Diversity of collective migration patterns of invasive breast cancer cells emerging during microtrack invasion

Understanding the mechanisms underlying the diversity of tumor invasion dynamics, including single-cell migration, multicellular streaming, and the emergence of various collective migration patterns, is a long-standing problem in cancer research. Here we have designed and fabricated a series of microchips containing high-throughput microscale tracks using protein repelling coating technology, which were then covered with a thin Matrigel layer. By varying the geometrical confinement (track width) and microenvironment factors (Matrigel concentration), we have reproduced a diversity of collective migration patterns in the chips, which were also observed in vivo. We have further classified the collective patterns and quantified the emergence probability of each class of patterns as a function of microtrack width and Matrigel concentration to devise a quantitive “collective pattern diagram.” To elucidate the mechanisms behind the emergence of various collective patterns, we employed cellular automaton simulations, incorporating the effects of both direct cell-cell interactions and microenvironment factors (e.g., chemical gradient and extracellular matrix degradation). Our simulations suggest that tumor cell phenotype heterogeneity, and the associated dynamic selection of a favorable phenotype via cell-microenivronment interactions, are key to the emergence of the observed collective patterns in vitro.

Figure 1

Illustration of the migration-on-track experiments. (a) The procedure to synthesize AAM PRC on glass substrates: An allyltrichlorosilane (ATC)-grafted glass substrate (a1) is further modified by AAM and BIS via photopolymerization under UV; subsequently the formed gridlike structure on the substrate (a2) is again cross-linked with PEG via BIS by photopolymerization; finally, the IPN structure (a3) is formed; (a4) the modified surface repels the cell from adhesion. (b) Schematic illustration of the experimental procedures. The tracks with various widths were established on the microchip (b1), and then cells were seeded in the pool for incubation. Meanwhile a PDMS block was applied to the tracks to prevent the cells from running into the tracks (b2). Matrigel with a specific concentration was dropped onto the tracks to mimic the variations of ECM in vivo, when the pool was filled by cells due to proliferation (b3). Cells invaded the tracks and exhibited various migration patterns (b4). (c) Before the migration experiment, the cells were held in the pool with a PDMS block on the tracks. (d) A typical fluorescence image by confocal microscopy showing a snapshot of the tumor cells migrating on microtracks with varying widths. The MDA-MB-231 cells marked with green fluorescent protein in panels (c) and (d) were obtained from China Infrastructure of Cell Line Resources (Beijing, China). The scale bar = 100 μm.

Figure 2

Categorization and clinical pathological slice probability statistics of the most commonly seen collective migration patterns of invasive tumor cells based on the analysis of pathological slices of breast cancers: (a) images of clinical slices and (b) corresponding cartoons showing four types of collective patterns—(1) stream, (2) blunt tip, (3) cone tip, and (4) lumen—and (c) the probability of the four types of patterns in the clinical pathological slices from 60 patients.

Figure 3

Quantitative analysis of different collective migration patterns emerging in the in vitro experiments with 60 repetitions, which resemble the collective patterns observed in vivo. The left column is the steady-state collective migration patterns from the migration-on-track experiments, followed by an illustration of the quantitative analysis of the patterns: (a) a stream pattern (the migration front width $d<2d_0$); (b) a blunt-tip pattern (the front angle $\theta >{\theta }_0$, and $d>2d_0$); (c) a cone-tip pattern ($\theta <{\theta }_0$ and $d>2d_0$); and (d) a lumen pattern (the area of the cavity $S>S_0$, which is roughly the area one and a half cells). Here $tan{\theta }_0=0.6, d_0=15\mu \mathrm{m}$, and $S_0\sim 1072\mu {\mathrm{m}}^2$. The right column is the corresponding color map for the probability of emergence for the four types of patterns vs track width $\omega$ and Matrigel concentration $\rho$, in which dark colors indicates higher probability. The scale bar shown is $100\mu \mathrm{m}$.

Figure 4

Collective pattern diagrams from migration-on-track experiments. Four distinct pattern regions are identified based on experimental data. The red, green, blue and purple regions represent the stream, blunt-tip, cone-tip, and lumen patterns, respectively.

Figure 5

Statistics of the speed at the invasive front of cancer cells. (a)–(d) The difference in speed according to pattern; the green points are the experimental data, and the pink line indicates the average speed. (e), (f) The energy cost in the cone-tip and blunt-tip patterns, $d$ represents the movement distance of the collective cells via Matrigel degradation, $l$ is the width of the collective cells. The red arrow shows the movement direction of collective cells. $S_c$ and $S_b$ (the shaded part surrounded by the dashed line) represent the area of the degraded Matrigel in the cone-tip and blunt-tip patterns, respectively. This assumes that the Matrigel is degraded by the high-activity cells or the leader of the collective cells. This shows that the area of $S_c$ is equal to $S_b$, indicating no difference in energy cost in both cases along the same distance of invasion.

Figure 6

Formation of distinct collective migration patterns associated with different microenvironmental factors (microtrack width and Matrigel density) via cellular automaton simulations.