All experiments were conducted in accordance with the Nagoya University Regulations on Animal Care and Use in Research and were approved by the Institutional Animal Care and Use Committee, Nagoya University (approval number R23026). Mice were group-housed and kept under 12-h light/dark cycle (lights on at 8:00) with food and water provided ad libitum unless otherwise noted. Wild-type C57BL/6J mice were purchased from SLC Japan. Prkcd-cre mice {Tg(Prkcd-glc-1/CFP,-cre)EH124Gsat; stock #011559-UCD} and Ai14 mice {B6. Cg-Gt(ROSA)26Sor tm14(CAG-tdTomato)Hze/J; stock #007914} were obtained from the Mutant Mouse Resource & Research Center and the Jackson Laboratory, respectively, and were bred on a C57BL/6J genetic background. Only male mice were used in this study.

Corticosterone (27840; Sigma-Aldrich) were suspended in corn oil (8001-30-7; Nacalai Tesque) to a concentration of 20 mg/mL, thoroughly sonicated, and administered subcutaneously at a dose of 40 mg/kg once daily between 10:00 and 12:00, starting at 9–10 weeks of age, according to the schedule shown in Figure 1A. Mice treated with vehicle (VEH, corn oil) on the same schedule were used as controls. Plasma corticosterone levels were measured using the AssayMax Corticosterone ELISA Kit (EC3001-1; AssayPro) according to the instructions provided. Blood samples were collected through the inferior vena cava under deep anesthesia with isoflurane at the time points for measurement, and anticoagulated with sodium citrate. Subsequently, the blood was centrifuged at 3,000 × g at 4°C for 10 min, and the supernatant was collected as plasma.

Mice were euthanized by cervical dislocation immediately after blood collection, and the brains were isolated and immediately frozen in crushed powder dry ice. Fresh frozen brains were coronally sliced at a 20-μm-thick using Cryostat (CM3050S; Leica), mounted on polyphenylene sulfide Frame Slide (11600294; Leica), and immediately dried up by air blowing. Microtissue dissection was performed using a 110-μm-inner diameter punching needle guided by tdTomato fluorescence under a fluorescence microscope as previously described (Yoda et al., 2017; Ueda et al., 2021). The isolated microtissues were dispensed into the PCR tubes filled with 99.5% ethanol and stored at −80°C until library preparation.

RNA-seq library preparation was performed as previously described (Yamazaki et al., 2020). Briefly, after removing the ethanol using a vacuum evaporator, the isolated microtissues were lysed in 5.3 μL of cell lysis mixture [PKD buffer (QIAGEN): Proteinase K (QIAGEN) = 16:1] at 56°C for 1 h, and the poly(A) RNA was purified with the oligo dT magnetic beads (61,005; Thermo Fisher Scientific) according to the instruction manual. The purified mRNA were directly processed according to the Smart-seq2 protocol (Picelli et al., 2014). PCR products were purified with 0.8 × volume of AMPure XP beads (A63880; Beckman Coulter).

Sequencing and data analysis were conducted as previously described (Yoda et al., 2017). The amplified cDNA (0.25 ng) was used for the preparation of the sequencing library, using the Nextera XT DNA library prep kit (Illumina). The libraries were sequenced with 150-bp paired-end read on an Illumina Hiseq. Adapter sequences were trimmed off from the Raw data (raw reads) of fastq format by flexbar (ver. 3.4.0). The resulting reads were aligned to the Ensembl mouse reference genome (GRCm38 ver. 92) by hisat (ver. 2.1.0) with the default parameters. The number of reads assigned to genes was calculated using featureCounts (ver. 1.6.4). To normalize for the differences in sequencing depth across samples, protein coding gene counts were rescaled to counts per million (CPM) by Trimmed Mean of M-values normalization (TMM) from edgeR. Low-expression genes with an average fewer than 10 CPM in all subnuclei were excluded from analysis. Principal component analysis (PCA) was performed using the prcomp function in R software (ver. 4.5.1) with option scale = TRUE. Differential gene expression analysis was performed using the R/Bioconductor package edgeR (ver. 3.32.0). Differentially expressed genes (DEGs) between VEH and CORT administered mice were defined as p-value < 0.01 and |log₂ FC| > 1.

