Environmental enrichment and physical exercise prevent stress-induced social avoidance and blood-brain barrier alterations via Fgf2

Abstract

Chronic stress promotes blood-brain barrier (BBB) integrity loss leading to passage of inflammatory mediators in mood-regulating brain areas and establishment of depressive behaviors. Conversely, neurovascular adaptations favoring stress resilience and preventive strategies to promote them are undetermined. We report that environmental enrichment dampens stress-induced loss of endothelial tight junction Claudin-5 (Cldn5) along with anxiety- and depression-like behaviors in male mice via an increase in fibroblast growth factor 2 (Fgf2). Coping with voluntary physical exercise also protects the BBB from stress deleterious effects by increasing Fgf2. Fgf2 is mostly expressed by glial cells, and viral-mediated astrocyte-specific Fgf2 upregulation prevents stress-induced social avoidance while downregulation increases stress susceptibility and blunts physical exercise benefits. Treatment of mouse and human endothelial cells with Fgf2 prior an immune challenge reduces BBB dysfunction, Cldn5 loss, and altered signaling supporting its protective role. Circulating FGF2 level is linked with depression severity and symptomatology in men and women reinforcing involvement of this growth factor in mood disorders.

Fig. 1. Environmental enrichment dampens stress-induced social avoidance and blood–brain barrier alterations in male mice.

A Experimental timeline for chronic social defeat stress (CSDS) with enriched environment (EE). Male mice were housed with a nestlet, plastic chew toy, and shelter beginning 3 d prior to CSDS and continuing until the last defeat, followed by social interaction (SI) testing. B Compared to previously published results from CSDS with standard cages, stressed EE mice show less deficits in social behavior measured by the SI test, and a greater percentage of resilience. SI ratio was calculated by dividing the time spent in the interaction zone in presence vs absence of a novel CD-1 aggressor (AGG). Mice with SI < 1 were classified as stress-susceptible (SS), while SI > 1 were resilient (RES). C Stressed EE mice also show less time in corners of the SI test than those stressed in plain cages with no access to a nestlet, plastic chew toy, and shelter. Representative heatmaps of SI test in the second trial (AGG present) show differences between CTRL and CSDS mice with standard caging and EE. D Heatmap showing transcription of BBB-related genes in the nucleus accumbens (NAc) and prefrontal cortex (PFC) after stress for EE mice. Cldn5, Ocln, and Fgf2 are upregulated in the NAc of all stressed mice. E Increased Cldn5 expression and decreased Il-6 in SS EE mice compared to published data from SS mice in plain cages. F CTRL, SS, and RES behavioral groups form distinct clusters based on principal component analysis of NAc gene expression data in standard CSDS but are grouped together in EE. G Fgf2 immunofluorescent labeling is increased in all stressed mice from the EE cohort, while Cldn5 relative to blood vessel area is diminished specifically in SS mice (scalebar = 50 μm). Fgf2 area correlates with the degree of Cldn5 loss suggesting a protective response. Data represent mean ± s.e.m., the number of animals is indicated on graphs. Group comparisons were evaluated with one- or two-way ANOVA followed by Bonferroni’s post hoc tests and correlation with Pearson’s correlation coefficient where appropriate; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Fig. 2. Environmental enrichment dampens stress-induced anxiety and blood-brain barrier alterations in the prefrontal cortex of female mice.

