Abstract
Rats quickly learn to fear a stimulus (e.g., a light) that signals brief but painful footshock. The consolidation of this first-order conditioned fear requires transcription and translation of specific genes in the basolateral amygdala complex (BLA). Rats also learn to fear the associates of first-order conditioned stimuli, such as a sound paired with the already-conditioned light. The consolidation of this second-order conditioned fear also requires transcriptional processes in the BLA, but unlike consolidation of first-order fear, it is unaffected by BLA infusions of the protein synthesis inhibitor, cycloheximide. Accordingly, this study sought to identify genes/pathways that regulate second-order conditioned fear in male rats. It focused on the involvement of immediate early genes, as these can be transcribed in the presence of protein synthesis inhibitors. In Experiment 1, we used a principal components analysis of PCR data and found that second-order fear conditioning involves region- and time-specific changes in a network of genes including the immediate early genes bdnf and egr1. In Experiments 2–4, we used BLA infusions of antisense oligonucleotides and found that bdnf is required for consolidation of both first- and second-order fears but is not required for reconsolidation of either and egr1 is not required for consolidation of first-order fear but is required for acquisition of second-order fear and reconsolidation of both types of fear. Hence, bdnf and egr1 regulate different aspects of first- and second-order fear conditioning in the BLA. These findings are discussed with respect to novel forms of information processing in the mammalian brain.
Significance Statement
Consolidation of second-order fears is unaffected by infusions of the protein synthesis inhibitor, cycloheximide, into the basolateral amygdala complex (BLA). This study sought to identify immediate early genes that might consolidate second-order fears in the BLA, as these genes can be transcribed in the presence of cycloheximide. It found that, within the BLA, second-order fear is regulated by bdnf and egr1: both regulate the acquisition/consolidation of second-order fear, and the latter additionally regulates its reconsolidation. These findings are consistent with previous studies of information processing in the hippocampus and are significant in showing how fear-related information can be regulated by immediate early genes in the BLA. They are considered in relation to current theories of information processing in the amygdala.
Introduction
There are two types of dangerous experiences: those that involve physical pain and those that involve threat of pain without themselves being painful. In the laboratory, these experiences can be studied using a protocol in which rats are first exposed to pairings of one stimulus (S1; e.g., a light) and brief but painful footshock (Stage 1) and, some days later, to pairings of a second stimulus (S2; e.g., a sound) and the already-conditioned S1 (Stage 2). Subsequent presentations of either stimulus evoke a range of responses indicative of fear in people, including changes in heart rate and blood pressure, release of stress hormones, potentiated startle, and freezing (Bolles and Collier, 1976; Iwata and LeDoux, 1988; Davis et al., 1993). Importantly, the responses evoked by S2 are not due to generalization from the already-conditioned S1 as they do not occur among controls exposed to unpaired presentations of the relevant stimuli in either stage of training (Rizley and Rescorla, 1972; Parkes and Westbrook, 2010; Holmes et al., 2013; Michalscheck et al., 2021; Leake et al., 2024). Instead, fear of the S2 is associatively mediated, or put another way, a result of its conditioning. Pavlov (1927) referred to this conditioning of S2 as second-order to distinguish it from the prior conditioning of S1, which he referred to as first-order.
The fears produced by first- and second-order conditioning are similar in many ways (Todd and Holmes, 2022). Both require neuronal activity in the basolateral amygdala complex (BLA), including activation of N-methyl-d-aspartate (NMDA) receptors for their acquisition (Campeau et al., 1992; Gewirtz and Davis, 1997; Parkes and Westbrook, 2010), CaMK signaling, and DNA methylation for their consolidation (Rodrigues et al., 2004; Maddox et al., 2014; Lay et al., 2018). However, the two types of fear also differ in important respects. These differences include the ways in which they are consolidated in the BLA. Whereas first-order fears are disrupted by a BLA infusion of the protein synthesis inhibitor, cycloheximide, second-order fears are unaffected by these infusions (Lay et al., 2018; Williams-Spooner et al., 2019; Leake et al., 2024). This has led to the view that second-order fear conditioning requires only a subset of the molecular processes that consolidate first-order conditioned fear in the BLA; notably, processes that are unaffected by the actions of a protein synthesis inhibitor. We hypothesized that these processes might involve activation of immediate early genes in the BLA, as a defining feature of these genes is that they can be transcribed in the presence of protein synthesis inhibitors (Cole et al., 1989; Lanahan and Worley, 1998). Specifically, we hypothesized that second-order fear conditioning would be accompanied by changes in expression of immediate early genes that have been shown to regulate first-order fear conditioning in the hippocampus and BLA (Lee et al., 2004; Maddox et al., 2010) and disrupted by manipulations that prevent transcription of these genes in the BLA.
Accordingly, the aim of the present study was to identify the immediate early genes that regulate second-order fear conditioning in the BLA. In Experiment 1, rats in groups of interest were exposed to S1–shock pairings in Stage 1 and S2–S1 pairings in Stage 2. The parameters of these exposures were identical to those used to produce second-order conditioned fear in our previous studies (Parkes and Westbrook, 2010; Holmes et al., 2013, 2014; Fam et al., 2023). Rats were killed 30, 90, or 270 min after the Stage 2 conditioning session, and their brains were processed for analyses of gene expression in the BLA and neighboring perirhinal cortex (PRh; control). These analyses identified second-order conditioning with time-dependent and BLA-specific changes in a network of genes including brain-derived neurotrophic factor (bdnf) and endothelin-like growth factor 1 (egr1). Subsequent experiments then used BLA infusions of antisense oligonucleotides (ASOs) to assess the effect of inhibiting bdnf or egr1 mRNA on consolidation and reconsolidation of both first- and second-order conditioned fear.
Materials and Methods
Subjects
Subjects were experimentally naive, Long–Evans rats, 10–12 weeks old (350–415 g) at the start of the experiment. All rats were obtained from a breeding colony maintained by the School of Psychology at the University of New South Wales (UNSW) and housed in plastic tubs (with four rats per tub) in an air-conditioned colony room. The lights in the colony room were maintained on a 12 h light/dark cycle, with lights on at 7.00 A.M. Chow and water were continuously available in the tubs which also contained plastic tunnels and wood pieces for enrichment. All experimental procedures were conducted between 7.00 A.M. and 5.00 P.M. and were approved by the ethics committee of UNSW.
All rats in the present study were males. This was for two reasons. First, body weight often correlates with freezing data and, thus, can influence the expression of genes that are relevant to fear learning and memory (unpublished data in lab). As such, we decided to investigate gene expression in one sex only so that the range of body weights for the ∼96 rats in Experiment 1 was as small as possible. Second, in previous studies, the genes that regulate first-order fear conditioning in the BLA were identified in male rats. As we wanted to compare the results of the present study, in which we assessed the genes involved in second-order fear conditioning, with the results of previous studies, we decided to investigate gene expression in male rats only. Where sex differences have been assessed in relation to the role of the BLA in pavlovian fear conditioning, including the involvement of NMDA receptors in the acquisition of conditioned fear and protein synthesis in its consolidation, no sex differences have been observed (Williams-Spooner et al., 2022; Keidar et al., 2023; Leake et al., 2024). We hypothesize that the same will be true in relation to the gene networks that regulate second-order conditioned fear in the BLA, but this awaits confirmation in future studies.
Apparatus
All behavioral procedures took place in four identical chambers (30 cm wide × 30 cm high × 30 cm deep). The walls and ceiling of the chamber were made of clear Plexiglass, and the floor consisted of stainless-steel rods (7 mm in diameter, spaced 1.8 mm apart) connected to a shock generator which delivered a 0.5-s-long, 0.8 mA intensity shock. Each chamber was located in its own light- and sound-attenuating cabinet. A 2 × 3 array of light-emitting diodes, mounted on the back wall of each cabinet, was used to present the visual stimulus (light flashing at a rate of 3.5 Hz, ∼8 lux at the center of the chamber). A speaker, mounted on the side wall of each cabinet, was used to present the auditory stimulus (1,000 Hz tone, ∼75 dB, against a background noise of ∼45 dB). A camera and an infrared light, mounted on the back wall of each cabinet, were used to record and visualize the behavior of each rat. The presentation of stimuli and footshock was controlled with MATLAB software on a computer located in another room of the laboratory.
Surgical procedures
Rats were surgically implanted with cannulas targeting the BLA. Briefly, rats were anesthetized with 5% inhalant isoflurane, positioned in a stereotaxic apparatus, maintained under 2% isoflurane, and injected subcutaneously (s.c.) with the nonsteroidal anti-inflammatory, carprofen (5 mg/kg). Following the onset of stable anesthesia, an incision was made over the skull and two holes drilled through the skull to permit implantation of a 26-gauge guide cannula into the BLA in each hemisphere (coordinates: 2.4 mm posterior to bregma, 4.9 mm lateral to the midline, and 8.2 mm ventral to bregma). The cannulas were secured to the skull with dental cement and four jeweler's screws. To prevent contamination of the guide cannulas, dummy cannulas were inserted into each guide and remained there at all times except during drug or vehicle infusions. After surgery, rats received an intraperitoneal (i.p.) injection of penicillin (0.05 ml, 33 mg/kg) as a prophylaxis. They were then returned to their home cage and allowed to recover for 7 d, on each of which they were handled and weighed.
