Cortisol promotes breast‐to‐brain metastasis through the blood‐cerebrospinal fluid barrier

Abstract Background Elevated basal cortisol levels are present in women with primary and metastatic breast cancer. Although cortisol's potential role in breast‐to‐brain metastasis has yet to be sufficiently studied, prior evidence indicates that it functions as a double‐edged sword—cortisol induces breast cancer metastasis in vivo, but strengthens the blood‐brain‐barrier (BBB) to protect the brain from microbes and peripheral immune cells. Aims In this study, we provide a novel examination on whether cortisol's role in tumor invasiveness eclipses its supporting role in strengthening the CNS barriers. We expanded our study to include the blood‐cerebrospinal fluid barrier (BCSFB), an underexamined site of tumor entry. Methods and Results Utilizing in vitro BBB and BCSFB models to measure barrier strength in the presence of hydrocortisone (HC). We established that lung tumor cells migrate through both CNS barriers equally while breast tumors cells preferentially migrate through the BCSFB. Furthermore, HC treatment increased breast‐to‐brain metastases (BBM) but not primary breast tumor migratory capacity. When examining the transmigration of breast cancer cells across the BCSFB, we demonstrate that HC induces increased traversal of BBM but not primary breast cancer. We provide evidence that HC increases tightness of the BCSFB akin to the BBB by upregulating claudin‐5, a tight junction protein formerly acknowledged as exclusive to the BBB. Conclusion Our findings indicate, for the first time that increased cortisol levels facilitate breast‐to‐brain metastasis through the BCSFB—a vulnerable point of entry which has been typically overlooked in brain metastasis. Our study suggests cortisol plays a pro‐metastatic role in breast‐to‐brain metastasis and thus caution is needed when using glucocorticoids to treat breast cancer patients.


| INTRODUCTION
Metastatic brain tumors bear a dismal 6-month median survival time, constituting greater than 90% of all brain tumors. 1 The most common primary tumors sites are lung (50%), followed by breast (20%) which together comprise roughly two-thirds of cases. Unfortunately, despite the high frequency of metastatic brain tumors, there remains no universally accepted paradigm for chemotherapy treatment. 2 Conventional chemotherapy has historically played a muted role in brain metastasis management due to inherent barrier properties of the central nervous system (CNS). 3 Furthermore, we and others have shown that the brain's microenvironment has a promoting effect on metastatic tumors. [4][5][6][7][8][9][10][11][12] The CNS is tightly separated from the dynamic milieu of blood by the blood-brain-barrier (BBB) and the blood-cerebrospinal fluid barrier (BCSFB). The BBB is a highly specialized structure connected by the interactions of vascular brain endothelial cells (BECs), an ensheathed basement membrane housing pericytes, and projecting astrocytic foot processes. Unlike peripheral endothelial cells, BECs are nonfenestrated and inhibit the untethered paracellular diffusion of water-soluble molecules by an interconnected network of tight junctions (TJs). 13,14 Analogous to the endothelial barrier, the BCSFB correlate are apical TJs expressed within choroid plexus epithelial cells (CPEs) preventing the free flow of water-soluble molecules. These choroid plexus cells have a supplemental secretory functionproduction of cerebrospinal fluid (CSF) which extravasates into the brain's ventricles and is dispersed throughout the CNS.
Tumor colonization into the brain parenchyma requires successful breaching of these CNS barriers. Prior in vitro transendothelial/epithelial electrical resistance (TEER) assays have demonstrated that the BBB (80-100 Ω × cm 2 ) is stronger and less permeable than the BCSFB (30-40 Ω × cm 2 ). 15,16 This discrepancy can be attributed to their roles in CNS homeostasis regulation; the BBB's primary function is to inhibit all paracellular diffusion to protect the brain from invading pathogens whereas the BCSFB requires increased permeability to support water intake into the CSF. 17,18 The BBB and BCSFB differ in their TJ architecture, employing different occludin, claudin, and junctional adhesion molecules. 19 Claudin-5 is the most abundant and primary TJ protein for barrier formation in the BBB whereas the BCSFB correlate is claudin-1 and claudin-3. 17,20,21 Recently, expression of claudin-5 in CPEs was observed which previously was reported to be specific to BBB. 22 Cortisol, a glucocorticoid steroid hormone, targets BBB endothelial cells and upregulates claudin-5 expression generating a tighter barrier. 23 Interestingly, elevated basal cortisol levels are present in women with early stage breast cancer (0.49 μmol/L) and metastatic breast cancers (0.70 μmol/L) when compared to healthy women (0.29 μmol/L). However, cortisol exhibits a stark dualityglucocorticoids such as dexamethasone induce breast cancer metastasis in vivo, 24 albeit with no mention to brain metastasis.
Our study determines whether elevated levels of cortisol, induced by tumor formation, will effect brain metastasis by reinforcing the tight junctions of the BBB and BCSFB. HCMECs were grown on plastic while CPEs, astrocytes, and pericytes were grown on collagen I, rat tail (Thermo Fisher Scientific, Waltham, Massachusetts) coated flasks. All cell lines were grown and maintained in a humidified incubator at 37 C and 5% CO 2 . All experiments were performed in triplicates, all data are expressed as the mean ± SD. All cell lines have been authenticated using STR profiling and that all experiments were performed with mycoplasma-free cells should be included. All patient-derived cell lines were obtained through approved University of Southern California Institutional Review Board consent protocol.

