Primary tumor versus metastasis: new experimental models for studies on cancer cell homing and metastasis in melanoma

Abstract

Cutaneous melanoma is a neoplastic disease of increasing incidence and the deadliest type of skin cancer with limited therapeutic options, particularly in advanced tumors (Orgaz and Sanz-Moreno, 2013). Early detection, surgical excision, and adjuvant therapy of primary melanoma continue to be the key variables for long-term survival and eventual cure. However, the main cause of death in patients with malignant melanoma results from metastases that are resistant to conventional therapies, but not from the primary tumor from which these malignant lesions arise (Gray-Schopfer et al., 2007; Steeg, 2006). Thus, the ability to effectively treat cancer is critically dependent on the capacity to interdict the metastatic process. Metastasis can be divided into a number of complex cell biological events, collectively termed the ‘invasion–metastasis cascade’. The first step occurs in the primary tumor, where a subpopulation of cancer cells loose their cell–cell contact, exit the tumor mass, and invade locally through the extracellular matrix and tumor-associated stromal cell layers. Afterward, cancer cells intravasate into the microvasculature of the blood or lymphatic system, survive in the circulation, extravasate into the parenchyma of distant organs, and adapt to the foreign microenvironment in order to form metastases (Valastyan and Weinberg, 2011). Extensive work in experimental models has tried to define the multistep metastatic process that appears to be rate limiting. Whereas >80% of intravenously implanted carcinoma cells succeed in extravasation, only <3% survive in the parenchyma of the foreign tissue and start to form micrometastases (Luzzi et al., 1998). However, biological heterogeneity as well as drastic alterations in growth factor-signaling, cell–cell adhesion, gene expression, motility or cell shape of disseminated cancer cells in contrast to cancer cells of the primary neoplasm supports escape mechanisms for metastases to overcome conventional therapies. To gain mechanistic insights regarding metastatic cancer cells versus cancer cells of the primary tumor, we have generated two melanoma lines from (i) a primary cutaneous melanoma (CM) and (ii) a metastatic lymph node (LN) from the ret-transgenic melanoma model, which spontaneously develops primary malignant melanoma and distant metastasis (Helfrich et al., 2010; Kato et al., 1998). We provide here a detailed in vitro characterization but also describe the metastatic potential and homing specificity in vivo of the newly established melanoma cell lines. Histologically based analyses of 128 ret-transgenic, tumor-bearing mice revealed infiltration of highly pigmented melanoma cells into all lymph nodes of all animals as well as an 80% frequency for distant spleen metastasis. For the establishment of new cell lines, we isolated the first macroscopically visible primary skin tumor from a 25-day-old ret-transgenic mouse and a metastatic inguinal lymph node from a tumor-bearing mouse with a tumor load of >20 tumors (Figure 1A, left). Both cell lines analyzed by phase contrast showed similar cell morphology (Figure 1A, middle). Immunohistological analyses revealed in both lines comparable expression of melanocytic markers such as gp-100, tyrosinase-related protein-2 (TRP-2), and tyrosinase (data not shown), enzymes, all involved in regulating pigment production in melanocytes (Figure 1A, right). To address the functional behavior of both cell lines, we used a real-time in vitro measurement technique to overcome the problem of end-point observations. Dynamic monitoring of adherent cell proliferation was performed by impedance measurement using the xCELLigence system. This analysis identified higher proliferative activity of the CM as compared to the LN cell line (P = 0.008, Figure 1B, C). In addition, we observed significantly enhanced migratory capacity (P < 0.0001, Figure 1D, E) in the CM line but, to our surprise, a clear lower invasive potential (P = 0.003, Figure 1F, G) as compared to the LN cells. In consequence, we asked whether the expression or activity of matrix metalloproteinases (MMPs), enzymes known to be key regulators of tumor invasion (Airola et al., 1999; Shuman Moss et al., 2012), and their endogenous inhibitors, the tissue inhibitor of metalloproteinases (TIMPs) (Stetler-Stevenson, 2008) are responsible for this functional