PPI networks for sets of proteins encoded by all DEGs (including the up and downregulated DEGs) of each subnucleus were constructed using the Cytoscape stringApp (ver. 2.0.2) (Doncheva et al., 2019) with the following settings: network type = full STRING network; confidence score cutoff = 0.4. DEGs (nodes) without any interaction partners within the network are omitted from the visualization. To extract densely interconnected regions in the networks, we used the MCODE plugin (ver. 2.0.3) (Bader and Hogue, 2003) to find clusters with the following settings: include loops = true; degree cutoff = 2; haircut = false; fluff = false; node score cutoff = 0.2, max. Depth = 5, k-core = 2.

The GO enrichment analyses for sets of up or downregulated DEGs of each subnucleus were carried out and functionally grouped GO term networks were constructed using the ClueGO (version 2.5.10)/CluePedia (ver. 1.5.10) plugins (Bindea et al., 2009) in the Cytoscape (ver. 3.10.1) software (Shannon et al., 2003) with the following settings: ontology source = GO_BiologicalProcess-EBI-UniProt-GOA-ACAP-ARAP_25.05.2022, GO_CellularComponent-EBI-UniProt-GOA-ACAP-ARAP_25.05.2022, and GO_Molecularunction-EBI-UniProt-GOA-ACAP-ARAP_25.05.2022; evidence codes = All_Experimental; GO term fusion = true; statistical test used = Enrichment (Right-sided hypergeometric test); p-value cutoff = 0.05; GO tree interval level = 4 to 8; minimum number of genes for GO term selection = 3; minimum % of genes for GO term selection = 0; network connectivity Kappa score cutoff = 0.4; GO term grouping = true.

All values were expressed as mean ± s.e.m. Two-way RM analysis of variance followed by Sidak’s multiple comparison tests or unpaired one-tailed t-tests were performed using the GraphPad Prism software (ver. 10). One-tailed tests were used for behavioral comparisons because we hypothesized that chronic CORT administration should induce depression-like behavioral changes in a specific, predicted direction. p-values smaller than 0.05 were considered statistically significant.

To deepen the understanding of the pathophysiology of MDD, various rodent models have been developed, which recapitulate the psychological and/or physical stress exposure or reflect the physiological changes observed in patients with depression (Wang et al., 2017; Planchez et al., 2019). Among these, chronic CORT administration model mimics the elevated blood glucocorticoid concentration resulting from HPA axis activation and has been reported to display depressive symptoms in both rats and mice (Gregus et al., 2005; Johnson et al., 2006; Zhao et al., 2008). We first confirmed that a single dose of CORT significantly elevated plasma CORT levels under our experimental conditions (Supplementary Figure 1). Subsequently, as shown in Figure 1A, chronic CORT administration was initiated at 9–10 weeks of age and continued daily for 24 days, followed by an evaluation of depression-like symptoms. During the 24-day administration period, mice treated with vehicle (VEH) displayed a gradual increase in body weight, whereas CORT-treated mice exhibited suppressed weight gain (Figure 1B). In behavioral experiments, CORT-treated mice exhibited a decrease in sucrose preference in the SPT, indicative of anhedonia, a reduction in spontaneous locomotor activity in the OFT, suggesting reduced mobility and lethargy, and an increase in immobility time during the FST, indicative of despair-like behavior (Figures 1C,D,F). On the other hand, no significant differences were observed in anxiety-related behavior in the EPM test (Figure 1E). Taken together, these physical and behavioral alterations demonstrate that our chronic CORT-administered mouse model exhibit the face validity as a model of MDD.