A Experimental timeline for subchronic variable stress (SCVS) with enriched environment (EE). Female mice were housed with a nestlet, plastic chew toy, and shelter beginning 3 days prior to stress and continuing until the last session, followed by elevated plus maze (EPM) and tissue collection 24 h later or further behavioral testing depending on the cohorts. B Compared to previously published results from female SCVS with plain cages and no access to a nestlet, plastic chew toy, and shelter, stressed EE mice show greater exploratory behavior characterized by open arm time in the EPM. Representative heatmaps show differences in EPM behavior between standard and EE SCVS. C Heatmaps showing transcription of BBB-related genes in the nucleus accumbens (NAc) and prefrontal cortex (PFC) after stress. Cldn5 deficits are seen in the NAc, but not the PFC. D Increased Cldn5 expression in stressed EE mice compared to published data from stressed mice in plain cages. E CTRL and SCVS mice form distinct clusters based on principal component analysis of PFC gene expression data when performed in standard cages, but in the EE cohort they are more closely grouped. F BBB-related genes in the male NAc and female PFC respond differently to stress. G No changes in immunofluorescent staining of Fgf2 or Cldn5 following SCVS in female mice with EE (scalebar = 50 μm). Data represent mean ± s.e.m., the number of animals is indicated on graphs. Group comparisons were evaluated with two-way ANOVA followed by Bonferroni’s post hoc tests or two-tailed t-tests with Welch’s correction and correlation with Pearson’s correlation coefficient where appropriate; *p < 0.05, **p < 0.01, ***p < 0.001.

Fig. 3. Physical exercise protects the blood-brain barrier from deleterious effects of stress with light-cycle running promoting resilience.

A Experimental timeline for chronic social defeat stress (CSDS) with physical exercise (PE). Male mice were habituated with a battery-powered running wheel prior to CSDS and had voluntary access to wheel running until the last defeat, which was followed by social interaction (SI) testing. B Compared to previously published results from CSDS with plain cages, stressed PE mice show similar deficits in social behavior measured by the SI test, but a greater percentage of resilience. SI ratio was calculated by dividing the time spent in the interaction zone in presence vs absence of a novel CD-1 aggressor (AGG). Mice with SI < 1 were classified as stress-susceptible (SS), while SI > 1 were resilient (RES). C Stressed PE mice show substantially less time in corners of the SI test than those stressed in plain cages with no running wheel (n = 11–41). Representative heatmaps of SI test in the second trial (aggressor present) show differences between CTRL and SS mice in CSDS with standard caging and EE. D Representative graph showing running activity per hour throughout CSDS. E Stressed mice run slightly more than controls. F Stress phenotype is associated with running during the light cycle, with RES mice running more during the day. G RES mice run more in the hour following stress (****p < 0.0001). H Heatmaps showing transcription of BBB-related genes in the nucleus accumbens (NAc) and prefrontal cortex (PFC) after stress with PE. Cldn5 is upregulated following 10 d CSDS in both brain regions of mice with access to PE. I Increased Cldn5 expression and decreased Il-6 in SS EE mice compared to published data from SS mice in plain cages. J No loss of Cldn5 immunofluorescent labeling in SS mice with PE access (scalebar = 50 μm). Fgf2 immunofluorescent labeling is increased in all stressed mice from the PE cohort. Data represent mean ± s.e.m., the number of animals is indicated on graphs. Group comparisons were evaluated with one- or two-way ANOVA followed by Bonferroni’s post hoc tests, or two-tailed t-tests with Welch’s correction; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Fig. 4. Upregulation of astrocytic Fgf2 prevents stress-induced social avoidance.