Drug preparation and infusion
Antisense and missense oligonucleotides (MSO; Integrated DNA Technologies) were used to knock down the expression of the immediate early genes, bdnf and egr1. Antisense oligonucleotides are a common way by which genes are knocked down, through the use of ribonuclease H-dependent cleavage of mRNA to prevent translation (Achenbach et al., 2003). The oligonucleotide sequences, concentration, and volumes were taken from Lee et al. (2004). The sequences were as follows: bdnf antisense oligodeoxynucleotide (ODN), 5′-TCT TCC CCT TTT AAT GGT-3′; bdnf missense ODN, 5′-ATA CTT TCT GTT CTT GCC-3′; egr1 antisense ODN, 5′-GGT AGT TGT CCA TGG TGG-3′; egr1 missense ODN, 5′-GTG TTC GGT AGG GTG TCA-3′. Importantly, these sequences have been validated previously (Lee et al., 2004). Phosphorothioate bonds were included across all oligonucleotide sequences to prevent degradation caused by endonucleases. Oligonucleotides were resuspended in an IDTE (pH 8.0) vehicle.
Rats were randomly allocated to groups that received a BLA infusion of either antisense (ASO) or missense (MSO) oligonucleotides (1 nmol/μl; 0.5 μl per hemisphere). To perform infusions, dummy caps were removed from the guide cannulas, and 33-gauge internal cannulas were connected to Hamilton syringes. Oligonucleotides were delivered at a constant rate of 0.25 μl/min by an infusion pump (Harvard Apparatus). After the infusion, the internal cannulas were left in the guide cannulas for 2 min and then removed.
Histology for cannula placements
Following behavioral procedures, rats were killed with an intraperitoneal injection of sodium pentobarbital (1 ml). The brains were removed, frozen at −20°C, and then sectioned coronally at 40 µm on a cryostat. Sections containing the BLA were mounted onto glass slides and stained using cresyl violet. The cannula placements were determined using a light microscope, and any placements outside the boundaries of the BLA were excluded from analysis. Figure 1 shows the placements for rats that were included in the study.
Cannula placements for all rats included in this study. Black dots represent the most ventral point of the cannula track for each rat.
Sample collection for PCR
In Experiment 1, gene expression in the BLA and PRh was assessed in four groups (Table 1) at three time points: 30, 90, or 270 min after the training session in Stage 2. Rats in Group Second-Order were exposed to light–shock pairings in Stage 1 followed by tone–light pairings in Stage 2 in the manner described by Lay et al. (2018). Changes in gene expression in this group of interest were assessed relative to three other groups. Group Context received context-alone exposure during both Stages 1 and 2 and served as a baseline against which changes in gene expression could be assessed. Group First-Order Only were exposed to light–shock pairings in Stage 1 followed by context-alone exposure in Stage 2. Finally, Group Sensory Preconditioning (SPC) received context-alone exposures in Stage 1 followed by tone–light pairings in Stage 2. These groups were chosen as they are matched to Group Second-Order in terms of the training experience in Stage 1 (Group First-Order Only) or Stage 2 (Group Sensory Preconditioning). Thus, genes that are differentially expressed in Group Second-Order relative to Group First-Order Only must be involved in second-order conditioning, whether this involvement relates to the novelty of the tone or the absence of the expected shock. Genes that are differentially expressed in Group Second-Order relative to Group Sensory Preconditioning might be involved in second-order fear conditioning or sensory preconditioning. In order to differentiate between these possibilities, we additionally assessed gene expression in the perirhinal cortex (PRh) as the available evidence shows that it is required for sensory preconditioning in the protocol used here (Holmes et al., 2013, 2018, 2022; Qureshi et al., 2023).
Experimental design for Experiment 1
Either 30, 90, or 270 min after the final training session in Experiment 1, rats were killed as above, and their brains were rapidly extracted, frozen in liquid nitrogen, and stored at −80°C. Brains were sectioned into 1-mm-thick coronal slices using a brain matrix, and tissue was collected using four 1 mm punches in two consecutive brain slices for the BLA and four 2 mm punches in two consecutive slices for the PRh.
qRT-PCR quantification of gene expression
Gene expression in the BLA and PRh was determined using qRT-PCR. Genes of interest were selected based on previous work in which they were implicated in the consolidation of pavlovian conditioned fear (Goosens et al., 2000; Schafe et al., 2000; Schafe and LeDoux, 2000; Maren et al., 2003; Rattiner et al., 2004; Rodrigues et al., 2004; Keeley et al., 2006; Maddox et al., 2010; Ploski et al., 2010; Monsey et al., 2011; Lay et al., 2018). RNA was isolated using the TRIzol extraction method in accordance with the manufacturer's guidelines (Invitrogen). Genomic DNA contamination was removed using DNase I (Sigma-Aldrich), and samples were then reverse transcribed in a C1000 Touch Thermal Cycler (Bio-Rad Laboratories) using the QuantiTect Reverse Transcription Kit (Qiagen) according to the manufacturer's instructions. Quantitative PCR was performed in a StepOnePlus System (Applied Biosystems) using either SYBR Select Master Mix (Applied Biosystems) or TaqMan probes (Thermo Fisher Scientific). Primers for SYBR detection were designed based on previous studies and purchased from Integrated DNA Technologies. Primer sequences are listed in Table 2 with GAPDH as a housekeeper gene. The ΔΔCt method was used to determine relative expression of genes of interest (Pfaffl, 2001), with the context-only group serving as the baseline for calculation of the fold change.
Primer sequences used to assess changes in gene expression in the BLA and PRh
Behavioral procedures
Experiment 1
Context preexposure
All rats were familiarized with the experimental chambers (context) on Days 1 and 2. There were two, 20 min sessions of context exposure each day, one in the morning and the other 3 h later in the afternoon.
First-order conditioning
On Day 3, rats in Groups Second-Order and First-Order Only were exposed to four presentations of a flashing light (3.5 Hz). Each light presentation lasted for 10 s and coterminated with footshock (0.8 mA, 0.5 s). The first light presentation occurred 5 min after placement in the context, the interval between light presentations was 5 min, and rats remained in the context for an additional 2 min after the final light presentation. Rats in Group Sensory Preconditioning and Group Context received equivalent exposure to the context without stimulus presentations.
Context extinction
On Day 4, rats received two context-alone exposures in the absence of any scheduled events, one in the morning and the other in the afternoon. Each was 20 min in duration. These exposures were intended to extinguish any freezing elicited by the chambers that would otherwise obscure detection of second-order conditioned freezing responses to the tone in the next stage of training.
Second-order conditioning/sensory preconditioning
On Day 5, rats in Group Second-Order and Group Sensory Preconditioning received four pairings of a tone (1,000 Hz, 75 dB) and the flashing light. Each presentation of the tone lasted for 30 s, and its offset co-occurred with the onset of the 10 s flashing light. The first tone presentation occurred 5 min after placement in the context. The interval from light offset to the next tone onset was fixed at 5 min. Rats remained in the context for an additional 2 min after the final light presentation. Rats in Groups First-Order Only and Context Only received equivalent exposure to the context without stimulus presentations.
Experiment 2
The behavioral procedures for Days 1–5 were identical to those described for Group Second-Order in Experiment 1. Rats in this experiment received 2 additional days of testing with S2 and S1 on Days 6 and 7. Briefly, all rats were pre-exposed to the context on Days 1 and 2. On Day 3, rats received four light–shock pairings with an intertrial interval of 5 min. The first light–shock pairing occurred 5 min after placement in the chamber, and the rats were removed from the chamber 2 min after the final light–shock pairing. One hour prior to the session containing these pairings, rats received an intra-BLA infusion of either an antisense oligonucleotide (ASO) or a scrambled missense oligonucleotide. In Experiment 2a, rats were infused with a bdnf ASO or MSO prior to first-order conditioning. In Experiment 2b, rats were infused with an egr1 ASO or MSO prior to first-order conditioning. They then received context extinction (Day 4) followed by second-order conditioning (Day 5). During second-order conditioning, rats received four tone–light pairings, with a 5 min interval between pairings. Rats were removed from the chambers 2 min after the final pairing. On Days 6 and 7, rats were tested with the tone (S2) and the light (S1), respectively. The order of testing was not counterbalanced as any extinction of S1 on an initial test could undermine freezing to S2 on a subsequent test. Each test session consisted of eight stimulus-alone presentations, either S2 or S1. The duration of S2 and S1 remained 30 and 10 s, respectively, and the intertrial interval was set at 3 min. After the final stimulus presentation, rats stayed in the chambers for another minute and were then returned to their home cage.
Experiment 3
The behavioral procedures for Experiment 3 were identical to those described for Experiment 2 with the exception that the intra-BLA infusion took place 1 h prior to second-order conditioning rather than 1 h prior to first-order conditioning. Briefly, rats received 2 d of context exposure (Days 1 and 2), followed by light–shock pairings (Day 3) and context extinction (Day 4). On Day 5, rats received intra-BLA infusions of either the bdnf ASO or MSO (Experiment 3a) or egr1 ASO or MSO (Experiment 3b) 1 h prior to second-order conditioning. During second-order conditioning, rats received four tone–light pairings, separated by 5 min. Rats were tested for their level of freezing to the tone (S2) and light (S1) on Days 6 and 7, respectively.
