Department of Chemical and Biological Engineering, Iowa State University, Iowa, USA

Department of Genetics, Development and Cell Biology, Iowa State University, Iowa, USA

Department of Biochemical Engineering, University College London, London, UK

Abstract

Background

Cell migration plays an essential role in many biological processes, such as cancer metastasis, wound healing and immune response. Cell migration is mediated through protrusion and focal adhesion (FA) assembly, maturation and disassembly. Epidermal growth factor (EGF) is known to enhance migration rate in many cell types; however it is not known how FA maturation, FA dynamics and protrusion dynamics are regulated during EGF-induced migration. Here we use total internal reflection fluorescence (TIRF) microscopy and image analysis to quantify FA properties and protrusion dynamics under different doses of EGF stimulation.

Results

EGF was found to broaden the distribution of cell migration rates, generating more fast and slow cells. Furthermore, groups based on EGF stimulation condition or cell migration speed were marked by characteristic signatures. When data was binned based on EGF stimulation conditions, FA intensity and FA number per cell showed the largest difference among stimulation groups. FA intensity decreased with increasing EGF concentration and FA number per cell was highest under intermediate stimulation conditions. No difference in protrusion behavior was observed. However, when data was binned based on cell migration speed, FA intensity and not FA number per cell showed the largest difference among groups. FA intensity was lower for fast migrating cells. Additionally, waves of protrusion tended to correlate with fast migrating cells.

Conclusions

Only a portion of the FA properties and protrusion dynamics that correlate with migration speed, correlate with EGF stimulation condition. Those that do not correlate with EGF stimulation condition constitute the most sensitive output for identifying why cells respond differently to EGF. The idea that EGF can both increase and decrease the migration speed of individual cells in a population has particular relevance to cancer metastasis where the microenvironment can select subpopulations based on some adhesion and protrusion characteristics, leading to a more invasive phenotype as would be seen if all cells responded like an “average” cell.

Background

Cell migration plays an important role in tumor progression

FAs are dynamic, macromolecular structures that serve as both mechanical linkages and centers of intracellular signal transduction

Protrusion is mediated by actin polymerization, whereas retraction is driven through myosin II activity and actin depolymerization

To understand the relationship between EGF-stimulated cell migration, FA properties and protrusion dynamics, we imaged metastatic (MTLn3) and non-metastatic (MTC) cell lines. We analyzed the cell migration speed and persistence under various EGF stimulation conditions and found that EGF moderately increased the median migration rate and persistence of MTLn3 cells, whereas it had no significant effect on the speed and persistence of MTC cells. Interestingly, higher concentrations of EGF broadened the distributions and increased the coefficient of variation of both the migration rate and persistence of MTLn3 cells, but not MTC cells. When data was binned based on EGF stimulation conditions, FA intensity and FA number per cell showed the largest difference among stimulation groups. FA intensity decreased with increasing EGF concentration and FA number per cell was highest under intermediate stimulation conditions. No difference in protrusion behavior was observed. However, when data was binned based on cell migration speed, FA intensity and not FA number per cell showed the largest difference among groups. FA intensity was lower for fast migrating cells. Additionally, waves of protrusion tended to correlate with fast migrating cells. Consequently, low FA intensity and waves of protrusion are markers for fast migrating cells, but these characteristics are only partially predictive of EGF stimulation conditions because of the large cell-to-cell variability in response to EGF.

Results

EGF stimulation broadens the distributions of migration speed and persistence of MTLn3 cells

In many cell types EGF has been reported to enhance the mean migration speed. However, the long term migration response of individual cells after challenge with EGF in this model system is not known. Consequently, we examined cell migration under various doses of EGF in both adenocarcinoma cells (MTLn3) and non-metastatic cells taken from the same tumor (MTC) (Figure

Figure S1.Migration trajectories of typical cells under different EGF concentrations. Three cell tracks were chosen randomly under each EGF concentrations and labeled with different colors. All trajectories were aligned to the starting point (0, 0).

