Introduction

Stress-related psychiatric disorders are the most common mental health conditions, with up to 33% of the population experiencing anxiety over the course of their lifetimes [1]. In humans, chronic psychosocial stressors, including low socioeconomic status, caregiver stress, and low perceived social status, increase the risk of developing anxiety and the frequency of anxious symptoms [2,3,4,5]. Moreover, inflammation after stress is associated with these behavioral deficits [6]. Prolonged stress exposure promotes inflammation evident in increased circulating inflammatory cytokines (interleukin(IL)-6, IL-1β), pro-inflammatory transcriptome of blood monocytes, glucocorticoid resistance, and hypothalamic–pituitary–adrenal (HPA) axis dysfunction [7,8,9,10,11]. These behavioral and immunological alterations with chronic stress are recapitulated in mice exposed to Repeated Social Defeat (RSD). RSD causes activation of neurons, endothelia, microglia; increases production of pro-inflammatory and glucocorticoid-insensitive monocytes; increases HPA axis activation and accumulation of monocytes in the brain; and induces anxiety-like behaviors [12,13,14,15].

We previously reported region-specific activation of microglia and neurons after RSD in the prefrontal cortex, hippocampus, hypothalamus, amygdala, lateral septum, and bed nucleus of the stria terminalis [15]. These are regions associated with fear, threat appraisal, and behavior changes [16, 17]. Additionally, RSD induced a reactive brain endothelium with increased expression of cell adhesion molecules (ICAM-1, VCAM-1), selectins (E-selectin, P-selectin), and IL-1 Receptor (IL-1R1) [17, 18]. The reactive endothelium responded to inflammatory signals (i.e., IL-1β) from peripheral monocytes through IL-1R1 and propagated these signals to cells of the CNS [19, 20]. The interaction between inflammatory monocytes (IL-1β) and brain endothelia (IL-1R1) was critical for development of anxiety after RSD [17, 19]. Nonetheless, the pathways induced in immune-activated brain endothelia that regulate the physiological and behavioral responses to social stress are unclear.

One potential mechanism is that monocytic IL-1β induces cyclooxygenase-2 (COX-2) in brain endothelia. COX-2 synthesizes prostaglandin (PG)H2 from membrane arachidonic acid that is converted to other prostaglandins (PGD2, PGE2, PGF2), prostacyclin (PGI2), or thromboxane A2 [21]. Previous reports indicate COX-2 was induced in brain endothelia following peripheral challenge with LPS or IL-1β and the resulting prostaglandins mediated sickness responses and HPA axis activation [22,23,24,25]. Stressors including crowding, isolation, chronic unpredictable stress, and social defeat, increased COX-2 and prostaglandins across several brain regions (prefrontal cortex, hippocampus, hypothalamus, subcortical regions) [26,27,28,29]. COX-2 inhibition during restraint and chronic unpredictable stress also prevented anxiety- and depressive-like behaviors [27, 30]. Thus, we sought to determine the RNA profile of activated endothelia and delineate the pathways by which these endothelia communicate within the brain to influence physiological and behavioral responses to social stress.

Materials and methods

Mice

C57BL/6 and CD-1 mice were obtained from Charles River Laboratories. Male C57BL/6 mice were housed in cohorts of three, while CD-1 aggressors were individually housed. Transgenic lines (Tie2-Rpl22-HA, Cdh5-Rpl22-HA, Il1r1loxP/loxP, and Il1r1r/r) were bred in-house. All procedures were in accordance with NIH Guidelines and were approved by OSU-IACUC.

Repeated social defeat

Mice were subjected to RSD as described in Supplementary Materials [31, 32]. CD-1 aggressor mice were placed into the home cages of experimental mice for 2 h (16:00–18:00) per night for six consecutive nights. During each cycle, submissive behaviors (upright posture, fleeing, crouching) were observed to ensure defeat.

RiboTag HA-immunoprecipitation

Immunoprecipitation of polysome-bound mRNA from brain endothelia was performed in Tie2-Rpl22-HA and Cdh5-Rpl22-HA mice as described [33, 34] and outlined in Supplementary Materials. Tissue was homogenized and HA-ribosomes were immunoprecipitated using Protein A/G magnetic beads with anti-HA. RNA was eluted with high salt buffer and isolated using the RNeasy Micro-Kit.

RNA sequencing

Quality control was performed using PicoAssay and samples with RIN > 8.0 were processed for sequencing. cDNA libraries were generated with the Ovation SoLo RNA-Seq System and sequenced at 40 M single-end bp100 reads. Data was aligned to the mouse mm9 reference genome and differential expression was analyzed using DESeq2 [35,36,37]. Pathway analyses were performed using Ingenuity Pathway Analysis, Cytoscape EnrichmentMap, and PANTHER Overrepresentation analysis.