Experiment 4
The behavioral procedures were similar to those previously described. Rats received context exposure on Days 1 and 2, followed by light–shock pairings (Day 3), context extinction (Day 4), and tone–light pairings (Day 5). On Day 6, rats received an intra-BLA infusion of either the bdnf ASO or MSO (Experiment 4a) or egr1 ASO or MSO (Experiment 4b) and, 1 h later, were re-exposed to two presentations of the tone (S2) alone. Finally, on Days 7 and 8, rats were tested for their level of freezing to the tone (S2) and light (S1), respectively.
Experimental design and statistical analyses
qRT-PCR gene data
The ΔΔCt method was used to determine the relative level of gene expression in each experimental group relative to the context-only control. The results are presented as a mean fold change value with standard error of the mean. The distribution of ΔCt values was assessed using the Shapiro–Wilk test, and equality of group variances was assessed using Levene's test. Between-group differences were then analyzed using ANOVA with post hoc Tukey HSD. As our study involved three behavioral training conditions (the context-only condition was used solely as a reference for calculation of the ΔΔCt values), two brain regions (BLA, PRh), three time points (30, 90, and 270 min), and 12 genes of interest (see below), we used the “false discovery rate” (FDR) procedure to control the chance of a Type 1 error at 0.05. In the reporting of the results below, we indicate genes that were significantly differentially expressed between the different groups in the absence of any correction for multiple comparisons, as well as those genes that remained significantly differentially expressed after the FDR correction for multiple comparisons.
A principal component analysis (PCA) was also used to determine the relative contribution of individual genes to differences between Groups Second-Order and First-Order Only. PCA can be used to confirm a priori group allocations based on differences in the abundance of genes/proteins in individual samples (Reich et al., 2008). Importantly, PCA can identify hidden patterns in datasets and provide a way to visualize between-group differences (Ringnér, 2008). Previous work has demonstrated the utility of such an approach in analyzing and interpreting mRNA expression data (Macdonald et al., 2001; Notingher et al., 2004; Modlin et al., 2009). For the PCA analysis, PCR data from the BLA of rats in Groups Second-Order and First-Order Only were presented as an n × m matrix where rows were associated with the samples and columns with genes of interest (repeated for each time point). The differences between Groups Second-Order and First-Order Only in principal components (PC) were then assessed using a Mann–Whitney U Test, and the contributions of individual genes to the principal components were presented and assessed as PCA loadings.
Finally, subsequent to the PCA, we conducted a final series of experiments in which we knocked down genes that our transcriptional analyses and PCA implicated in second-order conditioned fear and examined the impact of these knockdowns on second-order conditioned fear. The results of these experiments confirmed the indications of our gene transcription and PCA results. They are described below.
Behavioral data
Scoring
Freezing was used to quantify conditioned fear and was defined as the absence of all movements except those required for breathing (Fanselow, 1980). Each rat was scored as either “freezing” or “not freezing” every 2 s. A percentage score was calculated as the ratio of observations scored as freezing divided by the total number of observations. All data were scored by the experimenter and then cross-scored by an experienced observer who was blind to the purposes of the experiment. The correlation between the two sets of scores was high (Pearson’s product–moment correlation >0.9). Any discrepancies between the scores were resolved in favor of those by the naive observer.
Statistical analysis of behavioral data
The acquisition data in each experiment were analyzed using mixed-model ANOVA with a between-subject factor of group and a within-subject factor of trial. The test data in each experiment were analyzed using a one-way ANOVA with a between-subject factor of group (ASO vs MSO). For all statistical analyses, the criterion for rejection of the null hypothesis was set at α = 0.05.
Results
Experiment 1: second-order fear conditioning involves time-dependent and amygdala-specific changes in gene expression, including the immediate early genes bdnf and egr1
The aim of this experiment was to identify the genes that regulate second-order conditioned fear in the BLA. The 12 genes of interest were selected for two reasons. The first reason was based on our prior work investigating the substrates of second-order fear conditioning in the BLA (Lay et al., 2018) and/or sensory preconditioning in the PRh (Holmes et al., 2018; included as a control region to determine the specificity of gene involvement in second-order conditioned fear). Specifically, our prior work indicated that second-order conditioned fear requires neuronal activity in the BLA (Parkes and Westbrook, 2010; Holmes et al., 2013; Lay et al., 2018). The activity requirement includes activation of NMDA receptors, CaMK signaling, and the methyltransferases that catalyze DNA methylation—this was demonstrated by showing that drugs that target each of these processes (DAP5 for NMDA receptors, KN-62 for CaMK signaling, and Rg108 and 5-AZA for DNA methylation) disrupt the consolidation of second-order conditioned fear when injected directly into the BLA (Lay et al., 2018). The activity requirement does not include ERK/MAPK or PKA/PKC/PKI signaling as evidenced by our findings that drugs targeting each of these processes (U0126 for ERK/MAPK signaling, H7 for PKA/PKC/PKI signaling) have no effect on the consolidation of second-order conditioned fear when injected into the BLA (Lay et al., 2018). Hence, the first nine genes of interest included camKIIa, camKIIb, camK4, dnmt3a, mapk1, map2k7, PKCb1, PKIa, and Prkaca. The second reason for selecting genes of interest was based on their immunity to the effects of cycloheximide. That is, based on our findings that a BLA infusion of the protein synthesis inhibitor, cycloheximide, has no effect on consolidation of second-order conditioned fear (Lay et al,., 2018; Leidl et al., 2018; Williams-Spooner et al., 2019; Leake et al., 2024), genes in the second category included the immediate early genes arc, bdnf, and egr1, as (1) each of these genes is differentially expressed in the amygdala 30–180 min after first-order fear conditioning (Rattiner et al., 2004; Keeley et al., 2006) and (2) the defining feature of these genes is their ability to be transcribed in the presence of protein synthesis inhibitors (Cole et al., 1989; Hughes et al., 1993; Lauterborn et al., 1996; Lanahan and Worley, 1998; Okuno, 2011; Bahrami and Drablos, 2016). Our general hypothesis was that immediate early genes may regulate second-order conditioned fear by coordinating the activity of CaMKs and DNMTs in the BLA. We speculate about how or why this might occur in the Discussion.
Gene expression in the BLA was assessed 30, 90, or 270 min after behavioral training in rats that underwent second-order conditioning, relative to rats in three other groups (Table 1). Group Second-Order received light–shock pairings in Stage 1 and tone–light–shock pairings in Stage 2. Group First-order Only received context-only exposure in Stage 1 and light–shock pairings in Stage 2. Group Sensory Preconditioning received tone–light pairing in Stage 1 and light–shock pairings in Stage 2. Finally, Group Context received context-alone exposure in both sessions and served as a baseline against which changes in gene expression were measured. In addition to assessing gene expression in the BLA, we also assessed gene expression in a control region, the PRh. This region was selected as the available evidence suggests that it is selectively involved in sensory preconditioning (Holmes et al., 2013, 2018, 2022; Qureshi et al., 2023).
Acquisition of first- and second-order conditioned fear
Figure 2 shows the acquisition of freezing to the first-order conditioned light (S1) and second-order conditioned tone (S2). Averaged across groups, freezing to S1 increased across its four pairings with shock in Stage 1 (F(3,126) = 1,292.38, p < 0.001). There was no significant difference between Group Second-Order and Group First-Order Only in overall levels of freezing to S1 and no significant group × trial interactions (Fs < 1). Freezing to S2 in Group Second-Order increased across its four pairings with conditioned S1 in Stage 2 (F(3,126) = 20.98, p < 0.0001) while freezing to S1 was maintained across these pairings (F(3,126) = 1.04, p = 0.38). There were no significant differences in freezing between the groups that would be killed either 30, 90, or 270 min after the second-order conditioning session (Fs < 1).
Acquisition of first-order fear (freezing) to S1 across its pairings with shock in Stage 1 and second-order fear to S2 across its pairings with S1 in Stage 2. A, Overview of the protocol used to generate first- and second-order fear conditioning in Groups First-Order Only and Second-Order. B, Mean (±SEM) percentage freezing to S1 across the four S1–shock pairings in Stage 1. C, Mean (±SEM) percentage freezing to the S2 and S1 across the four S2–S1 pairings in Stage 2.
