Background
Malignant melanoma frequently metastasizes to the brain and is the third most common cause of brain metastases after lung and breast cancer [
1]. Clinical studies show that 30–50 % of patients with advanced-stage melanoma are diagnosed with brain metastases during the course of their disease, and the incidence of brain involvement increases to 50–75 % in autopsy studies [
2‐
5]. Current treatment options for brain metastases, including surgical resection, stereotactic radiosurgery, and whole-brain radiotherapy, are mostly palliative and show low efficacy in melanoma patients [
5,
6]. Consequently, the prognosis of melanoma patients diagnosed with brain metastases is generally poor, and many studies report a median survival of less than 6 months after diagnosis [
4,
7,
8].
Development of improved treatment strategies for patients with melanoma brain metastases requires detailed understanding of the underlying pathobiology. Most clinical data are derived from retrospective studies based on either diagnostic CT or MRI images or on histological analysis of autopsy specimens. Such studies have shown that multiple lesions and intratumoral hemorrhage are characteristic features of melanoma brain metastases and they have identified important negative prognostic factors, including multiple brain lesions, the presence of extracranial metastases, and meningeal involvement [
4,
7‐
10]. Studies of patient biopsies and autopsy specimens have also suggested that diffuse infiltration and growth along brain microvessels are common growth patterns in melanoma brain metastases [
11,
12].
Although clinical studies have provided valuable insight into the biology of melanoma brain metastases, preclinical studies are required to study biological properties at a detailed level and to elucidate underlying mechanisms. Melanoma brain metastases have been studied preclinically by using animal models that include spontaneous brain metastases after orthotopic implantation, direct intracerebral tumor cell implantation, intra-arterial injection of tumor cells into the heart or the internal carotid artery, and cranial window chambers [
13‐
15]. Two different patterns of melanoma metastases after intracarotid artery injection have been identified. Thus, some melanoma cell lines produce lesions in the brain parenchyma whereas others produce lesions preferentially in the meninges and ventricles [
16,
17]. This site-specificity has been associated with expression of the transforming growth factor-β2 in a murine melanoma model [
18]. Data on the microscopic invasion patterns within and between different brain compartments are however sparse. Preclinical studies have also demonstrated that vascularization of melanoma lesions within the brain parenchyma occur by both angiogenesis, i.e. sprouting of new blood vessels from pre-existing vessels, and by vascular co-option, i.e. growth of melanoma cells along pre-existing vessels [
15,
19‐
21]. Angiogenesis within the brain has been shown to depend on the vascular endothelial growth factor A (VEGF-A) in melanoma as well as lung, colon, and breast carcinoma models [
19,
22,
23]. However, so far, most studies demonstrating vascular co-option in melanoma brain metastases have used melanoma cell lines with low expression of VEGF-A [
15,
20].
The human melanoma cell lines A-07, D-12, R-18, and U-25 have been established and extensively studied in our laboratory [
24], and they have been shown to differ substantially in angiogenic signature, angiogenic potential, and metastatic properties when inoculated orthotopically in nude mice [
25,
26]. The angiogenic potential of these melanoma cell lines is associated with expression of VEGF-A and interleukin 8 (IL-8) [
25]. Furthermore, the invasiveness of A-07 and D-12 cells in vitro has been shown to depend on the matrix metalloproteinases MMP-2 and MMP-9 [
27]. MMP-2 and MMP-9 are proteolytic enzymes that degrade type IV collagen and other components of the extracellular matrix, and they have been associated with tumor cell invasion, metastasis, and angiogenesis [
28].
The purpose of the study reported here was to use A-07, D-12, R-18, and U-25 cells as models to study intertumor heterogeneity in vascularity and invasiveness of melanoma brain metastases. Artificial brain metastases were established by intracerebral and intra-arterial injection of melanoma cells transfected with green fluorescent protein (GFP), and vascularization and invasion patterns in the brain parenchyma, ventricles, meninges, and scull bone were studied in detail by fluorescence imaging and histological analysis. We report that the melanoma cell lines showed highly different patterns of vascularization and invasion in the brain and, furthermore, these differences were associated with differences in expression of the angiogenic factors VEGF-A and IL-8 and the matrix metalloproteinases MMP-2 and MMP-9.