Gene expression analysis

Total RNA was extracted with Tri-Reagent and reverse transcribed using a High-Capacity cDNA Reverse Transcription Kit. Quantitative real-time (q)-PCR was performed using the TaqMan Gene Expression Assay.

RNAscope

RNAscope labeling of Ptgs2 and Cdh5 was performed using the RNAscope Multiplex Fluorescent Reagent Kit v2 (ACDBio). Brains were fixed with 4% formaldehyde, saturated in sucrose, and sectioned (10 μm). Sections were heated, post-fixed, dehydrated, and subjected to antigen retrieval. Probes (Ptgs2-C1 and Cdh5-C2) were hybridized. Probe amplification was performed according to manufacturer’s instructions. Probe signals were developed using Opal dyes (Ptgs2-Opal 690; Cdh5-Opal 520) and counterstained with DAPI. Images were collected using a Leica SP8 confocal microscope at 40× magnification and were quantified using ImageJ.

Celecoxib treatment

The dosage of Celecoxib (CCB) was selected based on previous reports [38,39,40]. Mice received 10 mg/kg CCB or vehicle i.p. 30 min before each cycle of RSD to ensure CCB concentration peaked during RSD [41].

PGE2 ELISA

Mice were euthanized with CO2 asphyxiation immediately after the final cycle of RSD. PGE2 concentration was determined in clarified brain lysates by ELISA (Cayman). ELISA values were normalized to total protein concentration measured via Bradford assay.

Immunohistochemistry for ΔFosB

Mice were transcardially perfused with PBS and 4% formaldehyde. Brains were post-fixed, saturated in sucrose, sectioned (30 μm), and labeled for ΔFosB. Slides were imaged at 20× magnification and the number of ΔFosB+ cells was determined using ImageJ.

Corticosterone ELISA

Mice were euthanized with CO2 asphyxiation immediately after the final cycle of RSD. Plasma corticosterone was assessed using the corticosterone ELISA kit (Enzo) according to manufacturer’s instructions.

Flow cytometry

Labeling of cell surface antigens in the bone marrow, blood, and brain was performed as described [42]. In brief, Fc receptors were blocked, cells were washed, and cells incubated with select antibodies. Antigen expression was determined using a FACSCalibur cytometer with Cytek nine-color upgrade. Data were analyzed using FlowJo software and positive labeling for each antibody was determined based on isotype-labeled controls.

Pain sensitivity

Tactile mechanical sensitivity was determined using von Frey filaments 14 h after RSD. Mice were placed on a mesh platform in a compartment (8 × 12 × 5.5 cm) and habituated. Filaments were applied to the mid-plantar area of the hind paw and measurements were determined by the up-down method [43]. Withdrawal threshold was determined by averaging the results from the left and right hind paws. Logarithmic transformation of the 50% withdrawal threshold was performed before statistical analysis [44].

Anxiety-like behavior

Anxiety-like behavior was determined using the open field test as described [17, 45] 14 h after the 6th cycle of RSD. In brief, mice were placed in the corner of the test apparatus (40 × 40 × 25 cm Plexiglas box) and activity was recorded for 5 min. Behavior was recorded and analyzed using Fusion software (Omnitech).

Social avoidance

Social avoidance was performed as described 14 h after the 6th cycle of RSD [20, 46]. First, an empty 18 × 8 × 8 cm metal wire enclosure was placed on one side of the open field arena and experimental mice were permitted to explore the enclosure (2.5 min). Then, an unfamiliar CD-1 mouse was placed inside the wire enclosure. Experimental mice were placed into the arena and activity was recorded (2.5 min). The duration of social investigation of the caged mouse was determined using Fusion.

Statistical analysis

GraphPad Prism software was used for analysis of variance (ANOVA) of histological and behavioral data. One-way or two-way ANOVA were used to determine main effects and interactions between factors. Tukey’s HSD was used for post-hoc analysis when main effects and/or interactions were determined.

Results

RSD induced a unique RNA profile in brain endothelia consistent with increased cell adhesion, leukocyte migration, and prostaglandin synthesis

IL-1R1 signaling in endothelia (from monocytes) is critical for prolonged anxiety after RSD [17,18,19]. This study investigated if immune-activated endothelia regulate physiological and behavioral responses to stress. Here, endothelial-specific RiboTag mice were generated by crossing Rpl22-StoploxP/loxP-Hemagglutinin (HA) mice with endothelial-specific Tie2-Cre mice (Fig. 1A). HA expression was confirmed in CD31+ endothelia of Cre+ mice (Fig. 1B). Next, Tie2-Rpl22-HA mice were exposed to RSD and endothelial-specific mRNA from the brain was sequenced (Fig. 1C). RSD caused differential expression of 355 genes (DEGs; 176 decreased & 179 increased) in brain endothelia (Fig. 1D, padj < 0.05). Notably, RSD increased Icam1 (intercellular adhesion molecule 1), Lrg1 (leucine rich alpha-2-glycoprotein 1), Lcn2 (lipocalin 2), Ackr1 (atypical chemokine receptor 1), and Ptgs2 (prostaglandin-endoperoxide synthase 2) in endothelia (Fig. 1E–I, p < 0.02, for each).

