Research ArticlePSYCHOLOGY

Corticosteroid signaling at the brain-immune interface impedes coping with severe psychological stress

See allHide authors and affiliations

Science Advances  29 May 2019:
Vol. 5, no. 5, eaav4111
DOI: 10.1126/sciadv.aav4111


The immune system supports brain plasticity and homeostasis, yet it is prone to changes following psychological stress. Thus, it remains unclear whether and how stress-induced immune alterations contribute to the development of mental pathologies. Here, we show that following severe stress in mice, leukocyte trafficking through the choroid plexus (CP), a compartment that mediates physiological immune-brain communication, is impaired. Blocking glucocorticoid receptor signaling, either systemically or locally through its genetic knockdown at the CP, facilitated the recruitment of Gata3- and Foxp3-expressing T cells to the brain and attenuated post-traumatic behavioral deficits. These findings functionally link post-traumatic stress behavior with elevated stress-related corticosteroid signaling at the brain-immune interface and suggest a novel therapeutic target to attenuate the consequences of severe psychological stress.


Psychological stress is a common life event, which often affects brain function and cognitive performance. Severely stressful conditions can have a long-term effect on the brain and contribute to a variety of chronic diseases, ranging from depression to post-traumatic stress disorder (PTSD) (1). When changes in the environment are perceived as a threat to homeostasis that may challenge the organism’s well-being, the body responds by initiating a protective mechanism through a cascade of events, which culminates with the release of glucocorticoids from the adrenal cortex (2). Such a response, if persists, can cause suppressive effects on peripheral immune responses (3, 4). This, together with the fact that adaptive immunity plays a role in the maintenance and functional plasticity of the central nervous system (CNS) (57), including in experimental animal models of coping with stress (811), emphasize that mental stress is reciprocally intertwined with immune system activity, although the extent of this relationship is not fully understood.

The cross talk between the immune system and the brain is tightly regulated by the restrictive nature of the CNS anatomy and is thus confined to specialized interfaces, such as the brain’s choroid plexus (CP) and the meningeal spaces (12, 13). Specifically, the CP was shown to function as an immunological interface through which blood-borne leukocytes can enter the CNS territory both for immunosurveillance in homeostasis and in the cases of brain pathologies, in which it was demonstrated that circulating immune cells contribute to brain repair (1417). Recently, CP-mediated interaction of immune cells with the CNS was shown to dysfunction in aging and under certain neurodegenerative conditions, partly due to peripheral immune changes (18, 19) and/or due to intrinsic signals from within the CNS (7). Furthermore, augmenting leukocyte trafficking across the CP by systemic immunomodulation was functionally linked with a beneficial effect on brain pathology (7, 1922). On the basis of these findings, we hypothesized that the fate of the CP as an immunological interface that supports leukocyte trafficking and thereby brain immunosurveillance might have a critical role in mediating the effects of psychological stress on anxiety-related behaviors.

Here, we found that severe stress in mice results in reduced immunosurveillance within the cerebrospinal fluid (CSF), which we found to be associated with corticosteroid (CORT) signaling at the brain-immune interface. Partial elimination of glucocorticoid receptor (GR) expression by CP epithelial cells by genetic manipulation enhanced its gateway activity for leukocyte trafficking during the stress response and promoted an anxiolytic effect, which was accompanied by accumulation of nonpathogenic Gata3 and Foxp3 T cells in the brain.


Severe psychological stress negatively affects CP gateway activity

On the basis of past findings showing that peripheral immune cells play an important role in stress resilience (10, 11, 23), we first assessed whether severe psychological stress affects spontaneous cell trafficking to the CNS. To this end, we adopted a mouse model of stress that leads to sustained behavioral deficits (24). In this model, mice are subjected to two sessions of inescapable electrical foot shocks, where the first session serves as a severe stressful event (“Trauma”) and the second as a boosting trigger (“Trigger”) (Fig. 1A). Mice exposed to such manipulation were previously shown to exhibit hypervigilant behavior, impaired attention, increased risk assessment, and sleep disorders (24). Quantitative polymerase chain reaction (qPCR) analysis of the CP, isolated from the brain ventricles at different time points following stress induction, revealed a sustained decrease in mRNA levels of genes encoding for trafficking molecules, such as icam1, cxcl10, and ccl2, following the combination of Trauma and Trigger, but not Trauma alone (Fig. 1B). These trafficking-related molecules were previously demonstrated to be essential for transepithelial migration of leukocytes across the CP (17, 25, 26), and their reduced expression levels were associated with decreased immunosurveillance of the CNS (7, 19, 20).

Fig. 1 CP gateway dysfunction for immunosurveillance following severe psychological stress.

(A) Scheme depicting the experimental design, showing the schedule of two sessions of stress induction by electrical foot shock followed by the testing period. (B) Gene expression levels, as measured by qPCR, in CP of mice following Trauma only or Trauma + Trigger. Gray area represents the SE margins of the nonstressed group (time: −24 hours, before Trauma) (n = 6 to 14 per group; Student’s t test; data represent combined analysis of results obtained from seven independent experiments). A.U., arbitrary units. (C) Quantitative analysis of leukocyte numbers (total number of CD45+ leukocytes; CD4+ and CD11b+ cells out of total CD45+) in the CSF of mice following Trauma + Trigger. d, day; each dot represents number of cells per microliter CSF of a single mouse [n = 12 to 13 per group; one-way analysis of variance (ANOVA) (F = 7.808; P = 0.0016 for CD45+, F = 4.516; P = 0.182 for CD11b+, F = 5.544; P = 0.0082 for CD4+) followed by Newman-Keuls post hoc test; data represent combined analysis of five independent experiments]. (D) Correlation between mRNA levels and total number of CD45+ leukocytes in the CSF of a separate cohort of stressed mice 24 hours following stress induction. Each dot represents a single mouse. In all panels, error bars represent mean ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001.

To further evaluate the effect of psychological stress on CNS immunosurveillance, we analyzed the cellular composition of the CSF using flow cytometry. We found a transient reduction in CD45+ leukocyte numbers, including both CD11b+ and CD4+ cells, occurring 24 to 48 hours following the last session of foot shocks (counting after the Trigger session in the sequential combination of the two sessions) (Fig. 1C). Analyzing the brain tissue, excluding the CP and most of the peripheral meninges, revealed a similar decrease in the numbers of CD4+ T cells (fig. S1). Further analysis revealed a positive correlation between the number of leukocytes in the CSF and gene expression levels of icam1, cxcl10, and ccl2 at the CP (Fig. 1D), substantiating the functional linkage between CP expression of genes encoding for leukocyte trafficking and CNS immunosurveillance.

Boosting the peripheral immune response before psychological stress enhances brain immunosurveillance and attenuates stress-related behavioral deficits

The observed reduction of immunosurveillance following severe stress, coupled with the linkage between CP dysfunction and the pathophysiology of neurodegenerative diseases (7, 19), led us to suggest that boosting CP gateway activity to support leukocyte trafficking would improve the ability to cope with stress via increased immunosurveillance of the CNS. To test this working hypothesis, we first employed a method that was previously used to transiently eliminate peripheral regulatory T cells (Tregs), thereby activating the immune system and evoking an interferon-γ (IFN-γ)–mediated immune response, which is associated with augmented CP gateway activity for leukocyte trafficking (19). Specifically, we used Foxp3.LuciDTR mice that express a diphtheria toxin (DTx) receptor under the FoxP3 promotor (FoxP3-DTR) (27), which enables the transient depletion of Tregs upon administration of DTx (19). Notably, wild-type mice lack DTx receptor; thus, administration of the toxin to wild-type mice in this paradigm was not found to lead to any alterations in CP activity or blood-brain barrier (BBB) integrity (19, 28, 29). FoxP3-DTR mice that were subjected to both Trauma and Trigger were injected daily with DTx for four consecutive days, starting from 2 days before the Trauma session, to achieve the depletion of Tregs at a time point coinciding with the end of stress induction, namely, right after the Trigger session (Fig. 2A and fig. S2). Analysis of gene expression 24 hours following the end of stress induction (24 hours after the Trigger session), a time point that showed a marked decrease in CP gateway activity in response to stress (Fig. 1), revealed an increase in the expression of leukocyte trafficking–related genes following Treg depletion, among which cxcl10 was the most pronounced (Fig. 2B).

