Tumor Cell Plasticity and Angiogenesis in Human Melanomas

Recent molecular studies provide evidence for a significant transcriptional plasticity of tumor cell subpopulations that facilitate an active contribution to tumor vasculature. This feature is accompanied by morphological changes both in vitro and in vivo. Herein, we investigated the morphological plasticity of tumor cells with special focus on vasculogenic mimicry and neovascularisation in human melanoma and mouse xenografts of human melanoma cell lines. In melanoma xenograft experiments, different vessel markers and green fluorescent protein expression were used to show how melanoma cells contribute to neovascularization. Additionally, we analyzed neovascularization in 49 primary melanomas and 175 melanoma metastases using immunostaining for blood (CD34) and lymphatic (D2–40) vessel-specific markers. We found significantly more lymphatic vessels in primary melanomas than in melanoma metastases (p<0.0001). In contrast to the near absence of lymphatic vessels within metastases, we found extensive blood micro-neovascularization. Blood micro-neovascularization was absent in micro metastases (less than 2 mm). A significant inverse correlation between Glut-1 expression (implying local hypoxia) and the presence of microvessels indicates their functional activity as blood vessels (p<0.0001). We suggest that the hypoxic microenvironment in metastases contributes to a phenotype switch allowing melanoma cells to physically contribute to blood vessel formation.


Introduction
Melanoma is a highly aggressive neoplasm. Even very thin primary tumors (Breslow tumor thickness ,1 mm) may seed metastases and precipitate rapid death [1,2]. Several cancer studies, including melanoma, have identified primary tumor lymphangiogenesis and its predictive relevance for lymph node metastasis [3,4,5,6]. Interestingly, animal tumor models for melanoma and breast cancer have revealed vasculature reorganisation in regional lymph nodes prior to invasion by metastatic cancer cells [7,8,9,10].
The introduction of high through put assays such as gene expression arrays have offered new tools for cancer classifications including melanoma. [11]. Recently, we have demonstrated two opposing gene expression profiles in melanoma short term cell cultures associated with proliferative or invasive cell phenotypes and provided evidence for phenotype switching favouring disease progression [12,13]. This transcriptional plasticity is accompanied by morphological changes in vitro, and involves many factors governing angiogenesis. In 1999, Maniotis and co-workers introduced the concept of vasculogenic mimicry (VM) in melanoma [14]. He defined VM as melanoma cells with endothelial-like morphology embedded in a PAS and collagen IV positive extracellular matrix. The matrix with endothelial-like melanoma cells forms loops or networks which may be solid or hollow. VMs have been seen in additional malignant tumor types including breast, liver, prostate and other cancers [15,16,17,18]. The presence of VMs have been associated with a poor clinical outcome [14,19,20,21]. Furthermore, mosaic blood vessels in which both endothelial cells and tumor cells form the luminal surface have been described for both melanoma and colon cancer [22,23,24]. Most recently, sophisticated investigations have shown that tumor cells directly participate in blood vessel formation [25,26].
In the present study we investigated melanoma metastases for neovascularization in both human biopsies and mouse xenografts of human melanoma cell lines. We found VMs and a tremendous blood micro-neovascularization in human biopsies as well as in xenografts. In the mouse xenografts of human melanoma cell lines we showed melanoma cell participation in capillary formation.

Tissue microarray analysis
Immunohistochemistry was performed on three different tissue microarrays (TMAs) representing a total of 49 primary melanomas and 127 melanoma metastasis. The metastasis array included 21 lymph node and 106 organ metastases. Construction of TMAs was performed as previously described [27]. Furthermore, in order to avoid sampling error immunohistochemistry was performed on whole tissue sections of 35 lymph node and 13 organ metastases. Approval for the use of melanoma TMAs and melanoma metastases was obtained from the official ethical authorities of the Canton Zurich. The consent statement was written (StV , permission submitted in the supplement).

