Article Text
PDF
Abstract
Background To analyse the expression of melanoma chondroitin sulfate proteoglycan (MCSP) and the preferentially expressed antigen of melanoma (PRAME) in conjunctival melanoma (CoM), lymph node (LN) metastases of cutaneous melanoma (CM) and conjunctival nevi (CoN) by immunohistology.
Methods Immunohistology was performed in 70 samples of CoM, 25 of LN metastases of CM and 12 of CoN, and assessed by an immunoreactive score (0–12 points). Statistical analysis was performed to disclose relevant differences in the expression pattern. The diagnostic value of the markers was tested by receiver operating characteristics (ROC) analysis.
Results MCSP and PRAME were expressed at significantly higher levels in CoM and LN metastases of CM than in CoN (p<0.0001). Within CoM, an MCSP expression <9.0 points meant higher risk for recurrences (Cox HR=3.1) and a shorter recurrence-free survival (p=0.002) than an MCSP expression >9.0 points. ROC analysis showed an area under the curve of 91.3% for MCSP (p=0.0002) and 93.8% for PRAME (p<0.0001).
Conclusions MCSP and PRAME are differentially expressed in conjunctival melanomas and nevi. MCSP might have an impact on the risk for recurrence in being inversely correlated to the event. Both markers have high potential to discriminate CoM from CoN. The results indicate that immunohistological characteristics gain relevance in the assessment of CoM.
- MCSP
- PRAME
- conjunctiva
- melanoma
- immunohistochemistry
- conjunctiva
- pathology
- neoplasia
- diagnostic tests/investigation
Statistics from Altmetric.com
- MCSP
- PRAME
- conjunctiva
- melanoma
- immunohistochemistry
- conjunctiva
- pathology
- neoplasia
- diagnostic tests/investigation
Introduction
In conjunctival melanoma (CoM) most predictable factors are based on clinical or histopathological characteristics,1–6 but little is known about the tumour's biology and protein expression profile.7 Concerning growth pattern and metastatic pathways, CoM resembles its cutaneous counterpart rather than other ocular melanomas (ie, uveal).8–11 Data regarding the origin of CoMs and their rates of local recurrence and metastatic disease show that the tumour's biology needs to be further illuminated in order to understand the underlying mechanisms. New therapeutic strategies are required to reduce recurrence rates and improve survival.12–14
Melanoma chondroitin sulfate proteoglycan (MCSP) became a target in melanoma research when immunotherapy became relevant in cutaneous melanoma (CM). It was identified in the late 1970s as a transmembrane proteoglycan with two components.15 As a cell surface proteoglycan, MCSP serves as an adhesion receptor influencing cell adhesion and migration.16 Additionally, it influences cell invasion17 and proliferation.18 19 It has been found that the MCSP expression correlates to the metastatic potential of a soft tissue sarcoma and melanoma.18 20–22 A limited expression of MCSP is shown in normal tissues, but it is expressed in a high percentage of melanoma lesions23 (reviewed by Campoli et al24).
The tumour antigen ‘preferentially expressed antigen of melanoma’ (PRAME) is a non-mutated cancer–testis antigen that is expressed in fetal tissue, testes and in cancer, mainly in CM. PRAME is expressed in 88% of primary CMs and 95% of melanoma metastases.25 26 As it is expressed neither in benign cutaneous nevi nor normal skin, it is likely that PRAME expression occurs under melanocyte transformation.27 It supports cell survival and suppresses differentiation by interfering with retinoic acid (RA) signalling.28 No data exist regarding PRAME expression in CoM.
In this study we analysed the expression of MCSP and PRAME in 70 cases of CoM by immunohistochemistry and compared the results with lymph node (LN) metastases of CM and conjunctival nevi (CoN). The differential diagnosis between CoM and nevi often gives rise to controversy between pathologists or pathologists and clinicians, and has an important impact on the therapeutic approach. The established antigens in melanoma diagnostics cannot clarify the diagnosis in all cases. Therefore, we examined the potential of MCSP and PRAME to differentiate between these two conjunctival entities. Both antigens have an impact on melanoma progression and may represent therapeutic targets. While MCSP is of special interest as a melanoma-specific antigen, PRAME is connected to a very important pathway which is a target for modern treatments (the RA signalling pathway). We chose these two antigens to cover these issues for CoM. No information about the two antigens exists in CoM research so far.
