Tumor cell invasion resulting in metastasis is the main cause of cancer mortality rather than primary tumor growth, and the tumor microenvironment plays a critically important role in thisinvasion process [21]. In order to metastasize, the tumor cells undergo epithelial-to-mesenchymal transition (EMT)-like events whereby they lose their polarity, and cell-cell and cell-matrix contacts. The acquired mesenchymal, de-differentiated and motile characteristics facilitate cell movement and invasion to novel metastatic locations. The molecular hallmarks of EMT are downregulation of the cell-cell adhesion molecule E-cadherin and upregulation of many mesenchymal markers [22?4]. ECM composition and remodeling affect the differentiation state and behavior of tumor cells [25,26]. For example, increased expression and crosslinking of collagen I and IV are suggested to promote EMT, tumor progression and metastasis [27?0]. EMT is a reversible process; during mesenchymal-to-epithelial transition (MET) the cells become again non-motile [22,31]. The complex interactions between cells and ECM molecules are largely regulated through integrins and other cell surface receptors [32,33]. Particularly collagen IV has been shown to be the binding substrate of integrins in many cell types, including tumor cells, and its binding to different integrin subtypes may vary depending on its remodeling state [34]. Integrin binding triggers intracellular signaling events that contribute to cancer progression. The pathways leading to EMT via regulation of cadherins requires co-operative signals from integrins [32,33]. As arresten has effects on other cell types in the tumor microenvironment besides endothelial cells [18], we focused here on its impact on highly metastatic human tongue squamous cell carcinoma HSC-3 cell line. By using in vitro cell culture assays, organotypic invasion and in vivo mouse xenograft models, we show that overexpression of arresten promotes epithelial morphology, and efficiently inhibits proliferation, migration and invasion of carcinoma cells, and induces their apoptosis, leading to suppression of tumor growth and progression.
To confirm that the observed significant change in the Arr-HSC cell motility was not due to an artifact of overexpression, but rather to the secretion of arresten into the culture medium we collected CM from the Arr-HSC cells, transferred it to Ctrl-HSC cells and measured the effect on cell migration by Transwell assay. The migration of Ctrl-HSC cells decreased approximately 40% in the presence of conditioned Arr-HSC medium (p,0.001) (Figure S4A). To verify that the secreted arresten did not become degraded during the co-culture period, we collected CM for Western blot analysis at various time points of culture. This analysis showed that no protein degradation occurred during the 72 h culture period (Figure S4B).
Arresten Reduces Tumor Vasculature and Suppresses Growth and Invasion of HSC-3 Xenografts
Ctrl-HSC or Arr-HSC carcinoma cells were injected subcutaneously into nude mice and tumor growth was monitored for 16 days. The Arr-HSC tumors grew significantly more slowly than the control tumors (Figure 2A). In addition, some differences in local tumor invasion were noted between Arr-HSC and Ctrl-HSC xenografts upon histopathological examination (Figure 2B). Most (,80%) of the arresten tumors had not invaded into the surrounding tissue, whereas half of the control tumors showed at least minor score of invasiveness (Figure 2B). Our observation of the less invasive phenotype of Arr-HSC xenografts was supported by an in vitro experiment, where the Arr-HSC cells invaded less through Matrigel than the Ctrl-HSC cells (Figure S5). Immunostaining of HSC-3 xenografts for Ki-67 revealed almost 70% reduction in the amount of proliferative cells in arresten tumors (p,0.001) (Figure 2D), at least partly explaining the smaller size of these tumors. Since arresten is a potent inhibitor of angiogenesis, the amount of tumor blood vessels was determined. The blood vessel density reduced almost 50% (p,0.001) in the arresten xenografts relative to the control tumors (Figure 2F). Histological analysis of HSC-3 xenografts revealed that besides being smaller the Arr-HSC tumors also more often contained central keratinized areas and keratin pearls, indicating higher degree of differentiation, and the proportion of the surrounding poorly differentiated tumor cell layer was smaller than in the control tumors (Figure S6A). E-cadherin staining showed either diffuse cytoplasmic signals in the poorly differentiated tumor areas, or membranous staining within the keratinized areas in all xenografts (Figure S6E).
Results Arresten Inhibits Carcinoma Cell Migration in vitro
After stable transfections, the expression of recombinant arresten was verified in three separate clones of HSC-3 tongue squamous cell carcinoma cells, and also in two MDA-MB-435 breast carcinoma cell clones. By comparison to the parental cells, these stable cell lines showed a substantial increase in arresten expression at mRNA level as ascertained by qPCR (Table S1). More importantly, a ,29 kDa Flag-tagged arresten was detected by Western blotting in the conditioned medium (CM) collected from Arr-HSC and Arr-MDA cells (Figure S1A�B). The following experiments were performed using Ctrl-HSC(1) and Arr-HSC(1) (Figure S1) clones unless otherwise stated. To study the effects of arresten on carcinoma cells, we first performed Transwell migration experiments and found that the Arr-HSC cells migrated significantly less than the control cells (p,0.001) (Figure 1A). The addition of exogenous human recombinant arresten had a similar inhibitory and dose-dependent effect on Ctrl-HSC cell migration in Transwell assay (Figure 1B). Furthermore, the Arr-HSC clones showed a clear non-migratory phenotype in the scratch wound healing assay, whereas the control cells almost closed the wound within 48 h (Figure 1C�D, Figure S2A and S2C). Also the Arr-MDA breast carcinoma cells were statistically less motile than the Ctrl-MDA cells in the wound healing assay (Figure S2B and S2D). HSC-3 cell proliferation, measured by BrdU incorporation into the DNA-synthesizing cells, was not affected by the overexpression of arresten within 24 h (Figure S3A), but a reduced number of viable arresten cells was observed in the MTT assay in a longer experimental set-up (68 h) in monolayer culture (p = 0.001) (Figure S3B).