Human tumor cell isolation and xenotransplantation
Studies described in this publication were performed according to Swiss Animal Welfare laws and specifically as described in animal licenses No. 77/2008 (Int. 3816) issued and approved by the Cantonal Veterinary Office (official state office), of the Canton of Zurich (permission submitted in the supplement). The approval was signed by Mrs. Dr. Simone Gilg, scientific board deputy and responsible for the review board of the Cantonal Veterinary Office, of the Canton of Zurich. The approval ID/permit number is 2008077.
The melanoma cell line (M010308) and the melanoma cell lines used for chromosome 17/HER-2 probe were obtained from consenting patients, according to the guidelines of the University Hospital of Zurich. The permission to produce cell cultures from fresh tumor tissue was given from the cantonal ethic commission in Zürich. The consent statement was written (EK-No. 647, see attached permission in the supplement). Melanoma cell lines used for the chromosome 17/HER-2 probe are additionally described in more detail in a recent publication [28]. HEK293 cells were bought from GenHunter (www.genhunter.com; cat. number Q401).
In order to demonstrate mosaic microvessels 2 xenotransplantations with male/female constellation were established as follows: a male melanoma cell culture (M010308) established 4 years ago which went through at least 20 passages, which was frozen and re thawed several times and finally was used for xenotransplantations in female athymic nude mice. The cell line (M010308) was proven to be negative for CD31 and CD34 by immunohistochemistry (data not shown). From the melanoma culture, a total of 3610E6 cells were injected into both flanks of 8-week-old female athymic nude mice. Mice were kept in individually ventilated cages for a maximum of 75 days postinjection. Tumour volume was measured using vernier calipers once every 1 or 2 days during the linear growth phase. If at least one xenograft tumor reached 1 cm 3 , the mouse was sacrificed and the tumors were investigated with the ychromosome probe.
In addition using melanoma patients cell line [28] 4 xenotransplantations were established in the following way: Immediately after surgical resection, solid metastatic lesions were dissociated into single-cell suspensions using HBSS (without Ca 2+ and Mg 2+ , Invitrogen) containing Collagenase III (1 mg/ml, Worthington Bioch.) and Dispase (0.5 mg/ml, Roche). Incubation at 37uC for 1 h with concurrent mincing allowed complete digestion. The resulting cell suspension was filtered through 40 mm nylon mesh and single cells were harvested. Afterwards the cells were cultured in neuro-sphere conditions and passed 8 times. Due to the limited amount of cells, patient tumor xenografts were established by subcutaneous injection of dissociated bulk melanoma cells into nude BALBc/Swiss mice. Resulting patient tumor xenografts were harvested, dissociated as described above. Thereafter tumor cells were labelled with the neuroectodermal marker p75, sorted by FACS and again injected into nude athymic mice. The resulting tumors were investigated by chromogenic in situ hybridization for the human HER-2 probe localized on chromosome 17.

Human X/Y-Chromosome/CD31 (FICTION)
FISH analysis for CEP X (DXZ1) Spectrum Green/CEP Y (DYZ3) Spectrum Orange Probe (ABBOTT Molecular) was performed according to the manufacturers' protocol. FISH procedure was followed by the detection of human endothelial cells: Mouse anti-human CD31 (DAKO, dilution 1:10) was coupled with a fluorescence labeled secondary antibody: Goat anti-mouse Alexa488 (Molecular Probes, dilution 1:100), counterstain with DAPI (Incubation at room-temperature for 30 min each).
A cell was defined as human-derived as follows: on confocal microscopy, the nucleus had to present typical endothelial morphology, being positive for the Y-probe, and the CD31 positive cytoplasm had to be fully discernible.

Human ALU-repeat probe and CD31
All procedures were done on BondMax (Leica) with Kits and solutions from Leica according to the manufacturer's guidelines.

Fluorescence Imaging
Confocal image acquisition was done using sequential mode on a Leica TCS SP5 confocal laser scanning microscope (Mannheim, Germany). Co-localization studies were conducted using IMARIS software (Bitplane, Zurich, Switzerland).