Fig. 2 Breaking peripheral immune tolerance prevents post-traumatic stress behavior.

(A) Scheme depicting the experimental design, showing DTx injections superimposed on the schedule of stress induction. IHC, immunohistochemistry. (B) qPCR analysis of CP gene expression in mice 24 hours after the last stress episode, with (DTx) or without Stress Treg depletion (n = 6 per group; Student’s t test). (C) Representative fluorescence microscopy images (side panels show an enlargement of marked area) and quantification of CD3+ (red) cells at the (D) dentate gyrus (DG) of the HC and the (E) third ventricle CP 14 days following stress induction (n = 8 per group; Student’s t test). (F) Flow cytometry quantitative analysis of CD45+/TCRβ+ cells/brain (excluding CP and meninges) 7 days following stress induction [n = 5 per group; one-way ANOVA (F = 7.145; P = 0.0103) followed by Newman-Keuls post hoc test]. (G) Hypervigilant behavior, measured by acoustic startle reaction time (left) and amplitude (right), of stressed mice 14 days following stress induction [n = 16 per group; one-way ANOVA (F = 7.778; P = 0.0015 for reaction time, F = 7.053; P = 0.0029 for amplitude) followed by Newman-Keuls post hoc test; data are representative of two independent experiments]. In all panels, error bars represent mean ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001.

The enhanced expression by the CP of genes encoding for leukocyte trafficking molecules, following the depletion of Tregs, prompted us to assess whether this manipulation could subsequently increase the homing of immune cells into the CNS parenchyma. Immunohistochemical analysis of brain tissues 2 weeks following stress induction revealed that transient depletion of Tregs resulted in the accumulation of CD3+ cells, which were mostly observed along the interventricular foramina adjacent to the CP and in the hippocampus (HC) (Fig. 2, C to E). Moreover, about a quarter (26 ± 4%) of the CD3+ T cells in the HC were found to be positive for FoxP3+ (fig. S3), which might indicate an outcome of local conversion of T cell phenotype from effector cells to regulatory cells (30). Further flow cytometry analysis of whole brains (excluding the CP and most of the meningeal tissues) at 2 weeks following stress induction confirmed the increased numbers of CD45+/TCRβ+ T cells in mice that were transiently depleted of Tregs (Fig. 2F). To exclude potential cell entry through a breached BBB, we examined the endothelial basal lamina in both treated and untreated stressed mice and found no signs of perivascular cuffing associated with BBB breakdown and endothelial leukocyte transmigration (fig. S4) (31).

Peripheral immunomodulation was previously shown to affect behavioral outcome in the context of mental stress (10, 23). To assess whether augmented immune-brain trafficking through the CP might affect stress-related behavior, we repeated the experiment and tested the mice for post-traumatic hypervigilance 2 weeks following stress induction. Vehicle-treated stressed mice exhibited an increased startle response to acoustic stimulus, as previously reported in this model (24). We found that the transient depletion of Tregs in the context of stress resulted in a startle response similar to that observed in the nonstressed littermates (Fig. 2G). Together, these findings suggest that increased CP gateway activity for leukocyte trafficking following a severe stressful experience might contribute to the enhanced ability to cope with psychological stress.

Systemic blockade of GR signaling restores CP gateway activity and mitigates stress-related behavioral deficits

The observed results that CP gateway activity might be involved in stress resilience prompted us to further investigate which component(s) of the physiological stress response could affect the CP (Fig. 1). Psychological stress classically leads to the activation of the hypothalamic-pituitary-adrenal (HPA) axis, which culminates in increased secretion of CORTs (32). This response affects a broad range of peripheral tissues, mediated by the activation of GRs, which, among other functions, can suppress the activity of nuclear factor κB (NFκB) (33), a signaling pathway found to be relevant for CP activity in mediating transepithelial migration of leukocytes to the CNS territory (18). We therefore first tested whether blood levels of CORT were elevated in this model of stress, and whether subjecting the mice to the second stressor (Trauma followed by Trigger) would maintain high CORT levels. We found that 1 hour following exposure of the mice to Trauma stimulus, serum CORT levels were significantly elevated, and that the level remained high following the subsequent subjection to Trigger stimulus on the next day. After 24 hours, the levels of serum CORT returned to their baseline (Fig. 3A).

Fig. 3 Systemic blockade of GR signaling restores CP gateway activity and mitigates stress-related behavioral deficits.

(A) Serum CORT levels (pg/ml), measured at 1 hour following Trauma, 1 hour following Trigger preceded by Trauma, and 24 hours following Trigger preceded by Trauma [n = 4 per group; one-way ANOVA (F = 18.47; P < 0.0001) followed by Newman-Keuls post hoc test]. (B) qPCR analysis of CP gene expression 4 hours following intracerebroventricular injection of TNFα (100 ng per mouse) with or without intraperitoneal injection of hydrocortisone (125 μg per mouse) [n = 4 per group; one-way ANOVA (F = 5.695; P = 0.0116 for Icam1, F = 11.56; P = 0.0007 for Cxcl10, F = 6.205; P = 0.0087 for Ccl2) followed by Newman-Keuls post hoc test]. n.s., not significant. (C) qPCR analysis of CP gene expression in mice following 2 weeks of oral administration of corticosterone (100 μg/ml) via drinking water (n = 6 per group; Student’s t test). (D) qPCR analysis of CP gene expression 24 hours following stress induction in mice treated with mifepristone (RU-486; GR blocker) or vehicle [n = 6 to 12 per group; one-way ANOVA (F = 5.792; P = 0.0081 for Icam1, F = 2.535; P = 0.0979 for Cxcl10, F = 6.992; P = 0.0036 for Ccl2) followed by Newman-Keuls post hoc test]. (E) Quantitative analysis of CXCR3+ and CXCR3 CD45+/TCRβ+ cells within the CSF 5 days following stress induction with administration of mifepristone (RU-486; GR blocker) [n = 4 per group; one-way ANOVA (F = 0.038; P = 0.9624 for CXCR3, F = 5.938; P = 0.0227 for CXCR3+) followed by Newman-Keuls post hoc test]. (F) Immunohistochemical analysis of CD3+ cells/mm3 in the dentate gyrus of mouse HC [n = 5 to 10 per group; one-way ANOVA (F = 6.556; P = 0.0058) followed by Newman-Keuls post hoc test]. (G) Percentage of time engaged in risk assessment (right) and total time spent exploring the lit compartment of the Dark/Light test (middle) measured for nonstressed (CTRL) or stressed mice treated with GR blocker or vehicle [n = 9 to 10 mice per group; one-way ANOVA (F = 4.665; P = 0.0182 for time, F = 5.740; P = 0.0084 for risk) followed by Newman-Keuls post hoc test]. Left: Representative heat maps (top) and tracks (bottom). (H) Correlation analysis between CD3+ cells in the HC and time spent in risk assessment (left) and in the lit arena (right) in the Dark/Light test. Each dot represents a single mouse. Data are representative of two independent experiments. In all panels, error bars represent mean ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001.

