Introduction

Understanding and regulation of the motility of cancer cells is an important basis for strategies for cancer treatment. Most cancers emerge from epithelial tissue, where they proliferate, break through the underlying basement membrane, invade connective tissue, and finally metastasize to other organs via the bloodstream [1]. In this process, the phenotype of cancer cells typically transitions from the epithelial to the mesenchymal type in the so-called epithelial-to-mesenchymal transition [2,3,4]; this phenotype exhibits characteristic motility in a 2-dimensional (2D) flat environment such as the basement membrane and in a 3-dimensional (3D) fiber matrix such as the collagen matrix in connective tissues [5, 6]. Normal epithelial cells or nonmalignant tumor cells show a 2D motile phenotype, while malignant cancer cells have distinctive 3D motility [6, 7].

These different types of cell motility are not only the result of biological interactions among intracellular biomolecules but also regulated by the mechanical and structural conditions of the extracellular milieu, which govern, in principle, the mechanical behaviors of the cytoskeleton and focal adhesion [8,9,10]. Therefore, to better understand and control cancer cell invasion into connective tissue after the initial dissolution of the 2D basement membrane by MMPs, it is essential to address what mechanical and structural conditions of a 3D matrix determine the motility of cancer cells in the matrix and to distinguish the effects of these conditions from those of MMPs in the 3D matrix.

To explore these issues, a biomimetic model matrix with independent tunability for mechanical and structural conditions is required [11, 12]. It should be emphasized that conventional simple chemical gels and physical gels are not suitable for this purpose. In the case of chemical gels, the typical pore size is on the order of nanometers, which is too small for cells to invade the matrix; cells cannot enter nanometer-scale-crosslinked chemical gels without the decomposing activity of MMPs, just as healthy basement membranes prohibit cell invasion. On the other hand, physical gels can accept the ingrowth of cells; however, it is difficult to tune their mechanical properties except by changing the matrix density, which is not suitable for checking the effects of mechanical and structural conditions of the matrix. To better discriminate the effects of physical parameters on 3D cell motility, gels should have independent tunability for elasticity and porosity. Recent studies have also highlighted the importance of avoiding concurrent changes in these parameters to better interpret cell behaviors in fibrillar microenvironments [13, 14]. However, it has not been independently characterized how purely the mechanics of fibers and the porosity of a matrix influence the invasion of cancer cells apart from the effects of MMPs. After the first step of MMP-dependent dissolution of the basement membrane, what general mechanical and structural conditions of the underlying matrix control the invasive behaviors of cancer cells? This is the central question of the present study.

To answer this question, we developed elasticity/porosity-independent tunable matrices composed of electrospun photocrosslinked gelatinous microfibers, which enabled us to independently design the mechanical properties of the microfibers (nanometer-scale-crosslinked chemical gels) and the porosity of the matrices with well-regulated fiber bonding (tens-of-micron-scale fiber-bonded gels); i.e., the elasticity of each fiber could be regulated by changing the conditions of photocrosslinking, while the porosity of the matrix could be modulated by regulating the degree of bonding between fibers (see Scheme 1). Gelatin is a component of collagen and a significant constituent of the natural ECM [15]. This situation mimics that of collagen-formed physical gels in connective tissue, which also has a hierarchical matrix structure of physically bonded microfibrils (diameter of 100–300 nm) [16].

Scheme 1
scheme 1

Preparation of the microfiber gel matrix with elasticity-tunable gelatinous microfibers. Electrospun nanofibers (a) were photocrosslinked in the solid state (b) and finally swollen in aqueous solution (c)

Full size image

By using this elasticity/porosity-independent tunable fibrous gel matrix, we investigated the effects of both the porosity of the matrix and the elasticity of the component fibers on the motility of highly malignant MDA-MB-231 breast cancer cells with or without MMP inhibitors and of nontumorigenic MCF-10A mammary epithelial cells. The mechanical and structural conditions of the ECM that give rise to cancer cell invasion are discussed.