Melanoma cell infection with GFP-labelled lentivirus
Freshly isolated melanoma cells were injected into nude athymic mice. Patient tumour xenografts were harvested and cells dissociated as described above. Some of those cells were cultured in neuro-sphere condition and passed 8 times. At the 8th passage the cells were infected with a GFP driving virus.
GFP-expressing lentivirus was produced by transient fourplasmids cotransfection into HEK 293 [29] with VSV-pseudotyped third-generation lentiviral vectors (LV) by using calcium phosphate method. Supernatants were collected 48 hours after transfection, filtered through a 0.22 mm membrane [30] and transferred to target cells. Resulting GFP-tagged tumor cells were checked by FACS and subcutaneously injected as described above.

Statistical analyses
CD34 and D2-40 expressions are presented as median with range and compared between different patient groups using the Mann-Whitney and the Krustal-Wallis test. Glut-1 expression was compared using Fisher's exact test. Differences between CD34 and D2-40 expressions were compared using the Wilcoxon signed rank test. P-values below 0.05 were considered as significant. SPSS 16.0 for Windows (SPSS Inc., Chicago, IL) was used for statistical analyses.

Primary melanomas show a higher lymphatic vessel density than melanoma metastases
We investigated 127 melanoma metastases (21 lymph nodes and 106 other sites) and 49 primary melanomas for vascularisation. The number of vessels was counted per core. There were significantly more D2-40 positive lymphatic vessels in primary melanoma than in melanoma metastases (p,0.0001; Figure 1A).
Only 15 of 127 (12%) metastases had lymphatic vessels whereas 39 of 49 (80%) primary melanomas were positive for lymphatic vessels. Metastases had a mean of 0.6 lymphatic vessels per core (median 0; range 0-15) in contrast to primary tumors with a mean of four lymphatic vessels per core (median 3; range 0-15; Table 1). Analyzing a separate collection of 48 complete paraffin-embedded metastases (35 lymph node and 13 organ metastases) confirmed negativity for lymphatic vessels. 18 of 48 (38%) metastases were completely negative for lymphatic vessels, while 23 (48%) showed foci of dilated lymphatic vessels at the interphase to normal organ parenchyma. All but three of these were lymph node metastases ( Figures 1B, C). Only seven (14%) had few lymphatic foci within the metastatic body. However, normal skin 0,5 cm adjacent to the primary melanoma showed an average of 8 vessels per mm 2 which is congruent with the literature [31]. In contrast lymph node tissue 0,5 cm adjacent to the metastases showed no lymphatic vessels. However, in all 48 cases blood neovascularization was prominent. The lymph node metastases comprised 20 macro-and 15 micrometastases (,2 mm). Congruent with Folkman's findings [32] we found neither lymphatic nor blood-micro-vessels in micrometastases.
There were significantly more blood than lymphatic microvessels in primary melanomas (p,0.0001) as well as in metastases (p,0.0001). In addition primary melanomas had significantly more blood vessels than metastases (p = 0.012). Primary melanomas had an average of 20 blood vessels per core (median 19; range 1-60) and metastases had an average of 15 blood vessels (median 11; range 0-69) per core (Table 1). There was no statistically significant difference in vessel density or composition between lymph node and organ metastases. Interestingly, there was a virtual homogenous distribution of blood neovascularization over cross section of metastases. In large metastases there was no difference between central and peripheral regions. We could not find blood vessel concentrations at the interface with normal lymphatic tissue as found for lymphatic vessels ( Figure 1D). Glut-1 expression in melanoma cells is an indicator of a hypoxic environment. We analyzed Glut-1 expression semi quantitatively and differentiated between Glut-1 expression in less or more than 50% of melanoma cells per core. There was a significant difference between primary melanoma and metastases (p = 0.002). 45 of 127 (35%) melanoma metastases showed Glut-1 expression in more than half of the cells whereas this was true for only 6 of 49 (12%) primary melanomas. A significant inverse correlation between blood vessel number and Glut-1 expression in melanoma (p = 0.0002) indicates the functionality of blood neovascularization.

