Examining Go-or-Grow Using Fluorescent Cell-Cycle Indicators and Cell-Cycle-Inhibiting Drugs

Abstract

The go-or-grow hypothesis states that adherent cells undergo reversible phenotype switching between migratory and proliferative states, with cells in the migratory state being more motile than cells in the proliferative state. Here, we examine go-or-grow in two-dimensional in vitro assays using melanoma cells with fluorescent cell-cycle indicators and cell-cycle-inhibiting drugs. We analyze the experimental data using single-cell tracking to calculate mean diffusivities and compare motility between cells in different cell-cycle phases and in cell-cycle arrest. Unequivocally, our analysis does not support the go-or-grow hypothesis. We present clear evidence that cell motility is independent of the cell-cycle phase and that nonproliferative arrested cells have the same motility as cycling cells.

Significance

Under the go-or-grow hypothesis, a cell is either migrating or proliferating, but never both simultaneously; the migrating cell is not expending energy proliferating, so it is more motile than the proliferating cell. Here, we test go-or-grow for adherent melanoma cells and find that our data do not support the hypothesis.

Main Text

The “go-or-grow” hypothesis, also referred to as the “phenotype switching model” or the “migration/proliferation dichotomy,” proposes that adherent cells reversibly switch between migratory and proliferative phenotypes (1), exhibiting higher motility in the migratory state because motile cells are not using free energy for proliferation (1, 2, 3, 4, 5). Previous experimental investigations of the go-or-grow hypothesis are conflicting because some studies support the hypothesis (1,6,7), whereas others refute it (8, 9, 10).

Go-or-grow was initially proposed as an explanation for the apparent mutual exclusivity of migration and proliferation for astrocytoma cells, first in two-dimensional (2-D) in vitro experiments (7) and later for in vivo investigations (6). In these early studies, claims for evidence of go-or-grow are based on the comparison of the subpopulation of cells at the perimeter of the cell population, where cells are considered to be invasive, with the subpopulation of cells in the central region, where cells are considered noninvasive. Data suggest that the proliferation rate is lower at the perimeter and higher in the center, leading to the assertion that the more migratory cells are less proliferative. The experimental data, however, only indicate that the subpopulation at the perimeter is less proliferative as a whole compared with the center, and therefore, we cannot conclude definitively that the more migratory cells are less proliferative.

To test for evidence of go-or-grow, it is necessary to look at the single-cell level, as is done in subsequent studies (8, 9, 10) in which single-cell tracking is used with single-cell migration, measured in terms of the net displacement of the cell trajectory. These three studies, none of which support go-or-grow, involve 2-D and three-dimensional (3-D) in vitro experiments with medulloblastoma cells (10); 2-D in vitro experiments with mesothelioma, melanoma, and lung cancer cells (9); and 2-D and 3-D in vitro experiments with melanoma cells (8). Studies of tumor heterogeneity in melanoma suggest that cells may reversibly switch between invasive and proliferative phenotypes (1). Because melanoma is highly metastatic, forms tumors that are very heterogeneous, and is well known to respond to mitogen-activated protein kinase (MAPK) inhibitors that induce G1 arrest (11,12), melanoma cells are a prime candidate for studying the go-or-grow hypothesis.

Confirmation of go-or-grow would have important implications for anticancer treatments employing cell-cycle-inhibiting drugs. For most eukaryotic cells, the cell cycle is a sequence of four discrete phases (Fig. 1 a)—namely, gap 1 (G1), synthesis (S), gap 2 (G2), and mitosis (M). Cell-cycle arrest (Fig. 1 d), which occurs when progression through the cell cycle halts (13), can be induced by cell-cycle-inhibiting drugs (8,14,15). An arrested cell is not proliferative, so the cell’s free energy could be utilized for migration, potentially leading to an exacerbation of metastasis (3).

Figure 1.

Experimental data and mean diffusivities. (a) The cell cycle, indicating the color of FUCCI in each phase. (b and c) Experimental images of cycling C8161 cells; cell counts at 0 and 48 h are 331 and 1878, respectively. (d) The cell cycle, indicating the color of FUCCI in each phase together with arrest in G1. (e and f) Experimental images of C8161 cells in G1 arrest (30 nM trametinib); cell counts at 0 and 48 h are 261 and 469, respectively. (g–i) 50 cell trajectories of G1 cycling, S/G2/M cycling, and G1-arrested (30 nM trametinib) C8161 cells, respectively. (j–l) There is no difference in mean diffusivity, $\langle D\rangle$, for C8161, WM983C, and 1205Lu cells, respectively. For each 2-h time interval, $\langle D\rangle$ is the mean of all individual diffusivities D corresponding to cells with trajectories within the time interval. In each case, we show $\langle D\rangle$ and report the variability using $\langle D\rangle$ plus or minus the sample standard deviation. Data for each experimental condition are offset with respect to the time-interval axis for clarity. Scale bars, 200 μm. To see this figure in color, go online.