Materials and methods

Fabrication of electrospun StG nano/microfiber mesh sheets

Photocurable styrenated gelatin (StG) was synthesized as previously reported [17] and electrospun into nano/microfiber mesh sheets. First, StG (16.6 wt%) was dissolved in 1,1,1,3,3,3-hexafluoro isopropanol with stirring at 45 °C while protected from light. Next, the photopolymerization initiator sulfonyl camphorquinone (SCQ, Toronto Research Chemicals, ON, Canada; 10% of StG) and Fluolid-PM (succinimidyl ester type; IST, Co., Ltd, Fukuoka, Japan; 1.5 mg) were added to the StG solution and dissolved as described above. Electrospinning was performed with a custom-designed apparatus composed of a high-voltage power supply (HSP-30k-2; Nippon Stabilizer Ind. Co., Ltd, Osaka, Japan), an infusion pump (KDS 100; KD Scientific, Inc., MA USA), a syringe equipped with a stainless needle, a conductive nozzle, and an aluminum collector. The prepared StG solution was delivered to the nozzle at a constant flow rate (F) using the infusion pump and then ejected at a voltage sufficient to induce an unstable jet (E) with an appropriate air gap (AG) between the nozzle and the collector. These experimental parameters are described in the figure captions. A mesh sheet was obtained by accumulating the StG nano-/microfibers on the collector. To eliminate the remaining solvent, the resulting mesh sheets were dried in vacuo at least overnight in the dark.

Solid-phase photocrosslinking of StG nano/microfiber mesh sheets

The StG nano-/microfiber mesh sheets were cut into small pieces (ca. 5 × 5 mm) and mounted on a cover glass (18 × 24 mm, Matsunami Glass Ind., Ltd, Osaka, Japan) with a pair of spacers of arbitrary thickness (18 × 6 mm pieces of stainless shim; EA440FD-0.01, 0.02, 0.03, 0.04, 0.05, 0.06, 0.08, and 0.1; ESCO Co., Ltd, Osaka, Japan). Another cover glass was then placed over the sheets and spacers with 60 µL of 99% ethanol. The glass-sandwiched StG fiber mesh sheets were placed between a pair of ferrite magnets (Φ20 × Φ10 × 3 mm, clamping force: 0.45 kg, Niroku Seisakusho Co., Ltd, Kobe, Japan). By changing the spacer thickness, the degree of compression could be regulated, which was calculated from the mesh thickness (D) before and after compression according to the following equation:

$${\rm{Compression}}\; {\rm{rate}}=\frac{D_{0}-D}{D_{0}}$$

The glass-sandwiched samples were photoirradiated at a light intensity of 80 mW/cm2 (at 488 nm) for a few minutes to several tens of minutes and immersed in 99% ethanol. The cover glass was removed, and the samples were swollen in phosphate-buffered saline (PBS), which ultimately produced microfiber gel matrices. The obtained samples were fixed on adhesive-coated glasses (Biobond, Electron Microscopy Sciences, PA, USA).

Measurement of the elastic modulus of a single fiber

To measure the elastic modulus for the microscopic fiber elasticity of the fiber mesh sheets, force-indentation curves were measured with an atomic force microscope (AFM, Nano Wizard 4, JPK Instruments, Berlin, Germany). The force loaded on the matrix and the z-displacement of the probe were fixed at 0.5 nN and 1000 nm, respectively. The fibers in the gel matrices were sparsely fixed on a glass substrate (Fig. S1). The microfiber gel matrices fabricated without compression were placed on Biobond-coated glasses, unfolded with sharp tweezers, and left to dry. The samples were again immersed and completely swollen in PBS, and then indentation tests were performed as described previously. This procedure allows measurement of the deformation of the fiber surface under an appropriate indentation depth less than the fiber thickness, thus giving the apparent microscopic elastic modulus of each single fiber, which was calculated through nonlinear least-squares fitting of the force-indentation curves to the Hertz model in the case of a conical indenter (semivertical α: 30°; Poisson ratio μ: 0.5) [18,19,20] with the following equation:

$$F=\frac{2E\,\tan\alpha}{\pi(1-\mu^{2})}\delta^{2}$$

Cell culture

MDA-MB-231 human metastatic breast cancer cells were cultured in L-15 medium supplemented with 10% FBS, 50 U/mL penicillin, and 50 µg/mL streptomycin at 37 °C in the absence of CO2. To check the interference of MMPs, the broad-spectrum MMP inhibitor GM6001 (Abcam, Cambridge, MA, USA) was used (5 μM) [21]. MCF-10A mammary epithelial cells were cultured in serum-free MEGM supplemented with 10 ng/mL cholera toxin at 37 °C in a humidified atmosphere containing 5% CO2. The medium was replaced every 2–3 days.