Melanoma cells participate in neovascularization during tumorigenesis
We investigated six xenografts of human melanoma cell lines labelled with p75 and sorted by FACS as described in material and methods for MECA32, a marker specific for mouse vessel endothelia. There was significant neovascularization within the tumor (Figures 2A, B). Most endothelial cells surrounding microvessels were MECA32 positive, indicating their mouse origin. However, fluorescence double-labelling showed that some were negative for MECA32 and positive for human chromosome 17 and HER-2, indicating their melanoma origin ( Figure 2C). As they have the same staining properties and a similar endothelial morphology as melanoma cells of VMs we introduced for melanoma cells in VMs as well as for melanoma cells contributing to capillaries the term endothelial-like cell. For quantifying the percentage of endothelial-like tumor derived cells contributing to capillaries, chromogenic in situ hybridization for HER-2 gene probe and centromeric probe for chromosome 17 was used ( Figure 2D). Capillaries and fibroblasts of tumor adjacent soft tissue were used as a negative internal control ( Figure 2E). We identified 6% of all capillary lining cells as endothelial-like cells. Endothelial-like cells of VMs are not included in this account. In addition, tumor cell contribution to neovascularization was also confirmed by combined in situ-and immunofluorescence labelling in the mouse model displaying double-positivity of endothelial-like cells for the vessel marker CD31 and the human Y-chromosome ( Figure 2G) as well as for CD31 and the human-DNA specific ALU-repeat sequences ( Figure 3A-D). Mouse vessels of adjacent tissues served as internal negative controls (2F). Table 2

Vascular mimicries in human metastases and xenograft tumors
Vascular mimicries (VM) were observed in both human metastases (Figures 1E, F, G, 4A, B) and mouse xenografts ( Figure 2B, H). Melanoma cells contributing to VM were embedded in a PAS (1F) and collagen IV (1G) positive extracellular matrix, were small with endothelial-like cell morphology and highly condensed nuclei (Figures 1E, F, G). Most of these cells were negative for PanMelanoma (HMB-45 and MART-1). However, some contained melanin (Masson Fontana positive) or were positive for PanMelanoma ( Figures 4A, B) indicating their melanocytic origin. Most importantly, VM in xenografts were positive for the chromosome 17 and HER-2 probe ( Figure 2H) and GFP fluorescence labelling, which proofed their melanoma derivation (Figure 2 I, J).
In contrast to their endothelial-like morphology, they were negative for CD31, CD34 and D2-40. In xenografts tumor cell VM were associated with MECA32-positive mouse derived microvessels ( Figure 2B) whereas in human melanoma metastases they associated and connected with CD34-positive blood vessels ( Figure 1E).

