Wild-type C57BL/6J mice (DBL, Republic of Korea) were utilized at 5 weeks of age at the onset of the experiments. The animals were maintained in group housing conditions, with 4–5 per cage, under a 12-h light/dark cycle at room temperature. They had access to a standard diet and water ad libitum. RARE-LacZ mice (TG(RARE-Hspa1b/lacZ)12Jrt/J, JAX stock #008477) were bred to maintain hemizygosity (Rossant et al., 1991). These transgenic mice were between 8 and 15 weeks old at the start of the experiments. All experimental procedures were approved by the DGIST Animal Care and Use Committee, Republic of Korea (DGIST-IACUC-19040202-0006).

Dietary treatment commenced when the C57BL/6J mice reached 5 weeks of age. The mice were housed in groups of 4–5 per cage and were provided with either a vitamin A deficient (VAD) diet (TD.86143, Envigo) or a control diet (TD.91280, Envigo). Body weight and pellet consumption for all mice were recorded every week. Behavioral assessments were initiated after 12 weeks on the VAD diet. Subsequently, vitamin A replenishment began with the introduction of a standard diet for 3 weeks.

To prepare the LE135 solution, 10 mg of LE135 was dissolved in 2.3 ml DMSO to make a 10 mM stock solution. On the injection day, LE135 was diluted with saline at 10 μM. Stereotaxic infusion of retinoic acid receptor beta (RARβ) antagonist LE135 was performed using an Angle Two™ stereotaxic frame (Leica, Grove, IL, USA). Primary anesthesia was induced by adding 5% isoflurane to the chamber, and 1%–3% isoflurane was given to anesthetized mice using a nose mask with a stereotaxic instrument. Depending on the drug group, each mouse was delivered vehicle (0.1% DMSO in saline) or LE135 500 nl per injection site into bilateral dorsal DG region (AP: −2.00; ML: ±1.35; DV: −1.95) using a nanofil syringe. The flow rate was controlled at 100 nl/min by a Legato^® 130 controller (KD Scientific, USA). Following the injection, 5 min waiting period was observed before pulling out the needle and closing the incision; then, the mice were returned to their home cage for recovery. Mice were given a 3-day rest before proceeding with the electrophysiological recording experiments.

Acute hippocampal slices (thickness, 300 μm) were prepared from the brains of C57BL/6J mice of either sex. Mice were anesthetized with isoflurane and decapitated immediately. All samples were obtained coronally for the dorsal hippocampus. Slices were prepared in an oxygenated ice-cold physiological saline using a vibratome (VT1200S, Leica), incubated at ∼32°C for 30 min, and subsequently maintained in the artificial cerebral spinal fluid (ACSF) at room temperature until the recordings. Recordings were performed at near-physiological temperature (∼32°C) in an oxygenated ACSF.

Patch pipettes were obtained from borosilicate glass capillaries (outer diameter = 1.5 mm, inner diameter = 1.1 mm) with a horizontal pipette puller (P-1000, Sutter Instruments). The open-tip resistance of patch pipettes was 4.5–6.5 MΩ for recordings. Current clamp recordings were performed with an EPC-10 USB Double amplifier (HEKA Elektronik) and stored using Patchmaster software. In current-clamp recordings, series resistance was 10–20 MΩ. All experiments were performed on visually identified GCs under DIC optics. GCs located in the middle regions of the suprablade GC were purposely targeted. Around 3 min after patch break-in, resting membrane potential (RMP) was measured, and input resistance (Rin) was determined by applying Ohm’s law to the steady-state voltage difference resulting from a hyperpolarizing current (−10 pA, 500 ms). Pipette capacitance and series resistance (Rs) compensation (bridge balance) were done at the beginning of the clamp recordings.

The extracellular solution for dissection was a choline chloride-based solution (25 mM NaHCO₃, 2.5 mM KCl, 1.25 mM NaH₂PO₄, 7 mM MgCl₂, 0.5 mM CaCl₂, 25 mM glucose, 11.61 mM ascorbic acid, 3 mM pyruvic acid, and 110 mM choline chloride). Physiological saline for experiments was standard ACSF (119 mM NaCl, 26 mM NaHCO₃, 2.5 mM KCl, 1.25 mM NaH₂PO₄, 1 mM MgSO₄, 2 mM CaCl₂, 0.4 mM ascorbic acid, 2 mM pyruvic acid, and 20 mM glucose). For whole-cell recording, we used K⁺ rich intracellular solution that contained 125 mM K-gluconate, 20 mM KCl, 10 mM HEPES, 0.5 mM EGTA, 4 mM ATP, 10 mM phosphocreatine, and 0.3 mM Tris GTP, pH adjusted to 7.2–3 with KOH (∼300 mOsm).