Methods
Mice
Adult (8–10 weeks of age) female BALB/c nu/nu mice were used as host animals. The mice were bred at our institute and maintained under specific pathogen-free conditions at a temperature of 22–24 °C and a humidity of 30–50 %. The animal experiments were approved by the Institutional Committee on Research Animal Care and were performed according to the Interdisciplinary Principles and Guidelines for the Use of Animals in Research, Marketing, and Education (New York Academy of Sciences, New York, NY).
Cell lines
A-07, D-12, R-18, and U-25 human melanoma cells [
24] were obtained from our frozen stock and maintained as monolayers in RPMI 1640 (25 mM HEPES and L-glutamine) supplemented with 13 % bovine calf serum, 250 μg/ml penicillin, and 50 μg/ml streptomycin. The cultures were incubated at 37 °C in a humidified atmosphere of 5 % CO
2 in air and subcultured twice a week. For in vivo experiments, cells constitutively transfected with GFP were used [
29]. Cells were harvested from exponentially growing cultures and re-suspended in Ca
2+-free and Mg
2+-free Hanks’ balanced salt solution (HBSS) before injection into animals.
Anesthesia
Intracerebral and intra-arterial injection of tumor cells were carried out with anesthetized mice. Fentanyl citrate (Janssen Pharmaceutica, Beerse, Belgium), fluanisone (Janssen Pharmaceutica), and midazolam (Hoffmann-La Roche, Basel, Switzerland) were administered intraperitoneally (i.p.) in doses of 0.63 mg/kg, 20 mg/kg, and 10 mg/kg, respectively. The body core temperature of the mice was maintained at 37–38 °C by using a heating pad.
Intracerebral tumor cell injection
The mice were fixed in a stereotactic apparatus (Model 900; Kopf Instruments, Tujunga, CA) for injection of tumor cells into the right cerebral hemisphere. The injection point was 2 mm anterior to the coronal and 1 mm lateral to the sagittal suture lines. A 100 μl Hamilton syringe with a 26-gauge needle was used to inject 6 μl of cell suspension at a depth of ~3 mm below the scull. To minimize tumor cell reflux, the cells were injected slowly and the needle was left in place for 2 min before it was slowly retracted. Three different cell numbers, 5×102, 3×103, and 1×104, were included in the experiments, and all results were independent of cell number within this range. The mice were examined daily for up to 70 days after tumor cell injection. Moribund mice were killed and autopsied, and the brain was removed for subsequent fluorescence imaging or histological analysis.
Intra-arterial tumor cell injection
A tuberculin syringe with a 26-gauge needle was used to inject 105 cells suspended in 0.1 ml HBSS into the left ventricle of the heart. A small air-bubble was left in the plunger side of the syringe before injection to facilitate blood entrance. The ribs were visualized by a 1 cm skin incision along the sternum. The needle was inserted vertically into the third intercostal space 1–2 mm to the left of the sternum. Spontaneous and continuous entrance of pulsatile oxygenated blood into the needle hub was used as an indication of proper positioning in the left ventricle of the heart. The cells were injected slowly over a period of ~30 s. The mice were examined daily after tumor cell injection and moribund mice were killed and autopsied. The brain, spinal cord, axial and inguinal lymph nodes, lungs, kidneys, adrenal glands, pancreas, and liver were examined by visual inspection and by fluorescence microscopy.
Fluorescence imaging
Imaging was performed with an inverted fluorescence microscope equipped with a filter for green light (IX-71; Olympus, Munich, Germany), a black-and-white CCD camera (C4742-95, Hamamatsu Photonics, Hamamatsu, Japan), and appropriate image acquisition software (Wasabi, Hamamatsu Photonics). The fresh brain was imaged immediately after autopsy. The dorsal and ventral surfaces of the brain as well as 3 coronal brain sections with a thickness of approximately 2 mm were imaged at low (x2) and high (x4-x10) magnifications.
Histological analysis
The brain was fixed in phosphate-buffered 4 % paraformaldehyde. Histological sections were stained with hematoxylin and eosin (HE) by using a standard procedure or immunostained by using a peroxidase-based indirect staining method [
30]. An anti-GFP rabbit polyclonal antibody, an anti-CD31 rabbit polyclonal antibody, an anti-MMP-2 rabbit polyclonal antibody, or an anti-MMP-9 rabbit polyclonal antibody (all from Abcam, Cambridge, United Kingdom) was used as primary antibody. Diaminobenzidine was used as chromogen, and hematoxylin was used for counterstaining.
