In vitro analysis of the multistep cascade of T-cell migration across the BBB model under physiological flow. A. In vitro under-flow assay setup. T cells were perfused on top of the pMBMEC monolayer, and their migration under flow was observed using phase-contrast imaging modality. Imaging was performed with a time step of 10 seconds/timeframe. B. In vitro flow assay timeline. During the accumulation phase under flow with the shear stress of 0.1 dynes/cm2, T cells adhered to the pMBMEC monolayer. After 5 minutes (timeframe 30), the shear stress was increased to 1.5 dynes/cm2, leading to rapid detachment of not firmly adhering T cells. Analysis of the post-arrest T-cell behavior was thus starting at 5.5 minutes (timeframe 33). tf = timeframe. C. Example of phase-contrast imaging data. Red arrow . crawling T cell; yellow arrow . fully transmigrated T cell; yellow arrow-head . transmigrated part of a partially transmigrated T cell; red V arrowheads . pMBMECs. D. Schematic representation of distinct T-cell behavior regimes detected and analyzed using the developed UFMTrack framework. Crawling cells migrate continuously while probing cells interact with the pMBMECs and move around the interaction point within two cell-size (20 ....) as indicated by the arrows. Side and top views are shown.

Example of data used for training of the T-cell segmentation model. From left to right: Phase-contrast microscopy image of T cells migrating on top of the pMBMEC monolayer; annotated T cell mask; annotated mask of transmigrated part of T cells; T cell centroids; wcell weight map. The bottom row shows zoom-in on the highlighted area.

Comparison of the performance of the 2D and 2D+T-cell segmentation models.

The T-cell tracking pipeline. A. Segmentation and centroids of two migrating T cells under flow over time. B. Nodes corresponding to segmented T cells. Node sizes and colors denote the multiplicity estimate of a node. C. Nodes, together with links between adjacent in space and time nodes, form a graph. D. Selected links are obtained with global optimization on the graph. E. Graph of segments and vertices obtained according to node multiplicity. F. Resolving global segment multiplicity consistency with global optimization on the graph. Additionally, a search for rapid displacement of segmented T cells due to under-segmentation or flow is performed. G. Extension of tracks of individual T cells into the track intersection. H. Resolving track intersections.

Analysis of CD4+ T-cell tracks. A. Start and end time distribution of CD4+ T-cell tracks on non-stimulated or TNF or IL-1β stimulated pMBMECs. More T cells detach from the non-stimulated pMBMECs after 5 minutes when the flow is increased to physiological levels. B. Quantification of CD4+ T-cell behavior in the respective categories obtained on non-stimulated and TNF and IL-1β stimulated pMBMECs. Error bars show the statistical error of the mean. (see text for details). C-H. Motility parameters of the crawling CD4+ T cells were obtained for the three endothelial stimulatory conditions. Distributions of T-cell path length (C), displacement (D), crawling speed (path/time) (E), migration speed (displacement/time) (F), variability of instantaneous T-cell crawling speed along the track (standard deviation, G), and meandering index (H). I. Distribution of CD4+ T-cell accelerated movement (AM) speed is a proxy metric for the T-cell adhesion to the healthy or inflamed endothelium. J. Migration speed distribution of the transmigrated CD4+ T cells. Stimulation applied to the luminal side of pMBMECs affects T-cell migration at the abluminal side of pMBMECs after their transmigration. FT - T cells performed full transmigration; UT – T cells performed uncompleted transmigration.

Comparison of automated analysis with UFMTrack and manual analysis of the CD8+ T-cell tracks. A-C: CD8+ T-cell behavior statistic obtained for non-stimulated (NS) and cytokine-stimulated pMBMECs as obtained manually by four experimenters, as well as automatically with UFMTrack. A. Quantification of CD8+ T-cell behavior in the respective categories obtained on non-stimulated and TNF/IFN-γ stimulated pMBMECs is consistent with results obtained by manual frame-by-frame analysis. Cytokine stimulation of pMBMECs increases T-cell probing behavior (B) and T-cell transmigration rate (C). Error bars show the standard deviation of the manual analysis and the statistical error of the mean for automated analysis. Points correspond to individual experimenters. D. Counts of CD8+ T cells were obtained manually by four experimenters and automatically by UFMTrack. The T-cell detection efficiency is above 90%. Error bars show the standard deviation of the manual analysis and the statistical error of the mean for the automated analysis. Points correspond to individual experimenters. E. Detection efficiency of crawling CD8+ T cell tracks between the manual and automated analysis. F-I. Comparison of CD8+ T-cell crawling speed (path/time) (F, G) and migration speed (displacement/time) (H, I) on non-stimulated (F, H) and cytokine-stimulated (G, I) pMBMECs. The T-cell position assignment error in manual tracking leads to biased crawling speed estimation. J. Comparison of the analysis time (per cell) required for behavior analysis and tracking of CD8+ T cells. K. Total analysis time (per dataset) for behavior analysis and tracking of CD8+ T cells. Comparison is shown for manual analysis with tracking of crawling cells only (Crawling manual), the time estimate for manual analysis with tracking of all cells (All manual), and the in-depth automated analysis of all cell tracks with UFMTrack (All auto).

FT - T cells performed full transmigration; UT – T cells performed uncompleted transmigration; NT - T cells did not perform transmigration.

Analysis of different BBB models with UFMTrack. A-C. Human PBMCs interacting with HBMEC. A. Raw image frame, segmentation, and tracking results. B. PBMCs behavior distribution. C. Crawling and migration speed distribution. D-F. BMDM interacting with the pMBMEC monolayer. D. Raw image frame, segmentation, and tracking results. E. BMDM crawling vs probing behavior distribution. F. Crawling and migration speed distribution. G-K. Human T cells interacting with the EECM BMEC monolayer. G. Raw image frame, segmentation, and tracking results. H. Crawling and migration speed distributions for the control condition (ctrl) and with the blocking anti-bodies (bab). I. X and Y relative cell displacement. J. Crawling vs probing behavior comparison. K. Transmigrations behavior comparison. L-O. Human T cells interacting with immobilized recombinant BBB adhesion molecules. L. Raw fluorescent image frame, thresholding-based segmentation, and tracking results. M. Crawling vs probing behavior distribution. N. X and Y relative displacement O. Crawling vs probing speed comparison.

Scale bar 50 µm. Statistical tests are performed with the Mann-Whitney U test.