We next wished to see whether the exposure to CORT outside the context of stress would be sufficient to induce an inhibitory effect on the CP. We envisioned that even a short exposure to CORT would have an effect on CP gateway activity, if it occurs concomitantly when a “danger signal” emerges from the brain parenchyma as a result of the stress. To this end, we used a paradigm that simulates danger signaling in the brain and induces the elevation of trafficking molecules by the CP. Specifically, mice received an intracerebroventricular injection of tumor necrosis factor α (TNFα) and an intraperitoneal injection of CORT with respective vehicle controls. While the administration of TNFα alone, as previously observed (17, 18), markedly increased CP epithelium expression of genes encoding for leukocyte trafficking molecules, cotreatment with CORT inhibited TNFα induction of icam1 and ccl2, but not of cxcl10 (Fig. 3B). The short exposure to CORT, administrated to control animals that were injected intracerebroventricularly with phosphate-buffered saline (PBS), appeared not to affect the expression levels of these genes, suggesting that a short exposure to exogenous CORT is enough to negate the effects of a danger signal coming from the brain, but might not be sufficient to decrease the basal expression levels of trafficking-related molecules.

We further tested whether the mere prolonged systemic elevation in CORT would be sufficient to induce an inhibitory effect on the basal CP expression of leukocyte trafficking–related genes. To overcome the relatively short biological half-life of endogenous CORT following a single dosage (34) and to avoid potential stress induced by repeated injections, CORT was provided to the mice through drinking water, similar to a method previously reported (35). Following 2 weeks of continuous CORT administration, we found a decrease in icam1, cxcl10, and ccl2 expression levels at the CP (Fig. 3C), confirming that a long-term exposure to CORT can potentially lead to suppressive effects on basal CP gateway activity.

The results shown above led us to explore whether a pharmacological inhibition of GR, in the context of a severe stressful event, could restore CP activity and alleviate the behavioral consequence of the stress. To this end, we systemically treated mice with the GR inhibitor mifepristone (RU-486; 20 mg/kg) 30 min before each electrical shock session. qPCR analysis of the CP for immune cell trafficking–related gene expression at 24 hours after stress induction showed increased cxcl10, ccl2, and icam1 mRNA levels in mice treated with a GR blocker relative to vehicle-treated stressed mice (Fig. 3D). Assessment of the cellular composition of the CSF by flow cytometry revealed an increase in TCRβ+/CXCR3+ cells (Fig. 3E), a cell population that binds to the chemoattractant CXCL10 and was previously shown to be associated with CNS immunosurveillance (36). Immunohistochemical analysis of brain tissues 2 weeks following stress induction revealed an increase in the number of CD3+ cells in the HCs of GR-blocked mice relative to vehicle-treated stressed animals (Fig. 3F and fig. S5). Together, the presence of a higher number of CD3+ cells in the HC, along with increased numbers of leukocytes in the CSF and the elevated expression of leukocyte trafficking genes by the CP, was in line with the results obtained following a transient depletion of Tregs (Fig. 2), although the effect was somewhat less intense.

To determine whether the enhancement of leukocyte-mediated immunosurveillance of the CSF via GR blockade would be linked with an improved behavior following the stress, mice were tested in a Dark/Light exploratory and risk assessment behavioral maze, as previously used to document the consequences of stress in this murine experimental model (24). Anxiety-like behavior manifested in the stressed vehicle-treated mice by reduced time spent visiting the bright chamber of the Dark/Light maze, as well as greater proportion of time spent engaging in risk assessment, when crossing over to the bright chamber. In contrast, systemic blockade of GR at the time of stress had an anxiolytic effect, demonstrated by exploratory behavior resembling that of the control group (Fig. 3G). Additional behavioral assessment using the open-field arena revealed that GR blockade led to a trend toward an increase in the cumulative time spent exploring the arena relative to vehicle-treated stressed mice (fig. S6).

Last, combined analysis of the histological data together with the behavioral indices obtained in the Dark/Light test per individual mouse revealed that the number of HC-infiltrating CD3+ cells negatively correlated with the proportion of time engaged in risk assessment and positively with time spent in the bright arena (Fig. 3H), suggesting the possible involvement of the infiltrating cells in the attenuation of the stress-related behavior.

CORTs can directly impair brain-immune communication

We next assessed whether the observed effect of CORT on CP gateway activity in vivo could be, at least in part, an outcome of a direct effect on the CP epithelium or whether it primarily reflected a suppressive effect on the peripheral immune system. To this end, we generated a primary culture of CP epithelial cells from C57BL/6J mice, grown to form a confluent monolayer, and challenged it for 24 hours with TNFα, as an NFκB-dependent inducer of CP gateway activity (17, 18), with or without CORT. While the administration of TNFα alone, as previously observed (17), markedly increased expression of genes encoding leukocyte trafficking molecules, the cotreatment with CORT inhibited TNFα induction of trafficking genes (Fig. 4A). Similar effects were also found in a human cell line of retinal pigment epithelium (fig. S7), a tissue with partial functional similarities to the CP epithelium.

Fig. 4 Corticosteroids can directly suppress CP gateway activity for leukocyte trafficking.

(A) qPCR analysis of CP gene expression in primary CP epithelial cultures, treated with either TNFα or hydrocortisone (CORT) at 100 ng/ml [n = 3 per group; one-way ANOVA (F = 547.2; P < 0.0001 for Icam1, F = 72.88; P < 0.0001 for Cxcl10, F = 235.6; P < 0.0001 for Ccl2)] followed by Newman-Keuls post hoc test; error bars represent mean ± SEM. ***P < 0.001. Data are representative of at least three independent experiments. (B) Representative microscopic images of ZO-1 (red) epithelial tight junction disruption in CP cell cultures. Boxes highlight the contours of tight junction morphology. (C) TEER measured over time in primary CP epithelial cultures treated with TNFα (100 ng/ml) (red), hydrocortisone (100 ng/ml) (green), or both (orange). The x axis of the graph was adjusted to omit TEER fluctuations associated with medium changes (n = 3 per group; gray area represents SE for each group).

Immunocytochemical examination of murine-cultured CP epithelial monolayer morphology revealed that the cotreatment with CORT reversed the remodeling of tight junctions, which was induced by TNFα (Fig. 4B). To further evaluate the relevance of CORT signaling in CP epithelial cells to the physiological barrier properties of the blood-CSF barrier, primary CP epithelial cells were cultured on Transwell membrane inserts, which enabled the constant monitoring of transepithelial electric resistance (TEER) of the established monolayer, a marker for the integrity of intercellular tight junctions (37). Administration of TNFα to the culture medium after the CP monolayer reached confluence caused an immediate reduction of TEER, while cotreatment with CORT completely reversed this effect, resulting in TEER even higher than that measured for the cultures treated with medium alone (Fig. 4C). Moreover, the administration of CORT alone was sufficient to increase TEER, suggesting that the effect on tight junction integrity may be independent of exogenous TNFα-related signaling. Overall, the direct effects of CORT on CP epithelial cells in culture suggest that the observed effect of CORT in vivo, in addition to its systemic effect on the peripheral immune system, could be an outcome of direct effect on the CP gateway.

