isoflurane) and decapitated, their brains removed and transferred into a frozen slurry of high magnesium/low calcium ACSF solution containing (in mm): 118 NaCl, 25 NaHCO3, 10 glucose, 0.5 CaCl2, 1.2 NaH2PO4, and 2.6 MgCl2; 350-µm-thick coronal sections were then prepared using a Leica VT1000S vibratome and transferred to a holding chamber containing the same ACSF with 0.3 mm kynurenic acid at 32.5°C for 30 min, then maintained for several hours at room temperature. All procedures were performed with the approval of the Institutional Animal Ethics Committees of Charles Sturt University and conducted in accordance with the approved guidelines.
Whole-cell voltage clamp recordings were made using a patch clamp amplifier (Multiclamp 700A, Axon Instruments). During recordings individual brain slices were maintained at 32–33°C in a recording chamber continuously perfused with oxygenated ACSF (in mm; 118 NaCl, 25 NaHCO3, 10 glucose, 2.5 CaCl2, 1.2 NaH2PO4, and 1.3 MgCl2) and visualized using IR/DIC techniques. Patch electrodes (3–5 MΩ) were filled with pipette solution containing (mm): 135 CsMeSO4, 8 NaCl, 10 HEPES, 2 Mg2ATP, and 0.3 Na3GTP (pH 7.2 with CsOH, osmolarity 290 mOsm/kg). Current signals were filtered at 4–8 kHz and digitized at 20 kHz (National Instruments, USB-6221 digitizer), acquired, stored, and analyzed on Toshiba Satellite Pro L70 PC using Axograph software. Access resistance (5–15 MΩ) was monitored throughout the experiment.
During the experiments, drugs were added to the ACSF solution used to perfuse the slice, including NASPM (Tocris), NBQX, and bicuculline (Abcam). Postsynaptic currents were evoked using a bipolar stimulating electrode rotated perpendicular to the slice with a single electrode tip placed onto the surface of the slice. Responses shown are averages of 10–50 individual trials. Student's t tests were used for statistical comparisons between groups (except where indicated). All results are expressed as mean ± SEM.
Results
Experiments 1a and 1b
Experiments 1a and 1b constituted a first test of our hypothesis that the role of BLA NMDAR in Pavlovian fear conditioning varies with the predictability of the aversive US. In both experiments, rats were exposed to S2-S1-shock sequences under a BLA infusion of D-AP5 or vehicle. The experiments differed in the training that rats had previously received. In experiment 1a, rats had been previously exposed to S1-shock pairings: thus, the shock was predicted by S1 during the S2-S1-shock sequences. In experiment 1b, rats received context exposure before serial-order conditioning, meaning that the shock occurred unexpectedly during S2-S1-shock sequences (for schematics, see Figs. 2a, 3a).
Experiment 1a: activation of NMDAR in the BLA is not required for conditioning to S2 when it is shocked in compound with an already-conditioned S1. a, Experimental timeline of conditioning and tests of freezing to S2 and S1 for rats injected with vehicle (n = 8) or D-AP5 (n = 9) in stage two. Percentage freezing to (b) presentations of S1 across its pairings with shock in stage one and (c) presentations of S2 and S1across S2-S1-shock sequences in stage two. Percentage freezing during (d) counterbalanced tests of freezing to S2 and S1 and (e) to presentations of S2 across its pairings with shock and test presentations of S2 as a first-order CS. Closed squares represent rats treated with vehicle. Open squares represent rats treated with D-AP5. Data are mean ± SEM; *p < 0.05.
Experiment 1b: activation of NMDAR in the BLA is required for conditioning to S2 when it is shocked in compound with a novel S1. a, Experimental timeline of conditioning and tests of freezing to S2 and S1 for rats that received a BLA infusion of D-AP5 or vehicle before serial-order conditioning. Percentage freezing to (b) presentations of S2 and S1 across serial-order conditioning and (c) across counterbalanced tests of freezing to S2 and S1. Following extinction of freezing to S2 and S1, rats were exposed to S1-shock pairings followed by S2-S1-shock sequences under a BLA infusion of D-AP5 (n = 11) or vehicle (n = 11). Re-acquisition of freezing to S2 and the maintenance of freezing to S1 across serial-order conditioning were unaffected by the BLA infusion of D-AP5. The remaining panels of the figure show percentage freezing to (d) presentations of S1 across its pairings with shock during first-order conditioning, (e) presentations of S2 and S1 during serial-order conditioning, and (f) during counterbalanced tests of freezing to S2 and S1. Closed squares represent rats treated with vehicle. Open squares represent rats treated with D-AP5. Data are mean ± SEM; **p < 0.01, ***p < 0.001.
Experiment 1a
Training was successful. Rats in groups vehicle (n = 8) and D-AP5 (n = 9) acquired fear to S1 at similar rates on day 3 (see Fig. 2b). There was a significant main effect of trials (F(1,15) = 38.4, Fc = 4.5, p < 0.001), but no significant effect of group (F(1,15) = 1.7, Fc = 4.5, p = 0.21) or trial × group interaction (F(1,15) = 1.6, p = 0.23), although rats in group D-AP5 exhibited slightly less freezing to S1 than did the rats in group vehicle. Similarly, during S2-S1-shock sequences on day 5, which occurred under a BLA infusion of vehicle or D-AP5, both groups acquired fear to S2 at similar rates and maintained equivalent levels of fear to the already-conditioned S1 (see Fig. 2c). This was evidenced by a main effect of trials for fear to S2 (F(1,15) = 33.7, Fc = 4.5, p < 0.001), no effect of trials for fear to S1 (F(1,15) = 1.9, p = 0.19), and no effects of group (F(1,15) = 0.8, Fc = 4.5, p = 0.39 for S1; F(1,15) = 1.1, p = 0.18 for S2) or trial × group interactions (F(1,15) = 0.004, Fc = 4.5, p = 0.95 for S1; F(1,15) = 0.6, p = 0.46 for S2) for fear to either stimulus.
The critical results of this experiment are those from the tests of S2 and S1 on days 7 and 8 (Fig. 2d). Importantly, both groups froze at similar levels when tested with S2 or S1. The statistical analyses confirmed that there were no main effects of group on either test (F(1,15) = 0.6, p = 0.47 for S1; F(1,15) = 0.2, p = 0.69 for S2). It revealed significant effects of trials (F(1,15) = 4.8, Fc = 4.5, p = 0.045 for S1; F(1,15) = 20.0, p < 0.001 for S2), indicating that freezing declined across the test presentations of S2 and S1, but with no significant trial × group interactions for either S2 (F(1,15) = 0.4, Fc = 4.5, p = 0.56) or S1 (F(1,15) = 3.3, p = 0.09), indicating that freezing to S2 and S1 declined at a similar rate in the two groups.
Experiment 1a, part 2
To confirm the efficacy of the BLA D-AP5 infusion, we sought to replicate its well-established effect on first-order Pavlovian fear conditioning (Rodrigues et al., 2001; Bauer et al., 2002). Therefore, freezing to S2 was completely extinguished and then rats were exposed to S2-shock pairings under a BLA infusion of vehicle or D-AP5 (see Fig. 2a). Performance during S2-shock pairings was complicated by the emergence of escape behaviors, but the subsequent S2 alone test revealed a clear effect of NMDAR blockade on fear conditioning of S2 (see Fig. 2e). Specifically, the analysis of these data confirmed significant main effects of group (F(1,14) = 8.6, Fc = 4.5, p = 0.011, d = 1.5, 95% CI = 8.0, 51.7) and trials (F(1,14) = 6.3, p = 0.025), but no group × trials interaction (F(1,14) = 1.5, p = 0.24). Thus, NMDAR activation in the BLA is required for conditioning of S2 when it is paired with an unexpected US; thereby highlighting the significance of the finding that NMDAR activation in the BLA is not required for conditioning of S2 when it is paired with an expected US.