Cell invasion assay

The microfiber gel matrices were sterilized with 70% ethanol and then rinsed three times with medium. The cells were labeled with DiI (Molecular Probes, Thermo Fisher Scientific, OR, USA) and seeded at 6 × 103 cells/cm2 on the gel matrices. After 7 days of culture, the medium was replaced, and samples were rinsed with PBS three times, fixed in 4% paraformaldehyde (PFA)/PBS for 10 min, rinsed with PBS three times, and finally stored in PBS. Z-stack observation was performed using a confocal laser scanning microscope (CLSM, LSM 510META MAITAI, Carl Zeiss SMT, Inc., Oberkochen, Germany). Multifocus side views were analyzed using ImageJ software (ImageJ 1.48v, https://imagej.nih.gov/ij/), and the fluorescence brightness values were quantified. Cell localization was evaluated as the distribution of brightness values in the depth direction.

Scanning electron microscopy

The nano/microfiber mesh sheets were cut into ~5 × 5 mm pieces, fixed on specimen mounts with carbon tape and then surface-coated with osmium. The surface structures were observed by scanning electron microscopy (SEM, Real Surface View VE-7800, Keyence, Osaka, Japan). The diameters and distributions were analyzed for at least 30 fibers using ImageJ software.

Confocal laser scanning microscopy

The samples were prepared, fixed on adhesive-coated glasses and then observed with a confocal laser scanning microscope. The images were acquired using a 20x objective lens (Plan NeoFluar 20×/0.5) and a confocal aperture of 1 Airy disk. A 543 nm HeNe laser and a 488 nm Argon laser were used for imaging with the fluorescent dyes DiI and Fluolid-W Green 520, respectively. Z-stacks were collected at 3-µm intervals. The diameters and distributions were analyzed for at least 30 fibers using ImageJ software.

Porosimetry

The porosimetry of the microfiber gel matrices was analyzed using a mercury porosimeter (Poresizer 9320, Micromeritics Instrument Co., Norcross, GA, USA) at Toray Research Center (Tokyo, Japan). The microfiber gel matrices were cut into 1 × 1 cm rectangular pieces and weighed. Three pieces of each sample were placed in the sample cup. The pressure used during mercury intrusion was in the range of 3 kPa to 400 MPa.

Results

Fabrication of nano/microfiber mesh sheets and their solid-phase photocrosslinking

To evaluate the influence of the micromechanical environment on the invasive behavior of cancer cells, nano/microfiber mesh sheets with variable elasticity were fabricated from photocurable StG. First, StG was electrospun to form solid-phase nanofiber mesh sheets (Scheme 1a); next, the formed sheets were photoirradiated to produce intrafiber crosslinking of StG and interfiber bonding of the StG fibers (Scheme 1b). Finally, the photoirradiated mesh sheets were swollen in an aqueous environment (Scheme 1c).

The morphologies of the mesh sheets before and after photoirradiation/swelling were observed by SEM (Fig. 1a) and CLSM (Fig. 1b–d), respectively. From the SEM observations, the electrospun solid fibers were confirmed to have smooth surfaces without any bead formation under the optimal electrospinning parameters, and the average diameter of the solid fibers was measured as 0.70 ± 0.10 μm. From the CLSM observations, the average diameter of the swollen gel fibers was estimated from the contour of fluorescence to be 2.2 ± 0.5 μm, 2.2 ± 0.4 μm and 2.1 ± 0.4 μm for compression ratios of 15%, 30%, and 60%, respectively. Although these fibers were prepared from different compression ratios, they showed similar average diameters and distributions under our observations, which suggested the successful solid-phase crosslinking and swelling of each StG fiber. Notably, the estimated diameter of a swollen gel fiber was an approximation due to the limitation of the resolution and imaging principle of CLSM, even though the inevitable blurring effect was minimized through optimization of the CLSM parameters.