| BBB and BCSFB in vitro model
Twelve-well plate transwell inserts (665 635, Greiner Bio-One, Monroe, North Carolina) with 8.0-μm 2 pores were used to measure barrier integrity of the BBB and BCSFB in vitro. The transwell inserts were inverted and placed into a six-well plate where the exterior side of the transwell was coated with rat tail collagen I. For BBB assays, the goal was to culture HCMECs on top of the transwell with astrocyte pericyte mixture on the bottom. In order to achieve this, 1:1 ratio of 100 000 astrocytes and pericytes were seeded onto the exterior side of the inverted transwells. After 24 hours, transwells were reinverted and 150 000 HCMECs in 1 mL of medium were seeded into the top chamber. BBB models were used 5 days after this setup. For BCSFB assays, 80 000 CPE cells were seeded similarly onto inverted transwells. Cells were seeded with 300 μL of their appropriate culture medium. Lids were placed onto the six-well plates making contact with the 300 μL of medium on the inverted transwells which prevents evaporation.
The cells were allowed to settle and adhere to the transwells overnight. The following day the transwells were reinverted to their normal positioning and placed into a 12-well plate containing 1 mL of their appropriate medium. A total of 1 mL of medium without cells was placed into the top chamber of the BCSFB transwells. BCSFB models were used 5 days after this setup. Additionally, for co-culture experiments with tumor cells, breast or lung cancer cells were seeded at 150000 cells on the bottom of transwell. where it was centrifuged at 5000 × g for 20 minutes. The filtrate was discarded and the retentate (400 μL) was used for treatments or stored at -80 C for later use.

| TEER assay
Transwell cultures were replenished with fresh medium after 2 days of initial seeding. The in vitro barrier TEER was measured in Ω × cm 2 twice in a day according to the manufactures protocol until it reached the maximum resistance potential according to the manufacturers protocol of MilliCell ERS-2 (MilliporeSigma, Burlington, Massachusetts).
Confluency was observed when resistance peaked and leveled off (>40 Ω × cm 2 for BCSFB, and >70 Ω × cm 2 for BBB) which occurred after 4 days of initial seeding.  Later, the transwells were gently washed twice with 1× PBS and placed into a new 12-well plate containing 1 mL of 1% paraformaldehyde and allowed to fix for 10 minutes. After fixation, the transwells were further washed twice in 1× PBS and allowed to air-dry for 3 minutes. The transwell mesh was cut out of the insert using a razor and placed onto a microscope slide with the cells facing upward. A total of 10 μL of Pro-Long Gold Antifade Mountant with DAPI (Thermo Fisher Scientific, Waltham, Massachusetts) was added to the fixed sample, and a 22-mm square cover glass was placed onto the sample. Clear nail polish was used to seal the cover glass. Seven fields from each mesh filter were imaged using a confocal microscope at 20× magnification. Images were acquired using only the GFP channel. GFP positive cells were counted and averaged among the seven fields. All fluorescent imaging was done in the Cell and Tissue Imaging Core of the USC Research Center for Liver Diseases.

