CD147 and matrix-metalloproteinase-2 expression in metastatic and non-metastatic uveal melanomas

Paraffin sections from 49 of primary uveal melanoma, which were pathologically well characterized (See Table 1), were analyzed retrospectively with respect to the immunohistological expression of CD147 and MMP2. Staining for von Willebrand Factor (vWF), a marker for mature vasculature, was performed to allow the count of vessels. For visualisation of tumor induced angiogenesis, immunohistochemical staining for Endoglin (CD105) was carried out. During neoplasm-related angiogenesis the transforming growth factor-β1 and –β3 (TGF- β1, TGF-β3) binding receptor Endoglin is strongly up-regulated on the endothelium [16, 17]. All the primary and secondary antibodies used are listed in Table 2, with the information on their type, source, and dilution factor, as well as the corresponding pretreatment and visualisation methods.

The uveal melanomas were obtained by enucleation from 1982 to 1999 at the University Eye Hospital of Tübingen (specimens’ use authorized by JMR, head of ophthalmopathology at the University Eye Hospital of Tübingen). In 28 cases metastatic disease was found in the restaging examination. The remaining 21 patients had a mean follow-up of 10 (6–18) years. Irradiation was conducted previously in 16 tumors. The non irradiated tumors received enucleation as primary therapy.

For CD147 and MMP-2 double-immunostainings, serial paraffin sections were deparaffinized in three changes of xylol and rehydrated in a graded series of alcohol (100–50 %). Antigen retrieval was performed by cooking the slides in citrate puffer for 25 min, followed by cooling down at room temperature for 20 min and washing three times with phosphate-buffered saline (PBS). Afterwards, sections were incubated with blocking buffer (2 % bovine serum albumin (w/v) and 20 % rat serum (v/v) in PBS) for one hour at room temperature followed by the coincubation with primary CD147 and MMP-2 antibodies (diluted as stated in Table 2 in 3%BSA-0.1 % Triton X-100 in PBS) overnight at 4 °C. Negative controls were incubated in the blocking buffer alone. Sections of mamma carcinoma were incubated with the primary antibody cocktail as positive control. After rinsing with PBS, sections were incubated with the corresponding fluorescence-conjugated secondary antibodies (Table 2) diluted in 3 % BSA-PBS for 1 h at room temperature under protection from light. After rinsing with PBS three times for 5 min, nuclei were counterstained with DAPI (0.5 μg/ml in PBS, 150 μl per section) for 8 min in the dark. After rinsing with PBS for 5 min, the slides were mounted with Mowiol-coated coverslips.

Triple fluorescence-immunostainings were performed by following the procedure above, by incubating the sections with a cocktail of primary antibodies against CD147, MMP-2 and CD34, CD68 or GFAP, followed by the cocktail of secondary antibodies, which were diluted in 3 %BSA-0.1 % Triton X-100 in PBS as described in Table 2.

The Endoglin and vWF immunostainings were performed as described previously [18]. In brief, the sections were deparaffinized and rehydrated as above. In order to better visualize the staining, melanin was bleached by incubating the sections in 3.0 % (v/v) hydrogen peroxide, 0.5 % potassium hydrogen phosphate, and 1.0 % (w/v) disodium hydrogen phosphate for 12 h at room temperature as described previously [19]. The majority (n = 26) of slides received a bleaching due to the high level of pigmentation with ranging bleaching times (mean: 65.2 min) depending on the pigmentation grade. No background staining arising from endogenous peroxidases was detected in the negative controls of the remaining unbleached slides (data not shown). Antigen retrieval by proteolytic digestion was performed on the sections that would be processed for Endoglin immunostaining by incubation in 0.5 % Pronase (Sigma-Aldrich, Munich, Germany) in PBS for 30 min. Sections were incubated with horse serum (15 μl in 1 ml PBS) for 30 min at room temperature, followed by the primary antibodies against human Endoglin or vWF (Table 2) for 12 h at 4 °C. Appropriate normal serum instead of the primary antibodies was added for negative controls. Paraffin sections of eyes that have shown a high positivity in a previous staining served as positive controls. After three washes for 5 min in PBS, sections were incubated with horseradish peroxidase-conjugated secondary antibodies for 1 h at room temperature. Before counterstaining with Mayer’s hematoxylin and coverslipping the slides were washed in PBS and developed with 3-Diaminobenzidine (Fluka, Buchs, Switzerland).

