Detailed statistical comparison of interleukin-1β-treated and untreated platelets within a clot.

Cells were labelled with Alexa647-labelled anti-CD62p-antibody. Images were taken at the edge of an artificial clot, where platelets are more sparsely distributed than within the clot. Therefore, segmentation (ROI extraction) is required. Here, we present a statistical comparative analysis of the full image, with sparsely distributed platelets and we compare the extracted regions. The full (not regionalized) sample analysis is depicted in the left column of the Fig (a’-e’), the ROI analysis on the right ((a-e); parts of the ROI analysis are included in Fig 3 in the main text). From the analysis, the advantage of segmentation (region extraction) is clearly visible. The first level analysis of the comparison of the KS/WX- tests presented in (a) (and Fig 3E) and (a’) indicates a significant similarity for full images. For ROI-based comparison, density and curvature of the clusters shows similarity solely for cluster dimensions >700 nm (sim_M/sim_L = 0.47/0.31 and 0.45/0.24 for cluster density/curvature distributions, in contrast: sim_M/sim_L = 0.87/0.75 and 0.87/0.80 for a full sample). Fig 3C (main text) depicts two extracted regions of an untreated (upper) and IL-1β treated (lower) clot; a difference in the CD62p distribution can be observed. The divergence of the results between ROI and full image analysis is due to numerous small clusters, which are homogenously distributed in the images; in particular in the platelet-free area of the interleukin-1β treated clot. Not-cell-related signals add a significant additional cluster population, which is less significant within a ROI. For full image analysis, clusters for a specific cluster dimension originate from various regions due to hierarchical clustering, which influences the overall cluster density/curvature distribution. This is not the case for ROI-based analysis. Similar results can be observed for the second level analysis: the mean-cross comparison of the statistical data on cluster density and curvature show divergent results. As before, a dissimilarity for the ROI comparison is shown (sim_M/sim_L = 0.07/0.07 (Fig 3F) and 0.91/0.50 (b, c) for aggregated cluster density/curvature comparison, in contrast: sim_M/sim_L = 0.54/0.2 and 0.85/0.35 for cluster density/curvature comparison on a full sample (c’)). The results are confirmed by the K-/H-Ripley’s statistics. The mean-cross analysis of all values, shown in (d, d’), indicates that many platelets within the extracted clot regions are sparsely distributed. Again, solely K-/H-Ripley statistics determined for the extracted ROI’s indicate a difference in CD62p protein secretion after interleukin 1β treatment (sim_M/sim_L = 0.03/0.01 and 0.04/0.01 for K-/H-Ripley comparison (ROI analysis), in contrast: sim_M/sim_L = 0.94/0.78 and 0.94/0.79 for K-/H-Ripley comparison on a full sample).

(TIF)