Cancer Cells Display Increased Migration and Deformability in Pace with Metastatic Progression

Abstract

In this study, we explored the relation between metastatic states versus the capacity of confined migration, amoeboid transition and cellular stiffness. We compared across an isogenic panel of human breast cancer cells derived from MDA-MB-231 cells. It was observed that cells after lung metastasis have the fastest migration and lowest stiffness, with a significantly higher capacity to transition into amoeboid mode. Our findings illustrate that metastasis is a selective process favoring motile and softer cells. Moreover, the observation that circulating tumor cells resemble the parental cell line, but not lung-metastatic cells, suggests cells with higher deformability and motility are likely selected during extravasation and colonization.

Introduction

Metastasis, a process during which cancer cells dissociate from a primary tumor to navigate through interstitial tissues and ultimately colonize distant organs, is responsible for about 90% of cancer-associated deaths (1). To establish metastatic foci, cancer cells of solid tumors disseminate from the primary tumor site, intravasate into blood or lymphatic vessels, circulate through the body, extravasate to distant sites and proliferate to form secondary tumors (2). Certain phenotypes are required to overcome the physical barriers presented to cancer cells during metastasis. For example, intravasation and extravasation require cells to deform considerably in order to pass through the narrow interstitial space in the endothelium and epithelium of the vessel wall. Moreover, cell locomotion through confining spaces is a pivotal step in the metastatic dissemination of cancer cells from a primary tumor to distant organs in the body. Thus, cell motility, an active process dependent on cytoskeletal dynamics, and the cellular deformability, an attribute dependent on the cytoskeletal organization, are crucial for metastasis. We hypothesize that cells which manifest the favorable phenotypes in cell motility and deformability in terms of metastasis are selected and drive tumor progression. Indeed, it has been reported that during the dissemination stage of metastasis in carcinoma of epithelial origins, cancer cells undergo epithelial-mesenchymal transition (EMT). During this process, cells adopt a more elongated cell shape and develop actin-rich protrusions such as lamellipodia and filopodia, traits belonging to highly motile mesenchymal cells (3).

In addition to cell mobility, deformability is equally critical for cancer cells to navigate the confined space in the tissues during metastasis (4). Cellular deformability correlates to the viscoelasticity of the cell, which could be probed by atomic force microscopy (AFM), micropipette aspiration and magnetic tweezers (5–7). These studies demonstrated an inverse trend between cell stiffness and metastatic potential. However, the subjects compared in these studies are cells from different individuals or cell lines of heterogeneous genetic backgrounds. Thus, the involvement of inter-individual genomic heterogeneity calls into question the certainty of the correlation between cell stiffness and metastatic potential.

To address this uncertainty, we studied an isogenic cell line panel to represent three progressively metastatic stages: parental human breast cancer cells MDA-MB-231, circulating tumor cells (CTCs) and lung-metastatic cells (LMCs). The latter two were derived from the parental cell line using a xenograft approach, as reported previously (8). We examined the migration of the three cell lines under varied confinement, in mimicry of the interstitial space in the tissue. The cellular stiffness was also measured. We found that LMCs exhibit significantly increased motility and decreased stiffness, which is in line with our postulation that metastasis is a selective process favoring motile and softer cells.

Materials and Methods

Cell culture

The human breast cancer cell line MDA-MB-231 was used to generate CTCs and LMCs in a mouse model as previously described (8). Cells were cultured in RPMI 1640 with L-glutamine (Corning Cellgro) supplemented with 10% fetal bovine serum (Corning Cellgro), 100 U/ml penicillin and 100 U/ml streptomycin (Corning Cellgro) and maintained in a humidified incubator in 5% CO2 at 37 °C.

Magnetic tweezers assay

The magnetic tweezers were a generous gift from the Superfine group at the University of North Carolina (9). The operation of magnetic tweezers was performed by a customized user interface written in MATLAB. The field strength was calibrated prior to experiments by recording the displacement of the 2.8-μm dynabeads (Thermofisher) suspended in DMPS (MilliporeSigma) as a function of the distance from the tip of the tweezers. A 40x objective on Nikon Eclipse TE2000-U microscope (Nikon), Rolera EM-C² EMCCD camera (QImaging) and Micro-Manager (open source software program) were used to record movies of particles (10). The velocity of the particles was calculated based on the displacement. Assuming low Reynolds number the force magnitude was computed using Stokes’ Law: F=6πηRv, where η represents the dynamic viscosity of the fluid (known for DMPS), R the bead radius and v bead velocity. The force magnitude and the distance from the tip were plotted against each other and fitted to a power equation (Fig. S4).