To visualize VM CD31/PAS-immunostaining was performed as previously described [18]. In brief, sections were deparaffinized and bleached as above. Antigen retrieval was performed by cooking the sections in Tris-buffered saline (TBS) for 30 min. After blocking with horse serum (15 μl in 1 ml TBS) for 60 min, sections were incubated with the primary antibodies against CD31, followed by the horseradish peroxidase-conjugated secondary antibodies (Table 2) for 1 h at room temperature each. After three rinses for 5 min in PBS, sections were developed with 3-Diaminobenzidine (Fluka, Buchs, Switzerland), counterstained with periodic acid-Schiff, and coverslipped. Negative controls of the tumor sections and gastric mucosa were treated with normal sera instead of antibodies and all exhibited a negative staining (data not shown).

Three days after immunohistochemistry, images were acquired using a fluorescence microscope (Leica DMI 6000B, Solms, Germany connected to a digital camera (Leica DFC 290, Solms, Germany) and the corresponding filters (A4: Ex: 360/40, Em: 470/40 nm; L5: Ex: 460/40, Em: 527/30 nm; Y3: Ex: 545/30, Em: 610/75 nm; Y5: Ex: 620/60, Em: 700/75 nm). All tumor samples were initially observed under 100X magnification to evaluate the homogeneity of CD147 and MMP-2 stainings. For the quantification of CD147 and MMP-2 expression, images from n = 10 random fields of 0.04 mm² (400X magnification) were acquired, ensuring that the images are representative of the areas exhibiting different intensities in the case of tumors having inhomogeneous antigen expression. Quantifications were performed using the ImageJ Software (National Institute of Mental Health, Bethesda, Maryland, USA) by two independent observers (JL, VV), who were totally masked to the follow-up of the patients. Evaluation of the antigen expression in tumor samples was conducted in 10 random fields of 0.04 mm² (400X magnification) with the ImageJ Software (National Institute of Mental Health, Bethesda, Maryland, USA) by two independent observers (JL, VV) who were totally masked to the follow-up of the patients. Additionally, the presence of clustered CD147 positive cells was evaluated: (no: 0, moderate number: 1, or high number: 2).

The objective evaluation of the fluorescence intensity was performed using the ImageJ software. Ten images were summarized as image stack and a grid was placed over the pictures with an edge length of 3.5 in. resulting in 680 rectangles. A rectangle was counted as positive if 50 % of the area was positive. The intensity of the pixels in stack was calculated according to the formula: V = (R + G + B)/3. At the end of the calculation, the percentage mean intensity was calculated from the mean of the intensity of stained areas for each marker. The intensity was then classified: (no expression: no positive cells (0), low intensity (1): 1–24 %, moderate intensity (2): 25–50 %, high intensity (3): 51–75 %, very high intensity (4): 76–100 %). The following formula was used for calculating the objective expression rate: 100/16 x cell count x intensity score.

To evaluate the status of mature vasculature and angiogenesis the number of all vWF and Endoglin positive cells was evaluated in n = 3 of the 10 images (0.04 mm²) exhibiting the highest level of expression according to the following grading system: no expression: no positive cells, low/moderate expression: 1–50 %, high expression: 51–100 %) as reported previously [18].

The analysis of VM two independent observers (JL, VV) analyzed the CD31 and PAS stained uveal melanoma specimens under 200x and 400x magnification (Leica DMI 6000B, Solms, Germany) and assigned the tumors to be VM positive or VM negative. A positive rating was given in case of vascular channel formation as reported previously [1]. An example is shown in Fig. 1. Sections with different rates by first and second observer were re-evaluated to come to a final decision.