Materials and methods
Demographic data
In this study we analysed the clinical data of 70 patients with histologically confirmed malignant CoM and performed immunohistology of their tumour samples. The patients were treated between 1974 and 2006 in the Department of Ophthalmology of the University Hospital Essen or Tübingen. The samples were acquired from these departments and the Department of Pathology and Neuropathology of the University Hospital Essen. The mean age was 63.7 ± 16.0 years at the time of diagnosis. The mean follow-up was 69.6 months (minimum 0.33, maximum 317.16, median 38.83). Forty cases (57.1%) were presented at the time of first diagnosis and then treated primarily in one of the participating centres. Twenty-two cases (31.4%) had already had a local recurrence at the time of presentation; in 8 cases (11.5%) we had no information about the earlier course of the disease. The tumours were staged following the 6th edition of the TNM staging system for conjunctival melanoma (American Joint Committee on Cancer 2002). For demographic and TNM data, see table 1. LN metastases of CM (n=25) and benign CoN (n=12) served as control groups. Clinical data of these control groups were not available. Primary CM samples served as positive controls for MCSP and PRAME. Therefore, we did not use primary CM as a separate group in this study. We were interested in the expression of the antigens in metastases. Metastases of CoM are very rare and not usually excised. Therefore, we decided to include LN metastases of CMs as a control group.
Demographic and pathological data of the conjunctival melanoma (CoM) group with the number of tumours of the whole group
All experiments and procedures were conducted in accordance with the Declaration of Helsinki. An informed consent was obtained from the patients for the analysis of collected samples. The local Ethics Committee of the Medical Faculty of the University Hospital Essen, University of Duisburg-Essen, Germany approved the study.
Immunohistochemistry
Diagnosis of melanoma was confirmed by H&E slides and S-100, HMB-45 and Melan-A immunostaining. Tissue microarrays (TMAs) of CoMs (n=70), LN metastases of CMs (n=25) and CoN (n=12) were assembled with a core size of 600 μm and cut into 5 μm thick slices. Suitable donor tissue for the construction of TMAs was identified and marked on H&E sections. TMA blocks were constructed with a manual tissue arrayer (Beecher Instruments, Sun Prairie, Wisconsin, USA) with three punch specimens (diameter 0.6 mm) per tumour from corresponding blocks of formalin-fixed and paraffin-embedded tissue. The area of trephination was defined under microscopic control. In small samples, 5 μm slices were cut and placed directly on object holders. Afterwards, the samples were allowed to dry at 52°C overnight.
For antigen demasking and deparaffinisation, sections were incubated in xylol and rehydrated through graded decreasing concentrations of alcohol (ethanol 100%, 90%, 70%). Afterwards, sections were rinsed with aqueous solution and washing buffer (DCS, Hamburg, Germany: WL583CO500). As pretreatment, citrate buffer (pH 6.0) (Zytomed Systems, Berlin, Germany; ZUC028-500) was applied to a water bath at 98°C for 20 min. Cooling had been allowed for 10 min before the samples were transferred to the washing buffer. Afterwards, peroxidase block was performed for 5 min (Dako, Glostrup, Denmark: S2001), and then the specific antibodies were applied for 30 min (PRAME, rabbit, polyclonal, dilution 1:200; MCSP, mouse immunoglobulin G1 (IgG1), LHM2, dilution 1:400; both Biozol (Eching, Germany), both dilution media Zytomed Systems ZUC025-500). A detection system of Zytomed Systems (POLHRP-100) was used. DAB (3,3′-diaminobenzidine, Zytomed Systems DAB 530) provides brown staining to visualise the complexes. The sections were counterstained with haematoxylin 1:8. Endogenous melanin pigment was identified in H&E-stained samples and was valuable for the estimation of the immunostaining. A primary CM (PRAME+, MCSP+) served as positive control for the TMA investigation. For negative controls, specific primary antibodies were replaced by normal serum of the same species. The stained samples were analysed and graded by two independent examiners (HW and FG) using the immunoreactive score (IRS) (also known as the Remmele score).29 Interobserver agreement was assessed by the Kappa test.30 In cases of interobserver differences, a third examiner re-estimated the sections and the results were discussed between all three examiners to find a congruent result. The Remmele score is used to estimate the grade of immunoreaction. The product of the points for staining intensity (no staining, 0; weak staining, 1; moderate, 2; strong, 3) and percentage of stained cells (0%, 0; <10%, 1; 11–50%, 2; 51–80%, 3; 81–100%, 4) results in a score from 0 to 12.31 32
Statistical analysis
Statistical analysis was performed using Microsoft Excel 2000, SPSS (SPSS for Windows, version 17.0; SPSS Chicago, IL, USA) and StatView for Windows (Version 4.55, Abacus Concepts, Berkeley, California, USA).
