The Rostral Anterior Cingulate Cortex Modulates the Efficiency of Amygdala-Dependent Fear Learning

Abstract

Background

The rostral anterior cingulated cortex (rACC) and the amygdala consistently emerge from neuroimaging studies as brain regions crucially involved in normal and abnormal fear processing. The aim of this study is to investigate a possible top-down modulation of a specific rACC sub-region on the acquisition of auditory fear conditioning.

Methods

We performed excitotoxic lesions, temporal inactivation and activation of a specific sub-region of the rACC that we identified by tracing studies as supporting most of the connectivity with the basolateral amygdala (r_Amy-ACC). The effects of these manipulations over amygdala function were investigated using a classical tone-shock associative fear conditioning in the rat.

Results

Excitotoxic lesions and transient inactivation of the r_Amy-ACC pre-training selectively produced deficits in the acquisition of the tone-shock associative learning (but not context). This effect was specific for the acquisition phase and did not required protein synthesis in the r_Amy-ACC. However, it could be overcome if an overtraining procedure was applied. Conversely, pre-training transient activation of the r_Amy-ACC facilitated tone-shock associative learning and interfered with further extinction processes.

Conclusion

Our results suggest that a subregion of the rACC is key to gating the efficiency of amygdala-dependent auditory fear conditioning learning. Since r_Amy-ACC inputs were confirmed to be glutamatergic, we propose that recruitment of this brain area might modulate overall BLA excitatory tone during CS-US concomitant processing. In the light of clinical research, our results provide new insight on the effect of inappropriate rACC recruitment on emotional behaviours.

Introduction

It is now well established that anxiety disorders arise as a consequence of alterations in amygdala-dependent fear learning processes [1–5]. Fear learning has been investigated using classical Pavlovian fear conditioning paradigms, which consists of pairing a neutral conditioned stimulus (CS) such as a tone, with an aversive unconditioned stimulus (US) such as a footshock. Upon subsequent exposure, the CS is then perceived as aversive and provokes a fear response [6]. Using this paradigm, brain areas engaged in the perception and response to threat were identified [1–4,7,8]. While the amygdala was found to be central for the CS to acquire aversiveness, the hippocampus primarily processes information associated with the fear context [7,9,10]. In addition, cortical areas such as the medial prefrontal cortex (mPFC) have been shown to play a critical role in the processing of emotional and cognitive information [11–15]. Of all cortical areas, the rostral anterior cingulate cortex (rACC), the affective division of the ACC [16–18], is ideally positioned between limbic and cortical structures to integrate emotion and cognition and is thereby primed to influence amygdala-dependent learning [17–25]. Neuroimaging studies have revealed a key role of the rACC in the normal processing of emotion and learning [16], but abnormal responses following provocation studies in patients suffering from various psychopathologies were also reported [26–30]. Interestingly, in fear conditioning studies, activation of the rACC occurs following the learning of repeated CS-US pairings [31–33], suggesting some functional interplay between the rACC and the amygdala specifically during the CS-US association. That the rACC may somehow control aspects of amygdala-dependent fear learning is further indicated by the coincident changes in neuronal activity detected within both these structures in affect-related processing [15,17,31,32,34–39].

Previous studies reported that the rACC is involved in the consolidation of inhibitory avoidance memories [40,41], in the acquisitions of fear avoidance learning [42,43], conditional visual discrimination [44] and appetitive conditioning [45]. Although there is a the large corpus of anatomical data on ACC-amygdala connectivity [18–22,46–49], very little is known about the rACC-amygdala interaction during the acquisition of auditory fear conditioning may be due to the fact that the ACC is involved in several processes such as anticipation, pain, attention, error monitoring and effortful recall [17,29,50–54]. Furthermore, lesions of this area led to different outcomes depending on the site and extent of the lesion, as well as the behavioural task used. However, in light of the above mentioned clinical reports studies focused on such interactions are especially prescient.

Here, we focused on the effects of permanent or temporary inactivation of a subregion of the rACC on the acquisition or expression of amygdala-dependent fear learning. We also sought to confirm that this subregion encompasses the full area of reciprocal connectivity between the two structures. Together our data demonstrate that a subregion of the rACC, which we term r_Amy-ACC, is key for establishing the efficacy and strength of auditory-associated fear learning.