Fig. 1
figure 1

Representative morphology of StG nano-/microfiber mesh sheets before (a, SEM image) and after (b–d, CLSM image) photocrosslinking and swelling. The mesh sheets were prepared under room temperature with a humidity of 45.4~59.5%, a flow rate of 1.0 ml/h, an air gap of 15 cm and accelerating voltage of 21.0 kV. The gel matrices in b–d were photoirradiated for 5 min but had different compression ratios of 15%, 30%, and 60%, respectively. Scale bars: 10 μm

Full size image

Influence of porosity on the invasive behavior of cells

In our present system, matrix porosity was modulated by changing the preparation conditions with regard to the compression of the mesh sheets during photoirradiation (Scheme 2). Spacers with different thicknesses were used to change the compression ratios for the solid-phase fiber mesh sheets and thus to control the degree of interfiber bonding in the matrices. Compression of the fiber mesh forces fibers inside the mesh together during photoirradiation, which increases the efficiency of photocured bonding between the attached fibers. After photoirradiation, the compression was removed, and then the fiber matrix almost recovered its original thickness, indicating that the fiber density in the matrix was almost constant. Since the porosity of the matrices should modulate cancer cell motility during invasion and metastasis depending on the phenotypes, we first examined the effects of matrix porosity.

Scheme 2
scheme 2

Control of the degree of interfiber bonding in the matrix through regulation of the compression ratio during the photoirradiation process. Spacers with a defined thickness were used to alter the compression ratio

Full size image

Four kinds of matrices with compression ratios of 10, 30, 60 and 90% were prepared, and the pore size distributions were measured with a mercury porosimeter (Fig. 2). The results indicated the presence of three types of void for each matrix condition: voids ca. 0.01–0.03 μm (type A), 2–3 μm (type B), and over 10 μm (type C). These sizes are typical for electrospun fiber meshes, as we have previously reported [22]. These matrices showed a similar void distribution, and the type B void exhibited the highest peak. The peak heights for the Type B and C voids were decreased in the matrices prepared with higher compression ratios of 60 or 90%, suggesting that greater interfiber bonding in a matrix tends to reduce the numbers of voids larger than several micrometers.

Fig. 2
figure 2

Porosimetry for the matrices prepared with different compression ratios ranging from 10 to 90% [22]. Electrospun mesh sheets were photoirradiated for 5 min to crosslink StG and produce microfiber gel matrices. *p < 0.05; **p < 0.01

Full size image

Based on the above data on matrix porosity, the 3D motility of the highly malignant cancer cell line MDA-MB-231 was characterized on matrices prepared with different compression ratios (15%, 30%, and 60%) 7 days after seeding the cells (Fig. 3). To examine the effect of MMPs, the inhibitor GM6001 (GM) was applied. Fluorescence detected deeper than −16.3 μm, which is the average size of MDA-MB-231 and MCF-10A cells (Fig. S2), was defined as the fluorescence emitted by invading cells, and representative CLSM Z-stacked side views are shown in Fig. 3a, b. Invaded MDA-MB-231 cells were only observed under the lowest compression ratio of 15%, irrespective of the presence of the MMP inhibitor. Based on a series of stacked observations by CLSM for different Z positions, the distribution of invaded cells in the Z-direction was quantitated from the sum of the fluorescence intensity of cells localized in each Z-layer with a 3.0 μm-thick resolution (Fig. 3c). Some populations of cells were detected near +20 μm in all the conditions due to the topography of the surface-loosened fiber matrix. Under the lowest compression ratio of 15%, cells showed the greatest fluorescence distribution below −16.3 μm, and the distributions of cells cultured in medium with and without GM6001 (indicated as GM (+) and GM (−), respectively) were similar to each other. With an increase in the compression ratio to 30%, the fluorescence distribution below -16.3 μm was significantly reduced. These results indicate that greater interfiber bonding hinders the 3D movement of cells.

Fig. 3
figure 3

Influence of the compression ratio (i.e., degree of fiber bonding) of the microfiber gel matrix on the invasive behavior of MDA-MB-231 cells. The matrices were prepared under the same conditions as in Fig. 2. Representative CLSM images of cells cultured on matrices with different compression ratios for 7 days without inhibitor (a) and with inhibitor (b). c Distribution of the Z positions of cells (red and blue: cells in medium without inhibitor (−) and with inhibitor (+), respectively). Lateral summation of the fluorescence intensities of DiI-labeled cells was performed for each stacked image of the z-layer, and the distributions of those in the Z-direction are plotted. Three images of 450 × 450 μm obtained from three different samples of gel sheets were analyzed for each degree of compression, which included 110~270 cells. Scale bars: 50 μm (Color figure online)

Full size image

Influence of fiber elasticity on the invasive behavior of cells

Next, we investigated the influence of microscopic fiber elasticity on the mesh sheets. In our system, microscopic fiber elasticity could be regulated by changing the degree of intrafiber crosslinking through alteration of the duration of photoirradiation. Since lower interfiber bonding is suitable for the invasion of cells, as clarified above, the compression ratio was fixed at 0%. The AFM results showed that with an increase in the duration of photoirradiation from 5 to 20 min, the mean Young’s modulus of the fibers increased from 20 to 80 kPa (Fig. 4, purple line). The three groups showed a significant difference in microscopic fiber elasticity, while the matrices had the same porosity.