The go-or-grow hypothesis also has important implications for mathematical models of collective cell invasion in a population of migratory and proliferative cells. Such models of cell invasion are often based on the Fisher-Kolmogorov-Petrovskii-Piskunov equation (16, 17, 18, 19),

$$\frac{\partial u}{\partial t}=D\frac{{\partial }^{2}u}{\partial x^2}+\lambda u\left(1-\frac{u}{K}\right),$$ (1)

where x is position, t is time, $u\left(x,t\right)>0$ is cell density, $D>0$ is the diffusivity, $\lambda >0$ is the proliferation rate, and $K>0$ is the carrying-capacity density. Equation 1 and related adaptations, including stochastic analogs (20,21), have been successfully used to model cell migration in vitro and in vivo (22, 23, 24, 25, 26). A key assumption underlying these models is that D is independent of the cell-cycle phase, which may not hold if cells are subject to go-or-grow because a cycling, and therefore nonarrested, cell may then become less motile as it progresses through the cell cycle and nears cell division (8).

In this work, we rigorously examine the go-or-grow hypothesis for adherent melanoma cells, for which phenotype switching between migratory and proliferative states is proposed to occur (1). We use melanoma cell lines in this study because melanoma is the prototype for the phenotype switching model and is highly responsive to G1 arrest-inducing mitogen-activated protein kinase kinase (MEK) inhibitors, such as trametinib. Melanoma cells are therefore an ideal candidate for studying go-or-grow (1,3,27). Our experimental data are obtained from single-cell tracking in 2-D in vitro assays. We conduct our experiments in 2-D before utilizing the knowledge gained in more complicated 3-D or in vivo experiments because it is the natural situation in which to commence a new experimental study. Indeed, experimental studies of cell migration are often conducted in 2-D in vitro assays for several reasons: the observed cell migration is partly representative of cell migration in vivo; the assays are amenable to standard laboratory techniques, such as live-cell microscopy; and image analysis, such as cell counting and single-cell tracking, is relatively easy (28, 29, 30). Furthermore, cell migration in 3-D may be affected by the properties of a 3-D matrix, which is not present in 2-D assays. For example, cell migration in 3-D through constricting pores can damage the nucleus and thereby cause a delay in cell division as the nucleus undergoes repair, which could be interpreted incorrectly as evidence for go-or-grow (31).

We employ fluorescent ubiquitination-based cell cycle indicator (FUCCI) (32), which consists of two reporters enabling visualization of the cell cycle of individual live cells: when the cell is in G1, the nucleus fluoresces red, and when the cell is in S/G2/M, the nucleus fluoresces green (Fig. 1 a). During early S, both of the red and green reporters are active producing yellow. FUCCI allows us to study cell motility in G1 separately from cell motility in S/G2/M (8,22,33,34). Specifically, we investigate cycling cells for differences in motility when the cells are in G1 compared with S/G2/M. Furthermore, given the potential for an arrested cell to become more motile, we use a cell-cycle-inhibiting drug to effect G1 arrest and compare the motility of the arrested cells with cycling cells. Note that FUCCI does not provide delineation of S, G2, and M, so our motility measurements for these phases are combined into S/G2/M.

Our methodology for examining go-or-grow is novel in a number of ways. We induce G1 arrest in cells to determine whether nonproliferative cells have higher motility than cycling cells. We use experimental data to show that our three cell lines have distinctly different cell-cycle durations, ratios of duration in G1 to S/G2/M, and migration characteristics, all of which may affect motility under the go-or-grow hypothesis. Importantly, the data set we generate and analyze is large: for each cell line and experimental condition, we randomly sample 50 single-cell trajectories for analysis out of more than ${10}^3$ trajectories. In total, we analyze 450 carefully collected trajectories for evidence of go-or-grow. Using these trajectories, we carefully estimate diffusivities by first accounting for anisotropy in the cell migration so that our estimates are based on time frames for which the cells are undergoing free diffusion.

Our data consist of time-series images, acquired every 15 min for 48 h, from 2-D proliferation assays using the melanoma cell lines C8161, WM983C, and 1205Lu (8,22,35,36), which have respective mean cell-cycle durations of ∼21, 23, and 37 h (8). The cell lines have very different ratios of durations in G1 to S/G2/M (Supporting Material; Data S1, S2, and S3). Fig. 1, b and c shows images of an assay with cycling C8161 cells at 0 and 48 h, illustrating the red, yellow, and green nuclei, corresponding to cells in G1, early S, and S/G2/M, respectively. For comparison, Fig. 1, e and f shows images of an assay with G1-arrested C8161 cells treated with the cell-cycle-inhibiting drug trametinib (30 nM), illustrating that most cells are arrested in G1, appearing red. We use the lowest possible concentration of trametinib to induce G1 arrest for the experiment duration to minimize other effects. Consequently, each cell eventually returns to cycling, illustrated by the small proportion of green cells (Fig. 1, e and f). These few green cells will eventually divide, with both daughter cells arresting in G1. We quantitatively confirm the G1 arrest by comparing the cell counts between the experiments with cycling cells and arrested cells. For the cycling cells, there is a 5.7-fold increase in the number of cells over 48 h (Fig. 1, b and c), whereas there is only a 1.8-fold increase in the number of arrested cells over 48 h (Fig. 1, e and f). The 1.8-fold increase in the population of G1-arrested cells is expected because we use the lowest possible concentration of trametinib. Consequently, a small subpopulation of cells may not be arrested at the start of the experiment, and cells may recommence cycling during the experiment, producing a small increase in the population.