The amygdala is a central brain region that plays a critical role in the emotional processing such as fear and anxiety in response to external stimuli. It is located upstream of the paraventricular nucleus of the hypothalamus and is involved in the activation of the HPA axis, thus playing a significant role in the stress response (Feldman and Weidenfeld, 1998; Herman et al., 2003; Sah et al., 2003; Roozendaal et al., 2009; Janak and Tye, 2015). Glucocorticoids released by the activation of the HPA axis act on both mineralocorticoid and glucocorticoid receptors in the amygdala, providing a concentration-dependent feedback mechanism that modulates amygdala activity (Herman et al., 2005). Consistently, recent studies have reported abnormalities in the activity and connectivity of the amygdala in patients with depression, strongly suggesting a relevance between depression symptoms and dysregulation of amygdala functions (Drevets, 2001; Ressler and Mayberg, 2007; Price and Duman, 2019). However, the amygdala is not a single functional unit. When the central extended amygdala, which encompasses the CeA and the lateral part of the BNST (BNSTL), is included, it consists of numerous subnuclei that are further interconnected through intra- and inter-subnuclear circuits (Fox et al., 2015; Lebow and Chen, 2016). The molecular mechanisms underlying depression within these amygdala subregions remain poorly understood due to this anatomical and functional complexity.

To unravel the molecular responses occurring in the amygdala with such complex subnuclear organization, we previously established a microdissection punching system with a punching needle (inner diameter of 110 μm) under a fluorescence microscope, enabling subnucleus-targeted tissue collection and RNA extraction from small tissue samples. We then conducted RNA-seq analysis and successfully captured the transcriptome profiles of each subnucleus (Yoda et al., 2017; Yamazaki et al., 2020; Ueda et al., 2021). Using the same approach, we analyzed gene expression profiles in five amygdala-associated subnuclei—BLA, CeL, CeM, BNSTov, and BNSTfu—following chronic CORT administration, using three biological replicates per group (VEH: n = 3, CORT: n = 3) (Figures 2A,B, and Supplementary Figure 2A). These subnuclei were selected based on their known involvement in emotional processing and stress-related circuits. For precise subnuclei-targeted tissue collection in this method, we utilized Prkcd-cre; Ai14 mice, in which PKCδ+ neurons were labeled with tdTomato fluorescence. This allowed us to identify the CeL and BNSTov, where PKCδ+ neurons are abundant, and the CeM and BNSTfu, which receive abundant axonal projections from PKCδ+ neurons (Dong et al., 2001b; Ueda et al., 2021). Similar to wild-type mice, Prkcd-cre; Ai14 mice also exhibited suppressed body weight gain and depression-like behaviors following chronic CORT administration (Supplementary Figures 2B–D). Since tissue collection was performed 6 days after the final CORT administration, when plasma CORT concentrations had decreased to levels comparable to or lower than those in the VEH group (Figure 1G and Supplementary Figure 2E), it is likely that the direct effects of CORT on gene expression profiles were minimized at the time of sampling. This delayed collection was intentional, as our study aimed to investigate transcriptional changes accompanying established depression-like phenotypes induced by chronic stress exposure, rather than to assess transient pharmacological responses to CORT.

Principal component analysis (PCA) of the RNA-seq data revealed that samples from each amygdala-related subnucleus formed separate clusters, indicating both the high precision of our subnucleus-targeted tissue collection and the robustness of the resulting molecular profiling (Supplementary Figure 3A). Moreover, consistent with our previous findings (Ueda et al., 2021), PCA revealed overall molecular similarities in gene expression profiles between CeL and BNSTov, as well as between CeM and BNSTfu. Next, from the results of RNA-seq, we identified differentially expressed genes (DEGs) between VEH and CORT administered mice (defined as p-value < 0.01 and |log₂ FC| > 1) in each of the five subnuclei. As a result, 154 upregulated and 123 downregulated DEGs in the BLA, 34 upregulated and 80 downregulated DEGs in the CeL, 54 upregulated and 65 downregulated DEGs in the CeM, 17 upregulated and 66 downregulated DEGs in the BNSTov, and 77 upregulated and 51 downregulated DEGs in the BNSTfu were extracted, respectively (Figure 2C, Supplementary Figures 3B–F, and Supplementary Table 1). Further examination of overlapping DEGs across subnuclei revealed minimal commonalities. Only four genes were shared among upregulated DEGs: one between the CeL and BNSTov (Egr2), two between the CeM and BNSTov (Arsk and Nr4a1), and one between the CeM and BNSTfu (Kcnf1). Among downregulated DEGs only two genes were shared: one between the BLA and CeL (F3), and another between the CeL and CeM (Ttn) (Figure 2C). These findings suggest that chronic CORT administration induces specific molecular responses in each amygdala subnucleus.

