Extracellular Matrix Rigidity-dependent Sphingosine-1-phosphate Secretion Regulates Metastatic Cancer Cell Invasion and Adhesion

Dynamic interaction between cancer cells and the surrounding microenvironment is critical for cancer progression via changes in cellular behavior including alteration of secreted molecules. However, the molecular mechanisms underlying the influence exerted by the cancer microenvironment on secretion of molecules during cancer progression remain largely unknown. In this study, we report that secretion of spingsine-1-phosphate (S1P) and its regulator, SphK1 expression is dependent of the substrate rigidity, which is critical for the balance between cancer cell invasion and adhesion. Conditioned media (CM) of MDA-MB-231, an aggressive breast cancer cell obtained from soft substrate (~0.5 kPa) induced chemo-attractive invasion, while CM obtained from stiff substrate (~2.5 kPa) increased cell adhesion instead. We found that the expression of SphK1 is upregulated in the stiff substrate, resulting in an increase in S1P levels in the CM. We also found that upregulation of SphK1 expression in the stiff substrate is dominant in metastatic cancer cells but not in primary cancer cells. These results suggest that alterations in the mechanical environment of the ECM surrounding the tumor cells actively regulate cellular properties such as secretion, which in turn, may contribute to cancer progression.

Scientific RepoRts | 6:21564 | DOI: 10.1038/srep21564 carcinogenesis in a paracrine and/or autocrine manner 16,17 . S1P generated by sphingosine kinase 1 (SphK1) is secreted by the cell via ABCC1 transport and binds to the S1P receptor (S1PR) to promote cellular proliferation, migration, and contraction [18][19][20] . NIH3T3 fibroblasts overexpressing SphK1 acquired the transformed phenotype, including colony growth in soft agar and the ability to form tumors in NOD/SCID mice 21 . In addition, level of SphK1 is upregulated in various forms of cancer including breast cancer 18,22 and correlates with poor prognosis 23 and resistance to chemotherapy 24 . Several heterotrimeric, G-protein-coupled receptors have been identified as S1PRs, and their presence determines the differential cellular function of S1P 25,26 . However, for the aggressive breast cancer cell line MDA-MB-231, S1P shows anti-migratory and invasive effects in a receptor-independent manner, via an unknown molecular mechanism 27 .
In this study, we compared the effect of conditioned medium (CM) derived from MDA-MB-231 human breast cancer cells (MDA-CM) and MCF10A normal breast epithelial cells (10A-CM) on cell migration and invasion using the collagen-coated Transwell system. The results indicated that the serum-induced migration and invasion of MDA-MB-231 cells was significantly decreased by MDA-CM. CM produced in the presence of pharmacological inhibitors of protein secretion and exosome formation did not rescue the inhibitory function of MDA-CM. However, depleting the lipid growth factor from MDA-CM by activated charcoal as well as CM obtained from cells with siRNA-mediated SphK1 silencing did not show inhibition of cell invasion. We also found that SphK1 expression is upregulated in breast tumors with increased stiffness (approximately 2.5 kPa) compared with that in normal breast tissue (approximately 0.5 kPa). Additionally, MDA-MB-231 cell invasion was unaffected by CM obtained from cells cultured on soft matrix, whereas CM obtained from stiff matrix seemed to promote cell adhesion. Finally, regulation of SphK1 expression and S1P secretion by ECM stiffness is dependent on cancer cell origin. In primary cell lines, increasing ECM stiffness reduced SphK1 expression. On the contrary, in aggressive metastatic cell lines, increasing ECM stiffness induced SphK1 expression. Additionally, CM harvested from cells with upregulated SphK1 expression cultured on stiff or soft matrix enhanced cell adhesion. Thus, our data suggest that the temporal regulation of S1P secretion by the differential mechanical conditions is one of key regulators of the balance between cancer cell invasion and adhesion during metastasis.

CM derived from MDA-MB-231 cells shows an inhibitory effect on invasion and migration.
