A) Dorsal laminectomy and B) diagram of the imaging window in the dorsal white matter of lumbar level 1 (L1) spinal cord in mice. C–D) Examples of eGFP-positive tight junction (TJ) protrusions, tortuosity, and gaps. E–L) Maximum intensity projections of two sequential time-lapse images of spinal cord venules in healthy (E,F, n = 6), pre-clinical (G,H, n = 5, EAE score 0), mild EAE (I,J, n = 6, score 1) and severe EAE (K,L, n = 6, scores 2–3) Tg eGFP-Claudin5^+/− mice at 0 and ~45-minute intervals. Yellow arrows show static TJ protrusions, red arrows point to dynamic TJ protrusions that appear or disappear during the imaging period, blue arrows show TJ gaps and magenta open arrows point to hematopoietic cells. (M–P) Bar graphs show percentages of TJ segments with protrusions (M), dynamic protrusions (N), gaps (O), and tortuosity (P). Each dot represents the mean of multiple fields from a single mouse. Bar graphs show mean ± SEM, *p<0.05, **p<0.01, ***p<0.001, one-way ANOVA. Scale bars are 20 μm (C,E–L) and 10 μm (D). See also .

A) Clinical EAE scores in wild-type (n = 34) and Cav1^−/− (n = 32) littermate mice. Cav1^−/− mice have less severe clinical scores at the indicated times [12–15 dpi, **p<0.01; ***p<0.001; two-way repeated measure ANOVA]. The graph excludes mice that died during the experiment (n = 0 wild-type, n = 2 Cav1^−/−). B) Scatter plot of maximum clinical scores attained by each mouse. Cav1^−/− mice had less severe scores (p<0.001, Mann-Whitney comparison of medians). C) Bar graph of disease incidence. D–E) Maximum intensity projections of time-lapse imaging recordings (45 minutes) for spinal cord venules in Tg eGFP-Claudin5^+/−, Cav1^−/− mice at EAE score 0 (n=5–6 group). Yellow arrows show static protrusions and red arrows point to dynamic protrusions that appear or disappear during the imaging period. Scale bars are 20 μm. F–I) Bar graphs show percentages of TJ segments with protrusions (F), dynamic protrusions (G), gaps (H), and tortuosity (I) between Tg eGFP-Claudin5^+/−, and Tg eGFP-Claudin5^+/−, Cav1^−/− mice. No differences were found for TJ structural features between Tg eGFP-Claudin5^+/− and Tg eGFP-Claudin5^+/−, Cav1^−/− mice. See also .

Biocytin-TMR was intravenously injected into Tg eGFP-Claudin5^+/− and Tg eGFP-Claudin5^+/−, Cav1^−/− mice prior to in vivo imaging with two-photon microscopy. (A–D) Healthy and CFA/pertussis toxin control Tg eGFP-Claudin-5^+/− mice had continuous TJ segments (A,C) and biocytin-TMR was confined within the lumen of spinal cord blood vessels (B,D). (E,F) Tg eGFP-Claudin5^+/− mice with EAE score 0 had some TJ gaps (E; blue arrowheads) and minimal biocytin-TMR leakage outside of blood vessels (F; yellow arrows). (G–J) Tg eGFP-Claudin5^+/− mice with EAE scores 1 (G–H) and 2 (I–J) had multiple TJ gaps (G,I; blue arrowheads) and extensive extravascular biocytin-TMR in the spinal cord (H,J; yellow arrows). (K–L) Tg eGFP-Claudin5^+/−, Cav-1^−/− mice with EAE score 1 had multiple TJ gaps (blue arrowheads) and extravascular biocytin-TMR. (M–N) Graphs of biocytin-TMR fluorescence intensity in the lumbar (M) and thoracic (N) spinal cord normalized to fluorescence intensity in the liver ex vivo (n = 4 CFA/pertussis toxin control, n = 7 EAE score 0, n = 6 EAE score 1, n = 8 EAE score 2–3). Scale bars are 20 μm. See also .