Methods and Materials

Animals

Male Sprague-Dawley rats (Charles River, France) weighing 275–300g were group-housed on a 12-hour light-dark cycle with food and water ab libitum. The animals were allowed to habituate for 4 days before surgery. All experiments were carried out in accordance with the Veterinary Authority of Basel-Stadt, Switzerland.

Reagents

Ibotenic Acid (1.5M, IBO, Sigma Aldrich) was dissolved in PBS (0.1M) and adjusted to a PH of 7.2–7.4 using 1.0M NaOH. Muscimol (10nmol/μl, Sigma Aldrich) and Bicuculline Methiodine (50pmol/μl, Sigma Aldrich) were dissolved in NaCl. Anisomycin (50ug/0.5 μl, Sigma Aldrich) was diluted in equimolar HCL, diluted 50/50 with ACSF and the PH was adjusted to 7 with NaOH. DiI crystals were from Molecular Probes (USA) and Fast Blue (3–5% solution in ddH₂O) from Polysciences (Germany).

Surgery

Rats were anesthetized with a isoflurane-oxygen/nitrous oxide mixture. Surgeries were performed using a TSE Systems stereotaxic apparatus (Germany). For Tracing experiments, a 28 gauge injection cannula (Plastics One, Roanoke, VA) connected via a 0.28mm ID polythene tubing (SIMS Portex) to a microinfusion pump (CMA, Sweden) was either filled with DiI crystals diluted in PBS or a Fast Blue solution. The cannula was directed towards the rACC using the following coordinates from bregma (Paxinos and Watson – 4^th edition): AP: +2.7, +1.7, +0.7, 0; ML: ± 0.5, DV: 2.2, 2.8 and a total volume of 0.3μl was injected over a 5 minute period bilaterally. Tracers were allowed to diffuse for 3–5 weeks before the animals were sacrificed. For lesion studies, IBO (0.2μl) or 0.1M PBS were microinjected at the following coordinates: AP: +2.7, +1.7; ML: ±0.5; DV: 2.2, 2.8, over 5 minutes. The cannula was left in place for another 5 minutes to allow for complete diffusion. For intra-rACC injections, two chronic guide cannulae per side (33-gauge, Plastic ones) were inserted at the above AP and ML coordinates until they reached the surface of the Dura. The guide cannulae were then secured to the skull using surgical screws and bone cement. Stainless steel dummy cannulae reaching out to the tip of the guide were inserted to keep it free of contamination during recovery time. The dummies were checked everyday to avoid blockage and to habituate the animals with insertion and removal procedures. All animals (tracing, lesion, sham and microinjection) were handled and monitored for their weight and behavioural testing started a minimum 10 days later.

Behavioural training and microinjections

Apparatus

The fear conditioning and testing experiments were performed using a computerized fear-conditioning system (TSE, Bad Homburg, Germany). The conditioning test chamber A (transparent Perspex) was placed inside a detection sensor frame. Each chamber is located into a sound-attenuating box equipped with loud speaker (for auditory stimulus delivery and background white noise), light (10 watts), a ventilation fan and a camera mounted on the ceiling. The floor of the conditioning test chambers consisted of a removable stainless foot-shock grid connected to a shock delivery unit.

Conditioned stimulus (CS) and unconditioned stimulus (US) were controlled via the main computer. To dissociate between context and cue-induced freezing different chambers made of black Perspex with different grids and floors, were used on testing days (Chamber B). Freezing behaviour was automatically recorded and was defined as the absence of any movement for 1 sec.

Locomotor activity

The possible effects of manipulating rACC functionality on locomotor activity was assessed as previously described [55]. Briefly the animals were put in transparent Plexiglas boxes (dimensions: 19 × 31 × 16 cm) and the horizontal activity was detected using infrared light beam along the x-, y-, and z-axes (TSE Moti system TSE, Bad Homburg, Germany) registering the animal’s total immobility time and total distance traveled over ninety minutes.

Pain Behaviour

Changes in pain sensitivity to the US following all treatments were determined by measuring the induction of rapid movements to increasing footshock intensities from 0 to 0.6 mA. Cumulative probability plots correspond to the probability that a specific footshock intensity induces movement reaction (1 being 100% of the time).