Fig. 4
figure 4

Box plots for Young’s modulus of the microscopic fiber elasticity in microfiber gel matrices prepared with different photoirradiation durations. The matrices were prepared through photoirradiation without compression. The box limits, blue lines, and purple horizontal lines indicate the range, median, and mean of Young’s modulus, respectively. Microindentation tests were performed for ten fibers from three different pieces of fibrous gel sheets, and Young’s modulus was measured for n > 100 points. *p < 0.01 (Color figure online)

Full size image

The 3D motility of MDA-MB-231 cells and MCF-10A cells on these matrices was observed by CLSM 7 d after cell seeding (Fig. 5). MDA-MB-231 cells cultured without GM6001 (Fig. 5a) invaded deeper into the matrix with 20 kPa fibers; in matrices of 50 and 80 kPa, the cells did not show invasive behaviors. On the other hand, in the presence of GM6001, the invasion behavior of MDA-MB-231 was almost the same as that of GM (−) (Fig. 5b). In contrast, MCF-10A cells were arranged tightly, distributed mainly on the surfaces of the matrices, and did not show any invasive behavior due to the strong epithelial intercellular interactions, irrespective of the fiber elasticity (Fig. 5c).

Fig. 5
figure 5

Representative CLSM images of MDA-MB-231 cells (a, b, in medium without inhibitor (−) and with inhibitor (+), respectively) and MCF-10A cells (c) cultured on microfiber gel matrices with different fiber elasticities of 20, 50, and 80 kPa for 7 days. The matrices were prepared under the same conditions as in Fig. 4. The cross-sectional views are shown beside the horizontal view. Scale bars: 50 μm

Full size image

To quantitatively characterize the above observations, the distributions of fluorescence brightness in the depth direction were analyzed from Z-stack images in terms of the effects of the MMP inhibitor (Fig. 6a) and of malignancy (Fig. 6b). According to the above definition of invaded cells, the invasion rate was defined as the proportion of fluorescence emitted by invaded cells to that emitted by all the cells observed (Fig. 6a). After 7 days of incubation, regardless of whether GM6001 was added, MDA-MB-231 cells on the matrix with 20 kPa fibers showed the greatest fluorescence distribution below −16.3 μm (Fig. 6a), which indicated that they had the highest invasion rate. On the other hand, the detected intensities of invaded MCF-10A cells on all the matrices (Fig. 6b, green curves) were very small and much lower than those of MDA-MB-231 cells. The quantitative results are summarized in Fig. 6c. For MDA-MB-231 cells cultured with or without GM6001, the invasion rate increased with decreasing fiber elasticity, and no significant difference in invasive behavior was observed in the presence versus absence of the MMP inhibitor. The invasion rate was much lower for MCF-10A cells than for MDA-MB-231 cells, and the most significant decrease was observed in the matrix with 20 kPa fibers. Under these conditions, MDA-MB-231 cells showed greater invasion rates in the matrices with a fiber elasticity of 20 kPa than in the other matrices.

Fig. 6
figure 6

Distributions of the Z positions of cells (a, b) and statistical invasion ratios (c) of MDA-MB-231 cells (red and blue: cells in medium without inhibitor (−) and with inhibitor (+), respectively) and MCF-10A cells (green) cultured on microfiber gel matrices with different fiber elasticities for 7 days of culture. The fluorescence intensities of DiI-labeled cells were measured in stacked images of z-layers. Lateral summation of the fluorescence intensities of DiI-labeled cells was performed for each stacked image of the Z-layer, and the distributions of those in the Z-direction are plotted. Three images of 450 × 450 μm obtained from three different samples of gel sheets were analyzed for each degree of compression, which included 130~400 cells. *p < 0.01 (Color figure online)

Full size image

Discussion

Cancer cells can sense and respond to multiple biological signals to alter their motility. Many studies have focused on the influences of biochemical signals, and recent evidence has emphasized that biomechanical signals also play roles in regulating the motility of cancer cells. For example, a stiffer ECM activates the subcellular protrusions of invadopodia to secrete ECM-degradable matrix metalloproteinases (MMPs) [23], drives a change to a malignant phenotype through biomechanical adhesion signaling [24], and ultimately promotes the invasion and metastasis of cancer. ECM stiffness-dependent regulation of the invasive properties of cancer has been verified in different types of cancer cells in terms of mechanobiological signaling pathways [25,26,27].