In this tumor series CD147 was expressed by 47 (96.0 %) primary uveal melanomas especially at the cell surface. The percentage fraction of positive cells according to the objective cell-count varied between 1.07 % and 100 % (mean: 37.12 %). Particularly in large tumors (LTD > 12 mm), most cells at the tumor margin exhibited a higher degree of expression compared to the cells in the inner regions (Fig. 2). In the primary tumor a weak colocalisation of CD147 and MMP-2 was observed in some CD68 positive cells, which were probably tumor-infiltrating macrophages (Fig. 3c). There was no significant difference according to the overall objective or subjective expression of CD147 in the uveal melanomas which did or did not develop metastatic disease in the further clinical course. The complete data and p-values of the objective evaluation are summarized in Table 3. The correlation with known prognostic factors (age, gender, radiation, cell type, invasion of the ciliary body, largest tumor diameter (LTD), height, TNM-classification, invasion of Bruch’s membrane, intrascleral invasion, extrascleral extension, invasion of the optic nerve) revealed no significant association with the overall expression analysis (Pearson correlation, p < 0.05).

A closer look at the distribution of CD147 expression at higher magnification (400X) revealed that the majority of positive cells had a homogenous distribution and intensity within randomly selected fields (0.04 mm², Fig. 2b and c). However, in 29 uveal melanomas (59.2 %) nested neoplastic cells were highly positive for CD147. The vast majority of these tumors with clustered CD147 positive cells (n = 24) developed metastasis in the further clinical course but the level of significance was not reached (p = 0.145) (Fig. 4). Nevertheless, in the non-metastatic sub-group, a significant correlation of clustered CD147 positive cells with LTD (p = 0.039), height (p = 0.047) and TNM-stage (p = 0.013) was evident. In the metastatic sub-group the presence of nested CD147 positive cells was correlated with ciliary body involvement (p = 0.042).

In 35 % of the examined eyes a low or moderate CD147 expression was observed in the retina especially in the nuclear layers (Fig. 5). Coimmunostainings for CD147 and GFAP or CD68 demonstrated that some astrocytes in the nerve fiber layer, Müller cells, and CD68-positive microglia were also expressing CD147 in the retina, whereas the retinal endothelial cells (detected by antibodies against CD34) exhibited a weaker CD147 expression (Fig. 5). The epithelium of the ciliary body expressed CD147 in 15 % of the cases. The corneal endothelium and epithelium revealed a very low CD147 expression in 21 % of the documented cases. The remaining cases expressed no CD147 in the above mentioned ocular structures (Fig. 5).

MMP-2 was expressed by 47 of the 49 examined uveal melanomas in this study (96.0 %). An average of 89.3 % tumor cells showed MMP-2 expression (range: 3.26–100 %). In some large tumors (LTD > 12), cells at the tumor margin exhibited a higher degree of immunoreactivity compared to the cells in the inner regions (Fig. 1). A difference in the MMP-2 expression of tumors with or without metastatic disease could not be shown (Table 4).

The correlation analysis of MMP-2 expression with known prognostic factors (age, gender, radiation, cell type, invasion of the ciliary body, largest tumor diameter, tumor height, TNM-classification, invasion of Bruch’s membrane, intrascleral invasion, extrascleral extension, invasion of the optic nerve) revealed no significant association (Pearson-correlation, p < 0.05). Although an increased MMP-2 expression was observed in tumors which formed VM the level of significance was not reached (p = 0.073).

In the detailed analysis of the distribution of MMP-2 positive cells, an up-regulation was observed in the area of the nested uveal melanoma cells which were highly positive for CD147 (Fig. 4). Beside the overall expression analysis revealed a positive correlation of those two markers (p < 0.001).

A low MMP-2 positivity was evident in some cases of the documented corneal endothelium (40 %) and epithelium (29 %). The remaining cases expressed no MMP-2. In 69 % of the examined eyes a low or moderate MMP-2 expression was observed in the retinal pigment epithelium and the detached retina associated with the uveal melanoma (Fig. 5). The retinal staining was evident particularly in the nuclear layers but also in some astrocytes, Müller cells, and retinal endothelial cells (Fig. 5). The epithelium of the ciliary body expressed MMP-2 in 40 % of the cases (Fig. 5).