Conditioning

Rats were handled for 3 days before behavioural manipulations. On day 1, the rats were habituated for 15 min to chamber A prior to conditioning. Conditioning trial consisted of a 15 sec, 4.5 kHz, 85 dB tone (CS) that co-terminated with a 0.6mA, 1 sec footshock (US). Pairing was repeated 6 times at random inter-trial intervals between 30 and 90 sec. After the last pairing, rats were returned to their home cage. Conditioning test chambers were cleaned with 70% ethanol after each trial. Unless stated otherwise each result expressed in bar graphs in the same graph figure result from separate experiments.

Testing

On day 2, rats that underwent training on day 1 were split in two groups. One group was placed in chamber B and presented with 3 tones and the other group in the original training chamber A for contextual testing. Rats were videotaped for subsequent behavioural scoring. Chambers B were cleaned with Thedra (a window cleaner) between each trial and Chamber A with ethanol.

Extinction

Each extinction trial (one per day over 3 days) consisted of 6 presentations of the CS alone (see above). On day 4 % freezing to 3 tone presentations was assessed and averaged.

Microinjections

For all experiments, after removal of the dummy, injectors which were connected to a microinfusion pump (CMA, Sweden) were inserted in the 4 guide cannulae (2 per side). Drugs or vehicle were microinjected at a rate of 0.2μl/2 min (0.2μl total volume par injection site). The cannula was left in place for a remaining of 5 min. For intra-ACC inactivation, activation, vehicle and protein synthesis blockade experiments, microinjections were performed 5 minutes prior to the training session. The habituation phase of the training protocol was then used as a baseline activity and locomotor measurement between the drug and control groups. For experiment 4, muscimol and PBS were injected 5 min prior to testing.

Tracing and histology

Rats were anesthetized with a combination of 75 mg/kg ketamine (Narketan, Vétoniquol, Switzerland) and 5 mg/kg xylazin (Rompun Bayer, Switzerland) and then perfused transcardially with 0.1 M PBS followed by 4% parafolmadehyde. The brains were post-fixed 4 hours in cold 4% paraformaldehyde, cryoprotected in sucrose gradients and frozen for cryosectioning. Twenty μm thick coronal sections were collected onto pre-coated slides, rehydrated in PBS, incubated for 1 hour with blocking solution containing 10% normal donkey serum, 0.2% bovine serum albumen and 0.3% Triton-X100 in PBS, and then incubated overnight with anti-vGLUT-1 or anti-vGAT (Synaptic Systems, 1,500) in blocking solution. Sections were then repeatedly rinsed in PBS and incubated with species-specific Cy3-conjugated antibodies (1:1,000; Molecular Probes) combined with Hoechst 33342 (10 μg/m, Sigma) in PBS and mounted. Pictures were collected with a Zeiss CCD digital camera connected to a Zeiss Axioplan2 microscope and processed using Adobe Photoshop 7.0. Hoechst and ethidium bromide (EthBr) were used to label all nuclei. The extent of DiI and fast blue labelling within the anterior-posterior axis of the BLA was examined using a Zeiss microscope interface with camera lucida. Each tenth Nissl-staining section was used to define boundaries of BLA cytoarchitecture and was superimposed to the adjacent section containing the fluorescent labelling for counting. For microinjection experiments toluedine blue was injected intra-rACC prior to killing for verification of the injection sites. Only animals with clearly defined bilateral lesions in the region of interest as shown by gliosis and neuronal loss were included in the analysis.

Statistical analysis

All reported statistical analyses were performed using the program SYSTAT (SPSS Inc, version 11). Effect of the lesions/drug microinjections on the mean percent freezing scores across each test trial were analyzed using a two-factors ANOVA (treatment × time) with time as a repeated measure. If significant, each time point was analyzed using a Student’s t-test. Freezing scores were averaged across trials for each rat and expressed as a percentage of the total tone presentation time. Total amount of time spent freezing during the conditioning and the memory retention test (24hrs and 5 days later following extinction) were compared using a Student’s t-test. All values are expressed as means ± s.e.m. statistical significance for all experiments was P< 0.05.