A–L) BBB leakage of intravenously injected albumin-AlexaFluor 594 (red) in lumbar spinal cord sections from healthy (A–B), CFA/pertussis toxin control (C–D), EAE scores 1 (E–F) and 3 (G–H) wild-type mice and healthy (I–J) and EAE score 1 (K–L) Cav1^−/− mice. The vascular marker Glut-1 (green) labels endothelial cells and DAPI (blue) shows nuclei. Yellow arrows indicate perivascular albumin. M,N) Bar graphs of the fraction of endothelium-associated (M) or parenchyma-associated (N) albumin that traversed the BSCB. Each dot represents the mean of multiple fields from a single animal (n = 5–6 healthy, n = 7 CFA/pertussis toxin control, n = 6 EAE score 0, n = 6 EAE score 1–1.5, n = 13–16 EAE scores 2–3, n = 3 healthy Cav1^−/−, n = 7 EAE score 1–1.5 Cav1^−/−). Bar graphs show mean ± SEM, *p<0.05, one-way ANOVA. Scale bars are 20 μm. See also .

A–F) Immunofluorescence for CD4, laminin and DAPI in the ventral (A,B,D,E) and lateral funiculus (C,F) of wild-type and Cav1^−/− EAE thoracic spinal cords 15 dpi. White arrows point to T cells that have crossed the basal lamina (white dotted line in A,D), whereas yellow arrowheads point to T cells that are located in meningeal or vascular spaces. G) Bar graphs of CD4⁺ T cell numbers in parenchymal or meningeal compartments of thoracic spinal cords in wild-type and Cav1^−/− EAE mice 15 dpi. There is no significant difference between genotypes (n = 9 wild-type; n = 9 Cav1^−/−; unpaired two-way Student’s t-test). Scale bars in (A,B,D,E) are 200 μm and (C,F) are 10 μm. H–K) FACS plots for mononuclear cells isolated from spinal cords of wild-type (n = 5) or Cav1^−/− (n = 6) mice 15 dpi. (L–P) Scatter plots of leukocytes gated on viable dye exclusion and high CD45 expression. There were no differences in percentages of infiltrating CD4⁺ CD45^hi (L) or CD8⁺ CD45^hi (M) between the two genotypes. N–O) Significantly fewer IFN-γ-producing CD4⁺ Th1 cells and significantly more IL-17-producing CD4⁺ Th17 cells were observed in spinal cords of Cav1^−/− compared to wild-type mice (n = 8 wild-type, n = 8 Cav1^−/− mice, *p<0.05, **p<0.01; Student’s t-test). P) There is no significant difference in the proportion of IL-17⁺ IFN-γ⁺ CD4⁺ T cells. Q) There was no difference in the fraction of infiltrating CD19⁺ B cells.

A) Schematic of T cell transmigration states in the in vitro assay. Wild-type Th1 or Rorc(γt)^GFP/+ GFP⁺ Th17 cells were applied to monolayers of wild-type or Cav1^−/− primary brain ECs (BECs). Based on their positions relative to endothelial ZO-1 in confocal image stacks, T cells were scored as either loosely adherent (non-transmigrating), or transmigrating via transcellular or paracellular routes. B–E) Apical and basal confocal images and orthogonal views of Th1 in vitro transmigration assay immunostained for ZO-1 (red), CD45 (green) and DAPI (blue). Loosely adherent Th1 cells are indicated with white arrows and transmigrating Th1 cells with yellow arrowheads on wild-type (B–C) or Cav1^−/− (D–E) BECs. F–I) Apical and basal confocal images and orthogonal views of Rorc(γt)^GFP/+ Th17 (eGFP⁺) in vitro transmigration assay immunostained for ZO-1 (red), GFP (green), CD45 (white) and DAPI (blue). Scale bar = 10 μm. J) Stacked bar graph indicating numbers of transcellular and paracellular transmigrating Th1 or Th17 cells across wild-type or Cav1^−/− BECs. Significantly fewer Th1 cells transmigrate across Cav1^−/− BECs via either transcellular (p<0.05) or paracellular routes (p<0.05) as compared with wild-type BECs. Significantly fewer Th1 cells transmigrate across Cav1^−/− BECs as compared with Th17 cells crossing Cav1^−/− BECs (p<0.05). K) Model for lymphocyte trafficking across the BBB in EAE. Th1 cells preferentially deploy caveolae to cross the BSCB (transcellular permeability). In contrast, Th17 cells efficiently cross the BSCB through disrupted TJs (paracellular permeability) as well as caveolae. TJ remodeling involves formation of membrane invaginations (protrusions) that fuse with EEA1⁺ endosomes independently of caveolae. See also .