A p value of ≤0.05 was considered statistically significant. Results were presented as arithmetic means, SD, medians, ranges and percentage rates.
Survival analyses to test the relationship of antigen expression, clinical features and recurrences or metastatic disease were performed by the Kaplan–Meier analysis. Risks were assessed using univariate Cox regression analysis.
Mann–Whitney U test was used for the comparison of antigen expression and nominal variables (two variables); the Kruskal–Wallis test was performed for nominal variables with ≥3 variables. Bonferroni correction was applied to multiple testing with an α level set to 0.05. For correlation of continuous variables the Spearman correlation was performed, the results being visualised by linear regression.
The unknown cases of the clinical and histopathological parameters were not included in the calculations and were treated as missing values. The cut-off points for MCSP and PRAME of an IRS of 9.0 points were generated using the mean and median of all cases of CoM (MCSP, mean 8.9; median 9.0/PRAME, mean 9.4).
Receiver operating characteristic (ROC) analysis was performed to analyse the diagnostic value of the antigen expression (sensitivity and specificity) in the distinction of CoM and CoN. In general, the area under the curve is given as an indicator of the accuracy of a tool (here antigen). All points above the diagonal line are better than a random guess. The asymptomatic significance (asymptomatic p value) of <0.05 indicates that a tool is better than a guess in predicting an event (null hypothesis: true area=0.5). The cut-off points for each antigen are suggested by the statistical program and they include values from the minimum observed test value minus 1 to the maximum observed test value plus 1.
Results
Demographic data and therapy
Clinical and histopathological characteristics of the CoM group are summarised in table 1. In our cohort, we could not identify clinical or histopathological risk factors for local recurrence or metastatic disease (data not shown). Analysing therapeutic approaches, 12 patients had only been treated by tumour excision and had not received adjuvant treatment. Five patients underwent adjuvant mitomycin C local treatment, 8 cryocoagulation, 12 proton radiotherapy and 18 brachytherapy with ruthenium-106 plaques. Fifteen patients were given other adjuvant treatment. Cox regression analysis revealed an increased risk for local recurrence in patients who received local chemotherapy with mitomycin C eye drops (p=0.04; HR 5.7; 95% CI 1.1 to 30.1). In Cox regression analysis, tumours originating from primary acquired melanosis (PAM) with atypia had a 2.4-fold increased risk for recurrence compared with de novo tumours (HR for PAM with atypia=1.3; p=0.58, 95% CI 0.52 to 3.2; HR for de novo tumours=0.5; p=0.46, 95% CI 0.11 to 2.7).
Expression of MCSP and PRAME in CoM, CoN and LN metastases of CM
The Kappa test revealed an overall interobserver reproducibility of 64.6%, which is a substantial agreement.30
Both antibodies showed a cytoplasmic and nuclear immunoreaction. For MCSP the cytoplasmic staining exceeded nuclear staining. For PRAME nuclear staining was much more intense than cytoplasmic staining. This pattern was observed in CoMs, CoN and LN metastases of CM in equal measure. The highest MCSP expression with a mean IRS of 11.1 points and a very homogenous expression profile (SD ±1.69) was found in LN metastases of CM. CoM expressed MCSP at a high level, as well, but the mean IRS was significantly lower (8.9 ± 2.96 points). CoN showed a minor MCSP expression with a mean IRS of 3.8 points (SD ±2.15; figure 1A–C). The difference in MCSP expression was statistically significant between each of the groups (p=0.0005 CoM to LN metastases; p<0.0001 CoM to CoN and LN metastases to CoN; figure 2A).
Immunohistology of melanoma chondroitin sulfate proteoglycan (MCSP) (A–C) and the preferentially expressed antigen of melanoma (PRAME) (D–F). Magnification ×200. Both antibodies showed a cytoplasmic and nuclear immunoreaction. MCSP, cytoplasmic > nuclear; PRAME, nuclear >> cytoplasmic. This pattern was observed in conjunctival melanomas (CoM), conjunctival nevi (CoN) and lymph node metastases of cutaneous melanoma (LN-met. CM) in equal measure.