For cellular stiffness measurements, cells were seeded on a fibronectin-coated coverslip and cultured overnight. Tosyl-activated 2.8 μm dynabeads coated with Fibronectin (MilliporeSigma) were then added to bind through integrins on the plasma membrane to the cytockeleton. After 30-min incubation, unbound particles were removed by a gentle wash with PBS. During experiments, only beads satisfying the following criteria were selected for further study: 1) only one bead is attached to the cell under study; 2) the bead is not situated on the cell nucleus; 3) after applying magnetic forces, the bead moves toward the magnetic tweezer tip, and shows a proper viscoelastic displacement profile (Fig. 3b). The displacement of the particles r(t) was recorded by time-lapse imaging during the application of the magnetic forces (191±20 pN) for 5s. Cell compliance was calculated as J(t)=6πRr(t)/F (11), where R represents the bead radius, r(t) the particle displacement, F the force applied. The calculated compliance was then fitted to a Jeffrey model (Fig. 3c), and the spring constant in the fitted model was reported as cell stiffness.

Fig. 3.

Stiffness measurement of the cells by magnetic tweezers. (a) A cell-bound magnetic particle is being pulled toward the tip of the magnetic tweezers. The contour marked by the dashed line traces the cell. (b) The magnetic force applied to the particle and the corresponding particle displacement profile over time are shown. (c) Cell compliance, directly calculated from the particle displacement, was fitted to Jeffrey’s model to calculate the elastic modulus and the viscosity of the cell. (d) The elastic moduli are reported as stiffness for the three cell lines. (Scale bar: 15 μm)

Microchannel migration assay

Microfluidic devices were designed and fabricated as described previously (12–14). Briefly, silicon wafers (Wafer World, Inc.) with microchannel patterns of varying dimensions (width x height: 3x5, 3x10, 6x10, 10x10, 20x10 μm²) were made by photolithography. The channels all have the same length (200 μm). On the silicon wafers, polydimethylsiloxane (PDMS, Sylgard 184) was polymerized at 85 °C for about 1.5 hours after all air bubbles were removed by vacuum. A 5-mm punch was used to make six holes in the polymerized PDMS devices for seeding cells and adding media, and excess PDMS was cut off using a razor blade. The patterned PDMS devices were sonicated in 100% ethanol, cleaned by 100% ethanol and DI water, activated by plasma cleaner (Harrick PDC-32G plasma cleaner) and bonded to 25x75 mm glass slides (Electron Microscopy Sciences) to create microchannels (Fig. S5).

The channels were first coated with 20 μg/mL type-I rat tail collagen (BD Biosciences) for 1 hour at 37 °C, 5% CO₂, and 100% humidity. 20 μL of cell-containing medium with a density of 1,000 cells/μL was added to the cell seeding well. The seeded cells were cultured at 37 °C, 5% CO2, and 100% humidity for °20 minutes until the cells were fully spread on the substrate close to the microchannel entrances. After briefly aspirating media in all wells, the cell migration experiments were performed with serum free, 1% (v/v) penicillin/streptomycin (P/S. Gibco) RPMI added to the two wells on the entrance side of the channels, and with 10% (v/v) fetal bovine serum (FBS, Life Technologies/Gibco), 1% (v/v) P/S RPMI added to the four wells on the exit side of the channels, creating a chemogradient. Time-lapse images were taken by an inverted Eclipse Ti Microscope every 20 min for 24 hours at 37 °C, 5% CO₂, and 100% humidity.

To calculate the percentage entry of cells, the number of cells that formed protrusions into the channel interior was first counted, then the percentage among those cells that went on to fully enter the channels was quantified. Cell tracking was performed in ImageJ (U.S. National Institutes of Health, Bethesda MD, USA) with a manual tracking method, and data from ImageJ were extracted to Excel files and were further analyzed with MATLAB codes to obtain average velocity, average speed, and persistence for each cell. The velocity was calculated as the total net cell displacement divided by the total time; the speed was calculated as the total distance traveled divided by the total time; the persistence was calculated as total displacement divided by the total distance traveled, or equivalently, velocity divided by speed.