Experiment 1b
This experiment provided a further test of the hypothesis that the role of BLA NMDAR in Pavlovian fear conditioning varies with the predictability of the shock US. It did so by omitting the S1-shock training before serial-order conditioning on day 3, which again occurred under a BLA infusion of vehicle (n = 11) or D-AP5 (n = 11; see Fig. 3a). This infusion impaired the acquisition of freezing to S2 and S1 (see Fig. 3b): rats in group vehicle froze more than those in group D-AP5 to both stimuli across the S2-S1-shock sequences. This was confirmed in the analyses of freezing to S2 and S1 which revealed significant main effects of group (F(1,20) = 35.1, Fc = 4.4, p < 0.001 for S1; F(1,20) = 20.2, Fc = 4.4, p < 0.001 for S2) and trials F(1,20) = 91.1, Fc = 4.4, p < 0.001 for S1; F(1,20) = 115.4, Fc = 4.4, p < 0.001 for S2), as well as significant trial × group interactions F(1,20) = 8.2, Fc = 4.4, p = 0.01 for S1; F(1,20) = 8.6, Fc = 4.4, p = 0.008 for S2).
The tests on days 5 and 6 (see Fig. 3c) confirmed that rats in group vehicle again froze more than those in group D-AP5 to S2 and S1. This was supported by the separate analyses of freezing to S2 and S1, which revealed significant main effects of group (F(1,20) = 11.4, Fc = 4.4, p = 0.003, d = 1.5, 95% CI = 9.7, 41.2 for S2; F(1,20) = 37.9, p < 0.001, d = 3.0, 95% CI = 34.0, 68.8 for S1). The analysis of freezing to S2 additionally revealed a significant effect of trials (F(1,20) = 8.0, Fc = 4.4, p = 0.01) and a significant trial × group interaction (F(1,20) = 11.2, p = 0.003), indicating that freezing to S2 remained low in group D-AP5 and extinguished across test presentations in group vehicle. There was no effect of trials (F(1,20) = 3.2, p = 0.09) or trial × group interaction (F(1,20) = 1.4, p = 0.25) in the analysis of freezing to S1, indicating that the difference between the two groups persisted across the entire test session.
Experiment 1b, part 2
These results show that, in rats with no prior conditioning history, BLA infusions of D-AP5 impair fear conditioning to both the S2 and S1 elements of a serial-order compound. They differ from those of experiment 1a where the BLA D-AP5 infusions failed to affect fear conditioning of S2 (or S1) in rats that had been previously exposed to S1-shock pairings. The second part of experiment 1b sought to replicate the main finding from the previous experiment. Following testing, residual freezing to S2 and S1 was extinguished, freezing to S1 was reconditioned through its pairing with foot shock, and any freezing elicited by the chambers was extinguished. Rats were then reallocated to group vehicle (n = 11) and group D-AP5 (n = 11) such that each contained equal numbers (within one) of rats that had previously received BLA infusions of vehicle or drug. The two groups were also matched for freezing to S1 across its pairings with foot shock. Rats in these new groups then received BLA infusions of vehicle or D-AP5 before serial-order conditioning, exposure to the chambers to extinguish any freezing, and counterbalanced tests with S2 and S1 (for a schematic, see Fig. 3a).
Figure 3d shows levels of freezing to S1 across its pairings with foot shock (day 11). The statistical analysis confirmed that both groups acquired freezing to S1 at similar rates, as there was a significant main effect of trials (F(1,20) = 213.2, Fc = 4.4, p < 0.001), but no significant main effect of group (F(1,20) = 0.2, p = 0.71) or trial × group interaction (F(1,20) = 0.9, p = 0.35). The acquisition of freezing to S2 during serial-order conditioning was also successful and unaffected by the BLA D-AP5 infusion (see Fig. 3e). The analysis of this freezing revealed a significant effect of trials (F(1,20) = 20.1, Fc = 4.4, p < 0.001), but no effect of group (F(1,20) = 0.02, p = 0.90) or trial × group interaction (F(1,20) = 0.07, p = 0.79). Further, the analysis of freezing to S1 during serial-order conditioning found no effects of trials F(1,20) = 0.3, p = 0.61) or group (F(1,20) = 0.02, p = 0.90) but a significant trial × group interaction (F(1,20) = 5.2, Fc = 4.4, p = 0.034): this was because of variability in levels of fear to S1 across trials by rats in group vehicle (see Fig. 3e).
The critical results from this stage of the experiment are those from the tests of S2 and S1 on days 15 and 16 (Fig. 3d). Inspection of the figure suggests that both groups froze at similar levels to S2 and S1, with freezing to S2 declining across the session and freezing to S1 remaining high throughout the test. These impressions were confirmed by the statistical analysis, which revealed no main effects of group (F(1,20) = 0.5, p = 0.50 for S2; F(1,20) = 0.6, p = 0.46 for S1) or trial × group interactions (F(1,20) = 0.4, p = 0.53 for S2; F(1,20) = 0.9, p = 0.37 for S1) for either S2 or S1. There was a significant effect of trials for freezing to S2 (F(1,20) = 15.0, Fc = 4.4, p = 0.001) but not for freezing to S1 (F(1,20) = 1.1, p = 0.31), indicating that freezing to S2 declined across the session but that freezing to S1 remained high throughout. Thus, these results replicate the main finding in experiment 1a: when rats are exposed to S2-S1-shock sequences under a BLA infusion of D-AP5, the acquisition of fear to S2 is spared among rats that have been previously conditioned to fear S1 (i.e., previously exposed to S1-shock pairings). More generally, the results of experiments 1a and 1b show that, when rats are exposed to S2-S1-shock sequences, the involvement of BLA NMDARs in fear conditioning to S2 is regulated by what has been learned about S1. The activation of these receptors is necessary for conditioning to S2 when S1 has not been previously conditioned, but is not necessary for conditioning to S2 when S1 has been previously conditioned.
There are two further points to note in relation to experiments 1a and 1b. First, the results of these experiments appear to conflict with those of a prior study which claimed that activation of NMDAR in the amygdala is necessary for context fear conditioning in animals that had been previously conditioned to fear a different context (Lee and Kim, 1998). Specifically, given our finding that rats exposed to S1-shock pairings in stage 1 and S2-S1-shock sequences in stage 2 acquire fear to S2 independently of NMDAR in the BLA, one might have expected conditioning of the second context in the Lee and Kim (1998) study to have also occurred independently of NMDAR in the BLA (but see Finnie et al., 2018; Hardt et al., 2009; Sanders and Fanselow, 2003; Wang et al., 2012). One explanation for these apparently discrepant findings is that, in experiments 1a and 1b, the S2-S1-shock sequences were conducted under a BLA infusion of the NMDAR antagonist, D-AP5, which blocks NMDAR but spares neuronal activity (for discussion, see Bauer et al., 2002; Matus-Amat et al., 2007; Walker and Davis, 2008), whereas in the Lee and Kim study, conditioning to the second context was conducted under a BLA infusion of the NMDAR antagonist, DL-AP5, which blocks both NMDAR and neuronal activity in the BLA (for discussion, see Chapman and Bellavance, 1992; Matus-Amat et al., 2007; Walker and Davis, 2008). As such, we take the present findings to show that NMDAR activity in the BLA is not required for acquisition of fear to S2 among rats exposed to S1-shock pairings in stage 1 and S2-S1-shock sequences in stage 2, and the Lee and Kim findings to show that neuronal activity in the BLA is required for context fear conditioning among animals that have been previously conditioned to fear a different context. The involvement of BLA NMDAR in conditioning under these circumstances remains to be established.
