Contextual Fear Learning and Extinction in the Primary Visual Cortex of Mice

Abstract

Fear memory contextualization is critical for selecting adaptive behavior to survive. Contextual fear conditioning (CFC) is a classical model for elucidating related underlying neuronal circuits. The primary visual cortex (V1) is the primary cortical region for contextual visual inputs, but its role in CFC is poorly understood. Here, our experiments demonstrated that bilateral inactivation of V1 in mice impaired CFC retrieval, and both CFC learning and extinction increased the turnover rate of axonal boutons in V1. The frequency of neuronal Ca²⁺ activity decreased after CFC learning, while CFC extinction reversed the decrease and raised it to the naïve level. Contrary to control mice, the frequency of neuronal Ca²⁺ activity increased after CFC learning in microglia-depleted mice and was maintained after CFC extinction, indicating that microglial depletion alters CFC learning and the frequency response pattern of extinction-induced Ca²⁺ activity. These findings reveal a critical role of microglia in neocortical information processing in V1, and suggest potential approaches for cellular-based manipulation of acquired fear memory.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12264-022-00889-8.

Introduction

Accurate detection and evaluation of danger are crucial to survival. Classical fear conditioning (CFC), as one of the most powerful behavioral models used to study the neural mechanisms underlying danger prediction, has been widely explored [1]. In CFC, associative learning is established between an unconditioned noxious stimulus (US; for example, foot shock) and a neutral conditioned stimulus (CS; for example, context, oriented grating, and tone), which elicits a conditioned response (for example, freezing) to the neutral stimulus [1, 2]. Oriented grating, looming, or tone-cued fear conditioning has been confirmed to be closely related to multiple brain regions, including the visual and auditory cortex [3–11]. By contrast, the neural mechanisms of CFC induced by complex multiple sensory stimuli are more complicated.

Previous studies on CFC in experimental animals have pointed to several distinct neural circuit mechanisms [7, 12–16]. The formation and retrieval of memory are accomplished through different hippocampal pathways [13], and parvalbumin hippocampal interneurons are critical cellular elements involved in the persistence of fear memory [17]. Extinction of fear memory requires the formation of new engram cells which are formed and stored in basolateral amygdala (BLA) protein phosphatase 1-regulatory inhibitor subunit 1B-expressing neurons [12]. In addition to the well-explored bottom-up entorhinal/hippocampal and amygdala system, the top-down medial prefrontal cortex projections to the hippocampus also reveal a sparsely implemented memory-retrieval mechanism and support bidirectional communication during memory consolidation [18]. However, vision is a rich source of information about the presence of contextual threats, and its essential function in CFC is poorly known.

V1 is known to participate in processing information about static and moving objects, and is excellent in pattern recognition [19–23]. In experiments related to human vision, fear cues lead to enhanced performance on visual tasks [24] and stronger functional magnetic resonance imaging (fMRI) signals in various visual cortices [25]. As the emotional center, the amygdala sends extensive feedback projections directly to visual cortices in primates and cats, including the high-order cortex and primary visual cortex [26, 27]. Using intrinsic signal optical imaging, a study reported that activation of the amygdala enhances the visually-evoked responses of the primary visual cortex in the cat [28]. In rodents, visually-cued fear learning leads to the activation of a neuronal population in V1. Following activation of this visual engram population, mice respond in a manner similar to that following the visual presentation of fear cues [3]. These results suggest that V1 plays an important role in CFC training in an experience-dependent manner.

Experience-dependent cortical responses cause changes in synaptic structure and neuronal activity [5, 29–32]. For example, monocular deprivation causes rapid spine elimination and neuronal activity reduction in V1 [30], motor learning and novel sensory experience promote rapid dendritic spine formation in the motor cortex [33], and auditory-cued fear conditioning induces an increase in bouton formation in lateral amygdala axons that terminate in layer I of the auditory cortex [5]. Therefore, we hypothesized that CFC causes structural and functional changes in V1. In the present study, we tested this hypothesis using chemogenetics, presynaptic axonal bouton imaging, and Ca²⁺ imaging. Given the role of microglia in mediating memory [32, 34, 35], we investigated whether microglia are involved in the plasticity of neuronal activity induced by CFC. Our results revealed a change in presynaptic structure and neuronal activity pattern involving microglia in V1 as a mechanism underlying CFC learning and extinction.