RESULTS

Effects of r_Amy-ACC lesions on fear learning

To investigate the physiological relevance for the r_Amy-ACC amygdala connectivity, we performed excitotoxic lesions of the r_Amy-ACC region and subjected the animals to an auditory fear conditioning protocol. Two bilateral infusions of ibotenic acid (IBO) resulted in specific fibre sparing lesions (Fig. 1a, b). Only animals that exhibited no damage to motor, limbic and prelimbic areas, and in which, the extent of the lesion was similar to that of the fluorescent dyes encompassing the area of r_Amy-ACC connectivity, were selected for analysis. The fear conditioning training consisted of six CS-US pairings (CS; 4.5 kHz, 15 s, 85dB – US; 0.6mA, 1s) separated by random time intervals to avoid anticipatory behaviours. The time course of learning was measured by the amount of freezing induced during each tone presentation. While both groups developed freezing behaviour as the number of associative pairing increased (Fig. 1c), the lesion group exhibited a slower acquisition across pairings (Fig. 1c). ANOVA analysis revealed a significant effect of pairings (F_5,100 = 5.5, *P < 0.05) in which pairing 5 and 6 were significantly different (*P < 0.05) as well as a treatment effect (F_1,20 = 6.044, *P < 0.05) and a significant interaction between pairings × treatment (F_5,100 =7.977, *P < 0.05). The learning impairment was also reflected by a deficit in the total time spent freezing (averaged over the 6 pairings) across the entire acquisition session (lesion: 14.2 ± 2.5%; n = 10 sham: 31.2 ± 4.1%; n = 12; *P < 0.05; Fig. 1d). 24 hours later, rats were tested for either contextual or cue fear expression. Consistent with reduced levels of initial learning, fear expression to the CS+ presentation was lower in the lesion group compared to sham (18.1 ± 4.5%; n = 10 vs. 48.3 ± 5.4%; n = 12; **P < 0.05; Fig. 1e). However, the expression of contextual fear was not affected by lesion of the r_Amy-ACC (sham: 51.1 ± 3.4% vs. lesioned: 49.2 ± 4.7%; n = 10; P >0.05; Fig. 1f). The rACC has an important role in motor function and in pain processing [16–18], we then asked whether changes in locomotor activity or in US perception could account for the learning impairment following r_Amy-ACC lesion. No differences were observed in immobility behaviour and total distance travelled between groups (Fig. 1g). Similarly, US-induced movements across pairings were not affected by r_Amy-ACC the lesion (Fig. 1h). Finally, no difference was found in the amount of exploratory behaviour during the baseline periods, prior to training, and on day two, when introduced to a new context (baseline groups of the cue expression test, Fig. 1d, e). Taken together, these results suggest that pre-training lesions of the r_Amy-ACC specifically modulates amygdala-dependent CS-US acquisition without interfering with the further expression of contextual fear.

Figure 1.

Lesion of the r_Amy-ACC produce deficit in fear acquisition. (a) Examples of the larger (light grey) and smaller (dark grey) extend of the lesions. (b) Representative phase contrast image showing a site specific lesion of the Cg1 (c) Lesion of the r_Amy-ACC (black squares: n = 10) exhibit a slower acquisition time course than sham (white dots: n = 12). (d) The mean time spent freezing (± s.e.m) during the acquisition phase was significantly reduced in lesioned animals (black dots, n = 10) compared to sham (n = 12). (e) Fear expression in lesioned animals was also impaired on day 2 There was no difference during baseline habituation (no tone) between groups. (f) Lesions of the r_Amy-ACC did not change locomotor activity as measured by the total immobility and total distance traveled. (g) There was no difference between the two groups in the response to the six consecutive shocks given at 0.6mA.