Gene expression in the BLA and PRh 30 min after training in Stage 2
There were substantial changes in gene expression in the BLA 30 min after the Stage 2 training session in Groups Second-Order, First-Order Only, and Sensory Preconditioning (evident as deviation from the baseline in Fig. 3A). In the analysis with no correction for multiple comparisons, two genes were significantly differentially expressed between these groups: mapk1 (F(2,24) = 12.71, p < 0.001) and camkIV (F(2,24) = 5.49, p < 0.05). The expression of mapk1 was lower in Groups Second-Order (p < 0.01) and Sensory Preconditioning (p < 0.001) relative to Group First-Order Only, and the expression of camkIV was lower in Group Second-Order relative to Group First-Order Only (p < 0.05). However, after the FDR correction for multiple comparisons, the only significant between-group difference in gene expression related to mapk1: the expression of mapk1 was lower in Group Sensory Preconditioning relative to Group First-Order Only. There were few changes in gene expression in the PRh 30 min after the Stage 2 session, and, in the analysis with no correction for multiple comparisons, only one gene was differentially expressed among the different groups: mapk1 (F(2,24) = 5.49, p < 0.05; Fig. 3B). Specifically, the expression of mapk1 was significantly lower in Group Sensory Preconditioning that received pairings of the affectively neutral S2 and S1 relative to both Groups Second-Order (p < 0.05) and First-Order Only (p < 0.01). After the FDR correction for multiple comparisons, the only significant between-group difference in gene expression was again related to mapk1: the expression of mapk1 was lower in Group Sensory Preconditioning relative to Group First-Order Only.
Expression of mRNA after the training session in Stage 2 for Groups Second-Order, First-Order Only, and Sensory Preconditioning. Expression is shown as fold change relative to the context-alone group ± SEM. mRNA expression in the BLA (A, C, E) and PRh (B, D, F) 30 min after training (A, B), 90 min after training (C, D), or 720 min after training (E, F). All ns = eight rats/group. a, any group different from Group Second-Order; b, any group different from Group Sensory Preconditioning; c, any group different from Group First-Order Only. A fold change of zero indicates that gene expression levels were equivalent to those observed in Group Context, a fold change greater than zero indicates that gene expression was increased relative to Group Context, and a fold change below zero indicates that gene expression was suppressed relative to Group Context.
Gene expression in the BLA and PRh 90 min after training in Stage 2
In the BLA, the uncorrected analysis suggested that five genes were significantly differentially expressed among the groups 90 min after the training session in Stage 2 (Fig. 3C): camkIIb (F(2,24) = 3.60, p < 0.05), camkIV (F(2,24) = 8.81, p < 0. 0.01), pkia (F(2,24) = 6.31, p < 0.01), dnmt3a (F(2,24) = 15.94, p < 0.001), and egr1 (F(2,24) = 5.57, p < 0.05). For four of these genes, camkIV, pkia, dnmt3a, and egr1, the difference was due to lower levels of expression in Groups Second-Order and Sensory Preconditioning relative to Group First-Order Only (all ps < 0.05). In the case of the fifth gene, camkIIb, its expression was lower in Group Second-Order relative to Group First-Order Only (p < 0.05). After the FDR correction for multiple comparisons, the only between-group difference that remained significant related to dnmt3a: the expression of dnmt3a was lower in Group Second-Order relative to Group First-Order Only.
In the PRh, the uncorrected analysis suggested that four genes were significantly differentially expressed among the groups: camkIV (F(2,24) = 8.01, p < 0.01), dnmt3a (F(2,24) = 10.1, p < 0.01), pkia (F(2,24) = 3.47, p < 0.05), and mapk1 (F(2,24) = 4.39, p < 0.05, Fig. 3D). For two of these genes, camkIV and dnmt3a, the difference was due to higher levels of expression in Group Sensory Preconditioning relative to Groups Second-Order and First-Order Only (all ps < 0.05). For pkia, the difference was due to higher levels of expression in Group Sensory Preconditioning relative to Group Second-Order (all ps < 0.05). Finally, for mapk1, the difference was due to lower levels of expression in Group Sensory Preconditioning relative to Group First-Order Only. After the FDR correction for multiple comparisons, the only between-group difference that remained significant related to camkIV: the expression of camkIV was higher in Group Sensory Preconditioning relative to Group Second-Order.
Gene expression in the BLA and PRh 270 min after training in Stage 2
In the BLA, the uncorrected analysis suggested that just two genes were significantly differentially expressed among the groups 270 min after training in Stage 2: mapk1 (F(2,24) = 9.47, p < 0.01) and egr1 (F(2,24) = 3.47, p < 0.05). For mapk1, the difference was due to lower levels of expression in Groups Second-Order and Sensory Preconditioning relative to Group First-Order Only (all ps < 0.05). For egr1, the difference was due to higher expression in Group Sensory Preconditioning compared with Group First-Order Only (p < 0.05; Fig. 3E). However, after the FDR correction for multiple comparisons, none of these between-group differences remained significant.
Relative to both the 30 and 90 min timepoints, the uncorrected analysis showed that there was a greater number of differentially expressed genes in the PRh 270 min after the training session in Stage 2 (Fig. 3F): camkIIa (F(2,24) = 12.75, p < 0.001), camkIV (F(2,24) = 7.50, p < 0.01), map2k7 (F(2,24) = 6.58, p < 0.01), prkaca (F(2,24) = 4.27, p < 0.05), arc (F(2,24) = 4.65, p < 0.05), and egr1 (F(2,24) = 7.37, p < 0.01). For camkIIa, map2k7, and egr1, the differences were due to higher expression in Group Sensory Preconditioning than Groups Second-Order and First-Order Only (all ps < 0.05). For prkaca and arc, the differences were due to higher expression in Group Sensory Preconditioning than Group Second-Order (all ps < 0.01). For camkIV, the difference was due to lower expression in Group Sensory Preconditioning than Group Second-Order (p < 0.01). After the FDR correction for multiple comparisons, the between-group differences that remained significant were those relating to camkIIa and camkIV: the expression of camkIIa was higher in Group Sensory Preconditioning relative to Group Second-Order, and the expression of camkIV was lower in Group Sensory Preconditioning relative to Group Second-Order.
Notes about results obtained using qRT-PCR
There are two things to note about the results obtained using qRT-PCR. The first thing to note is that none of the 12 genes selected for analysis were ever upregulated in Group Second-Order. This could be taken to imply that the gene expression programs that regulate second-order conditioning occur in sparse ensembles of BLA neurons that are not captured by the methods used here. Future work will assess changes in gene expression in just those BLA neurons that are activated during the second-order conditioning session. However, it should also be noted that at the 30 and 90 min time points, differentially expressed genes were suppressed in Groups Second-Order and SPC relative to Group First-Order Only and, at the 90 min time point (where changes in gene expression were most pronounced), the suppression was evident as a selective upregulation of gene expression in Group First-Order Only. This finding is important because it shows that re-exposure to a previously shocked context (as occurred for rats in Group First-Order Only) initiates changes in gene expression in the BLA and, critically, that exposures to the tone and light in Group Second-Order interrupt these changes or prevent them from occurring. That is, whereas past studies have identified first-order fear conditioning with increased expression of genes that regulate different cell functions in the BLA, the present findings show that second-order fear conditioning is different: relative to rats with a common training history, genes that have been shown to be upregulated in the BLA following first-order conditioning are suppressed following second-order conditioning.
The second thing to note about results obtained using qRT-PCR is that the changes produced by second-order conditioning were very similar to those produced by sensory preconditioning: in Groups Second-Order and SPC, the pattern of gene expression was identical from 30 to 90 min after the session of tone–light pairings, with divergences in gene expression only occurring after a further 180 min (e.g., camkIIa and egr1 270 min after Stage 2). Such findings suggest that genes in the BLA are responsive to tone–light pairings independently of their conditioning history, which is consistent with recent work showing that (1) novel stimuli are processed in the BLA regardless of their affective significance and (2) learning about the relations between such stimuli requires neuronal activity, including activation of NMDA receptors, in the BLA (Qureshi et al., 2023; Leake et al., 2024). That is, when rats are exposed to relatively few tone–light pairings (as they were in the present study), they are processed in the BLA regardless of whether they are presented as part of a second-order conditioning protocol or a sensory preconditioning protocol. However, as the number of tone–light pairings is increased (i.e., as the stimuli become familiar across their repeated exposures), they continue to be processed in the BLA when presented as part of a second-order conditioning protocol but cease to be processed in the BLA when presented as part of a sensory preconditioning protocol (Holmes et al., 2013, 2018; Qureshi et al., 2023).
Principal components analysis
The evidence for gene suppression following second-order fear conditioning reinforces the point that learning and memory should not be identified with a simple increase in gene expression relative to some baseline (here, the context-only control). Rather, it is the pattern of expression across genes in a network that determines how new information is encoded and stored in the BLA. This was supported by a principal component analysis (PCA) of the data obtained through PCR at the 90 min time point, as this was when we observed the greatest number of between-group differences in the individual gene expression levels in the BLA. Given the established similarities between Groups Second-Order and Sensory Preconditioning, the PCA compared Groups Second-Order and First-Order Only, as these groups were matched for their prior experience of footshock and differed only with respect to whether they did (Group Second-Order) or did not (Group First-Order Only) receive second-order conditioning. This comparison revealed a network of genes that discriminated between these two cases (Fig. 4). As shown in Figure 4B, the PCA plot revealed distinct clustering among rats in Groups Second-Order and First-Order Only. The first three principal components (PC) in this analysis accounted for 83.6% of the overall variance in the dataset (Fig. 4A). Among the three principal components, PC1 (W = 49, p = 0.01) and PC3 (W = 10, p = 0.04) were significantly different between Groups Second-Order and First-Order Only. Contributors to the difference in clustering along PC1 included camKIIa, camkIIb, and dnmt3a (Fig. 4D). This is consistent with previous findings by Lay et al. (2018) who reported that consolidation of second-order fear is disrupted by a BLA infusion of drugs that inhibit CaMK signaling (KN-62) or methyltransferases that catalyze DNA methylation (Rg108, 5-AZA). In contrast, contributors to the difference in clustering along PC3 included pkcb1, mapk1, and map2k7 (Fig. 4E). This is most likely due to the involvement of these genes in some aspect of first-order conditioned fear (Group First-Order Only) as Lay et al. (2018) reported that consolidation of second-order fear is unaffected by a BLA infusion of drugs that inhibit PKA/PKC (H7) or ERK/MAPK (U0126) signaling.