Tissue-specific knockdown of GRs in the CP boosts its gateway activity and reduces anxiety-like behaviors

To determine the functional role of CP CORT signaling in mediating the consequences of severe stress, we established a genetic model for the specific knockdown of GR expression at the CP. Site-specific knockdown was achieved by administration of TAT-CRE recombinase to the brain ventricles of mice expressing the GR gene under a loxP cassette (GRloxP/loxP) (38); a similar manipulation was previously reported to enable CP-specific knockdown of a floxed gene (39). To confirm that the effect of an intracerebroventricular injection of TAT-CRE remains confined to the CP, we first verified this method in mice expressing yellow fluorescent protein (YFP) under a loxP cassette (40). The consequent recombination of genes, as detected by immunohistochemistry, was found to be restricted to the ventricles and occurred almost exclusively at the CP (Fig. 5A).

Fig. 5 Knockdown of GR expression at the CP increases its gateway activity and attenuates symptoms of PTSD-like behavior.

(A) Representative fluorescence microscopy image of R26R-EYFP mice injected intracerebroventricularly with TAT-CRE to demonstrate that the expression of YFP is restricted to the CP. DAPI, 4′,6-diamidino-2-phenylindole; Lat.v, lateral ventricle. (B) qPCR analysis of NR3C1 (Gr) and trafficking-related gene expression levels at the CP of GRloxP/loxP animals 2 weeks following intracerebroventricular TAT-CRE administration (n = 6 per group; Student’s t test). (C) Immunohistochemistry analysis and representative images of CPs from GRloxP/loxP mice after TAT-CRE injection. Images show ICAM-1 (green, top left) and GR (red, top right), merged with nuclear staining (blue, bottom left), and an enlargement of the marked area (bottom right). Arrowheads show epithelial cells expressing decreased levels of GR (n = 2 animals per group; Student’s t test). (D) CP gene expression 24 hours following stress induction in GRloxP/loxP mice pre-injected with TAT-CRE or vehicle (n = 6 per group; Student’s t test). (E) Acoustic startle reaction time (left) and amplitude (right) of stressed mice 14 days following stress induction (n = 6 to 7 animals per group; Student’s t test). (F) Time spent exploring the open arms of an elevated plus maze 12 days following stress induction (n = 6 to 7 animals per group; Student’s t test). (G) Percentage of time mice engaged in risk assessment (middle) and total time spent exploring the lit compartment of the Dark/Light maze (right) 7 days following stress induction (n = 5 per group; Student’s t test; representative graph of two independent experiments). Left: Representative heat maps (top) and tracks (bottom). In all panels, error bars represent mean ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001.

Next, to measure the effect of GR signaling ablation on CP gateway activity, we injected GRloxP/loxP mice intracerebroventricularly with TAT-CRE recombinase or vehicle (PBS) and allowed them to recover from the procedure for a period of 2 weeks. qPCR analysis of the mouse CP revealed a twofold reduction in GR expression following TAT-CRE recombination, relative to PBS-injected mice, which was accompanied by elevation in trafficking-related gene expression (Fig. 5B). Immunohistochemical analysis of GRloxP/loxP mice injected with TAT-CRE confirmed the decrease of GR at the protein level, which was accompanied by an increase in the intercellular adhesion molecule–1 (ICAM-1) expression (Fig. 5C). While epithelial GR expression was down-regulated following the administration of TAT-CRE, no loss of GR was detected in resident leukocytes that inhabit the CP, an observation that supports the notion that a local administration of TAT-CRE has no effect on the immune system.

Having established that tissue-specific knockdown of GR at the CP is sufficient to increase its gateway activity for leukocyte trafficking, we then assessed the potential of a similar manipulation in the context of stress. To that end, following the 2-week recovery period after TAT-CRE or vehicle injection, mice were subjected to the stress paradigm, and their CPs were analyzed by qPCR. We found that even in the context of stress, the knockdown of GR allowed increased expression of trafficking-related genes 24 hours following stress induction (Fig. 5D).

We next measured the effects of CP-specific knockdown of GR on post-traumatic anxiety-like behaviors. We found that knockdown of the GR in the context of stress partially attenuated startle responses to acoustic stimuli (Fig. 5E), showed a trend toward increased time spent exploring the open arms of an elevated plus maze (Fig. 5F), and diminished anxiety behavior as indicated by both reduced risk assessment and increased time spent in the bright arena of a Dark/Light maze (Fig. 5G). Together, these findings further substantiate the notion that CORT signaling in CP epithelial cells plays a role in modulating CP gateway activity for leukocyte trafficking from the blood circulation to the CNS territory.

Increased CP gateway activity under psychological stress leads to recruitment of immune-modulating T cells into the brain

Last, we wished to determine whether GR reduction was associated with increased immune cell trafficking in the context of stress and the nature of these cells. Flow cytometry analysis of the brains of TAT-CRE and vehicle-injected stressed and naïve mice, excluding the CP and most of the meningeal tissues, identified an increase in the number of CD45+TCRβ+ T cells following GR knockdown. This increase was reflected both in absolute cell number and in the percentage of T cells out of total CD45+ cells (Fig. 6, A and B). Knockdown of GR alone, in the absence of stress, was not sufficient to induce trafficking of T cells; however, in stressed mice, the recruitment of T cells correlated with the degree of GR knockdown (Fig. 6C). We further compared the number of brain-infiltrating T cells in TAT-CRE–injected mice, relative to their anxiety indices, and found that their presence correlated with the degree of attenuation in the anxiety-like behavior (Fig. 6D), suggesting that these cells play a positive role in coping with stress.

Fig. 6 Increased CP gateway activity under psychological stress leads to recruitment of nonpathological T cells into the brain.

(A) Flow cytometry of T cell (CD45+CD11bTCRβ+) ratio out of total CD45+ cells in the brains (excluding CP and most of the meningeal tissues) of TAT-CRE or vehicle intracerebroventricularly injected stressed and unstressed (naïve) GRloxP/loxP mice 17 days following stress induction [n = 3 to 7 mice per group; one-way ANOVA (F = 6.170; P = 0.0061) followed by Newman-Keuls post hoc test]. (B) Flow cytometry quantification of CD45+/CD4+ T cells in the brains (excluding CP and meningeal tissues) of TAT-CRE or vehicle-injected mice at day 14 following stress (n = 5 per group; Student’s t test). (C) Correlation analysis between brain T cell to microglia ratio and the relative expression of the GR in TAT-CRE (black) and vehicle (red) intracerebroventricularly injected stressed GRloxP/loxP mice. (D) Correlation between the number of CD4+ cells in the brain and time spent in risk assessment (left) and in the lit arena (right) of the Dark/Light maze (red, vehicle treated; black, TAT-CRE intracerebroventricularly injected GRloxP/loxP mice). Statistical analysis of the curve fit was measured only for the TAT-CRE group, excluding the PBS-treated group. Each dot represents a single mouse. (E) Heat map showing the average expression of selected genes by 1600 brain-infiltrating CD45+CD11bTCRβ+ cells, clustered into 18 meta-cells. Meta-cells are ordered using hierarchical clustering of T cell–specific gene lists. Colors represent a spectrum of each row relative minimal (blue) to maximal (red) gene expression. (F) Proportion of T cells classified as CD4 or CD8 according to their gene expression pattern. (G) Assessed ratios between classical helper T cell phenotypes, as derived by the product of transcription factor expression and the number of cells in each cluster. (H) Abundance of each meta-cell cluster (calculated as the average unique molecular identifiers per cell in each meta-cell) in stressed, vehicle (sham, red)–injected, or TAT-CRE (CRE, black)–injected mice. In all panels, error bars represent mean ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001.