Transient inactivation of the r_Amy-ACC

To distinguish between a role of the r_Amy-ACC on the initial steps of learning (acquisition) or on the expression of conditioned responses, we microinjected the GABA_a receptor agonist muscimol (4.4 nmol/side, 0.20μl/side)[56,57] in the r_Amy-ACC either pre-training or 24 hours later before the expression test (Fig. 2a, b). Similarly to the effect of permanent lesions, transient inactivation of the r_Amy-ACC reduced the time course for the acquisition across pairings (two-way ANOVA, pairing effect F_5,70 = 13.176, *P < 0.05) with pairings 4,5 and 6 being significantly different (P < 0.05). There was also a treatment effect: F_1,14 =7.275, *P < 0.05, n = 8; and a significant interaction between the two: pairings × treatment F_5,70 = 8.14, *P < 0.05 (Fig. 2c). The averaged time spent freezing over the acquisition phase was reduced in the lesion group (muscimol: 13% ± 4.3; n = 8; Vehicle: 35.8% ± 5.1; n = 8, *P < 0.05; Fig. 2d). Fear expression measured on day two was also reduced in the muscimol group (18% ± 2.2; n = 8; vehicle 52.3% ± 6.5; n = 8; **P < 0.05; Fig. 2e). Blocking neuronal activity in the r_Amy-ACC did not change locomotor activities or the threshold for US induced movements compared to the vehicle group (Fig. 2f). We next tested whether the inactivation of the r_Amy-ACC could also affect fear expression. In a separate experiment, two untreated groups (group 1 and group 2) were first fear conditioned (Fig. 2g). 24 hours later the two groups either received muscimol or vehicle injection. No differences were observed on the levels of conditioned responses (Muscimol: 52.6% ± 4.7 Vehicle: 54.1% ± 5.2; P > 0.05, Fig. 2h). Taken together, these data suggest that inactivation and lesion of the r_Amy-ACC specifically impair fear acquisition but not fear expression.

Figure 2.

Fear learning is impaired by blocking neuronal activity in the r_Amy-ACC. (a) Example of injections sites for the muscimol microinjected group. (b) Phase contrast image showing two toluedine dark blue dots indicating the location of the tip of the cannulae. (c) Inactivation of the r_Amy-ACC also impaired fear acquisition. There was no effect of muscimol injection on baseline activity. (d) 24 hours after learning fear expression was also reduced without affecting habituation to a new context (no tone group). (e) Both groups showed no difference in the threshold for foot-shock induced behaviour. (f and g) Muscimol injection post-training did not affect freezing levels during fear acquisition and on day two during CS+ presentation.

Longer training exposure overcomes the effect of r_Amy-ACC inactivation

Previous studies suggested that the rACC may facilitate the process of “learning the rule” by changing the representation associated with the CS during learning and adapting for the optimal behavioural outcome [44,52,58]. It is therefore possible that the learning deficit observed following r_Amy-ACC inactivation, reflect a slower capacity for learning the CS-US associative rule. To test for this hypothesis, we doubled the number of pairings during training, subjecting both groups to 12x CS-US instead of 6x. During the acquisition, this protocol induced a ceiling effect in the vehicle group indicating that maximum learning had been reached (6x: 44 ± 5.1, n = 8 vs 12x: 55.2 ± 4.3, n = 8, Fig. 3a). The muscimol group, similar to the previous experiment (Fig. 2) showed a slower acquisition curve compared to vehicle, ANOVA analysis significant effect of pairing number (F_11,154 =12.73 *P < 0.05, n = 8), treatment (F_1,14 = 11.22 *P < 0.05, n = 8 both groups) and a significant interaction between pairing × treatment interaction (F_11,154 = 25.8 *P < 0.05) with most of the pairings except pairing number one and the last three, being significantly different in the muscimol group. However, for the last pairing (12) the muscimol group exhibited a similar percentage of freezing as for the vehicle group following 6 pairings (6^th pairing in the vehicle group: 44.5 ± 5.1 vs 12^th pairing in the muscimol group 43 ± 3.9, Fig. 3a) suggesting some learning recovery. There was still a significant difference across the whole acquisition session, between the two groups mainly due to the deficits observed in the first part of the acquisition. Consistent with the idea that more pairings would overcome the effect of r_Amy-ACC inactivation over learning, ear expression to the CS+ presentation was similar in both groups (vehicle: 58% ±4, muscimol: 53.5% ± 5.1 both groups n = 8). This further suggested that injection of muscimol in the r_Amy-ACC during acquisition does not impair but rather delays learning.

Figure 3.