Principal components analysis of gene expression in the BLA at 90 min after training in Stage 2. A, Scree plot of explained variance in the first 10 principal components. B, PCA plot with three principal components. C, The difference between Groups Second-Order and First-Order Only in three principal components. D, E, Relative percentage contribution of each gene to PC1 (D) and PC3 (E). Red dashed line indicates the expected value if all 12 genes contributed at the same level.
Importantly, the gene network that discriminated between Groups Second-Order and First-Order Only also included the immediate early genes bdnf and egr1 (PC1 and PC3). Specifically, bdnf explained 35% of the covariation among genes in PC3 (Fig. 4E) whereas egr1 explained 12% of the covariation among genes in PC1 (Fig. 4D) and 16% of the covariation among genes in PC3 (Fig. 4E). These results are a little surprising given that neither gene was differentially expressed among the different groups 30, 90, and 270 min after the Stage 2 training session (Fig. 3). By way of resolution, we argue that the absence of differential expression could and should be taken to indicate that bdnf and egr1 were equally engaged for the learning that is likely to have occurred in each of the groups relative to the context-only reference group: e.g., Group First-Order Only was re-exposed to the context where they had been shocked during Stage 1 and, as such, is likely to have extinguished context conditioned fear through activation of bdnf and/or egr1 in the BLA, and Group SPC was exposed to S2–S1 pairings and, as such, is likely to have encoded this information through activation of both bdnf and egr1 in the BLA.
The potential involvement of bdnf and egr1 in second-order fear conditioning is interesting for two reasons. First, bdnf and egr1 are activated downstream of neurotransmitter systems and signaling pathways that regulate both first and second-order conditioning in the BLA (e.g., glutamate via NMDA receptors and ERK/MAPK and CaMK signaling; Wisden et al., 1990; Williams et al., 2000; Duclot and Kabbaj, 2017; Gallo et al., 2018). Second, in the hippocampus, bdnf and egr1 have dissociable roles in consolidating and reconsolidating context conditioned fear (Lee et al., 2004). Specifically, hippocampal infusion of a bdnf antisense oligonucleotide (ASO) disrupts the consolidation of context conditioned fear but spares the reconsolidation of this fear. Conversely, hippocampal infusion of an egr1 ASO spares the consolidation of context conditioned fear but disrupts the reconsolidation of this fear. Hence, in the hippocampus, the roles of bdnf and egr1 in the consolidation and reconsolidation of context conditioned fear are doubly dissociable. Consolidation of context fear requires expression of bdnf but not egr1, and, conversely, reconsolidation of context fear requires expression of egr1 but not bdnf.
As we had not previously examined the involvement of bdnf and egr1 in the consolidation of second-order conditioned fear, we proceeded from the PCA to a direct examination of the roles of bdnf and egr1 in consolidating and reconsolidating first- and second-order conditioned fear in the BLA. That is, as we had already investigated each of the other processes that the PCA identified as being of potential interest in second-order fear conditioning (Lay et al., 2018), we used an antisense oligonucleotide (ASO) strategy to examine the roles of bdnf and egr1 in consolidating and reconsolidating first- and second-order conditioned fear in the BLA. Before proceeding to these experiments, there are two things that are important to note. First, the analyses of qRT-PCR data did not detect upregulation of bdnf or egr1 at any point after the Stage 2 training session (relative to the context-only reference group): if anything, egr1 was downregulated in the BLA 90 min after second-order conditioning in Stage 2. The failure to detect upregulation of bdnf or egr1 after second-order conditioning is likely due to the fact that our analyses were based on tissue punches collected from the BLA and PRh, which will have included cells that had been activated during the Stage 2 training session as well as cells that had not been activated. That is, our analyses were not confined to just those cells that were activated by the Stage 2 training session; hence, they were not as sensitive as they could have been to detecting upregulation of bdnf or egr1 after Stage 2 and/or between-group differences in the expression of bdnf and egr1. This highlights the significance of the next experiments in which we used ASOs to examine the involvement of bdnf and egr1 in the consolidation and reconsolidation of first- and second-order conditioned fear. The ASOs will be highly effective in blocking any upregulation of bdnf or egr1 in BLA cells activated during first- or second-order fear conditioning: hence, they are appropriate for examining whether bdnf and/or egr1 regulate the consolidation/reconsolidation of first- and second-order conditioned fear in the BLA.
The second thing to note is that, in contrast to cycloheximide which inhibits de novo protein synthesis by blocking the elongation phase of translation (Rao and Grollman, 1967; Obrig et al., 1971; see also Alberini, 2007), ASOs bind to specific RNA targets resulting in their degradation (i.e., they inhibit gene expression; Lee et al., 2004). As such, ASO infusions can be used to confirm the involvement of bdnf and egr1 in aspects of second-order fear conditioning in the BLA, and any effects of these infusions can be contrasted with the absence of a cycloheximide effect to identify ways in which bdnf and egr1 support information processing in second-order fear conditioning (Lay et al., 2018; Leidl et al., 2018; Williams-Spooner et al., 2019; Leake et al., 2024). For example, bdnf mRNA may have increased stability or enhanced translation efficiency under certain conditions, allowing for its continued synthesis even in the presence of cycloheximide (Lu et al., 2007; see also Lauterborn et al., 1996). Alternatively or additionally, bdnf and/or egr1 may function to incorporate second-order fears into existing first-order fear memories and do so by interacting with existing stores of cellular proteins (e.g., those that typically maintain short-term memory; Schafe et al., 2000; Bekinschtein et al., 2008; Leidl et al., 2018; Williams-Spooner et al., 2019; Leake et al., 2024); hence, the consolidation of second-order conditioned fear is unaffected by a BLA infusion of the protein synthesis inhibitor, cycloheximide (see below, Caveats and considerations re the effects of cycloheximide infusions in the BLA).
Our general hypothesis was that the bdnf and egr1 ASOs would disturb the network of genes that regulate second-order fear in the BLA and, thereby, disrupt its consolidation and reconsolidation. More specifically, given the results of Experiment 1 (notably, the PCA) and prior findings by Lee et al. (2004), we hypothesized that the consolidation of both first- and second-order conditioned fear would require expression of bdnf but not egr1 in the BLA and, conversely, that the reconsolidation of both first- and second-order conditioned fear would require expression of egr1 but not bdnf in the BLA. Hence, we predicted that the consolidation of both first- and second-order conditioned fear would be disrupted by a BLA infusion of the bdnf ASO but unaffected by a BLA infusion of the egr1 ASO and, conversely, that the reconsolidation of both first and second-order conditioned fear would be disrupted by a BLA infusion of the egr1 ASO but unaffected by a BLA infusion of the bdnf ASO.
Experiment 2: the consolidation of first-order conditioned fear requires expression of bdnf but not egr1 in the BLA
The next set of experiments examined the roles of bdnf and egr1 in consolidating and reconsolidating first- and second-order fear in the BLA. We began by examining their roles in consolidating first-order conditioned fear. In Experiment 2a, rats received a BLA infusion of the bdnf ASO or a scrambled missense oligonucleotide (MSO) 1 h prior to a session of S1–shock pairings. In Experiment 2b, rats received a BLA infusion of the egr1 ASO or MSO 1 h prior to the session containing the S1–shock pairings. In both experiments, rats were tested for fear to the S1 48 h after its conditioning. Given the results of previous studies by Lee et al. (2004) and Maddox et al. (2010), we predicted that consolidation of first-order fear to S1 would require expression of bdnf but not egr1 in the BLA. Hence, we expected that test presentations of the S1 would evoke less freezing among rats in Group bdnf ASO compared with rats in Group bdnf MSO (Experiment 2a), but equivalent levels of freezing among rats in Group egr1 ASO and egr1 MSO (Experiment 2b).
The baseline levels of freezing in the context during conditioning and test sessions were low (<10%) and did not significantly differ between the groups in this experiment, or in any subsequent experiment (Fs < 2). Figure 5 shows the levels of freezing to S1 across its conditioning and at test. In both experiments, rats showed a significant increase in freezing to S1 across its pairing with shock (Experiment 2a, F(3,51) = 34.14, p < 0.001; Experiment 2b, F(3,54) = 389.14, p < 0.001). The levels of freezing during the conditioning session did not differ between ASO and MSO groups (Fs < 1, p > 0.05), and there was no significant group × trial interaction (Fs < 2, p > 0.05), indicating that the oligonucleotides did not have an acute effect on freezing per se. At test, rats that received the bdnf ASO infusion showed significantly less freezing to S1 than those that received the corresponding MSO infusion (F(1,17) = 5.43, p < 0.05). Conversely, there was no significant difference between rats that received the egr1 ASO or MSO infusion (F(1,18) = 2.21, p = 0.15). Together, these results indicate that the consolidation of first-order conditioned fear involves expression of bdnf, but not egr1, in the BLA.