Materials and Methods

Experimental Animals

Thy1-YFP-H transgenic mice (Jackson Laboratory, Bar Harbor, USA) and C57BL/6J mice (Shanghai Laboratory Animal Center, Shanghai, China) were housed 2–5 per cage under a 12-h light-dark cycle (08:00 lights on, 20:00 lights off) at constant temperature (23°C) and humidity (~50%) with ad libitum access to food and water. Eight to ten-week-old male mice at the initiation of the experimental procedure were used. The appropriate guidelines and protocols for the care and use of laboratory animals were approved by the Committee for Animal Experiments of Zhejiang University.

Behavior Protocol

Fear conditioning and extinction were applied in a training chamber equipped with stainless-steel shocking grids connected to a precision feedback current-regulated shocker within a sound-attenuating box (Coulbourn Instruments, Allentown, USA). Aversive foot-shock stimuli were automatically presented using FreezeFrame software (Coulbourn Instruments). The chamber was cleaned with water before experiments. Behavior was recorded by low-light video cameras and measured by an automated scoring system.

Fear Training

Mice were trained twice, 24 h apart. During each training session, 2 min after mice were placed in the chamber they received 4 weak foot-shocks (0.5 mA, 2 s duration, 1 min apart). Mice were taken out of the training chamber 1 min after the last foot-shock and returned to their home cages. After the last session, mice were housed under standard conditions until the retrieval test 24 h later.

Retrieval Test

Mice were placed in the fear conditioning chamber and left for 5 min.

Fear Extinction Training

Mice were trained twice, 24 h apart. During each training session, mice were placed in the fear conditioning chamber and left for 15 min and then housed under standard conditions until the extinction test 24 h later.

Extinction Test

Mice were placed in the fear conditioning chamber and left for 5 min.

Untrained Control Group

Mice were habituated to the training chamber without shocks 6 times, 5 min each time, 24 h apart.

Virus Injection and Cranial Window Implantation

Mice were anesthetized with isoflurane (4% induction, 1–2 min; 1%–2% maintenance throughout the surgery) and fixed in a stereotaxic frame (Reward Technology Co., Ltd, Shenzhen, China). Body temperature was maintained at 37°C by a heating pad. Ophthalmic ointment was applied to prevent the eyes from drying. Lidocaine was injected subcutaneously as a local anesthetic. After shaving and sterilizing the skin with 75% ethanol, a midline incision was made with a sterile scalpel. Then surgical scissors were used to further expose the skull. The skull surface was wiped and cleaned with cotton swabs. Bregma and lambda were identified and leveled to the same Z-axis. Using a dental drill (0.5 mm in diameter), a small craniotomy (0.5 mm × 0.5 mm) was made above the primary visual cortex at 1.0 mm anterior to lambda, −2.5 mm lateral to the midline. We inserted a glass micropipette with a tip diameter of 10–20 μm to infuse the virus, starting at a depth of 0.35 mm from the pial surface.

For bilateral AAV2/9-hSyn-hM4D(Gi)-mCherry or AAV2/9-hSyn-mCherry (Taitool Bioscience, Shanghai, China) injection, ~150 nL of virus solution was slowly injected. After the injection, the micropipette was maintained in place for 10 min before retraction to prevent leakage. The scalp incision was closed with tissue glue (3M Animal Care Products, Saint Paul, USA). The experiments were performed 3–6 weeks after virus injection.

For AAV2/9-hSyn-jGCaMP7s (Taitool Bioscience) injection followed by cranial window implantation, ~80 nL of virus solution was slowly injected. Following the injection, the micropipette was maintained in place for 10 min before retraction. Then a craniotomy (~3 mm × 3 mm) was performed centered on the virus injection site. A double-layered coverslip consisting of a small coverslip (3 mm in diameter) attached to a larger one (6 mm in diameter) was embedded and sealed with dental cement. The small coverslip fitted snugly into the craniotomy, and the larger one was attached to the polished skull. The last step used dental cement to fix a head plate on the skull. The mice were placed on a heating pad until they were fully awake. Antibiotics were injected for at least one week after surgery. The experiments were performed 2–4 weeks after virus injection.

Cranial windows were implanted in Thy1-YFP-H transgenic as above. Antibiotics were injected for at least one week after surgery. Mice were allowed ~3 weeks to recover from the surgery.

CNO Application

Clozapine-N-oxide (CNO) from Sigma-Aldrich (Cat# C0832, St. Louis, USA) was dissolved in DMSO (10 mg/mL) and then diluted in 0.9% saline to make a final concentration of 0.5 mg/mL. CNO was injected intraperitoneally (5 mg/kg) 30–40 min before the fear learning or retrieval test.