Although this qualitative tendency for cancer cell invasion in response to ECM stiffness has been clarified, the pure MMP-independent effects of mechanical and structural conditions of the ECM on cancer cell invasion have not yet been quantitatively and systematically established. To elucidate this issue, macroscopic matrix porosity and microscopic fiber elasticity should be independently controlled. Thus, we focused on the respective influences of these properties on the invasive behaviors of highly malignant MDA-MB-231 cancer cells with or without MMP inhibitors in comparison with normal MCF-10A epithelial cells. The broad-spectrum MMP inhibitor GM6001 was used to exclude the interference of MMPs secreted by cancer cells [25, 28, 29]. In this system, both types of cells attached to the gelatinous matrix in a manner dependent on physically adsorbed serum-based fibronectin and vitronectin from the culture medium. The independent characterizations of both the matrix porosity and the individual fiber elasticity revealed the following: (1) MDA-MB-231 cells preferentially invaded the matrices with enough void space within 7 days, and (2) MDA-MB-231 cells strongly invaded matrices composed of fibers with an elastic modulus of 20 kPa, while matrices composed of stiffer fibers (50–80 kPa) inhibited invasion under the same conditions of interfiber bonding. These results raise two important questions for discussion. First, why can MDA-MB-231 cells invade a matrix, while MCF-10A cells cannot enter a matrix with the same mechanical properties. Second, why does a matrix with fibers of 20 kPa show the greatest invasion of malignant cancer cells? Below, we discuss these questions.

In general, the invasion of cancer cells can be divided into two types: MMP-dependent and MMP-independent invasion. In the former, the first 2D basement membrane is dissolved by MMPs secreted by cancer cells, and some holes required for invasion are made; then, the 3D matrix is also degraded, which typically generates cell-scale tracks [30]. In our experiment, signs of matrix degradation were not observed over the period of 7 days (Figs. 3 and 5). Furthermore, there were no significant differences in cell invasive behavior regardless of whether MMPs were inhibited with GM6001. Thus, during the short period of 7 days of culture, the invasion of MDA-MB-231 cells with an MMP-independent mesenchymal-ameboid transition phenotype is considered to occur [23, 31, 32].

During this process, to pass through voids in the gel fiber matrix, cells adopt a rounded morphology and undergo deformation to adapt to the void size. The cytoplasm of cells is flexible and can pass through a small void, while the much stiffer nuclei limit the voids that a cell can invade [33] and critically affect cell migration [34]. Although a larger cytoplasmic volume facilitates cell migration through pores because it generates larger forces [35], the similar cell sizes of MDA-MB-231 and MCF-10A cells (Fig. S2) exclude the influence of cytoplasmic volume. On the other hand, considering that the nuclei of MDA-MB-231 cells are larger than those of MCF-10A cells [36], the observed greater invasion of MDA-MB-231 cells indicates that the effect of nucleus size is weak in the present matrix. Thus, the influences of void size and fiber gel matrix deformability as well as those of cellular deformability should be carefully considered.

For cancer cells, the void size for cell invasion must be larger than 7 μm2; otherwise, cell invasion strongly depends on MMP activity [33, 37]. In our study, due to the limited culture duration, the influences of MMPs were diminished. Thus, the invasive behavior of MDA-MB-231 cells was mainly influenced by the size distribution of voids. Under an increased compression ratio, the decrease in the proportion of larger voids (described above as type C, with sizes greater than 10 μm) restricted the invasion of cancer cells. This critical size for an invadable void should be larger for normal cells than for cancer cells because the nuclei of normal cells are less deformable than those of cancer cells, which makes it harder for the normal cells to deform their nuclei to adapt to the void [38]. This difference thus makes it hard for normal cells to invade a matrix with very poor interfiber bonding prepared without compression, while cancer cells can invade the matrix with the same microenvironment (Scheme 3a, b).