Consolidation of first-order conditioned fear requires expression of bdnf but not egr1 in the BLA. A, Experimental timeline of conditioning and testing for rats that received the bdnf MSO/ASO or egr1 MSO/ASO. B, D, Mean (±SEM) levels of freezing to S1 across its pairings with shock. C, E, Mean (±SEM) levels of freezing to test presentations of the S1 alone.
Experiment 3: the reconsolidation of first-order fear requires expression of egr1 but not bdnf in the BLA; acquisition/consolidation of second-order fear requires expression of both bdnf and egr1 in the BLA
Experiments 3a and 3b examined the roles of bdnf and egr1 in consolidating second-order fear and reconsolidating first-order fear in the BLA. In each experiment, rats were exposed to a session of S1–shock pairings in Stage 1 (first-order conditioning) followed 48 h later by a session of S2–S1 pairings in Stage 2 (second-order conditioning). One hour prior to the second-order conditioning session, rats received a BLA infusion of either the bdnf ASO or MSO (Experiment 3a) or egr1 ASO or MSO (Experiment 3b). Over the next 2 d, rats were then tested for fear to the S2 alone and S1 alone. Given the results of the previous experiment and those reported by Lee et al. (2004) and Maddox et al. (2010), we predicted that consolidation of second-order fear to S2 would require expression of bdnf but not egr1 and that the reconsolidation of first-order fear to S1 would require expression of egr1 but not bdnf. Hence, we expected that (1) relative to rats in Group bdnf MSO, rats in Group bdnf ASO would show less freezing to S2 but equivalent freezing to S1 (Experiment 3a) and (2) relative to rats in Group egr1 MSO, rats in Group egr1 ASO would show equivalent freezing to S2 but less freezing to S1 (Experiment 3b).
The results of Experiment 3a are shown in the top panel of Figure 6. Freezing to S1 increased across its pairings with shock in Stage 1 (F(3,51) = 31.10, p < 0.001) and to S2 and S1 across their pairings in Stage 2 (S2, F(3,51) = 17.42, p < 0.001; S1, F(3,51) = 3.18, p < 0.05). There were no significant differences between bdnf ASO and bdnf MSO groups in levels of freezing to S1 or S2 during first- and second-order conditioning and no significant group × trial interactions (Fs < 1.5, p > 0.05). At test, rats in Group bdnf ASO froze significantly less to S2 than rats in Group bdnf MSO (F(1,17) = 13.93, p < 0.01). In contrast, the levels of freezing to S1 did not significantly differ between the two groups F(1,17) = 0.26, p = 0.62). Thus, the consolidation of second-order conditioned fear requires expression of bdnf in the BLA, but the reconsolidation of first-order conditioned fear occurs independently of this expression.
The acquisition/consolidation of second-order fear requires expression of both bdnf and egr1 in the BLA. In contrast, the reconsolidation of first-order fear requires expression of egr1 but not bdnf in the BLA. A, Experimental timeline of conditioning and testing for rats infused with the bdnf MSO/ASO (B–E) or egr1 MSO/ASO (F–I). Mean (±SEM) levels of freezing to the S1 across its first-order conditioning in Stage 1 (B, F), to the S2 and S1 across second-order conditioning in Stage 2 (C, G), and to test presentations of the S2 alone (D, H) and S1 alone (E, I).
The results of Experiment 3b are shown in the bottom panel of Figure 6. In Stage 1, freezing increased to S1 across its pairings with shock (F(3,69) = 65.60, p < 0.001). There was no overall difference in freezing between the two groups and no significant group × trial interaction (Fs < 3, p > 0.05). In Stage 2, freezing increased to S2 across its pairings with the conditioned S1 (F(3,69) = 3.58, p < 0.05) while freezing to the S1 was maintained (F(3,69) = 2.20, p = 0.11). There was, however, a significant difference in overall freezing between the two groups with rats in Group egr1 ASO showing significantly less freezing to both S2 (F(1,69) = 10.49, p < 0.01) and S1 (F(1,69) = 21.84, p < 0.001) than those in Group MSO. These differences between the two groups persisted across trials, with the statistical analyses showing no significant group × trial interactions (Fs < 1.5, p > 0.05). Importantly, the effect of the egr1 ASO infusion on freezing to both S2 and S1 was also observed during the final test sessions, with rats in Group egr1 ASO again showing significantly less freezing to both S2 (F(1,23) = 19.02, p < 0.001) and S1 (F(1,23) = 10.43, p < 0.01) than those in Group MSO.
The results of these experiments show that, within the BLA, bdnf and egr1 regulate different aspects of first- and second-order conditioned fear. The expression of bdnf is required to consolidate second-order fear to S2 but is not required to retrieve/express or reconsolidate first-order fear to S1. In contrast, the expression of egr1 is required to retrieve/express and reconsolidate first-order fear to S1: knockdown of this gene produced a persistent disruption of freezing to S1 in the final test session. It is also required for the acquisition/consolidation of second-order fear to S2, though this may be a consequence of its disruptive effect on the first-order conditioned S1.
Experiment 4: the reconsolidation of second-order conditioned fear requires expression of egr1 but not bdnf in the BLA
Experiments 4a and 4b assessed the involvement of bdnf and egr1 in the reconsolidation of second-order fear. In each experiment, rats were exposed to a session of S1–shock pairings in Stage 1, a session of S2–S1 pairings in Stage 2, and a session in which the second-order conditioned S2 was briefly re-exposed in Stage 3. One hour prior to the brief re-exposure session, rats received a BLA infusion of either the bdnf ASO or MSO (Experiment 4a) or egr1 ASO or MSO (Experiment 4b). All rats were subsequently tested for fear to S2 alone and then S1 alone. Given the findings by Lee et al. (2004) that hippocampal bdnf and egr1 have dissociable roles in consolidating and reconsolidating context conditioned fear, we hypothesized that reconsolidation of fear to the second-order conditioned S2 would require expression of egr1 but not bdnf in the BLA. Hence, we expected that test presentations of the S2 would elicit equivalent freezing among rats in Groups bdnf ASO and bdnf MSO (Experiment 4a), but less freezing among rats in Group egr1 ASO compared with rats in Group egr1 MSO (Experiment 4b).
Figure 7 shows the levels of freezing across first- and second-order conditioning and averaged across the two re-exposures to the second-order conditioned S2 in Experiments 4a (top panel) and 4b (bottom panel). In both experiments, there was a significant increase in freezing to S1 across its pairings with shock in Stage 1 (Experiment 4a, F(3,45) = 59.63, p < 0.001; Experiment 4b, F(3,57) = 71.08, p < 0.001) and to S2 across its pairings with S1 in Stage 2 (Experiment 4a, F(3,45) = 3.57, p < 0.05; Experiment 4b, F(3,57) = 9.10, p < 0.001). Freezing to S1 was maintained across the four S2–S1 pairings and did not differ significantly across trials in either experiment (Fs < 1.6, p > 0.05). There were no significant group differences in levels of freezing to the S1 or S2 across their conditioning and no significant group × trial interactions in either experiment (Fs < 1.5, p > 0.05). During S2 re-exposure, there were no significant differences between groups, indicating that there was no acute effect of the infusions on levels of freezing to the S2 (Experiment 4a, F(1,15) = 0.13, p = 0.72; Experiment 4b, F(1,19) = 1.62, p = 0.22). This is notable given the acute effect of the egr1 ASO infusion on freezing to S1 in the previous experiment (see Discussion).
Reconsolidation of second-order conditioned fear requires egr1 but not bdnf in the BLA. A, Experimental timeline of conditioning and testing for rats infused with bdnf MSO/ASO (B–F) or egr1 MSO/ASO (G–K). Mean (±SEM) levels of freezing to presentations of the S1 across first-order conditioning (B, G), to the S2 and S1 across second-order conditioning (C, H), to the S2 across reactivation (D, I), and to test presentations of the S2 (E, J) and S1 (F, K).
The test results for Experiments 4a and 4b are shown in the rightmost panels of Figure 7. In Experiment 4a, rats in Groups bdnf ASO and bdnf MSO displayed equivalent levels of freezing to the S2 (F(1,15) = 0.07, p = 0.80) and the S1 (F(1,15) = 1.26, p = 0.28). In Experiment 4b, rats in Group egr1 ASO showed significantly less freezing to the S2 than rats in Group egr1 MSO (F(1,19) = 9.08, p < 0.01). There was no significant difference between these two groups in levels of freezing to the S1 (F(1,19) < 0.00, p > 0.05). Together, these results indicate that, within the BLA, the expression of bdnf is not required for the reconsolidation of second-order conditioned fear, whereas expression of egr1 is required for the reconsolidation of this fear.