Quantitative PCR
RNA isolation, cDNA synthesis, and quantitative PCR were performed as described in detail previously for cells in culture [
29]. Briefly, gene expression was assessed by using the RT
2 Profiler PCR Array Human Angiogenesis (PAHS-024A) from SABiosciences (Frederick, MD). Real-time PCR was performed on an ABI 7900HT Fast Real-Time PCR instrument (Applied Biosystems, Carlsbad, CA). Each tumor line was run in three biological replicates. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and β-actin (ACTB) were used as normalization genes because these housekeeping genes showed stable expression across the melanoma lines studied here. Thus, each replicate C
T-value was normalized to the mean C
T-value of GAPDH and ACTB (ΔC
T = C
T
gene of interest – C
T
mean of GADPH and ACTB).
ELISA
Medium samples from cell cultures in exponential growth were collected 24 h after change of medium. Commercial ELISA kits (Quantikine; R&D Systems, Abingdon, United Kingdom) were used according to the manufacturer’s instructions to measure the concentrations of VEGF-A, IL-8, and MMP-2 in the medium samples as described previously [
25].
Statistical analysis
Statistical comparisons of data were performed by one-way analysis of variance followed by the Student-Neuman-Keuls test when the data complied with the conditions of normality and equal variance, and under other conditions by the Kruskal-Wallis one-way analysis of variance on ranks. Statistical comparisons of survival curves were performed using the log-rank test. Probability values of P < 0.05 were considered significant. Statistical analysis was performed with SigmaStat statistical software (SPSS, Chicago, IL).
Discussion
The human melanoma cell lines A-07, D-12, R-18, and U-25 showed aggressive intracranial growth following intracerebral implantation, and the survival of the mice was limited by tumor growth in the meninges and ventricles rather than by local tumor growth at the injection site. A similar growth pattern was observed after intra-arterial injection, suggesting that a preference for growth in the meninges and ventricles was a characteristic feature of the melanoma cells. All four cell lines are established from subcutaneous metastases [
24], and the results are thus in accordance with previous studies showing that human melanoma cell lines derived from non-CNS metastases produce lesions mainly in the meninges and ventricles after intracarotid artery injection, whereas cell lines derived from brain metastases produce lesion mainly in the brain parenchyma [
16,
17]. In the present study, the melanoma cells induced angiogenesis in the meninges, whereas angiogenic activity was not detected within the brain parenchyma. Blood vessels in the brain parenchyma form a tight blood brain barrier, whereas meningeal capillaries are lined by fenestrated endothelium and are thus more similar to capillaries found in most extracranial organs [
16,
17]. The melanoma cells’ preference for growth in the meninges may therefore reflect differences between meningeal and parenchymal blood vessels in their susceptibility to angiogenic factors secreted by the melanoma cells.
The aggressiveness of meningeal tumors differed substantially among the melanoma lines, and these differences were associated with differences in angiogenic activity and in expression of VEGF-A and IL-8. Thus, A-07, D-12, and U-25 cells grew more aggressively in the meninges, showed higher microvascular density in leptomeningeal tumors, and showed higher expression and secretion of VEGF-A and IL-8 than R-18 cells. In contrast, all four melanoma lines appeared to rely primarily on vascular co-option for growth and invasion of the brain parenchyma regardless of VEGF-A expression. Our data thus suggest that also melanoma cells with high expression of VEGF-A may show a co-optive and infiltrative growth pattern within the brain. The observation time for tumor growth in the brain parenchyma was limited by the aggressiveness of the meningeal tumors and, therefore, it is possible that also parenchymal lesions induce angiogenesis in a later growth stage. For example, parenchymal lesions may develop regions with hypoxic tissue during expansion, and changes in the extent of hypoxia during growth may cause dynamic changes in the expression of angiogenic factors influencing the form of vascularization, leading to vascularization by angiogenesis in addition to co-option in large lesions.
The correlations reported here suggest that VEGF-A and IL-8 are required for the growth of brain metastases in our artificial metastasis models. This suggestion is in accordance with an earlier work with the same melanoma lines showing that treatment of tumor-bearing mice with neutralizing antibody against VEGF-A or IL-8 decreased the microvascular density and the growth rate of the primary tumor and reduced the incidence of spontaneous metastases, including lymph node, pulmonary, hepatic, and brain metastases [
26]. Our suggestion is also consistent with a study by Yano et al. [
22], reporting that VEGF-A expression is necessary but not sufficient for production and growth of brain metastases.
