Distinct Contribution of Granular and Agranular Subdivisions of the Retrosplenial Cortex to Remote Contextual Fear Memory Retrieval

Abstract

The retrieval of recent and remote memories are thought to rely on distinct brain circuits and mechanisms. The retrosplenial cortex (RSC) is robustly activated during the retrieval of remotely acquired contextual fear memories (CFMs), but the contribution of particular subdivisions [granular (RSG) vs agranular retrosplenial area (RSA)] and the circuit mechanisms through which they interact to retrieve remote memories remain unexplored. In this study, using both anterograde and retrograde viral tracing approaches, we identified excitatory projections from layer 5 pyramidal neurons of the RSG to the CA1 stratum radiatum/lacunosum-moleculare of the dorsal hippocampus and the superficial layers of the RSA in male mice. We found that chemogenetic or optogenetic inhibition of the RSG-to-CA1, but not the RSG-to-RSA, pathway selectively impairs the retrieval of remote CFMs. Collectively, our results uncover a specific role for the RSG in remote CFM recall and provide circuit evidence that RSG-mediated remote CFM retrieval relies on direct RSG-to-CA1 connectivity. The present study provides a better understanding of brain circuit mechanisms underlying the retrieval of remote CFMs and may help guide the development of therapeutic strategies to attenuate remote traumatic memories that lead to mental health issues such as post-traumatic stress disorder.

SIGNIFICANCE STATEMENT The RSC is implicated in contextual information processing and remote recall. However, how different subdivisions of the RSC and circuit mechanisms through which they interact to underlie remote memory recall remain unexplored. This study shows that granular subdivision of the RSC and its input to hippocampal area CA1 contributes to the retrieval of remote contextual fear memories. Our results support the hypothesis that the RSC and hippocampus require each other to preserve fear memories and may provide a novel therapeutic avenue to attenuate remote traumatic memories in patients with post-traumatic stress disorder.

Introduction

The retrosplenial cortex (RSC), a dorsomedial parietal area of the brain, is uniquely positioned as an interface between sensory cortices and various functional components of the hippocampal memory system (Sugar et al., 2011; Haugland et al., 2019). The RSC is not a unitary structure but instead can be divided into two main subregions, ventral granular subdivision (RSG) and dorsal agranular subdivision (RSA), in rodents, according to their cytoarchitectural organization and connectivity. Anatomically, the RSG has widespread connections with the lateral dorsal and anterior dorsal thalamic nuclei, subiculum, postsubiculum, and area CA1, whereas the RSA is densely interconnected with the anterior medial and lateral posterior thalamic nuclei, visuosensory cortical areas, and postrhinal, anterior cingulate, and prefrontal cortices (Jones et al., 2005; Sigwald et al., 2019). The RSC has been broadly implicated in multiple cognitive processes, including spatial navigation, orienting, learning and memory, and planning for the future (Vann et al., 2009; Mitchell et al., 2018). In terms of its role in memory processes, emerging evidence supports the involvement of the RSC in spatial, contextual, and episodic memories. In mice, temporary inactivation of RSC neurons at test impairs the retrieval of spatial memory in the Morris water maze task (Czajkowski et al., 2014). Evidence from lesion and pharmacological inactivation studies showed that the RSC is necessary in retrieving recent and remote contextual fear memories (CFMs; Keene and Bucci, 2008; Corcoran et al., 2011; Robinson et al., 2012) and inhibitory avoidance memories (Bouton and Todd, 2014; Katche et al., 2013). There is also evidence that the RSC is important for the extinction of both recently and remotely acquired CFMs (Corcoran et al., 2013). These results support the involvement of the RSC in fear-context association during memory retrieval. Despite being associated with the retrieval of spatial and contextual memories, a study suggests a significant role of the RSC in retrieving remotely acquired tone-based fear memories (Todd et al., 2016). However, an important caveat is that previous studies have considered the entirety of the RSC, and, hence, the specific contributions of individual RSC subdivisions and the circuit mechanisms through which they interact in the processing of CFMs remain unexplored.

Episodic memories are thought to be initially stored in the hippocampus and as time passes, are gradually transferred to the neocortex for long-term storage (Squire and Alvarez, 1995; Frankland et al., 2006; Varela et al., 2016; Albo and Gräff, 2018). Although there is still great debate about a time-limited role of the hippocampus in remote memory retrieval (Barry and Maguire, 2019), studies have shown that neocortical areas, particularly the medial prefrontal cortex, anterior cingulate cortex (ACC), and RSC, become more engaged during the retrieval of remote memories (Kitamura et al., 2017; Wirt and Hyman, 2019). The RSC, in particular, is of interest because of its anatomic connectivity with hippocampal-parahippocampal memory structures and its role in spatial, contextual, and episodic memories (Smith et al., 2018; Todd et al., 2019). The critical finding that the RSC can support contextual fear learning and memory without the hippocampus further emphasizes the importance of the RSC in processing hippocampal-dependent memories (Coelho et al., 2018). Notably, patients with RSC damage exhibited severe retrograde amnesia, suggesting that the RSC may serve as a critical cortical component for remote memory storage or retrieval (Vann et al., 2009; Miller et al., 2014; Ferguson et al., 2019). However, less is known about how RSC outputs regulate remote memory retrieval.

Here, we propose that distinct subdivisions of the RSC could differentially contribute to CFM processing. Employing a combined approach of viral tracing, chemogenetics, and optogenetics, we demonstrated that a subset of RSG layer 5 pyramidal neurons project directly to the CA1 of dorsal hippocampus (DH) and are recruited in the retrieval of remotely acquired CFMs in mouse brain. Our findings indicate a critical role for the RSG, but not the RSA, in remote CFM recall and reveal a novel RSG-to-CA1 pathway underlying remote CFM recall.

Materials and Methods

Animals

Adult male C57BL/6 mice (2–3 months old) were originally obtained from Charles River Laboratories, and CaMKIIα-Cre mice (B6.Cg-Tg(Camk2a-cre)T29-1Stl/J, catalog #JAX:005359; RRID:IMSR_JAX:005359) and Vgat-Cre mice (Slc32a1tm2(cre)Lowl, catalog #JAX:016962; RRID:IMSR_JAX:016962) were derived from The Jackson Laboratory and crossed with the C57BL/6 background. All mice were bred in the Laboratory Animal Center at National Cheng Kung University. Mice were socially housed in numbers of three littermates in humidity- and temperature-controlled (25 ± 1°C) rooms on a consistent 12 h/12 h light/dark cycle (lights on at 07:00) with food and water provided ad libitum. All behavioral procedures were conducted during the light cycle of the day (10:00–15:00). Mice were acclimatized to the testing room for at least 1 h before testing. All experimental procedures were approved by the Institutional Animal Care and Use Committee at National Cheng Kung University and conducted in accordance with the guidelines by National Institutes of Health for the Care and Use of Laboratory Animals. All efforts were made to minimize animal suffering and the number of animals used. The experimenters were blind to the treatment.

Recombinant adeno-associated virus vector production

DNA plasmids encoding adeno-associated virus (AAV), pAAV-human synapsin-1 (hSyn)-mCherry (plasmid #114472), pAAV-hSyn-hM4D(Gi)-mCherry (plasmid #50475), pAAV-calcium/calmodulin-dependent protein kinase II α (CaMKIIα)-mCherry (plasmid #114469), pAAV-CaMKIIα-hM4D(Gi)-mCherry (plasmid #50 477), pAAV-distal-less homeobox (Dlx)-GFP (plasmid #83900), pAAV-Dlx-hM3D(Gq)-GFP (plasmid #83897; modification by Addgene), pAAV-CaMKIIα-EGFP (plasmid #50469), and pAAV-CaMKIIα-enhanced Natronomonas pharaonis halorhodopsin 3.0, (eNpHR3.0)-EYFP (plasmid #26971) were obtained from Addgene. Plasmid DNA was amplified, purified, and collected using a standard plasmid maxiprep kit (Qiagen). The purified plasmids were mixed into CaCl₂ solution with the DNA plasmid coding AAV-DJ and cotransfected into HEK293GP cells using the calcium phosphate precipitation method. Transfected cells were harvested 72 h after transfection, and the virus was purified using the AAV Purification Mega Kit (Cell Biolabs). Viral titers were 5 × 10¹² particles/ml and stored in aliquots at −80°C until use.

Stereotactic surgery and drug infusion

Stereotactic surgery was conducted as described previously (Tsai et al., 2019). Briefly, mice were anesthetized with a mixture of 50 mg/kg zolazepam (Zoletil, Virbac) and 5 mg/kg xylazine hydrochloride (Rompun, Bayer) in the stereotaxic frame (David Kopf Instruments) for the entire surgery, and body temperature was maintained with a heating pad. Mice were bilaterally implanted with 26 gauge cannula guides (RWD Life Science) aimed at the RSG [−2.0 mm anteroposterior (AP), ± 0.2 mm mediolateral (ML), −1.0 mm dorsoventral (DV) to bregma] and RSA (−2.0 mm AP, ± 0.5 mm ML, −0.6 mm DV to bregma) according to the stereotaxic atlas of adult mouse brain (Franklin and Paxinos, 2008). Cannula guides were kept in place using BioGlue (CryoLife). Dummy cannulas (30 gauge) were inserted into the guides following the surgery to prevent clogging. Mice were intraperitoneally administrated with ketoprofen (5 mg/kg) for postoperative analgesia and given 2 weeks to recover. Fluorescent muscimol (0.1 µg/µl, Sigma-Aldrich) was microinfused bilaterally into the RSG at the rate of 0.25 µl/min (0.5 µl/side) 15 min before contextual fear conditioning (CFC) training or testing by using a 30 gauge needle that connected via polyethylene tubing to a Hamilton syringe. Fluorescent muscimol was dissolved in the vehicle (PBS). Drug dose was selected on the basis of the published study (Chiou et al., 2016). The infusion cannulas were kept in place for an additional 5 min to avoid pulling out the injectant when removing the cannulas. For virus injections, a beveled injection pipette was inserted at the desired coordinate, and 0.2 µl of the virus preparation was slowly injected using a microprocessor-controlled injector over a period of 10 min. The pipette was left in place for an additional 5 min to allow for diffusion of the virus solution and then withdrawn. Histologic verification of the locations of cannula tip was performed at the end of behavioral testing. Only data from animals with correct cannula implants (95% of the mice) were included in statistical analysis.