Selective Elimination of Microglia

PLX3397 from Selleck (Cat# S7818, Shanghai, China) was formulated in American Institute of Nutrition 76A (AIN-76A) standard chow (475 mg/kg) by Moldiets (Beijing, China). Mice were fed a PLX3397 diet for 4 weeks to eliminate microglia.

In vivo Two-photon Imaging

Images were captured on an Olympus two-photon microscope (FVMPE-RS, Tokyo, Japan) equipped with a mode-locked Ti:Sapphire laser (MaiTai DeepSee, Spectra-Physics, San Francisco, USA) set at 920 nm and a water-immersion objective lens (25×, N.A. 1.05; Nikon, Tokyo, Japan). The imaging protocol comprised 3 sessions. Session 1: baseline recording 1 day before fear learning. Session 2: functional or morphological plasticity recording after fear retrieval. Session 3: functional or morphological plasticity recording after fear extinction.

For axonal bouton plasticity imaging, Thy1-YFP-H transgenic mice were imaged through the craniotomy as above. Z-stacks (170 µm × 170 µm; 1024 pixels × 1024 pixels, 0.7-µm step size) of fluorescently labeled neuronal processes were taken by galvanometer raster scanning and used as a map for the relocation of the same area at later time points, in addition to the marked brain vasculature map. Images within a depth of 350 µm from the pial surface were collected at each session, yielding a full three-dimensional data set of axons in the area of interest.

For Ca²⁺ imaging, mice injected with GCaMP7s were imaged through the craniotomy as above. Z-stacks (509 µm × 509 µm; 512 pixels × 512 pixels, 1-µm step size) were taken by galvanometer raster scanning and used as a map for the relocation of the same area within a depth of 650 µm from the pial surface, in addition to the marked brain vasculature map. Time-series imaging (7.60 Hz, 1500 frames) was taken by resonant scanning within the Z-stack.

Data Analysis of Bouton Structural Plasticity

In transgenic Thy1-YFP mice, imaged axons were thinner and produced less fluorescence than dendrites and were thus easily identified. All images were analyzed using ImageJ (National Institutes of Health, USA). Only fields of view showing a good signal-to-noise ratio in all imaging sessions were considered for data analysis. For bouton identification, strict criteria were adopted as described in previous studies [36, 37]. En passant boutons (EPBs) are analyzed based on their intensity relative to the axon backbone. The fluorescence intensity was profiled along identified axons using ImageJ. Bright axonal varicosities were identified as EPBs when the peak intensity was >3 times brighter than the axonal backbone. For bouton elimination, the intensity of boutons had to drop below 1.3 times the neighboring axon backbone intensity. For formation and elimination rate analysis of boutons, reconstructed 3D images were used to minimize image distortion caused by movements and rotation between imaging intervals. The formation and elimination rates of axonal boutons are defined as the number of boutons formed and eliminated divided by the total number of axonal boutons in the previous imaging session. We defined the turnover rate of boutons as the combined number of formed and eliminated boutons divided by the total number of axonal boutons in the previous imaging session.

Calcium Imaging Data Analysis

All data were analyzed using ImageJ and customized MatLab software (The MathWorks, Natick, USA). We first realigned the images from each session to a reference image (the average image of 20 frames in the middle of an imaging session) using a customized cross-correlation-based translation algorithm, to correct the X-Y offset of images caused by the slight head movements of the head-fixed awake mice.

The neuronal cell bodies (regions of interest, ROIs) were recognized from both their maximal intensity projection images and their activity in time-scan imaging. The ratio of fluorescence change (ΔF/F₀) of these ROIs was calculated for each activated neuron. ΔF/F₀ = (F – F₀)/F₀, where F was calculated by averaging the corresponding pixel values in each specified ROI and the baseline fluorescence F₀ was assessed as the 25^th percentile of the entire fluorescence recording.

To identify significant Ca²⁺ events, we used a peak-detection algorithm that identified maxima in the derivative of the ΔF/F₀ signal implemented in MatLab (peak finder). The identified maxima must be above a threshold, defined as >3 SD. of the entire ΔF/F₀ distribution. Then frequency was calculated by the number of events per min.

Histology

To confirm the viral expression and microglia depletion, all of the animals were anesthetized with ketamine (120 mg/kg) and xylazine (10 mg/kg) intraperitoneally and perfused transcardially with 20 mL phosphate-buffered saline (PBS) and 40 mL 4% paraformaldehyde (PFA). The fixed brains were removed and postfixed in 4% PFA for 24 h, dehydrated in 30% sucrose in PBS, and cut into 50-μm coronal sections. To check virus expression, the sections were imaged using a VS120 Virtual Slide Microscope (Olympus). For immunofluorescence staining, brain sections were incubated in primary antibody (anti-Iba1, rabbit, 1:500, Wako, Japan) overnight at 4ºC, and then in secondary antibody (Alexa Fluor 488 goat anti-rabbit, 1:500; Abcam, England) for 2 h at room temperature (25ºC), in the presence of 1% bovine serum albumin (BSA) Albumin Fraction V and 0.2% (weight/volume) Triton X-100. The sections were mounted on slides with an anti-fade medium containing DAPI (Beyotime, Shanghai, China).