(A) Boxplot of the results of melanoma chondroitin sulfate proteoglycan (MCSP) expression in the three diagnosis groups. (B) Boxplot of results of preferentially expressed antigen of melanoma (PRAME) expression in the three diagnosis groups. The horizontal line within the box marks the median. The box itself spans from the 25th percentile (lower end) to the 75th percentile (upper end). The T-lines below and above the box mark the 10th percentile (below) and the 90th percentile (above). CoM, conjuctival melanoma; CoN, conjunctival nevi; LN-met. CM, lymph node metastases of cutaneous melanoma.
PRAME expression was highest in LN metastases of CM (mean IRS 9.84 ± 2.4) and CoM (mean IRS 9.41 ± 3.4). Both groups expressed PRAME at a similar high level (figure 2B). CoN showed significantly lower PRAME expression with a mean IRS score of 2.6 points (IRS score minimum 0; maximum 6; figure 1D–F). The difference in PRAME expression was statistically significant between CoM and CoN (p<0.0001) and between LN metastases of CM and CoN (p<0.0001).
A highly significant correlation was found between the expression of MCSP and PRAME in CoM (R2=0.4, p<0.0001).
Expression of MCSP and PRAME in CoM
Statistically, we examined a possible correlation with gender, tumour origin, location of primary tumour (cTNM), pigmentation, infiltration depths (pTNM), predominant cell type, recurrence rate and time of first recurrence, metastatic disease, and occurrence of a second recurrence. We found no statistically significant correlation between these parameters and MCSP expression (data not shown). Tumours with an MCSP expression with an IRS of <9.0 had a higher risk for local recurrence than tumours with an IRS ≥9.0 (Cox regression analysis, HR=3.1; p=0.003; 95% CI 1.5 to 6.6; Wald2=8.6). Additionally, tumours with an MCSP expression with an IRS <9.0 had a shorter recurrence-free interval than tumours with an IRS of MCSP expression ≥9.0 (Kaplan–Meier analysis; log-rank p=0.002; χ2=9.7) (figure 3). Analysing the subgroups, we found that tumours with a lower MCSP expression had a higher risk of developing local recurrence and a shorter recurrence-free survival independent of their clinical characteristics (eg, for epibulbar/non-epibulbar tumours, p=0.005, χ2=7.8). The same trend was observed in other characteristics such as pigmentation (pigmented/amelanotic; p=0.002; χ2=9.8), predominant cell type epithelioid (p=0.02; χ2=5.1) or occurrence of a second recurrence (p=0.03; χ2=4.9). We also observed a trend that tumours with the origin PAM with atypia (mean IRS 7.9 ± 2.7) had a lower mean MCSP expression than tumours appearing de novo (mean IRS 9.5 ± 2.6). This did not reach statistical significance (p=0.18) but gave us a link to a clinical explanation for our findings.
Kaplan–Meier analysis for recurrence-free survival in patients with conjunctival melanoma (CoM). The group was divided into melanoma chondroitin sulfate proteoglycan (MCSP) expression <9.0 points immunoreactive score (IRS) and ≥9.0 points IRS.
For PRAME, we did not find significant correlations to most parameters (data not shown). In tumours that were confined to the epithelium (pT1, melanoma in situ), PRAME expression was significantly lower than in tumours with subepithelial or deeper invasion (pT2–pT4) (p=0.04). Tumours classified as pT2, pT3 and pT4 had no significant difference in PRAME expression between the groups. We observed a trend that tumours with an IRS score for PRAME <9.0 points had a lower risk for metastatic disease (Cox regression, HR=0.44, p=0.3; 95% CI 0.1 to 2.1) and tended to develop metastases later than tumours with an IRS score ≥9.0 points (log-rank p=0.3, χ2=1.2). This did not reach statistical significance, possibly due to the limited number of cases with metastatic disease in this cohort.
ROC analysis and predictive value of MCSP and PRAME
We performed a ROC analysis in order to assess the power of MCSP and PRAME to discriminate CoN from CoM. For MCSP, the area under the curve was 91.3% (asymptomatic p value=0.0002; 95% CI 0.84 to 0.99, figure 4). For PRAME, the area under the curve was 93.8% (asymptomatic p value<0.0001; 95% CI 0.87 to 1.0; figure 4). These findings make both parameters suitable to differentiate CoN from CoM. Table 2 shows the cut-off points within the IRS score revealing that MCSP and PRAME were diagnostically valuable in the intermediate and high scores (2.5–8.5 points IRS).