Contextual fear conditioning

Mice were handled for 2–3 min per day for 3 d before training. The CFC procedures were performed using a computer-controlled context conditioning system (ENV-307A, MED Associates) as previously described (Tsai et al., 2019). The conditioning chamber was placed inside a ventilated and sound-dampening isolation cubicle. Baseline freezing levels were measured during the 3 min of context exposure before training. During conditioning, mice were individually placed into a rectangular Plexiglas conditioning chamber (15.9 × 14.0 × 12.7 cm) and allowed to freely explore the apparatus for 2 min followed by three aversive electrical footshocks (0.6 mA, 2 s duration, 30 s intershock interval, constant current) through a stainless steel grid floor. After the last shock, mice were allowed to explore the apparatus for an additional 2 min before returning to their home cages. Control mice underwent the same experimental protocol as conditioned mice except without footshocks. The behavior of the mice was recorded using a digital near-infrared video camera on the wall of the cubicle. Context-dependent freezing responses were measured 1, 14, or 28 d after CFC training. Freezing was scored as the total time spent freezing in the conditioning during the 3 min test session. Significant motion pixel (SMP) values are a linear measure of mouse motion between frames captured at 7.5 Hz. Freezing, defined as SMP < 20 for longer than 1 s, was scored using the FreezeView program (MED Associates). The chamber was thoroughly cleaned with 70% ethanol after each trial to prevent bias based on olfactory cues.

Chemogenetic manipulations

For chemogenetic manipulation of RSC neuronal activity, mice were bilaterally injected with AAV_DJ-hSyn-hM4D(Gi)-mCherry, AAV_DJ-CaMKIIα-hM4D(Gi)-mCherry, or AAV_DJ-Dlx-hM3D(Gq)-GFP into the RSG or RSA, respectively. The AAV_DJ-hSyn-mCherry, AAV_DJ-CaMKIIα-mCherry, or AAV_DJ-Dlx-GFP was used as the control. Two weeks later, mice were injected intraperitoneally with vehicle [5% dimethyl sulfoxide (DMSO) in PBS] or clozapine-N-oxide (CNO; 3 mg/kg in 5% DMSO; catalog #4936, Tocris Bioscience) 30 min before memory retrieval testing. For terminal-specific silencing of RSG projections to the CA1, AAV_DJ-CaMKIIα-mCherry or AAV_DJ-CaMKIIα-hM4D(Gi)-mCherry was bilaterally injected into the RSG, and then mice were bilaterally implanted with 26 gauge cannula guides (RWD Life Science) aimed at the CA1 [−2.0 mm AP, ± 0.15 mm ML, −2.0 mm DV to bregma] to deliver vehicle or CNO. Twenty-eight days later, 0.5 µl vehicle (1% DMSO in PBS) or CNO (1.4 µg/µl) was infused bilaterally into the CA1 30 min before CFC testing. The dose of CNO was selected on the basis of published studies (Stachniak et al., 2014; Chiou et al., 2016; Meira et al., 2018). After the behavioral test, brains were dissected, and serial slices were imaged to verify correct viral expression.

Optogenetic stimulation

For pathway-specific optogenetic silencing of RSGγCA1 or RSGγRSA projections, mice were bilaterally injected with AAV_DJ-CaMKIIα-eNpHR3.0-EYFP into the RSG. The AAV_DJ-CaMKIIα-EGFP was used as the control. One week later, mice were implanted with optical cannula into the RSA or CA1. For in vivo optogenetic stimulation, the implanted optical cannula was linked to optic rotary joint path fiber (200 μm core diameter, Thorlabs) via a ferrule connector, which was connected to a red LED light source (Prizmatix) through an FC/PC adapter. LED output was controlled by a Pulser USB to TTL interface box (Prizmatix) to deliver light one train with a total duration of 30 s for 625 nm red light. Power output was determined using a PM100A power meter coupled to a S130C photodiode sensor (Thorlabs) and analyzed using LabVIEW 8.5 software. The estimated light intensity was ∼5 mW/mm², calculated using a model predicting irradiance in mammalian tissues (https://web.stanford.edu/group/dlab/cgi-bin/graph/chart.php). Light pluses at the same intensity were delivered for silencing of axonal terminals from the RSG to the RSA and CA1.

Anterograde and retrograde tracing

To map output projections of RSG or RSA projecting neurons, AAV_DJ-CaMKIIα-mCherry (0.2 μl) was injected bilaterally into the RSG (−2.0 mm AP, ± 0.2 mm ML, 1.0 mm DV to bregma) or RSA (−2.0 mm AP, ± 0.5 mm ML, −0.6 mm DV to bregma), respectively. Two weeks later, the mice were perfused with 4% paraformaldehyde and processed for mCherry labeling. For retrograde tracing of projecting neurons, 4% Fluorogold (0.1 μl; Santa Cruz Biotechnology) was injected unilaterally into the dorsal CA1 (−2.0 mm AP, ± 1.5 mm ML, 1.8 mm DV to bregma rostral) or the RSA (−2.0 mm AP, ± 0.5 mm ML, 0.6 mm DV to bregma) at 0.04 µl/min by using a 1 µl Hamilton syringe. The syringe was slowly retracted after an additional 5 min solution diffusion. For retrograde double labeling of RSG projecting neurons, the AAVrg-CaMKIIα-mCherry (0.2 μl) was injected into the dorsal CA1, and AAVrg-CaMKIIα-EGFP (0.2 μl) was injected into the RSA, respectively. The mice were perfused for tissue processing 5 d after injection. The proportion of double-projecting RSG neurons to either the dorsal CA1 or RSA was calculated from each mouse.

Rabies virus tracing

Rabies virus-based monosynaptic retrograde tracing was conducted as previously described (Wall et al., 2010; Wickersham et al., 2007). A viral cocktail, including AAV₅-FLEX-TVA-GFP and AAV₅-FLEX-RG (1:1 mixed, 0.5 μl), was injected into a single hemisphere of the dorsal CA1 of either CaMKIIα-Cre or Vgat-Cre mice. After allowing 21 d for expression of these helper viruses, EnvA-ΔG-mCherry (0.5 μl) rabies virus was injected at the same anatomic location. After another 10 d to allow for monosynaptic retrograde labeling, mice were killed and perfused and slices prepared for fluorescence microscopy. To preserve the location of surface membrane TVA receptor, we captured GFP signal without antibody enhancing GFP in the CA1. To preserve the mCherry signal, we used a chicken polyclonal antibody against mCherry (1:1000; catalog #ab205402; Abcam; RRID:AB_2722769) and amplified with a goat anti-chicken secondary Alexa Fluor 594 antibody (1:1000; catalog #A-11042, Thermo Fisher Scientific; RRID:AB_2534099). A mouse monoclonal antibody against CaMKIIα (1:200; catalog #ab22609, Abcam; RRID:AB_447192) was used to label excitatory neurons in the RSG.

Immunohistochemistry

Immunohistochemistry follows protocols as previously described (Tsai et al., 2019). Ninety minutes after the CFC test, mice were deeply anesthetized with 5% isoflurane and perfused transcardially with cold (4°C) PBS, followed by 4% paraformaldehyde. After the perfusion, brains were removed and continued to fix in 4% paraformaldehyde for 24 h at 4°C and then immersed in the solution containing 30% sucrose and kept at 4°C for at least 48 h before sectioning. Coronal slices containing the RSC were sectioned to a 40 µm thickness in serial order using a vibratome, washed with 0.3% Triton X-100, and then incubated for blocking with solution containing 3% bovine serum albumin in PBS. After blocking, the sections were transferred to wells and incubated in the primary antibodies against c-fos (1:500; catalog #2250, Cell Signaling Technology; RRID:AB_2247211), CaMKIIα (1:200; catalog #ab22609, Abcam; RRID:AB_447192), GAD67 (1:500; catalog #SAB2500443, Sigma-Aldrich; RRID:AB_10602450) and neuronal nuclei (NeuN, 1:2000, catalog #MAB377, Millipore; RRID:AB_2298772), for 24 h at 4°C in blocking solution. Sections were then washed with 0.4% Triton in PBS, followed by 3 h incubation with the secondary Alexa Fluor 488 (1:200; catalog #A-11094; Molecular Probes; RRID:AB_221544) and Alexa Fluor 594 antibodies (1:2000, #A-11032, Thermo Fisher Scientific; RRID:AB_2534091) at room temperature. The sections were collected on separate gelatin-subbed glass slides, rinsed extensively in PBS, and mounted with ProLong Gold Antifade Reagent (Invitrogen). Nuclei were counterstained with diamidino-2-phenylindole (DAPI, 1:20,000; catalog #D9542; Sigma-Aldrich). The slides were coverslipped and allowed to dry overnight. Fluorescence microscopic images of neurons were obtained using an Olympus FluoView FV3000 confocal laser scanning microscope. For quantification of c-fos immunopositivity, c-fos⁺ neurons were determined only when cells were colocalized with NeuN staining and visual-based semiquantitative estimation was used every sixth coronal section containing the RSG or RSA. All images were imported into NIH ImageJ software (National Center for Microscopy and Imaging Research: ImageJ Mosaic Plug-ins; RRID:SCR_001935) for analysis, and all the parameters used were kept consistent during capturing. The fluorescence intensity of c-fos⁺ cells is at least 10-fold above the background.

Electrophysiology and photostimulation in brain slices

Mice were deeply anesthetized with 5% isoflurane and decapitated, and brains were dissected quickly and chilled in ice-cold oxygenated sucrose cutting solution containing the following (in mm): 234 sucrose, 2.5 KCl, 0.5 CaCl₂, 7 MgCl₂, 25 NaHCO₃, 1.25 NaH₂PO₄, and 11 glucose, pH 7.3–7.4, and equilibrated with 95% O₂-5% CO₂. Coronal brain slices (250 µm) containing the RSC were prepared using a vibrating microtome (VT1200S, Leica; RRID:SCR_018453) and transferred to a holding chamber of artificial CSF (ACSF) containing (in mm): 117 NaCl, 4.7 KCl, 2.5 CaCl₂, 1.2 MgCl₂, 25 NaHCO₃, 1.2 NaH₂PO₄, and 11 glucose, pH 7.3–7.4, and equilibrated with 95% O₂-5% CO₂ and maintained at room temperature for at least 1 h before use.

For recording, slices were placed in a submersion-type recording chamber and fixed at the glass bottom of the chamber with a nylon grid on a platinum frame. The chamber was constantly perfused with ACSF at 32.0 ± 0.5°C at a rate of 2–3 ml/min. Whole-cell patch-clamp recordings were made from RSG or CA1 pyramidal neurons by using a patch-clamp amplifier (Axopatch 200B, Molecular Devices, RRID:SCR_018866) under infrared differential interference contrast microscope. Data acquisition and analysis were performed using a digitizer (Digidata 1440, Molecular Devices; RRID:SCR_021038) and pCLAMP 9 software (Molecular Devices). The patch electrode (2–3 mΩ resistance) was filled with pipette solution containing the following (in mm): 130 CsMeSO₄, 8 CsCl, 1 MgCl₂, 0.3 EGTA, 10 HEPES, 4 Mg-ATP, 0.3 Na-GTP, 10 Na-phosphocreatine, 1 QX-314, and 5 μm biocytin (pH 7.2, adjusted with CsOH, 280–290 mOsm). To examine spike patterns, cesium methanesulfonate was replaced with potassium methanesulfonate, and QX-314 was removed. To confirm the expression of engineered hM3D(Gq) or hM4D(Gi) receptors in RSG neurons, a depolarizing current pulse (150 pA, 500 ms) was injected into the control, hM3D(Gq)⁺ or hM4D(Gi)⁺, (hM3D(Gq) or hM4D(Gi) expression neurons to induce spiking in current-clamp mode. Following 10 min continuous recordings, CNO (50 μm) was applied into the ACSF, and a second depolarizing current pulse (+150 pA, 500 ms) was injected into the hM3D(Gq)⁻, hM3D(Gq)⁺, hM4D(Gi)⁻ and hM4D(Gi)⁺ neurons to induce spiking to compare spike activity before and after CNO bath application.