Cell staining and blebbing quantification

Cells were seeded onto glass-bottom dishes (Cellvis) and cultured for ~36h. At around 30% confluence, the cells were fixed with 4% Paraformaldehyde (PFA) for 10 min and permeabilized with 0.1% Triton X-100 (MilliporeSigma) in PBS for 10 min. Then F-actin was stained using Phalloidin-488 (Cytoskeleton, Inc.). Cells were also stained with DAPI for reference. Z-stack fluorescence images were taken under a confocal microscope (Leica TCS SP8) using a 63x oil-immersion objective (NA 1.4). Acquired images were processed and analyzed using ImageJ. To quantify bleb size, the largest bleb on each cell was manually identified in the F-actin fluorescent images. The bleb was then treated as an approximate sphere and the diameter was measured. A detailed step-by-step procedure of image analysis is provided in the Supplementary Information.

Actin dynamics

For live cell measurements, cells were transfected with F-tractin-GFP, following the standard protocol of Lipofectamine 3000 (Thermofisher). Time-lapse images were taken using the same confocal microscope at the sampling rate of 3 seconds per frame.

Ki67 immunofluorescence

Cells were seeded onto glass-bottom dishes (Cellvis) with 40% to 60% confluent and incubated for 24 hours. The cells were fixed for 10 minutes with 4% paraformaldehyde and permeabilized for 10 minutes with 1% Triton X-100. The permeabilized cells were blocked with 2% BSA, 1% goat serum, and 0.1% Triton X-100 in PBS−/− for one hour at room temperature. The cells were stained with the primary anti-Ki67 antibody (clone 8D5, Cell Signaling Technology, product #9449, and 1:800 dilution) in the same blocking buffer overnight at 4^oC. The cells were stained with the Alexafluor 488 goat anti-mouse secondary antibody (Invitrogen, 1:100) and Hoeschst 33342 (Invitrogen, 1:2000) in the same blocking buffer for one hour at room temperature. The samples were imaged on an epifluorescent microscope (Eclipse Ti; Nikon) using Plan Fluor 10x/0.30. The images were analyzed with the NIS element software and ImageJ by manually counting the number of cells showing with or without Ki67 signals.

Statistical analysis

Statistical significance of differences between two groups was calculated using a Wilcoxon rank sum test in MATLAB. Effect size was reported using the formulation of Cohen’s d: d=(μ₁-μ₂)/s, where μ₁ and μ₂ are means of the two samples, and s is the pooled standard deviation.

Results

LMCs display a higher propensity for entry into microchannels and faster migration velocities

Cells in vivo travel through confining three-dimensional (3D) pores varying from 1 to 20 µm in diameter, or fiber- and channel-like tracks ranging from 3 to 30 µm in width (15, 16). Thus, we first examined whether there is any difference in motility between parental cells, CTCs and LMCs over a range of physiologically relevant dimensions. To this end, PDMS-based microchannels (Fig. 1a) of different cross-sectional areas (width x height (WxH): 3x5, 3x10, 6x10, 10x10, 20x10 μm²) were fabricated, and a chemogradient of fetal bovine serum (FBS) was established along the channel lengths. As the channel size gets smaller in both the width and the height, different degrees of confinement are imposed on cells. Cells inside 3x10 µm² channels contact all four channel walls, and are thus defined as confined (17, 18). Hence, we refer to the smallest channels tested here (3x5 µm²) as tightly confining channels. It is worth noting that this microfluidic assay is physiologically relevant for all three cell types. The parental cells and LMCs go through stromal invasion at primary or metastatic tumor sites. CTCs need to arrest at a distant organ site and extravasate out of the blood vessels, both of which are critical milestones of metastasis, and require them to be adherent. We noted that the parental cells, CTCs and LMCs entered the microchannels (Movie S1, S2, S3) with different propensities. While 75 ± 7 % LMCs entered the tightly confining (WxH=3x5 µm²) microchannels, significantly lower percentages were observed for the parental cells (51 ± 9 %) and CTCs (44 ± 12 %) (Fig. 1b). Among the cells that fully entered the channel, the time and the net displacement between initial positions and final positions were measured for each cell and are used to calculate the migration velocity. While all three cell types showed a positive velocity toward the other side of channels in the direction of chemogradient of FBS, LMCs were generally faster than parental cells and CTCs in all the channels of various dimensions (Fig. S1). In particular, for the 3x5 μm² microchannel, LMCs migrated at 44.9 ± 28.6 μm/h, 46% faster than the parental cells (30.8 ± 21.6 μm/h) and 63% faster than CTCs (27.6 ± 19.4 μm/h), respectively (Fig. 1c). This agrees with the notion that migration velocity represents a key aspect of metastatic capacity (19).