All mice were anesthetized with Avertin (250 mg/kg) via intraperitoneal injection and perfused with PBS, followed by 4% paraformaldehyde (PFA). The brains were then extracted and post-fixed in 4% PFA overnight, dehydrated in 15% sucrose for 12 h, and finally in 30% sucrose overnight at 4°C. The fully saturated brains were sectioned coronally at 40 μm using a Cryostat (CM3050S, Leica). The sections underwent a free-floating treatment process.

For immunostaining, sections were first blocked with blocking buffer (5% goat serum in PBS) for 1 h at room temperature (RT). They were then incubated with primary antibodies diluted in the blocking buffer for 24 h at 4°C. The primary antibodies used included β-galactosidase (chicken, polyclonal IgY, 1:1,000, ab9361, Abcam), calbindin1 (mouse, monoclonal IgG1, 1:750, CB300, Swant), and c-Fos (rabbit, polyclonal, 1:500, ab102499, Abcam), Prox1 (rabbit, polyclonal, 1:500, #925201, BioLegend), Parvalbumin (mouse, monoclonal, 1:250, PV235, Swant). After incubation, sections were washed thrice with washing buffer (0.2% Triton X-100 in PBS) for 5 min each at RT. Sections were then incubated with Alexa Fluor-conjugated secondary antibodies (1:400, Life Technologies) or DRAQ5 (1:2,000, 62251, ThermoFisher) for 3 h at RT. Following this, sections were rewashed thrice with washing buffer and stained with DAPI (1:1,000, Sigma-Aldrich). Finally, they were mounted with Prolong Gold (Life Technologies, USA) anti-fade mounting medium. Imaging and analysis of the sections were performed using a Zeiss LSM800 confocal microscope. Fluorescence-positive cells were quantified through hand-counting and using MetaMorph software, and regions of interest (ROIs) within the DG were evaluated for their area size. The cell count was normalized by the area size of the ROI, and the cell density is presented as [counting number per mm²].

To monitor RA depletion in the VAD model, livers were harvested post-euthanasia under deep anesthesia. Each liver was rapidly sectioned into eighths at 4°C and immediately flash-frozen in liquid nitrogen. Total RNA was extracted from the liver sections using the Axygen^® AxyPrep MicroRNA (miRNA) Miniprep Kit and subsequently reverse-transcribed into cDNA. Quantitative PCR (qPCR) was performed using 10 ng of cDNA with the following primers for RARβ: 5′-TATGAGATGACAGCGGAGCTAGAC-3′ (forward) and 5′-GGCTTTCCGGATCTTCTCAGT-3′ (reverse). Each sample was assayed in triplicate. The qPCR analysis was conducted on an AriaMx Real-Time PCR system (Agilent Technologies, CA, USA) using the following thermal cycling conditions: an initial denaturation at 95°C for 30 s, followed by 40 cycles of denaturation at 95°C for 5 s and annealing/extension at 60°C for 30 s (Basu et al., 2015). The relative quantification of mRNA levels was determined, including normalization to the gapdh housekeeping gene.

Environmental novelty-evoked GC activation was assessed after acclimating all mice in the test room for a sufficient period. The NE consisted of a box (30 cm × 30 cm × 30 cm) outfitted with toys serving as novel objects. As indicated in the figures, each mouse was introduced into the box and kept there for the duration specified in each figure. Subsequently, the mice were quickly perfused, and immunohistochemistry was performed with c-Fos staining to identify activated GCs. The control group remained in their home cages until the time of perfusion. To assess c-Fos expression in RA-responsive GCs, animals were exposed to NE for 1.5 h, coinciding with peak expression of immediately early genes as a neural activation marker in active neurons (Chaudhuri et al., 2000; Paul et al., 2020). Additionally, a 30-min NE exposure was used to examine the effect of VAD on GC activation at the time point when GC activation is on the rise but has not yet reached its peak.

General behavior tests in the Phenotyper™: Cages were acclimated to the test room for 1 h during a light cycle. Each mouse was placed in a Phenotyper™ (Noldus, USA) enclosure for 24 h, equipped with a feed tray, a water bottle, and an overhead camera. Locomotion and feeding frequency were monitored using EthoVision™ software (Noldus, USA) (Jankovic et al., 2019; Rhine et al., 2019).