Statistical Analysis

The statistical test used, test statistics, and the P-values are shown in the figure legends. Five test methods were used throughout this work: unpaired Student’s t-test, one-way analysis of variance (ANOVA) followed by Tukey’s multiple comparisons tests or Fisher’s least significant difference (LSD) multiple comparisons tests, Mann–Whitney test, Kruskal–Wallis test followed by multiple comparisons tests, and two-way ANOVA test. Significance levels were set at P ≤0.05.

Results

V1 Plays an Important Role in Contextual Fear Memory Retrieval

In the neocortex, the neural mechanisms of CFC induced by complex multiple sensory stimuli remained unclear. Previous work has suggested that V1 is closely related to visually-evoked locomotion [3], so we used CFC to train mice to confirm whether V1 is involved in contextual fear memory. On day 1, conditioned mice were habituated to the context (CS) for 2 min and a baseline for freezing was established over the 2 min (described as naïve in the following text). Then mice were subjected to four 2-s long foot shocks (US, 0.5 mA) at 60-s intervals on two consecutive days. On day 3, mice were exposed to the context for 5 min to evaluate fear retrieval. On days 4 and 5, mice underwent extinction training by exposure to the context for 15 min per day in the absence of foot-shock. An extinction test was applied on day 6, when mice were exposed to the context for 5 min (Fig. 1A). The group of 6 mice started with a very low freezing response (Fig. 1B, trial #1, 0.60%), which then reached 59.30% (trial #9) after 2 days of fear learning. The freezing response was 71.11% (trial #10) in the fear retrieval test. After 2 days of extinction training, the freezing response dropped to 11.29% (trial #13) in the extinction test (Fig. 1B).

Fig. 1.

Inactivation of the primary visual cortex impairs contextual fear memory retrieval. A The behavioral test paradigm for contextual fear conditioning. Conditioned mice (including control, mCherry-injected, and hM4Di-injected mice) receive fear training for two consecutive days followed by retrieval on day 3, and then receive extinction training for two consecutive days followed by extinction test on day 6. mCherry or hM4Di mice are injected with CNO 30 min before fear retrieval or fear learning. B Contextual fear learning and extinction performance curve (n = 6 mice; mean ± SEM). C Bilateral AAV2/9-hSyn-hM4Di or AAV2/9-hSyn-mCherry viral injection sites in the primary visual cortex (dashed line, magnification of injection site; scale bars, left: 1 mm; right: 50 μm). D Freezing responses of conditioned mice on days 1 and 2 (no significant difference among the four groups). E Freezing responses in the retrieval test. CNO application in hM4Di-injected mice before retrieval induces a significant decrease in contextual freezing in the fear retrieval test, compared to the other three groups. F Freezing responses in the extinction test (no significant difference among the four groups). Data are presented as the mean ± SEM. Control, n = 6; mCherry-CNO before retrieval, n = 5; hm4Di-CNO before retrieval, n = 5; hm4Di-CNO before training, n = 4. *P <0.05, **P <0.01, ns, no significant difference, one-way ANOVA followed by Fisher’s LSD multiple comparisons tests.

To explore the involvement of V1 in CFC, we inactivated the visual cortex through bilateral AAV2/9-hSyn-hm4Di injection with intraperitoneal CNO injection before retrieval (hm4Di-CNO before retrieval) (Fig. 1A, C). Two control groups were set up as follows. The CFC group went through the behavioral protocol without viral injection or any other surgery. The mCherry-CNO before retrieval group was injected with AAV2/9-hSyn-mCherry and CNO was injected intraperitoneally before retrieval (Fig. 1A). At the same time, we were interested in whether inactivation of the visual cortex during fear learning had an impact on fear retrieval. Therefore, CNO was injected before learning (hm4Di-CNO before training). Freezing responses of the four groups in fear learning showed no significant difference (Figs 1D and S1), which excluded differences in fear retrieval caused by the differences in individual learning levels. The results also show that inactivation of V1 before fear learning did not affect the conditioned fear response. Subsequently, in the fear retrieval test, the hm4Di-CNO before retrieval group showed significantly fewer freezing responses than the other three groups (Fig. 1E. CFC, mCherry-CNO before retrieval and hm4Di-CNO before training vs hm4Di-CNO before retrieval, P = 0.02, P = 0.003, and P = 0.02, respectively). However, there was no significant difference in the extinction test among the four groups (Fig. 1F). Together, these results showed that V1 is required for contextual fear memory retrieval.

