Integrative analysis of T cell motility from multi-channel microscopy data using TIAM

Abstract

Integrative analytical approaches are needed to study and understand T cell motility as it is a highly coordinated and complex process. Several computational algorithms and tools are available to track motile cells in time-lapse microscopy images. In contrast, there has only been limited effort towards the development of tools that take advantage of multi-channel microscopy data and facilitate integrative analysis of cell-motility. We have implemented algorithms for detecting, tracking, and analyzing cell motility from multi-channel time-lapse microscopy data. We have integrated these into a MATLAB-based toolset we call TIAM (Tool for Integrative Analysis of Motility). The cells are detected by a hybrid approach involving edge detection and Hough transforms from transmitted light images. Cells are tracked using a modified nearest-neighbor association followed by an optimization routine to join shorter segments. Cell positions are used to perform local segmentation for extracting features from transmitted light, reflection and fluorescence channels and associating them with cells and cell-tracks to facilitate integrative analysis. We found that TIAM accurately captures the motility behavior of T cells and performed better than DYNAMIK, Icy, Imaris, and Volocity in detecting and tracking motile T cells. Extraction of cell-associated features from reflection and fluorescence channels was also accurate with less than 10% median error in measurements. Finally, we obtained novel insights into T cell motility that were critically dependent on the unique capabilities of TIAM. We found that 1) the CD45RO subset of human CD8 T cells moved faster and exhibited an increased propensity to attach to the substratum during CCL21-driven chemokinesis when compared to the CD45RA subset; and 2) attachment area and arrest coefficient during antigen-induced motility of the CD45A subset is correlated with surface density of integrin LFA1 at the contact.

Keywords: Integrative analysis; Multi-channel microscopy; T cell motility; Tracking.

Fig. 1

Overview of the schema for data integration in TIAM. Transmitted light images are used for detecting and tracking cells. Several parameters quantifying the motility characteristics are calculated and stored in MATLAB  ‘cell arrays’. Individual tracks are considered for extracting information from reflection and fluorescence images that are part of multi-channel time-lapse data. Centroids from track positions are used for local segmentation and outlining that would correspond to the cell under consideration. Features are computed from the outlined regions and stored along with rest of the track-related information.

Fig. 2

Detection and tracking of cells by TIAM. TIAM uses transmitted light images for detecting and tracking cells. Illustration of detection by TIAM is provided with an example (a–d). A DIC image of human primary CD8 T cells is used (a). The panels, b to d, represent sequential stages during cell detection. In the first step, the Canny edge filter is applied to generate a binary image of cell boundaries (b). Then, a circular Hough transform (CHT) is applied to this binary image. This operation maps cell outlines to points in a parameter space based on a voting scheme (c). Local maxima in the parameter space are used to pick centroids of cells (d). TIAM has a graphical user interface that walks the user through the choice of parameters for edge filtering and CHT to allow for accurate detection of cells.

Fig. 3

Evaluation of performance of tracking T cells by TIAM. a) Tracks of cells obtained after manually establishing the ground truth (in green) are overlaid on tracks of cells obtained from TIAM (in red). The overlap between the tracks is shown in yellow. These correspond to frames 11–40 of Experiment 2 (Table 1). b) ATA values at different thresholds for the nearest neighbor association (parameter r) in both experiments. ATA values suggest that tracking results are relatively robust to changes in the threshold value for the nearest neighbor association, a critical parameter in the tracking algorithm. Thresholded ATA values are plotted here. Jaccard Similarity of 0.4 or more is considered as 1 (see Supplementary methods) during the calculation of thresholded ATA. This is done to ensure that minor localization inaccuracy is not penalized.

Fig. 4

Evaluation of performance of extracting features from DIC (a), reflection (b) and fluorescence images (c). Aspect ratio (readout of morphological polarity), contact area, and mean fluorescence intensity were measured from DIC, reflection and fluorescence channels, respectively. Outlines were drawn along cell-boundaries in either a manual or semi-automated manner using ImageJ to establish the ground truth for respective channels. Performance of extracting features was evaluated by quantitative comparisons with the ground truth after establishing one-to-one pairing between TIAM results and the respective ground truth. The measured values for each pair are plotted: 1389 for DIC, 4005 for reflection and 5973 for fluorescence. Overall, the data hovered around the diagonal line implying reasonable accuracy for measurement of polarity from DIC and good accuracy for measurement of contact area and fluorescence intensity from reflection and fluorescence channels, respectively.

Fig. 5

Examples of integrated analysis of human CD8 T cell motility enabled by TIAM. a) Effects of pharmacological inhibitors of PKCs on CCL21 driven chemokinesis in CD45RA+ve cells. Population median values of different motility characteristics from mean values of individual cell tracks were calculated first. These have been normalized to the median motility characteristics in the ‘control’ data and shown in a colored heat map. Statistical significance of differences in the population was calculated. Unless specified otherwise in the heat map cell, p-value was below 0.0001. b) Average speed and average contact area of CD45RA+ve and CD45RO+ve cells subjected to CCL21 driven chemokinesis is shown. The number of tracks that had reflection footprint out of the total tracked cells is given for both subsets. Even when the reflection footprint existed in a portion of the track, it was counted as a cell track with attachment. The motility experiments were conducted with a mixed population wherein the subsets were isolated, loaded with different vital dyes and then mixed in equal ratio. Statistical significance was assessed by Mann–Whitney U-test in both (a) and (b). c) Average surface density of LFA1 (measured by binding of Alexa Fluor 488 labeled Fab fragment of TS2/4 non-blocking antibody) is plotted against average contact area and arrest coefficient for individual CD45RA+ve cell tracks. Pearson correlation coefficient values are shown at the top of the plots. Arrest coefficient was calculated based on a threshold instantaneous speed of 0.5 μm/min. All results are representative of two or more independent sets of experiments.