Figure 1.
Zebrafish experimental setup and neutrophil analysis procedure.
A: Zebrafish larva from the transgenic line, Tg(mpx:GFP)i114. Neutrophils are visualised by excitation of green fluorescent protein, as previously described (Renshaw et al., 2006). The zebrafish were prepared by transection of the tailfin at the site indicated to elicit an inflammatory response, which caused recruitment of the neutrophils to the site of injury. B: The chemoattractant field inference framework. Firstly, images of neutrophil recruitment to the zebrafish wound site were acquired by video microscopy. The neutrophil centroid positions were then obtained from a segmentation and tracking algorithm. Velocities of the neutrophils were estimated from the neutrophil centroid tracks using a Kalman smoother and lastly, the velocity estimates were used in the inference of the chemoattractant field.
Figure 2.
Neutrophil centroid position tracks.
The neutrophil tracks (colour lines) were obtained from a segmentation and tracking algorithm and are shown here in relation to the zebrafish image (greyscale), where the zebrafish image of dimension 10001000 pixels has been zoomed on the vertical axis to the 100–900 pixel range.
Figure 3.
Typical examples of neutrophil tracks and neutrophil velocity estimates.
A and D: The image on the left shows a highlighted red track that is zoomed in the plot on the right, in which the centroid positions extracted from the tracking algorithm (black) and smoothed track estimate (red) are compared (the open circle indicates the track start point and the filled circle indicates the track end point). B and E: X-Y cell centroid position estimates corresponding to tracks highlighted in A and B are shown as signals with respect to time produced by the tracking algorithm (black) and estimates from the smoothing algorithm (red). C and F: X-Y velocity estimates (raw estimates in black and smoothed estimates in red), corresponding to position signals in B and E. Raw estimates of velocity were obtained by numerical differencing (central difference method) applied to the tracker position estimates.
Figure 4.
A: Histogram of neutrophil velocities in the X-direction at each sample time (histograms are zoomed to the −10 to 10 m/min range for an effective visualisation and data are aggregated over all fish). B: Histogram of neutrophil velocities in the Y-direction.
Figure 5.
Chemoattractant field inference in vitro.
A: Cell tracks of human neutrophils in vitro chemotaxing due to presence of the chemokine interleukin-8, which increases in concentration from left to right. B: Inferred chemoattractant field, normalised to the range (0,1). The chemoattractant field estimate is dimensionless hence the scale of the colormap is in arbitrary units (a.u.). C: Comparison of inferred chemoattractant field averaged over the Y-direction, to the level of chemokine interleukin-8 reported in [30]. D: Circular histogram of neutrophil angles, demonstrating a directional bias of the tracks shown in panel A towards the lower right corner.
Figure 6.
Chemoattractant field inference in the zebrafish.
For each zebrafish, 1–15, the estimate of the chemoattractant field (colour) is overlayed with transparency on the fish image (grayscale). Each colormap is scaled to the range −20 to 40 to provide an effective visual comparison over all fish. The chemoattractant field estimate is dimensionless hence the scale of the colormap is in arbitrary units.