to BioTechniques free email alert service to receive content updates.
An algorithm to quantify correlated collective cell migration behavior
 
Benjamin Slater*1, Camila Londono*2, and Alison P. McGuigan1,2
1Department of Chemical Engineering and Applied Chemistry
2Institute of Biomaterials and Biomedical Engineering, University of Toronto, Toronto, Ontario, Canada


*B.S. and C.L. contributed equally to this work
BioTechniques, Vol. 54, No. 2, February 2013, pp. 87–92
Full Text (PDF)
Supplementary Material
Abstract

Collective cell migration is an important process that determines cell reorganization in a number of biological events such as development and regeneration. Random cell reorganization within a confluent monolayer is a popular in vitro model system for understanding the mechanisms that underlie coordination between neighboring cells during collective motion. Here we describe a simple automated C++ algorithm to quantify the width of streams of correlated cells moving within monolayers. Our method is efficient and allows analysis of thousands of cells in under a minute; analysis of large data sets is therefore possible without limitations due to computational time, a common analysis bottleneck. Furthermore, our method allows characterization of the variability in correlated stream widths among a cell monolayer. We quantify stream width in the human retinal epithelial cell line ARPE-19 and the fibroblast cell line BJ, and find that for both cell types, stream widths within the monolayer vary in size significantly with a peak width of 40 µm, corresponding to a width of approximately two cells. Our algorithm provides a novel analytical tool to quantify and analyze correlated cell movement in confluent sheets at a population level and to assess factors that impact coordinated collective cell migration.

Collective cell migration is critical in numerous biological processes such as embryo morphogenesis (1, 2), angiogenesis (3), and wound healing (4). It also plays a central role in pathological situations such as metastasis, allowing some tumor cells to invade surrounding tissues (5). However, despite the importance of collective cell migration, how cell coordination occurs during group migration is not well understood (6). A better understanding of the underlying behaviors during collective cell migration will provide insight into embryo development and tissue reorganization during regeneration and disease. Furthermore, the ability to control collective migration will provide novel tools for engineering tissues with reproducible structures.

There are a number of in vivo (7-9) and in vitro model systems (5, 10-12) used to study collective cell migration, both directed migration and random migration. Two common in vitro models are confluent sheets of cells (13) with or without a wound intended to induce directed migration. These simple 2-D model systems are powerful because they allow tracking of individual cells within the sheet and careful characterization of the group behavior (14) and cell-cell coordination (15) that arises from cell-cell interactions between neighboring cells. Specifically, cells within the sheet randomly exchange places with their neighbors, resulting in a net displacement from their original location (16) and the formation of randomly oriented cellular “streams” characterized by correlated cell motion (16, 17). When a wound is present, the streams at the wound edge become oriented and directed cell movement into the free space occurs to close the wound. Within a cellular stream, during both random and directed migration, motion is coordinated between neighboring cells (i.e., across the width of the stream) and between cells moving behind and ahead of each other (i.e., along the length of the stream). While cell coordination along the stream length could arise passively from the creation of free space behind a cell, coordination across the width of the stream requires more complex cellular interactions. Therefore, understanding the behavior of cells along the width of a stream can potentially provide more insight into the mechanisms of cell-cell coordination during sheet reorganization. Furthermore, quantifying stream width is potentially useful for designing culture surfaces (18) with adhesive or topographic features (unpublished data) that allow manipulation of this streaming behavior, in which feature dimensions must be on the scale of the stream width (19).

A common method for quantifying cell coordination is the velocity correlation function (16, 20-22), which evaluates the overall correlation of cell trajectories as a function of distance from a cell using the Equation 1:



where N refers to the total number of objects being analyzed, n is the number of neighboring objects within a radius, r, from the current object i, and j refers to all other objects within radius r. By averaging all dot product values for all cell pairs, the function provides an overall description of neighboring cell correlation as a function of distance from a cell. When the velocity correlation function falls to ~0 (or θaverage = π/2), cells are no longer considered correlated. The radius at which this occurs provides a measurement ofthe correlated domain (a combination of stream width and length). A major drawback of the velocity correlation function is that it is inherently biased by the behavior of the cells both in front and behind the cell of interest. For example, when streams extend over large distances compared with the stream width, such as in fast moving cell types, the velocity correlation function reports positive correlation over a larger distance than the stream width because correlation parallel to the direction of motion offsets the diminishing correlation in the axis perpendicular to the direction of motion. Furthermore, cell behavior within a confluent sheet is highly variable, and since the velocity correlation function is often presented as the average for all cells, information about the variation in behavior among the population is lost. Here, we describe an algorithm to explicitly quantify the width of cellular streams and the variability in the width of stream within a confluent cell sheet, with or without a wound.

Materials and methods

Cell culture

We conducted experiments using human retinal epithelial cells (ARPE-19; ATTC, Manassas, VA, USA) and human foreskin fibroblast cells (BJ; ATTC). ARPE-19 cells were grown in Dulbecco's modified Eagle's medium/nutrient F-12 (DMEM/F-12; Invitrogen, Burlington, ON, Canada) supplemented with 10% fetal bovine serum (FBS; VWR International, Mississauga, ON, Canada) and 1% penicillin/streptomycin (pen/strep; VWR International). BJ cells were grown in DMEM (ATCC) supplemented with 10% FBS and 1% pen/strep. All cells were maintained in a humidified atmosphere at 37°C and 5% CO2. For some experiments we overexpressed GFP-N-cadherin in ARPE19 cells using a lentivirus overexpression vector, followed by sorting of the cells by flow cytometry to ensure a homogenous population.

Live cell imaging and tracking

We seeded 64,000 cells/cm2 in a custom-made 96-well polydimethylsiloxane (PDMS) flat bottom plate to generate confluent cell sheets (Figure 1A). For wound healing studies, 90,000 cells/cm2 were seeded in a 24-well tissue culture polystyrene (TCPS) plate coated with 314 µg/mL PureCol (Advanced Biomatrix, San Diego, CA, USA) for 1 h prior to seeding and wounds were made using a P10 tip. We tracked cells a day later using a previously established protocol (15, 23). Briefly, cells were stained with 500 ng/mL Hoechst 33342 (Invitrogen) in culture medium for 30 min, followed by a phosphate-buffered saline (PBS) wash and addition of culture medium. An ImageXpress Micro high-content screening microscope with a live-cell imaging module (Molecular Devices, Sunnyvale, CA, USA) was used to take images of the cell nuclei every 20 min for 8 h for tracking in confluent sheets (Supplementary Figure S1 and Supplementary Movie S1) and for 23 h for wound healing (Supplementary Figure S2 and Supplementary Movie S2). Throughout the experiment, cells were maintained in a humidified atmosphere at 37°C and 5% CO2. Positional tracking of cells in each well was performed using the Multi-Dimensional Motion Analysis application module in the MetaXpress software package (Molecular Devices). This tracking algorithm uses nuclear shape and fluorescent intensity measurements to track the position of individual cells (Supplementary Figures S1 and S2) (24, 25).



  1    2    3    4