The second point to note in relation to experiment 1a is that, in part two of this experiment, the rats used to confirm the efficacy of the BLA D-AP5 infusion were not naive: they had been used to demonstrate the NMDAR independence of conditioning to S2 in part one, and arrived in part two having been extensively extinguished to both S2 and S1. Accordingly, the finding that conditioning to S2 in this part of the experiment (achieved via S2-shock pairings) required NMDAR activation in the BLA reinforces the point that these receptors are required for Pavlovian fear conditioning across a broad range of circumstances, i.e., NMDAR activation is required for the acquisition of fear to a CS regardless of whether it is novel (well established in the literature) or familiar (Laurent et al., 2008), regardless of whether it is paired with an innate or learned source of danger (Parkes and Westbrook, 2010), and regardless of whether it has been previously conditioned and extinguished (i.e., activation of NMDAR is required for reacquisition of fear to an extinguished CS; Laurent et al., 2008). Moreover, the results from part two of experiment 1a replicate the disruption of reconditioning by a BLA infusion of D-AP5; and the results from part two of experiment 1b replicate the results from part one of experiment 1a (no disruption of serial-order conditioning to S2 when S1 has been pretrained). Thus, while the reuse of animals may sometimes produce confusing or “noisy” results, this was not the case in part two of either experiment. Instead, the results confirm that: (1) reconditioning of a Pavlovian CS requires activation of NMDAR in the BLA and, hence, that reinstatement of the memory produced by the initial CS-US pairings does not permit conditioning to occur independently of these receptors (experiment 1a); and (2) Pavlovian conditioning can occur independently of NMDAR activation in the BLA when S2 is shocked in sequence with an already-conditioned S1 (experiment 1b).
Experiments 2a and 2b
Experiments 2a and 2b examined how the prior conditioning of S1 influenced the NMDAR activation requirement for conditioning to S2. In both experiments, rats were exposed to S1-shock pairings in stage one. The two experiments differed from the previous ones in the conditioning of S2 in stage two. In experiment 2a, S2 was conditioned using a trace protocol: each S2 presentation was followed by a 10-s trace period that co-terminated with delivery of foot shock. In experiment 2b, S2 was conditioned using a second-order protocol in which each S2 presentation terminated in the onset of the already conditioned S1 but the shock was omitted. In both experiments, the conditioning to S2 occurred under an infusion of vehicle or D-AP5 into the BLA. Finally, rats were tested with S2 and S1 (for a schematic, see Fig. 4a). If, in the previous experiments, conditioning of S1 per se rendered subsequent conditioning of S2 independent of BLA NMDAR, then conditioning of S2 in the current experiments should again occur independently of NMDAR activation because rats are conditioned to S1 before they enter the trace and second-order protocols. Alternatively, if conditioning to S2 in the previous experiments occurred independently of BLA NMDAR because S1 already signaled the shock across the S2-S1-shock sequences, then conditioning to S2 in the current experiments will require NMDAR activation because S1 is omitted in the trace protocol (experiment 2a) and the shock predicted by S1 is omitted in the second-order protocol (experiment 2b).
Experiment 2a: trace fear conditioning to S2 requires NMDAR activation in the BLA. a, Experimental timeline of conditioning and tests of freezing to S2 and S1 for rats given a BLA infusion of D-AP5 (n = 8) or vehicle (n = 8) before trace conditioning of S2. b, Percentage freezing to presentations of S1 during first-order conditioning on day 3 (left), and to presentations of S2 during trace conditioning on day 5 (right). c, Percentage freezing to test presentations of S2 on day 7 (left), and S1 on day 8 (right). Closed squares represent rats treated with vehicle. Open squares represent rats treated with D-AP5. Data are mean ± SEM. N.S., nonsignificant; *p < 0.05.
Experiment 2a
Conditioning was successful (Fig. 4b). In stage one, freezing to S1 increased linearly across the four S1-shock pairings (F(1,14) = 81.3, Fc = 4.6, p < 0.001), and did so at the same rate among rats in groups vehicle (n = 8) and D-AP5 (n = 8; F(1,14) = 0.3, p = 0.62 for main effect of group, and F(1,14) = 0.03, p = 0.86 for trial × group interaction). In stage two, there was no acute effect of the BLA D-AP5 infusion on acquisition of trace conditioned fear: freezing to S2 increased linearly across the four S2-shock (trace) pairings (F(1,14) = 76.0, Fc = 4.6, p < 0.001), and did so at the same rate in groups vehicle and D-AP5 (F(1,14) = 2.9, p = 0.11 for main effect of group, and F(1,14) = 0.2, p = 0.69 for trial × group interaction).
The critical results of this experiment are those from the test of S2 on day 7 (Fig. 4c, left panel) and S1 on day 8 (Fig. 4c, right panel). Rats in group vehicle froze more than those in group D-AP5 when tested with S2; but rats in both groups froze at the same level when tested with S1. These impressions were confirmed by the statistical analyses. The analysis of freezing to S2 revealed a significant main effect of group (F(1,14) = 5.3, Fc = 4.6, p = 0.037, d = 1.1, 95% CI = 1.4, 39.6), but no effect of trials (F(1,14) = 4.0, p = 0.07) or trial × group interaction (F(1,14) = 0.1, p = 0.83). The analysis of freezing to S1 showed that freezing declined across the testing (F(1,14) = 5.4, Fc = 4.6, p = 0.036), but there was no main effect of group (F(1,14) = 0.01, p = 0.95) or trial × group interaction (F(1,14) = 4.0, p = 0.07). Hence, NMDAR activation in the BLA is necessary for trace fear conditioning to S2 among animals that had been previously conditioned to fear S1.
Experiment 2b
Conditioning was again successful. In stage one (Fig. 5b), freezing to S1 increased linearly across the four S1-shock pairings (F(1,18) = 135.0, Fc = 4.4, p < 0.001), and did so at the same rate among rats in groups vehicle (n = 9) and D-AP5 (n = 11; F(1,18) = 0.4, p = 0.53 for main effect of group, and F(1,18) = 0.1, p = 0.79 for trial × group interaction). In stage two, there was no acute effect of the BLA D-AP5 infusion on acquisition of second-order conditioned fear to S2 (Fig. 5c): freezing to S2 increased linearly across the four S2-S1 pairings (F(1,18) = 11.6, Fc = 4.4, p = 0.003), and did so at the same rate in groups vehicle and D-AP5 (F(1,18) = 0.9, p = 0.35 for main effect of group, and F(1,18) = 1.0, p = 0.33 for trial × group interaction). There was, however, an unexpected effect of the D-AP5 infusion on maintenance of fear to S1 (Fig. 5d): it reduced freezing to the already-conditioned S1 on trials 1–2 but not across trials 3–4, as evidenced by a significant trial × group interaction (F(1,18) = 11.8, Fc = 4.4, p = 0.003). The main effect of trials (F(1,18) = 11.1, Fc = 4.4, p = 0.004) was also significant, indicating that freezing to S1 in group D-AP5 returned to the same level as that displayed by vehicles on trials 3–4. Importantly, the overall level of freezing to S1 did not differ between the two groups (F(1,18) = 1.1, p = 0.31).
Experiment 2b: second-order fear conditioning to S2 requires activation of NMDAR in the BLA. a, Experimental timeline of conditioning and tests of freezing to S2 and S1 for rats treated with vehicle (n = 9) or D-AP5 (n = 11) before second-order conditioning. Percentage freezing to presentations of (b) S1 across its pairings with shock in stage one, and presentations of (c) S2 and (d) S1 across S2-S1 pairings in stage two. Percentage freezing during tests of freezing to (e) S2 and (f) S1. Closed and open squares represent rats treated with vehicle and D-AP5, respectively. Data are mean ± SEM; ***p < 0.001.