Statistical analysis

No statistical methods were used to predetermine sample size, but the number of animals used in each experiment was based on previous work of a similar nature by our laboratory (Tsai et al., 2019). No specific randomization method was used. Animals were randomly allocated into different experimental groups. Results are mean ± SEM and were analyzed using Prism 6 software (GraphPad). The sample size per group and statistical tests are indicated in the figure legends. Normality of data distribution was verified using the Shapiro–Wilk test. Two-tailed unpaired Student's t test or Mann–Whitney U test was used to examine statistical difference between two groups. The significance of the difference among multiple groups was calculated by two-way ANOVA followed by Bonferroni's post hoc test when significant main effects or interactions were detected. The number of animals used is indicated by n. Statistical significance was defined as p < 0.05. The sample sizes and statistical analysis details for each figure are reported in the figure legends.

Results

Pharmacological inactivation of the RSC impairs the acquisition and retrieval of CFMs

To get first insights into the role of the RSC in the acquisition and retrieval of CFMs, we performed pharmacological inactivation experiments using the fluorescent muscimol-bodipy, a potent agonist of the GABA_A receptors. The localization of vehicle and muscimol-bodipy injection sites in the RSC is illustrated in Figure 1A, left. The muscimol-bodipy remained confined to the RSC of infusion and did not spread to adjacent structures (Fig. 1A, middle, right). To determine whether neuronal activity in the RSC is required for fear acquisition, muscimol-bodipy or vehicle control was bilaterally infused into the RSC before CFC. No differences in freezing levels across conditioning trials were observed between vehicle- and muscimol-bodipy-treated mice. During the memory test, mice with pretraining infusion of muscimol-bodipy expressed less freezing compared with vehicle group 1 d after CFC (Fig. 1B). We next examined the contribution of the RSC in the retrieval of recent (1 d) and remote (14 d) CFMs by bilaterally infusing muscimol-bodipy into the RSC before memory retrieval testing. In comparison with vehicle control group, inactivation of the RSC, which did not affect fear learning, significantly reduced freezing to the conditioning context 1 and 14 d after CFC (Fig. 1C,D). These results suggest that the RSC is involved in both the acquisition and retrieval of recent and remote CFMs.

Figure 1.

The RSC is required for the acquisition and retrieval of recent and remote CFMs. A, Left, Schematic coronal sections (between bregma −2.30 and −1.82 mm) showing the intra-RSC injection sites of vehicle (Veh; gray circle) and muscimol-bodipy (Mus; red circle) for mice tested. Middle, A representative image of Mus expression in the RSC. Scale bar, 200 µm. Right, Magnification of boxed area (left) showing overlap between Mus (red) and NeuN (green). Scale bar, 20 µm. B, Left, Schematic representation of the experimental design. Mice were bilaterally injected into the RSC with Veh (PBS, 0.5 µl/side) or Mus (50 ng/0.5 µl/side) 15 min before CFC training (3 CS × US paired training). Middle, The learning curve for three acquisition trials of CFC training for Veh- and Mus-treated mice (Veh, n = 6; Mus, n = 6; interaction (trial × treatment): F_(3,40) = 0.28, p = 0.84; trial variable: F_(3,40) = 28.43, p < 0.001; treatment variable: F_(1,40) = 1.03, p = 0.32; two-way ANOVA). Right, Summary bar graphs depicting the fear memory retention test at 1 d after receiving CFC training in Veh- and Mus-treated mice (t₍₁₀₎ = 5.55, p = 0.0002, unpaired two-tailed Student's t test). The freezing behavior during 3 min of context exposure before training was measured as baseline. C, Left, Schematic representation of the experimental design. Mice were bilaterally injected into the RSC with Veh or Mus 15 min before the retrieval of recent (1 d) CFMs. Middle, The learning curve for three acquisition trials of CFC training in mice receiving Veh or Mus before retrieval test (Veh, n = 7; Mus, n = 7; interaction (trial × treatment): F_(3,48) = 0.07, p = 0.97; trial variable: F_(3,48) = 19.20, p < 0.001; treatment variable: F_(1,48) = 0.22, p = 0.64; two-way ANOVA). Right, Summary bar graphs depicting the fear memory retention test 1 d after CFC training in mice receiving Veh or Mus treatment 15 min before retrieval test (t₍₁₂₎ = 2.49, p = 0.029, unpaired two-tailed Student's t test). D, Left, Schematic representation of the experimental design. Mice were bilaterally injected into the RSC with Veh or Mus 15 min before the retrieval of remote (14 d) CFMs. Middle, The learning curve for three acquisition trials of CFC training in mice receiving Veh or Mus before retrieval test (Veh, n = 7; Mus, n = 7; interaction (trial × treatment): F_(3,48) = 0.48, p = 0.70; trial variable: F_(3,48) = 18.78, p < 0.001; treatment variable: F_(1,48) = 0.46, p = 0.50; two-way ANOVA). Right, Summary bar graphs depicting the fear memory retention test 14 d after CFC training in mice receiving Veh or Mus treatment 15 min before retrieval test (t₍₁₂₎ = 3.04, p = 0.01, unpaired two-tailed Student's t test). Data are presented as mean ± SEM; *p < 0.05 and ***p < 0.001.

A separate cohort of mice was used to assess the specificity of the pairing of neutral context with aversive electrical footshocks. To do this, control mice were trained without footshocks, and all mice were tested 1 or 14 d later. In comparison with control unshocked mice, paring mice showed significantly higher freezing levels 1 d (recent memory; Fig. 2A) and 14 d (remote memory; Fig. 2B) after CFC training. At 90 min following testing, mice were perfused, and we analyzed the expression of the c-fos immunoreactivity (a functional marker for neuronal activity) in the RSG (Fig. 2C). We found that the retrieval-induced c-fos expression was significantly higher in pairing mice compared with control unshocked mice 1 d (Fig. 2D) and 14 d (Fig. 2E) after CFC training.

Figure 2.

c-fos Expression in the RSG following the retrieval of recent and remote CFMs. A, Left, Schematic representation of the experimental design. Right, Summary bar graphs depicting the fear memory retention test in unshocked control and pairing groups 1 d after CFC training (unshocked, n = 5; pairing, n = 5; t₍₈₎ = 19.46, p < 0.001, unpaired two-tailed Student's t test). B, Left, Schematic representation of the experimental design. Right, Summary bar graphs depicting the fear memory retention test in unshocked control and pairing groups 14 d after CFC training (unshocked, n = 5; pairing, n = 5; t₍₈₎ = 13.45, p < 0.001, unpaired two-tailed Student's t test). C, Representative images of c-fos labeling in RSG layer 5 neurons from unshocked control and pairing mice 90 min after remote (14 d) memory retrieval test. Scale bar, 200 µm. Right, Magnification of boxed area (left) showing c-fos-expressing RSG cells colocalized with a neuronal marker, NeuN. Scale bar, 20 µm. D, Summary graphs depicting the percentage of c-fos⁺/NeuN⁺ cells in the RSG from unshocked control and pairing mice 90 min after recent (1 d) memory retrieval test (unshocked, n = 5; pairing, n = 5; t₍₈₎ = 4.34, p = 0.003, unpaired two-tailed Student's t test). E, Summary graphs depicting the percentage of c-fos⁺/NeuN⁺ cells in the RSG from unshocked control and pairing mice 90 min after remote memory retrieval test (unshocked, n = 5; pairing, n = 5; t₍₈₎ = 4.17, p = 0.003, unpaired two-tailed Student's t test). Data indicate mean ± SEM; **p < 0.01 and ***p < 0.001.

Chemogenetic silencing of RSG neuronal activity at test impairs the retrieval of remote CFMs

We then investigated whether distinct subdivisions of the RSC have different roles in retrieving CFMs by using an approach based on a designer receptor exclusively activated by designer drugs (DREADD) to attenuate RSG or RSA neuronal activity during the memory retrieval test. To do so, we bilaterally injected AAV-DJ expressing an engineered G_i-coupled receptor hM4D(Gi) tagged with a fluorescent reporter mCherry under the control of pan-neuronal promoter human synapsin-1 (AAV_DJ-hSyn-hM4D(Gi)-mCherry) into the RSG or RSA, respectively. Control mice received infusion of a virus expressing only the fluorescent reporter (AAV_DJ-hSyn-mCherry). The experimental procedure is depicted in Figure 3A. The localization of viral vector injection sites in the RSG and RSA is illustrated in Figure 3B. At 14 d following AAV injection, mice were subjected to CFC and were tested for memory retention 1 or 14 d later. Mice were treated with CNO, a synthetic ligand of hM4D(Gi) (Armbruster et al., 2007), before the memory retrieval test. A post hoc histologic examination of brain sections revealed robust and bilateral coexpression of hM4D(Gi) with the neuronal marker NeuN in the RSG or RSA, respectively (Fig. 3C). In a subset of mice, we performed ex vivo electrophysiological assessment in acute brain slices to verify the effectiveness of CNO in AAV-infected neurons. Application of CNO (50 μm) inhibited spiking responses to injected square current pulses in hM4D(Gi)-expressing (hM4D(Gi)⁺) neurons, but not in neurons that expressed mCherry alone (Fig. 3D). We found that inactivation of the RSG or RSA by systemic administration of CNO before memory retrieval testing did not affect freezing levels on day 1 after CFC (recent memory), compared with hM4D(Gi)/vehicle or mCherry/CNO treatment groups (Fig. 3E,F). In contrast, inactivation of the RSG, but not the RSA, by systemic administration of CNO before the memory retrieval test significantly decreased freezing levels on day 14 after CFC (remote memory), compared with hM4D(Gi)/vehicle or mCherry/CNO treatment groups (Fig. 3G,H). Silencing specificity and effectiveness were verified by the analysis of c-fos immunoreactivity in the RSG and RSA 90 min following the memory retrieval test (Fig. 4A–E). These results suggest that neuronal activity of the RSG is selectively required for the retrieval of remote CFMs.

Figure 3.

