Metastatic pathway and the microvascular and physicochemical microenvironments of human melanoma xenografts

Adult (8–12 weeks of age) female BALB/c nu/nu mice, bred and maintained at our research institute [32], were used as host animals for xenografted tumors. Four CDX models (C-10, D-12, R-18, T-22) and two PDX models (E-13, N-15) of metastatic melanoma were included in the study. The CDX models and the patients from whom permanent cell lines were established in culture have been described elsewhere [36]. The E-13 PDX model was obtained from a lymph node metastasis in the axilla of a 52-year-old female melanoma patient. The donor developed multiple pulmonary metastases and died from metastatic growth in the brain 18 months after presentation at the Norwegian Radium Hospital. The N-15 PDX model was derived from a lymph node metastasis in the abdomen of a 60-year-old male admitted to the Norwegian Radium Hospital for the treatment of metastatic melanoma. The donor patient responded poorly to chemotherapy and died from metastatic growth in the liver 27 months after the initiation of treatment. Both PDX models were established in BALB/c nu/nu mice by implanting small tumor pieces subcutaneously, and they have been maintained solely in vivo by serial subcutaneous transplantation of tumor cell aliquots. In the present study, we used orthotopic tumors initiated by intradermal inoculation of aliquots of ~ 3.5 × 10⁵ to ~ 1.0 × 10⁶ cells suspended in 10 μl of Hanks’ balanced salt solution. Tumors were included in experiments when having grown to a volume of 500–600 mm³. The volume (V) and the volume doubling time (T _d) of the tumors were calculated as V = π/6 × a × b × c and T _d = ln2 × t/(lnV _t − lnV ₀), where a, b, and c are three orthogonal tumor diameters measured with calipers, and V _t and V ₀ are tumor volume at time t and time zero, respectively.

Tumor-bearing mice were anesthetized with fluanisone (20 mg/kg body weight), midazolam (10 mg/kg body weight), and fentanyl citrate (0.63 mg/kg body weight), and IFP was measured in the tumor center with a Millar SPC 320 catheter equipped with a 2F Mikro-Tip transducer (Millar Instruments, Houston, TX, USA) [28]. The catheter was connected to a computer, and data were acquired with the LabVIEW software. Two independent measurements were carried out in each tumor by inserting the IFP probe from two opposite directions, and these measurements provided highly similar IFP values.

Pimonidazole [1-[(2-hydroxy-3-piperidinyl)-propyl]-2-nitroimidazole], a marker of tumor hypoxia, was administered to the mice in a dose of 30 mg/kg body weight immediately after tumor IFP was measured. The mice were euthanized 3–4 h later, and the primary tumor was removed for immunohistochemical assessment of hypoxic fraction (HF_Pim), vascular density, and the levels of angiopoietin-2, coagulation factor-III, tie1, and neuropilin-2. CD31 and LYVE-1 were used as markers of blood and lymph vessel endothelial cells, respectively. Histological sections were prepared by standard procedures, and immunostaining was carried out by using a peroxidase-based indirect staining method [29]. An anti-pimonidazole rabbit polyclonal antibody (provided by Professor Raleigh, University of North Carolina, Chapel Hill, NC, USA), an anti-mouse CD31 rabbit polyclonal antibody (Abcam, Cambridge, UK), an anti-mouse LYVE-1 rabbit polyclonal antibody (Abcam), an anti-human angiopoietin-2 rabbit polyclonal antibody (Abcam), an anti-human coagulation factor-III rabbit polyclonal antibody (Abcam), an anti-human tie1 rabbit polyclonal antibody (Abcam), or an anti-human neuropilin-2 rabbit polyclonal antibody (Abcam) was used as primary antibody. Diaminobenzidine was used as chromogen, and hematoxylin was used for counterstaining. Quantitative studies were carried out on preparations cut through the central regions of tumors and the surrounding tissue, and three sections of each staining were analyzed for each tumor. Staining densities were quantified by using the ImageJ software. Microvessels were defined and scored manually as described by Weidner [15]. Blood vessel density (BVD) in the invasive front (peripheral BVD) was assessed by counting vessels located within a 1-mm-thick band in the tumor periphery [32]. Peritumoral lymph vessel density (LVD) was determined by counting vessels in the surrounding skin located within a distance of 0.5 mm from the tumor surface [32]. HF_Pim was assessed by image analysis and was defined as the area fraction of the non-necrotic tissue showing positive pimonidazole staining [27].

