Epithelial‐mesenchymal transition softens head and neck cancer cells to facilitate migration in 3D environments

Abstract The biological impact and signalling of epithelial‐mesenchymal transition (EMT) during cancer metastasis has been established. However, the changes in biophysical properties of cancer cells undergoing EMT remain elusive. Here, we measured, via video particle tracking microrheology, the intracellular stiffness of head and neck cancer cell lines with distinct EMT phenotypes. We also examined cells migration and invasiveness in different extracellular matrix architectures and EMT‐related signalling in these cell lines. Our results show that when cells were cultivated in three‐dimensional (3D) environments, the differences in cell morphology, migration speed, invasion capability and intracellular stiffness were more pronounced among different head and neck cancer cell lines with distinct EMT phenotypes than those cultivated in traditional plastic dishes and/or seated on top of a thick layer of collagen. An inverse correlation between intracellular stiffness and invasiveness in 3D culture was revealed. Knock‐down of the EMT regulator Twist1 or Snail or inhibition of Rac1 which is a downstream GTPase of Twist1 increased intracellular stiffness. These results indicate that the EMT regulators, Twist1 and Snail and the mediated signals play a critical role in reducing intracellular stiffness and enhancing cell migration in EMT to promote cancer cells invasion.

EMT have attracted much less attention, and many important questions in the mechanobiology of EMT remain unanswered.
Cell mechanics plays an important role in determining the metastatic potential of cancer cells. [3][4][5] Compared with the normal cells, the metastatic cancer cells are often softer, have less amount of cytoskeletons and are more invasive in trespassing through blood vessels. 6 Recently, it has been reported that cell stiffness can serve as a biomarker of the metastatic potential of ovarian cancer cells. 7 Most of the previous biomechanical studies focused mainly on cells cultured in non-coated dishes, where cells are seated on a twodimensional (2D) glass substrate with stiffness on the order of 3GPa.
The natural physiological environment of cells is, however, often three-dimensional (3D), where cells are embedded in a complex dynamic environment of 3D extracellular matrixes (ECM) 8 with stiffness in the range of tens of Pa to a few KPa, depending on the specific microenvironments. 9 Extracellular matrix is an important mediator of cellular physiological and pathological functions. [10][11][12] Depending on the specific physiological and pathological scenarios, cells sense different chemical and physical cues from the ECM (including its stiffness and architecture) via different signal transduction pathways to convert the physicochemical stimuli into biochemical signals 13 and respond dynamically. [14][15][16][17][18][19][20][21] In general, the cellular responses may include changes in focal adhesion, 14 cytoskeletal organization, 14,16 migration speed and trajectory, 19 cellular differentiation, 17,18 proliferation 20 and intracellular stiffness. 11,22 It has been well-established that when cells are cultured in more rigid ECM, their focal adhesion, cytoskeletal organization and stiffness tend to increase. 11,14,16,22 In addition, the ECM microenvironment also regulates phenotypes of cancer cells. These studies imply that there is an intimate correlation between cell mechanics and cellular response to different ECM stiffness and architectures. However, the interplays among intracellular stiffness, geometry and stiffness of ECM and EMT progression have not yet been well characterized.
We previously demonstrated that the EMT regulator Twist1 induces mesenchymal migration of head and neck squamous cell carcinoma (HNSCC) cells through suppression of let-7i to activate Rac1, and that the effect could only be observed in 3D cells culture. 23 We therefore speculate that EMT-induced cellular migration is associated with the changes in biophysical and biomechanical properties in 3D environment. In this study, we investigate the interplay between cell stiffness and EMT in HNSCC cells.

| Cell lines and shRNA experiments
The human head and neck squamous cell carcinoma (HNSCC) cell lines (FaDu, CAL-27, SAS, OEC-M1) were cultured in RPMI-1064 supplemented with 10% FBS and 1% penicillin-streptomycin solution at 37°C, 5% CO 2 . For shRNA experiments, we used two independent sequences to knock-down Snail or Twist1 in OEC-M1 or SAS cells. The first system for shRNA experiment was the pLKO.1 system. For gene silencing, pLKO.1-shLuc, Snail shRNA (TRCN0000063819 and TRCN0000063821), and Twist1 shRNA (TRCN0000020540 and TRCN0000020542) were obtained from the National RNAi Core Facility of Taiwan. The second system for shRNA experiment was the pSUPER system. A hairpin sequence containing the shRNA against Snail or Twist1 was cloned into pSU-PER.puro vector, and a scrambled sequence was used as a control.
In OEC-M1 cells, both pLKO.1 and pSUPER system were used. In SAS cells, we used pLKO.1 system with two independent sequences to knock-down Snail or Twist1. The sequences of shRNA are given in Supplementary Table S1.

| Culturing cells in different collagen matrix architectures
The 2D, 2.5D and 3D collagen type 1 matrices were prepared based on the method & procedure reported by Yang et al 23 The concentration of collagen type 1 (ECM675, Millipore) is 1.71 mg/mL, and the corresponding matrix stiffness is~259 Pa.