For each cell line, we employ single-cell tracking to obtain 50 trajectories of cells for each experimental condition: 1) G1 cycling, 2) S/G2/M cycling, and 3) G1 arrest. Each trajectory is selected randomly without replacement from the set of all trajectories for a given cell line and experimental condition. For the cycling cells, trajectories are recorded for the complete duration of the G1 or S/G2/M phase. For the G1-arrested cells, the duration of the trajectory corresponds to the maximal duration that the cell is arrested within the 48-h duration of the experiment (Supporting Material).

In Fig. 1, g–i, we visualize the trajectories for cycling C8161 in G1 and S/G2/M, and C8161 in G1 arrest. The trajectories are translated so that their initial positions are at the origin. The trajectories of the G1-arrested cells are generally much longer than those for the cycling cells because the arrested cells reside in G1 for a much longer duration than cycling cells reside in G1 or S/G2/M. Specifically, the approximate mean duration of cycling C8161 cells in G1 is 5 h; for cells in S/G2/M, it is 6 h (8); and for cells in G1 arrest during the 48 h of the experiment, it is 34 h (Supporting Material). Therefore, to easily compare the trajectories of G1-arrested cells with cycling cells in G1, we show within the inset the truncated trajectories of the G1-arrested cells. The trajectories are truncated to a duration equal to the mean duration of the corresponding trajectories for cycling cells in G1. Based on these data, the migration is isotropic, without any drift, and independent of the cell-cycle phase. We now quantify these observations.

For each cell line and experimental condition, we find that the cell migration is isotropic and that directional persistence dissipates within a relatively short lag time of 1 h (Supporting Material). From each individual cell trajectory, we estimate D using the mean-square displacement as a function of lag time within 2-h time intervals. The intervals begin at the initial point of the trajectory, t = 0 h, with successive intervals offset by 1 h. We always use lag times from 1 to 2 h to guarantee the absence of persistence (Supporting Material). We then calculate the mean diffusivity $\langle D\rangle$ for each time interval by averaging our estimates of D for those trajectories that extend to the end of that interval.

Fig. 1, j–l shows, for each cell line, $\langle D\rangle$ for successive time intervals. From these data we arrive at clear conclusions (Supporting Material), none of which are consistent with the go-or-grow hypothesis:

• •For each cell line and experimental condition, there is little variation in $\langle D\rangle$ over time, indicating that there is no appreciable change in motility during each cell-cycle phase and during G1 arrest (Supporting Material).
• •For each cell line, there is little variation in $\langle D\rangle$ between cycling cells in G1, cycling cells in S/G2/M, and G1-arrested cells. The lack of variability in $\langle D\rangle$ is remarkable and clearly demonstrates that cells in G1 are not more motile than cells in S/G2/M and that G1-arrested cells at no time become more migratory than the cycling cells.
• •Even though our three cell lines have very different proliferation and migration characteristics (Supporting Material), our estimate of $\langle D\rangle$ is remarkably consistent across the three very different cell lines.

In summary, our analysis of cell migration in 2-D assays using three melanoma cell lines does not support the go-or-grow hypothesis. We find that cell motility is independent of the cell-cycle phase, so the implication from go-or-grow that cells are more motile in G1 than in S/G2/M when they are nearing cell division is not supported by our data. Notably, there is no change in cell motility when we effect drug-induced G1 arrest in the cells, again displaying a lack of support for the go-or-grow hypothesis.

Author Contributions

All authors designed the research. S.T.V. performed the research. All authors contributed analytic tools and analyzed the data. S.T.V. wrote the manuscript, and all authors approved the final version of the manuscript.

Editor: Dunn Alexander.

Supporting Material

Data S1. Trajectory Data for the C8161 Cell Line

50 cell trajectories for each condition: cycling cells in G1; cycling cells in S/G2/M; G1-arrested cells; G1-arrested cells (truncated trajectories).

Data S2. Trajectory Data for the WM983C Cell Line

50 cell trajectories for each condition: cycling cells in G1; cycling cells in S/G2/M; G1-arrested cells; G1-arrested cells (truncated trajectories).

Data S3. Trajectory Data for the 1205Lu Cell Line

50 cell trajectories for each condition: cycling cells in G1; cycling cells in S/G2/M; G1-arrested cells; G1-arrested cells (truncated trajectories).