Bouton Turnover Increases After Contextual Fear Learning and Extinction in V1

That learning experience leads to changes in the dynamics of postsynaptic structure has been confirmed in many studies [29, 31]. Compared with the widely-studied postsynaptic structural changes, much less is known about the plasticity of presynaptic axonal terminals in vivo. Therefore, we investigated the impact of CFC on axonal bouton plasticity in V1.

We implanted cranial windows to image YFP-labeled axons of layer I and layers II/III in V1 with two-photon microscopy (Fig. 2A). The CFC group underwent a standard behavioral experiment as described above, while the control mice were exposed to the context for 6 consecutive days, for 5 min per day. The freezing responses and imaging of untrained control mice on days −1, 3, and day 6, corresponded to the naïve, retrieval, and extinction tests of the CFC group, respectively (Fig. 2A). To facilitate analysis and understanding, the results of untrained control mice are also described as naïve, retrieval, and extinction in this work. Starting with a very low freezing response, the CFC group displayed significantly more freezing responses after fear learning and identical freezing responses after extinction training compared to the untrained control group (Fig. 2B, control vs CFC: naïve, P = 0.06; retrieval, P <0.0001; extinction, P = 0.06; unpaired t-test). To image the same field of view, we injected SR101 intraperitoneally into mice, marking blood vessels for positioning (Fig. 2C). Our results showed that boutons on these axons were mostly EPBs [37], and very few terminal boutons were found in the field of view (FOV). So, EPBs of the same segmented axon were analyzed in the naïve, retrieval, and extinction test groups (Fig. 2D).

Fig. 2.

Contextual fear learning and extinction induce increased bouton turnover in the primary visual cortex. A Experimental timeline of the behavioral test paradigm and repeated in vivo axon bouton imaging in Thy1-YFP mice. Imaging before behavioral training (day −1) is regarded as the baseline. Imaging in the behavioral manipulation process is performed immediately after each behavioral test. Scale bar, 0.5 mm. B Freezing responses in CFC training mice and control mice. Control, n = 6; CFC, n = 7. C Examples of imaging showing YFP-labeled axons in control and CFC mice. Scale bar, 10 μm. D High magnification repeated imaging of the same axons corresponding to the boxed regions in C in control and CFC mice. Bouton elimination (compared with the next image, red arrowheads) and bouton formation (compared with the previous image, blue arrowheads) are identified. Scale bar, 5 μm. E Percentages of bouton turnover (including newly formed and eliminated boutons) in control and CFC mice in the retrieval test. Control, n = 75 axons from 5 mice. CFC, n = 79 axons from 7 mice. F Percentages of bouton turnover (including newly formed and eliminated boutons) in control and CFC mice in the extinction test. Control, n = 75 axons from 5 mice. CFC, n = 79 axons from 7 mice. Data are presented as the mean ± SEM. *P <0.05, ***P <0.001, ****P <0.0001, ns, no significant difference, unpaired t-tests.

We calculated the percentages of bouton turnover, formation, and elimination by comparing the images from naïve mice with those from retrieval mice, and by comparing the images from retrieval mice with those from extinction mice. A comparison between the CFC group and the untrained control group revealed a difference in axonal bouton dynamics. In the retrieval test, CFC mice displayed higher turnover, formation, and elimination than control mice (Fig. 2E, control vs CFC: turnover, 19.75 ± 1 vs 32.98 ± 1, P <0.0001; formation, 8.86 ± 1 vs 17.29 ± 1, P = 0.0002; elimination, 10.89 ± 1 vs 15.68 ± 1, P = 0.01). Consistent with the retrieval results, CFC mice that received fear training also displayed higher turnover, formation, and elimination than control mice in the extinction test (Fig. 2F, control vs CFC: turnover, 20.06 ± 1 vs 31.92 ± 1, P <0.0001; formation, 10.76 ± 1 vs 16.81 ± 1, P = 0.02; elimination, 9.30 ± 1 vs 16.75 ± 1, P = 0.0003). But no significant difference in bouton turnover, formation, and elimination was found either in control mice or in CFC mice (Fig. S2). Taken together, these results indicate that contextual fear learning and extinction induce an increase in axonal bouton turnover in V1, which is due to increases in both bouton formation and bouton elimination.