A Experimental timeline for chronic social defeat stress (CSDS) with physical exercise (PE). Male mice were habituated with a battery-powered running wheel prior to CSDS and had voluntary access to wheel running until the last defeat, which was followed by social interaction (SI) testing and tissue collection 24 h later. B Increased Fgf2 and astrocyte marker immunoreactivity is observed along blood vessels of resilient mice with access to PE when compared to unstressed PE controls (scalebar = 50 μm and 25 μm for high magnification images). C Experimental timeline for AAV5-gfaABC1D bilateral injection in the NAc. Mice were injected and 4 weeks later, NAc brain tissues were collected. D Mice injected with AAV5-mFgf2-P2A virus have higher Fgf2 signal intensity (left), precisely colocalized on blood vessels (CD31), compared to controls AAV5-GFP (middle). Fgf2 signal is strongly colocalizing with CD31+ endothelial cells (representative images on the right) (scalebar = 50 μm and 25 μm for high magnification images). E Experimental timeline for AAV5-gfaABC1D bilateral injection in the NAc, followed by CSDS, then Social Interaction (SI) test. F SI ratio is similar between AAV5-GFP and AAV5-mFgf2-P2A, either for control unstressed mice (CTRL, n = 5/group) or stressed mice (CSDS, n = 12/group). AAV5-mFgf2-PA2 mice from the stressed group show higher susceptibility compared to AAV5-GFP stressed mice (G) no difference in locomotion (H) however, they spend more in the interaction zone (I) and less time in the corners when the aggressor (AGG) is present, particularly in susceptible (SS) mice that received AAV5-mFgf2 virus (n = 5–12/group). J After CSDS, AAV5-mFgf2 mice show high levels of Fgf2 mRNA, no matter their CSDS group, compared to mice with control virus AAV5-GFP. Data represent mean ± s.e.m., the number of animals is indicated on graphs. Group comparisons were evaluated with one- or two-way ANOVA followed by Bonferroni’s post hoc tests, or two-tailed t-tests with Welch’s correction; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Fig. 5. Downregulation of astrocytic Fgf2 in the NAc increases stress susceptibility and blunts physical exercise benefits.

A Experimental timeline for AAV5-gfaABC1D-shRNA bilateral injection in the NAc. Mice were injected and 4 weeks later, NAc brain tissues were collected. B Mice injected with AAV5-shRNA-mFgf2 virus have lower Fgf2 signal intensity in astrocytes compared to controls AAV5-GFP (representative images of NAc infected area on the left, scalebar = 2 mm, and cell specificity on the right, scalebar = 50 μm with 25 μm for high magnification images below). C Experimental timeline for AAV5-gfaABC1D-shRNA bilateral injection in the NAc, followed by chronic social defeat stress (CSDS), then Social Interaction (SI) test. D SI ratio is lower for stressed mice with downregulation of astrocytic Fgf2 in the NAc. AAV5-gfaABC1D-shRNA-mFgf2 mice from the stressed group display higher susceptibility compared to AAV5-gfaABC1D-shRNA-scramble stressed mice (E) but no difference in locomotion (F). G Experimental timeline for AAV5-gfaABC1D-shRNA bilateral injection in the NAc, followed by CSDS with access to running wheels, then SI test prior tissue collection. H Chronic social stress promotes social avoidance as measured by lower SI ratio and increased number of entries in the corners in both AAV5-gfaABC1D-shRNA-mFgf2- and AAV5-gfaABC1D-shRNA-scramble-injected mice (I) and this is not related to locomotion impairment (J). K Running distance increases across days for mice injected with the control virus but not for AAV5-gfaABC1D-shRNA-mFgf2-injected animals (L). The dotted line is set at maximum running distance during the first 24 h. M After CSDS, AAV5-gfaABC1D-shRNA-mFgf2 mice show lower levels of Fgf2 mRNA without or with access to physical exercise (PE), compared to mice injected with control AAV5-gfaABC1D-shRNA-scramble virus. PE also elevates Fgf2 in this group. Data represent mean ± s.e.m., the number of animals is indicated on graphs. Group comparisons were evaluated with one- or two-way ANOVA followed by Bonferroni’s post hoc tests, or two-tailed t-tests with Welch’s correction; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Fig. 6. Treatment with Fgf2 reduces TNF-α-induced Cldn5 loss, endothelial cell signaling alterations, and barrier hyperpermeability.

