Fig 2: S1P3 overexpression in the astrocytic microenvironment of highly permeable brain metastases of breast cancer. a S1P3 is expressed by a proportion of activated astrocytes. Brain sections containing 231-BR experimental brain metastases were stained for S1P3 (green), GFAP+-activated astrocytes in the neuroinflammatory response (red), CD31 endothelial cells to visualize the location of astrocyte end feet (gray). All nuclei are stained with DAPI (blue). Right panel, merge. Scale bar: 50 µm. b–e Representative TRD permeability, S1P3, and GFAP staining in brain metastasis sections from four model systems (left); ratio of normalized astrocytic S1P3 in highly permeable/poorly permeable metastases (right). For each model system, distinctly permeable and poorly permeable metastases to TRD (red) were identified (scale bar = 100 µm); successive sections were stained for S1P3 (green) and GFAP (red) (scale bar = 50 µm). For quantification, S1P3 staining area was normalized to GFAP staining area. Staining in all distinctly highly permeable metastases/poorly permeable metastases in a single mouse brain was calculated (dot). The red line at unity would indicate no trend in highly/poorly permeable metastases. b Independent 231-BR cohort. Astrocytic S1P3 in highly/poorly permeable metastases, median = 1.73 (n = 12). c HER2+ JIMT-1-BR brain metastases4. Astrocytic S1P3 in highly/poorly permeable metastases, median = 1.36 (n = 10). d HER2+ inflammatory breast cancer SUM190-BR metastases19. Astrocytic S1P3 in highly/poorly permeable metastases, median = 1.80 (n = 8). e 4T1-BR brain metastases. Astrocytic S1P3 in highly/poorly permeable metastases, median = 1.59 (n = 8). From (b) to (e): Graphs represent the median, the interquartile range (box), and the min/max values (error bars). Wilcoxon matched-pairs signed rank test, two-tailed, *P < 0.05

Fig 3: Knockdown of astrocytic S1PR3 strengthened the BBB and BTB in vitro. Immortalized human astrocytes were transduced with either control shRNA (shControl-1, shControl-2) or shS1PR3 (shS1PR3-1, shS1PR3-2), and used in TEER and permeability assays of the BBB and BTB (see legend to Fig. 5). a RT-PCR for S1PR3 mRNA (top) and immunofluorescent staining (bottom) for S1P3 protein (green). Nuclei are DAPI stained (blue) (scale bar = 50 µm). b TEER assays using either shControl-1, -2 or shS1PR3-1, -2 astrocytes in the cultures. S1PR3 knockdown elevated TEER at 24 h, indicative of a tighter BBB and BTB, (n = 6). c Permeability of 100 µg mL-1 doxorubicin from luminal to abluminal side of the inserts in the BBB and BTB, (n = 6). Astrocytic S1P3 knockdown (using shS1PR3-1 and shS1PR3-2) reduced permeability. d Staining of endothelial cell monolayers in TEER assays for VE-cadherin (red) demonstrating increased membrane staining (white arrows) when shS1PR3-1, -2 astrocytes were included. Nuclei were DAPI stained (blue) (scale bars = 10 µm for set-1 and 20 µm for set-2). For (b) and (c), Mann–Whitney statistical analysis was performed. Graphs represent the median and the error bars the 95% confidence intervals, from independent transwell BBB/BTB cultures (n) from multiple experiments. **P < 0.01, ***P < 0.001

Fig 4: S1P3 is expressed in the neuroinflammatory response of patients’ brain metastasis specimens. Fresh frozen sections of nine resected brain metastases of breast and lung cancers, nine of which were previously described19, were stained for S1P3 (green) and GFAP (red). Nuclei are DAPI stained (blue). a Specimen #4, ER-, PR-, HER2+ breast cancer. b Specimen #5, ER-, PR-, HER2+ breast cancer. Co-staining of S1P3 and activated astrocytes in the tumor microenvironment is seen. A summary of all staining is presented in Supplementary Data 3. (Scale bar = 50 µm)

Fig 5: S1P3 antagonists strengthen the blood–tumor barrier in vitro. a An in vitro model of the BBB was constructed with confluent brain endothelial cells on the luminal side of a coated porous membrane, brain pericytes on the abluminal side of the membrane, and astrocytes in the bottom of the abluminal side of the chamber. To create a blood–tumor barrier (BTB) model, spheres of 231-BR tumor cells were added to the bottom of the chambers. Barrier integrity was quantified by transendothelial electrical resistance (TEER), or diffusion of a drug (doxorubicin) from the luminal to abluminal side. Cells were cultured to confluence for 4 days, at which time antagonists (b) (EX26: 1 nM, JTE-013: 17 µM, TY: 2 µM, CAY10444: 10 µM, and CYM50358: 25 nM) or vehicle were added. In BBBs and BTBs, both S1P3 antagonists selectively showed increased TEER, indicative of a tighter BBB (c) and BTB (d) (n = 5). TY-52156 and CAY10444 decreased permeability of 100 µg mL-1 doxorubicin significantly in BBB (e) and BTB (f) models (n = 5). g After 24 h post-treatment with TY-52156 and CAY10444, the membranes containing brain endothelial cells were removed and stained for VE-cadherin (red) (white arrows). Nuclei were DAPI stained (blue). Increased staining and membranous localization was observed with TY-52156 and CAY10444 (scale bar = 20 µm). h The effect of vehicle, 2 µM TY-52156 or 10 µM CAY10444 on BBB TEER in cultures with or without astrocytes. Significant TEER elevation by TY-52156 and CAY10444 was dependent on the presence of astrocytes (n = 4). For (c–f) and (h), Kruskal–Wallis test and Dunn’s multiple comparison test were performed. Graphs represent the median and the error bars the 95% confidence interval, from independent transwell BBB/BTB cultures (n) from multiple experiments. *P < 0.05, **P < 0.01, ns non-significant