Cancer cell sedimentation in 3D cultures reveals active migration regulated by self-generated gradients and adhesion sites

Abstract

Cell sedimentation in 3D hydrogel cultures refers to the vertical migration of cells towards the bottom of the space. Understanding this poorly examined phenomenon may allow us to design better protocols to prevent it, as well as provide insights into the mechanobiology of cancer development. We conducted a multiscale experimental and mathematical examination of 3D cancer growth in triple negative breast cancer cells. Migration was examined in the presence and absence of Paclitaxel, in high and low adhesion environments and in the presence of fibroblasts. The observed behaviour was modeled by hypothesizing active migration due to self-generated chemotactic gradients. Our results did not reject this hypothesis, whereby migration was likely to be regulated by the MAPK and TGF-β pathways. The mathematical model enabled us to describe the experimental data in absence (normalized error<40%) and presence of Paclitaxel (normalized error<10%), suggesting inhibition of random motion and advection in the latter case. Inhibition of sedimentation in low adhesion and co-culture experiments further supported the conclusion that cells actively migrated downwards due to the presence of signals produced by cells already attached to the adhesive glass surface.

Fig 1

Investigation of migration mechanisms (a) Schematic representation of 3D cultures (center). The cells were uniformly distributed across the semi-ellipsoid on day 0 and actively migrated towards the glass bottom on day 14 (right panel). The examined hypotheses include passive and active migration mechanisms (left). (b) Cell distribution across different cell culture heights on days 0 and 14 in the presence (four different doses) and absence of Paclitaxel. Migration was inhibited for the three highest Paclitaxel doses, ruling out the possibility of passive migration. The glass bottom is located at z = 0μm. (c) (left) Cellular anatogram relating the gene expression activity to the corresponding cellular components. (right) Genes related to actin filaments, protein secretion, focal adhesion sites, and plasma membrane exhibited a monotonic decrease in their expression with respect to Paclitaxel dose. These cellular components are also involved in cell migration. (d) TGF-β (left) and MAPK (right) signalling pathways were downregulated as a function of the Paclitaxel dose. The two columns per condition belong to the two corresponding replicates. The inhibition of migration in combination with the downregulation of these pathways suggests that collective cell migration is more likely to occur than individual cell migration (i.e., epithelial-to-mesenchymal transition (EMT); Fig E of S1 File).

Fig 2. Schematic representation of the hypothesis of signal induced migration towards the bottom of the space.

Cells from the initial stages of the experiment that are close to the bottom attach to it and secrete signals to stimulate aggregation. Signal diffusion in the 3D space forms a gradient that decreases from the bottom to the top. Floating cells orient and migrate towards the direction of increased signal concentration.

Fig 3. Continuum model calibration.

(a) Posterior marginal distributions of the estimated model parameters for the examined experimental conditions. The NRMSE values for these conditions (bottom right) suggest overall low errors for all conditions, with the model yielding overall higher errors for the non-treatment condition during cell aggregation. (b) Surface plots of the in vitro and in silico density profiles for representative datasets on days 2 and 14. Paclitaxel was administered on day 5; hence, day 2 was not affected by treatment, while on day 14, we observed a difference in the spatial cell density distributions across the non-treatment and treatment conditions. The distributions remained relatively constant for the three highest doses because cell migration was inhibited, while for the lowest dose and control conditions. the distribution changed due to cell sedimentation. The green coloured profiles correspond to experiments and the blue to simulations.

Fig 4. Hybrid modelling results and spatial analysis.

(a) 3D plots of cell locations on day 14 for the control and treatment conditions. Red: experiments/Blue: simulations. (b) Predicted and experimentally observed cell numbers. The total cell population tended to decrease with increasing doses. (c) Histograms of the proportion of cells as a function of position in the culture across time for the experimental (top row) and simulated (bottom row) data, respectively. (d) Nearest Neighbour distances across the examined conditions with respect to time, for the experimental and simulated data. (e) Complete Spatial Randomness test; average values of the K-function across all samples and the corresponding standard error of mean (SEM).

Fig 5. Coating of the glass bottom with agarose.

(a) Schematic representation of the experimental procedure. The gray area represents the agarose deposited on the glass surface. The white disks denote the air bubbles injected to create pockets, the second row represents the removal of the inflated agarose surface, and the blue disks represent the cell/Matrigel mixture deposited on the pockets of the polymerized agarose. (b) Violin plots of the distribution of cells across the z-dimension of the cell culture for days 0 and 14 of the agarose coating experiment. Cell positions were overall maintained throughout the course of the experiment. (c) Boxplots of the cell distributions across the z-dimension of the control and the agarose coating experiments for day 14. The three asterisks denote p-value < 10⁻³ that was calculated using the Kruskal-Wallis test between the two distributions.

Fig 6. TNBC cells and fibroblasts co-culture.

Fibroblasts were suspended in the area surrounding the Matrigel. Cancer cells (green) and fibroblasts (magenta) were mixed at (a) the bottom of the space and (b) at 195 μm height. (c) Boxplots of cancer cell distribution across different culture heights for the co-culture and cancer cell monoculture for day 14. The three asterisks denote p-value < 10⁻³ that was calculated using the Kruskal-Wallis test between the two distributions.