The melanoma lines showed highly different invasion patterns, and differences in invasiveness was associated with differences in expression of MMP-2 and MMP-9. Thus, R-18 cells invaded locally both within the brain parenchyma at the injection site and within the leptomeninges and ventricles, but showed low ability to invade across the barriers separating different brain compartments. In contrast, A-07, D-12, and U-25 cells frequently invaded the brain parenchyma from the meninges or ventricles. The brain parenchyma and the pia mater are separated by the glia limitan, consisting of astrocyte foot processes attached to a parenchymal basal lamina [
31]. MMP-2 and MMP-9 have previously been shown to facilitate the migration of leukocytes across the glia limitans [
31,
32]. In this study, we found higher expression of MMP-2 in the A-07, D-12, and U-25 lines than in the R-18 line, and higher expression of MMP-9 in the A-07 line than in the other lines. Lower expression of these matrix metalloproteinases could thus potentially explain the low invasiveness of R-18 cells across the glia limitan.
A-07, D-12, R-18, and U-25 cells growing in the meninges also differed considerably in their ability to invade and destruct the overlying scull bone. These differences were not directly related to angiogenic activity in the meninges. Thus, although A-07 and U-25 mice showed similar aggressiveness and angiogenic activity in the meninges, the ability to invade and destruct the scull bone was substantially higher for U-25 cells than for A-07 cells. Interestingly, the fraction of mice with macroscopic scull bone destruction was closely associated with the expression of MMP-2. Overexpression of MMP-2 has previously been shown to promote both brain and bone metastasis in a preclinical model of breast cancer and to promote bone absorption in a prostate cancer model [
33,
34]. Our data suggest that MMP-2 may be associated with bone invasion also in melanoma, and this warrants further investigations.
In a previous study, we revealed that A-07, D-12, and T-22 melanoma cells cultured at acidic extracellular pH show increased secretion of MMP-2 and MMP-9, enhanced invasiveness in vitro, and enhanced potential to develop experimental pulmonary metastases in BALB/c
nu/nu mice [
27]. Moreover, acidity-induced experimental pulmonary metastasis was inhibited by treatment with the general MMP inhibitor GM6001 and the general cysteine proteinase inhibitor E-64 [
27], suggesting that MMP-2 and MMP-9 are required for the development of pulmonary metastases in our melanoma models. Because brain metastases frequently originate from pulmonary metastases, it is likely that MMP-2 and MMP-9 are required for the development of spontaneous brain metastases also. Other members of the MMP family may also be involved in melanoma brain metastasis; however, we have no evidence from our models that MMP-1, MMP-13, or MT1-MMP is required for metastatic growth in the brain.
Melanomas have a high tendency to produce multiple brain lesions, and Kienast et al. have suggested that this tendency is the result of a high motility of melanoma cells within the brain [
15]. Our data are consistent with this suggestion, showing that melanoma cells injected into a specific location in the right cerebral hemisphere are able to produce multiple brain lesions involving brain compartments far from the injection site. Moreover, our data show that the melanoma cells used the meningeal surfaces of the brain as transport routes. Leptomeningeal involvement has been shown to be correlated with an increased number of parenchymal lesions in a large retrospective study of patients with stage IV melanoma [
8]. Furthermore, autopsy studies have suggested that the true incidence of leptomeningeal metastasis may be substantially higher than the clinical incidence [
2,
3]. Consequently, efficient treatment strategies for patients with melanoma brain metastases may need to target melanoma cells in the meninges as well as in the brain parenchyma.
New treatment strategies for patients with melanoma brain metastases are highly needed, and antiangiogenic therapy targeting the VEGF-A pathway has been suggested as a potential strategy [
35]. However, patients who harbor brain metastases with a co-optive growth pattern will most likely not respond to antiangiogenic agents. Although VEGF-A-dependent angiogenesis is a characteristic feature of malignant melanoma [
36], anti-VEGF-A therapy may have limited potential in patients with melanoma brain metastases due to a tendency of melanoma cells to grow by vascular co-option in the brain despite high expression of VEGF-A. Metastases with a co-optive growth pattern are also more difficult to treat with surgical resection or stereotactic radiosurgery due to the diffuse borders between tumor and normal tissue.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
TGS, JVG, and EKR conceived and designed the study. TGS performed the experiments. TGS, JVG, and EKR analyzed and interpreted the data. TGS wrote the manuscript. All authors read and approved the final manuscript.