The critical results of this experiment are those from the test of S2 on day 6 (Fig. 5e) and S1 on day 7 (Fig. 5f). Rats in group vehicle froze more than those in group D-AP5 when tested with S2; but rats in both groups froze at the same level when tested with S1. These impressions were confirmed by the statistical analyses. The analysis of freezing to S2 revealed a significant main effect of group (F(1,18) = 22.5, Fc = 4.4, p < 0.001, d = 2.3, 95% CI = 7.5, 19.3). The main effect of trials (F(1,18) = 12.8, Fc = 4.4, p = 0.002) and trial × group interaction were also significant (F(1,18) = 8.5, Fc = 4.4, p = 0.009, 95% CI = 5.0, 31.0), reflecting the fact that group vehicle displayed lower levels of freezing across the initial S2 alone presentations than the more substantial levels across the later presentations, whereas group D-AP5 froze at relatively low levels across all trials. The analysis of freezing to S1 confirmed that the main effects of group (F(1,18) = 0.1, p = 0.78) and trials (F(1,18) = 4.3, Fc = 4.4, p = 0.053) were not significant. There was, however, a significant trial × group interaction (F(1,18) = 5.1, Fc = 4.4, p = 0.037, 95% CI = −42.8, −1.5), indicating that group D-AP5 froze less than group vehicle across the initial S1 presentations, but more than group vehicle across the later ones. Hence, NMDAR activation in the BLA is necessary for the acquisition of second-order fear to S2 but not for the maintenance of first-order conditioned fear to S1 (see also Gewirtz and Davis, 1997; Parkes and Westbrook, 2010; Holmes et al., 2013).
Experiments 2a and 2b found that, in rats that have already received S1-shock pairings, NMDAR activation in the BLA is required for both trace and second-order conditioning to S2. These findings add to those of the previous experiments in showing that the NMDAR activation requirement for conditioning to S2 is not determined by the prior training of S1 per se (experiment 2a), the presence of the pretrained S1 during conditioning to S2 (experiment 2b) or the interval between the S2 and shock (experiment 2a); this interval was equivalent in the trace protocol used in experiment 2a and the serial-order protocol used in experiments 1a and 1b, but the BLA infusion of D-AP5 disrupted conditioning to S2 in the former case but not the latter. Rather, the NMDAR activation requirement for conditioning to S2 appears to be determined by surprise or error in relation to the shock US. Activation of BLA NMDAR is necessary for conditioning to S2 when the occurrence of shock is not predicted by S1 (experiment 1b) or when the shock predicted by S1 does not occur (experiment 2b). Activation of BLA NMDAR is not necessary for conditioning to S2 when predictions about the shock are correct, as is the case when the S2 element of the serial-order compound contains an already-conditioned S1 (experiments 1a and 1b).
It is worth noting that the BLA infusion of D-AP5 disrupted freezing to the already conditioned S1 in stage 2 of experiment 2b (second-order conditioning; S2-S1 pairings in Fig. 5d) but not in stage 2 of experiment 1a (serial-order conditioning; S2-S1-shock sequences in Fig. 2c). This is despite the fact that S1 had been conditioned identically in stage 1 of each experiment and was presented under identical conditions on the first trial in stage 2. The exact reason for this difference is not clear; however, the acute D-AP5 effect in experiment 2b also contrasts with the absence of any such effect in the next two experiments [3a (Fig. 6d) and 3b (Fig. 7c)], and in part two of experiment 1b (Fig. 3e, although the drug free tests in each of these experiments revealed a clear effect of the D-AP5 infusion on acquisition of freezing to S2). As such, the acute disruption of S1 in experiment 2b appears to be a spurious finding; and reinforces our general approach of placing greater emphasis on the findings from the drug-free tests of freezing to S2 and S1 in each experiment.
Experiment 3a: a context shift between S1-shock pairings in stage one and S2-S1-shock sequences in stage two restores the NMDAR activation requirement for conditioning to S2. a, Experimental timeline of conditioning and tests of freezing for rats treated with D-AP5 or vehicle before serial-order conditioning. b, Percentage freezing to presentations of S1 during first-order conditioning on day 3. c, Percentage freezing in context A versus context B for group vehicle (n = 8) and group D-AP5 (n = 8) during context discrimination tests on day 4. d, Percentage freezing to presentations of S2 (left panel) and S1 (right panel) across serial-order conditioning on day 5. Percentage freezing to test presentations of (e) S2 and (f) S1 during counterbalanced tests on days 7–8. Closed squares represent rats treated with vehicle. Open squares represent rats treated with D-AP5. Data are mean ± SEM; **p < 0.01.
Experiment 3b: an increase in the interval between S1-shock pairings in stage one and S2-S1-shock sequences in stage two restores the NMDAR activation requirement for conditioning to S2. a, Experimental timeline of conditioning and tests of freezing for rats treated with D-AP5 or vehicle before serial-order conditioning. b, Percentage freezing to presentations of S1 during first-order conditioning on day 3. c, Percentage freezing to presentations of S2 (left panel) and S1 (right panel) across serial-order conditioning on day 17. d, Percentage freezing to test presentations of S2 (left panel) and S1 (right panel) during counterbalanced tests on days 19–20. Closed squares represent rats treated with vehicle. Open squares represent rats treated with D-AP5. Data are mean ± SEM; *p < 0.05.
Experiments 3a and 3b
The previous experiments suggest that, when S2 is shocked in compound with an already-conditioned S1, there is little or no error associated with the occurrence of the shock; hence, S2 is conditioned independently of NMDAR activation in the BLA. Experiments 3a and 3b examined whether the presence of the already-conditioned S1 in the compound is sufficient to render fear to S2 independent of BLA NMDAR. They did so by examining whether a context shift or increase in time between S1-shock pairings in stage one and S2-S1-shock sequences in stage 2 restores this error, and thereby, the requirement for BLA NMDAR activation in conditioning of S2. If the presence of the already conditioned S1 is sufficient, then learning about S2 will not require NMDAR in the BLA; if, however, error is additionally computed on the basis of the context where the S1-shock pairings and the S2-S1-shock sequences occurred and the time that elapsed between those training experiences, then learning about S2 will require NMDAR in the BLA.
In experiment 3a, rats were familiarized with two contexts, labeled A and B (days 1–2), that differed in several ways (floors, location in the laboratory, and smell; B was the standard context used in the previous experiments). They were then exposed to S1-shock pairings in context A on day 3, a context discrimination test on day 4; S2-S1-shock sequences in context B under a BLA infusion of vehicle (n = 8) or D-AP5 (n = 8) on day 5, and finally, counterbalanced tests of S2 and S1 in context B on days 6 and 7 (for a schematic, see Fig. 6a). Experiment 3b used the standard context in stages 1 and 2 and the same design as experiment 1a except that the interval between S1-shock pairings in stage one and S2-S1-shock sequences in stage two was extended to 14 d rather than the 2 d used in the previous experiments. That is, rats were exposed to S1-shock pairings on day 3, S2-S1-shock sequences on day 17 under a BLA infusion of vehicle (n = 14) or D-AP5 (n = 15), and finally, counterbalanced tests of freezing to S2 and S1 on days 19 and 20 (for a schematic, see Fig. 7a).
Experiment 3a
Conditioning was successful. In stage one (Fig. 6b), freezing to S1 increased linearly across the four S1-shock pairings in context A (F(1,14) = 36.5, Fc = 4.6, p < 0.001), and did so at the same rate among rats in groups vehicle (n = 8) and D-AP5 (n = 8; F(1,14) = 0.1, p = 0.72 for main effect of group, and F(1,14) = 1.0, p = 0.33 for trial × group interaction). During the subsequent context extinction/exposure sessions (Fig. 6c), all rats froze more in the context where they had been shocked, A, than the context where they had not been shocked, B (F(1,14) = 12.6, Fc = 4.6, p = 0.003, d = 0.9, 95% CI = 4.2, 17.1), and the level of context discrimination was equivalent in each of the two groups (F(1,14) = 0.1, p = 0.76 for main effect of group, and F(1,14) = 0.03, p = 0.87 for context × group interaction). During S2-S1-shock sequences in the standard context (labeled B) in stage two, there was no acute effect of the BLA D-AP5 infusion on acquisition of fear to S2 (Fig. 6d, left panel): freezing to S2 increased linearly across the four trials (F(1,14) = 27.3, Fc = 4.6, p < 0.001), and did so at the same rate in groups vehicle and D-AP5 (F(1,14) = 2.0, p = 0.18 for main effect of group, and F(1,14) = 3.3, p = 0.09 for context × group interaction). There was also no acute effect of the BLA D-AP5 infusion on the maintenance of freezing to S1 (Fig. 6d, right panel): freezing to S1 did not change across the four trials, and there was no main effect of group (F(1,14) = 1.6, p = 0.23) or trial × group interaction (F(1,14) = 0.2, p = 0.70).