To test the effect of molecules secreted from cancer cells on cell invasion and migration, we compared the ability of CM derived from MDA-MB-231, an aggressive breast cancer cell line and MDF10A, a normal breast epithelial cell line, on invasion of MDA-MD-231 cells. A cell migration and invasion assay using the non-collagen-coated or collagen-coated Transwell system showed that chemo-attractive invasion of MDA-MB-231 cells was significantly increased when serum was loaded in the lower chamber. 10A-CM appeared to increase cell migration and invasion slightly compared to the control conditions. However, MDA-CM decreased cell migration and invasion significantly ( Fig. 1a and Supplementary Fig. S1).
MDA-CM was obtained from the cells incubated for 24 h in a serum-free medium in which cells are likely to induce self-decomposition. To rule out this possibility, MDA-CM was prepared with serum and its ability to induce cell invasion was examined. Figure 1b shows that chemo-attractive invasion of MDA-MB-231 cells was significantly inhibited in the MDA-CM prepared with serum. Thus, these results indicate that the molecules secreted from MDA-MB-231 cells negatively regulate cell migration and invasion.
To further confirm whether the factors secreted from MDA-MB-231 cells directly inhibited cell invasion, the Transwell invasion assay was carried out, where MCF10A or MDA-MB-231 cells were seeded in the lower chamber. Interestingly, MCF10A cells seeded in the lower chamber showed a similar extent of cell invasion to 10A-CM. However, the invasion of MDA-MB-231 cells while MDA-MB-231 cells were present in the lower chamber was dramatically inhibited (Fig. 1c). These results clearly demonstrate that the molecules secreted from MDA-MB-231 cells contain inhibitory factor/s for cell invasion.

MDA-MB-231 cell-derived CM induced focal adhesion and cellular contractility.
It has been previously reported that invading cells possess a slender morphological form with numerous lamellipodia and weak stress fiber formations 28 . We attempted a comparison of morphological characteristics of MDA-MB-231 cells in response to MDA-or 10A-CM treatment. Figure 2a,b show representative images of cells and statistical analysis of cell polarity with MDA-and 10A-CM, respectively. Most of the cells treated with MDA-CM possessed a round form with developed actin stress fibers, whereas 10A-CM induced the cells to acquire more slender and elongated morphologies. In addition, MDA-CM-induced vinculin-positive focal adhesions along the cell periphery, while cells under 10A-CM did not develop focal adhesions or stress fibers (Fig. 2a). Furthermore, immunoblotting for pFAK and pMLC showed that the cells treated with MDA-CM profoundly induced the phosphorylation of FAK and MLC compared to control cells or 10A-CM (Fig. 2c). Thus, these results suggest that MDA-CM induces cellular contractility resulting from the development of focal adhesion and stress fibers.
The increase in cell contractility with MDA-CM was confirmed by the measurement of stress-related collagen matrix contraction (SRMC) which is a typical method for evaluating cell contractility induced by mechanical tension 29 . SRMC under MDA-CM increased profoundly, while the extent of SRMC with 10A-CM was similar to that of control medium (Fig. 2d). If the inhibition of cell invasion by MDA-CM is due to an increase in cellular contractility, we anticipate that a decrease in cellular contractility induced by MDA-CM may reverse cell invasion. Figure 2e shows that the invasion of MDA-MB-231 cells in the Transwell system under the MDA-CM conditions was increased significantly with various pharmacological inhibitors for cellular contractility. Taken together, MDA-CM contains pro-contractile materials that induce the abnormal increase in cellular contractility, which might be the reason for the inhibition of cell invasion.  Figure 3a showed that none of the chemical inhibitors of exocytosis (brefeldin A; BFA), exosome secretion (dimethyl amiloride; DMA), and protein synthesis (cycloheximide; CHX) rescued the inhibitory function of MDA-CM. The only exception was treatment with activated charcoal, which significantly restored the invasive ability of MDA-MB-231 cells. Lipid-depleted MDA-CM decreased the size of focal adhesions visualized by vinculin but increased cellular polarity (Fig. 3b). These results indicate that the lipid compounds secreted from MDA-MB-231 cells might be a factor that increases cellular contractility and involves an inhibitory role of the invasive ability of MDA-MB-231 cells.
