Efficient deformation mechanisms enable invasive cancer cells to migrate faster in 3D collagen networks

Abstract

Cancer cell migration is a widely studied topic but has been very often limited to two dimensional motion on various substrates. Indeed, less is known about cancer cell migration in 3D fibrous-extracellular matrix (ECM) including variations of the microenvironment. Here we used 3D time lapse imaging on a confocal microscope and a phase correlation method to follow fiber deformations, as well as cell morphology and live actin distribution during the migration of cancer cells. Different collagen concentrations together with three bladder cancer cell lines were used to investigate the role of the metastatic potential on 3D cell migration characteristics. We found that grade-3 cells (T24 and J82) are characterized by a great diversity of shapes in comparison with grade-2 cells (RT112). Moreover, grade-3 cells with the highest metastatic potential (J82) showed the highest values of migration speeds and diffusivities at low collagen concentration and the greatest sensitivity to collagen concentration. Our results also suggested that the small shape fluctuations of J82 cells are the signature of larger migration velocities. Moreover, the displacement fields generated by J82 cells showed significantly higher fiber displacements as compared to T24 and RT112 cells, regardless of collagen concentration. The analysis of cell movements enhanced the fact that bladder cancer cells were able to exhibit different phenotypes (mesenchymal, amoeboid). Furthermore, the analysis of spatio-temporal migration mechanisms showed that cancer cells are able to push or pull on collagen fibers, therefore producing efficient local collagen deformations in the vicinity of cells. Our results also revealed that dense actin regions are correlated with the largest displacement fields, and this correlation is enhanced for the most invasive J82 cancer cells. Therefore this work opens up new routes to understand cancer cell migration in soft biological networks.

Figure 1

Effect of cell metastatic potential and collagen concentration on (A) migration speed, (B) power law exponent, (C) cell effective speed, (D) Mean directionality (at 0.95 mg/mL). Three cancer cell lines of increasing invasiveness (RT112 < T24 < J82) and collagen gel concentration (0.95 mg/mL, 1.8 mg/mL and 4.5 mg/mL) were used. Power-law exponents $\alpha$ in (B) were obtained from MSD(t) = D ${(t/t_0)}^{\alpha }$. Box-whisker plots indicate median value, 25% and 75% quartiles and whiskers extend to the 5% and 95% percentiles (A–C). Statistical significance from another collagen concentration * corresponds to p < 0.05, ** corresponds to p < 0.01, *** corresponds to p < 0.001. Significant difference (+) from RT112 cells (+ p < 0.05, ++ p < 0.01). Cell number: (A) and (C) For 0.95 mg/mL, N = 9, 10, 7 respectively for J82, T24, RT112 cells. For 1.8 mg/mL, N = 8, 9, 7. For 4.5 mg/mL, N = 10, 13, 9. (B) For 0.95 mg/mL, N = 8, 9, 7 respectively for J82, T24, RT112 cells. For 1.8 mg/mL, N = 7, 8, 7. For 4.5 mg/mL, N = 8, 9, 9. (D) Graph representing the mean directionality vs. time for the three cancer cell lines at 0.95 mg/mL.

Figure 2

(A) Mean sphericity index ($\psi$) for different cancer cells (RT112, T24 and J82) in various collagen gels (0.95 mg/mL to 1.8–4.5 mg/mL). *p < 0.05 (Kruskal Wallis test). Error bars represent mean ± SEM. (B) Mean major radius (R) of different cancer cells at low, intermediate and high collagen concentrations. (C) Graph representing contour fluctuations ($\nu$) vs. sphericity index $\psi$ . Insets show various shapes used by cancer cells of various invasiveness (blue for RT112 cells, green for T24 cells, red for J82 cells). (D) graph of contour fluctuations ($\nu$) vs. major radius (R) in μm. Solid line: power fit of $\nu$ = A + B * (R − C)${}^{1/2}$ with A = 10, 10, 10; B = 7.5, 10, 11; C = 14, 13, 11 respectively for J82, T24 and RT112 cells. (E) Migration speed (V) vs. contour fluctuations ($\nu$). Fits V = a*$\nu$ + b with a = 0.5, 0.25, 0.1 μm/h; b = 9.6, 7.3, 5.5 μm/h, for J82, T24 and RT112 cells.

Figure 3

Boxes (A–C): Z-projections (left panels) of confocal fluorescence images of the actin cytoskeleton for three migrating bladder cancer cells (RT112, T24 and J82 respectively) embedded in a 0.95 mg/mL collagen gel, at 10 min time interval (first position in yellow, second one in cyan). Superposition of collagen fiber images (right panels) at these times, for one slice level in the image stack, with indication of cell contours. Scale bar = 20 μm. The white arrow indicates the rear of the T24 cell (B) or the long cylindrical protrusion displayed by the J82 cell (C). The red dotted arrow indicates the migration direction. Boxes (D–I): Corresponding 3D collagen fiber displacements around each migrating bladder cancer cells: RT112 (D,G), T24 (E,H), J82 (F,I). Two viewing angles have been selected. The initial 3D cell shape is shown in grey levels. The vector length and color indicate the displacement magnitude in μm. The x, y and z grids are in μm. The red dotted arrow indicates the migration direction. (J–L) Angle distributions of displacement vectors—with respect to the direction of migration—shown in red (resp. blue) for vectors located at the front (resp. at the rear) edge.

Figure 4

T24 cell migrating in 0.95 mg/mL collagen. (A) and (C) Two successive images of F-actin (projected Z-stack fluorescence). White half-circles indicate regions of interest. The white arrow indicates the direction of migration whereas the white dotted arrow indicates actin-rich regions inside the nucleus. (B) and (D) show the collagen fiber 3D-displacement field around the migrating cell. For more visibility, the initial 3D cell shape is shown in grey levels and the displacement field is shown for a half space. Vector lengths and colors indicate the displacement magnitude in μm. Scale bar = 20 μm. (E) and (G) Box and whisker plots of fiber displacement vector angles versus cell migration direction for regions of interest. (F) Maximum displacement amplitudes as a function of cell-fiber distance for the actin-rich regions 1 and 4 in (A).

Figure 5

Correlation between actin intensities and displacement magnitudes (J82 cell). (A,B) Values of IWDs are shown at two different times, 20 min, and 30 min. Symbols represent low and high IWDs, respectively [0–25%] and [75–100%] ranges. (C) Violin plots of the displacement norm distributions corresponding to low (blue) and high (green) IWD values in time. (D) Median of the norms (low and high IWDs) vs. time, with significant difference from Kruskal-Wallis test. (E) Corresponding cell shapes. (F) Displacement norms vs. weighted actin intensities (IWD) for t = 20 min. The red line is the regression curve. (G) Slopes of regression curves at times for which Norm${}_{\mathit{high}}$/Norm${}_{\mathit{low}}$> 1, for the three cell types. Significance of Kruskal-Wallis test, followed by Dunn’s test with Bonferroni adjustment.