The test levels of freezing to S2 and S1 are shown in Figure 6e,f, respectively. Inspection of the figures suggests that rats in group D-AP5 froze less to S2 than those in group vehicle and that both groups froze at similar levels to S1. These impressions were confirmed by the statistical analyses which found a significant main effect of group for freezing to S2 (F(1,14) = 12.1, Fc = 4.6, p = 0.004, d = 1.7, 95% CI = 6.9, 28.9), but not S1 (F(1,14) = 0.4, p = 0.53). There were no significant effects of trials in either analysis (F(1,14) = 2.8, p = 0.12 for S1, and F(1,14) = 0.4, p = 0.55 for S2) and no trial × group interactions (F(1,14) = 0.7, p = 0.42 for S1, and F(1,14) = 0.3, p = 0.60 for S2). Taken together, these results show that a BLA infusion of D-AP5 disrupts conditioning to S2 when the context is shifted between S1-shock pairings in stage one and S2-S1-shock sequences in stage two; and does so without affecting the levels of freezing to the already conditioned S1. Thus, NMDAR activation in the BLA is required for conditioning to S2 when the physical context where the S2-S1-shock sequences occur is different to the one where the S1-shock pairings had occurred.
Experiment 3b
Conditioning was again successful. In stage one (Fig. 7b), freezing to S1 increased linearly across its four pairings with shock (F(1,27) = 55.3, Fc = 4.2, p < 0.001), and did so at the same rate among rats in groups vehicle (n = 14) and D-AP5 (n = 15; F(1,27) = 0.003, p = 0.96 for main effect of group, and F(1,27) = 0.67, p = 0.42 for trial × group interaction). Two weeks later, in stage two, there was no acute effect of the BLA D-AP5 infusion on acquisition of freezing to S2 across the four S2-S1-shock sequences (Fig. 7c, left panel): freezing to S2 increased linearly across the four shocked sequences (F(1,27) = 86.3, Fc = 4.2, p < 0.001), and did so at the same rate in groups vehicle and D-AP5 (F(1,27) = 0.4, p = 0.53 for main effect of group, and F(1,27) = 2.9, p = 0.10 for trial × group interaction). There was also no acute effect of the BLA D-AP5 infusion on maintenance of freezing to S1 (Fig. 6d, right panel): there was no main effect of group (F(1,27) = 0.1, p = 0.82) or trial × group interaction (F(1,27) = 3.1, p = 0.09). Freezing to S1 did, however, increase across the four trials (F(1,27) = 14.1, Fc = 4.2, p = 0.001), which may reflect some forgetting of the S1-shock association across the 14-d interval between stages one and two.
The critical results of this experiment are those from the counterbalanced tests of S2 and S1 on days 19 and 20 (Fig. 7d, left and right panels, respectively). Inspection of Figure 7d, left panel, suggests that rats in group D-AP5 froze less to presentations of S2 than rats in group vehicle, with freezing in both groups declining across the session. These observations were confirmed by the statistical analysis, which found significant effects of group (F(1,27) = 7.7, Fc = 4.2, p = 0.010, d = 1.3, 95% CI = 6.0, 39.7) and trials (F(1,27) = 6.9, Fc = 4.2, p = 0.014), but no trial × group interaction (F(1,27) = 2.0, p = 0.17). Inspection of the right panel suggests that the two groups exhibited similar levels of freezing to S1: there was no effect of group (F(1,27) = 0.2, p = 0.64) and no trial × group interaction (F(1,27) = 0.01, p = 0.91). There was, however, a significant effect of trials (F(1,27) = 4.9, Fc = 4.2, p = 0.035), indicating that freezing to S1 increased slightly across the test session. Taken together, these results show that a BLA infusion of D-AP5 disrupts conditioning to S2 when a 14-d interval is interpolated between the S1-shock pairings in stage one and S2-S1-shock sequences in stage two; and does so without affecting freezing to the already conditioned S1.
The results of experiments 3a and 3b show that NMDAR activation is required for conditioning to S2 when the context where the S2-S1-shock sequences occurred was different from that where the prior S1-shock pairings had occurred, and when the interval between the S2-S1-shock sequences and the prior S1-shock pairings was increased to 14 d. These results contrast with those of the earlier experiments which showed that conditioning to S2 was independent of NMDAR activation when the S2-S1-shock sequences and S1-shock pairings occurred in the same context 2 d apart (experiments 1a and 1b). We take these findings to mean that a context shift or the lapse of time restores error in relation to the shock, and thereby, the NMDAR activation requirement for conditioning to the S2. When S2 is conditioned in the presence of an S1 that had been conditioned remotely (two weeks earlier) or more recently (2 d), but in a different context, the uncertainty as to what S1 signals engages NMDAR in the BLA for conditioning of S2, but when S2 is conditioned in the presence of an S1 that was conditioned recently (2 d earlier) and in the same context, there is little or no uncertainty as to what S1 signals and conditioning to S2 occurs independently of NMDAR activation in the BLA.
Importantly, in each of these experiments, the BLA infusion of D-AP5 again spared freezing to S1 when rats were exposed to S2-S1-shock sequences, confirming that activation of NMDAR in the BLA is not required for expression of already-conditioned fear (see also experiments 1a and 1b). This appears to conflict with the results of previous studies which claimed a role for NMDAR in fear expression on the basis of results obtained using infusions of the NMDAR antagonist, DL-AP5 (Lee and Kim, 1998; Lee et al., 2001). As noted above, unlike D-AP5 which blocks activation of NMDAR without disrupting neuronal activity in the BLA (for discussion, see Bauer et al., 2002; Matus-Amat et al., 2007; Walker and Davis, 2008), DL-AP5 blocks activation of NMDAR as well as neuronal activity in the BLA (for discussion, see Chapman and Bellavance, 1992; Matus-Amat et al., 2007; Walker and Davis, 2008). As such, previous findings that have been taken to suggest a role for NMDAR in fear expression likely reflect an effect of the DL-AP5 infusion on neuronal activity in the BLA, which is necessary for fear expression; and the present findings which show no effect of the D-AP5 infusion on fear expression likely reflect the fact that this infusion spared neuronal activity in the BLA. That is, when taken together, the results of these studies show that activation of NMDAR in the BLA is not required for expression of already-conditioned fear even when that fear had been conditioned in a different context or two weeks earlier (for consistent findings, see Do-Monte et al., 2015).
Experiments 4a, 4b, and 4c
What are the substrates of NMDAR-independent fear conditioning in the BLA? Experiments 4a and 4b addressed this question by examining whether such conditioning involves activation of CP-AMPAR as: (1) blockade of these receptors in the BLA impairs first-order Pavlovian fear conditioning (Clem and Huganir, 2010; Hong et al., 2013; Ferrara et al., 2019; Torquatto et al., 2019); and (2) changes in neurotransmission at interneurons that express these receptors have been shown to occur independently of NMDAR activation (Mahanty and Sah, 1998; Polepalli et al., 2010, 2020). Finally, experiment 4c examined whether CP-AMPAR in the BLA regulate conditioning to S2 when rats are exposed to S2-S1-shock sequences in the absence of any prior conditioning. This was intended to assess the generality of the role played by these receptors in Pavlovian fear conditioning.
The design of experiments 4a and 4b was identical to that used in experiment 1a (for schematics, see Figs. 8a, 9a): rats were exposed to S1-shock pairings (day 3), S2-S1-shock sequences (day 5), and finally, tests of S2 and S1 (days 7 and 8). In experiment 4a, rats received a BLA infusion of vehicle (n = 10) or the competitive, nonselective AMPAR antagonist, NBQX (n = 9) before the session containing the S2-S1-shock sequences; in experiment 4b, rats received a BLA infusion of vehicle (n = 11) or the selective CP-AMPAR antagonist, NASPM (n = 9) before this session. The design of experiment 4c was identical to that used in experiment 1b (for a schematic, see Fig. 10a): rats were exposed to S2-S1-shock sequences in training (day 3) followed by tests of S2 and S1 (days 5 and 6). The session of S2-S1-shock sequences was preceded by a BLA infusion of vehicle (n = 13) or NASPM (n = 10).