The finding that augmenting the gateway activity of the CP in the context of stress leads to attenuated post-traumatic behavioral aberrations prompted us to further analyze the population of cells that are recruited into the brain. Because the overall number of infiltrating immune cells was too low for further in-depth assessment by conventional flow cytometry, we opted for a more sensitive analysis method using single-cell RNA sequencing (RNA-seq). To this end, the brains of GR knockdown and sham-treated stressed mice were processed into a single-cell suspension for flow cytometry, and the CD45+CD11bTCRβ+ cells were sorted for single-cell RNA sequencing using massively parallel single-cell RNA sequencing (MARS-seq). Unsupervised clustering analysis of T cells from both GR knockdown and sham-treated mice into meta-cells (see Materials and Methods) revealed the presence of both CD4+ and CD8+ cells. Classification of the CD4+ cells according to the expression of transcription factors associated with different T helper cell subsets identified these as mainly T helper 2 (TH2) cells (Gata3) and Tregs (Foxp3), with no detectable expression of either TH1 (Tbx21) or TH17 (Rorc) (Fig. 6, E to G). Comparison between the cell populations found in the brains of vehicle and TAT-CRE–injected stressed mice showed no significant differences in their meta-cell distribution profile, except for the overall number of recruited cells (Fig. 6H and fig. S10). These findings suggest that the level of CP gateway activity affects the number, but not the nature, of T cells recruited into the brain, which might indicate that entry of leukocytes into the brain is a physiological response that occurs spontaneously but is apparently insufficient under severe stress. Moreover, it supports the contention that improved recovery from the stress, achieved by reducing CORT signaling at the CP, enhanced a natural process of coping with stress.


The present study demonstrates a functional linkage between the ability to cope with mental stress and CNS immunosurveillance, which is mediated, at least in part, via the brain’s CP. We show that under severe psychological stress, CP gateway activity for leukocyte trafficking is suppressed, whereas restoration of this activity, either by transiently breaking peripheral immune tolerance or by neutralizing CORT signaling, recruits TH2 and Tregs to the CNS territory and attenuates anxiety-like behaviors. Notably, by demonstrating both in vitro and in vivo that CORT signaling could directly inhibit CP gateway activity, we identified a novel GR-dependent stress-induced pathophysiology of the CP that affects the ability to cope with mental stress.

The activity of the CP as a gateway for leukocyte trafficking was previously shown to be affected by signals derived from both sides of the blood-CSF barrier—apically from the CSF compartment, and basally from the CP stroma or blood circulation—rendering it susceptible to rapid changes in the levels of cellular and soluble factors (7, 18, 41, 42). Therefore, the decrease in CP gateway activity, reported here in the context of severe mental stress, could result not only from direct GR-mediated signaling within the CP epithelium but also via CORT-mediated immune suppression of peripheral immune activity that would affect the overall availability of IFN-γ. Accordingly, alleviation of peripheral immune suppression by transient depletion of Tregs before exposure to the stress, as well as systemic GR blockade, prevented CP dysfunction and partially attenuated the stress-induced increase in anxiety-like behaviors. We found that the mere blocking of CORT signaling at the CP, without any systemic immunomodulation, was sufficient to attenuate the stress-induced increase in anxiety-like behaviors, a finding that emphasizes the direct involvement of CP epithelium in coping with psychological stress.

In this study, the CPs from all four brain ventricles were taken together as a uniform functional cohort; however, it is possible that CP tissues from different ventricles have distinct immunological functions and different susceptibility to the effects of CORT. In addition, other roles of the CP, which are not related to leukocyte trafficking, were not investigated in this work. It is thus possible that CORT signaling might affect functions of the CP unrelated to leukocyte trafficking, such as CSF formation and clearance.

The engagement of CORTs with their receptor (GR) can form a transrepressive complex with NFκB, which prevents it from binding to its target genes (43). Recently, the NFκB signaling pathway in the CP epithelium was shown to be essential for the induction of leukocyte trafficking activity (18). Accordingly, the potent ability of CORTs to antagonize TNFα-induced gene expression and enhance tight junction integrity in primary CP epithelial cultures supports the possibility that sustained stress-associated GR signaling could render the CP unresponsive to pathology-derived CNS danger signals. Glucocorticoid secretion in the organism is constantly maintained at a basal level that fluctuates over the course of the day. Nevertheless, in our experience, we could not detect any changes in CP gateway activity when comparing the expression of trafficking-related molecules at different times of the circadian cycle.

The GR, in addition to its transrepression activities, can serve as a transcription factor, with a wide array of transactivation targets that can lead to diverse effects throughout many organs of the body. One of the suspected functions of CORT that extends beyond immune-mediated functions is neuronal memory consolidation, which was suggested to be affected during the stress response (44). Accordingly, exogenous administration of high doses of CORTs at the time of stress was shown to facilitate fear extinction and to attenuate post-traumatic stress behavior (45). In the same study, lower doses of exogenous CORT were ineffective in attenuating stress-induced abnormalities and significantly increased the propensity of mice to develop extreme behavioral responses. In our experience, a single injection of high dose of hydrocortisone can lead to an immediate transient decrease in CP expression of trafficking-related molecules, which rapidly rebounds to higher than baseline levels (fig. S9). A possible explanation for this phenomenon might lie with the HPA axis activity being subjected to negative feedback control by glucocorticoids. Accordingly, acute high levels of exogenous glucocorticoid administration could lead to the suppression of both basal and stress-induced HPA axis activities.

Exposure to a single stressor (Trauma alone) did not lead to the suppression of the CP and even led to a slight trend toward an elevation of gene expression of trafficking-related molecules (Fig. 1B, ccl2). These results are in line with previous studies by our group (23), in which it was shown that a single short episode of stress could induce, rather than suppress, trafficking of T lymphocytes into the CNS territory. Note that while previous studies from our team reported a spontaneous increase in trafficking molecules at the CP under mild stress (23), rather than the shutdown reported here, this increase was shown to be beneficial for coping with the stress, which is consistent with our present findings. Moreover, coping with stress was further improved by augmenting immune cell trafficking at the CP using myelin oligodendrocyte glycoprotein immunization (23). Together, these findings might suggest that suppression of trafficking through the CP occurs only beyond a certain “threshold,” which is reached upon extremely severe or repeated episodes of stress. Thus, while different levels of stress might have different outcomes on CP function, in either case, augmenting CP gateway activity for leukocyte trafficking had a beneficial effect on stress resilience.

Augmenting CP gateway activity was previously associated with the recruitment of immunoregulatory monocytes and T cells that homed to the sites of brain pathology (19, 20). In this study, we found that increasing CP expression of leukocyte trafficking molecules was associated with an increase in leukocyte abundance in the CSF and the brain parenchyma. Among the cells detected, a dominant population was positive for the chemokine receptor CXCR3, which coincided with increased expression levels of its ligand, cxcl10, by the CP. Similarly, CXCR3+ memory T cells were previously reported to be the primary population of T cells in the CSF under physiological conditions and were suggested to infiltrate through the CP (36). In parallel to the increased number of T cells within the CNS, both systemic and local blockage of GR signaling resulted in a clear reduction of anxiety-related behavior. The lower anxiety levels correlated with an increase in the number of T cells recruited to the brain parenchyma, supporting the contention that these cells play a beneficial role in buffering stress responses (46). Notably, while the knockdown of CP GR alone increased the expression of trafficking-related molecules, no recruitment of cells into the brain parenchyma was detected in the absence of stress. This finding may suggest that the expression of trafficking-related molecules by the CP reflects the availability of CSF-borne immunosurveillance of the brain but will not by itself lead to unnecessary infiltration of leukocytes into the brain parenchyma.