difference. Previous studies revealed association of MMPs, in particular the gelatinases MMP-2, and MMP-9, with tumor invasion and progression, demonstrating enhanced expression in tumor cells, including melanoma (Airola et al., 1999; Stetler-Stevenson, 2008). Whereas activation of proMMP-9 (the inactive form of MMP-9) can be mediated by different serine proteases, activation of proMMP-2 occurs primarily via the formation of a trimolecular complex consisting of membrane bound MT1-MMP, TIMP-2, and MMP-2 (Birkedal-Hansen, 1995; Butler et al., 1998). Thus, the expression of the inhibitor is necessary for regulating both enzyme activation and activity. Quantitative expression analyses revealed significant differences in expression of MMP-9 (~11.5-fold) and MT1-MMP (~6.5-fold), but also of their corresponding inhibitors TIMP-1 (~6.5-fold) and TIMP-2 (~2.2-fold) in CM cells (Figure 2A), whereas all transcripts were barely detected in LN cells (Figure 2A). In agreement with these data, zymographic analyses displayed well-detectable proMMP-9 in CM, but not in LN cells supernatants, whereas proMMP-2 was very little detected in both cells (Figure 2B). We additionally analyzed the quantitative differences of TIMP-1 expression by Proteome Profiler analyses and revealed a drastic increase in TIMP-1 secretion (P = 0.0004) as well as a ~3.5-fold higher protein expression (P = 0.0046) in lysates of CM compared with the LN cell line (Figure 2C). Thus, increased TIMP-1 expression may reduce the invasive capacity of the generated melanoma cell lines likely via inhibition, among other proteases, of MMP-9 activity. To prove this hypothesis, we inhibited TIMP-1 in the CM cell line using siRNA technology and analyzed the invasion potential. Transfection of TIMP-1 siRNA in CM cells reached a ~ 80% transfection efficiency (Data S1A) and led to 93–98% reduced transcript expression at 24 h and up to 3 days after transfection (Data S1B and Figure S2D). In consequence, reduction in TIMP-1 expression in CM cells resulted in a significant increase in the invasive capacity (P = 0.006, Figure 2E, F). Thus, taken together, increased extracellular TIMP-1 may block the activity of MMP-9 and that of other TIMP-1-sensitive proteases leading to the reduced invasion in CM cells in vitro. Cancer cells exhibit substantial phenotypic heterogeneity when measured for their capacity to grow under various conditions. The ability to exhibit anchorage-independent growth (colony-forming capacity in semisolid media) has been connected with tumor cell aggressiveness but also utilized as a marker for in vitro transformation (Mori et al., 2009). Two days after seeding the cells, both cell lines started to proliferate. To our surprise, despite a ~ 90% decrease in number of colonies with a diameter of <50 μm (Figure 3B), LN cells completely failed to develop spherical colonies above 50 μm as compared to the CM cell line (P = 0.0045) (Figure 3A, B). Multiple genetic factors for anchorage independency have been identified (Cifone and Fidler, 1980; Mori et al., 2009); however, the detailed molecular signature for this phenotypic behavior is largely unknown. One pathway described to be critically involved is the interplay of the colony-stimulating factor 1 (CSF-1) with its corresponding receptor (CSF-1R) whose overexpression leads to increased cell proliferation, rapid anchorage-independent growth, and aggressive tumor formation in ovarian (Keshava et al., 1999) and breast cancer cells (Sapi et al., 1996). Expression analyses of our newly generated cell lines revealed a nearly 10-fold higher transcript expression of CSF-1 in CM compared with LN cells (Figure 3C), which we could confirm also on the protein level (data not shown). Importantly, expression of the CSF-1R receptor could exclusively be detected in the CM cell line (Figure 3C). In agreement with the aforementioned studies, our results implicate that the signature of anchorage-independent growth may be mediated by this ligand/receptor molecule complex. To analyze this, we blocked the CSF-1R in CM cells using the selective and well-established CSF-1R inhibitor PLX6134 (GW2580) (Hume and MacDonald, 2012; Priceman et al., 2010). Daily treatment of PLX6134 (1 μM in DMSO) or DMSO alone (ctrl) over a period of 12 days resulted in ~98% inhibition of CSF-1R in CM cells (Figure 3D) without affecting cell morphology or viability (data not shown) and, in consequence, resulted in complete failure of colony formation (Figure 