Experiment 4a: activation of AMPAR in the BLA is required for conditioning to S2 when it is shocked in compound with an already-conditioned S1. a, Experimental timeline of conditioning and tests of freezing to S2 and S1 for rats treated with vehicle (n = 10) or NBQX (n = 9) before serial-order conditioning. b, Percentage freezing to presentations of S1 across its pairings with shock in stage one. c, Percentage freezing to presentations of S2 (left panel) and S1 (right panel) across serial-order S2-S1-shock sequences in stage two. d, Percentage freezing to test presentations of S2 (left panel) and S1 (right panel) on days 7–8 collapsed across the counterbalanced order of testing. Closed squares represent rats treated with vehicle. Open squares represent rats treated with NBQX. Data are mean ± SEM. N.S., nonsignificant; **p < 0.01, *p < 0.05.
Experiment 4b: activation of CP-AMPAR in the BLA is required for conditioning to S2 when it is shocked in compound with an already-conditioned S1. a, Experimental timeline of conditioning and tests of freezing to S2 and S1 for rats treated with vehicle (n = 11) or NASPM (n = 9) before serial-order conditioning. b, Percentage freezing to presentations of S1 across its pairings with shock in stage one. c, Percentage freezing to presentations of S2 (left panel) and S1 (right panel) across serial-order S2-S1-shock sequences in stage two. d, Percentage freezing to presentations of S2 (left panel) and S1 (right panel) during counterbalanced tests on days 7–8. Closed squares represent rats treated with vehicle. Open squares represent rats treated with NASPM. Data are mean ± SEM. N.S., nonsignificant; **p < 0.01, *p < 0.05.
Experiment 4a
Conditioning was successful. In stage one (Fig. 8b), freezing to S1 increased linearly across the four S1-shock pairings (F(1,17) = 23.7, Fc = 4.5, p < 0.001), and did so at the same rate among rats in groups vehicle (n = 10) and D-AP5 (n = 9; F(1,17) = 0.3, p = 0.61 for main effect of group, and F(1,17) = 0.1, p = 0.78 for trial × group interaction). During S2-S1-shock sequences in stage two, BLA infusions of NBQX appeared to have impaired acquisition of freezing to S2 (Fig. 8c, left panel) and reduced freezing to the already-conditioned S1 (Fig. 8c, right panel). This was partially confirmed by the statistical analyses. Freezing to S2 increased linearly across the four trials (F(1,17) = 20.9, Fc = 4.5, p < 0.001), and there was a main effect of group (F(1,17) = 11.2, Fc = 4.5, p = 0.004), indicating an acute effect of the NBQX infusion on freezing to S2 (the trial × group interaction was not significant, F(1,17) = 0.4, p = 0.54). In contrast, freezing to S1 did not change across the four trials (F(1,17) = 0.03, p = 0.88), and the main effect of group (F(1,17) = 3.3, p = 0.09) and trial × group interaction (F(1,17) = 0.8, p = 0.40) were not statistically significant.
The critical results of this experiment are those from the counterbalanced tests of S2 and S1 on days 7 and 8 (Fig. 8d, left and right panels, respectively). It is clear that the BLA infusion of NBQX had impaired fear conditioning of S2 but left intact conditioned fear of S1. The statistical analysis confirmed that rats in group NBQX froze significantly less to S2 than those in group vehicle (F(1,17) = 7.8, Fc = 4.5, p = 0.012, d = 1.3, 95% CI = 6.9, 50.2). It also revealed a significant main effect of trials, indicating a decline in freezing across the S2 alone presentations (F(1,17) = 6.5, Fc = 4.5, p = 0.02), but no trial × group interaction (F(1,17) = 1.7, Fc = 4.5, p = 0.21), indicating that the between-group differences between the levels of freezing persisted across test presentations. The analysis of freezing to S1 confirmed that there was no effect of group (F(1,17) = 0.1, p = 0.74), trials (F(1,17) = 2.0, p = 0.18) or trial × group interaction (F(1,17) = 0.4, p = 0.54). These results indicate that AMPAR-mediated signaling is required for conditioning to S2 during S2-S1-shock sequences: disturbances of this signaling impair conditioning to S2 without affecting freezing to the already-conditioned S1.
Experiment 4b
Conditioning was successful. In stage one (Fig. 9b), freezing to S1 increased linearly across the four S1-shock pairings (F(1,18) = 42.8, Fc = 4.4, p < 0.001), and did so at the same rate among rats in groups vehicle (n = 11) and NASPM (n = 9; F(1,18) = 0.2, p = 0.66 for main effect of group, and F(1,18) = 0.08, p = 0.78 for trial × group interaction). During S2-S1-shock sequences in stage two, the BLA infusions of NASPM appeared to have impaired the development of freezing to S2 and reduced freezing to the already-conditioned S1. The statistical analysis of the S2 data confirmed a significant main effects of group (F(1,18) = 10.6, Fc = 4.4, p = 0.004) and trials (F(1,18) = 20.9, Fc = 4.4, p < 0.001) but no trial × group interaction (F(1,18) = 0.1, p = 0.74). With respect to S1, the analysis found a significant main effect of group (F(1,18) = 4.9, Fc = 4.4, p = 0.04), but no main effect of trials (F(1,18) = 1.1, p = 0.31) or trial × group interaction (F(1,18) = 0.3, p = 0.57).
The critical results are those from the counterbalanced tests of S2 and S1 on days 7 and 8 (Fig. 9d, left and right panels, respectively). The statistical analysis confirmed what is clear from inspection of the figure: rats in group vehicle froze significantly more to S2 than those in group NASPM (F(1,18) = 7.2, Fc = 4.4, p = 0.015, d = 1.2, 95% CI = 5.1, 41.7). There was no effect of trials (F(1,18) = 0.2, p = 0.70) or trial × group interaction (F(1,18) = 1.9, p = 0.18), confirming that the between-group differences in freezing were present across the test presentations. Inspection of the figure suggests that group vehicle also froze more than group NASPM to test presentations of S1. However, this difference was not statistically reliable, as the analysis found no significant effects of trials (F(1,18) = 0.1, p = 0.74), group (F(1,18) = 2.5, p = 0.13), or trial × group interaction (F(1,18) = 0.004, p = 0.95). Hence, activation of CP-AMPAR is necessary for conditioning to S2 during S2-S1-shock sequences: blockade of these receptors impairs conditioning to S2 but leaves the already-conditioned S1 intact.
Experiment 4c
Figure 10b shows levels of freezing to presentations of S2 (left panel) and S1 (right panel) across the four S2-S1-shock sequences. In contrast to its acute effect in the previous experiment, the BLA infusion of NASPM did not affect freezing to S2 or S1. The analysis of freezing to S2 showed a significant main effect of trials (F(1,21) = 193.1, Fc = 4.3, p < 0.001, but no effect of group (F(1,21) = 0.7, p = 0.40) or trial × group interaction (F(1,21) = 0.2, p = 0.63). Similarly, the analysis of freezing to S1 showed a significant main effect of trials (F(1,21) = 508.1, Fc = 4.3, p < 0.001), but no main effect of group (F(1,21) = 0.5, p = 0.51) or trial × group interaction (F(1,21) = 0.4, p = 0.53).