Inactivation of the RSG impairs the retrieval of remote CFMs. A, Schematic representation of the experimental design. Two weeks after stereotaxic injection of AAV_DJ-hSyn-hM4D(Gi)-mCherry or AAV_DJ-hSyn-mCherry into the RSG or RSA, mice were trained in a CFC paradigm, and memory retention was tested 1 d (recent) or 14 d (remote) after training. Mice were injected intraperitoneally with vehicle (Veh) or CNO (3 mg/kg) 30 min before the retrieval test. B, Schematic coronal sections (between bregma −2.54 and −1.82 mm) showing the intra-RSG and intra-RSA injection sites of viral vectors for mice tested. C, Representative images showing the expression of hSyn-hM4D(Gi)-mCherry in the RSG and RSA. Scale bars: 200 µm (top); 20 µm (rectangle amplification, bottom). D, Representative traces showing responses of uninfected (hM4D(Gi)⁻) and infected (hM4D(Gi)⁺) neurons to depolarizing current pulse (200 pA) under whole-cell current clamp before and after bath application of CNO (50 μm) in the ex vivo RSG slices. E, Summary of experiments showing the effects of systemic Veh and CNO injections on the retrieval of recent CFMs in mice that received bilateral injections of AAV_DJ-hSyn-hM4D(Gi)-mCherry or AAV_DJ-hSyn-mCherry into the RSG (hM4D(Gi) + Veh, n = 8; mCherry + CNO, n = 8; hM4D(Gi) + CNO, n = 8; F_(2,21) = 3.17, p = 0.06, one-way ANOVA). F, Summary of experiments showing the effects of systemic Veh and CNO injections on the retrieval of recent CFMs in mice that received bilateral injections of AAV_DJ-hSyn-hM4D(Gi)-mCherry or AAV_DJ-hSyn-mCherry into the RSA (hM4D(Gi) + Veh, n = 9; mCherry + CNO, n = 8; hM4D(Gi) + CNO, n = 10; F_(2,24) = 1.92, p = 0.17, one-way ANOVA). G, Summary of experiments showing the effects of systemic Veh and CNO injections on the retrieval of remote CFMs in mice that received bilateral injections of AAV_DJ-hSyn-hM4D(Gi)-mCherry or AAV_DJ− hSyn-mCherry into the RSG (hM4D(Gi) + Veh, n = 8; mCherry + CNO, n = 8; hM4D(Gi) + CNO, n = 8; F_(2,21) = 23.30, p < 0.0001, one-way ANOVA followed by Bonferroni's post hoc test). H, Summary of experiments showing the effects of systemic Veh and CNO injections on the retrieval of remote CFMs in mice that received bilateral injections of AAV_DJ-hSyn-hM4D(Gi)-mCherry or AAV_DJ-hSyn-mCherry into the RSA (hM4D(Gi) + Veh, n = 8; mCherry + CNO, n = 8; hM4D(Gi) + CNO, n = 10; F_(2,23) = 2.76, p = 0.08, one-way ANOVA). Data indicate mean ± SEM; ***p < 0.001.

Figure 4.

Validation of CNO/hM4D(Gi)-mediated inhibition of RSG or RSA neurons. A, Representative images of c-fos labeling in RSG layer 5 neurons from AAV_DJ-hSyn-mCherry + CNO and AAV_DJ-hSyn-hM4D(Gi)-mCherry + CNO mice 90 min after remote memory retrieval test. Scale bar, 200 µm. Right, Augmented figures showing c-fos-expressing RSG neurons in rectangle area. Scale bar, 20 µm. B, Summary graphs depicting the percentages of c-fos⁺/mCherry⁺ (left) and mCherry⁺/NeuN⁺ (right) cells in the RSG from AAV_DJ-hSyn-hM4D(Gi)-mCherry + Veh (n = 5), AAV_DJ-hSyn-mCherry + CNO (n = 5), and AAV_DJ-hSyn-hM4D(Gi)-mCherry + CNO (n = 5) mice 90 min after recent memory retrieval test (c-fos⁺/mCherry⁺: F_(2,12) = 20.58, p = 0.0001, one-way ANOVA followed by Bonferroni's post hoc test; mCherry⁺/NeuN⁺: F_(2,12) = 1.64, p = 0.23, one-way ANOVA). C, Summary graphs depicting the percentages of c-fos⁺/mCherry⁺ (left) and mCherry⁺/NeuN⁺ (right) cells in the RSA from AAV_DJ-hSyn-hM4D(Gi)-mCherry + Veh (n = 5), AAV_DJ-hSyn-mCherry + CNO (n = 5), and AAV_DJ-hSyn-hM4D(Gi)-mCherry + CNO (n = 5) mice 90 min after recent memory retrieval test (c-fos⁺/mCherry⁺: F_(2,12) = 7.19, p = 0.009, one-way ANOVA followed by Bonferroni's post hoc test; mCherry⁺/NeuN⁺: F_(2,12) = 1.90, p = 0.19, one-way ANOVA). D, Summary graphs depicting the percentages of c-fos⁺/mCherry⁺ (left) and mCherry⁺/NeuN⁺ (right) cells in the RSG from AAV_DJ-hSyn-hM4D(Gi)-mCherry + Veh (n = 5), AAV_DJ-hSyn-mCherry + CNO (n = 5), and AAV_DJ-hSyn-hM4D(Gi)-mCherry + CNO (n = 5) mice 90 min after remote memory retrieval test (c-fos⁺/mCherry⁺: F_(2,12) = 53.28, p < 0.0001, one-way ANOVA followed by Mann–Whitney U test; mCherry⁺/NeuN⁺: F_(2,12) = 0.30, p = 0.75, one-way ANOVA). E, Summary graphs depicting the percentages of c-fos⁺/mCherry⁺ (left) and mCherry⁺/NeuN⁺ (right) cells in the RSA from AAV_DJ-hSyn-hM4D(Gi)-mCherry + Veh (n = 5), AAV_DJ-hSyn-mCherry + CNO (n = 5), and AAV_DJ-hSyn-hM4D(Gi)-mCherry + CNO (n = 5) mice 90 min after remote memory retrieval test (c-fos⁺/mCherry⁺: F_(2,12) = 12.66, p = 0.001, one-way ANOVA followed by Mann–Whitney U test; mCherry⁺/NeuN⁺: F_(2,12) = 0.77, p = 0.49, one-way ANOVA). Data indicate mean ± SEM; **p < 0.01 and ***p < 0.001.

In the above experiments, we used the pan-neuronal hSyn promoter to drive hM4D(Gi) gene expression, so we could not distinguish excitatory and inhibitory neuronal populations altered by chemogenetic manipulations. We then performed additional inactivation experiments to validate how the RSG and RSA regulate CFMs by using cell-type-specific neuronal promoters. We first bilaterally injected the RSG or RSA, respectively, with an AAV-DJ vector expressing hM4D(Gi)-mCherry under the control of the CaMKIIα promoter (AAV_DJ-CaMKIIα-hM4D(Gi)-mCherry), favoring expression to excitatory neurons (Liu and Jones, 1996). As shown in Figure 5A, 14 d after AAV injections, mice were subjected to CFC, and memory tests were conducted 1 or 14 d after training in the presence of vehicle or CNO treatment. A post hoc histologic examination of brain sections revealed robust and bilateral coexpression of hM4D(Gi) with the neuronal maker NeuN in the RSG or RSA (Fig. 5B). Application of CNO (50 μm) significantly decreased spiking responses to injected square current pulses in hM4D(Gi)⁺ neurons but not in neurons that expressed mCherry alone in ex vivo brain slices (Fig. 5C). We observed that chemogenetic silencing of RSG but not RSA pyramidal neurons led to a significant decrease in freezing levels on day 1 after CFC, compared with hM4D(Gi)/vehicle or mCherry/CNO treatment groups (Fig. 5D,E). Similar to our hSyn-hM4D(Gi) data (Fig. 3), we found that systemic CNO treatment in mice expressing CaMKIIα-hM4D(Gi) in RSG but not RSA neurons significantly decreased freezing to the conditioning context on day 14 after CFC, compared with hM4D(Gi)/vehicle or mCherry/CNO treatment groups (Fig. 5F,G). Analysis of c-fos expression 90 min after the memory test revealed that hM4D(Gi) stimulation decreased activation of RSG and RSA neurons in CNO-treated mice compared with vehicle-treated group (Fig. 6A–E).

Figure 5.

Chemogenetic silencing of excitatory neurons in the RSG impairs the retrieval of recent and remote CFMs. A, Schematic representation of the experimental design. Two weeks after stereotaxic injection of AAV_DJ-CaMKIIα-hM4D(Gi)-mCherry or AAV_DJ-CaMKIIα-mCherry into the RSG or RSA, mice were trained in a CFC paradigm, and memory retention was tested 1 d (recent) or 14 d (remote) after training. Mice were injected intraperitoneally with vehicle (Veh) or CNO (3 mg/kg) 30 min before the retrieval test. B, Representative images showing the expression of CaMKIIα-hM4D(Gi)-mCherry in the RSG and RSA. Scale bars: 200 µm (top); 20 µm (rectangle amplification, bottom). C, Representative traces showing responses of uninfected (hM4D(Gi)⁻) and infected (hM4D(Gi)⁺) neurons to depolarizing current pulse (200 pA) under whole-cell current clamp before and after bath application of CNO (50 μm) in the ex vivo RSG slices. D, Summary of experiments showing the effects of systemic Veh and CNO injections on the retrieval of recent CFMs in mice that received bilateral injections of AAV_DJ-CaMKIIα-hM4D(Gi)-mCherry or AAV_DJ-CaMKIIα-mCherry into the RSG [mCherry + Veh, n = 7; hM4D(Gi) + Veh, n = 7; mCherry + CNO, n = 7; hM4D(Gi) + CNO, n = 7; interaction (treatment × drug): F_(1,24) = 5.72, p = 0.025; treatment variable: F_(1,24) = 8.12, p = 0.009; drug variable: F_(1,24) = 2.48, p = 0.13; two-way ANOVA followed by Bonferroni's post hoc test]. E, Summary of experiments showing the effects of systemic Veh and CNO injections on the retrieval of recent CFMs in mice that received bilateral injections of AAV_DJ-CaMKIIα-hM4D(Gi)-mCherry or AAV_DJ-CaMKIIα-mCherry into the RSA [mCherry + Veh, n = 8; hM4D(Gi) + Veh, n = 8; mCherry + CNO, n = 8; hM4D(Gi) + CNO, n = 8; interaction (treatment × drug): F_(1,28) = 0.4499, p = 0.51; treatment variable: F_(1,28) = 1.48, p = 0.23; drug variable: F_(1,28) = 0.96, p = 0.33; two-way ANOVA]. F, Summary of experiments showing the effects of systemic Veh and CNO injections on the retrieval of remote CFMs in mice that received bilateral injections of AAV_DJ-CaMKIIα-hM4D(Gi)-mCherry or AAV_DJ-CaMKIIα-mCherry into the RSG [mCherry + Veh, n = 11; hM4D(Gi) + Veh, n = 13; mCherry + CNO, n = 13; hM4D(Gi) + CNO, n = 11; interaction (treatment × drug): F_(1,44) = 10.66, p = 0.002; treatment variable: F_(1,44) = 14.36, p < 0.001; drug variable: F_(1,44) = 2.62, p = 0.11; two-way ANOVA followed by Bonferroni's post hoc test]. G, Summary of experiments showing the effects of systemic Veh and CNO injections on the retrieval of remote CFMs in mice that received bilateral injections of AAV_DJ-CaMKIIα-hM4D(Gi)-mCherry or AAV_DJ-CaMKIIα-mCherry into the RSA [mCherry + Veh, n = 8; hM4D(Gi) + Veh, n = 8; mCherry + CNO, n = 8; hM4D(Gi) + CNO, n = 8; interaction (treatment × drug): F_(1,28) = 0.003, p = 0.95; treatment variable: F_(1,28) = 2.70, p = 0.11; drug variable: F_(1,28) = 2.15, p = 0.15; two-way ANOVA]. Data indicate mean ± SEM; *p < 0.05, **p < 0.01, and ***p < 0.001.