3E). To verify the impact of this mechanism for the disease of malignant melanoma, we additionally used the experimental metastatic B16 melanoma variants (Fidler, 1975; Nakamura et al., 2002; Poste et al., 1980). First, we observed a high heterogeneity in pigmentation (Figure S2A) and expression of CSF-1 and CSF-1R (Figure S2B) of the metastatic melanoma cell lines B16F0 (parental line), -F1 (derived from pulmonary metastasis after i.v. injection of -F0 into a syngeneic C57BL/6), and -F10 [ten-times passage of lung colonies using in vivo–in vitro selection based on Fidler's method (Fidler, 1975)]. Nevertheless, all three cell lines express CSF-1 and CSF-1R and presented colony formation and growth in semisolid media. In addition, blockage of CSF-1R using PLX6134 resulted, as shown for the CM cells, in loss of anchorage-independent growth (Figure S2C). Highly aggressive melanoma cells have been shown to be able to form perfusable, matrix-rich, vasculogenic-like networks in three-dimensional matrices in vitro, and corresponding structures has been found in aggressive tumors of patients with melanoma (Maniotis et al., 1999a). This phenomenon is called ‘vasculogenic mimicry’ in which some melanoma cells appear to acquire the capability to form blood channels in the absence of endothelial cells (Hendrix et al., 2001; Maniotis et al., 1999a). Using three-dimensional matrices, resembling the matrix-rich networks observed in aggressive patients with tumor (Seftor et al., 2012), individual cells of both cell lines started to change their morphology and created a tubular-like phenotype three hours after seeding (Figure 4A). After initial development of vasculogenic channel formation, LN cells failed to establish well-defined tubular networks in comparison with the established cell line CM, which we measured and statistically quantified by the number of branching points per growth area (Figure 4B, P ≤ 0.0001 for all indicated time points). In agreement to the ability of anchorage-independent growth, this assay pointed again the high-aggressive potential of the CM cell line, established from the primary melanoma. We have then analyzed the vasculogenic potential of the metastatic B16 variants, generated by serial transplantation, and well characterized for their increase in invasive potential from B16F0 to B16F10 (Fidler, 1975; Fidler and Kripke, 1977). Interestingly, the cell line B16F0 showed initial tubular structure formation but failed to build well-defined vascular channels as detected in B16F1 matrigel cultures (Figure S3). Similarly to the LN cells, the highly metastatic cell line B16F10 did not display any potential for vascular mimicry. These data indicate that melanoma cells retain in vitro an individual signature which defines them as aggressive, tumor forming, or metastasizing, that are able to disseminate to distant organs. Based on our results, we asked whether the new established cell lines CM and LN would show tumor formation after re-implantation in vivo. Syngeneic transplantation of both novel melanoma cell lines using s.c. injection into wild-type recipient mice resulted in tumor initiation and growth progression. However, the cell line CM showed a significantly higher growth kinetic starting after 3 days post-transplantation and for all indicated time points (*P < 0.05, ** P < 0.001) as compared to the LN cell line (Figure 5A). In parallel, we could also detect significantly higher tumor weights (P = 0.0073) at the end of the experiment in CM as compared to LN tumors (Figure 5B). The growth of solid tumors and the process of lymph node and organ metastases beyond a size of 1–2 mm3 are always accompanied by the initiation of angiogenesis for maintenance of expanding malignant cell population with oxygen and nutrients (Folkman, 1971). The vascular endothelial growth factor (VEGF)/VEGF receptor (R) and the angiopoietin (Ang)-2/Tie2 system are key regulators of the angiogenic cascade and therefore critically involved in the process of tumor progression and metastasis (Helfrich and Schadendorf, 2011). A few years ago, we identified Ang-2 as a secreted product of melanoma cells (Helfrich et al., 2009), which are also a well-established source of VEGF production (Kerbel, 2008). Quantitative analyses of both new established cell lines in vitro revealed the expression of VEGF-A and the blood vessel-destabilizing Tie2 ligand Ang-2 (Figure S4A) without detection of corresponding receptors, known to be primarily