(a) A longer training protocol could overcome the impairment in fear acquisition induced by r_Amy-ACC inactivation. (b) On day two, there was no difference between the lesion and vehicle group in conditioned responses to the CS+ presentation

Transient activation of the r_Amy-ACC

Our results provided evidence that activating GABA_A receptors in the r_Amy-ACC delayed the early stages of fear learning. Thus, we next sought to examine whether inhibiting GABA_A receptors in the r_Amy-ACC could facilitate learning. Injections of bicuculline [59] (50pmol/0.2μl) in the r_Amy-ACC pre-training resulted in a faster acquisition time course compared to vehicle (Fig. 4a, b). A two-way ANOVA showed a significant effect of pairings (F_5,90 = 21,49, *P < 0.05) with pairing numbers 2, 3 and 5 being significantly different (*P < 0.05), a treatment effect (F_1,18 = 6,49, *P < 0.05) and an interaction between the two (F_5,90 = 7,12 *P < 0.05, Fig. 4b) were also seen. The averaged freezing time over the whole acquisition was increased in the bicuculline injected group (Lesion: 58.4% ± 3.1 vs. vehicle: 35.9% ± 4.4, P < 0.05; n = 10 Fig. 4c). On day 2, higher conditioned fear expression was also found in this group compared to vehicle (70.7% ± 5.2 vs. 48.5% ± 4.4 P < 0.05 n = 10 Fig. 4d). These effects were unlikely to be due to a bicuculline-induced hyperactivity behavior as no difference in locomotor activity (data not shown) or baseline freezing measurements were observed between groups (Fig. 4d). In addition, the thresholds for US-induced movements were similar in both groups (Fig. 4e).

Figure 4.

Activation of the r_Amy-ACC facilitates and strengthens fear learning. (a) Example of bicuculline injection targeting the r_Amy-ACC. (b) The time course of the fear learning was facilitated by neuronal activation of the r_Amy-ACC. (c) There was a significant increase in the total amount of freezing across the conditioning phase in bicuculline treated group compare to vehicle. (c) On day two, freezing levels to the CS+ presentation were significantly increased in the bicuculline group compared to vehicle Bicuculline injection did not induce changes in baseline activity. Following three days of extinction protocol, the vehicle group showed significant reduction in freezing behavior to the CS+. In the contrary, the bicuculline group exhibited the same levels of freezing to the tone 24 hours and 4 days later suggesting that extinction was impaired. (d) No threshold difference in US-induced movements were observed between the vehicle and the bicuculline groups.

Both groups were then submitted to an extinction protocol consisting of 6x CS presentation each day for three days. The extinction of the conditioned response was assessed on day 4. While the vehicle group showed a significant reduction in freezing levels in response to the CS (before extinction 48.5% ± 4.4 vs. after extinction 34.5% ± 3.5, n = 10 Fig. 4d), the group that had learned under r_amy-ACC activation expressed similar freezing levels as when tested 24hrs after learning (before extinction 70.7% ± 5.2 vs. after extinction 68.5% ± 7.3 after extinction Fig. 4d). Collectively, these results support the idea that the r_Amy-ACC when recruited, allows fear to the CS+ to be acquired faster but also seems to be interfering with mechanisms underlying extinction processes.

Protein synthesis is not required in the r_Amy-ACC

Previous studies reported that besides the amygdala other brain areas exhibited protein synthesis dependent neuronal plasticity during the acquisition and consolidation of fear memories [60]. To investigate whether protein synthesis in the r_Amy-ACC was also required during learning we infused anisomycin [61,62] (50ug/0.2 μl) or vehicle pre-training. Levels of fear acquisition and expression did not differ between the two groups (Fig. 5a and 5b) suggesting that long-term plasticity in the r_Amy-ACC is not required for its modulatory action over learning.

Figure 5.

Blocking protein synthesis in the r_Amy-ACC pre-training did not affect fear learning (a) or fear expression (b)

Connectivity between the rACC and the amygdala

As our data indicate a key role of the rACC in the circuitry of fear, it was important to define if the area within the rACC (r_amy-ACC) which we are manipulating is encompassing the entire area of connectivity with the BLA and further define the nature of the top-down connections. We injected fluorescent neuronal tracers at four different coordinates unilaterally along the anterior-posterior axis of the rACC and examined sections across the amygdala for fluorescent signals three weeks later. Injection of Fast blue (FB) performed at AP: +2.7, +1.7 coordinates within the rACC (r_Amy-ACC) region, resulted in a clear labelling of neuronal cell bodies in the basolateral amygdala (BLA) (Fig. 6a and b), consistent with a retrograde tracing of axonal tracts. Injection of the tracer DiI resulted almost invariably in the labelling of disperse puncta in the BLA and unlike FB, it never labelled cell bodies, indicating that in our experiments DiI diffused in an anterograde manner [63,64] (Fig. 6a and b). No labelling was observed in the amygdala on the contraleral side of dye injections, suggesting that lateralization occurs (Fig. 6a). Injections of either tracer at AP: +1 and 0 did not show fluorescence in the BLA.