Figure 6.

Validation of CNO/hM4D(Gi)-mediated inhibition of RSG or RSA CaMKIIα-expressing excitatory neurons. A, Representative images of c-fos labeling in RSG layer 5 CaMKIIα-expressing excitatory neurons from AAV_DJ-CaMKIIα-mCherry + CNO and AAV_DJ-CaMKIIα- hM4D(Gi)-mCherry + CNO mice 90 min after remote memory retrieval test. Scale bar, 200 µm. Right, Augmented figures showing c-fos-expressing RSG neurons in rectangle area. Scale bar, 20 µm. B, Summary graphs depicting the percentages of c-fos⁺/mCherry⁺ (left) and mCherry⁺/NeuN⁺ (right) cells in the RSG from AAV_DJ-CaMKIIα-mCherry + CNO (n = 5) and AAV_DJ-CaMKIIα-hM4D(Gi)-mCherry + CNO (n = 5) mice 90 min after recent memory retrieval test (c-fos⁺/mCherry⁺: t₍₈₎ = 2.9, p = 0.02, unpaired Student's t test; mCherry⁺/NeuN⁺: t₍₈₎ = 0.4, p = 0.67, unpaired Student's t test). C, Summary graphs depicting the percentages of c-fos⁺/mCherry⁺ (left) and mCherry⁺/NeuN⁺ (right) cells in the RSA from AAV_DJ-CaMKIIα-mCherry + CNO (n = 5) and AAV_DJ-CaMKIIα-hM4D(Gi)-mCherry + CNO (n = 5) mice 90 min after recent memory retrieval test (c-fos⁺/mCherry⁺: t₍₈₎ = 8.2, p < 0.0001, unpaired Student's t test; mCherry⁺/NeuN⁺: t₍₈₎ = 1.8, p = 0.10, unpaired Student's t test). D, Summary graphs depicting the percentages of c-fos⁺/mCherry⁺ (left) and mCherry⁺/NeuN⁺ (right) cells in the RSG from AAV_DJ-CaMKIIα-mCherry + CNO (n = 5) and AAV_DJ-CaMKIIα-hM4D(Gi)-mCherry + CNO (n = 5) mice 90 min after remote memory retrieval test (c-fos⁺/mCherry⁺: t₍₈₎ = 7.3, p < 0.0001, unpaired Student's t test; mCherry⁺/NeuN⁺: t₍₈₎ = 0.4, p = 0.71, unpaired Student's t test). E, Summary graphs depicting the percentages of c-fos⁺/mCherry⁺ (left) and mCherry⁺/NeuN⁺ (right) cells in the RSA from AAV_DJ-CaMKIIα-mCherry + CNO (n = 5) and AAV_DJ-CaMKIIα-hM4D(Gi)-mCherry + CNO (n = 5) mice 90 min after remote memory retrieval test (c-fos⁺/mCherry⁺: t₍₈₎ = 6.0, p = 0.0003, unpaired Student's t test; mCherry⁺/NeuN⁺: t₍₈₎ = 0.3, p = 0.80, unpaired Student's t test). Data indicate mean ± SEM; *p < 0.05 and ***p < 0.001.

We next tested the effects of chemogenetic activation of GABAergic neurons with the excitatory hM3D(Gq) DREADD in the RSG or RSA on CFM retrieval. To do this, we bilaterally injected the RSG or RSA with an AAV-DJ vector expressing hM3D(Gq)-GFP under the control of the Dlx promoter (AAV_DJ-Dlx-hM3D(Gq)-GFP), favoring expression to forebrain GABAergic neurons (Dimidschstein et al., 2016). The experimental design is shown in Figure 7A. At 14 d following AAV injection, mice were subjected to CFC, and memory tests were performed 1 or 14 d after training in the presence of vehicle or CNO treatment. A post hoc histologic analysis revealed robust and bilateral coexpression of hM3D(Gq) with the neuronal marker NeuN in the RSG or RSA (Fig. 7B). Application of CNO (50 μm) significantly increased spiking responses to injected square current pulses in hM3D(Gq)⁺ neurons but not in neurons that expressed GFP alone in ex vivo brain slices (Fig. 7C). We found that neither the activation of GABAergic neurons in the RSG nor RSA by systemic administration of CNO affected freezing levels on day 1 after CFC, compared with hM3D(Gq)/vehicle or GFP/CNO treatment groups (Fig. 7D,E). Consistent with the notion that RSG activity is necessary for remote memory retrieval, stimulation of GABAergic neurons in the RSG, but not the RSA, with systemic administration of CNO before the memory test significantly decreased freezing levels on day 14 after CFC, compared with hM3D(Gq)/vehicle or GFP/CNO treatment groups (Fig. 7F,G). To confirm that activation of hM3D(Gq)-expressing GABAergic neurons decreased the activity of pyramidal neurons during fear memory recall, we quantified c-fos immunoreactivity in the RSG and RSA 90 min after the memory retrieval test. As expected, there was a significant decrease in c-fos labeling pyramidal neurons in CNO-treated mice compared with the vehicle-treated group (Fig. 8A–E). These results provided evidence for the involvement of RSG excitatory pyramidal neurons in remote CFM retrieval. Our results extend previous experiments in which neurotoxic lesion or pharmacological inactivation of the RSC impaired recall of recent and remote CFMs (Keene and Bucci, 2008; Corcoran et al., 2011; Robinson et al., 2012).

Figure 7.

Chemogenetic activation of GABAergic neurons in the RSG impairs the retrieval of remote CFMs. A, Schematic representation of the experimental design. Two weeks after stereotaxic injection of AAV_DJ-Dlx-hM3D(Gq)-GFP or AAV_DJ-Dlx-GFP into the RSG or RSA, mice were trained in a CFC paradigm, and memory retention was tested 1 d (recent) or 14 d (remote) after training. Mice were injected intraperitoneally with vehicle (Veh) or CNO (3 mg/kg) 30 min before the retrieval test. B, Representative images showing the expression of Dlx-hM3D(Gq)-GFP in the RSG and RSA. Scale bars: 200 µm (top) and 20 µm (rectangle amplification, bottom). C, Representative traces showing responses of uninfected (hM3D(Gq)⁻) and infected (hM3D(Gq)⁺) neurons to depolarizing current pulse (200 pA) under whole-cell current clamp before and after bath application of CNO (50 μm) in the ex vivo RSG slices. D, Summary of experiments showing the effects of systemic Veh and CNO injections on the retrieval of recent CFMs in mice that received bilateral injections of AAV_DJ-Dlx-hM3D(Gq)-GFP or AAV_DJ-Dlx-GFP into the RSG [GFP + Veh, n = 8; hM3D(Gq) + Veh, n = 8; GFP + CNO, n = 8; hM3D(Gq) + CNO, n = 8; interaction (treatment × drug): F_(1,28) = 0.28, p = 0.60; treatment variable: F_(1,28) = 0.003, p = 0.95; drug variable: F_(1,28) = 2.15, p = 0.58; two-way ANOVA]. E, Summary of experiments showing the effects of systemic Veh and CNO injections on the retrieval of recent CFMs in mice that received bilateral injections of AAV_DJ-Dlx-hM3D(Gq)-GFP or AAV_DJ-Dlx-GFP into the RSA [GFP + Veh, n = 8; hM3D(Gq) + Veh, n = 8; GFP + CNO, n = 8; hM3D(Gq) + CNO, n = 8; interaction (treatment × drug): F_(1,28) = 0.057, p = 0.81; treatment variable: F_(1,28) = 0.70, p = 0.41; drug variable: F_(1,28) = 0.14, p = 0.71; two-way ANOVA]. F, Summary of experiments showing the effects of systemic Veh and CNO injections on the retrieval of remote CFMs in mice that received bilateral injections of AAV_DJ-Dlx-hM3D(Gq)-GFP or AAV_DJ-Dlx-GFP into the RSG [GFP + Veh, n = 9; hM3D(Gq) + Veh, n = 9; GFP + CNO, n = 9; hM3D(Gq) + CNO, n = 9; interaction (treatment × drug): Interaction: F_(1,32) = 12.56, p = 0.001; treatment variable: F_(1,32) = 7.06, p = 0.012; drug variable: F_(1,32) = 2.62, p = 0.12; two-way ANOVA followed by Bonferroni's post hoc test]. G, Summary of experiments showing the effects of systemic Veh and CNO injections on the retrieval of remote CFMs in mice that received bilateral injections of AAV_DJ-Dlx-hM3D(Gq)-GFP or AAV_DJ-Dlx-GFP into the RSA [GFP + Veh, n = 8; hM3D(Gq) + Veh, n = 8; GFP + CNO, n = 8; hM3D(Gq) + CNO, n = 8; interaction (treatment × drug): F_(1,28) = 0.004, p = 0.95; treatment variable: F_(1,28) = 5.83, p = 0.023; drug variable: F_(1,28) = 0.55, p = 0.46; Two-way ANOVA]. Data indicate mean ± SEM; **p < 0.01 and ***p < 0.001.

Figure 8.