Fig. 1.

Microchannel migration assay. (a) A series of phase contrast time-lapse images showing a LMC migrating through a 3x5 μm² microchannel. Red arrows indicate the front end of the cell, and green arrows indicate the rear end of the cell. (b) Quantification of cell entry into the microchannels. (c) Cell migration velocity within microchannels. (d) Average migration speed of cells within microchannels. (e) Quantification of directional persistence as velocity divided by speed. * P<0.05, ** P<0.01, *** P<0.001, **** P<0.0001, n.s. not significant. d is the effect size calculated as Cohen’s d. These indications apply to all the figures. (Scale bar: 30 μm)

Cells do not always move in the forward direction in the microchannels, as they sometimes stall and sometimes move backward. Therefore, the migration velocity is determined by the average speed and the directional persistence of the cells while moving. We evaluated both average speed and directional persistence to dissect how these two parameters contribute to the overall higher velocity observed in LMCs. We found that LMCs migrated with higher average speed (48.6 ± 29.0 μm/h), compared to the parental cells (32.8 ± 18.4 μm/h) and CTCs (32.0 ± 16.3 μm/h) (Fig. 1d), as well as higher directional persistence when moving through 3x5 μm² microchannels. Specifically, 82% LMCs exhibited persistence value > 0.95, whereas only 67% parental cells and 54% CTCs did (Fig. 1e). Interestingly, when moving through microchannels of larger dimensions, the persistence of parental cells and CTCs increased and became comparable to that of LMCs. (Fig. S2).

Amoeboid motility mode is more prevalent in LMCs

Cancer cells are known to transition from focal adhesion (FA)-mediated, mesenchymal mode to FA-independent, amoeboid mode of cell motility, a process known as mesenchymal-amoeboid transition (MAT). Cells in the mesenchymal mode migrate by forming protrusions composed of branched actin filament meshwork in the lamellipodia at the leading edge (20), while cells in the amoeboid mode form blebs that are initiated by local detachment between the plasma membrane and cortical actin, and expand forward under intracellular pressure (21). It was reported that amoeboid migration allows cells to squeeze through pre-existing gaps without degrading extracellular matrix (ECM) (22).

Based on the observation that LMCs entered the microchannels with higher probability, we postulated that LMCs are more prone to adopt the amoeboid motility. To verify our postulation, we proceeded to quantify the proportion of cells with amoeboid morphology in each cell line. Previous studies have shown that amoeboid mode is mostly observed when cells are in confinement, such as microchannels, or on the substrate of low adhesion (23). Microchannels, simulating the confined interstitial spaces during metastasis, might impose pre-selection on amoeboid cells in terms of the bleb size and the bleb number. Since we aimed to survey the whole population of each cell line without preselection imposed by microchannels, we chose to measure blebs of the cell on low-adhesion substrates to achieve unbiased bleb characterization of each cell line. Cells were cultured on glass without ECM coating for 24~36 hours before being fixed and stained with the F-actin dye phalloidin and imaged. Phalloidin staining facilitated the detailed visualization of the distinct amoeboid morphology, where the plasma membrane detaches from the actin cortex and forms blebs (Fig. 2a). In addition, to exclude the possibility that cells might secrete ECM protein to enhance cell-substrate adhesion, immunofluorescence was performed to confirm that the ECM protein, fibronectin, was not deposited by cancer cells after the time of culture (Fig. S3). Image analysis results showed that 71.6% of LMCs exhibited the blebbing amoeboid morphology, whereas only 45.7% and 47.2% of the parental cells and CTCs, respectively, were amoeboid (Fig. 2b). Moreover, we measured the diameters of the blebs, and found that larger blebs formed in the amoeboid LMCs (3.33 ± 1.51 μm) than the parental cells (1.82 ± 0.70 μm) and CTCs (2.27 ± 0.89 μm) (Fig. 2c).

Fig. 2.