The lungs and six pairs of lymph nodes (i.e., popliteal lymph nodes, inguinal lymph nodes, proper axillary lymph nodes, accessory axillary lymph nodes, medial iliac lymph nodes, and renal lymph nodes) were resected and examined for metastatic growth by light microscopy. Histological sections were cut at 100-μm intervals throughout the resected tissue and stained with hematoxylin and eosin. A group of five or more melanoma cells was scored as metastatic growth. Usually, the metastatic deposits had grown to a diameter of more than 100 μm when the lungs and lymph nodes were resected.

The RT² Profiler PCR Array Human Angiogenesis (PAHS-024Z; SABiosciences, Frederick, MD, USA) was used for expression profiling of angiogenesis-related genes. The genes included in the array are presented in Additional file 1: Table S1. Total RNA was isolated from tumor tissue stabilized in RNAlater RNA Stabilization Reagent (Qiagen, Hilden, Germany). RNA isolation, cDNA synthesis, and real-time qPCR were carried out as described earlier [26]. Fold difference in gene expression was calculated by using the ∆∆C_T-method [37]. A C_T-value of 35 (15 cycles above the positive PCR control) was set as detection limit, and consequently, genes with C_T-values above 35 were not included in the analysis. The arrays included five house-keeping genes [β-actin (ACTB), β-2-microglobulin (B2M), glyceraldehyde-3-phosphate dehydrogenase (GAPDH), hypoxanthine phosphoribosyltransferase-1 (HPRT1), ribosomal protein lateral stalk subunit P0 (RPLP0)], and each C_T-value of a tumor was normalized to the mean C_T-value of these genes (∆C_T = C _T ^{gene of interest}  − C _T ^{mean of housekeeping genes} ). Normalized gene expression levels were calculated from three tumors as \(2^{{ - \,{\text{mean}}\;\Delta {\text{C}}_{\text{T}} }}\).

Statistical analysis was carried out with the SigmaStat statistical software. Data are shown as mean ± standard error unless otherwise stated. The Kolmogorov–Smirnov method and the Levene’s method were used to test for normality and equal variance. Comparisons of data were carried out by using the Student t test (single comparisons) or by one-way ANOVA followed by the Bonferroni’s test (multiple comparisons) when the data complied with the conditions of normality and equal variance. Under other conditions, comparisons were carried out by non-parametric analysis using the Mann–Whitney rank-sum test (single comparisons) or by Kruskal–Wallis ANOVA on ranks followed by the Dunn’s test (multiple comparisons). Probability values of p < 0.05, determined from two-sided tests, were considered significant.

Metastatic frequency was determined by studying a total of fifty tumor-bearing mice of each melanoma model. Four experiments were carried out, each involving 10–15 mice per model, and metastatic growth was observed frequently, both in lymph nodes (Fig. 1a) and lungs (Fig. 1b). The incidence metastasis (i.e., the percentage of mice that showed metastatic growth) and the metastatic pattern differed significantly among the six melanoma models, whereas the number of metastatic lesions in each metastasis-positive mouse did not. The N-15, R-18, and T-22 models showed higher incidence of lymph node metastasis than the C-10, D-12, and E-13 models (p = 0.016; Fig. 1c), whereas the incidence of pulmonary metastasis was higher in the C-10, D-12, and E-13 models than in the N-15, R-18, and T-22 models (p = 0.0012; Fig. 1d). Lymph node metastases develop primarily from tumor cells disseminated through lymphatics associated with the primary tumor, whereas the development of pulmonary metastases can be a result of tumor cell dissemination through primary tumor-associated blood vessels as well as tumor cell dissemination from lymph node metastases [6, 14, 16]. The incidence of lymph node metastasis was higher than the incidence of pulmonary metastases in the N-15, R-18, and T-22 models (p = 0.00075), and consequently, N-15, R-18, and T-22 tumors disseminated primarily through the lymphogenous route. The incidence of pulmonary metastasis was higher than the incidence of lymph node metastasis in the C-10, D-12, and E-13 models (p = 0.022), and consequently, indicative that the hematogenous route was the primary metastatic route in these tumors.