Among bioactive lipids, LPA-and S1P-induced cellular contractility via RhoGTPase signaling pathway have been reported 30,31 . Therefore, we examined the effect of LPA or S1P on the invasion of MDA-MB-231 cells. As shown in the Fig. 3c, LPA did not inhibit the invasion of MDA-MB-231 cells induced by serum. However, medium containing S1P profoundly decreased the invasion of MDA-MB-231 cells. Depletion of S1P from the medium by activated charcoal treatment restored the serum-induced MDA-MB-231 cell invasion (Fig. 3d). Immunofluorescence images also revealed that treatment of MDA-MB-231 cells with S1P increased focal adhesions along the cell periphery, but treatment of the S1P-containing medium with activated charcoal inhibited the development of focal adhesion and cellular contractility (Fig. 3e). To compare the levels of S1P between 10A-and MDA-CM, we attempted to measure levels of S1P in the CM using ELISA 32 . Figure 3f showed that the secretion of cells (1 × 10 5 cells/well) were placed at the bottom of the lower chamber, as shown in the right panels in a schematic diagram of the co-culture chemotaxis assay. The average number of invading cells/photographic field was counted. Values represent mean ± s.e.m., ***p < 0.001. S1P from MDA-MB-231 cells was significantly increased compared to 10A-CM (Fig. 3f). Therefore, S1P is likely to be a strong candidate among the substances secreted from MDA-MB-231 cells to induce cellular contractility and inhibit invasive activity of the cell. Although it has been previously reported that S1P synthesis is mediated by two sphingosine kinases (SphKs), SphK1 and SphK2, the level of extracellular S1P is regulated by SphK1 in MDA-MB-231 cells 33 . Therefore, we explored the effect on MDA-MB-231 cell invasion of the CM obtained from cells where  SphK1 activity was blocked by a pharmacological inhibitor, PF-543 34 . Figure 4a shows that MDA-MB-231 cells incubated with MDA-CM obtained from PF-543-treated cells exhibited an elongated morphology with extensive lamellipodia at the leading edge; MDA-CM obtained from PF-543-treated cells did not prevent the invasive activity of MDA-MB-231 cells (Fig. 4b), which was consistent with the results obtained thus far.
To obtain more direct evidence for the requirement of SphK1 activity for the inhibitory effect of MDA-CM in cell invasion, SphK1 was silenced in MDA-MB-231 cells by siRNA transfection. Treatment of cells with siRNA decreased the SphK1 transcript specifically without affecting the level of GAPDH transcript (Fig. 4c). The amount of S1P in the CM obtained from the SphK1 knockdown MDA-MB-231 was reduced by 40% compared with that of Mock-siRNA-treated cells (Fig. 4d). MDA-MB-231 cells incubated with MDA-CM obtained from SphK1-silenced cells exhibited elongated cell bodies with massive lamellipodia compared with cells that were incubated with MDA-CM generated from mock siRNA-transfected cells (Fig. 4e). MDA-CM obtained from SphK1-silenced cell also did not show an inhibitory effect on MDA-MB-231 cell invasion (Fig. 4f).
