Integrating transient cellular and nuclear motions to comprehensively describe cell migration patterns

Various subcellular activities, such as protrusion and detachment, compose a cell migration process. The molecular mechanisms of these subcellular activities have been elucidated. However, there is no method that can assess the contributions of these subcellular activities to the global cell migration pattern of a given cell type. Hence, we develop a powerful approach based on CN correlations that quantitatively profiles the cell migration pattern of a given cell type in terms of assembled subcellular activities. In this way, we bridge migration data at the cellular level with underlying molecular mechanisms. The CN correlation profile is found to uniquely and consistently represent the cell migration pattern of each cell type probed. It can clearly reveal the effects of molecular perturbations, such as Y27632 and Cdc42 knockdown on each subcellular migratory activity. As a result, the CN correlation approach serves as a cell dynamic descriptor that can extract comprehensive quantitative data from cell migration movies for integrative biological analyses.


SI Text Cell turning collectively resembles the CN correlation distributions of assembled subcellular events
A cell turning event is composed of several subcellular migratory events. As a result, the overall distribution of CN correlations of a cell turning event would be expected to be a composition of the distributions from the assembled subcellular migratory activities. Yet, from an example movie we found that a whole-cell contraction process could distribute some CN correlation data to the polar angle zone greater than 130° (Fig. S1).
The example movie showed that the whole cell abruptly contracted at the beginning of the turning event (1 st -8 th minute). This contraction led the momentary leading-edge retraction and generated a sequence of CCD in a direction opposite to previous movement. Meanwhile, the nucleus still performed a forward motion toward the leading edge, suggesting that a contractile force continuously existed between the leading edge and the nucleus. Hence, the coupled NCD// were opposite to the CCD direction and the corresponding CN correlations were distributed in the polar angle zone greater than 130°. In the case a tensile force did exist between the nucleus and the original leading edge, the leading edge and the nucleus underwent a so-called "tug-of-war" process. Hence, the tensile force would be released quickly when one of them withdraws significantly. Therefore, a greater CCD would lead to a smaller NCD// and vice versa.
Afterwards, the cell mainly performed sampling (9 th -25 th minute) and gave rise to CN correlations within the 60°-120° polar angle zone. Since the cell had already been released from the previous polarity, the previously elongated nucleus also exhibited a notable relaxation to become rounded. During the period, the cell eventually developed a more significant protrusion as the new leading edge. Hence, the nucleus rotated accordingly and the cell started protruding along the new direction (26 th -43 rd minute). In this period, the CN correlation data were distributed in a similar manner as protrusion under active migration cases. Following, the cell performed regular detachment under active motion (44 th -60 th minute). Figure S1. In a cell turning event the CN correlations distribution collectively resembles those of the assembled subcellular events. A NIH 3T3 fibroblast turning event (Movie S5) consists of 4 sequential and independent subcellular migratory events: leading edge retraction, cell sampling with notable nucleus shape adjustment, nucleus rotation and new polarized direction development, and active leading-edge protrusion (from left to right). The CN correlation distribution of the corresponding event (red dots) is displayed in the CN plot. The schematic diagram of each assembled subcellular migratory activity is also depicted in the associated panel. The overlaid time-lapse frames of fluorescence cell (green) and nucleus (blue) are displayed in grim graphs (bottom row).

The CN correlation profile of a cell type is consistent and unique
To further prove that the CN correlation profile of a cell type is representative and unique, we evaluated two CN correlation profiles extracted from different batches of the same cell type. Besides fibroblasts (NIH 3T3 fibroblasts) (in main text Fig. 2c), different cell types, such as osteosarcoma cells (U-2 OS cells) and adenocarcinoma cells (SKOV-3 cells), all constructed indistinguishable CN correlation profile (Fig. S2a). In addition, we also tested whether the CN correlation profiles among different cell types are distinguishable. Eight cell types were cross-compared, including 1 epithelial cell (OSE-10), 3 fibroblasts (NIH 3T3, Swiss 3T3 and Human Foreskin (HF) fibroblasts), 3 adenocarcinoma cells (OVCAR-3, SKOV-3, and MDA-MB-231 cells), and 1 osteosarcoma cell (U-2 OS cells). The CN correlation profiles of these 8 cell types were individually constructed using 25 one-hour cell movies recoded at one-minute time intervals. The occurrence diagrams and CCD diagrams were paired for comparison using the Sign test and the Lepage test, respectively. Since the comparisons of the other cell types have been shown in the later section of the main text, here we only showed the comparison between two cell types, U-2 OS cells and Swiss 3T3 fibroblasts, which have the most similar CN correlation profiles. The results clearly elucidated the significant difference between their profiles (Fig. S2b). Hence, among all the cell types considered, each individual cell type possesses a unique CN correlation profile describing its cell migration pattern and momentary dynamics.

Hoechst 33342 staining will not affect the assessment of subcellular activities
In the CN correlation approach, the imaging acquisition takes one-hour duration; the Hoechst 33342 stain was only applied to the sample 10 minutes prior to the acquisition. To explore whether the application of the Hoechst 33342 affects the imaging acquisition and the reliability of the CN correlation approach, we compared the subcellular morphology changes per minute by the axis ratio, the max length, the eccentricity and the circularity in individual cells within different one-hour durations using the twosample-T-test under the following conditions: without staining, the first hour after staining, and the second hour after staining (Table S2). In addition, we also evaluated the histogram of cell dynamic activities during the first hour and the second hour after the Hoechst stain was applied (Fig. S3). These results demonstrated that no noticeable subcellular dynamic differences among the control and after the sample was subjected to the Hoechst stain.

Supplementary movies
Movie S1-S5. Each video file displays a typical cell migration mode at 1-min intervals over a 20-minute period, including (Movie S1) detachment, (Movie S2) protrusion, (Movie S3) sampling, (Movie S4) side protrusion, and (Movie S5) large angle turning. The file was an overlay of two simultaneous movies of the same cell: one documented the red fluorescence protein (RFP)-labeled cell and the other documented its coupled Hoechst 33342-labeled nucleus.