Figure 6.

Connectivity between the rACC and amygdala. (a) The fluoresecent neuronal tracers FB and DiI were stereotaxically injected at different anterior-posterior positions along the ACC. Only injections at AP:+2.7, +1.7 coordinates resulted in specific labelling in the BLA. Images show injection sites in the ACC and amygdala three weeks later. Higher magnification images reveal FB-positive neuronal cell bodies and processes and punctuated DiI-positive axonal terminals. Double labelling with the nuclear markers ethidium bromide (EthBr) or Hoechst was used to indicate the position of all nuclei. (b) Coronal sections along the anterior-posterior axis of the BLA indicate the areas of FB (blue dots) and DiI (red dots) labelling. Note that the areas occupied by the two tracers and non-overlapping and that their position changes along the anterior-posterior axis of the BLA.

Projections from the r_Amy-ACC to the BLA are excitatory

We next finally examined the nature of the inputs from the r_Amy-ACC and the amygdala by performing immunostaining of FB- or DiI-traced sections throughout the amygdala, with antibodies against the antigens vesicular GABA transporter (V-GAT) and the vesicular glutamate transporter (V-GLUT). These markers are expressed by GABAergic interneurons and by glutamatergic neurons respectively and they exhibit a punctuated pattern of expression along their neuronal cell body and processes. The large majority of DiI-positive puncta in the amygdala co-localized with V-GLUT-positive, but not with V-GAT-positive cells (Fig. 7a, b), suggesting that afferents from the r_amy-ACC are excitatory. The majority of retrogradely labeled FB-positive neurons in the amygdala expressed V-GLUT and were mostly V-GAT-negative, indicating that these cells are also excitatory and exhibit morphological similarities to BLA pyramidal neurons [19,65] (Fig. 7d, e).

Figure 7.

Section through the amygdala traced with either DiI or fast blue (FB) were immunolabeled with antibodies against V-GAT or V-GLUT. DiI-positive puncta do not co-localize with V-GAT (a, c), but they do with V-GLUT, co-localization is seen in yellow (b, c). a′ and b′ are higher magnification images of a and b. Most FB-positive neurons are negative for V-GAT (d, f) and express V-GLUT, co-localization in white (e, f). d′ and e′ are higher magnification images of d and e. Bar, 50 μm in a, b, d, e and 25 μm in a′, b′, d′, e′.

DISCUSSION

In this study, we provide evidence that lesion or transient inactivation of a specific sub-division of the rACC (which we referred to as r_amy-ACC) can modulate the efficiency of amygdala-dependent fear conditioned learning. This effect was transient and could be reversed by increasing the number of pairings during learning. Conversely, transient activation of the r_amy-ACC enhanced fear learning to the CS+ and prevented memory extinction. Since long-range r_amy-ACC afferents were found to be excitatory, these results support a role for the r_amy-ACC in the early steps of amygdala-dependent fear learning by possibly modulating the excitatory tone of the BLA.

Pre-training lesions as well as temporal inactivation of the r_amy-ACC specifically impaired the acquisition but not the expression of auditory fear conditioning indicating specific r_amy-ACC/BLA interactions during the early stages of fear learning. Since activation of the rACC has previously been shown to be involved in pain [43,50,66,67], changes in CS or US processing per se could influence the effects observed. However, this is unlikely to be the case under our experimental conditions as both pain threshold responses and contextual fear expression were unchanged following treatment. The fact that the amount of contextual fear expression was intact following lesions of the r_amy-ACC is in line with recent studies demonstrating that the rACC is preferentially involved in remote contextual fear memory storage and retrieval rather than during the acquisition phase [68–71]. As we also found reciprocal connectivity between the r_amy-ACC and the hippocampus (data not shown), it is possible that the r_amy-ACC subserves diverse functions depending on the behavioural cue versus contextual requirements. Interestingly, whilst our data clearly demonstrate that conditioned fear responses are contingent on r_amy-ACC function, our previous results indicated that the r_amy-ACC does not play a role in the manifestation of unconditioned anxiety behaviours in the elevated plus maze [72]. This may be explained by reports from lesion studies which indicate that the amygdala plays a crucial role in learned fear responses [9,10,73], but not in approach-avoidance behaviour in the plus maze [74,75]. Moreover, it reinforces the concept that multiple circuits underlie the manifestation of various symptoms relevant to the diverse anxiety disorders [76].