Fig. 1: RSD promoted a unique RNA profile in brain endothelia consistent with increased cell adhesion, leukocyte migration, and prostaglandin synthesis.
figure 1

A Diagram shows generation of endothelial RiboTag (Tie2-Rpl22-HA) mice. B Representative images of HA (green) and CD31 (red) labeling in the hippocampus of Cre- and Cre+ RiboTag mice. C Experimental design with male Cre+ Tie2-Rpl22-HA mice subjected to Repeated Social Defeat (RSD) or left undisturbed as controls (CON). 14 h following the final cycle of social defeat, the brain was collected, polysome-bound endothelial mRNA was immunoprecipitated, and RNA was sequenced (n = 3; litters = 3). D Volcano plot of differentially expressed genes in brain endothelia between CON and RSD. The gray triangles represent genes that were significantly different between CON and RSD. The blue triangles represent genes that were significantly decreased with |log2FoldChange| >1. The red triangles represent genes that were significantly increased with |log2FoldChange| > 1. From the same RNAseq data set, normalized endothelial mRNA counts of E Icam1, F Lrg1, G Lcn2, H Ackr1, I and Ptgs2 were assessed. Values represent the mean ± SEM, and individual data points are provided. Values with (*) are different (p < 0.05) from controls. J Ingenuity Pathway Analysis of Canonical Pathways inhibited or activated by RSD in endothelia (Z-score). K Ingenuity Pathway Analysis of Upstream Regulators inhibited or activated (Z-score) in endothelia by RSD. L Gene Ontology Biological Processes determined by Cytoscape EnrichmentMap (CEM) that were increased in brain endothelia after RSD. Circle size represents the number of genes within the category and the color of the circle represents the relative strength of connections. (M) Heat map of differentially expressed genes organized by Gene Ontology Biological Processes determined by PANTHER Overrepresentation Analysis. Individuals within stress groups are represented in the heatmap.

Full size image

Ingenuity Pathway Analysis was used to assess Canonical Pathways and Upstream Regulators in endothelia after RSD. Canonical Pathways decreased by RSD included Integrin-Linked Kinase, Platelet-Derived Growth Factor, and Integrin signaling. Pathways increased by RSD included Endothelin-1, Oxidative Phosphorylation, and Chemokine Signaling (Fig. 1J; p < 0.05, for each). For predicted upstream regulators, α-catenin was decreased by RSD and myriad inflammatory regulators (IKBKB, IKBKG, CCR2, CHUK) and Insulin-Like Growth Factor 1 Receptor were increased (Fig. 1K; p < 0.05, for each). Next, DEGs increased by stress were input into Cytoscape EnrichmentMap to visualize Gene Ontology (GO) Biological Processes. Pathways associated with adhesion, vascular remodeling, and lipid biosynthetic processes (Fig. 1L) were increased in endothelia following RSD. DEGs were also used in PANTHER overrepresentation analysis and highly enriched GO Biological Processes were represented in a heat map (Fig. 1M). RSD influenced biological processes in endothelia associated with regulation of cell migration (Ptgs2, Lcn2), cellular respiration (Atp5k, Cycs), cell–cell adhesion (Cd200, Nrcam), leukocyte migration (Icam1, Cxcl12), and insulin stimulus (Gpld1, Igf2) (p < 0.05, for each). Overall, activation of brain endothelia after RSD was associated with increased leukocyte adhesion and migration, cellular respiration, and prostaglandin synthesis.

The Prefrontal Cortex (PFC) is a key region for stress-induced endothelial IL-1R1 expression and recruitment of peripheral monocytes [17]. In parallel experiments, endothelial-specific (inducible) RiboTag mice were generated by crossing Rpl22-StoploxP/loxP-HA mice and Cdh5-CreERT2 mice to prevent nonspecific HA expression during development [47,48,49] (Fig. S1A). HA expression was confirmed in CD31+ endothelia of Cre+ mice after tamoxifen injections (Fig. S1B). Next, Cdh5-Rpl22-HA mice were exposed to RSD and endothelial-specific mRNA from the PFC was sequenced (Fig. S1C). In the PFC, there was differential expression of 86 genes with RSD (9 increased and 77 decreased, Fig. S1D). Ackr1, Lrg1, Lcn2, and Il1r1 (Interleukin-1 Receptor Type 1) were increased by RSD (Fig. S1E–H, padj < 0.05, for each). Mfsd2a and Cldn5 (Claudin-5) were decreased by RSD (Fig. S1I, J, padj < 0.05, for each). PANTHER Overrepresentation Analysis showed that the DEGs after RSD were associated with Response to Cytokines, Cell Migration, and Regulation of Endothelial Proliferation pathways (Fig. S1K, p < 0.05). Thus, RSD elicited an RNA profile in PFC-endothelia consistent with increased cytokine signaling, cell migration, and IL-1R1 signaling.