| Immunocytochemisty
To verify tight junction expression, CPE cells or HCEMCs were stained

| Hydrocortisone and tumor conditioned medium influence barrier property of the BBB and BCSFB
Hydrocortisone increases barrier tightness of the BBB, 23 but its effect on the BCSFB remains to be investigated. Therefore, we now assess whether hydrocortisone can preserve the functional barrier integrity of the BBB and BCSFB in the presence of tumor conditioned medium.
Our results show that at baseline, BCSFB was significantly (P < .001) more permeable than the BBB. Treatments with hydrocortisone showed decreased leakiness in both BBB and BCSFB. However, in the presence of HC the BCSFB was still significantly leakier (P < .05) than BBB ( Figure 1A). Further, we examined leakiness of both barriers with tumor conditioned medium (CM). In the absence of HC treatment, the barrier performance of BBB co-cultured with tumor cell conditioned medium was found to be preserved. However, a significantly compromised barrier function was observed with BCSFB when co-cultured with CM MDA-MB-231 (P < .01) and CM BBM 3.1 (P < .05; Figure S1A). In contrast, BCSFB exhibited a minor change in permeability with CM A549 or CM LuBM5 ( Figure S1A).
The in vitro brain barriers displayed significant reduction in permeability when treated with HC and various cancer cell condition media (CM; Figure 1B). Barrier function of the BBB treated with HC was stable in all tumor conditioned medium ( Figure S1B). Similarly, BCSFB was able to retain its barrier property in co-culture with CM A549 or CM LuBM5 medium, however, loss of barrier function was evident with CM MDA-MB-231 (308.05 ± 85.32 FL, P < .05) co-culture compared to the control. Additionally, the HC-treated BCSFB along with CM BBM3.1 showed a similar tendency of increasing permeability with no cogent difference (ns, P > .05) compared to the BBB ( Figure S1B).

F I G U R E 4
Breast tumor co-cultures do not decrease tight junction gene expression. Bar graphs are fold changes relative to 0 nM hydrocortisone (HC) CP only. RPLPO gene expression was used to normalize data. For multiple group analysis, one-way analysis of variance (ANOVA) with Bonferroni tests was used followed by statistical significance. *P < .05, **P < .01, ***P < .001, ****P < .0001. ns, not significant. Significance test was relative to CP only control within 0 nM HC and 550 nM HC F I G U R E 5 Breast tumor co-cultures decrease claudin-5 protein expression in the presence of hydrocortisone (HC). Fluorescent intensity threshold was determined by adjusting the fluorescent intensity until no signal was detected on negative primary antibody controls. These settings were used for imaging corresponding positive antibody samples (claudin-5 or ZO-1). All the treatments were for 48 hours. A,B, 0 nM HC, C,D, 550 nM HC. ICC stains includes DAPI (blue) Claudin-5 (red), ZO-1 (green), and Phalloidin (white). Images 40×

| Breast tumor environments degrade BCSFB tight junctions on CPE cells
We next determined mRNA expression profiles of BCSFB tight junction proteins in tumor conditioned environment in the presence and absence of HC. qPCR results on CPE cells with HC treatment show increase of Claudin-5 (P < .01) and ZO-1 P < .05) expression and significant downregulation of Occludin (P < .001), Claudin-1 (P < .05) and Claudin-3 (P < .05) ( Figure 3A). The protein expression levels for Claudin-5 and ZO-1 were then confirmed ( Figure 3B,C). Results show, similarly to mRNA levels, CPE ZO-1 and Claudin-5 have increased expression with HC treatment relative to controls. Previous F I G U R E 6 Hydrocortisone (HC) increases migration independent of barrier resistance in LuBM5 and BBM3.1. A-D, GFP positive tumor transmigration through blank transwells ±550 nM HC for 12 hours. Graphs are quantifications of 0 nM HC vs 550 nM HC migrations, A, MDA-MB-231-GFP, B, BBM 3.1 (DAPI was used since GFP signal was very low the first 12 hours of initial BBM 3.1 migration), C, A549-GFP, and D, LuBM5-GFP. Seven fields were captured for all the images and cells were averaged among the fields. The unpaired Student's t test (two tailed) was used to detect statistically significant differences. *P < .05, **P < .01, ***P < .001, ****P < .0001. ns, not significant. Images 40× permeability and transmigration assays demonstrated that BCSFB breakdown was susceptible to MDA-MB-231 and BBM 3.1 cells.
Therefore, we evaluated mRNA expression profiles of various TJ proteins in presence and absence of HC co-cultured with tumor cell lines.
Results show there was a significant increase in Occludin and Claudin-1 in CPE cultures relative to co-cultures of CPE and tumor cells. However, when HC was added, there was no significant change in TJ proteins in CPE cultures relative to co-cultures of CPE and tumor cells ( Figure 4). Furthermore, in the absence of HC, Claudin-5 and ZO-1 protein expression remained consistent with tumor co-cultures ( Figure 5A,B). However, in the presence of HC, a decrease in Claudin-5 expression was observed ( Figure 5C). No change in ZO-1 TJ protein across both the co-cultures was visible ( Figure 5D).