A, Experimental timeline for inflammatory insult with TNF-α and FGF2 co-treatment. Once confluent (4 days), HBEC-5i or bEnd.3 were pretreated with FGF2 or vehicle for one hour and then stimulated with TNF-α or vehicle for up to 7 days. FGF2 alters mouse (B) and human (C) endothelial transcription in response to acute TNF-α stimulation. FGF2 promotes faster restoration of TNF-α-induced Cldn5 loss in mouse (D) and human (E) endothelial cells. Chronic stimulation with TNF-α leads to a reduction in endothelial monolayer integrity measured by trans-endothelial electrical resistance (TEER) in mouse (F) and human (G) endothelial cells. FGF2 co-treatment preserves normal TEER despite TNF-α. H 7 days of TNF-α treatment promotes spikes and discontinuities in Cldn5 tight junction strands in bEnd.3, which is reversed by FGF2 treatment (scalebar = 20 μm). Data represent mean ± s.e.m., and each experiment was replicated at least twice on independent samples. Group comparisons were evaluated with two-way ANOVA followed by Bonferroni’s post hoc tests; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Fig. 7. Fgf2 induces GSK3β phosphorylation and prevents β-catenin dissociation from tight junctions.

A 1 h pretreatment with Fgf2 increases serine-9 phosphorylation of GSK3β in HBEC-5i when compared to no treatment (CTRL 0 h). TNF-α treatment induces rapid, transient dephosphorylation of GSK3β, but this effect is not reversed by Fgf2 coadministration. Each dot represents a replicate (n = 3). B 1 h Fgf2 pretreatment diminishes basal β-catenin phosphorylation when compared to no treatment (CTRL 0 h). Further, while TNF-α induces a rapid reduction in phosphorylated β-catenin, Fgf2 reverses this dynamic upon inflammatory activation (n = 3). C In health control endothelial cells (top), β-catenin interacts with VE-Cadherin at the cell membrane, and this complex inhibits Cldn5 transcriptional suppression by FOXO1. Excess cytosolic β-catenin is phosphorylated by GSK3β, targeting it for degradation. When stimulated with TNFα, unbound β-catenin complexes with FOXO1, leading to suppression of Cldn5 expression (bottom, red arrow), while a small amount is targeted for degradation. Meanwhile, when FGF2 is co-administered with TNF-α (bottom, blue arrow), our results suggest that unbound β-catenin is strongly redirected toward GSK3β-mediated phosphorylation. D 30 min of TNF-α is sufficient to induce β-catenin distribution at tight junctions (n = 4 replicates) with representative images on the right (E) (scalebar = 20 μm). F Fgf2 attenuates TNF-α-induced reductions in the wound healing capacity of HBEC-5i (n = 4 replicates) (****p < 0.0001). Data represent mean ± s.e.m., and each experiment was replicated at least twice on independent samples. Group comparisons were evaluated with one or two-way ANOVA followed by Bonferroni’s post hoc tests or two-tailed t-tests with Welch’s correction when appropriate; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Fig. 8. The environment is a key variable determining BBB response to stress and FGF2 a biomarker of MDD severity and symptomatology.

Hierarchical clustering performed with Euclidean distances of blood-brain barrier (BBB)-related gene expression changes induced by stress exposure in the male nucleus accumbens (A, NAc) or female prefrontal cortex (B, PFC) in standard conditions (SE), with access to an enriched environment (EE), or voluntary physical exercise (PE). C Fibroblast growth factor 2 (FGF2) level in blood serum samples from men and women with a diagnosis of major depressive disorder (MDD) at various degrees of severity (D, minimal vs moderate severe: *p = 0.0171; minimal vs severe: *p = 0.0452). E FGF2 blood level vs MDD diagnosis, sex, and education level (men with no diploma: *p = 0.0172). F Proportion of individuals in each group with a university diploma (top) or employed (bottom). G Spearman correlation between circulating FGF2 and MDD symptoms according to the Patient Health Questionnaire (PHQ-9) item constructs. Data represent mean ± s.e.m., the number of individuals is indicated on graphs. Group comparisons were evaluated with two-way ANOVA followed by Bonferroni’s post hoc tests or two-tailed Mann-Whitney U-test; *p < 0.05, **p < 0.01.