Receiver operating characteristic (ROC) analysis of melanoma chondroitin sulfate proteoglycan (MCSP) (continuous line) and preferentially expressed antigen of melanoma (PRAME) (dotted line). The diagonal reference line is for orientation. Area under the curve for MCSP, 91.3% (p=0.0002); for PRAME, 93.8% (p<0.0001).
Results of the receiver operating characteristic (ROC) analysis
Discussion
Our results indicate that MCSP and PRAME are differentially expressed in benign CoN and malignant CoMs. CoMs and LN-metastases of CM express both MCSP and PRAME at a similar level, whereas CoN express both antigens at a significantly lower level. Both markers can discriminate CoN from melanoma with high significance.
In our cohort, MCSP had the surprising effect that CoM tumours with lower MCSP expression had a higher risk for local recurrence and developed their first recurrence earlier than tumours with a high MCSP expression. Factors that had turned out to be prognostically relevant to the outcome of CoM in clinical studies before did not show this trend in our study and might therefore be of lesser predictive relevance compared with immunohistochemical characteristics. However, despite our results, this can only be stated as a hypothesis and has to be confirmed by further studies. For explanation, one has to recall the mechanisms associated with MCSP. MCSP is able to influence intracellular signal cascades that have an impact on cell adhesion, motility and invasion.33 One relevant mechanism in the communication of melanoma cells with the extracellular matrix (ECM) is provided by the α4β1-integrin and its binding to fibronectin and to the vascular endothelial cell adhesion molecule VCAM,34 which is also expressed in CoM.35 It is known that the MCSP rat homologue NG2 collaborates with α4β1-integrin to stimulate focal contacts and stress fibre formation in melanoma cells.21 This may contribute to the metastatic potential of a tumour. On the other hand, it may also be possible that in cases of a low MCSP expression, the reduced communication of a tumour cell with the ECM and the immune cells helps to establish a tumour recurrence in the local environment without being ‘recognised’. This would explain the trend that a lower MCSP expression is linked to a higher risk for local recurrence in our cohort of CoM.
Our results could probably also be explained clinically, as follows: tumours originating from PAM with atypia had a lower mean MCSP expression and a 2.4-fold increased risk for recurrence compared with primary tumours in our cohort. The recurrence rate may be higher because clinical assessment of PAM with atypia carries the risk of missing a recurrent transformation to melanoma due to their diffuse and variable appearance on the ocular surface. It can be assumed that the lower mean expression of MCSP in this group leads to our finding that such tumours tend to have a higher risk for recurrence in the whole cohort.
In CoM, PRAME is expressed at a high level. Together with the very low expression of PRAME in CoN, our results correspond to the reported PRAME expression in CM and cutaneous nevi.25 27 Our results indicate that invasive tumours express higher levels of PRAME than intraepithelial tumours. This suggests that PRAME expression increases under melanoma progression. It is, however, not linear progression but rather seems to divide early (intraepithelial) from advanced (invasive) tumours. PRAME acts by interfering with RA signalling as a dominant inhibitor of the retinoic acid receptor (RAR).28 Binding of RA to its receptor initiates the activation of target genes that induce proliferation arrest, differentiation and apoptosis. This is the basis of the antitumoural activity of RA.36 37 Subsequently, the change to, and increase in, PRAME expression may support tumour progression as the suppression of RAR signalling promotes tumour growth and inhibits differentiation. The impact of PRAME on the prognosis of a tumour entity is variable.38 39 Our samples showed a trend towards earlier occurrence of metastases in cases with high PRAME expression combined with an increased risk for metastases. The statistical value was limited due to the limited number of cases in that subset.
In order to understand the role of PRAME and MCSP in CoM more completely, further genetic studies are required to elucidate the source and mechanism of the increased PRAME and MCSP expression in CoM.
MCSP and PRAME have a good diagnostic value in distinguishing between CoN and CoM. Whether the antigens are able to improve the immunohistological diagnosis in difficult cases has to be evaluated further.
In conclusion, in this study we reported on the expression of MCSP and PRAME in CoM and benign CoN. We showed that MCSP and PRAME were differentially expressed in CoM and CoN, and can help to differentiate the lesions diagnostically.
References
Footnotes
Funding The study was funded by the Medical Faculty of the University Duisburg-Essen.
Competing interests None.
Ethics approval This study was conducted with the approval of the ethics committee of the Medical Faculty of the University Hospital Essen, University of Duisburg-Essen, Germany.
Provenance and peer review Not commissioned; externally peer reviewed.