Given our previous report demonstrating the high similarity in the basal gene expression profiles between the CeL and BNSTov, but distinct gene expression responses following the stressful condition of fear conditioning (Ueda et al., 2021), we hypothesized that these regions might exhibit differential transcriptional responses to external stimuli. We visualized the magnitude and direction of gene expression changes induced by chronic CORT administration using scatter plots. Interestingly, most BNSTov-specific DEGs (blue) clustered near the line of symmetry (y = x) in the CeL plot, indicating that these genes were not differentially expressed in the CeL under CORT treatment. Similarly, most CeL-specific DEGs (red) also showed little variation in the BNSTov plot. Notably, while only one gene (Egr2) showed concordant upregulation in both regions (green), we identified four genes (Plp1, Mal, Opalin, and Tha1) that exhibited opposite regulation between the CeL and BNSTov (yellow).

Taken together, these findings underscore the complex and region-specific transcriptional adaptations in response to chronic CORT exposure in the amygdala-related subnuclei. In order to gain deeper insights into distinct biological alterations in each subnucleus, we performed protein–protein interaction (PPI) network analysis and gene ontology (GO) enrichment analysis on the identified DEGs in each amygdala-related subnucleus. The detailed results for each region are presented in the following subsections.

The BLA is a central hub for integrating affective, sensory, and cognitive information by possessing reciprocal connections with the prefrontal cortex, sensory cortices, and hippocampus. Through these connections, the BLA contributes to the emotional learning and stress regulation (Zhang et al., 2021). It also serves as a major upstream nucleus to the central extended amygdala including the CeA and BNSTL, a key output region of the amygdala involved in emotional expression. Protein–protein interaction (PPI) network analysis was performed on the DEGs extracted in the BLA using the Cytoscape stringApp (Doncheva et al., 2019), and clustering analysis was conducted with the MCODE plugin (Bader and Hogue, 2003). This analysis revealed eight distinct clusters within the network (Supplementary Figure 4A and Supplementary Table 2). The largest cluster (29 DEGs, nodes outlined in green) consisted mainly of ribosomal proteins, all of which were upregulated. The second-largest cluster (13 DEGs, nodes outlined in magenta) comprised genes involved in the oxidative phosphorylation pathway, with most showing increased expression. The third-largest cluster (13 DEGs, nodes outlined in orange) contained genes associated with synaptic function and the Ras/MAPK signaling pathway. Notably, among the synapse-related genes, those involved in inhibitory transmission, such as Gad1, Gad2, and Slc6a1, were downregulated, while excitatory synapse-associated genes, including Grin1, Nrxn2, and Cacna1a, were upregulated, indicating altered excitatory–inhibitory synaptic balance. Subsequently, GO enrichment analysis was conducted separately for the upregulated and downregulated DEGs using the ClueGO plugin in Cytoscape (Bindea et al., 2009). Consistent with the PPI findings, the upregulated DEGs were significantly enriched for GO terms related to ribosomal function and synaptic transmission (Figures 3A,B, Supplementary Figure 4B, and Supplementary Table 3). In contrast, the downregulated DEGs were significantly associated with GO terms linked to ion transporter activity and axonal regulation (Figures 3C,D, Supplementary Figure 4C, and Supplementary Table 3). These findings suggest that chronic CORT administration induces transcriptional changes in the BLA, including upregulation of ribosomal biogenesis, mitochondrial energy production, and synaptic remodeling, potentially shifting the excitatory–inhibitory balance in emotional processing circuits.

Subsequent analysis was conducted on the DEGs extracted in the CeL. The CeL is a subdivision of the CeA and serves as a major target of input from the BLA as well as from external and internal sensory pathways (Gilpin et al., 2015). It plays a critical role in the expression of emotional and stress-related responses by projecting either directly or via the CeM to brainstem and hypothalamic regions (Moscarello and Penzo, 2022). PPI network analysis revealed a cluster (five DEGs, nodes outlined in green) comprising genes associated with myelination, such as Opalin, Mal, and Plp1, all of which were upregulated (Figure 4A and Supplementary Table 2). Consistent with this, GO analysis of the upregulated DEGs identified terms related to myelination and glial cell differentiation (Figures 4B,C, and Supplementary Table 3). In contrast, GO analysis of the downregulated DEGs revealed terms related to regulation of cell fate specification (Figures 4D,E, and Supplementary Table 3). Notably, Wnt11, one of the DEGs annotated with this GO term, is a component of the Wnt/β-catenin signaling pathway, which is known to suppress oligodendrocyte differentiation (Shimizu et al., 2005; Ye et al., 2009). Taken together, these findings suggest that chronic CORT administration promotes oligodendrocyte differentiation and myelin formation in the CeL.