Immune-induced endothelial activation after RSD was dependent on IL-1R1

Endothelial-selective profiling after RSD showed a unique profile associated with immune activation, cell adhesion, and cell migration. This profile was regulated by activators of the transcription factor NF-κB (Figs. 1 and S1), which is key target of IL-1R1 signaling [50]. Thus, the influence of IL-1R1 signaling on endothelial genes induced by RSD in the brain and PFC was investigated. Here, Il1r1WT and Il1r1KO mice (Fig. 2A) were exposed to RSD (Fig. 2B) and the PFC, HPC, and HYP were collected. Expression of several endothelial genes associated with stress in RiboTag mice (Cldn5, Lrg1, Icam1, Lcn2, Ptgs2) were determined in Il1r1WT and Il1r1KO mice. Cldn5 was not influenced by RSD or IL-1R1 in the PFC (Fig. 2C), but was increased by RSD in the HPC (Fig. 2D; F(1,16) = 5.687, p < 0.03). Cldn5 was increased by RSD in the HYP and this was IL-1R1-dependent (interaction, F(1,18) = 7.005, p < 0.02). Lrg1 expression was increased by RSD in the PFC, HPC, and HYP (Fig. 2F–H; F(1,17) = 21.0, p < 0.001, for each) and was IL-1R1-dependent only in the HPC (interaction, F(1,16) = 9.0, p < 0.01). Icam1 and Lcn2 in the PFC, HPC, and HYP were increased by RSD (Fig. 2I–N; F(1,18) = 12.7, p < 0.003, for each) and were IL-1R1-dependent (interaction, F(1,18) = 4.2, p < 0.06, for each). Finally, Ptgs2 was increased by RSD in the PFC and HYP (Fig. 2O–Q, F(1,11) = 6.987, p < 0.03). This was IL-1R1-dependent only in the HYP (interaction, F(1,13) = 47.93, p < 0.001). Overall, most genes induced by RSD were region- and IL-1R1-dependent.

Fig. 2: Immune-induced endothelial activation after RSD was dependent on IL-1R1.
figure 2

A Diagram of “wild-type IL-1R1” (IL-1R1loxP/loxP) and IL-1R1KO (IL-1R1r/r) mice. Phenotypic WT controls have loxP sites inserted into the Il1r1 gene but express Il1r1 in all cells. IL-1R1KO mice have a disruption sequence in exon 9 of the Il1r1 gene leading to functional knockout of Il1r1 in all cells. B IL-1R1WT and IL-1R1KO mice were subjected to Repeated Social Defeat (RSD) or left undisturbed as controls (CON) (litters = 2 per genotype). 14 h following the final cycle of social defeat, the brain was microdissected and RNA was collected from the prefrontal cortex (PFC), hippocampus (HPC), and hypothalamus (HYP). RNA levels of C–E Cldn5, F–H Lrg1, I–K Icam1, L–N Lcn2, O–Q and Ptgs2 were determined from the prefrontal cortex, hippocampus, and hypothalamus (n = 3). Data are represented as fold change from the WT-CON group. Bars represent the mean ± SEM, and individual data points are provided. Letters (a,b,c) represent post-hoc analysis of multiple comparisons (Tukey’s HSD) and denote statistically different groups (p < 0.05). Italicized letters (b) denote groups that tended to be different from WT-CON (p < 0.10).

Full size image

PGE2 increased in the brain after RSD and Ptgs2 was increased in the PFC and PVN

Endothelial-selective profiling after RSD indicated that prostaglandin synthesis was increased (Fig. 1). PGE2 is a key prostaglandin generated by cyclooxygenase-2 (COX-2 or Ptgs2) and is associated with fever, edema, leukocyte recruitment, inflammation, and pain [25, 51,52,53]. Here, C57BL/6 mice were subjected to RSD and PGE2 concentration was determined in whole brain homogenates immediately after 1, 3, and 6 cycles of RSD. PGE2 was increased in the brain after 6 cycles of RSD (Fig. 3B; F(3,17) = 7.703, p < 0.002).