Click here for file

Increasing EGF broadens the distribution of both cell speed and persistence of MTLn3 cells while not changing MTC cell migration

**Increasing EGF broadens the distribution of both cell speed and persistence of MTLn3 cells while not changing MTC cell migration. A**. cell speed and (um/hr) in the axis description overlaps. **B**. Cell persistence of MTLn3 (left) and MTC (right) under different EGF concentrations. Data for individual cells are shown as black dots. On each box, the central marker is the median; the edges of the box are the 25^{th} and 75^{th} percentiles; the whiskers extend to the most extreme data points not considered outliers. (For MTLn3, _{0} = 16, _{0.01} = 24, _{0.1} = 30, _{1} = 26, _{10} = 65, _{100} = 60; For MTC, _{0.01} = 26, _{1} = 24, _{100} = 41.) **C**. Correlation between cell speed and persistence for MTLn3 cells (white circles, **D**. Coefficient of variation of cell speed for MTLn3 cells as a function of EGF concentration. **E**. Coefficient of variation of cell persistence for MTLn3 cells as a function of EGF concentration. Curves were fitted to a 2^{nd} order polynomial and meant only to guide the eyes.

Given that EGF did not dramatically affect the median migration, we grouped cells according to no (0 nM EGF), low (0.01 and 0.1 nM) and high (1, 10 and 100 nM) EGF stimulation. Additionally, since the variability in cell migration speed seems to be an additional feature of the data, we also grouped cells based on cell migration speed. A

The distribution of FA characteristics differ under different EGF stimulation conditions and between fast and slow migrating cells

In order to measure the FA characteristics of FA intensity, number per cell, size, speed, lifetime and elongation, we used TIRF microscopy to observe FAs within a single cell transfected with paxillin-EGFP, a component that marks FAs throughout their entire lifetime. Expression of paxillin-EGFP did not seem to alter the migration speed, nor was the expression dependent on EGF stimulation (Additional file

Figure S2.Paxillin-EGFP expression levels for cells with different cell speeds and under different EGF stimulation conditions. Mean intensity of individual cells as a function of A. cell speed, and B. EGF concentrations. _{cell} = 56.

Click here for file

Figure S3.Average paxillin-EGFP intensity in cells as a function of time. A. Mean FA intensity and B. mean intensity of the whole cell for all cells as a function of time. Mean FA Intensity: _{cells} = 55. Mean Intensity _{cells} = 53.

Click here for file

Time-lapse series of FA dynamics in MTLn3 cells

**Time-lapse series of FA dynamics in MTLn3 cells.** MTLn3 cell expressing paxillin-EGFP and stimulated with 0.01 nM EGF is shown. Images were taken at 0, 10, 20, 30 and 40 min. **A**. Original total internal reflection fluorescence (TIRF) images of cells with FAs. **B**. Binary images of whole cell masks after segmentation. **C**. Binary images of FA masks after segmentation. **D**. Composite images of tracked FAs within the cell, where green represents original images and red represents the segmented FA masks. The scale bar is 10 μm.

**Time-resolved data of mean FA intensity and FA number per cell. A.** Mean FA intensity and C. FA number per cell for the cell shown in Figure 2 under 0.01 nM EGF stimulation as a function of time. B. FA mean intensity and D. number per cell of five typical cells under different EGF concentrations as a function of time.

Click here for file

**EGF (nM)**

**CELL NUMBER**

**POOR**

**GOOD**

**EXCELLENT**

Percentages of poor, good or excellent FA tracking results under different EGF concentrations are shown. An example of a cell that was rated as good is shown in Figure

0

5

0

60 %

40 %

0.01

10

20 %

70 %

10 %

0.1

10

50 %

20 %

30 %

1

10

10 %

20 %

70 %

10

12

17 %

33 %

50 %

100

8

25 %

50 %

25 %

Total

55

22 %

40 %

38 %

FA characteristics can be ordered based on the magnitude in the difference between either the no, low and high EGF stimulation conditions or between slow and fast migrating cells. This magnitude was quantified by the Kolmogorov-Smirnov statistic (Figure