Migratory modes of cells on less adhesive 2D substrates. (a) Representative images of cells without blebs, with small blebs and with a large bleb. Images in each column are of one cell, where the first row shows bright field images. The second and third rows show F-actin staining by fluorescently labeled Phalloidin at different Z positions. Blebs are identified by clear, bright circular boundaries on the dorsal side of the cells, as highlighted by red arrows. (b) Ratio of cells in amoeboid and mesenchymal modes for the three cell types. (c) Quantification of bleb size for all cells in the amoeboid mode. (d) A representative image of a LMC in the amoeboid mode. The cell was transfected with F-tractin-GFP. The timestamped snapshots show the formation and retraction of the bleb in the region marked by the dashed square. (e) Brightfield and fluorescence images of a cell that transitioned from mesenchymal to amoeboid motility in 6 minutes 48 seconds, and went through the reverse transition in the next 57 seconds. The membrane ruffles, which are characteristic of mesenchymal modes, and blebs, which are characteristic of amoeboid modes, are marked by the dashed square, and their dynamics are shown in the timestamped snapshots in the third row. (Scale bar: 5 μm)

To study the dynamic transition between mesenchymal and amoeboid modes of cell motility, we transfected cells with F-actin binding protein, F-tractin. Under live cell time-lapse microscopy, the formation and retraction of the blebs can be visualized in detail (Fig. 2d). The mesenchymal and amoeboid motility modes can be distinguished by the presence of membrane ruffles and blebs, respectively, and we observed reversible transitions between the two modes within several minutes (Fig. 2e, Movie S4, S5). This shows that transition into amoeboid motility is a rapid and reversible process. Taken together, our results suggest LMCs have a relatively higher propensity to transition into the amoeboid mode of cell motility, a mode known to be faster than the FA-mediated mesenchymal migration (24).

LMCs have lower stiffness

The higher probability of LMCs to enter microchannels indicate higher deformability, or equivalently lower stiffness. To test the hypothesis, magnetic tweezers were used to measure the stiffness of the cells (9). 2.8-μm paramagnetic particles were conjugated with fibronectin to bind integrin on the cell surface, thus probing the stiffness of actomyosin cortex, which accounts for most of the cellular stiffness (25, 26). The deformation of the cells induced by the magnetic forces could then be recorded by tracking the paramagnetic particles. The displacement of particles (Fig. 3b) bound to the cells upon applying the external magnetic field (Fig. 3a), was used to calculate the stiffness of the cells by fitting the displacement data to Jeffreys model (Fig. 3c). As expected, the LMCs exhibited the lowest stiffness (28.4 ± 18.1 Pa), the parental cells the highest (36.5 ± 25.5 Pa), and CTCs intermediate (31.1 ± 20.8 Pa) (Fig. 3d).

Discussion

In this work, we studied the mechanical phenotype of breast cancer cells in pulmonary metastasis by comparing across the parental-CTC-LMC panel of breast cancer cells derived from the MDA-MB-231 cell line. We found that LMCs are the fastest and the softest, with a significantly higher capacity of transitioning to amoeboid mode. There are multiple approaches to measure the mechanical properties of the cells, among which AFM, magnetic tweezers, optical tweezers and micropipette aspiration are the most commonly used. Comprehensive comparisons between these methods are available in reviews by Wu et al. and Lee et al. (27, 28). It should be noted that the elastic moduli measured with these methods show very large variations, and it is therefore difficult to compare the absolute values directly. For example, the cell line used in this study, MDA-MB-231, has been shown by previous studies to have stiffness ranging from ~10 Pa by optical tweezers, to ~50 kPa by AFM. The variability results from whether the applied mechanical stress is stretching (magnetic/optical tweezers) or compressive (AFM) as well as the size and the geometry of the probe. In this study, we consistently used the fibronectin-coated magnetic beads to bind to the cell via integrin (29), measuring the stiffness of the cortical actin (30), which determines the dynamics of cell migration. Hence, the measurement results from the three cell lines provide relative but statistically reliable information regarding cortical stiffness.