SphK1 expression is dependent on ECM rigidity. Since MDA-MB-231 cells are known to possess aggressive metastatic characteristics, the fact that the secreted molecules from MDA-MB-231 rather inhibit its invasion is controversial. Interestingly, it has been previously reported that gene expression can be influenced by stiffness of the surrounding ECM environment 35 , therefore we speculated that SphK1 expression is also controlled by matrix stiffness. To test this hypothesis, MDA-MB-231 cells were cultured on polyacrylamide gel (PAG) and 3-dimensional (3D) collagen matrices mimicking the physiological rigidities of early and late-stage breast cancer stroma, and the expression of SphK1 mRNA was evaluated by qRT-PCR. Figure 5a shows that the relative expression of SphK1 mRNA is much higher at 2.5 kPa PAG, which is within the stiffness range of late breast cancer tissues compared with that of normal breast tissue and early breast cancer stroma (approximately 0.5 kPa) 36 . Expression of SphK1 mRNA in 3D collagen matrices also showed that a high-density (HD) 3D collagen matrix substantially induced SphK1 expression compared with a low-density (LD) 3D collagen matrix (Fig. 5b). In addition, the amount of S1P secreted from MDA-MB-231 cell lines cultured on 2.5 kPa PAG and HD 3D collagen matrix was increased by 2 and 1.3 times, respectively (Fig. 5c). Morphological analysis showed that MDA-CM obtained from 0.5 kPa PAG and LD 3D collagen matrices promoted cellular polarity similar to that of cells incubated with MDA-CM obtained from SphK1-silenced cells, while cells incubated with MDA-CM obtained from 2.5 kPa PAG and LD 3D collagen matrices presented a more rounded morphology. This indicated that the circumstance in which the cell inhibits SphK1 expression affects its polarity (Fig. 5d). Invasive activity of MDA-MB-231 cells also increased with MDA-CM obtained from 0.5 kPa PAG or LD 3D collagen matrices compared with that in with CM obtained from 2.5 kPa PAG or HD 3D collagen matrices. However, MDA-CM generated from the stiff environment with the SphK1-silenced cells did not show the inhibitory effect on invasive activity of MDA-MB-231 cells (Fig. 5e), indicating that SphK1 expression is critical for the regulation of cell invasion, likely mediated by the S1P level.
To determine the physiological role of SphK1 and its biological product, S1P in the media, we performed the cell-matrix binding assay, since it has previously been reported that increased cellular contractility induces cell-matrix interaction 37 . MDA-CM obtained from 2.5 kPa PAG or HD 3D collagen matrix conditions promotes adhesive activity, which is reversibly decreased by SphK1 silencing (Fig. 5f,g). Taken together, SphK1 expression and its biological product, S1P are critical for the regulation of invasiveness and adhesiveness in response to the mechanical status in cancer tissues.

An increase in SphK1 expression by ECM hardening is characteristic of aggressive metastatic cancer cells.
We investigated the expression of SphK1 depending on the ECM stiffness using various cancer cell lines. Interestingly, normal breast epithelial cell lines, MCF10A, and primary breast cancer cell lines, Hs578T and BT20, showed the expression of SphK1 expression in stiff substrates was decreased by 15%, 16%, and 23% compared with soft substrates, respectively ( Table 1). Levels of S1P in CM were also reduced in these cells. However, cancer cells originating from pleural effusion with metastatic characteristics including MDA-MB-231 increased SphK1 expression in stiff substrates and increased S1P secretion. The exception to this was the MCF7, non-aggressive breast cancer cell lines. These results indicate that aggressive metastatic cancer cells, not primary cancer cells, increased SphK1 expression in stiff ECM.
Next, we investigated whether the SphK1 expression in response to ECM stiffness influences cell invasion and adhesion (Fig. 6). MCF10A and MCF7, which did not change the SphK1 expression at different rigidities, showed no significant change in cell adhesion between CM harvested from 0.5 or 2.5 kPa PAG. CM obtained from Hs578T and BT20 cells cultured in 2.5 kPa PAG, with decreased SphK1 expression, reduced cell adhesion while CM derived from MDA-MB-231 and H1299 cells cultured in 2.5 kPa, with increased SphK1 expression, profoundly induced cell adhesion. Taken together, S1P secretion with an increase the SphK1 expression in stiff substrate is characteristic of aggressive metastatic cancer cells.

Discussion
Progression of malignant cancer is often associated with abnormal secretion of various cytokines by cancer cells, the process of which is closely related to the development of cancer, and influenced by the surrounding cells in an autocrine and a paracrine manner. For example, increased secretion of TGF-β can lead to epithelial-mesenchymal transition (EMT), a developmental process that forces metastatic properties upon cancer cells, facilitating migration and invasion 38 . Moreover, cancer cells trigger angiogenesis and enhance growth via secretion of vascular endothelial growth factor (EGF) and platelet-derived growth factor (PDGF). However, our data showed that MDA-CM inhibited its own cell migration and invasion. Cells treated with MDA-CM showed reduced polarity compared with plain media or CM obtained from a normal breast epithelial cell line, MCF10A. This is consistent with findings from previous studies showing growth factors secreted from normal breast epithelial cells can induce cell migration 39 . Moreover, CM obtained from a non-small lung metastatic cancer cell line, H1299, also exhibited similar results, where cell migration and invasion induced by serum is profoundly inhibited (Supplementary Fig. S2). Thus, these results indicate that most secreted molecules from cancer cells may not be involved in cell migration and invasion. Rather, the molecules secreted from cancer cells are responsible for cell survival and growth 40,41 .