Figure S5.Distributions of FA size, speed, lifetime and elongation for different EGF stimulation conditions. Histograms of FA size A.-C., speed E.-G., lifetime I.-K. and size M.-O. were generated by dividing cells into three EGF stimulation groups (A., E., I. and M., no EGF (0 nM), B., F., J. and N. low EGF (0.01 and 0.1 nM) and C., G.,K. and O. high EGF (1, 10 and 100 nM)). Histograms were fitted with lognormal probability distributions except for speed, which was fitted with a Weibull probability distribution. The mean values of FA D. size, H. speed, L. lifetime and P. elongation are also shown. The number of measurements of FA properties is the product of the average FA number and the cell number. Size, speed, lifetime and elongation: _{cell,no} = 5, _{FA,no} = 1963, _{cell,low} = 20, _{FA,low} = 13,024, _{cell,high} = 30, _{FA,high} = 14,648. Error bars are 95 % confidence intervals and asterisks denote

Click here for file

Figure S6.Distributions of FA size, speed, lifetime and elongation for slow and fast migrating cells. Histograms of FA size A.-B., speed D.-E., lifetime G.-H. and size J.-K. were generated by dividing cells into slow and fast migrating groups (A., D., G. and J. slow migrating cells and B., E., H. and K. fast migrating cells). Histograms were fitted with lognormal probability distributions except for speed, which was fitted with a Weibull probability distribution. The mean values of FA C. size, F. speed, I. lifetime and L. elongation are also shown. The number of measurements of FA properties is the product of the average FA number and the cell number. Size, speed, lifetime and elongation: _{cell,slow} = 21, _{FA,slow} = 12,773, _{cell,fast} = 34, _{FA,fast} = 16,842. Error bars are 95 % confidence intervals and asterisks denote

Click here for file

Quantification of the difference between experimental distributions of FA properties

**Quantification of the difference between experimental distributions of FA properties.** Cells were binned based on EGF concentration (no (0 nM), low (0.1 and 0.01 nM) and high (100, 10 and 1 nM)) or cell speed (low (<42 μm/hr) or high (>42 μm/hr)). The Kolmogorov-Smirnov statistic was the average of three pair-wise comparisons (EGF) or simply the pair-wise comparison (cell speed). A larger value for the Kolmogorov-Smirnov statistic signifies a higher probability that there are differences between groups.

**Properties**

**Fast Migrating Cells**

**Low EGF**

**High EGF**

The ranges indicates areas of higher probability of fast migrating cells compared to slow migrating cells, cells stimulated with low concentrations of EGF compared to those exposed to no EGF and cells stimulated with high concentrations of EGF compared to those exposed to no EGF. The ranges were determined by calculating crossover points between the fit distributions. N/A means no crossover points.

FA Number

> 110

> 86

> 97

FA Size (μm^{2})

0.30 - 3.0

0.18 - 3.0

0.24 – 3.0

FA Sliding Speed (μm/hr)

> 12

> 12

> 12

FA Lifetime (s)

0 - 440

160 - 530

160 - 620

FA Intensity (grayscale)

< 22,000

<25,000

<25,000

FA Elongation

N/A

1.3 – 2.1

1.3 – 2.2

FA intensity decreases with increasing EGF concentration and number per cell is maximal at intermediate EGF concentrations

**FA intensity decreases with increasing EGF concentration and number per cell is maximal at intermediate EGF concentrations.** Histograms of FA intensity A.-C. and number per cell E.-G. were generated by dividing cells into three EGF stimulation groups (A., E. no EGF (0 nM), B., F. low EGF (0.01 and 0.1 nM) and C., G. high EGF (1, 10 and 100 nM)). Histograms were fitted with lognormal probability distributions. The mean values of FA D. intensity and H. number per cell are also shown. The number of measurements of FA number per cell is the product of the average number of frames and the cell number. The number of measurements of FA properties is the product of the average FA number and the cell number. Intensity: _{cell,no} = 5, _{FA,no} = 1963, _{cell,low} = 20, _{FA,low} = 13,024, _{cell,high} = 30, _{FA,high} = 14,648. Number per cell: _{cell,no} = 5, _{FA,no} = 1545, _{cell,low} = 20, _{FA,low} = 5909, _{cell,high} = 30, _{FA,high} = 8937. Error bars are 95 % confidence intervals and asterisks denote