Our results suggest that soft cells of mesenchymal modes are more likely to transition to amoeboid modes. However, the underlying mechanisms behind the relationship between cytoskeletal stiffness and the propensity for amoeboid modes are yet to be identified, and are beyond the scope of this present study. Furthermore, our findings illustrate that metastasis is a selective process favoring motile and softer cells. We showed that as cancer cells progress through metastasis, the heterogeneity within the cell population in terms of stiffness is clearly reduced (Fig. 3d). This suggests that a subpopulation with lower stiffness might be selected during metastasis. The decreased stiffness in LMCs might provide these cancer cells with additional advantages such as evading phagocytosis of innate immune cells (31). Apart from cell mechanics, cell proliferation is also an important factor in tumor growth and metastasis. Ki-67, as a marker of cell proliferation widely used for breast cancer prognosis (32), was examined using immunofluorescence (Fig. S6). Consistent with the previous study (33), there is a high percentage of Ki-67-positive cells among the parental cells, and such a high level is maintained similarly in CTC and LMCs. Therefore, although high proliferation is indispensable for cancer cell metastasis, if considering the progressive adaptations along the steps of metastasis, it is not as distinguishing as the mechanical and migratory characteristics investigated in our study.

During cell migration, speed indicates the ability of the cell to rapidly organize its cytoskeleton in order to translocate, and persistence indicates the ability of the cell to maintain movement in one and only one direction by maintaining front-rear polarity. Velocity indicates the combination of these two behaviors to achieve overall effective migration. By dissecting velocity into speed and persistence, we get valuable insights into the cell migratory behavior. LMCs have both higher speed and higher persistence in microchannels, suggesting that they are more active in cytoskeleton remodeling for translocation, and steadier in maintaining front-rear polarity in confined environments. In comparison with the reduced persistence for parental cells and CTCs as the channel size gets smaller (Fig. S2), the maintained high persistence for LMCs might result from the less mechanical resistance due to lower cell stiffness, which results in less frequent polarity change. This could be another reason why decreased stiffness is more favorable for invasion and confined migration, and more broadly, metastatic progression.

Interestingly, in terms of migration, CTCs resemble the parental cell line, but not LMCs. This endows us with two important insights. First, this suggests the extravasation step might be more selective in cell migratory capabilities, as compared to the intravasation step. During invasion and intravasation, cells can either migrate singly or collectively as a cohort, with both types together forming the CTC population (34). A recent study showed that loss of the intercellular adhesion protein, E-cadherin, as in the case of single migrating cells, will put the disseminated cells at a higher risk of apoptosis (35). On the other hand, during collective migration, the majority of cells in the cohort are similar to the cancer cells in the primary tumor, with a very small subpopulation (~1.4%) of metastatic cancer cells at the leading front of the collective cohort being highly invasive (36). These findings, together with the similarity between parental cells and CTCs found in our results, support the hypothesis that collective migration predominates single-cell migration in primary tumor invasion (37). After entering circulation, only as little as 0.01% of CTCs survive and are able to form secondary tumors (38). This could result from the physical constraint encountered during extravasation, as well as the additional genomic aberrations induced in the cells even if they completed such constricted migration (39). Indeed, the aforementioned study also found the majority of lung metastases (>88%) are molecularly similar to the leading cells that are highly invasive (36). Based on these findings, our study suggests that only the most motile proportion of the CTCs successfully form secondary tumors.

Second, it has been shown that cancer cell extravasation is independent of the expression of membrane-type 4 matrix metalloproteinase (MT4-MMP) (40), but highly dependent on cell deformability (41). Moreover, when ECM degrading mechanisms are absent, amoeboid movement can function as a compensation strategy that sustains migration (22). Therefore, the decreased stiffness of LMCs, as well as the increased capacity to transition to amoeboid motility, is likely the result from the selective process during extravasation.

In conclusion, our findings demonstrate that cancer cells become more motile as they progress through metastasis, and such changes are associated with decreased stiffness. The fact that CTCs resemble parental cells rather than cells in metastatic pulmonary tumors indicates that the extravasation step is highly selective in cell migratory capabilities. It should also be noted that metastases to different organs might be mediated by different processes (42). Therefore, a more comprehensive study is required to unveil the relationship between cell stiffness and metastasis.

Nonstandard abbreviations

EMT

epithelial-mesenchymal transition

AFM

atomic force microscopy

CTCs

circulating tumor cells

LMCs

lung-metastatic cells

FBS

fetal bovine serum

FA

focal adhesion

MAT

mesenchymal-amoeboid transition

ECM

extracellular matrix

MT4-MMP

membrane-type 4 matrix metalloproteinase

PDMS

polydimethylsiloxane

PFA

paraformaldehyde