Fig. 3: PGE2 increased in the brain after RSD and Ptgs2 was increased in the PFC and PVN.
figure 3

A Experimental design with male C57BL/6 mice subjected to Repeated Social Defeat (RSD) or left undisturbed as controls (CON). Immediately following the final cycle of social defeat, brain samples were collected for prostaglandin (PGE2) and RNA analyses. B PGE2 concentration (μg/mg total protein) in whole brain homogenates collected after RSD. Graph contains pooled data from three separate cohorts (1, 3, 6 cycles; n = 3, 3, 5 respectively). C Representative images of RNAscope labeling of Cdh5 RNA (red) and Ptgs2 RNA (green) in the prefrontal cortex (40×). Arrows and inset show Cd5h+/Ptgs2+ (yellow) cells in the PFC. D Percent area of Ptgs2 RNA labeling in the PFC (n = 4). E Representative images of RNAscope labeling of Cdh5 RNA (red) and Ptgs2 RNA (green) and F percent area of Ptgs2 RNA labeling in the DG (n = 4). G Representative images of RNAscope labeling of Cdh5 RNA (red) and Ptgs2 RNA (green) in the PVN. Arrows and inset show Cdh5+/Ptgs2+ (yellow) cells in the PVN. H Percent area of Ptgs2 RNA labeling in the PVN (n = 4). Bars represent the mean ± SEM, and individual data points are provided. Values with (*) are different (p < 0.05) from CON.

Full size image

Next, RNAscope was used to assess region-specific expression of Ptgs2. C57BL/6 mice were subjected to RSD and Ptgs2 and Cdh5 (endothelial marker [48, 49]) was determined in three stress-responsive brain regions (PFC, HPC, and HYP). In controls, Ptgs2 was expressed in Cdh5− cells of the PFC and dentate gyrus (DG). Ptgs2 was increased by RSD in the PFC (Fig. 3C, D; p < 0.02). Increased PFC Ptgs2 labeling was evident in both and Cdh5- and Cdh5+ cells (inset). In the DG, Ptgs2 was highly expressed (Cdh5-, non-endothelia) at baseline and was not influenced by RSD (Fig. 3E, F). In the Paraventricular Nucleus (PVN) of the HYP, baseline expression of Ptgs2 was minimal. RSD increased Ptgs2 expression in the PVN only in Cdh5+ endothelia (Fig. 3G, H, p < 0.05). Overall, stress increased prostaglandin production and enhanced endothelial expression of Ptgs2 in the PFC and HYP.

COX-2 inhibition with Celecoxib attenuated RSD-induced PGE2 signaling and HPA axis activation

Ptgs2 encodes COX-2, the enzyme that generates PGE2 in the brain. Thus, we investigated if inhibition of COX-2 with Celecoxib (CCB) blocked the physiological and behavioral responses to social stress. C57BL/6 mice were injected (i.p.) with CCB (10 mg/kg) 30 min prior to each cycle of RSD (Fig. 4A). RSD increased PGE2 concentration in the brain immediately after RSD (Fig. 4B; F(1,35) = 14.2, p < 0.001) that tended to be CCB-dependent (interaction, F(1,35) = 3.3, p = 0.08). Post-hoc analysis confirmed that PGE2 was highest in the RSD-Veh group (p < 0.002) compared to all groups (p < 0.02). Thus, brain PGE2 induction after RSD was attenuated by CCB.

Fig. 4: COX-2 inhibition with Celecoxib attenuated RSD-induced PGE2 signaling and HPA axis activation.
figure 4

A Experimental design with male C57BL/6 mice receiving injections of vehicle (0.25% carboxymethylcellulose) or 10 mg/kg celecoxib (CCB) 30 min prior to each cycle of social defeat. In the first experiment, the brain was collected for PGE2 and RNA analyses and plasma was collected for corticosterone immediately after the final cycle of RSD. B PGE2 concentration (μg/mg of total protein) in whole brain homogenates normalized to Con-Veh group collected immediately after RSD (n = 9). In a parallel study, the brain was microdissected and RNA was collected from the hypothalamus (HYP). RNA levels of C Ptgs2 and D Crh were determined in the HYP (n = 6). E Concentration of corticosterone was determined in the plasma after RSD (n = 6). F Representative images of ΔFosB labeling (20×) in the prefrontal cortex (PFC) and G number of positive ΔFosB cells in the PFC (n = 4). H Representative images of ΔFosB labeling and I number of positive ΔFosB cells in the DG (n = 4). J Representative images of ΔFosB labeling and K number of positive ΔFosB cells in the PVN (n = 3). Bars represent the mean ± SEM, and individual data points are provided. Letters (a,b) represent post-hoc analysis of multiple comparisons (Tukey’s HSD) and denote statistically different groups (p < 0.05).

