Image analysis

The microscope images were processed and analyzed by custom scripts written with the macro language of Fiji/ImageJ (https://fiji.sc; Schindelin et al., 2012) and R language (https://www.r-project.org/). The four-dimensional images of NMY-2::GFP were deconvolved time frame by time frame with the DeconvolutionLab2 plugin (http://bigwww.epfl.ch/deconvolution/; Sage et al., 2017) using the Richardson-Lucy algorithm with total-variation regularization and a point spread function calculated by the PSF generator plugin (Kirshner et al., 2013) with the Born and Wolf model. The coordinates of the cell periphery were determined frame by frame in the bleach-corrected, average z-projections of nondeconvolved images using the Trainable Weka Segmentation plugin (https://imagej.net/Trainable_Weka_Segmentation; Arganda-Carreras et al., 2017). The positions of the spindle poles were determined by human visual detection of the weak NMY-2 localization to the spindle and the spindle poles in anonymized videos.

The microtubule density near the cell cortex was quantified according to a procedure summarized in Fig. S2. The background outside of the cell was subtracted from the line profile along a curve of 1 µm width, placed 1.5 µm inside the cell boundary (Fig. S2 A) and standardized with the mean intensity (Fig. S2 B). The position along the curve from the anterior pole to the posterior pole was rescaled from 0 to 1. Data from the top and bottom sides of an embryo were pooled for 10 time points (16 s; Fig. S2 C). The intensity data that were ranked within the top 5% of the local area of 0.2 width window were treated as the signal from a microtubule (red dots in Fig. S2 D). After subtracting the threshold levels, the data from nine embryos were averaged (Fig. S2 E).

The velocity of the cytoplasmic NMY-2::GFP particles (Fig. 1 C) was determined by making kymographs of all the trajectories in four anaphase embryos and measuring their gradients. For automated scoring in Fig. 1 E, Fig. 5 D, and Fig. 7 C, cytoplasmic NMY-2::GFP particles were detected in each time frame of the deconvolved and average z-projected videos by finding maxima within the boundary of each embryo using MaximumFinder in Fiji/ImageJ. The particles detected in two consecutive time frames less than 7 pixels apart, which correspond to movement at 1.295 µm/s, were stitched into a trajectory by using a custom R script and overlaid on the original videos to be checked by visual inspection (not shown). The particles that had been trapped on the spindle from metaphase and the false signals derived from the ingressing cleavage furrow were manually omitted. The distance to the closest cortical point was measured for each time point of a trajectory. The direction of the movement of a trajectory relative to the nearby cortex was assessed by the average rate of the increase of the distance to the closest point on the cortex. For Fig. 1 E, the trajectories from 22 control(RNAi) embryos that appeared in six or more time frames (4.32 s) moving away from the cortex were overlaid with temporal color coding.

The activity of cytoplasmic transport of myosin II toward the centrosome was scored according to the procedure shown in Fig. S3. The force to move a particle of the same size and shape at velocity v in a viscous medium is proportional to v, and thus the power for the movement is proportional to v². Although we do not know the exact size and shape of individual particles, as an estimator of the transport activity of a trajectory per frame, we scored I × v², where I is the intensity of the fluorescent signal of each particle. To quantify the transport activity toward the spindle pole, the movement of the spindle pole needs to be considered. Although the position of the spindle pole was difficult to determine by automation, that of a chromosome linked to the pole, with the kinetochore microtubules of a constant length (anaphase A is absent in C. elegans embryos; Oegema et al., 2001), could be tracked by semi-automation, i.e., by manually specifying the dark spots corresponding to the chromosomes in the first time frame and by repeating identification of a nearby minimum in the next time frame. Thus, the movement of the particles was compensated for the movement of the spindle pole in the A-P direction using the A-P component of the chromosome movement (Fig. S3 A). The (x, y) coordinates of a trajectory (Fig. S3 B) were converted to the centrosome-directed components of the movement, z, considering the main axis of the movement and the centrosome/chromosome movement (Fig. S3 C). After smoothing z to z’, the I × dz'², where dz' is an increment of z' per time point, was calculated for each time point for which dz’ > 0 (Fig. S3 D) and summed up for all the particles moving away from the cortex (Fig. S3 E, thick lines). For Fig. 5 D and Fig. 7 C, the cumulative sum was calculated (Fig. S3 F). This is the total transport activity from anaphase onset (time 0) and is supposed to represent the overall work executed by dynein for the cytoplasmic transport of the myosin II particles.

For analysis of the temporal change of the cortical density and flow of NMY-2::GFP, first, the region 40 pixels (5.33 µm) inside and outside of the edge of the cell in the average z-projected image was straightened so that the periphery of the embryo that was traced counterclockwise starting from the anterior tip was placed from left to right, and thus the posterior tip ended at the center (Fig. S4). The intensity of the cortical NMY-2::GFP signal, which appeared now as a horizontal line in the middle of the straightened “edge” image, was normalized so that the local background (outside of the cell) is 0 and the local cytoplasmic level is 1. The kymograph of the normalized density at the cell cortex was then generated by reslicing the stack of the time series of the straightened edge images with horizontal lines at intervals of 1 pixel and averaging the five consecutive best focal planes, which corresponds to the peak of the NMY-2::GFP signal at the cell periphery (cortex) of 0.67 µm width. After averaging across multiple embryos, the kymograph was folded back at the center (at the posterior tip) so that the anterior and posterior tips were placed on the left and right ends, respectively. For Fig. 5, the average kymograph from the dataset obtained from 0 to 140 s p.a.o. and that from 70 to 215 s p.a.o. were merged by linearly changing the blending ratio for the overlapping period (70–140 s p.a.o.) after correction for photobleaching, which is more profound at the cortex than in the cytoplasm due to slower exchange with unbleached molecules.

One-dimensional PIV was performed by a custom R script that compares the one-dimensional distribution pattern of the cortical NMY-2 signal along the cell periphery in a time frame with that of the next time frame. The spatial resolution of the kymograph of the NMY-2 signal was increased fivefold (Fig. 5 and Fig. 6) or 10-fold (Fig. 7) by interpolation. The local velocity was determined as the spatial shift needed to maximize the cross-correlation between the windows of the size of 256 pixels (6.83 µm, Fig. 5 and Fig. 6) or 384 pixels (5.11 µm, Fig. 7) from the two consecutive time points. This was scanned along the cell periphery and repeated through the temporal dimension to make a kymograph of the flow of NMY-2, and the spatial resolution was set back to the original one by averaging. After averaging across multiple embryos, the kymograph of the flow was folded back so that the anterior and posterior tips were placed on the left and right ends, respectively. Positive (anterior to posterior) and negative (posterior to anterior) flow velocities were presented by pseudocoloring with green and magenta, respectively.