Validation of CNO/hM3D(Gq)-mediated activation of RSG or RSA Dlx-expressing GABAergic neurons. A, Representative images of c-fos labeling in RSG layer 5 neurons from AAV_DJ-Dlx-GFP + CNO and AAV_DJ-Dlx-hM3D(Gq)-GFP + CNO mice 90 min after remote memory retrieval test. Scale bar, 200 µm. Augmented figures (right) showing c-fos-expressing neurons in rectangle area. Scale bar, 20 µm. B, Summary graphs depicting the percentages of c-fos⁺/GFP⁻/DAPI⁺ (left) and GFP⁺/DAPI⁺ (right) cells in the RSG from AAV_DJ-Dlx-GFP + CNO (n = 5) and AAV_DJ-Dlx-hM3D(Gq)-GFP + CNO (n = 5) mice 90 min after recent memory retrieval test (c-fos⁺/GFP⁻/DAPI⁺: p = 0.008, Mann–Whitney U test; GFP⁺/DAPI⁺: t₍₈₎ = 1.0, p < 0.34, unpaired Student's t test). C, Summary graphs depicting the percentages of c-fos⁺/GFP⁻/DAPI⁺ (left) and GFP⁺/DAPI⁺ (right) cells in the RSA from AAV_DJ-Dlx-GFP + CNO (n = 5) and AAV_DJ-Dlx-hM3D(Gq)-GFP + CNO (n = 5) mice 90 min after recent memory retrieval test (c-fos⁺/GFP⁻/DAPI⁺: t₍₈₎ = 2.0, p = 0.08, unpaired Student's t test; GFP⁺/DAPI⁺: t₍₈₎ = 1.1, p = 0.31, unpaired Student's t test). D, Summary graphs depicting the percentages of c-fos⁺/GFP⁻/DAPI⁺ (left) and GFP⁺/DAPI⁺ (right) cells in the RSG from AAV_DJ-Dlx-GFP + CNO (n = 5) and AAV_DJ-Dlx-hM3D(Gq)-GFP + CNO (n = 5) mice 90 min after remote memory retrieval test (c-fos⁺/GFP⁻/DAPI⁺: t₍₈₎ = 4.0, p = 0.004, unpaired Student's t test; GFP⁺/DAPI⁺: t₍₈₎ = 0.5, p = 0.62, unpaired Student's t test). E, Summary graphs depicting the percentages of c-fos⁺/GFP⁻/DAPI⁺ (left) and GFP⁺/DAPI⁺ (right) cells in the RSA from AAV_DJ-Dlx-GFP + CNO (n = 5) and AAV_DJ-Dlx-hM3D(Gq)-GFP + CNO (n = 5) mice 90 min after remote memory retrieval test (c-fos⁺/GFP⁻/DAPI⁺: t₍₈₎ = 2.9, p = 0.02, unpaired Student's t test; GFP⁺/DAPI⁺: t₍₈₎ = 0.9, p = 0.41, unpaired Student's t test). Data indicate mean ± SEM; *p < 0.05 and **p < 0.01.

The RSG projects to the dorsal CA1 and the RSA

To investigate the neural circuit mechanisms by which the RSG mediates the retrieval of remote CFMs, we injected AAV_DJ-CaMKIIα-mCherry into the RSG. Fourteen days after injection, we observed abundant mCherry-positive cells in the RSG, and mCherry-labeled projections were detected in the CA1 stratum radiatum/stratum lacunosum-moleculare (SRLM) of the DH and the superficial layer 2 of the RSA of coronal brain sections (Figs. 9A, 10A). To identify the proportions of RSG neurons projecting to the CA1 region, we unilaterally injected the retrograde tracer Fluorogold (4%) into the dorsal CA1 and performed immunohistochemistry. We found that retrogradely labeled neurons were mainly located in layer 5 of the RSG and were immunopositive for the excitatory cell marker CaMKIIα but not the inhibitory cell marker glutamic acid decarboxylase of 67 kDa (GAD67; Figs. 9B, 10B). We also injected AAV_DJ-CaMKIIα-mCherry into the RSA to label RSA projection neurons. We identified mCherry-labeled projections in several brain regions, including the RSG, lateral entorhinal cortex (LEnt), lateral amygdala (LA), and postrhinal (POR) cortex (Figs. 9C, 10C). To define reciprocal connectivity between the RSG and RSA, we injected Fluorogold into the RSA and observed that neurons projecting from the RSG to RSA were prominently located in layer 5 and were immunopositive for CaMKIIα (Figs. 9D, 10D). We also determined the proportions of RSG neurons projecting to the CA1 and RSA using two different retrograde tracers. We unilaterally injected AAVrg-CaMKIIα-mCherry into the CA1 and AAVrg-CaMKIIα-EGFP into the RSA to retrogradely label the RSG neurons projecting to these two areas. Using this approach, the cell bodies of RSG neurons projecting to the CA1 and RSA were labeled by AAVrg-CaMKIIα-mCherry and AAVrg-CaMKIIα-EGFP, respectively. Confocal imaging showed that most of the labeled RSG neurons expressed either mCherry or EGFP alone, indicating that they project to either the dorsal CA1 or RSA alone (Figs. 9E, 10E). Additional analysis revealed that only ∼3% of all labeled RSG neurons were coexpressed mCherry and EGFP (Fig. 9E). These results indicate that distinct populations of RSG neurons send excitatory projections to the dorsal CA1 and RSA.

Figure 9.

Distinct projection patterns of the RSA and RSG. A, Left, AAV_DJ-CaMKIIα-mCherry was injected into the RSG (top). Right, mCherry signals (red) were expressed in RSG neurons under the control of CaMKIIα promoter (top). Scale bar, 200 μm. The mCherry signals of axonal projections were observed in the CA1 SRLM (left and middle, bottom) and RSA layer 2 area (right, bottom). Scale bar, 100 μm. Right, Magnified image of rectangle (bottom). Scale bar, 20 μm. Sections were costained with NeuN (green). SO, Stratum oriens; SP, stratum pyramidale; SR, stratum radiatum; SLM, stratum lacunosum moleculare. Data were replicated in four mice. B, Left, Fluorogold (FG, 4%) was injected into the dorsal CA1 area (top). Right, FG signal (gray) was expressed in pyramidal cell layer of the dorsal CA1 (top). Scale bar, 100 μm. Bottom, The FG ⁺ cell bodies (blue and indicated by blue arrow) were found in layer 5 of the RSG and were immunopositive for CaMKIIα (red, indicated by red arrow). Colocalized cells are indicated by white arrow. Scale bar, 20 μm. Data were replicated in four mice. C, Left, AAV_DJ-CaMKIIα-EGFP was injected into the RSA (top). Right, mCherry signals (red) were expressed in RSA neurons under the control of CaMKIIα promoter (top). Scale bar, 200 μm. The mCherry signals of axonal projections were observed in the RSG (left, bottom), LA, LEnt (middle, bottom) and POR (right, bottom). Sections were costained with NeuN (green). Scale bar, 100 μm. Data were replicated in four mice. D, Left, FG was injected into the RSA (top). Right, FG signals were expressed in the RSA (top). The FG ⁺ cell bodies (gray) were found in layer 5 of the RSA. Scale bar, 100 μm. Bottom, The FG ⁺ cell bodies (blue) were found in layer 5 of the RSG and were immunopositive for CaMKIIα (red, indicated by red arrow). Colocalized cells are indicated by white arrow. Scale bar, 20 μm. Data were replicated in four mice. E, AAVrg-CaMKIIα-mCherry was injected into the dorsal CA1 area, and AAVrg-CaMKIIα-EGFP was injected into the RSA to label RSG neurons projecting to these areas. Representative images show RSG neurons projecting to the CA1 (green) and RSA (red). Scale bar, 20 μm. Bar chart shows the average proportion of RSG neurons projecting to the dorsal CA1 and RSA of all labeled RSG neurons. Data were replicated in four mice.

Figure 10.

The replication examples show distinct projection patterns of the RSA and RSG. A, Left, AAV_DJ-CaMKIIα-mCherry was injected into the RSG (top). Right, mCherry signals (red) were expressed in RSG neurons under the control of CaMKIIα promoter (top). Scale bar, 200 μm. The mCherry signals of axonal projections were observed in the CA1 SRLM (left and middle, bottom) and RSA layer 2 area (right, bottom). Scale bar, 100 μm. Right, Magnified image of rectangle (bottom). Scale bar, 20 μm. Sections were costained with NeuN (green). SO, Stratum oriens; SP, stratum pyramidale; SR, stratum radiatum; SLM, stratum lacunosum moleculare. B, Left, Fluorogold (FG, 4%) was injected into the dorsal CA1 area (top). Right, FG signal (gray) was expressed in pyramidal cell layer of the dorsal CA1 (top). Scale bar, 100 μm. Bottom, The FG ⁺ cell bodies (blue and indicated by blue arrow) were found in layer 5 of the RSG and were immunopositive for CaMKIIα (red, indicated by red arrow). Colocalized cells are indicated by white arrow. Scale bar, 20 μm. C, Left, AAV_DJ-CaMKIIα-EGFP was injected into the RSA (top). Right,: mCherry signals (red) were expressed in RSA neurons under the control of CaMKIIα promoter (top). Scale bar, 200 μm. The mCherry signals of axonal projections were observed in the RSG (left, bottom), LA, LEnt (middle, bottom) and POR (right, bottom). Sections were costained with NeuN (green). Scale bar, 100 μm. D, Left, FG was injected into the RSA (top). Right, FG signal were expressed in the RSA (top). The FG ⁺ cell bodies (gray) were found in layer 5 of the RSA. Scale bar, 100 μm. Bottom, The FG ⁺ cell bodies (blue) were found in layer 5 of the RSG and were immunopositive for CaMKIIα (red, indicated by red arrow). Colocalized cells indicated by white arrow. Scale bar, 20 μm. E, AAVrg-CaMKIIα-mCherry was injected into the dorsal CA1 area, and AAVrg-CaMKIIα-EGFP was injected into the RSA to label RSG neurons projecting to these areas. Representative images show RSG neurons projecting to the CA1 (green) and RSA (red). Scale bar, 20 μm.

The RSG-to-CA1 pathway is required for the retrieval of remote CFMs

Given that the RSG projects to both the dorsal CA1 and RSA, we asked which RSG output pathway may orchestrate the retrieval of remote CFMs by using terminal-specific optogenetic silencing approaches. For this, we bilaterally injected AAV_DJ-CaMKIIα-eNpHR3.0-EYFP into the RSG and, during the same surgery, implanted bilateral optic cannula into the dorsal CA1. In control group, we injected AAV_DJ-CaMKIIα-EGFP and implanted optic cannula into the dorsal CA1. At 14 d following AAV injection, mice were subjected to CFC, and optogenetic stimulation was delivered during the memory retrieval test 14 d after CFC (Fig. 11A). A post hoc histologic analysis revealed the expression of eNpHR3.0-EYFP in the RSG and EYFP-labeled projections in the CA1 SRLM of the DH (Fig. 11B). Functional validation of AAV-infected neurons using ex vivo whole-cell patch-clamp recordings confirmed that constant red light illumination (625 nm) reliably elicited a strong hyperpolarization and reversibly abolished spiking responses to injected square current pulses in a sustained manner in eNpHR3.0⁺ neurons (Fig. 11C). No significant differences were observed in freezing levels across conditioning trials between CaMKIIα-EGFP- and CaMKIIα-eNpHR3.0-EYFP-treated mice (Fig. 11D). We observed no significant differences in freezing between EGFP- and eNpHR3.0-treated groups during light-OFF periods on day 14 after CFC. During light-ON periods, freezing was significantly lower in eNpHR3.0-treated group than in the EGFP-treated group (Fig. 11E). A pairwise comparison confirmed a significant group difference in light-induced change in freezing during the test (Fig. 11F).

Figure 11.