Contextual Fear Learning and Extinction Lead to Opposing Changes in the Frequency of Neuron Activity in V1

Synaptic structural dynamics are canonically used as a proxy for synaptic functional plasticity, and neuronal activity is critical for regulating synaptic plasticity [38]. Neuronal activity induced by CFC could be involved in promoting presynaptic turnover. Therefore, we explored fear learning and extinction-related neuronal activity using Ca²⁺ imaging of layer II/III neurons expressing the genetically-encoded Ca²⁺ indicator GCaMP7s in V1.

In this experiment, mice were injected with GCaMP7s in V1 and a cranial window was implanted at the same time (Fig. 3A, B). After 2–3 weeks, mice were either trained for CFC or untrained for control (Fig. 3C). To image the same field of view, we injected SR101 intraperitoneally to mark blood vessels for positioning. A standard slow galvo scan was used to obtain Z-stack images (Fig. S3A). The fast resonant scan (up to 32 frames per second) was used to obtain images of neuronal activity (7.60 frames per second by averaging every 4 frames). CFC and control mice were imaged immediately after naïve, retrieval, and extinction tests. We recorded the spontaneous somatic activity of layer II/III neurons for 197 s with mice in a head-restrained awake state, and show the Ca²⁺ signal traces of the same example neurons in Fig. 3D. In untrained control mice, the frequency of Ca²⁺ events was not significantly altered in the naïve, retrieval, and extinction tests (Fig. 3E, P = 0.36, Kruskal–Wallis test). By comparison, the frequency of Ca²⁺ events in CFC mice was significantly decreased after fear learning and increased after extinction (Fig. 3F, naïve vs retrieval, P = 0.03; retrieval vs extinction, P = 0.004; naïve vs extinction, P = 1.00; Kruskal–Wallis test followed by multiple comparisons test). Distribution histograms of the frequency of spontaneous Ca²⁺ events in control and CFC mice are shown in Fig. 3G, H. After the experiment was completed, the mice were perfused to confirm the virus expression (Fig. S3B). Thus, our results demonstrated that the frequency of Ca²⁺ activity in mice without learning experience remains at a constant level; however, mice with learning experience show changes in related patterns of neuronal activity.

Fig. 3.

The decrease in frequency of neuronal activity in the primary visual cortex induced by contextual fear learning is reversed by fear extinction. A Experimental design. C57 mice are injected with AAV2/9-jGCaMP7s in V1 and a cranial window is implanted at the same time. After 2–3 weeks, two-photon imaging is performed along with a control or CFC behavioral experiment. B Neurons in V1 expressing GCaMP7s. Analysis of neuronal activity in layers II/III of V1. Scale bar, 100 μm. C Freezing responses in CFC training mice and control mice. Control, n = 6; CFC, n = 6. Data are presented as the mean ± SEM. ****P <0.0001, unpaired t-tests. D Ca²⁺ activity of layer II/III neurons in the V1 after naïve, retrieval, and extinction tests in control (upper) and CFC (lower) mice. Three traces: Ca²⁺ activity of neurons indicated by the dashed circle. Ca²⁺ fluorescence traces for 197 s are shown. Scale bar, 30 μm. E Ca²⁺ signal frequency in naïve, retrieval, and extinction tests in control mice, n = 4. Kruskal–Wallis test, no significant differences are found. Error bars, mean ± SEM. F Ca²⁺ signal frequency in naïve, retrieval, and extinction tests in CFC mice, n = 6. A significant frequency change occurs after both fear learning and extinction. Error bars, mean ± SEM. *P <0.05, **P <0.01, Kruskal–Wallis test followed by multiple comparisons test. G, H Distribution histograms of the frequency of spontaneous Ca²⁺ events in control and CFC mice.

Absence of Microglia Alters Fear Learning and Extinction Induced Neuronal Activity Dynamics Pattern

Recent studies have suggested that mice depleted of microglia show deficits in many learning tasks, including auditory-cued fear conditioning [34]. Microglia-depleted animals exhibit significantly fewer freezing fear responses to the auditory cue during the recall test than non-depleted controls. To further understand the effect of microglial deletion on learning and memory, we trained microglia-depleted mice in CFC behavioral paradigms and recorded Ca²⁺ activity.