Our observation that MDA-CM induced focal adhesion formation and the phosphorylation of FAK and myosin light chain indicates that MDA-CM contains a potent agonist for cell contraction. Furthermore, an increase in SRMC supports the idea that molecules secreted from MDA-MB-231 cells have the potential to enable cell contraction instead of cell migration, since SRMC is often used as a model for cell contraction 42  Data obtained at 2.5 kPa differs compared to data obtained at 0.5 kPa. The Student's t-test was used for statistical analysis. *p < 0.05, **p < 0.01, ***p < 0.001. previously reported that cellular contractility accompanying tyrosine phosphorylation of FAK and Src kinase is critical for cell migration 43 , the spatio-temporally optimal level for cellular contractility is indispensable for cell migration 44 . Thus, our results suggest that maximal activation of cellular contractility by MDA-CM rather exerts an inhibitory effect on cell migration and invasion. Consistent with this possibility, we found that various pharmacological inhibitors for cellular contractility restored the inhibition of MDA-MB-231cell invasion by MDA-CM.
Previous reports indicate that the role of S1P in cell motility in MDA-MB-231 cells is independent of cell surface receptors, as there is no corresponding receptor on MDA-MB-231 cells 27 . Moreover, the effective range of S1P on chemotactic migration of MDA-MB-231 cells is > 10 μ M, which is far beyond that of S1P receptor affinity. Wang et al. proposed the involvement of a non-oscillatory increase in free calcium by S1P as a plausible mechanism for the its inhibitory role in cell motility, independent of receptor activity 27 . Our results also showed that treatment with ML-7, an inhibitor of myosin light-chain kinase (MLCK), did not inhibit MDA-CM-mediated cell invasion, indicating that calcium-mediated signaling such as MLCK activation is required for S1P-mediated cell motility 45 .
Next, we showed that expression levels of SphK1 mRNA, which is a primary regulator of S1P synthesis in the cell is dependent on the mechanical status of the cell's surrounding environment. A positive correlation is known to exist between SphK1 expression and cancer progression. Increased SphK1 expression has been observed in breast cancer 46 . However, in MDA-MB-231 human breast cancer cell lines, SphK1 expression was not observed in this study in cells with approximately 0.5 kPa stiffness, which is similar to the rigidity of normal breast tissue, and was profoundly increased in cells with approximately 2.5 kPa stiffness, corresponding to breast cancer stiffness 11 . Moreover, MDA-CM obtained with 0.5 kPa stiffness did not interfere with cell invasion induced by serum, as opposed to CM obtained from cells with 2.5 kPa stiffness. It has been previously reported that the genetic and epigenetic alterations, along with changes in the biophysical properties of fibrotic stroma surrounding the tumor cell gradually occur as tumorigenesis proceeds 47 . For example, expression of Siva1, which acts as negative regulator of EMT, and post-translational modification of stathmin, a microtubule destabilizer, are different in primary tumors and metastatic tumors in breast cancer patients 48 . In addition, in pancreatic and breast cancer cells, MMP activity, which depends on the stiffness of ECM, differs between metastatic and primary cancer cells 49 . In this regard, the expression of SphK1 is likely one of markers for the mechano-responsive genes for tumorigenesis.
