|
|
|
|
|
|
|
|
|||||
| Home Help Feedback Subscriptions Archive Search Table of Contents | |||||
| ||||||||||||||||||||
Files in this Data Supplement:
Adobe PDF
Adobe PDF
Fig. S1. Automated cell density measurements. (A) Sample original image. (B) Background-corrected image The original image is blurred (with a 40 pixel disk structuring element), and the blurred image is subtracted from the original to produce this image, which is corrected for lighting inconsistency across the field of view (typically our images are brighter in the center and dark around the edges). (C) Identified cells. The corrected image is then thresholded and regions which are significantly darker than background are identified. Regions that contain fewer than six pixels are discarded, and the remaining regions are identified as cells. This panel shows the original image with the automatically identified cells overlaid in color. (Inset) The procedure does a good job even in crowded areas. (D) Automatic compared to manual density determinations for three independent experiments. Before we developed the automated density routines, cell density information had been acquired through manual counting (mouse clicks) in sub-regions of a frame, so when the automated routines were developed, we validated them with the manual counts. Large open circles are from automatic density determinations; small filled circles are from hand-counts. Different colors represent different experiments. Data from separate experiments have been aligned with one another by rigid sliding along the time axis until the differences between them have been minimized. The zero time point on the x axis is arbitrarily placed.
Fig. S2. Processing for Wound Extraction. (A) Sample original image. (B) Locally blurred original image. Similar to the procedure for cell density measurements, we want to subtract out the effects of lighting inconsistency across the field of view. However, unlike in the density measurements, the distribution of image features is not approximately uniform (the wound edges are real features that we do not want to blur away), so we use a smaller structuring element for blurring this time (a 10 pixel rather than 40 pixel disk). (C) Background corrected image from subtracting the blurred image from the original. (D) Binary image after thresholding. A threshold value is chosen which does a good job of separating cells from the wound gap. The method is very robust, typically a wide range of threshold values works well, and a value chosen for one image in an experiment works for all images acquired in that experiment. (E) Binary image after further processing. The main step after panel D is the morphological operation of ‘opening’, which erodes, then dilates islands of foreground pixels in a binary image. The net result is removal of any small white areas while leaving large white areas essentially unchanged. After this step, any remaining holes (black pixels completely surrounded by white pixels, or vice versa) are removed, yielding a single large white area (the wound gap) and two large black areas (the two sides of the monolayer). (F) Extracted wound edge. The boundaries identified in E are overlaid in red on the original image. The quality of edge identification over an entire experiment can be evaluated in Movie S3.
Movie S1. L1 and simulated cell culture. Both cultures grow to confluence in roughly 4 days (∼100 hours). Note that the simulation mimics the tendency of L1 cells to grow as dispersed, rather than colony-based, cultures. Images in both movies are 30 minutes apart, played at 10 frames per second (18,000× real time). Total duration is ∼4 days (100 hours). Cell colors in simulation movies are primarily for contrast between neighboring cells. However, colors are assigned by birth order, always stretch in a spectrum from dark blue through dark red, and new cells from division are added at the red end of the spectrum, which means that the youngest cells are always the darkest red.
Movie S2. Zoomed view of culture growth. This is a zoomed view of the culture growth Movie S1 for a period after confluence has been established, but before complete cessation of division. Images are spaced by 5 minutes of simulation time, played at 15 frames per second (4500× real time). The movie duration is ∼20 hours of simulation time.
Movie S3. Effects of sensing radius and division on wound healing dynamics. Four simulations play simultaneously in this movie. In each simulation cells have a division radius of 9 um. From bottom-to-top the cells have sensing radii of 2 μm, 4 μm, 9 μm and 9 μm respectively. In the bottom three panels division is turned off. The only difference between the top two panels is that division is turned on in the upper-most panel. Note how the value of the sensing radius determines the time when the monolayer expansion transitions from collective to diffusive. Also note that cell division sustains collective expansion until closure in the case of 9/9 cells. Images in all simulations are 30 minutes apart, played back at 10 frames per second (18,000× real time). Total duration is 24 hours.
Movie S4. Side-by-side comparison of L1 and simulated wound healing. The simulation is populated with 4/4 cells operating with 80% obstacle avoidance. Note that division occurs immediately in both cases and occurs deep within the surviving monolayer. L1 cells become phase bright when they divide. Images are 30 minutes apart, played at 10 frames per second (18,000× real time). Total duration is 22.5 hours.
Movie S5. Demonstration of automated wound tracking. To quantify L1 monolayer expansion we developed tools that automatically detect the wound edge throughout time-lapse movies. The method for edge detection is outlined in Figure S3. This movie demonstrates the ability of the routine to track of the leading edge (red line) over time. Images are 1 hour apart, played at 10 frames per second (36,000× real time). Total duration is 22.5 hours.
| ||||||||||||||||||||