It is worth noting that, as an immediate early gene, egr1 has been claimed to influence the reconsolidation of pavlovian conditioned fear by regulating the expression of late genes in the BLA. It has not been identified with short-term changes that drive performance. However, the protocols used to study the involvement of egr1 in learning and memory are typically much simpler than those used here. That is, in previous studies, rats received an injection of an egr1 ASO into the hippocampus or BLA and were then re-exposed to a context or auditory (tone) conditioned stimulus that had been previously paired with shock. In both cases, expression of the first-order fear responses (e.g., freezing) that had been conditioned to the context or tone was unaffected by the ASO injection (Lee et al., 2004; Maddox et al., 2010). Consistent with this prior work, here we showed that an injection of the egr1 ASO into the BLA also spared the expression of second-order conditioned fear to S2: this injection had no effect on freezing to the already-conditioned S2 when rats were tested for its reconsolidation. The ASO injection did, however, disrupt the expression of first-order fear to S1 during the acquisition of second-order fear to S2: that is, when the already-conditioned S1 was preceded by presentations of the novel S2, rats that received this injection did not freeze to the S1 and acquired less freezing to S2 than the missense controls. The implication of these results is that the involvement of egr1 in the retrieval/expression of pavlovian conditioned fear is influenced by other stimuli that are present. When a pavlovian conditioned stimulus is presented on its own, the expression of conditioned fear responses occurs independently of egr1 activation in the BLA. When a pavlovian conditioned stimulus is preceded by some other stimulus (S2), the expression of conditioned fear responses requires activation of egr1 in the BLA. Future work will seek to determine how other stimuli that are present influence the substrates of fear expression in the BLA: e.g., whether the influence of other stimuli varies with their novelty/familiarity or associative history.
A note on the selection of groups/controls
It is important to recognize that additional control groups could have been included in our experiments.
Controls for second-order conditioning
In Experiment 1, we could have included a group exposed to S1–shock pairings in Stage 1 and S1 alone presentations in Stage 2 and a group exposed to S1–shock pairings in Stage 1 and S2 alone presentations in Stage 2. Together, these groups represent the unpaired control which is used to confirm that freezing to the tone among rats subjected to S1–shock pairings in Stage 1 and S2–S1 pairings in Stage 2 is associatively mediated and not due to generalization of fear across stimuli. While these groups are potentially interesting, they were not included here for two reasons. The first reason is that these groups will learn about the stimuli in Stage 2 (the first group is extinguished to the S1, and the second group learns that S2 occurs without consequence), and the available evidence shows that this learning is processed in the BLA. That is, while these groups serve as useful controls for confirming that freezing to S2 is associatively mediated, it is not the case that learning does not occur in these groups—it does, and as this learning requires cellular processes in the BLA, there is no sense in which these groups could serve as a point of comparison for identification of genes that might be involved in second-order fear conditioning in the BLA (or, at least, not with the present methods). To identify such genes in Experiment 1, we selected groups matched to Group Second-Order for the training experiences in Stage 1 (Group First-Order Only) or Stage 2 (Group Sensory Preconditioning). The second reason for not including these control groups is practical. The current design involves between-subject factors of training condition (four levels) and time of sacrifice (three levels), yielding a total of 12 groups. If it had been expanded to include an additional two training conditions, it would have required 18 groups for completion, which was not feasible.
Controls for state-dependent learning
In Experiments 2–4, rats were tested drug-free after having received an ASO infusion in the BLA prior to first-order conditioning (Experiment 2), second-order conditioning (Experiment 3), or the S2 reminder session (Experiment 4). One potential explanation for the ASO-induced disruptions is that (1) the ASO infusion induced a particular state in the animals and (2) the learning that occurred under the influence of the ASO infusion was conditional upon this state. Hence, when rats were tested ASO-free (outside of the state created by the ASO infusion), they failed to retrieve what they had learned about S2 and S1 and did not freeze. However, this explanation is challenged by the finding that the test level of freezing to S1 was disrupted when the bdnf ASO was infused into the BLA prior to first-order conditioning, but not when the egr1 ASO was infused into the BLA prior to this conditioning, and, conversely, the test level of freezing to S2 was disrupted when the egr1 ASO was infused into the BLA prior to the S2 reminder session, but not when the bdnf ASO was infused into the BLA prior to this session. That is, the dissociation reported here renders the controls for state-dependent learning redundant.
Nonreactivation controls
Nonreactivation controls are typically used to confirm that a drug or compound only affects consolidation of first-order conditioned fear when administered proximal to (or in combination with) the session of CS–unconditioned stimulus pairings and/or reconsolidation of first-order conditioned fear when administered proximal to (or in combination with) a session of CS re-exposure. The dissociation reported here renders these controls redundant as the bdnf ASO infusion disrupted consolidation of first- and second-order conditioned fear but spared the reconsolidation of both fears, showing that the effects of this compound were only evident when it was administered into the BLA proximal to the session in which rats first learned about the S1 or S2. Similarly, the egr1 ASO infusion disrupted retrieval/reconsolidation of first-order conditioned fear (when administered prior to the second-order conditioning session) but spared the initial expression and consolidation of this fear (when administered prior to the first-order conditioning session), showing that the effects of this compound were only evident when it was administered into the BLA proximal to a session in which rats were re-exposed to the already-conditioned S1.
Caveats and considerations re the effects of cycloheximide infusions in the BLA
How are we to explain the finding that consolidation of second-order conditioned fear is disrupted by knockdown of bdnf in the BLA but unaffected by an infusion of the protein synthesis inhibitor, cycloheximide? One explanation is that the consolidation of second-order conditioned fear involves both transcription and translation of bdnf in the BLA and that the cycloheximide infusions in our previous studies simply failed to produce a sufficient level of protein synthesis inhibition to disrupt the bdnf-dependent consolidation of second-order fear. However, this account of our past and present findings is unlikely. Subsequent to the first demonstration by Lay et al. (2018), we have consistently failed to find any effect of a BLA cycloheximide infusion on consolidation of second-order conditioned fear in the protocol that was used for the present study (Leidl et al., 2018; Williams-Spooner et al., 2019; Leake et al., 2024). Specifically, across the four published studies in which we used this protocol, there are eight experiments in which rats received a BLA cycloheximide infusion immediately before/after a session in which they were exposed to pairings of a novel S2 and an already-conditioned S1. In each case, we failed to find any effect of the cycloheximide infusion on consolidation of fear to S2: relative to controls, rats that received the cycloheximide infusion displayed just as much freezing to S2 during the subsequent drug-free test. Indeed, we have “never” observed any effect of a BLA cycloheximide infusion on consolidation of fear to S2 in the protocol used for the present study but have consistently replicated the well-documented effect of these infusions on consolidation of first-order fear to S1: relative to controls, rats that received the BLA cycloheximide infusion immediately after a session of S1–shock pairings exhibited less freezing to S1 during the subsequent drug-free test. These dissociable effects of the BLA cycloheximide infusion on first- and second-order conditioned fear were obtained using between-subject designs where the infusion failed to affect the levels of second-order conditioned fear in one group but clearly disrupted the consolidation of first-order conditioned fear in another group (Lay et al., 2018; Leidl et al., 2018; Williams-Spooner et al., 2019). Importantly, they were also obtained using within-subject designs in which the same BLA cycloheximide infusion disrupted consolidation of first-order fear to S1 but spared the consolidation of second-order fear to S2, thereby ruling out any explanation of the results in terms of between-experiment variation in the drug infusion procedure (Leake et al., 2024). That is, the failure of a BLA cycloheximide infusion to affect consolidation of second-order fear to S2 cannot be due to ineffectiveness of the drug infusion procedure as the exact same procedure clearly disrupted the consolidation of first-order fear to S1.
Instead, there are three potential explanations for the failure of a BLA cycloheximide infusion to affect consolidation of second-order conditioned fear. The first explanation is that second-order conditioned fear exploits cellular processes that are involved in the protein synthesis-independent maintenance of first-order conditioned fear. That is, after rats have formed a first-order fear memory across the session of S1–shock pairings, the second-order fear memory that forms across the subsequent session of S2–S1 pairings may be consolidated to long-term memory via protein synthesis-independent processes that maintain first-order fear to S1. The second (and possibly related) explanation is that second-order conditioned fear does not require new neuronal connections for its successful retrieval and expression in defensive responses (Ryan et al., 2015). That is, while new connections are clearly needed for S1-evoked retrieval of first-order conditioned fear (which is disrupted when the session of S1–shock pairings is followed by a BLA infusion of cycloheximide), it is not needed for S2-evoked retrieval of second-order conditioned fear (which is unaffected when the session of S2–S1 pairings is followed by a BLA infusion of cycloheximide). The third explanation is that, relative to first-order conditioned fear, second-order conditioned fear engages a unique circuitry for its encoding and consolidation: hence, it is consolidated independently of de novo protein synthesis in the BLA but requires this synthesis in other regions of the brain. For example, the prelimbic region of the medial prefrontal cortex is not typically involved in the acquisition or consolidation of first-order conditioned fear to an S1 that is paired with footshock (Corcoran and Quirk, 2007) but is widely implicated in the retrieval/expression of this fear when rats are re-exposed to the S1 at test (Vidal-Gonzalez et al., 2006; Corcoran and Quirk, 2007; Burgos-Robles et al., 2009; Laurent and Westbrook, 2009; Sierra-Mercado et al., 2011; Dejean et al., 2016). As such, this region may play a critical role in the acquisition and consolidation of second-order conditioned fear when the novel S2 is paired with the already-conditioned S1.