Optogenetic inhibition of RSGγCA1 projection impairs the retrieval of remote CFMs. A, Schematic representation of the experimental design. Two weeks after stereotaxic injection of AAV_DJ-CaMKIIα-eNpHR3.0-EYFP or AAV_DJ-CaMKIIα-EGFP into the RSG, mice were trained in a CFC paradigm, and memory retention was tested 14 d after training. Optic fibers were implanted 7 d after training. Mice were subjected to constant red light illumination (625 nm) onto the CA1 during the retrieval test. B, EYFP signals were expressed in the RSG, and EYFP signals of axonal projections were observed in strata radiatum (left) and lacunosum moleculare (right) of the CA1 area. Scale bar, 100 μm. Right, Magnified image of rectangle (bottom). Scale bar, 20 μm. SO, Stratum oriens; SP, stratum pyramidale; SR, stratum radiatum; SLM, stratum lacunosum moleculare. C, Top, Representative whole-cell current-clamp recording showing red-light-mediated (625 nm, 300 ms) inhibition of evoked spike firing in eNpHR3.0⁺ RSG neurons held at −70 mV. The square pulse below the voltage trace indicates the timing of current injection through the patch pipette that was sufficient to induce action potential firing. The red line above the voltage trace indicates time of 300 ms activation of eNpHR3.0. Bottom, Representative whole-cell voltage-clamp recording showing outward photocurrent in a eNpHR3.0⁺ RSG neuron exposed to a 200 ms, 625 nm light pulse. D, The learning curve for three acquisition trials of contextual fear conditioning training in mice with EGFP or NpHR before retrieval test [EGFP, n = 9; NpHR, n = 9; interaction (trial × treatment): F_{(3, 64)} = 0.12, p = 0.95; trial variable: F_(3,64) = 21.65, p < 0.001; treatment variable: F_(1,64) = 0.047, p = 0.83; two-way ANOVA]. E and F, Optogenetic inhibition of RSGγCA1 projection decreased freezing levels in mice expressing eNpHR3.0-EYFP. E, Summary of experiments showing the effect of optogenetic inhibition of RSGγCA1 projection on the retention of remote CFMs in mice that received bilateral injections of AAV_DJ-CaMKIIα-eNpHR3.0-EYFP or AAV_DJ-CaMKIIα-EGFP into the RSG [EGFP, n = 9; NpHR, n = 9; interaction (treatment × light): F_{(2, 48)} = 3.25, p = 0.05; treatment variable: F_(1,48) = 1.03, p = 0.31; light variable: F_(2,48) = 3.34, p = 0.04; two-way ANOVA followed by Bonferroni's post hoc test]. F, Summary bar graphs depicting the fear memory retention test at 14 d after receiving contextual fear conditioning training in mice receiving red light off or on during retrieval test [EGFP, n = 9; NpHR, n = 9; interaction (treatment × light): F_{(1, 32)} = 4.84, p = 0.04; treatment variable: F_(1,32) = 5.47, p = 0.03; light variable: F_(1,32) = 2.58, p = 0.12; two-way ANOVA followed by Bonferroni's post hoc test]. Data indicate mean ± SEM; *p < 0.05 and **p < 0.01.

Next, we examined the role of the RSG-to-RSA pathway in retrieving remote CFMs. We injected AAV_DJ-CaMKIIα-eNpHR3.0-EYFP into the RSG bilaterally and implanted bilateral optic cannula into the RSA (Fig. 12A). A post hoc histologic analysis revealed the expression of eNpHR3.0-EYFP in the RSG and EGFP-labeled projections in layer 2 of the RSA (Fig. 12B). There were no significant differences in freezing levels across conditioning trials between mice treated with CaMKIIα-EGFP and CaMKIIα-eNpHR3.0-EYFP (Fig. 12C). Unlike the RSG-to-RSA pathway inhibition, silencing of the terminals of the RSG-to-RSA projecting axons in the RSA during the memory retrieval test did not affect freezing levels in the mice treated with CaMKIIα-eNpHR3.0-EYFP compared with those treated with CaMKIIα-EGFP on day 14 after CFC (Fig. 12D). A pairwise comparison showed no significant group difference in freezing during the test (Fig. 12E). Together, these findings reveal a key role of the RSG-to-CA1 pathway in the retrieval of remote CFMs.

Figure 12.

Optogenetic inhibition of RSGγRSA projection does not affect the retrieval of remote CFMs. A, Schematic representation of the experimental design. Two weeks after stereotaxic injection of AAV_DJ-CaMKIIα-EGFP or AAV_DJ-CaMKIIα-eNpHR3.0-EYFP into the RSG, mice were trained in a CFC paradigm, and memory retention was tested 14 d after training. Optic fibers were implanted 7 d after training. Mice were subjected to constant red light illumination (625 nm) onto the RSA during the retrieval test. B, EYFP signals were expressed in the RSG, and EYFP signals of axonal projections were observed in layer 2 of the RSA. Scale bar, 100 μm. Right, Magnified image of rectangle (bottom). Scale bar, 20 μm. C, The learning curve for three acquisition trials of contextual fear conditioning training in mice with EGFP or NpHR before retrieval test [EGFP, n = 7; NpHR, n = 7; interaction (trial × treatment): F_{(3, 48)} = 1.68, p = 0.18; trial variable: F_(3,48) = 82.12, p < 0.001; treatment variable: F_(1,48) = 0.33, p = 0.57; two-way ANOVA]. D, Optogenetic inhibition of RSGγRSA projection did not affect freezing levels in mice expressing eNpHR3.0-EYFP. Summary of experiments showing the effect of optogenetic inhibition of RSGγRSA projection on the retention of remote CFMs in mice that received bilateral injections of AAV_DJ-CaMKIIα-eNpHR3.0-EYFP or AAV_DJ-CaMKIIα-EGFP into the RSG [EGFP, n = 7; NpHR, n = 7; interaction (treatment × light): F_{(2, 36)} = 0.07, p = 0.93; treatment variable: F_(1,36) = 1.53, p = 0.22; light variable: F_(2,36) = 0.07, p = 0.93; two-way ANOVA]. E, Summary bar graphs depicting the fear memory retention test at 14 d after receiving contextual fear conditioning training in mice receiving red light off or on during retrieval test [EGFP, n = 7; NpHR, n = 7; interaction (treatment × light): F_{(1, 24)} = 0.14, p = 0.71; treatment variable: F_(1,24) = 0.13, p = 0.72; light variable: F_(1,24) = 0.41, p = 0.52; two-way ANOVA]. Data indicate mean ± SEM.

Prior studies on the retrieval of remote fear memories have focused on the 28 d postconditioning time point (Goshen et al., 2011; Todd et al., 2016), and brain circuits underlying memory may differ at different time points. To test whether RSG-to-CA1 activity is necessary for retrieval of remote memory on day 28 after CFC, we bilaterally injected control AAV_DJ-CaMKIIα-mCherry or AAV_DJ-CaMKIIα-hM4D(Gi)-mCherry into the RSG of mice. After allowing 2 weeks for viral expression, mice underwent CFC training and were bilaterally implanted with cannula targeting the dorsal CA1 2 weeks later. A memory retention test was conducted 28 d after CFC training in mice that received local injection of vehicle or CNO into the dorsal CA1 (Fig. 13A). A post hoc histologic analysis revealed the expression of hM4D(Gi)-mCherry with the neuronal marker NeuN in the RSG (Fig. 13B) and mCherry-labeled projections in the SRLM of the dorsal CA1 (Fig. 13D). There was no significant difference in the percentage of mCherry-expressing cells in the RSG between hM4D(Gi)/CNO or mCherry/CNO treatment groups (Fig. 13C). Similar to our 14 d data, we observed that chemogenetic silencing of RSG-to-CA1 activity (hM4D(Gi)/CNO) resulted in a significant decrease in freezing levels on day 28 after CFC training compared with mCherry/vehicle, mCherry/CNO, or hM4D(Gi)/vehicle treatment groups (Fig. 13E).

Figure 13.

Chemogenetic silencing of RSGγCA1 projection impairs fear memory retention 28 d after CFC training. A, Schematic representation of the experimental design. Two weeks after stereotaxic injection of AAV_DJ-CaMKIIα-mCherry or AAV_DJ-CaMKIIα-hM4D(Gi)-mCherry into the RSG, mice were trained in a CFC paradigm, and fear memory retention was tested 28 d after training. Cannulas were implanted over dorsal CA1 2 weeks after CFC training to allow local and selectively delivery of vehicle (Veh) or CNO to achieve specific silencing of RSG projections to the dorsal CA1. Veh (1% DMSO in PBS, 0.5 µl) or CNO (1.4 µg/µl, 0.5 µl) was bilaterally infused into the dorsal CA1 through cannulas 30 min before the retrieval test. B, Representative images of mCherry and NeuN overlap in the RSG of mouse injected previously with AAV_DJ-CaMKIIα-hM4D(Gi)-mCherry. Scale bar, 100 μm. Right, Magnified image of rectangle. Scale bar, 20 μm. C, Summary graphs depicting the percentages of mCherry⁺/DAPI⁺ cells in the RSG from AAV_DJ-CaMKIIα-mCherry + CNO (n = 5) and AAV_DJ-CaMKIIα-hM4D(Gi)-mCherry + CNO (n = 5; t₍₈₎ = 1.25, p = 0.25, unpaired Student's t test). D, Representative images of mCherry-labeled RSG axonal projections (red) in the dorsal CA1 of mouse injected previously with AAV_DJ-CaMKIIα-hM4D(Gi)-mCherry. Scale bar, 100 μm. Right, Magnified image of rectangle. Scale bar, 20 μm. SO, Stratum oriens; SP, stratum pyramidale; SR, stratum radiatum; SLM, stratum lacunosum moleculare. E, Summary of experiments showing the effects of Veh or CNO infusion into the dorsal CA1 on the retrieval of 28-d-old CFMs in mice injected previously with AAV_DJ-CaMKIIα-hM4D(Gi)-mCherry or AAV_DJ-CaMKIIα-mCherry into the RSG [mCherry + Veh, n = 5; hM4D(Gi) + Veh, n = 5; mCherry + CNO, n = 6; hM4D(Gi) + CNO, n = 8; interaction: F_(1,20) =24.29, p < 0.001 interaction (treatment × light): F_(1,20) =24.29, p < 0.001; treatment variable: F_(1,20) = 0.32, p = 0.58; light variable: F_(1,20) = 17.94, p < 0.001; two-way ANOVA followed by Bonferroni's post hoc test]. Data indicate mean ± SEM; *p < 0.05 and ***p < 0.001.