FA intensity is lower in fast migrating cells and FA number per cell is slightly higher in fast migrating cells

**FA intensity is lower in fast migrating cells and FA number per cell is slightly higher in fast migrating cells.** Histograms of FA intensity A.-B. and FA number per cell D.-E. were generated by dividing cells into two cell migration speed groups (A., D. slow (<42 μm/hr) and B., E. fast (>42 μm/hr)). Histograms were fit with lognormal probability distributions. The mean values of FA C. intensity and F. number per cell are also shown. The number of measurements of FA number per cell is the product of the average number of frames and the cell number. The number of measurements of FA properties is the product of the average FA number and the cell number. Intensity: _{cell,slow} = 21, _{FA,slow} = 12,773, _{cell,fast} = 34, _{FA,fast} = 16,862. Number per cell: _{cell,slow} = 21, _{FA,slow} = 6146, _{cell,fast} = 34, _{FA,fast} = 10,245. Error bars are 95 % confidence intervals and asterisks denote

Unique spatial organization of protrusion and retraction is exhibited in fast migrating cells

We assessed differences in protrusion and retraction behavior under different EGF stimulation conditions and between the slow and fast migrating cells. The cell-average protrusion and retraction velocities would be faster in the fast migrating cells resulting in the increased migration rate; however different patterns of protrusion could lead to the same average value, so we analyzed local protrusion behavior. One prominent feature that we observed was traveling waves of protrusion along the edge of the cell. This traveling wave behavior had the effect of broadening of the protrusion velocity distribution. Upon qualitative examination, traveling waves did not seem to be linked to EGF stimulation conditions. Additionally, slow migrating cells usually showed large quiescent areas (green) and random, disorganized protrusion and retraction behavior (Figure

The spatial control of protrusion differs between slow and fast migrating cells

**The spatial control of protrusion differs between slow and fast migrating cells. A**. Protrusion velocity map for slow migrating cells at 0.01, 1, and 100 nM EGF from left to right. **B**. Protrusion velocity map for fast migrating cells at 0.01, 1, and 100 nM EGF from left to right. The cell edge was divided into 100 segments and the average protrusion rate in each segment was determined over time. Red represents fast protrusion, green represents quiescence and blue represents fast retraction. **C**. The Kolmogorov-Smirnov statistic was the average of three pair-wise comparisons (EGF) or simply the pair-wise comparison (cell speed). A larger value for the Kolmogorov-Smirnov statistic signifies a higher probability that there are differences between groups. **D**. The fraction of cells with lateral waves (gray bars) or high standard deviation (STD) of protrusion velocity (white bars) between slow and fast migrating cells. **E**. Histograms of protrusion/retraction velocity between slow migrating cells (gray bars) and fast migrating cells (white bars). Histograms of slow (dot lines) and fast migrating cells (solid lines) were fit with Gaussian distribution. **F**. The mean values of standard deviation (STD) of protrusion velocity between slow and fast migrating cells were also shown. _{slow} = 34, _{fast} = 24. Error bars are 95 % confidence intervals and asterisks denote

Discussion

Variability in cell response to environmental cues is becoming a more appreciated phenomenon that can drive how populations of cells respond to their environment. Cell-to-cell variability can arise from heterogeneity in protein level

EGF does seem to regulate some FA characteristics, namely FA intensity and number per cell. FA intensity decreases as EGF stimulation increases. FA number per cell is highest at low EGF concentrations, suggesting that either the assembly is maximized or disassembly is minimized at this point. EGF is known to alter actin cytoskeleton dynamics, perhaps resulting in enhanced assembly dynamics. Alternatively, EGF is also known to upregulate calpain, a protease involved in disassembly, which might be largest at high EGF concentrations. Interestingly, FA number per cell does not seem to affect cell migration rate. Rather, low FA intensity seems to correlate with fast migrating cells. This might be a direct link between EGF stimulation and cell migration speed regulation (Figure ^{2}) is not surprising. Focal complexes, small FAs are traditionally thought to occupy this range of areas