Experiment 4c: activation of CP-AMPAR in the BLA is required for conditioning to S2 when it is shocked in compound with a novel S1. a, Experimental timeline of conditioning and tests of freezing to S2 and S1 for rats treated with vehicle (n = 13) or NASPM (n = 10) before serial-order conditioning. b, Percentage freezing to presentations of S2 (left panel) and S1 (right panel) across serial-order S2-S1-shock sequences on day 3. c, Percentage freezing to presentations of S2 (left panel) and S1 (right panel) during counterbalanced tests on days 5–6. Closed squares represent rats treated with vehicle. Open squares represent rats treated with NASPM. Data are mean ± SEM; ***p < 0.001.
The critical results are those from the counterbalanced tests of S2 and S1 on days 5 and 6 (Fig. 10c, left and right panels, respectively). The statistical analyses confirmed what is clear from inspection of the figure. The analysis of freezing to S2 showed that rats in group vehicle froze significantly more to S2 than those in group NASPM (F(1,21) = 14.8, Fc = 4.3, p < 0.001, d = 1.6, 95% CI = 14.3, 47.9). There was no effect of trials (F(1,21) = 2.1, p = 0.17) or trial × group interaction (F(1,21) = 0.4, p = 0.56). Similarly, the analysis of freezing to S1 showed a significant main effect of group (F(1,21) = 16.0, Fc = 4.3, p = 0.001, d = 2.8, 95% CI = 16.5, 52.4), but no effect of trials (F(1,21) = 0.001, p = 0.98) or trial × group interaction (F(1,21) = 1.3, p = 0.28). Hence, activation of CP-AMPAR is necessary for conditioning to S2 during S2-S1-shock sequences, even among animals that have not been previously conditioned S1. In fact, blockade of these receptors impairs conditioning to both S2 and S1.
Taken together, the results of experiments 4a, 4b, and 4c show that activation of CP-AMPAR in the BLA is required for conditioning to S2 when it is shocked in compound with either a novel (experiment 4c) or previously conditioned S1 (experiment 4b). More generally, they show that activation of CP-AMPAR contributes to both NMDAR-dependent and NMDAR-independent forms of Pavlovian fear learning in the BLA. When S2 is shocked in compound with an S1 that is novel, conditioning to S2 requires activation of both NMDAR and CP-AMPAR in the BLA. When S2 is shocked in compound with an already conditioned S1, conditioning to S2 only requires activation of CP-AMPAR in the BLA.
A note on between-group differences in freezing during acquisition
In the previous experiments, between-group differences in freezing on trial 1 of acquisition were not significant; and these numerical differences were unique to experiments that used infusions of NBQX and NASPM to probe the involvement of AMPAR in Pavlovian fear conditioning. There were no between-group differences on trial 1 of fear conditioning to S2 in experiments that used infusions of D-AP5 to probe the involvement of NMDAR. Beyond trial 1, between-group differences in acquisition, where they occurred, are perfectly consistent with the disruptions that we discuss in relation to the test data. However, here and elsewhere, we take the test data to be a more reliable index of what the animals learned about the CSs across acquisition, as test sessions were conducted subsequent to extinction of context freezing which often masks deficits in conditioning to auditory and visual CSs across training sessions. These test data are reliable in showing that BLA infusions of D-AP5 disrupted fear conditioning to S2 in every scenario examined except for one: when the US was signaled by an S1 that had been pretrained just 2 d earlier in the same context (experiments 1a and 1b). It is also worth noting that BLA infusions of NBQX and NASPM disrupted fear conditioning to S2 in every scenario examined without exception.
Experiment 5
The previous experiments have shown that, when S2 is shocked in compound with an already-conditioned S1, conditioning to S2 requires activation of CP-AMPAR but not NMDAR in the BLA (experiments 1a, 1b, and 4b). Within the lateral amygdala (LA) region of the BLA, CP-AMPAR are preferentially expressed on inhibitory interneurons and have been shown to regulate changes in neurotransmission independently of NMDAR activation (Mahanty and Sah, 1998; Polepalli et al., 2010, 2020). Accordingly, we used electrophysiology to assess the effect of the CP-AMPAR antagonist, NASPM, on the activity of inhibitory interneurons in the LA. We hypothesized that the effect of NASPM on conditioning to S2 in experiment 4b was because of its effect on inhibitory interneurons; and hence, predicted that NASPM would preferentially modulate the activity of these interneurons relative to their effect on excitatory principal neurons.
We first assessed the effects of NASPM on the frequency and amplitude of spontaneous IPSCs (sIPSCs) recorded from LA principal neurons. These currents represent mixed input from the population of local BLA interneurons onto the principal neurons. We found a reduction in frequency and amplitude following application of NASPM (100 μm) indicating a reduction in inhibition of the principal neurons: average sIPSC frequency decreased from 8.21 ± 1.29 to 5.36 ± 0.95 Hz, and amplitude decreased from 22.66 to 18.82 pA (frequency 34.0 ± 5.6% reduction, amplitude 16.8 ± 2.7%; n = 10, p < 0.01 for both; Fig. 11a,b). It has been shown previously that a proportion of BLA interneurons express CP-AMPAR at cortical excitatory synapses (Mahanty and Sah, 1998). To test whether the NASPM effect on inhibition resulted from blocking cortical activation of these interneurons, we applied the NASPM while recording the disynaptic evoked IPSCs (eIPSCs) from principal cells. To activate these interneurons, we electrically stimulated cortical axons in the external capsule superior to the BLA (Fig. 11c). We found that NASPM reduced the amplitude of evoked postsynaptic inhibitory currents by an average of 55.5 ± 10.3% (n = 8, p < 0.01; Fig. 11d,f). On washout of NASPM, these currents were also blocked completely by 10 μm bicuculline, indicating that these were GABAergic currents; and by the nonspecific AMPAR antagonist NBQX (10 μm) following washout of the bicuculline. This indicated that these evoked currents were the result of synaptic activation of local BLA interneurons. In contrast, eIPSC resulting from low intensity direct stimulation of interneurons within the BLA in the presence of 10 μm NBQX (Fig. 11c) were unaffected by NASPM (1.5 ± 5.1% inhibition, n = 7, p = 0.43; Fig. 11e,f). Together, these results indicate that NASPM blocks cortical excitation of a proportion of BLA interneurons, resulting in reduced activation of these cells and, thereby, reduced inhibition of BLA principal neurons. As a further control, we tested the effects of NASPM on evoked EPSCs recorded from the LA principal neurons as these cells are known to not express synaptic CP-AMPAR. We saw no effect on these currents (0.45 ± 1.6% change, n = 5 p = 0.66; data not shown) indicating that NASPM effects were not because of nonspecific blockade of AMPAR generally. That is, these findings confirm that NASPM blocks inhibition of cortical inputs to interneurons in the BLA while sparing cortical activation of principal cells. The implication of these findings, which could not be derived from the behavioral studies alone, is that the drug acts selectively on CP-AMPAR that are expressed on GABAergic interneurons in the BLA, reduces inhibitory tone and, thereby, produces specific circuit changes rather than a broad disruption of neurotransmission.
Experiment 5: NASPM reduces inhibitory tone by blocking glutamatergic activation of interneurons. a, Spontaneous inhibitory events recorded at Vm −5, from LA principal cells in control conditions, and following bath application of 100 μm NASPM. b, Histograms showing average sIPSC current frequency and amplitude in control and following bath application of NASPM, n = 10. c, Recording configuration for recording disynaptic eIPSCs activated by cortical inputs to LA (left) and monosynaptic eIPSCs by stimulating LA interneurons directly (right). Objective lens magnification: 5×. d, Time plot showing IPSC amplitudes for a single experiment where a disynaptic IPSC was evoked and NASPM, bicuculline (10 μm) and NBQX (10 μm) were applied sequentially following washouts. Insets show average IPSCs in control/after wash and following application of each drug. e, Time plot showing eIPSC amplitude for monosynaptic IPSCs in control and After addition of NASPM. Average eIPSC is shown inset. f, Average inhibition of eIPSCs by NASPM, n = 8 and 7, respectively. LA = lateral amygdala; *p < 0.05.