| Hydrocortisone promotes tumor migration
We analyzed whether HC has any effect on tumor migration independent of barrier resistance. MDA-MB-231 cells show no significant difference in migration rates between untreated and HC-treated samples ( Figure 6A). However, a significant increase in BBM 3.1 cell migration (P < .0001) was observed when treated with HC ( Figure 6B). No difference was observed in A549 cell migration between untreated and HC-treated samples ( Figure 6C). Additionally, HC increased LuBM5 cell migration significantly (P < .001; Figure 6D).
Overall, these results suggest HC induce tumor cell migration.

| DISCUSSION
The glucocorticoid steroid, cortisol, increases in primary and metastatic breast cancer patients. 25 Studies attempting to elucidate its mechanistic role have demonstrated a conflicting duality-cortisol promotes metastases systemically only in breast cancer, with no observation of brain metastasis . 24 Additionally, the role of hydrocortisone was found to reinforces the blood-brain-barrier properties. 26  Prior studies have demonstrated that the BCSFB is weaker and more permeable than the BBB. 15  Surprisingly, we observe patient-derived breast-to-brain metastasis have increased capacity to traverse the BCSFB even in the presence of HC; while migration of primary breast cancer cells across the BCSFB was decreased. This initially puzzling finding is provided greater clarity when observing migration rates independent of barrier resistance. Our findings demonstrate that HC aids breast-to-brain metastases cell migration significantly, but primary breast cancers remains unaffected. Although MDA-MB-231, MDA-MB-231BR, and BBM 3.1 are all triple negative breast tumors, the latter two have formerly colonized the brain parenchyma and have likely developed the critical mechanisms that have allowed them to traverse the CNS' barriers. This leads us to speculate that MDA-MB-231 migration is unaided by HC, yet HC functions to strengthen the BCSFB barrier which thwarts barrier traversal as compared to an absence of HC. Consequently, we postulate that high cortisol levels are the basal conditions for breast-to-brain metastases. Validating this assertion would require a comparison of the levels of glucocorticoid receptors across primary breast and secondary breast-to-brain tumors.
Examining GR expression may hold a key to elucidating breast-tobrain metastases evolution.
Another unexpected finding was that HC increased migration rates, independent of barrier resistance, in LuBM5 but not A549 cells.
A cortisol dependent migration mechanism may potentially exist in lung-to-brain metastases. Our study provides some evidence when observing the migration of LuBM5 cells and A549 across the CNS barriers in the presence of HC. While HC fails to significantly curtail migration of LuBM5 cells, it impedes A549 cell migration across both the BBB and BCSFB. The relationship between cortisol and lung tumor metastases has yet to be investigated.
We initially theorized that elevated cortisol levels would decrease breast-to-brain metastasis by reinforcing the CNS barrier properties.
Although HC decreased migration through the BBB, metastatic cell migration through the BCSFB was sharply higher. In conclusion, we demonstrate that cortisol facilitates breast-to-brain metastasis by inducing a more invasive tumor phenotype and breaching the cortisol strengthened BCSFB. Acknowledging the common use of cortisol in the clinic to treat cancer patients, 27 encourages the scientific community to advance mechanistic studies on cortisol and tumor progression.