RSG projection neurons make monosynaptic excitatory synapses onto CA1 neurons

Finally, to further confirm that the RSG projection neurons form direct synaptic connectivity with CA1 neurons, we performed monosynaptic input tracing using a conditional rabies virus approach (Wall et al., 2010; Wickersham et al., 2007). To do this, we first unilaterally injected a viral cocktail (AAV₅-FLEX-TVA-GFP and AAV₅-FLEX-RG) into the dorsal CA1 of CaMKIIα-Cre and vesicular GABA transporter (Vgat)-Cre mice, followed 21 d later by injection of G-deleted pseudotyped rabies virus expressing mCherry (EnvA-ΔG-mCherry) to infect Cre-expressing neurons (Fig. 14A). Consistent with our tracing data (Figs. 9, 10), confocal imaging results from CaMKIIα-Cre mice showed that starter cells (GFP⁺/mCherry⁺) were primarily located in the dorsal CA1 pyramidal layer, with mCherry⁺ monosynaptic excitatory input neurons (CaMKIIα⁺/mCherry⁺) found in layer 5 of the RSG (Fig. 14B). In Vgat-Cre mice, we also found retrogradely labeled (CaMKIIα⁺/mCherry⁺) neurons in layer 5 of the RSG (Fig. 14C). These results suggest that RSG axons form monosynaptic excitatory synapses onto dorsal CA1 pyramidal neurons and GABAergic interneurons.

Figure 14.

Monosynaptic rabies tracing of excitatory inputs from the RSG to the CA1. A, Schematic depiction of viral injection. AAV₅-FLEX-TVA-GFP, AAV₅-FLEX-RG and EnvA-ΔG-mCherry were unilaterally injected into the dorsal CA1. Three weeks after stereotaxic injection of AAV₅-FLEX-TVA-GFP and AAV₅-FLEX-RG, EnvA-ΔG-mCherry was injected into the dorsal CA1 region of either CaMKIIα-Cre or Vgat-Cre mice. Ten days later, mice were killed and perfused, and slices were prepared for fluorescence microscopy. B, Left, Representative images (top) of mCherry⁺ presynaptic cells in layer 5 of the RSG with expression of excitatory marker CaMKIIα in CaMKIIα-Cre mice. Right, Representative images (top) of AAV₅-helper virus-infected starter cell (yellow) in the CA1 region. Scale bar, 100 μm. Bottom, Higher magnification images of the dashed white box (top). Scale bar, 20 µm. Left, White arrowhead (bottom) indicates cell doubled for CaMKIIα (green) and mCherry (red). Right, White arrowheads (bottom) indicate cell doubled for GFP (green) and mCherry (red). Data were replicated in five mice. C, Left, Representative images (top) of mCherry⁺ presynaptic cells in layer 5 of the RSG with expression of excitatory marker CaMKIIα in Vgat-Cre mice. Right, Representative images(top) of AAV₅-helper virus-infected starter cell (yellow) in the CA1 region. Scale bar, 100 μm. Bottom, Higher magnification images of the dashed white box (top). Scale bar, 20 µm. Left, White arrowhead (bottom) indicates cell doubled for CaMKIIα (green) and mCherry (red). Right, White arrowheads (bottom) indicate cell doubled for GFP (green) and mCherry (red). SO, Stratum oriens; SP, stratum pyramidale; SR, stratum radiatum; SLM, stratum lacunosum moleculare. Data were replicated in four mice.

Discussion

Systems consolidation theory proposes that as memories become more remote, neocortical structures become more important for storage and retrieval (Kitamura et al., 2017; Albo and Gräff, 2018; Barry and Maguire, 2019). The RSC is thought to be a critical site for long-term memory storage (Todd and Bucci, 2015; Todd et al., 2019). In this study, we attempted to dissect the specific contributions of RSC subdivisions and related circuit mechanisms underlying the retrieval of remote memories. To do so, we used a classic CFC paradigm, animal model widely used for exploring the neural substrates of learning and memory. Our results revealed that inactivation of the RSG, but not RSA, disrupts the retrieval of remotely acquired CFMs. In addition, we identified a novel RSG-to-CA1 pathway that is critically and specifically important for remote CFM recall. Therefore, our data uncover a previously unrecognized role for RSG and its projection to the dorsal CA1 in retrieving remote memories.

Current understanding of the involvement of the RSC in memory processes rests mainly on evidence from lesion and pharmacological inactivation studies (Keene and Bucci, 2008; Todd and Bucci, 2015). In this context, a previous study in mice found that local injection of NMDA receptor antagonists into the RSC before the memory retrieval test impaired the retrieval of recent and remote CFMs (Corcoran et al., 2011), indicating that the RSC is necessary for retrieving CFMs. Unfortunately, this study is limited by treating the RSC as a unitary structure and fails to distinguish the contribution of particular neuronal populations of the RSC in CFMs. Here, we provide convincing evidence supporting distinct roles for the RSG and RSA in CFM processing. The results of our chemogenetic manipulation experiments strongly suggest that activity in the RSG is specifically required for the retrieval of remote CFMs. This finding is in line with a previous study demonstrating that selective degeneration of layers 4–5 neurons of the RSG impaired freezing response during the fear recall on day 12 after CFC (Sigwald et al., 2019). Thus, the RSG and RSA appear to serve distinct functions in regulating memory processing. Further support comes from immediate-early gene expression assay studies confirming different contributions of the RSG and RSA to spatial working memories (Pothuizen et al., 2009). The present findings reinforce the concept that over time, memories become stored more permanently in distributed regions of the neocortex (Winocur et al., 2010; Albo and Gräff, 2018; Tonegawa et al., 2018).

A key finding of the present work is that a proportion of RSG layer 5 neurons projects to the dorsal CA1 and is recruited during the retrieval of remote CFMs. To our knowledge, this is the first functional description of the RSG-to-CA1 pathway underlying the retrieval of remote fear memories. In view of the connections between the RSC and hippocampus, previous studies have concentrated on the hippocampal inputs to the RSC. For example, a study has shown vGlut1- and vGlut2-expressing pyramidal neurons in the dorsal subiculum project to layer 3 of the RSC and distinctly regulate RSC local cellular networks, which differentially contribute to the formation and persistence of CFMs (Yamawaki et al., 2019a). In addition to receiving excitatory inputs from the hippocampus, a direct GABAergic projection from the hippocampus to the RSG has also been described, arising from neurons in the dorsal CA1 at the SRLM border that send long-range axons to layer 1 of the RSG (Yamawaki et al., 2019b). Our viral tracing results indicate that layer 5 pyramidal neurons in the RSG provide monosynaptic excitatory inputs to the SRLM of the dorsal CA1. This finding is consistent with mouse brain connectivity data from the Allen Mouse Brain Connectivity Atlas (https://connectivity.brain-map.org; Oh et al., 2014). As far as we know, this is the first time that this excitatory projection has been functionally examined. The RSG–CA1 glutamatergic connection that we identified here adds neuroanatomical evidence supporting the functional interaction between the RSC and the hippocampus during remote memory retrieval. Our data are also consistent with model of systems consolidation that emphasizes a prolonged role for the hippocampus in memory retrieval (Wirt and Hyman, 2019). Nonetheless, if interactions between the RSG and dorsal CA1 increase as memories progressed from recent to more remote remains to be investigated in the future. However, our findings should also be interpreted with caution because only males were used for test subjects.

How do subdivisions of the RSC communicate with each other? Previous studies have documented extensive interconnections between the RSG and RSA, arising from pyramidal neurons in layer 4/5 of the RSG that send ipsilateral and contralateral axonal collaterals to superficial layer 2/3 of the RSA (Sugar et al., 2011; Sigwald et al., 2019). It is noteworthy that although our results found that a proportion of RSG neurons project to the RSA, optogenetic silencing of the activity of the RSG-to-RSA pathway does not affect remote CFM recall. The specific requirement for activity in the RSG-to-CA1, but not the RSG-to-RSA, pathway during the retrieval of remote CFMs suggest that RSG output pathways are function segregated in the processing of CFMs. The functional dissociation between these two RSG output pathways is in line with our observation that only a small proportion of RSG neurons projects to both the dorsal CA1 and RSA. To the best of our knowledge, this is the first report to uncover the function heterogeneity of the RSG-to-CA1 and RSG-to-RSA pathways. Although a study proposes a functional relevance of RSG input to the RSA in driving freezing responses during CFM recall (Sigwald et al., 2019), our results do not support the involvement of the RSG-to-RSA pathway in retrieving remote CFMs. This discrepancy in findings may be because of differences in time points of memory test (recent vs remote) or animal species used (rats vs mice). It is noteworthy that although Sigwald et al. (2019) have suggested that efferent connections from RSG neurons are functionally required for activating the RSA during CFM retrieval, they did not directly examine the effect of inactivating the RSG-to-RSA pathway on CFM recall. Instead, our optogenetic silencing of the RSG-to-RSA projecting axons in the RSA does not affect remote CFM recall. Nevertheless, additional research is needed to delineate how the RSG-to-CA1 pathway mediates remote CFM retrieval at the level of synapses. It was surprising to find that in addition to excitatory pyramidal neurons, GABAergic interneurons in the CA1 also receive excitatory inputs from the RSG. As the generation of rhythmic activity in the hippocampal network is tightly controlled by GABAergic interneurons, future studies will have to determine whether this excitatory connection can modulate memory function by modulating oscillatory activity within the hippocampus or facilitating interregional temporal coordination of activity, which are crucial for memory function.

Unlike the ACC, which is only necessary for remote CFC recall (Frankland et al., 2004; Goshen et al., 2011; Einarsson et al., 2015), our results reveal that activity in the RSG is necessary for the retrieval of both recent and remote CFMs. This finding is consistent with the results of a previous study demonstrating that infusion of NMDA receptor antagonist into the RSC before testing impaired the retrieval of recent and remote CFMs (Corcoran et al., 2011). Like the RSG, pharmacological inactivation of the medial prefrontal cortex (Blum et al., 2006; Leon et al., 2010) and excitotoxic lesions of perirhinal and postrhinal cortices (Burwell et al., 2004) have been shown to impair the retrieval of recent and remote CFMs. These findings suggest that the RSG and some cortical areas may therefore serve as nodal points for time-independent CFM retrieval. Interestingly, a study by de Sousa et al. (2019) showed that optogenetic reactivation of memory engram cells in the RSC a day after CFC produced a recent memory with features normally observed in consolidated remote memories. This finding suggests that the RSC engram cells necessary for remote CFM storage form at an early point in learning but are initially immature, and the process of long-term CFM consolidation relies on the maturation of these RSC engram cells.

In conclusion, we used a CFC paradigm in mice to uncover a selective and pivotal role for RSG pyramidal neurons in retrieving remotely acquired fear memories. We also identify a novel RSGγCA1 pathway in mediating remote CFM recall. These results fit well with findings of functional magnetic resonance imaging studies in humans showing increased RSC activity during the retrieval of remote spatial (Rosenbaum et al., 2004), autobiographical memories (Steinvorth et al., 2006; Svoboda et al., 2006) in healthy subjects and traumatic memories in patients with post-traumatic stress disorder (Piefke et al., 2007). Our findings contribute to a better understanding of circuit mechanisms by which the RSG and its circuit contribute to remote CFM recall. Relevant information may provide a novel therapeutic avenue to attenuate remote traumatic memories, often observed in patients with post-traumatic stress disorder.