Scheme 3
scheme 3

Schematic explanation for the influences of fiber bonding and fiber stiffness on 3D cell migration. a Cells can invade a matrix with low interfiber bonding and enough void if the fibers are suitably deformed during 3D cell migration. b Cells cannot invade a matrix with greater interfiber bonding due to the reduced size of the voids that the cells must pass through. c Cells cannot invade a matrix with stiff fibers even if interfiber bonding is low enough because the fibers are less deformable; thus, the cells can less easily increase the size of the voids during 3D migration

Full size image

The motility of cells has been shown to be governed by the stiffness of the culture substrate through integrin-mediated adhesion and signaling by mechanosensor proteins, such as talin, P130CAS and vinculin [39,40,41]. An increase in stiffness of the substrate reinforces cell protrusions along fibers in the matrix to form stable focal adhesions and then leads to cell spreading and ultimately to elongation movement [42, 43]. On the other hand, a decrease in stiffness of the substrate reduces cell spreading and focal adhesion maturation [8]. In our study, the stiffness of each fiber was measured in terms of the elastic modulus (Fig. 4), which defines the stability and deformability of fibers in a matrix [44]. The cancer cells on the matrix with stiffer fibers (50 and 80 kPa) exhibited limited invasion during 1 week of incubation, which is attributable to the formation of stable focal adhesions with nearby fibers on the matrix surface. For breast cancer cells on softer matrices with stiffness ranges lower than those of tumor stroma at ca. 30 kPa, invadopodia-associated ECM degradation is weakened [45], which confirms that the invasion of MDA-MB-231 cells did not depend on degradation of the matrix. In addition, since the formation of focal adhesions on softer matrices is relatively weak [46], the reduced adhesion to softer fibers (20 kPa) made the cells slightly mobile and enabled them to easily spread into the matrix, which caused MDA-MB-231 cells to exhibit the greatest invasion (Fig. 6c).

A stiffer environment may also make it difficult for cells to deform and accommodate into local voids. Thus, greater cell deformation is required for MMP-independent invasion [47]. Deformation of cell shape leads to a change in the shape of the nucleus. The nuclei of cancer cells do not behave regularly like those of normal cells do [48] and are thus more flexible [49]. If the invasion of cells is not MMP-dependent but rather subject to matrix void size and fiber elasticity, the different invasion tendencies of MDA-MB-231 and MCF-10A cells are likely attributable to the mechanical properties of the cells themselves, such as cellular deformability and stiffness (Scheme 3a, c). In general, the stiffness of both the cytoplasm and nucleus is significantly less in metastatic cancer cells than in nonmetastatic or normal cells [50]. In addition, highly invasive cells are more deformable than those with less invasiveness [51, 52]. In fact, MDA-MB-231 cells are softer and more deformable than MCF-10A cells with regard to both the nuclei and cytoplasm [50, 53]. To completely prove the contribution of cell deformability to the MMP-independent invasion in our system, it would be necessary to characterize the invasive behaviors of the stiffened MDA-MB-231 cells through molecular perturbations such as genetic manipulation, which should be included in essential future work.

Finally, the fiber diameter in a gel fiber matrix is also an important factor, since it can influence cell migration and invasion in a curvature-dependent manner [54, 55]. The influence of this factor is enhanced on hundreds-of-nanometers scales of fiber diameters, while it is reduced on larger scales [56]. In this study, the diameters of fibers were in the range of several microns, which was larger than the scale at which the curvature could induce an effect.

Conclusion

By using biomimetic microfiber gels with elasticity-tunable matrices and fibers, we have shown that malignant cancer cells exhibit the greatest MMP-independent invasion into matrices composed of fibers with a mean Young’s modulus of 20 kPa in the presence of a suitable void structure, while normal epithelial cells do not invade the matrices under the same mechanical and structural conditions. These behaviors are considered to be due to the stiffness-dependent modulation of cell adhesivity on each fiber [43, 57] and to differences in deformability between cancer cells and normal cells in response to void structures. These findings should contribute to understanding of the roles of mechanical and structural factors in controlling the invasion of cancer cells in 3D fiber matrices of natural and artificial ECMs. Thus, these results could help lead to a mechanobiological strategy for cancer treatment, such as a strategy involving application of the intrinsic invasiveness of cells to selectively trap and eliminate cancer cells with a stiffness- and porosity-optimized matrix.