In addition, regulation of expression of SphK1 and S1P secretion appeared to be influenced by factors other than matrix stiffness. Since expression of SphK1 in the triple negative (TN) cancer cell line, MDA-MB-231 was upregulated in stiff substrate in our results, we anticipated that breast cancer cells would increase SphK1 levels in stiff substrates. However, several breast cancer cell lines including TN breast cancer cell lines such as Hs578T and BT20 decreased the SphK1 expression in stiff substrate. Thus, there might be another factor involved in the regulation of SphK1 expression in addition to matrix stiffness. When results were analyzed regarding cells as primary or metastatic cancer cells based on their origin, we found that cancer cells taken from primary tumor sites tend to show a decrease in SphK1 expression in stiff substrate, whereas cancer cells taken from metastatic sites (e.g. MDA-MB-231 from pleural effusion; H1299 from lymph nodes), showed an increase in SphK1 expression in stiff substrate. Thus, these results indicate that SphK1 expression in response to ECM stiffness is likely to be determined by the metastatic status of cells.
Considering the seed and soil hypothesis, cancer progression is closely related with the mechanical properties of surrounding microenvironment 50 . During the early stages of tumorigenesis, stroma create an environment permissive of cell proliferation and invasion with minor changes to the biophysical properties of ECM. Therefore, one of the important roles for cancer cells existing in the primary site of the tumor is to change the mechanical properties of the ECM surrounding the cancer cells to generate the permissive microenvironment 51,52 . Our results showing that cancer cells obtained from primary tumor sites increase SphK1 expression in soft matrix support the concept that S1P secreted by cancer cells originating from the primary site into the surrounding cells and tissue allows cells to develop a mechanically permissive environment for cancer proliferation and survival. However, metastatic cancer cells must colonize once they extravasate from blood vessels, and the cells also require the ability to promote adhesion between the cells and the substrate 53 . We also showed that MDA-MB-231 and H1299, the aggressive metastatic cancers, increased SphK1 expression and adhesive activity in the stiff substrate. Normally, breast and lung cancer cell metastasizes to the bone and liver 54,55 , in which mechanical properties are stiffer than those of the primary site. Metastatic cancer cell originating from breast and lung must increase the adhesive activity once they arrive at the stiff substrate 56 . Thus, it is plausible that the increase in expression of SphK1 on stiff substrate may be a critical property of a metastatic cancer cell for facilitating the cell-matrix adhesion. In this regard, development of a pharmacological inhibition therapy targeting SphK1 is a useful therapeutic strategy for both early and late stages of cancer, although no specific inhibitors targeting SphK1 have been developed as yet 57 . Preparation of conditioned media. MDA-MB-231, MCF10A, Hs578T, BT20, MCF7, and H1299 cells were grown in culture media until the cell density reached 70 to 80% confluence. The cells were then washed with DMEM/F12 thrice, before they were incubated in DMEM/F12 at 37 °C for 24 h. After incubation, CM was harvested, centrifuged, and filtered through a 0.2 μ m filter.

Methods
Cell invasion through Transwells. Transwells (Costar, Corning, NY, USA) with 8 μ m pore size were coated with 10 μ g type I collagen (BD Bioscience, Bedford, MA, USA) overnight, at 37 °C. Cells were seeded in the upper chamber and DMEM/F12 or CM supplemented with 10% FBS was added to the lower chamber. Non-invading cells were removed with a cotton swab. Cells were fixed and stained with 0.1% crystal violet. The number of invading cells was quantified by counting of least five random fields per filter.

Stress-released matrix contraction.
Collagen matrices (1.5 mg/mL collagen, 2 × 10 5 cells/matrix) were polymerized for 1 h at 37 °C and matrices were cultured for 24 h in DMEM/F12 supplemented with 10% FBS. After incubation and rinsing with PBS, the matrices were released from the plate and incubated for 1 h while floating in CM, as shown in figure 2e. At the end of incubation, matrices were fixed with 3.7% paraformaldehyde and matrix diameters were measured. S1P ELISA. S1P concentrations in CM were measured using an S1P ELISA kit (MyBioSource, San Diego, USA). All procedures were performed according to the manufacturer's protocols. S1P concentrations were quantified in triplicate and compared with an internal S1P standard (3.125~25 ng/mL). Optical Density (O.D.) at 450 nm was measured by using microplate reader. Statistical analysis. Differences between control and treatment groups were analyzed by the Student's t-test using GraphPad PRISM software (Graphpad Software, San Diego, CA, USA). The data are expressed as average ± s.e.m. of three independent experiments. A value of p < 0.05 was considered to be the threshold of significance.