Discussion
This series of experiments examined the mechanisms through which first- and second-order conditioned fears are consolidated and reconsolidated in the BLA. It followed from previous work demonstrating that there are differences in the molecular substrates of the two types of fear, e.g., whereas first-order fear is disrupted by a BLA infusion of the protein synthesis inhibitor, cycloheximide, second-order fear is unaffected by this infusion (Lay et al., 2018; Leidl et al., 2018; Williams-Spooner et al., 2019; Leake et al., 2024). Accordingly, we hypothesized that the consolidation of second-order fear may involve activation of immediate early genes that can be transcribed in the presence of protein synthesis inhibitors. We specifically anticipated the involvement of bdnf and egr1, as these genes have been shown to regulate the consolidation and/or reconsolidation of first-order fear in the hippocampus and BLA.
Experiment 1 used PCR to examine the expression of 12 genes in the BLA and PRh at multiple time points after second-order fear conditioning (light–shock pairings in Stage 1 and tone–light pairings in Stage 2). The 12 genes were selected based on their known involvement in first-order fear conditioning, their potential to regulate learning in the presence of cycloheximide, and a previous study which identified second-order fear conditioning with CaMK signaling and DNA methylation in the BLA (Lay et al., 2018). The results showed that second-order fear conditioning involves time-dependent changes in gene expression which differ from those produced by first-order fear conditioning alone. This was most evident in the principal components analysis of the 90 min PCR data which identified a network of genes that discriminated the rats that received second-order fear conditioning from those that received first-order conditioning only. This network included camKIIa, camkIIb, and dnmt3a, consistent with previous findings by Lay et al. (2018) which showed that consolidation of second-order fear is disrupted by a BLA infusion of drugs that inhibit CaMK signaling or the methyltransferases that catalyze DNA methylation. It also included pkcb1, mapk1, and map2k7. This is most likely due to the involvement of these genes in the extinction of context conditioned fear as Lay et al. (2018) showed that consolidation of second-order fear is unaffected by a BLA infusion of drugs that inhibit PKA/PKC or ERK/MAPK signaling.
The principal components analysis also identified second-order fear conditioning with expression of the immediate early genes, bdnf and egr1, in the BLA. The involvement of these genes was confirmed in Experiments 2–4, which used BLA infusions of antisense oligonucleotides (ASOs) to examine the effect of knocking down bdnf and egr1 on both first- and second-order fear conditioning. Experiment 2 examined the effect of these ASO knockdowns on the consolidation of first-order conditioned fear. Rats received a BLA infusion of either a bdnf ASO or egr1 ASO 1 h prior to a session of S1–shock pairings. Subsequently, all rats were tested for expression of first-order fear to S1. Relative to controls that received a BLA infusion of a missense sequence (MSO), rats that received the bdnf ASO showed less freezing to S1 at test, whereas those that received the egr1 ASO exhibited just as much freezing to S1. These results were taken to mean that, within the BLA, the consolidation of first-order fear is differentially dependent on bdnf and egr1: it requires the expression of bdnf but does not require the expression of egr1.
Experiment 3 examined the effect of the bdnf and egr1 ASO knockdowns on the retrieval and reconsolidation of first-order fear and acquisition/consolidation of second-order fear. Relative to MSO-infused controls, rats that received the bdnf ASO infusion showed less freezing to the S2 at test but just as much freezing to S1, whereas those that received the egr1 ASO infusion exhibited less freezing to both S2 and S1. Importantly, the egr1 ASO infusion also disrupted freezing to S2 and S1 across their pairings in second-order conditioning. Hence, these results have two major implications. The first is that, within the BLA, the retrieval and reconsolidation of first-order fear does not require the expression of bdnf but does require the expression of egr1. The second is that, within the BLA, the acquisition/consolidation of second-order fear requires the expression of both bdnf and egr1, though the egr1 requirement may be a corollary of its involvement in retrieving/expressing first-order fear to the S1.
Experiment 4 examined the effect of the bdnf and egr1 ASO knockdowns on the retrieval and reconsolidation of second-order conditioned fear. Relative to MSO-infused controls, rats that received the bdnf ASO infusion showed just as much freezing to both the S2 and S1 at test, whereas those that received the egr1 ASO infusion exhibited less freezing to the S2 but just as much freezing to the S1. Thus, within the BLA, the two genes have dissociable roles in yet another aspect of second-order fear conditioning: the reconsolidation of second-order fear does not require the expression of bdnf but does require the expression of egr1.
The findings in Experiments 2–4 are largely consistent with previous reports that, within the hippocampus, bdnf and egr1 have doubly dissociable roles in consolidating and reconsolidating a context fear memory. Specifically, knockdown of bdnf in the hippocampus disrupts consolidation but not reconsolidation of context conditioned fear, whereas knockdown of egr1 in the hippocampus disrupts reconsolidation but not consolidation of that fear (Lee et al., 2004; Monteggia et al., 2004; Rattiner et al., 2004; Besnard et al., 2013). The similarities between our findings and those by Lee et al. (2004) suggest that the specific roles of bdnf and egr1 in different aspects of memory are maintained across different brain regions. However, it should be noted that the present findings are only partially consistent with those of a prior study (Maddox et al., 2010) which showed that an ASO knockdown of egr1 in the lateral amygdala (LA) disrupts the initial consolidation of the fear memory produced by tone-shock pairings, as well as the reconsolidation of that memory when rats are briefly re-exposed to the already-conditioned tone. The difference between the present and past findings may be due to differences in the specific regions targeted in the two studies [here we targeted the BLA whereas Maddox et al. (2010) targeted the LA] and/or variation in the extent of egr1 knockdown: e.g., if the knockdown of egr1 in the LA was more effective in the prior study, this could explain why both consolidation and reconsolidation were disrupted in that study, and only reconsolidation was disrupted in the present study. It remains for future research to determine whether there are regionally specific roles for egr1 in consolidating versus reconsolidating conditioned fear in the LA and BLA and whether the two processes are differentially sensitive to knockdown of this gene in the LA.
The present findings could also be extended through a detailed assessment of the specific cells that are activated during first- and second-order fear conditioning. We have proposed that second-order fears are assimilated or incorporated into an already-established first-order fear memory (Leidl et al., 2018; Williams-Spooner et al., 2019; but see also Leake et al., 2024). This idea implies that the cells activated by S1–shock pairings in first-order conditioning may be preferentially reactivated during S2–S1 pairings in second-order conditioning: hence, the molecular requirements for consolidation of second-order fear are different from the molecular requirements for consolidation of first-order fear. Here, we extend this idea by proposing that, within the BLA, bdnf and egr1 regulate consolidation of first-order fear by directing changes in the morphology of activated cells and consolidation of second-order fear by regulating synaptic/behavioral “tag and capture”-type processes in cells that were activated by the prior first-order conditioning (Lu et al., 2011; Sajikumar and Korte, 2011; Almaguer-Melian et al., 2012; for reviews of “tag and capture” theories, see Frey and Morris, 1997; Morris, 2006; Ballarini et al., 2009; Viola et al., 2014; Moncada et al., 2015; Vishnoi et al., 2016; Nomoto and Inokuchi, 2018; Bin Ibrahim et al., 2024; see also Routtenberg and Rekart, 2005; Hernandez and Abel, 2008; Tanaka et al., 2008). This hypothesis is speculative but consistent with the findings reported here and previously, where second-order conditioned fear was disrupted by a BLA infusion of oligonucleotides that inhibit transcription of bdnf and egr1, but unaffected by a BLA infusion of the protein synthesis inhibitor, cycloheximide (Lay et al., 2018; Leidl et al., 2018; Williams-Spooner et al., 2019; Leake et al., 2024). It will be tested in future research.
In summary, the present study has identified the immediate early genes bdnf and egr1 as regulators of second-order conditioned fear in the BLA. It has specifically shown that these genes have distinct roles in consolidating and reconsolidating first- and second-order conditioned fears: bdnf is required for the consolidation of both first and second-order conditioned fear, whereas egr1 is required for the retrieval/expression of first-order fear and reconsolidation of both forms of fear. Future work will use high-throughput technologies to examine the full array of genes that regulate consolidation of second-order conditioned fear in the BLA (e.g., RNA sequencing) and how these genes regulate that consolidation (e.g., through “tag and capture”-type processes across multiple regions of the brain).
Footnotes
This work was supported by an Australian Research Council (ARC) Discovery Grant to N.M.H. and R.F.W. (DP200102969), an ARC Future Fellowship to N.M.H. (FT190100697), and an Australian Government Research Training scholarship to A.S.
The authors declare no competing financial interests.
- Correspondence should be addressed to Nathan M. Holmes at n.holmes{at}unsw.edu.au.