| Video particle tracking microrheology
We applied video particle tracking microrheology (VPTM) 24  were tracked and analysed via customized Matlab software (Math-Works, Natick, MA). 24,25 We calculated the ensemble-averaged mean-squared displacement (MSD), <Dr 2 (s)> = <[x(t + s) À x(t)] 2 + [y(t + s) À y(t)] 2 >, the effective creep compliance J(s) and the elastic modulus G 0 (x) from the trajectory of each particle. 11 G 0 ðxÞ ¼ ½JðsÞ À1 where "a" is the particle's radius, K B the Boltzmann constant and T the absolute temperature. The intracellular stiffness (in Pascal, Pa) was measured and compared in terms of the value of the elastic modulus G 0 (x) at frequency f = x/2p = 10 Hz. Although our measurements allow us to determine G 0 (x) in the frequency range of approximately 0.1-100 Hz, the frequency of 10 Hz was chosen because at lower frequency, the experimental results may be affected by the noise due to any possible low-frequency drift as well as the system 1/f noise, and the higher frequency is limited by the frame rate of the CMOS camera. Furthermore, 10 Hz is the typical frequency often used by many researchers in the cell mechanics community to compare the intracellular stiffness. [24][25][26] A schematic illustration of our experimental procedure for the measurement of intracellular stiffness in different extracellular matrix architectures based on VPTM is given in Figure 1.
Although VPTM provides not only the elastic modulus G 0 (x), but also the viscous modulus G 0 (x), we have deliberately excluded all the data related to the viscous modulus, because our results (not shown) indicate that, in the context of this paper, all the corresponding changes in G"(x) are much less significant and also much less consistent.

| Western blot analysis
For Western blots, cells were harvested and lysed in the pro- F I G U R E 1 A schematic illustration of the measurement of intracellular stiffness in different extracellular matrix architectures based on video particle-tracking microrheology. A, Carboxylated polystyrene particles (Invitrogen, fluorescence excitation/admission peaks: 580/605 nm, diameter = 200 nm, concentration = 1.35 x 10 12 particles/mL) were injected into the cells by a biolistic particle delivery system (PDS-100, Bio-Rad; pressure 450 psi). B, After particle injection, cells were washed with PBS twice and incubated in different extracellular matrix architectures (2D, 2.5D, and 3D) for 16 h. C, Brownian motion of individual intracellular particles were recorded for 10 s via a fluorescence microscope equipped with a cell incubation chamber, an oil-immersion objective lens (100x, NA = 1.45), and a CMOS camera, which enables us to record the images at a frame rate of 100 frames per second, and a spatial resolution of 0.13 lm/pixel. D, The two-dimensional projection of the trajectories of the Brownian motion of each particle [x(t) and y(t), as a function of time (t)] were tracked and analysed. E, The ensemble-averaged mean-squared displacement (MSD), and the elastic modulus G 0 (x) were deduced from x(t) and y(t) (#50581, Abcam plc., Cambridge, UK); and b-actin (MAB1501, Chemicon International Inc., Billerica, MA) was used as a loading control. ImageJ software was used for densitometric measurements of the Western blots.

| 2D cell migration assay
A 8 lm Boyden chamber was used for migration assay. with Hoechst 33342 dye. The condition of 2.5D transwell invasion assay was the same as that of the 2D migration assay, except that the migration period was 24 hours and the inserts were coated with Matrigel.

| Cell mobility speed
Cells were allowed to adhere on 1.7 mg/mL collagen for 16

| Statistical analysis
Statistical significance of the experimental results was evaluated by Student's t test and indicated by * for P < 0.05 and ** for P < 0.01.

| The epithelial-type head and neck cancer cells exhibit larger increment in stiffness in 3D ECM architecture
To investigate the impact of EMT phenotypes and different ECM architectures on cellular stiffness in HNSCC cells, we measured the intracellular stiffness via video particle-tracking microrheology

| An inverse correlation between cellular stiffness, EMT phenotype and migration capability in 3D environment
Next, we investigated the effect of different ECM architectures on cellular migration and stiffness. For cell migration measurements, we used non-coated transwells for 2D culture, Matrigel-coated transwells for 2.5D system and collagen-embedded wells for 3D. Our results show that the mesenchymal-type cancer cells exhibited a higher migration capability than the epithelial-type cells in all three systems. Notably, the differences were more pronounced in 3D system than in 2.5D and 2D. In 3D cells culture, the epithelial-type cancer cells were relatively stiff and barely migrated; in contrast, the mesenchymal cells were relatively soft and highly migratory ( Figure 3A-C). Consistently, the increment in cell stiffness was also more pronounced in 3D compared with 2.5D and 2D, but only in epithelial-type cells, and not in mesenchymal cells ( Figure 3D-G). A strong inverse correlation between EMT progression with invasiveness and intracellular stiffness of HNSCC cells was revealed in 2.5D & 3D environment ( Figure 3H).
Taken together, these results reveal that the biomechanical, physical and chemical behaviours of cells cultured in 2D environment differ drastically from those in 3D; hence, cells culture in 3D could be an important first step towards a cell model system for better mimicking cellular behaviour in vivo.

Snail increases intracellular stiffness and reduces cell migration capability
Of the several possible molecular mechanisms in EMT progression, Twist1 has been identified as a master regulator of morphogenesis which plays an essential role in tumour metastasis. 30,31 We previously reported that Twist1 cooperates with BMI1 to suppress the expression of microRNA let-7i, resulting in the activation of Rac1 and inducing mesenchymal movement in 3D environment. 31 In this study, we observed that, in 3D culture environment, Rac1 activation We used two independent shRNA systems (pSUPER, see Figure 4; and pLKO.1, see Figure S3) containing different sequences to knockdown Snail or Twist1 in OEC-M1 cells ( Figures 4A and S3A,B). The wild-type OEC-M1 cells exhibited a mesenchymal morphology in 3D culture. Knock-down of either Twist1 or Snail had a prominent effect in abrogating the elongated morphology, generated a round cell ( Figures 4B and S3C) and reduced the activation of Rac1 (Figure S3D). Inhibition of Rac1 activity by NSC23766 or ROCK activity by Y27632 also partially attenuated the cellular protrusion and