Different adhesion and protrusion characteristics correlate with EGF stimulation and cell speed

**Different adhesion and protrusion characteristics correlate with EGF stimulation and cell speed.** Cells can either be grouped based on EGF concentration or cell speed. EGF concentration is considered an input that acts to regulate adhesion and protrusion characteristics, whereas cell speed is an output that acts to integrate information determined by inputs such as EGF concentration. Both FA intensity and number per cell correlate with EGF concentration, whereas FA intensity and the presence or absence of protrusion waves correlate with cell speed. Cell speed could be regulated by EGF through changes in FA intensity, but other inputs are most likely needed to regulate the presence of protrusion waves, since EGF concentration correlates poorly with the presence or absence of protrusion waves.

Given that local protrusion is linked to FA intensity and that FA intensity was lowest in fast migrating cells we were encouraged to examine the protrusion dynamics under different stimulation conditions. These cells are known to respond acutely to EGF stimulation with two peaks of barbed end formation resulting in a robust protrusion response

Conclusions

EGF was found to broaden the distribution of cell migration rates, generating both faster and slower cells, but not dramatically affecting the average response. Several different adhesion and protrusion characteristics correlated with EGF stimulation and cell migration speed, however there is a hierarchy of these correlations. FA intensity and number per cell correlate with EGF stimulation conditions. FA intensity decreases with increasing EGF stimulation and FA number per cell is highest at low EGF stimulation conditions. In contrast, FA intensity and not number per cell as well as protrusion waves correlate with cell speed. Fast cells are marked by low FA intensity and protrusion waves. Consequently, while EGF stimulation could regulate FA intensity to modulate cell speed directly or by partially activating protrusion waves, other environmental factors most likely lead to protrusion waves. Adhesion and protrusion characteristics that do not correlate with EGF stimulation condition but do correlate with cell migration speed constitute the most sensitive outputs for identifying why cells respond differently to EGF. The idea that EGF can both increase and decrease the migration speed of individual cells in a population has particular relevance to cancer metastasis where the microenvironment can select subpopulations based on some adhesion and protrusion characteristics, leading to a more invasive phenotype as would be seen if all cells responded like an “average” cell.

Methods

Materials

Cell culture media was α-MEM medium with L-glutamine (Invitrogen) containing 5 % fetal bovine serum (Invitrogen) and 1 % penicillin-streptomycin (Invitrogen). Collagen and poly-L-lysine (PLL) solution contained 3 μg/ml of rat tail collagen I (Invitrogen) and 2 μg/ml of poly-L-lysine hydrochloride (Sigma), dissolved in 0.5 M acetic acid (Fisher) and sterilized under ultraviolet light for 30 minutes. Serum free imaging media was α-MEM medium without phenol red (Invitrogen) containing 1 mg/ml bovine serum albumin (Sigma), 12 mM HEPES (Fisher), and 1 % penicillin-streptomycin (Invitrogen), adjusted to pH 7.4 and filtered through 0.22 μm pore size filter (Millipore, Fisher).

Cell culture

Rat mammary adenocarcinoma cell lines (metastatic MTLn3 and non-metastatic MTC) were obtained from Dr. Jeffrey E. Segall (Albert Einstein college of Medicine). Cell lines were derived from the 13762NF rat mammary adenocarcinoma tumor. Cells were maintained in cell culture media at 37 °C in 5 % CO_{2} and were passed every 2 or 3 days. Collagen and PLL solution was incubated on 22 × 22 mm squeaky cleaned coverslips (Corning, Fisher) at room temperature for 1 hour. Cells were seeded on coverslips with collagen and PLL and incubated for 24 ~ 48 hours at 37 °C in 5 % CO_{2} (50,000 ~ 100,000 cells/coverslip).