Systemic T cells were previously suggested to support brain functional plasticity, including higher cognitive performance (7), neurogenesis (5, 47), and social behavior (48), via the production of cytokines and neurotrophic factors from the CNS outer borders. Moreover, the presence of T cells at the CP compartment was proposed to support the physiological process of coping with mild stress (23). Nevertheless, the role of T cells in the parenchyma under stress-induced anxiety-like behaviors has not been fully elucidated. Different subsets of peripheral immune cell populations have been linked to different outcomes in experimental paradigms of psychological stress. Thus, for example, monocytes were found to exacerbate neuroinflammation and anxiety-related behavior (49, 50), whereas T cells were suggested to mediate stress resilience (10, 11). Therefore, it is possible, as was previously demonstrated both in acute CNS injury and in chronic neurodegenerative diseases (20, 28), that T cell recruitment is an interim phase in T cell–mediated immunoregulation of the brain microenvironment, which culminates in either local conversion of effector T cells into Tregs or further recruitment of immunosuppressive cells to the brain parenchyma and mitigation of neuroinflammatory responses (20). Single-cell RNA-seq of the T cells that homed to the CNS did not detect helper T cells expressing the transcription factors associated with the TH1 (Tbx21) or TH17 (Rorc) lineages, suggesting that the nature of the recruited cells is non-encephalitogenic, with potential anti-inflammatory activity. Specifically, TH2 cells were previously suggested to support the intrinsic anti-inflammatory properties of the brain (51), and interleukin-4 (IL-4), the central cytokine secreted under the TH2 response, was previously suggested to play a vital role in the regulation of brain maintenance and repair (52, 53). The finding that the relative ratio between the different meta-cell populations was not affected by reducing GR expression at the CP, and only the total number of infiltrating T cells was increased, reinforces our contention that augmenting CP gateway activity enables the boosting of a natural process that is insufficient when the stress is severe (46). Notably, our results do not imply that CORT-mediated repression of CP gateway activity restricts the migration of specific subsets of T cells; rather, we observed that enhanced gateway activity in the context of stress was accompanied by recruitment of GATA3 and FoxP3, but not TBX21 or RORgT-expressing T cells, to the brain parenchyma. Note that our findings do not exclude the potential involvement of other “non-T” immune cell populations in the response to psychological stress. While TCRβ+ cells were the only cell population whose abundance we showed to significantly change, it is possible that smaller immune cell populations could have been overlooked.

Overall, the present study attributes a novel negative role to CORT signaling in coping with severe psychological stress and functionally links it to limited CP-mediated immune cell trafficking to the CNS. Our results further highlight the GR expressed in CP epithelial cells as a potential target for preventing stress-induced psychopathologies, including anxiety, PTSD, and depression.



Grl1loxP/loxP mice (38) were provided by G. Schütz (German Cancer Research Center, Heidelberg, Germany). Rosa26-EYFP (40) and Foxp3.LuciDTR mice (on a C57BL/6J background) (27) were bred and maintained by the Animal Breeding Center of the Weizmann Institute of Science. C57BL/6J mice were supplied by Harlan Biotech (Jerusalem, Israel). All behavioral tests were conducted during the dark hours in a dimly lit room. All experiments were in compliance with the regulations formulated by the Institutional Animal Care and Use Committee (IACUC) of the Weizmann Institute of Science.

Stress induction

Mice were exposed to two sessions of electrical foot shocks on consecutive days, as previously described (24). On day 1, mice received 14 shocks of 1 mA, 1 s in duration over 85 min at variable intervals, representing the “trauma” in “context A.” On day 2, the same mice received five shocks of 0.7 mA, 1 s in duration over 5 min at fixed intervals, representing the “trigger” in “context B.” Shocks were applied in a fear-conditioning apparatus (TSE Systems) in a transparent Plexiglas cage (21 cm by 20 cm by 36 cm) with a metal grid floor. Preexperimental habituation took place in a room adjacent to the experiment room and was illuminated with red light. Between animals, the grid and cage were cleaned with 10% ethanol solution. Animals were transferred to the experiment room in darkness, and experiment room lights were kept off. Context A consisted of 10-lux illumination, without background noise. Context B consisted of 70-dB background noise and no illumination.

Systemic blockade of GR

For the systemic blockade of GR, mifepristone (RU-486; Sigma-Aldrich) was administered intraperitoneally 30 min before each foot shock session at a dose of 20 mg/kg.

RNA purification, cDNA synthesis, and real-time qPCR

Total RNA of the CP tissues, or from cell cultures, was extracted using the ZR RNA MicroPrep Kit (Zymo Research). mRNA (1 μg) was converted to complementary DNA (cDNA) using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). The expression of specific mRNAs was assayed using fluorescence-based real-time qPCR. qPCRs were performed using Power SYBR Green PCR Master Mix (Applied Biosystems). Reactions were performed in triplicate for each sample using the ΔΔCT method. Peptidylprolyl isomerase A (ppia) and hypoxanthine-guanine phosphoribosyltransferase (hprt) were chosen as reference genes in murine and human tissues, respectively, and according to their stability in the target tissue. The amplification cycles were 95°C for 5 s, 60°C for 20 s, and 72°C for 15 s. At the end of the assay, a melting curve was constructed to evaluate the specificity of the reaction. All real-time qPCRs were performed and analyzed using the StepOnePlus PCR System (Applied Biosystems).

For all murine tissues, the following primers were used: Ppia, 5′-AGCATACAGGTCCTGGCATCTTGT-3′ (forward) and 5′-CAAAGACCACATGCTTGCCATCCA-3′ (reverse); Icam1, 5′-AGATCACATTCACGGTGCTGGCTA-3′ (forward) and 5′-AGCTTTGGGATGGTAGCTGGAAGA-3′ (reverse); Ccl2, 5′-CATCCACGTGTTGGCTCA-3′ (forward) and 5′-GATCATCTTGCTGGTGAATGAGT-3′ (reverse); Cxcl10, 5′-AACTGCATCCATATCGATGAC-3′ (forward) and 5′-GTGGCAATGATCTCAACAC-3′ (reverse); Gr, 5′-ACCACCTCCCAAACTCTG-3′ (forward) and 5′-GTAATTGTGCTGTCCTTCCA-3′ (reverse).

For the human ARPE cell line, the following primers were used: Hprt, 5′-CTGGCGTCGTGATTAGTG-3′ (forward) and 5′-TAAACACCCTTTCCAAATCCTC-3′ (reverse); Ccl2, 5′-AAAGAAGCTGTGATCTTCAAGACC-3′ (forward) and 5′-TTCAAGTCTTCGGAGTTTGGG-3′ (reverse); Cxcl10, 5′-GAACCTCCAGTCTCAGCA-3′ (forward) and 5′-GGTACTCCTTGAATGCCAC-3′ (reverse); Icam1, 5′-GTGACCATCTACAGCTTTCC-3′ (forward) and 5′-GCCTCACACTTCACTGTC-3′ (reverse).

CSF collection

CSF was collected by the cisterna magna puncture technique, as previously described (7). Briefly, mice were anesthetized and placed on a stereotactic instrument so that the head formed a 135° angle with the body. A sagittal incision of the skin was made inferior to the occiput, and the subcutaneous tissue and muscle were separated. A capillary was then inserted into the cisterna magna through the dura matter lateral to the arteria dorsalis spinalis. Approximately 10 μl of CSF could be aspirated from each mouse. The collected CSF was taken for analysis by flow cytometry.