Full size image

Ptgs2 was increased in PVN endothelia after stress (Fig. 3G). The PVN is critical in HPA axis activation by releasing corticotrophin releasing hormone (CRH) during stress. Ptgs2 and Crh levels were determined in the HYP. RSD increased Ptgs2 in the HYP (F(1,20) = 31.3, p < 0.001) which was unaffected by COX-2 inhibition (Fig. 4C). RSD also increased Crh in the HYP (F(1,20) = 8.171, p < 0.01) which was COX-2-dependent (Fig. 4D; interaction, F(1,20) = 10.07, p < 0.005). HPA axis activation results in release of glucocorticoids into circulation. Thus, corticosterone was determined in plasma after RSD. Plasma corticosterone was increased by RSD (F(1,19) = 8.4, p < 0.01) and tended to be COX-2-dependent (interaction, F(1,19) = 3.1, p = 0.1). Post-hoc analysis confirmed that RSD-Veh mice had the highest amount of plasma corticosterone compared to all other groups (Fig. 4E; p < 0.05).

Next, neuronal activation in stress-responsive brain regions was assessed. RSD increased the number of ΔFosB+ cells in the PFC (Fig. 4G, H; F(1,17) = 13.8, p < 0.003) and DG (Fig. 4I, J; F(1,16) = 17.2, p < 0.002) independent of COX-2. ΔFosB+ cells in the PVN were also increased by RSD and this induction was COX-2-dependent (Fig. 4K, L; interaction, F(1,15) = 9.9, p < 0.008). Overall, inhibition of COX-2 with CCB prevented stress-induced activation of PVN neurons which was associated with reduced Crh and plasma corticosterone.

COX-2 inhibition with Celecoxib attenuated RSD-induced anxiety

RSD increases microglial activation and recruitment of monocytes to the brain. This stress-dependent enhancement of neuroinflammation influences behavioral, pain, and cognitive processes [15, 17, 20, 46, 54]. In general, social avoidance after RSD was monocyte/IL-1β-independent [55], pain was microglia-dependent [54], and anxiety was monocyte/IL-1β/eIL-1R1-dependent [17, 19]. Thus, these interactions were investigated with COX-2 inhibition by CCB. Mice were injected with CCB prior to each cycle of RSD (Fig. 5A) and monocyte production, release, and accumulation were determined. RSD increased production of myeloid cells in the bone marrow including neutrophils (F(1,19) = 32.5, p < 0.001) and monocytes (F(1,19) = 16.3, p < 0.001, Fig. 5B, C). In parallel, RSD reduced B-cells and red blood cells (RBC) (Fig. 5B). All these bone marrow changes with RSD were COX-2-independent. RSD also increased inflammatory (Ly6Chi) monocytes in circulation (F(1,15) = 16.9, p < 0.002) and brain (F(1,14) = 11.84, p < 0.004). Monocyte accumulation in the blood and brain after RSD were both COX-2-independent. Overall, monocyte production, release, and brain accumulation after RSD did not require COX-2.

Fig. 5: COX-2 inhibition with Celecoxib attenuated RSD-induced anxiety.
figure 5

A Experimental design with male C57BL/6 mice receiving injections of vehicle (0.25% carboxymethylcellulose) or 10 mg/kg celecoxib (CCB) 30 min prior to each cycle of social defeat. In the first experiment, percentages of Ly6Chi monocytes were determined in the bone marrow, blood and brain (n = 6). B Percentage of B-cells (Ter119−/CD11b−/B220+), neutrophils (Ter119−/CD11b+/Ly6G+), and red blood cells (Ter119+) in the bone marrow. C Percentage of LyC6hi monocytes (Ter119-/CD11b+/Ly6Chi) in the bone marrow. D Percentage of CD11b+/Ly6Chi monocytes in the blood (n = 4) E Representative bivariate dot plots and gating strategy for inflammatory monocytes in the brain. F Percentage of inflammatory monocytes (CD11b+/CD45hi/Ly6Chi) in the brain (of total CD11b+ cells). Bars represent the mean ± SEM, and individual data points are provided. G In a separate study, mechanical allodynia was assessed (n = 6). H Body weights of mice assessed in Von Frey. I Normalized Von Frey withdrawal threshold after 0, 1, 6 cycles of RSD (logarithmic transformation). J In a separate study, social avoidance was assessed in a two-trial social interaction paradigm. Gray areas indicate interaction and corner zones. The duration within the interaction zone during K empty trial and L during social trial. Parallel to these experiments, anxiety-like behavior in the open field was assessed (n = 25). M Representative heat maps of duration in zones during open field test. N Distance traveled in the open field. O Total duration in center zone of open field. Bars represent the mean ± SEM. Letters (a,b,c) represent post-hoc analysis of multiple comparisons (Tukey’s HSD) and denote statistically different groups (p < 0.05). Italicized letters (b) denote groups that tended to be different from CON-Veh (p = 0.10).