Cell migration assay

MTLn3 and MTC cells were incubated on coverslips with collagen and PLL for 48 hours and were switched to serum free imaging media for 2 hours. Coverslips were mounted onto glass slide chambers in serum free imaging media with different concentrations of EGF (0, 0.01, 0.1, 1, 10 and 100nM EGF). Chambers were maintained at 37 °C for 2 hours and then imaged on a heated stage every 2 minutes for 8 hours. Phase contrast time-lapse images were captured at 20× (NA 0.50, Nikon) with a charge-coupled device (CoolSNAP HQ2, Photometrics) attached to an inverted microscope (Eclipse Ti, Nikon). Cell centroids were identified and tracked manually by MTrackJ plugins of ImageJ. Single cell speed,

using a non-linear least squares regression analysis. The sampling time is every two minutes for 6–8 hours. The mean-squared displacement was constructed using non-overlapping time intervals. Consequently, the model was fitted to data up to a 30 min time lag due to the small number of displacements (<12-16) at time lags greater than 30 min. To quantify protrusion rate we used a constrained optimization program to measure the protrusion and retraction rates from masked images as done previously

Fluorecence imaging

MTLn3 cells were incubated on coverslips with collagen and PLL for 24 hours and transfected with paxillin-EGFP and Fugene 6 (Roche) according to the manufacturer’s protocol (6 μl of Fugene 6 and 3 μg of EGFP-paxillin). After one hour transfection, the media was changed to cell culture media and the transfected cells were maintained at 37 °C in 5 % CO_{2} for 23 hours. Then the cells were switched to serum free imaging media for 2 hours. Coverslips were mounted onto glass slide chambers in serum free imaging media with different concentrations of EGF (0, 0.01, 0.1, 1, 10 and 100 nM EGF). Chambers were maintained at 37 °C for 2 hours and then imaged on a heated stage every 10 seconds for 40 ~ 60 minutes. TIRF images were captured at 60× oil objective (NA 1.49, Nikon) equipped with a TIRF illuminator and fiber optic-coupled laser illumination. The 488 nm laser line of an air-cooled tunable Argon laser (Omnichrome Model 543-AP-A01, Melles Griot) was reflected off a dichroic mirror (89000 ET-QUAD, Chroma). Camera and shutter were controlled by μManager 1.3. An automated segmentation and tracking algorithm was utilized for large-scale analysis of FA dynamics ^{2} and larger than 10 μm^{2} were excluded from our analysis because they represent either FAs consisting of less than three pixels or several FAs clustered together. FA fluorescence intensities were calibrated to the standard condition of 1 mW laser power with a 300 ms exposure time, so FA intensity should be directly proportional to protein level across all samples.

Statistical analysis

All graphs and statistical analyses were done using JMP and Matlab software. Distributions of FA properties were constructed in the following ways. FA number as described in the results section is more precisely a FA number per cell and consequently the distribution was generated by using the calculated FA number per cell at each time point during the experiment for each cell. Consequently, the number of measurements of FA number per cell is the product of the average number of frames and the cell number. All the other distributions of FA properties were generated by using the time-averaged FA property for each FA in each cell. Consequently, the number of measurements of FA properties is the product of the average FA number and the cell number. Differences between conditions under various EGF concentrations and cell migration speeds were quantified by calculating the Kolmogorov-Smirnov statistic using the Matlab function . Model distributions were fitted by minimizing the Kolmogorov-Smirnov statistic using the Matlab function between the experimental distribution and the model distribution. To determine the statistical differences of the mean values between conditions under various EGF concentrations and cell migration speeds, a student’s t-test was utilized and a

Competing interests

The authors declare that they have no competing interests.

Authors’ contributions

YH carried out the bulk of the experiments. SH conducted the long term migration experiments. YH and ICS analyzed the results and wrote the paper. All authors read and approved the final manuscript.

Acknowledgements

We thank the Roy J. Carver Charitable Trust for funding ICS and YH. We also thank Dr. Jeffrey Segall for cell lines, Dr. Antonio Sechi and Thomas Wurflinger for help with the FA analysis; Dr. Gaudenz Danuser and Shann-Ching Chen for help with the protrusion analysis and Nick Romsey for help with the cell migration analysis.