Intracerebroventricular injections

TNFα (100 ng) dissolved in PBS to a final volume of 10 μl was injected intracerebroventricularly (0.4 mm posterior to the bregma, 1.0 mm lateral to the midline, and 2.0 mm in depth from the brain surface), as described (7).

Flow cytometry sample preparation and analysis

Before all tissue collections, mice were intracardially perfused with PBS. Brains were dissected, dissociated using a gentleMACS dissociator (Miltenyi Biotec), and loaded on a Percoll gradient (GE Healthcare) to isolate leukocytes. CSF samples were labeled without further processing. The following fluorochrome-labeled monoclonal antibodies were used according to the manufacturers’ protocols: fluorescein isothiocyanate (FITC)–conjugated anti-CD45.2, FITC-conjugated anti-TCRβ, allophycocyanin (APC)–conjugated anti-CD11b, BV421 anti-TCRβ, BV421-conjugated anti-CD45, peridinin chlorophyll protein (PerCP)–Cy5.5–conjugated anti-CXCR3, and phycoerythrin (PE)-conjugated anti-CD45 (all from BioLegend) and BD Horizon v450–conjugated anti-CD4 and Alexa Fluor 700–anti-conjugated CD45.2 (BD Biosciences). Flow cytometry analysis was performed on each sample using a BD Biosciences LSRII flow cytometer, and the acquired data were analyzed using FlowJo software (Tree Star). Gating strategies for all flow cytometry analyses are shown in fig. S8.

Single-cell sorting

Cell populations were sorted with a SORP-Aria cytometer (BD Biosciences, San Jose, CA). Before sorting, all samples were filtered through a 40-μm nylon mesh. Samples were gated for CD45+CD11bTCRβ+ after exclusion of doublets. Isolated cells were single cell–sorted into 384-well cell capture plates containing 2 μl of lysis solution and barcoded poly(T) reverse transcription (RT) primers for single-cell RNA-seq (54). Four empty wells were maintained in each 384-well plate as a no-cell control. Immediately after sorting, each plate was spun down to ensure cell immersion into the lysis solution, snap-frozen on dry ice, and stored at –80ºC until processing.

MARS-seq library preparation

Single-cell libraries were prepared as previously described (54). Briefly, mRNA from cells sorted into cell capture plates was barcoded, converted into cDNA, and pooled using an automated pipeline. The pooled sample was then linearly amplified by T7 in vitro transcription, and the resulting RNA was fragmented and converted into a sequencing-ready library by tagging the samples with pooled barcodes and Illumina sequences during ligation, RT, and PCR. Each pool of cells was tested for library quality, and concentration was assessed as described (54). MARS-seq libraries, pooled at equimolar concentrations, were sequenced using an Illumina NextSeq 500 sequencer at a sequencing depth of 50,000 to 100,000 reads per cell.

Analysis of single-cell RNA-seq data

All MARS-seq libraries were sequenced using an Illumina NextSeq 500 system at an average sequencing depth of 50,000 reads per cell. Sequences were demultiplexed, mapped, and filtered as previously described (54), extracting a set of unique molecular identifiers per cell. Cells were then clustered using the MetaCell analysis package (55). Briefly, informative genes were used to compute cell-to-cell similarity and to build a K-nn graph (k = 50) to group cells into cohesive groups (or meta-cells). Last, the package used bootstrapping to derive strongly separated clusters. The MetaCell package is described in detail by Giladi et al. (55).

Conditional depletion of Tregs

DTx (8 ng g−1 body weight; Sigma-Aldrich) was injected intraperitoneally daily for four consecutive days to Foxp3.LuciDTR mice. The efficiency of DTx was confirmed by flow cytometry analysis of immune cells in the spleen, achieving almost complete depletion of the green fluorescent protein (GFP)–expressing FoxP3+ CD4+ Tregs (fig. S2).

Immunohistochemistry and immunocytochemistry

For staining of brain sections, two different tissue preparation protocols (paraffin-embedded or microtomed free-floating sections) were applied, as previously described (19). The primary antibodies used were rabbit anti-CD3 (1:500, Dako), rat anti-Foxp3 (1:20, eBioscience), rabbit anti-GR (1:100, Cell Signaling Technology), rat anti-ICAM (1:100, Abcam), rabbit anti-laminin (1:100, Abcam), and mouse anti–β-dystroglycan (1:100, BioLegend). Secondary antibodies included Cy2/Cy3 anti-rabbit/rat/mouse antibodies (1:200, all from Jackson ImmunoResearch). The slides were exposed to Hoechst for nuclear staining (1:10,000, Invitrogen Probes) for 30 s. For Foxp3 intracellular staining, antigen retrieval from paraffin-embedded slides was performed using the Retrievagen Kit (BD Pharmingen). For immunocytochemistry, CP cells were isolated and grown on coverslips to confluence, as described before (17). Cells were washed with PBS and fixed with methanol-acetone (1:1) for 10 min at −20°C followed by two washing steps with PBS. The coverslips of the cultured CP cells were blocked with an M.O.M. immunodetection kit reagent (Vector Laboratories) containing 0.3% Triton X-100 (Sigma-Aldrich) and stained with mouse anti–ZO-1 (1:100, Invitrogen). The secondary antibody used was Cy3-conjugated donkey anti-mouse antibody (1:200, Jackson ImmunoResearch). The coverslips were exposed to Hoechst stain (1:10,000, Invitrogen) for 30 s and mounted onto slides using Immu-Mount (Thermo Scientific). For immunocytochemistry of the ARPE cell line, cells were fixed with 2.5% paraformaldehyde for 20 min, and 20% horse serum was used as a blocking agent. Mouse anti-human CD54 (Thermo Scientific) was used as primary antibody. A fluorescence microscope (Nikon Eclipse 80i) was used for microscopic analysis. The fluorescence microscope was equipped with a digital camera (DXM 1200F, Nikon) and with 20× numerical aperture (NA) 0.50 and 40× NA 0.75 objective lenses (Plan Fluor, Nikon). Recordings were made using acquisition software (NIS-Elements, F3). To obtain an estimate of the number of labeled cells per cubic millimeter volume, the average number of cells counted in the selected sections (average surface area = 1 mm2, thickness = 0.006 mm) was multiplied by 166.66. Hoechst staining was routinely used for nuclear labeling, which served to verify quantification of the cells. Before quantification, slices were coded to mask the identity of the experimental groups, and cell number was quantified by an observer blinded to the origin of the sample. To avoid overestimation due to counting fragments of cells that spanned several sections, only cells that had an intact morphology and a nucleus that was >5 μm in diameter were counted. For the quantification of corrected total or mean fluorescence intensity, images were analyzed using ImageJ (National Institutes of Health) via designed macros specific for each staining. Representative images were cropped, merged, and optimized using Photoshop CS6 13.0 (Adobe) and were arranged using Illustrator CS5 15.1 (Adobe).

Enzyme-linked immunosorbent assay CORT measurements

For CORT measurements, animals were quickly anesthetized using isoflurane, and their retro-orbital blood was immediately collected into MiniCollect Z serum separator tubes (Greiner Bio-One). Serum was isolated according to the manufacturer’s protocol and snap-frozen at −80ºC until further processing. Serum CORT levels were measured using a Corticosterone ELISA kit (Enzo).