Data are expressed as the mean ± standard error of the mean (SEM). Statistical analyses were performed using GraphPad Prism software, Version 8.4. The data were compared using an unpaired, two-tailed t-test, two-way ANOVA was employed, with Šídák’s and Tukey’s post-hoc test applied for further analysis. The results of all statistical tests are described in Supplementary Table 1. Significance levels are denoted as follows: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, with “ns” indicating no significance.

Retinoic acid-responsive cells have been identified in the adult mouse hippocampus (McCaffery et al., 2006; Goodman et al., 2012). In this study, we characterized the spatial distribution of RA-responsive cells in the DG. Utilizing RARE-LacZ reporter mice, we visualized RA-responsive cells where β-galactosidase (β-gal) expression is induced by RARβ binding to RARE (Figure 1A). We found β-gal-positive RA-responsive cells localized to the GC layers along the dorsoventral axis of the DG, but not within the CA1 or CA3 subregions, corroborating previous findings (McCaffery et al., 2006; Goodman et al., 2012). Notably, β-gal expression coincided with the mature GC marker calbindin1 (Calb1) in the DG (Figure 1C). Approximately 40% of the calbindin1-positive mature GCs were RA-responsive (42.9% ± 4.1% in dGCL, 35.4% ± 3.3% in vGCL), a proportion maintained throughout the dorsoventral extent of the hippocampus (Figures 1C–E). RA-responsive cells were more prevalent in the infrablade compared to the suprablade of the DG, revealing an uneven spatial distribution within these subregions (Figures 1F, G). Furthermore, the β-gal signal colocalized with Prox1 an excitatory GC marker, but not with parvalbumin an inhibitory basket cell marker, indicating the cell type identity of RA-responsive cells in the DG (Supplementary Figure 1). Thus, we have delineated a distinct RA-responsive subpopulation among mature GCs and mapped their specific spatial distribution within the DG.

Granule cells in the DGs are known to be sparsely activated in response to spatial or contextual stimuli, a critical feature for the pattern separation process within the DG, which enables discrimination between similar contexts or spaces (Erwin et al., 2020; Lamothe-Molina et al., 2022). RARE-LacZ reporter mice were used to investigate the reactivity of RA-responsive cells to a NE. These mice were either exposed to NE or remained in their home cage (HC) as a control (Figure 2A). For immunohistochemical analysis, β-gal was used as a marker for RA-responsive GCs, and c-Fos served as an immediate early gene indicator of neuronal activation. The density of GCs exhibited no difference between the supra- and infrablade of GC layers (Supplementary Figure 2). Therefore, the area value of each subregion was used to calculate the density of RA-responsive and c-Fos-positive GCs in the supra- and infrablade of the GC layer. Notably, the relative number of β-gal-positive RA-responsive GCs remained constant, showing no change with NE exposure (Figures 2B, E). In contrast, a significant increase in c-Fos-positive activated GCs was observed following NE exposure compared to the HC group (Figures 2B, C). Notably, while RA-responsive GCs were distributed more in the infrablade of the DG than the suprablade (Figures 2B, F), c-Fos-positive GC activation was notably higher in the suprablade than in the infrablade (Figures 2B, D). High-resolution imaging revealed that c-Fos immunoreactivity rarely overlapped with the β-gal signal within the GC layer of the DG (Figures 2G, H). The density of activated GCs due to NE exposure increased from 125.8 ± 29.0 to 304.0 ± 41.0 cells/mm² (Figure 2C). However, the proportion of GCs positive for both c-Fos and β-gal constituted less than 0.8% of the total RARE-positive GCs (Figure 2H), suggesting that RA-responsive GCs are non-reactive to activation by NE stimuli. This finding underscores the specificity of RA signaling pathways in the DG and highlights the potential for distinct regulatory mechanisms governing the activation of RA-responsive GCs in response to environmental changes.

Prompted by our initial observations, we explored the potential causal relationship between RA signaling and GC excitability. We first assessed the impact of dietary vitamin A depletion on GC activation. Mice aged 5 weeks were fed either a VAD diet or a control diet for 18 weeks (Etchamendy et al., 2003). There was no observed difference in body weight gain between control and VAD mice (Supplementary Figure 3A). However, a significant reduction in the expression of the autoregulatory gene RARβ, an indicator of RA signaling, was noted in the livers of VAD mice, confirming systemic RA depletion in our model (Supplementary Figure 3B). Subsequent exposure to a NE or maintenance in the HC was followed by an assay for c-Fos immunoreactivity (Figure 3A). Notably, VAD mice exhibited a more significant increase in NE-induced activated GCs than controls (Figure 3B). High-resolution imaging and quantitative analysis further confirmed that chronic dietary RA depletion significantly increases GC activation when exposed to NE (Figures 3C, D).