Feeding a chow diet containing the colony-stimulating factor 1 receptor (CSF1R) antagonist [39] PLX3397 (475 mg/kg) for 4 weeks resulted in almost complete elimination of microglia from the brain, as confirmed by the absence of the microglia marker Iba1 (Figs 4A and S4). Quantitative analysis showed that after 4 weeks, 99% of microglia were eliminated from the cerebral cortex. Iba1⁺ cells were counted in a 10× field of view from the cortex (Fig. 4B, control vs PLX: 72 ± 1 vs 0.43 ± 1, P <0.0001). Based on the above results, we injected GCaMP7s and implanted a cranial window after two weeks of PLX3397 chow (PLX-CFC group) (Fig. 4C). Apart from feeding the PLX-CFC mice with PLX3397 throughout the experiment, the rest of the procedures were the same in the trained and standard chow-fed control mice (CFC group). We found that PLX-CFC mice exhibited significantly reduced freezing fear responses during the fear learning and fear extinction training as compared with non-depleted CFC mice (Fig. 4D). In addition, in the naïve test, the frequency of Ca²⁺ events in PLX-CFC mice was lower than that of CFC mice (Fig. 4E). The Ca²⁺ imaging results showed that the frequency of Ca²⁺ events in PLX-CFC mice was significantly increased after fear learning. Unexpectedly, after fear extinction training, the Ca²⁺ event frequency did not change compared to retrieval (Fig. 4F, G, naïve vs retrieval, P = 0.004; retrieval vs extinction, P = 1.00; naïve vs extinction, P = 0.002; Kruskal–Wallis test followed by multiple comparisons test). Distribution histograms of the frequency of spontaneous Ca²⁺ events in PLX-CFC mice are shown in Fig. 4H.

Fig. 4.

Absence of microglia increases the frequency of neuronal activity after contextual fear learning, but no change after fear extinction. A Feeding C57 mice a chow diet containing the CSF1R antagonist PLX3397 for 30 days results in almost complete elimination of resident brain microglia. The absence of microglia is confirmed by the lack of an Iba1 (green) signal. Scale bars, left: 500 μm; right: 50 μm. B Numbers of microglia in C57BL6/J mice after a control or PLX3397 diet. Control, n = 7; PLX, n = 7. Data are presented as the mean ± SEM. ****P <0.0001, unpaired t-tests. C After two weeks of PLX chow, mice are injected with AAV2/9-jGCaMP7s in V1 and a cranial window is implanted at the same time. Two-photon imaging is performed along with a control or CFC behavioral experiment 2–3 weeks after surgery. PLX-CFC mice are fed with PLX3397 chow throughout the experiment. Control mice are fed with standard chow throughout the experiment. D Contextual fear learning and extinction curves in control CFC and PLX-CFC mice. CFC, n = 10; PLX-CFC, n = 7. Error bars, mean ± SEM. ****P <0.0001, two-way ANOVA test. E Ca²⁺ signal frequency decreases significantly in PLX mice compared to that in control mice. Control, n = 10; PLX, n = 6. ***P <0.001, Mann–Whitney test. F Ca²⁺ activity of layer II/III neurons in the V1 in PLX-CFC mice. 3 traces: Ca²⁺ activity of neurons indicated by the dashed circle. Ca²⁺ fluorescence traces for 197 s are shown. Scale bar, 30 μm. G Neuronal Ca²⁺ signal frequency in naïve, retrieval, and extinction tests in PLX-CFC mice, n = 6. Error bars, mean ± SEM. **P <0.01, Kruskal–Wallis test followed by multiple comparisons test. H Distribution histograms of the frequency of spontaneous Ca²⁺ events in PLX-CFC mice.

In our experiments, the Ca²⁺ event frequency did not significantly change in untrained control mice, while the standard chow-fed CFC mice showed a decreased frequency after fear learning and an increased frequency after extinction. In contrast, the PLX-CFC mice showed changes in the neuronal activity pattern caused by learning experience with an increased frequency after fear learning and a constant frequency after extinction (Fig. 5). In addition, we found that the frequency in PLX mice with retrieval was lower than that of naïve CFC mice, and the frequency in CFC mice with retrieval did not differ from that of naïve PLX mice (Fig. S5). Whether this increase or decrease in frequency has a causal relationship with behavioral performance needs to be further investigated in future work.

Fig. 5.

Depletion of microglia alters contextual fear learning and the extinction-induced frequency dynamics of Ca²⁺ events. In control mice, the Ca²⁺ event frequency decreases after fear learning, and this is reversed by fear extinction to the naïve level. In microglia-depleted mice, the Ca²⁺ event frequency increases after fear learning, and this is maintained after fear extinction.