Full size image

RSD increased mechanical allodynia that was dependent on inflammation [54]. Because PGE2 is a mediator of pain [52], we assessed if COX-2 inhibition influenced pain sensitivity after RSD (Fig. 5G). Baseline (0 cycle) withdrawal threshold was determined. Next, C57BL/6 mice were injected with CCB prior to each cycle of RSD and mechanical allodynia was determined 14 h after the 1st and 6th cycle. There were no differences in body weight between groups (Fig. 5H). There was an exposure-dependent increase in pain sensitivity after RSD (F(6,40) = 4.046, p < 0.003) that was independent of COX-2 (Fig. 5I).

Next, C57BL/6 mice were injected with CCB prior to each cycle of RSD and social avoidance and anxiety-like behavior were determined 14 h later (Fig. 5A). Social avoidance was determined using a two-trial paradigm (Fig. 5J). During the empty trial, all mice explored the empty cage and spent most time within the interaction zone (Fig. 5K). During the social trial with the CD-1 mouse, however, time spent in the interaction zone was reduced by RSD (F(1,107) = 15.8, p < 0.001, Fig. 5L), independent of COX-2 inhibition. In the open field, representative heat maps of duration are provided (Fig. 5M). There were no differences in total locomotion across groups (Fig. 5N). Time spent in the center of the open field tended to be decreased by RSD (F(1,85) = 2.971, p = 0.09) and was prevented by COX-2 inhibition (Fig. 5O; interaction, F(1,85) = 5.2, p < 0.03). Post-hoc analysis confirmed that the RSD-Veh group had the least amount of center time compared to all other groups (p < 0.05). Overall, of the monocyte, pain, and social behavior related-parameters examined here, only anxiety in the open field required the COX-2/PGE2 pathway.

Discussion

Repeated Social Defeat (RSD) increases the accumulation of inflammatory monocytes within the brain vasculature, which corresponds with increased IL-1R1-mediated activation of endothelia, neuronal activation, and anxiety-like behavior [17, 19, 20]. Here, we focused on understanding immune-activated brain endothelia in regulating key responses to RSD (Fig. S2). RNA profiling of endothelia after RSD revealed an active role of endothelia in leukocyte adhesion (Lrg1, Lcn2), leukocyte recruitment (Icam1), IL-1R1 signaling (Il1r1) and prostaglandin synthesis (Ptgs2). These data are consistent with previous reports using global brain assessments [17, 18, 56]. Pathway analysis revealed these gene changes were mediated by activators of NF-κB, a key transcription factor associated with IL-1R1-activation. Furthermore, endothelial-associated transcripts induced by stress including Icam1, Lrg1, Lcn2, and Ptgs2 were IL-1R1-dependent. Notably, a previous study with social defeat showed reduced blood-brain barrier (BBB) integrity with decreased expression of the tight junction protein Claudin-5 (Cldn5). Decreased BBB integrity in the Nucleus Accumbens was associated with increased IL-6 and social avoidance [57]. Here, Cldn5 expression was decreased in the PFC after RSD (endothelial RNA-sequencing). The results from qPCR studies of PFC, HYP, and HPC were less consistent, but these evaluations were not cell-specific. Based on the presence of inflammatory monocytes, the endothelial profile, and previous evidence for decreased integrity, RSD may elicit localized BBB permeability. For example, Lcn2 is associated with glial activation, neuronal death, and BBB leakage after LPS challenge and stroke induction [58,59,60,61] and was increased in endothelia after RSD. One limitation was that only male mice were used. Nonetheless, RSD in female mice increased monocyte trafficking, IL-1β production in the brain, and anxiety [42]. Therefore, we predict that RSD in female elicits similar IL-1R1-mediated changes in brain endothelia. In fact, a recent study using female mice showed that social defeat increased expression of inflammatory mediators (Il18, Ccl19, Ccr5, Hsp90b1) in brain endothelia [62]. Overall, there was a unique RNA profile in brain endothelia that developed after RSD consistent with increased cell adhesion, leukocyte migration, IL-1R1 signaling, and reduced BBB integrity.