Discussion
This series of experiments has shown that prediction error determines the role of BLA NMDAR in Pavlovian fear conditioning. In the initial experiments, rats received a BLA infusion of the NMDAR antagonist, D-AP5 before a serial-order conditioning protocol which consisted in S2-S1-shock sequences. These sequences contained an S1 that was novel or already conditioned via its parings with shock. In both cases, rats acquired fear of the S2 but differed with respect to the effects of the D-AP5 infusion: this infusion disrupted conditioning of S2 when S1 was novel (experiment 1b) but spared that conditioning when S1 had been previously paired with shock (experiments 1a and 1b). The failure of D-AP5 to disrupt conditioning to S2 was not because of the prior S1-shock pairings per se Following exposure to S1-shock pairings in stage 1, a BLA D-AP5 infusion disrupted trace conditioning of S2 when S1 was omitted (experiment 2a), and second-order conditioning of S2 when the shock was omitted (experiment 2b). The BLA D-AP5 infusion also disrupted conditioning of S2 when the S2-S1-shock sequences occurred in a different context from that where the S1-shock pairings had taken place (experiment 3a), and when these sequences occurred in the same physical context as the S1-shock pairings but the interval of time between the two experiences was extended from 2 d to two weeks (experiment 3b). These results suggest that NMDAR activation in the BLA is not required for conditioning of S2 when the shock occurs as expected given the context, time and stimuli that are present, but is required for that conditioning when the occurrence of the shock or its omission is unexpected (Table 1).
The NMDAR activation requirement for fear learning in the BLA covaries with surprise, or error in relation to the US
The finding that NMDAR activation in the BLA is critical for fear conditioning under some circumstances but not others suggests that the role typically accorded these receptors in such conditioning is, at best, incomplete. The biophysical properties of NMDAR lend themselves to the idea that they are activated by the coincidence of the CS and US, thereby constituting the contiguity mechanism historically thought to underlie associative formation. Instead, the findings here suggest that NMDAR only encode CS-US coincidences when the US is surprising, it does not encode these coincidences when the US is expected on the basis of other stimuli that are present. However, this merely begs the question: Why do NMDAR encode the fear memory that forms when a novel CS is paired with an unexpected US but fail to encode the fear memory that forms when a novel CS is paired with an expected US?
One account of this dissociation is that information provided by S1 alters the way in which NMDAR process US information and, thereby, their involvement in conditioning to S2. NMDARs differ from other glutamate receptors in that they are ligand-gated and voltage-gated, and within the BLA, the voltage gate has been identified with processing of US information: US presentations depolarize amygdala neurons and, thereby, remove the magnesium block on the NMDAR complex. Thus, in the presence of a previously conditioned S1, NMDAR may not be engaged for encoding fear to S2 because the expected shock US is less effective in depolarizing amygdala neurons and/or fails to remove the magnesium block from the NMDAR channel (for supporting evidence, see Johansen et al., 2010). This account explains many of the present results, including the finding that fear conditioning is NMDAR independent when S2 is paired with shock in the presence of an already-conditioned S1 (experiments 1a and 1b), and NMDAR-dependent when S2 is paired with shock in the absence of the S1 (experiments 1b and 2a). However, it fails to explain the NMDAR dependence of second-order fear conditioning (experiment 2b): as the US is not physically present, this account incorrectly predicts that acquisition of second-order fear to S2 should not require NMDAR activation in the BLA. It also gives no account of how context and time regulate the effects of the S1 on conditioning to S2, other than to suggest that S1 may be “perceived” differently when it is presented in a different physical/temporal context, resulting in a reduced expectancy of the shock US, and thereby, continued engagement of NMDAR for acquisition of fear to S2.
Rather than altering the way that NMDAR process US information, an alternative explanation for the present findings is that an already-conditioned S1 alters the threshold for neurotransmission in the BLA, hence, fear to S2 does not require activation of NMDAR. This possibility is supported by the findings that: (1) GABAergic interneurons regulate Pavlovian fear conditioning in the BLA (Wolff et al., 2014; Krabbe et al., 2018); (2) the involvement of these interneurons is achieved through activation of CP-AMPAR, not NMDAR (Mahanty and Sah, 1998; Polepalli et al., 2010, 2020); and (3) blockade of these interneurons by the CP-AMPAR antagonist, NASPM, disrupts fear to S2 when rats are exposed to S2-S1-shock sequences (experiments 4b, 4c, and 5). One possibility is that, under the influence of interneurons that express CP-AMPAR, the already-conditioned S1 reactivates the BLA neurons that had been involved in its conditioning; hence, the S2 engages previously excited neurons which have the potential to encode fear independently of NMDAR. However, error or uncertainty in predictions regarding the US result in activation of a new population of BLA neurons, again under the influence of interneurons that express CP-AMPAR; hence, conditioning to S2 in these newly excited neurons requires activation of NMDAR. Put differently, when S2 is presented in compound with an already-conditioned S1, it engages previously potentiated pathways and, thereby, its conditioning bypasses the NMDAR activation requirement in the BLA. But if the US fails to occur as expected (as in second-order conditioning) or its occurrence is rendered uncertain (as in the context and temporal shift experiments), the S2 is processed in the same way as any other novel stimulus, thereby reinstating the NMDAR activation requirement for its conditioning.
Finally, it is important to recognize similarities between the present findings and those of previous studies that examined the involvement of hippocampal NMDAR in the formation of a context fear memory (Sanders and Fanselow, 2003; Hardt et al., 2009; Wang et al., 2012; Finnie et al., 2018). For example, rats in the Sanders and Fanselow study were first conditioned to fear context A and, later, conditioned to fear context B. The critical result was that the latter conditioning occurred independently of NMDAR activation in the hippocampus. This is very similar to the present finding that, among rats exposed to S1-shock pairings in stage 1 and then S2-S1-shock sequences in stage 2, acquisition of fear to S2 occurred independently of NMDAR activation in the BLA (experiments 1a and 1b). Our proposal, that prediction error determines the involvement of NMDAR in Pavlovian fear conditioning can also be applied to the results obtained by Sanders and Fanselow. The A and B contexts used by Sanders and Fanselow will have had some unique features (size, shape smell; denoted A' and B') but shared many features in common (grid floor, texture, lighting, location in laboratory; denoted X). As such, the initial conditioning of context A will have conditioned fear to the unique features of this context (A') as well as those that it shares in common with context B (X). Hence, when rats were placed in context B, the shock will have been at least partially predicted by the features that B shares in common with context A, in the same way that the shock in stage 2 of our critical experiments was already predicted by the pretrained S1. That is, we suggest that the common features of the A and B contexts in the Sanders and Fanselow studies played the same role as the S1 in our critical experiments: their presence meant that the US was expected and, hence, NMDAR activation in the hippocampus was not required for conditioning fear to the unique features of context B. More generally, our proposal can be applied to the full set of findings that demonstrate NMDAR-independent learning in the hippocampus. This includes the seminal findings by Bannerman et al. (1995) which showed that spatial learning in the water maze became independent of hippocampal NMDA receptors when the animals were pretrained in a different context, although the common features of the contexts in this prior study are not as easily specified.
In summary, the encoding of a new dangerous experience requires NMDAR activation in the BLA when it is surprising or unexpected; but does not require activation of these receptors when it occurs as expected based on the context, time and stimuli that are present. Under the latter circumstances, the new dangerous experience involves activation of CP-AMPAR, which are expressed on interneurons that effectively gate neurotransmission in the BLA (Wolff et al., 2014; Krabbe et al., 2018). Future work will examine the involvement of these interneurons in fear conditioning as well as parallels between the present findings and theories that allow prediction error to influence the substrates of learning and memory in the brain (Pouget et al., 2013; Aitchison et al., 2021).
Footnotes
This work was supported by the Australian Research Council Discovery Grant DP200102969 (to N.M.H. and R.F.W.) and by an Australian Government Research Training Fellowship (M.J.W.-S.).
The authors declare no competing financial interests.
- Correspondence should be addressed to Nathan M. Holmes at n.holmes{at}unsw.edu.au