Startle response

Startle response (TSE Systems) measurements were performed as previously described (24). Briefly, mice were placed in a small Plexiglas and wire mesh cage on top of a vibration-sensitive platform in a sound-attenuated, ventilated chamber. A high-precision sensor, integrated into the measuring platform, detected movement. Two high-frequency loudspeakers inside the chamber produced all the audio stimuli. The acoustic startle response session began with 5-min acclimation to white background noise [65 db (A)] maintained throughout the session. Two sets of 12 startle stimuli [120 db (A), 40 ms in duration, with randomly varying intervals of 12 to 30 ms] were presented, interspaced by 48 startle stimuli randomly preceded by prepulses of [65 db (A) (40 ms)], [74 db (A) (40 ms)], [78 db (A) (40 ms)], or [82 db (A) (40 ms)]. Together, 72 stimuli were measured for each mouse. Latency to peak startle amplitude was measured both in response to startle stimuli and in response to startle stimuli preceded by prepulses.

Dark/Light transfer test

The apparatus for the Dark/Light transfer test (TSE Systems) consists of a Plexiglas box divided by a partition into two areas: one dark (14 cm by 27 cm by 26 cm) and the second brightly illuminated (30 cm by 27 cm by 26 cm, 700 lux). These areas are connected by a sliding door located at the floor level in the center of the partition. Mice were placed in the dark area, and the connecting door was opened to initiate a 5-min test session. The animal’s movements were recorded and scored using a camera and automated software (Ethovision, Noldus). Time spent in the lit arena, number of visits to the lit arena, and total distance traveled in the lit arena were measured. In addition, risk assessment behavior was assessed. Percentage risk assessment time was calculated as the amount of time spent in the risk assessment arena (a rectangular zone, 3 cm by 6 cm adjacent to the passage into the lit section) as a percentage of total time spent in the lit arena outside of the risk assessment zone.

Elevated plus maze

The elevated plus maze apparatus consisted of a gray polyvinyl chloride maze, comprising a central part (5 cm by 5 cm), two opposing uncovered arms (30.5 cm by 5 cm), and two opposing covered arms (30.5 cm by 5 cm by 15 cm). The apparatus was elevated at a height of 53.5 cm, and the open arms were illuminated with 10 lux. Mice were placed in the center, facing an open arm to initiate a 5-min session test. The time spent and the number of entries to the uncovered arms were measured.

Open-field test

Mouse exploratory behavior was recorded and analyzed over a 5-min period using an automated video tracking system (VideoMot 2, V5.76) in an open-field box made of gray plastic with 50 cm by 50 cm surface area and 30-cm-high walls (TSE Systems). Tracing paths of the mice were recorded, and time spent in the corners versus the middle of the arena was evaluated.

Primary culture of CP cells

Cell cultures were produced as previously described (18). Briefly, CP tissues were dissected following perfusion, digested with 0.25% trypsin (Sigma), and manually dissociated into a single-cell suspension. Cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM)/Ham’s F12 (Invitrogen) supplemented with 10% fetal calf serum (FCS), 1 mM l-glutamine, 1 mM sodium pyruvate, penicillin (100 U/ml), streptomycin (100 mg/ml), insulin (5 μg/ml), 20 μM arabinosylcytosine, sodium selenite (5 ng/ml), and epidermal growth factor (EGF) (10 ng/ml) (Sigma-Aldrich). The medium was refreshed every 24 hours, and after 72 hours, TNFα (100 ng/ml) (PeproTech) was added to the cells with or without hydrocortisone (100 ng/ml) (SOLU-CORTEF, Pfizer).

Human ARPE cell line

The ARPE-19 cell line was cultured in 24-well plates at 1 × 105 cells per well in DMEM/Ham’s F12 medium (Invitrogen) supplemented with 10% FCS penicillin (100 U/ml) and streptomycin (100 mg/ml). Once cells reached confluence, they were treated with TNFα (100 ng/ml) (PeproTech) with or without hydrocortisone (100 ng/ml) (SOLU-CORTEF, Pfizer).

TEER measurement

For the TEER measurements, CP cells were isolated as above and plated on 0.4–μm–pore size, 6.5-mm-diameter polycarbonate filters (Transwell, Sigma) and analyzed in triplicate by impedance spectroscopy using the cellZscope (nanoAnalytics, Muenster, Germany). TEER was monitored until it reached a plateau at day 5, and then both the upper and lower chambers of the Transwells were filled with medium containing TNFα (100 ng/ml) and/or hydrocortisone (100 ng/ml), and TEER was monitored for an additional 72 hours.

CP-specific knockdown of GR

TAT-CRE recombinase (Millipore) (40 μg per mouse, 20 μg in each ventricle) or PBS was injected into each of the lateral ventricles, similar to a method described elsewhere (39). Briefly, for each ventricle, a 10-μl Hamilton syringe was used to deliver a volume of 2.11 μl 0.4 mm posterior to the bregma, 1.0 mm lateral to the midline, and 2.0 mm in depth from the brain surface of anesthetized mice. Injection was carried out at a rate of 1 μl/min, after which the needle remained inside the brain for additional 4 min before being slowly removed. Mice were given a period of 2 weeks to recover from the procedure before further manipulations.

Statistical analysis

The specific tests used to analyze each set of experiments are indicated in the figure legends. Data were analyzed using a two-tailed Student’s t test to compare between two groups, and one-way analysis of variance (ANOVA) was used to compare several groups, followed by the Newman-Keuls post hoc procedure for pairwise comparison of groups after the null hypothesis was rejected. Sample sizes were chosen with adequate statistical power based on the literature and past experience, and mice were allocated to experimental groups according to age and genotype. Investigators were blinded to the identity of the groups during experiments and outcome assessment. All inclusion and exclusion criteria were preestablished according to the IACUC guidelines. Results are presented as mean ± SEM. In the graphs, y-axis error bars represent SEM. Statistical calculations were performed using GraphPad Prism software (GraphPad Software, San Diego, CA).


Supplementary material for this article is available at

Fig. S1. Decreased numbers of CD4 T cells in the brain following stress.

Fig. S2. Systemic depletion of FoxP3+ Tregs.

Fig. S3. FoxP3+ T cells in the HC of Treg-depleted stressed mice.

Fig. S4. Augmenting CP gateway activity in the context of stress does not affect BBB integrity.

Fig. S5. Infiltration of CD3+ cells to the HC in stressed mice treated with GR inhibitor.

Fig. S6. Exploratory behavior in an open-field arena in stressed mice treated with GR inhibitor.

Fig. S7. The effect of CORTs on the expression of trafficking-related molecules by human retinal pigment epithelial cell line.

Fig. S8. Gating strategy for flow cytometry analyses.

Fig. S9. Effect of acute CORT administration on CP gene expression.

Fig. S10. Meta-cell tSNE plot.

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: We thank G. Schütz and A. Chen for providing and assisting us with GR-floxed mice, O. Matcovitch-Natan for consulting on flow cytometry sorting techniques, and S. Schwarzbaum for editing the manuscript. Funding: This study was supported by Advanced European Research Council grants (232835), the Minerva Foundation, and an ISF Legacy BioMed grant (1353/15). A.D. is supported by Steven and Eden Romick. M.S. holds the Maurice and Ilse Katz Professorial Chair in Neuroimmunology. M.T. is the incumbent of the Carolito Stiftung Research Fellow Chair in Neurodegenerative Diseases. Author contributions: A.K. and K.B., in equal contribution and under the mentoring of M.S., conceived the general ideas of this study, performed all of the experiments, analyzed the data, and prepared the data for presentation. A.D., M.K., and T.C. assisted in experimental procedures. A.W. and I.A. assisted in analyzing data. I.C. assisted in transepithelial resistance recording experiments and subsequent data analysis. M.T. helped planning and assisted with all animal models. S.B.-H. assisted in intracerebroventricular injections. A.K., K.B., and M.S. wrote the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.
View Abstract

Navigate This Article