Discussion

In this study, we demonstrated that bilateral inactivation of V1 during fear retrieval reduces the fear responses induced by CFC learning. CFC learning and extinction significantly increased axonal bouton turnover, and the rates of both bouton formation and elimination were responsible for the increased turnover. The frequency of neuronal Ca²⁺ activity in V1 decreased after fear learning, but fear extinction reversed the decrease and raised the frequency to the naïve level. In contrast, the pattern of neuronal activity dynamics changed in microglia-depleted mice, with the Ca²⁺ event frequency significantly increasing after fear learning and unaltered after fear extinction. These results demonstrated that CFC learning and extinction induce changes in presynaptic structures and the neuronal activity pattern in V1, a cortical region important for contextual fear retrieval. Moreover, microglia are involved in the mechanism underlying the change of neuronal activity pattern.

In contrast to the important progress that has been made in revealing postsynaptic structural dynamics and processing of information at the postsynapse [38], comparatively little is understood about the formation and dynamic remodeling of the presynaptic bouton. In the process of neurotransmission, an action potential propagates along the axonal shaft to the presynaptic bouton to open voltage-gated Ca²⁺ channels, triggering the release of neurotransmitters to activate postsynaptic receptors [40, 41]. Therefore, the experience-dependent structural dynamics of boutons are important for neural circuit remolding. The bouton turnover in adult mice is low in a short period of time (such as two days), which is important for reliable routine information processing [42]. In our study, the formation and elimination rates of boutons caused by fear learning and extinction were significantly higher than those of untrained mice. This suggests that the synaptic structure under the stimulus state undergoes more intense remodeling, which may be the anatomical mechanism of learning and memory.

Several studies have shown that locomotion is sufficient to drive activity in mouse V1 and to modulate visually-evoked activity [43–45]. One of the important mechanisms could be that A24b (a subdivision of the anterior cingulate cortex) and secondary motor cortex send a dense and topographically-organized projection to V1 that targets most neurons in layers II/III [22]. Since CFC learning and extinction lead to visually-driven freezing and unfreezing related to the motor cortex, it is reasonable to speculate that at least part of the change in neuronal activity is due to projections from the motor cortex. In addition, brain-wide single-cell tracing has revealed that layer II/III neurons in V1 project to more than one target area, including the entorhinal area, hippocampus, and lateral amygdala [46]. These target areas have been reported to be key structures in fear memory and extinction [1, 2]. Therefore, the decreased or increased frequency of Ca²⁺ events in layer II/III neurons in V1 may affect neuronal activity in V1 projection areas and modulate conditioned freezing responses.

Microglial activation elicits a negative affective state, and vice versa; an aversive stimulus can also activate microglia, resulting in decreased excitability of neurons [47]. Here, we demonstrated that fear learning as an aversive stimulus decreased the frequency of neuronal activity in V1, consistent with the above report [47]. Mice with depletion of microglia had a significantly lower frequency of neuronal activity at baseline than that of controls in our study. It has been reported that microglia depletion enhances the population activity of both CaMKII excitatory and parvalbumin inhibitory neurons in vivo [48]. Since hSyn-jGCaMP7s is widely expressed in both excitatory neurons and interneurons, this lower frequency induced by microglial depletion may be attributed to the comprehensive effect of different kinds of neurons. Microglial depletion decreases the threshold for neuron activation [49], which can be one of the reasons for the higher frequency of neuronal activity induced by fear learning. Neuronal activation is accompanied by the release of ATP, which recruits microglia through the surface-expressed purinergic receptor P2RY12 [50]. In the absence of microglia, neuronal activation leads to excitotoxicity with Ca²⁺ overload. Therefore, we speculate that there is a link between the deficits in CFC learning and excitotoxicity. In addition, it has also been reported that the neural circuits of poor learning task performance in microglia-depleted mice are mediated by brain-derived neurotrophic factor signaling [34]. Unexpectedly, in mice with microglial depletion, the frequency of neuronal activity increased after fear learning, and this was maintained after fear extinction. It has been shown that microglia phagocytose complement-dependent synapses, which leads to forgetting [35]. In microglia-depleted mice, the absence of microglial phagocytosis may contribute to constant neuronal activity after fear extinction. Therefore, our data demonstrate neuronal functional remodeling and provide new insights into the microglial role in neocortical information processing during CFC learning and extinction. Future studies are needed to investigate the Ca²⁺ activity of different kinds of neurons, how these dynamic neuronal activities are generated, and how microglia are involved in related signaling pathways during contextual fear learning and extinction, which may provide novel targets for the therapeutic manipulation of acquired fear memory.

Supplementary Information

Below is the link to the electronic supplementary material.

Conflict of interests

The authors declare that they have no competing interests.