expressed on endothelial cells (Figure S4B). Next we asked whether the generated cell lines are able to form metastases in vivo. After intracardiac re-implantation of the CM cell line, 70% (7/10) of the wild-type recipients developed spleen metastasis, whereas ~45% (3/7) of these mice showed additional nodal micrometastasis. These data suggest that the cell line generated from the single primary melanoma is able to reflect the metastatic profile of the endogenously driven, ret-transgenic mouse model, in which tumor cell dissemination into the spleen and lymph nodes occurs during development of multiple spontaneous melanoma (Helfrich et al., 2010; Kato et al., 1998). To our surprise, retransplantation of the LN cell line, which was generated from nodal metastasis of a tumor-bearing ret-transgenic mouse, exclusively resulted in organ-specific metastatic infiltration of lymph nodes in 80% (8/10) of wild-type recipients, the organ of primary localization, and no colonization to the spleen. Metastatic spread into the liver or lung could not be detected for both cell lines. We quantified the metastatic tumor load by counting the melanin-containing tumor cell infiltrates in both organs. To avoid false-positive results caused by melanin-containing macrophages, so called melanophages, we excluded pigmented cells co-expressing the macrophage marker F4/80. Immunohistochemical analysis confirmed the metastatic profile which we first observed on the basis of macroscopic evaluation (Figure 5C). Spleens of CM-transplanted animals showed a mean infiltration of ~36 tumor cells/mm2. Remarkably, we revealed a significantly higher amount of nodal micrometastasis load (P = 0.0052) in LN- compared with CM-transplanted mice (Figure 5D), implicating a preferential colonization of this cell line into the organ of primary localization. In 1889, Stephen Paget published the seminal ‘seed and soil’ hypothesis to explain the pattern of metastasis. He proposed an organ-preference pattern of tumor metastasis as a product of favorable interactions between metastatic tumor cells (the ‘seed’) and their organ microenvironment (the ‘soil’) (Fidler and Poste, 2008). Other challenged his ideas by postulating that the primary factor that determined the patterns of tumor metastasis was the anatomy of vascular and lymphatic drainage from the site of the primary tumor (Zlotnik et al., 2011). However, for a successful metastasis, tumor cells must respond to chemotactic signals that guide the cells to the new microenvironment, and they must survive and thrive upon arrival. Chemotactic cytokines have been shown to be involved in both processes (Bendall, 2005; Zlotnik et al., 2011), and recent reports suggest that the expression of chemokines and chemokine receptors by melanoma cells may contribute to a preferential pattern of metastasis via their ability to escape tumor surveillance (Payne and Cornelius, 2002). Despite having identified differentially expressed molecules contributing to in vitro cell invasiveness and colony-forming activity, namely TIMP-1 and CSF-1/CSF-1R, respectively, we still not know the underlying pathway forcing the metastatic profile in our new established murine melanoma cell lines. However, to our understanding, these are newly generated mouse melanoma lines that may offer a new in vitro platform for further investigations and detailed studies on cancer cell homing and metastasis in melanoma. In this context, it would be of considerable interest to analyze the impact of chemokine/chemokine receptor signaling for the process of organ-specific metastasis in melanoma by gain- and loss-of-function experiments using both novel melanoma cell lines. Results of such studies would lead to clinical studies trying to combine modalities targeting the primary tumor as well as metastatic dissemination of malignant melanoma in general. We kindly thank Mashasi Kato (Department of Occupational and Environmental Health, Nagoya University Graduate School of Medicine, Japan) for providing the MT/ret model, Christiane Breuer, Mohamed Benchelall and Nadine Hochhard for excellent technical assistance and André Scherag (Institute for Medical Informatics, Biometry and Epidemiology; University Hospital, University Duisburg-Essen) for support in statistical analysis. This work was partially supported by the Melanoma Research Network of the Deutsche Krebshilfe (P.Z.) and the Hiege Stiftung gegen Hautkrebs (I.H.). Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.