The CeM is another subdivision of the CeA, located medially to the CeL, and is the main output subnucleus of the amygdala, regulating behavioral responses to stress and fear (Gilpin et al., 2015). Among the clusters identified through PPI analysis for DEGs in the CeM, the second-largest cluster (five DEGs, nodes outlined in magenta) consisted of peptide hormones and their receptors, including Npy5r, Sst, Pdyn, Penk, and Cartpt, all of which were upregulated (Figure 5A and Supplementary Table 2). Consistent with this finding, GO analysis (Figure 5B-E) about the upregulated DEGs revealed terms associated with response to peptide hormone and G protein–coupled receptor signaling pathway (Figures 5B,C, and Supplementary Table 3). Because the CeM contains a diverse population of peptide-producing neurons, which communicate with downstream targets through peptidergic transmission (Wang et al., 2023), these results suggest that chronic stress may enhance peptidergic signaling, potentially modulating CeM output to hypothalamic and brainstem effectors involved in emotional and neuroendocrine control.

The BNSTL is anatomically continuous with the CeA, and is considered a core component of the central extended amygdala. Through dense reciprocal neural connections, the BNSTL and CeA are thought to function in a cooperative and complementary manner (Dong et al., 2001b; Fox et al., 2015). The BNSTov is a subdivision of the BNSTL and is known to share similar cellular composition and molecular properties with the CeL (Fox et al., 2015; Ueda et al., 2021). PPI analysis of DEGs extracted from the BNSTov revealed that the largest cluster (nine DEGs, nodes outlined in green) consisted of various well-known myelin-related genes, including Opalin, Mag, Cldn11, Mog, Mal, Gjc2, Mobp, Mbp, and Plp1, all of which were downregulated (Figure 6A and Supplementary Table 2). As shown in Figures 2D, 4A, several of these genes were also identified as upregulated DEGs in the CeL, indicating that these genes are regulated in the opposite direction in the two subnuclei. The second-largest cluster (three DEGs, nodes outlined in magenta) consisted of Egr4, JunB, and Nr4a1, all classified as immediate early genes, and all of which were upregulated. Consistent with these results, GO enrichment analysis of the upregulated DEGs highlighted terms associated with transcriptional regulation, while the downregulated DEGs were enriched for terms related to myelination and oligodendrocyte differentiation (Figures 6B–E and Supplementary Table 3). These findings suggest that the BNSTov and CeL may undergo region-specific and opposing transcriptional regulation in response to chronic stress, particularly in pathways related to oligodendrocyte differentiation and myelination.

The BNSTfu is also a subdivision of the BNSTL and receives dense projections from the BNSTov as well as from the CeA (Dong et al., 2001b). We have previously reported that the BNSTfu shares molecular properties similar to those of the CeM (Ueda et al., 2021). PPI analysis of DEGs identified in the BNSTfu revealed four clusters. The largest cluster (six DEGs, nodes outlined in green) included Anxa2 and Vim, known markers of reactive astrocytes, as well as Ptprb, Mmrn2, and Eng, which are markers of vascular endothelial cells. The fourth-largest cluster (three DEGs, nodes outlined in purple) consisted of modifier and regulatory subunits of voltage-gated potassium channels, including Kcng1, Kcnf1, and Kcnab1 (Figure 7A and Supplementary Table 2). GO analysis of the upregulated DEGs revealed significant enrichment of GO terms related to angiogenesis, interleukin-6 (IL-6) signaling, ion transporters, and synaptic vesicle function (Figures 7B–E and Supplementary Table 3). These findings suggest that, in the BNSTfu, chronic CORT administration promotes differentiation into reactive astrocytes through IL-6–mediated mechanisms, and further imply that changes in vascular structure and synaptic properties may also be occurring.