We then explored the effects of acute pharmacological inhibition of RARβ on GC activation. The RARβ antagonist LE135 (Yin et al., 2014), or vehicle was infused into the bilateral DG. Three days after stereotaxic surgery, we performed whole-cell patch clamp recordings and measured intrinsic properties on acute hippocampal slices. The passive properties, including RMP (VEH, −73.8 ± 0.9 mV; LE135, −77.0 ± 1.1 mV; p = 0.060) and input resistance (VEH, 316.4 ± 30.1 MΩ; LE135, 285.2 ± 30.4 MΩ; p = 0.496), were not significantly different between groups (Figure 3G). However, GCs in LE135-treated group (LE135) showed higher excitability than GCs in control group (VEH) in response to 1 s depolarization step current pulse (Figures 3E, F), supporting inhibitory effect of RA signaling on GC excitability in the DG. Through chronic dietary vitamin A depletion and acute pharmacological RARβ inhibition approaches, we have demonstrated that RA signaling suppresses RA-responsive GCs, thereby scaling down NE-induced GC activation in the DG. This abnormal increase in GC activation aligns with the observed non-reactive nature of RA-responsive GCs, suggesting a complex interplay between RA signaling and GC responsiveness to environmental stimuli.

Building on our discovery of RA’s role in modulating GC activation within the DG, we investigated the impact of RA depletion on hippocampus-dependent behaviors. Control and VAD mice were subjected to 18 weeks of dietary treatment before undergoing a battery of behavioral tests to assess general activity, anxiety, and spatial learning (Figure 4A). VAD did not significantly affect food consumption or voluntary movement in the general behavior tests (Supplementary Figures 4A, B) nor altered anxiety-related behaviors in the OFT, L&D, and EPM tests (Supplementary Figures 4C–E). Similarly, sucrose preference remained unchanged between VAD and control mice (Supplementary Figures 4F–H).

Spatial discrimination was evaluated using the Intellicage™ system and a place learning paradigm heavily reliant on hippocampal function (Galsworthy et al., 2005; Maroteaux et al., 2018; Voikar et al., 2018). The test involved distinguishing between identical chambers at the four corners of the cage, with access to water contingent upon visiting and nose-poking specific chambers (Supplementary Figures 5A–D). Each mouse was equipped with a RFID tag to monitor corner visits. Following habituation and adaptation to the Intellicage™, mice underwent place learning sessions, including initial learning and reversal learning phases. Corner preferences were measured before and after introducing sweetened water (4% sucrose) at a single corner (Figure 4B). Control mice gradually increased visits to the sucrose-designated corner during the initial learning session, whereas VAD mice demonstrated significantly poorer performance in locating the sucrose corner (Figure 4C). This pattern persisted during the reversal learning session, which evaluated the acquisition of new spatial memories (corner 3) and the extinction of previous ones (corner 1) (Figures 4D, E). Overall, control mice achieved a 52.8% ± 3.5% success rate in learning tasks, while VAD mice managed only a 31.1% ± 2.5% success rate (Figure 4F), suggesting that RA depletion via a VAD diet impairs spatial discrimination abilities in adult mice.

To determine if impaired spatial discrimination could be reversed, VAD mice were switched to a control diet immediately after the initial spatial discrimination test, creating a Dietary vitamin A replenishment (VAR) model (Bonhomme et al., 2014) over 3 weeks in their HC (Figure 4A). Subsequently, both control and VAR mice were retested for spatial discrimination. During the initial learning session, VAR mice displayed a complete restoration of their ability to locate the sucrose corner (corner 2), performing on par with control mice (Figure 4G). Similarly, during reversal learning, VAR mice successfully shifted their preference from the previously rewarded corner (corner 2) to the new sucrose corner (corner 4), mirroring the adaptability of control mice (Figures 4H, I). Overall, both groups showed comparable success rates in the total learning period, with control mice at 71.9% ± 5.3% and VAR mice at 65.7% ± 2.8% (Figure 4J). These results underscore the vital role of RA, the bioactive derivative of vitamin A, in hippocampus-dependent spatial discrimination.