IFP was measured in all primary tumors (i.e., 50 tumors per melanoma model), whereas 20 tumors of each model were selected randomly for immunostaining and assessment of HF_Pim, BVD, and LVD. The N-15, R-18, and T-22 models showed higher IFP than the C-10, D-12, and E-13 models (p = 0.0011; Fig. 2a), suggesting an association between lymphogenous metastasis and highly elevated tumor IFP. HF_Pim did not differ between the C-10, D-12, and E-13 models on the one hand and the N-15, R-18, and T-22 models on the other, as illustrated qualitatively by using a C-10 and a T-22 tumor as representative examples (Additional file 2: Figure S1a) and shown quantitatively by presenting HF_Pim data for all six models (p > 0.05; Fig. 2b). This observation suggests that the primary metastatic route was not determined by the extent of tumor hypoxia.

The C-10, D-12, and E-13 models did not differ from the N-15, R-18, and T-22 models in peripheral tumor BVD or peritumoral LVD. This is illustrated qualitatively in Additional file 2: Figure S1b, c, which show representative examples of histological preparations immunostained for blood vessels and lymphatics, and quantitatively in Fig. 3a (p > 0.05) and Fig. 3b (p > 0.05). However, C-10, D-12, and E-13 tumors grew faster (p < 0.0001; Fig. 3c) and, hence, had shorter T _d (p = 0.0022; Fig. 3d) than N-15, R-18, and T-22 tumors. The rate of induction of new blood vessels (i.e., the angiogenic activity) is different in experimental tumors that show the same microvascular density but have different growth rates [38]. Because C-10, D-12, and E-13 tumors grew faster than N-15, R-18, and T-22 tumors and these two tumor groups had similar peripheral BVD and peritumoral LVD, the C-10, D-12, and E-13 models induced higher angiogenic activity than the N-15, R-18, and T-22 models. From these data and the data reported in Fig. 1, it follows that hematogenous metastatic dissemination and the development of pulmonary metastases were associated with high angiogenic activity in the primary tumor.

In general, more angiogenesis-related genes showed high expression in the C-10, D12, and E-13 models than in the N-15, R-18, and T-22 models (Fig. 4a). Three genes showed more than twofold higher expression in C-10, D-12, and E-13 tumors than in N-15, R-18, and T-22 tumors: ANGPT2 (angiopoietin-2), F3 (coagulation factor-III), and TIE1 (tyrosine kinase with immunoglobulin-like and epidermal growth factor-like domains-1), whereas NRP2 (neuropilin-2) was the only gene that showed more than twofold higher expression in N-15, R-18, and T-22 tumors than in C-10, D-12, and E-13 tumors (Fig. 4b).

The protein levels of these genes were investigated by subjecting primary tumors and their metastases to immunohistochemistry. The immunohistochemistry data were fully in agreement with the qPCR data. The C-10, D-12, and E-13 models showed higher protein levels of the ANGPT2, F3, and TIE1 genes than the N-15, R-18, and T-22 models, whereas the NRP2 protein level was higher in the N-15, R-18, and T-22 models than in the C-10, D-12, and E-13 models, as illustrated in Fig. 5 by using a C-10 and a T-22 tumor as representative examples. Five tumors of each model were subjected to quantitative analysis, which revealed that the staining density of C-10, D-12, and E-13 tumors was higher than that of N-15, R-18, and T-22 tumors for angiopoietin-2 (p = 0.0028), coagulation factor III (p = 0.0091), and tie1 (p = 0.0032) and lower than that of N-15, R-18, and T-22 tumors for neuropilin-2 (p = 0.0056), as illustrated in Fig. 5. Thus, hematogenous metastatic spread and the development of pulmonary metastases were associated with high expression of ANGPT2, F3, and TIE1, whereas lymphogenous metastatic spread and the development of lymph node metastases were associated with high expression of NRP2.