Another relevant finding was that Ptgs2/COX-2 was induced in brain endothelia by RSD. Previous studies reported increased COX-2 expression in homogenates of the PFC and HPC with social isolation and chronic unpredictable stress [27, 28]. We extend these findings by showing vascular (Cdh5+) and non-vascular (Cdh5−, likely neuronal) induction of COX-2 after RSD. COX-2 induction was region-dependent and increased in the PFC (endothelia and neuronal) and PVN (endothelia only), but not the DG. Notably, neurons in the cortex and HPC express COX-2 constitutively with expression levels coupled to excitatory neuronal activity [63, 64]. Thus, increased neuronal activity in the PFC after RSD likely promoted COX-2 expression in these neurons [15, 17, 46]. Here, COX-2 induction in the HYP was dependent on IL-1R1. This is consistent with a previous report that (i.v.) IL-1β injection caused robust cerebrovascular expression of COX-2 [23]. Collectively, IL-1R1 promoted vascular COX-2 expression after RSD in the HYP and PFC while non-vascular COX-2 expression in the PFC is likely mediated by neuronal activity.

One novel finding was COX-2 inhibition attenuated stress-induced activation of the HPA axis. CCB prevented hypothalamic expression of Crh and increased corticosterone in circulation after RSD. These data are consistent with studies showing vascular COX-2 in the PVN mediates HPA axis activation through PGE2 signaling [65,66,67,68]. Neuronal activation (ΔFosB) after stress was prevented in the PVN by COX-2 inhibition. Moreover, COX-2 expression was localized to endothelia in the PVN after stress. This stress-associated neuronal activation in the PVN also indicates that COX-2 activation of the HPA axis is important to the response to social stressors.

CCB did not prevent production, release, trafficking, and accumulation of inflammatory (Ly6Chi) monocytes in the brain after stress. We previously reported that monocyte recruitment to the brain with RSD depended on microglia and IL-1R1 signaling [17, 19, 20]. Thus, the current findings indicate that COX-2 activity was downstream of monocyte-to-endothelial signaling via IL-1R1. Likewise, we previously reported that corticosterone influenced monocyte release from the bone marrow into circulation and subsequent recruitment to the brain [14]. Both pharmacologic inhibition of corticosterone synthesis and adrenalectomy before stress prevented monocyte release into circulation [14]. Here, COX-2 inhibition also reduced corticosterone levels in plasma after RSD, but monocyte accumulation in the brain was still evident. Explanations for these differences are related to level of inhibition and timing. In previous experiments, our interventions ablated corticosterone throughout RSD, while COX-2 inhibition with CCB attenuated corticosterone at one specific time point (immediately after the last cycle). The presence of monocytes was time-dependent and requires at least 3 repeated cycles of RSD [69]. In this study, PGE2 levels in the brain were increased after 6 cycles of RSD, but not 1 cycle. Collectively, inflammatory monocytes reinforced HPA activation and corticosterone release via a COX-2-dependent pathway.

Another key finding was RSD-induced anxiety-like behavior in the open field was prevented by CCB. These results indicate that prostaglandins were critical in promoting RSD-induced anxiety. These data are consistent with studies showing COX-2 inhibition modulated calcium currents in the amygdala and prevented anxiety-like behavior after restraint stress and chronic corticosterone administration [30, 70]. Furthermore, blocking PGE2 signaling with an EP1 receptor knockout prevented anxiety-like behavior after social defeat [71]. Moreover, recent clinical studies indicate that Celecoxib was an effective adjunct therapy for anxiety disorders in both men and women [72, 73]. Our data connect these findings to show that social stress-induced PGE2 synthesis via endothelial COX-2 modulated neuronal activity and HPA activation and was essential for anxiety, providing support for the effectiveness of CCB in humans with anxiety. These data are also consistent with previous data that anxiety-like behavior after stress was blocked by endothelial IL-1R1 knockdown [19], microglial inhibition/elimination, and in mice with caspase-1 deficient monocytes [17]. These interventions affected monocyte recruitment to the brain or their ability to communicate with endothelia through IL-1β/IL-1R1. The extension here is that CCB blocked COX-2 activity, presumably in brain endothelia, preventing endothelial interpretation of IL-1R1 activation and release of prostaglandins. Thus, anxiety was the only behavior tested here that was COX-2-dependent. For example, mechanical allodynia was increased by RSD, but was COX-2-independent. While COX-2 inhibitors are common pain relievers [74], the effects of CCB on RSD-induced anxiety were unrelated to pain. Next, CCB did not influence neuronal activation in the PFC and HPC after RSD, suggesting that threat appraisal and fear after social defeat were independent of COX-2. As such, RSD-induced social avoidance of an unfamiliar CD-1 was independent of CCB. This finding is in line with previous reports that social avoidance developed after 1 cycle of RSD, was independent of monocyte accumulation and communication with endothelia, and was COX-2-independent [17, 55, 71]. In conclusion, brain endothelia and COX-2 regulate neuroinflammation, neuroendocrine activity, and anxiety-like responses to RSD.