Discussion
The establishment of metastases is a complex process, which includes angiogenic development. Several studies have shown that increased lymphatic density in primary melanomas is associated with a higher incidence of regional lymph node metastases [5,33].  To date there are few studies analyzing neovascularization in human cancers and these are mostly restricted to breast cancer [6,34,35,36]. In our study, we examined 175 melanoma metastases for lymphatic and blood vessel formation using immunohistochemistry. We found significantly more lymphatic vasculature in primary than in metastatic melanomas (p,0.0001).
In contrast to the lymphangiogenesis seen in primary lesions, there were few lymphatic vessels observed within lymph node and organ metastases. The few lymphatic vessels identified were mostly found in foci at the metastatic lymph node interphase, or in the lymph node parenchyma close to metastases, and these were often angiectatic ( Figures 1B, C). Our histopathological observations correspond well with van den Eynden's reports on breast cancer [34]. In melanoma metastases lymphatic foci have also been reported, however no information on their in situ distribution was provided [6].
However, we observed a significant anatomical difference in lymphatic density with a mean of 8 lymphatic vessels in normal skin 0,5 cm adjacent to the primary melanoma and absence in lymph node tissue 0,5 cm adjacent to the metastases. However, despite those anatomical differences, there still remained a significant difference in lymphatic vascularization between primary melanoma and adjacent normal tissue indicating the lymphatic induction potential of primary melanoma, which was almost absent in metastases. Lymphatic induction in primary melanoma and its prognostic relevance is well documented. In addition lymphatic induction is also a part of regression in primary melanoma [31,33].
Melanoma metastases presented a highly complex neovascularization with immunoreactivity for CD34 and negativity for D2-40. Furthermore, the highly significant inverse correlation between expression of Glut-1 (implying local hypoxia) and the presence of micro vessels suggests that hypoxia is a major driver (p,0.0001; Figure 1A). Interestingly there was a homogenous distribution of micro blood vessels over whole metastases. In terms of their structure, we observed that these tumor vessels are abnormal as their walls have numerous 'openings'. Ultrastructure studies of tumor vessels have shown endothelial fenestrae, transcellular holes, widened interendothelial junctions, and discontinuous or absent basement membranes. Such defects are known to make tumor vessels leaky [37] [38]. Increasing neovascularization with leaky vessels may contribute to hematogenous spread. The vascular  endothelial growth factor (VEGF), predominantly contributed by tumor cells is a major stimulator of angiogenesis [39]. Considering the tremendous neovascularisation of melanoma metastases, it is not surprising that inhibition of the VEGF signalling pathway has shown promising therapeutic benefits. Unfortunately, adaptive tumor responses are often observed, indicating the need for further understanding of VEGF regulation. Recently, a novel G proteincoupled receptor, GPR56, which inhibits VEGF production from the melanoma cell lines [40], as well as heparanase, which induces VEGF, were reported as possible tumor targets for influencing VGEF expression [41]. In contrast, one could speculate that antiangiogenic therapies result in hypoxic condition and therefore benefit the growing of VM which represent the aggressive invasive form of melanoma.
Similar blood neovascularization was found in xenografts of human melanoma cell cultures. These vessels were mostly positive for the mouse-specific vessel marker MECA32. However, about 6% of endothelial-like cells were negative for this marker. We propose that these MECA32-negative cells are tumor-derived because they are positive for the human chromosome 17 and HER-2 (Figure 2  . Taken together, we conclude that melanoma cells acquire vascular endothelial-like morphology and are incorporated in capillaries and thus actively contribute to the establishment of neovascularization [42]. These features are in congruence with findings from the laboratory of Mary Hendrix [23]. This group injected aggressive melanoma cells containing a fluorescent label into the ischemic hind limbs of nude mice, which were re-perfused five days later. Histological cross-sections of newly formed vasculature in re-perfused limbs showed human melanoma cells adjacent to and overlapping with mouse endothelial cells in a linear arrangement.
Cells involved in VM express genes associated with embryonic stem cells including those associated with primordial vascular development [14,19,20,21,43]. In a recent study, we have provided evidence for molecular plasticity in melanoma switching between proliferative and invasive phenotypes [13]. Culturing invasive phenotype melanoma cells on matrigel causes them to adopt VM morphologies. In addition, Hess et al have shown that epithelial cell kinase is a critical molecular determinant for VM in aggressive melanoma cells. Inhibition of tyrosine phosphorylation abrogates the ability of the aggressive melanoma cells to form VMs [21]. Therefore we postulate that VM tumor cells represent the invasive phenotype in vivo. Very recently, two independent groups revealed that a proportion of the endothelial cells of blood vessels in glioblastoma originate from the tumour cell population as proven by tumor specific genetic alterations [25,26].
Approximately one million tumor cells are shed per g of tumor per day [44]. Chang and Maniotis have shown in colon cancer and melanoma the functional activity of VM and their implication for tumor shedding [14,22]. Our results indicate that the melanoma lymph node metastases with tremendous neovascularization, VM and leaky vessels are a direct contributor to  hematogenous spread. The low percentage (6%) of tumor cells directly contributing to the capillary network is interesting concerning tumor cell plasticity and tumor cell niche aspects. Looking at the high number of tumor shedding, they at first seem irrelevant. However, they have a remarkable plastic capacity and by lack of melanocytic differentiation, antigens escape normal immune surveillance. Therefore, their influence in metastatic disease or tumor relapse should not be underestimated. In summary, we found a complex neovascularization consisting of VMs and leaky blood micro vessels in human melanoma biopsies and melanoma xenografts. In addition, melanoma cells contributed to tumor capillaries.