The learning impairment induced by r_amy-ACC inactivation was transient and could be overcome by doubling the number of pairings during the training session. It is unlikely reflecting a slow elimination of drug effect since muscimol was previously shown to remain stable in cortical tissue for up to two hours [77]. While we cannot exclude the possibility that increasing the number of pairings changes the essence of the task from conditioned fear learning to a repetitive chronic stress experience [42,44,53], this experiment provides evidence that in the absence of a functional r_amy-ACC fear learning can be delayed and that subsequent to sustained training its recruitment is compensated by other circuits.

In mice, lesions of the entire rACC specifically impaired trace but not delay conditioning [78] however, this same study also reported c-fos expression in the rACC following delay conditioning. Whether specific r_amy-ACC lesions would induce similar effects in the mouse model, have not yet been tested [51,79–81]. Considering the functional heterogeneity of this area, different observations following lesions might be accounted for by the different subregion of the rACC targeted, which differ in positions and extent. These results further support the need for carefully defining the exact lesion area of the rACC before experimentation.

Bicuculline microinjection into the r_amy-ACC accelerated learning while enhancing the levels of conditioned responses. Similarly to our previous observation with muscimol activation of the r_amy-ACC did not change pain sensitivity or baseline measurements compared to sham. Our results further indicated that the fear memory acquired under r_amy-ACC activation was resistant to our extinction protocol. Since we did not test for longer protocols, we cannot at this point determine whether the extinction process was permanently impaired or simply delayed. A large body of evidence indicated a functional role for the medial prefrontal cortex (mPFC: infralimbic, prelimbic) in the extinction of conditioned fear [82–85]. Considering that in the rat this region is extensively connected to the rACC [48,86], activation of the r_amy-ACC during learning might interfere with mPFC-mediated extinction processes. Blocking protein synthesis in the r_amy-ACC had no effect over fear acquisition and expression indicating that while this region can modulate learning, it is not itself a site of protein synthesis-dependent fear memory storage. In agreement with this, it has been shown that the contribution of the auditory thalamus to the consolidation of fear memories is also independent of protein synthesis, whereas signalling via the ERK/MAPK pathway is necessary [60,87]. Whether such mechanisms are also activated at r_amy-ACC-LA synapses during learning would require additional studies. While the identity of the postsynaptic targets of r_amy-ACC input to the BLA remains unknown, the fact that the afferents were found to be excitatory supports the hypothesis that the excitatory tone within the BLA can be modulated as a function of r_amy-ACC activity. Since fear conditioning involves the induction of long-term potentiation (LTP) at excitatory synapses of the LA [6], it would be interesting to investigate whether the r_amy-ACC can modulate associative learning by interfering with such mechanisms.

Although coupling between the r_amy-ACC and the BLA has previously been documented [20,22,46,88], our tracing experiments reported that the subregion of the r_amy-ACC was necessary and sufficient to support the full connectivity area between the two structures. Furthermore, our study also revealed that in all cases the connectivity between the r_amy-ACC and the BLA was ipsilateral. In this context, it is noteworthy that during the processing of fear memories in both rodents and humans the amygdala was functionally lateralized [89–92]. Moreover, responses in both the left amygdala and the left rACC were greater in masked affect perception studies, supporting a physiological role for the lateralization of the r_amy-ACC-amygdala axis [17,93]. It would be interesting to further investigate whether such anatomical lateralization underlies hemispheric specialization, as this has been proposed to facilitate learning by reducing the redundancy of information processed by each hemisphere [94–96].

In conclusion, our data provide new insight into the physiological relevance of the r_amy-ACC- amygdala axis during fear conditioning. While we found that inappropriate r_amy-ACC recruitment could either disrupt or facilitate amygdala-dependent fear learning, we also found that other processes such as memory extinction were affected. Since psychiatric disorders such as anxiety or post-traumatic stress disorders are likely to result from an overall network dysregulation, it will be important to further investigate the relationship of the rACC with other structures such as the infralimbic, prelimbic and orbitofrontal cortex during the processing, integration and recall of emotional information.