Br J Ophthalmol 96:451-458 doi:10.1136/bjophthalmol-2011-300432
  • Laboratory science
  • Original article

Expression of cancer-testis antigens (MAGE-A1, MAGE-A3/6, MAGE-A4, MAGE-C1 and NY-ESO-1) in primary human uveal and conjunctival melanoma

  1. M C Madigan1,2
  1. 1School of Optometry & Vision Science, University of New South Wales, Sydney, New South Wales, Australia
  2. 2Save Sight Institute, University of Sydney, Sydney, New South Wales, Australia
  3. 3Ludwig Institute for Cancer Research, Melbourne Centre for Clinical Sciences, Austin Health, Heidelberg, Victoria, Australia
  1. Correspondence to Dr Michele C Madigan, Optometry & Vision Science, University of New South Wales, RMB North, 3.023, NSW 2052, Australia; m.madigan{at}
  1. Contributors JAE: preparation of figures, results, drafting manuscript ideas, writing and completion of the manuscript; RMC: conceived and formalised the project and writing and preparing final manuscript; NW-C: sectioning and organisation of the specimens and collation of results; JB: immunostaining, grading of specimens and photography of sections, writing the manuscript; CF: immunostaining, grading of specimens and photography of sections; JC: conceived and formalised the project, provided antibodies, grading of specimens and writing and preparing final manuscript; MCM: organised access to the specimens in Sydney, coordination and collation of data, additional photography and preparation of the figures, drafting and revising manuscript for submission.

  • Accepted 27 November 2011
  • Published Online First 20 December 2011


Aim Metastatic disease in ocular melanoma remains untreatable, is associated with late detection and is resistant to conventional systemic therapies. Many tumours including cutaneous melanoma express specific cancer-testis (CT) antigens and vaccines targeting these antigens can induce T-cell-mediated and humoural immune responses. The authors examined primary uveal and conjunctival melanomas for expression of CT antigens to assess their potential as targets for ocular melanoma immunotherapy.

Methods Paraffin-embedded uveal (n=32) and conjunctival (n=15) melanomas were assessed by immunohistochemistry for melanocyte differentiation antigens (gp100, Melan-A/MART-1 and tyrosinase), and CT antigens (MAGE-A1, MAGE-A3/6, MAGE-A4, MAGE-C1 and NY-ESO-1).

Results Melanoma differentiation antigens, gp100, Melan-A/MART1 and tyrosinase, were expressed in >75% of tumour cells in all uveal and conjunctival melanomas tested. Expression of all five CT antigens tested was low in uveal melanomas, and when present, stained <25% of the tumour cells. MAGE-A1, MAGE-A4 and NY-ESO-1 were expressed in <10% of tumour cells in conjunctival melanomas, while MAGE-C1 and MAGE-A3/6 were expressed in ∼20% and ∼35% of tumour cells in this malignancy, respectively, with variable expression levels.

Conclusions Uveal and conjunctival melanomas consistently expressed high levels of the differentiation antigens (gp100, Melan-A/MART1 and tyrosinase). However, compared with other tumours, including cutaneous melanoma, only low levels of CT antigens were found in ocular melanomas. These observations suggest that immunotherapy directly targeting the CT antigens studied may not be effective for ocular melanoma.


Uveal melanoma, which involves the iris, ciliary body and/or choroid, occurs in about 8 cases per million per year,1 and is the most frequently occurring primary eye cancer in adults.2 Conjunctival melanomas are less common but display a high likelihood of local recurrence following treatment, seen in more than one in three cases.3 4

The morbidity and prognosis of ocular melanoma is generally poor. The mortality rates following diagnosis of uveal melanoma are ∼30% after 5 years and ∼45% after 10 years, with corresponding rates in conjunctival melanomas of ∼15% and ∼30%, respectively.2 About 50% of uveal melanoma patients die from the disease mainly due to metastatic spread,5 which remains the main cause of early death in all types of ocular melanoma. The outcomes from clinical studies using intravenous chemotherapy, immunotherapy or novel therapies for primary and metastatic ocular melanoma are generally disappointing.6 7 More effective therapies are currently needed for treating locally recurrent and metastatic ocular melanoma.

Melanoma-induced immune responses are mediated by CD8 cytotoxic T lymphocyte (CTL) recognition of melanoma cells in association with human leucocyte antigen (HLA) presentation of tumour antigens expressed by melanoma cells, and the potential for immunotherapy as a treatment option for cutaneous and ocular melanoma continues to be explored.8 Recent research suggests that uveal melanomas would be ideal targets for vaccine therapies. In vitro studies using cell-based uveal melanoma vaccines showed that CD80 and major histocompatibility complex class II (MHC II)-matched allogeneic cells can activate CD4 lymphocytes via interferon λ (IFN-λ) and CTLs, and effectively lyse uveal melanoma tumour cells.9 10 Since only MHC II alleles need to be matched to the vaccine, a bank of cell-based vaccines expressing the most common HLA-DR alleles could be established, and HLA-DR genotyping for each patient could be used to determine the most appropriate vaccine for use in treating patients both before and after metastases.9 10

Tumour antigen-derived peptides may also be used as vaccines. Promising target molecules for melanoma include differentiation antigens specific to melanocyte lineage cells and cancer-testis (CT) antigens. CT antigens are restricted in normal tissue, only being expressed on testis germ cells and on cells of the placenta during pregnancy.11 They are also commonly expressed in a variety of different tumour types, including cutaneous melanoma.12 Within this large family of antigens, about 50% are encoded by genes located on the X-chromosome (CT-X antigens) and include the MAGE family genes and NY-ESO-1.11 Spontaneous anti-tumour immune responses can be elicited by CT antigens. Several studies have demonstrated the potential for using these antigens in the study and treatment of cutaneous melanoma. For example, there is a strong association between NY-ESO-1 expression and cancer/metastatic stage, melanoma size, metastatic frequency and prognosis.13 The NY-ESO-1 ISCOMATRIX vaccine, that induces T-cell mediated (CD8) and humoural immune responses, is currently being trialled for cutaneous melanoma.14

Given the differences in growth patterns, gene expression and microenvironment of cutaneous and ocular melanomas,15 16 it is important to consider whether vaccines designed to target CT antigens in cutaneous melanoma may also be effective in the treatment of ocular melanoma. Several studies have looked for MAGE gene expression in primary uveal melanomas and uveal melanoma cell lines, with variable results. Mulcahy et al17 found very low/no expression of MAGE-1, MAGE-2, MAGE-3 and MAGE-4 genes in primary uveal melanomas and concluded that MAGE-peptide based vaccine therapies would be impractical. However, other studies have reported MAGE-1, MAGE-2 and MAGE-3 gene expression in uveal melanoma cells.18–20 The expression patterns of CT antigens in conjunctival melanoma remain to be defined.

In the current study we used immunohistochemistry to study the distribution and expression of differentiation and CT antigens in a series of primary uveal and conjunctival melanomas, to assess their potential as targets for ocular melanoma immunotherapy.

Materials and methods


Formalin-fixed, paraffin-embedded uveal and conjunctival melanomas were obtained from Anatomical Pathology, South Eastern Area Laboratory Service, Prince of Wales Hospital, with approval from the South Eastern Sydney & Illawarra Health Human Research Ethics Committee. All other specimens were either from patients who attended the Ludwig Institute Melanoma Clinic at Austin Health, Melbourne, Australia, or from archival paraffin blocks, with approval from the Austin Health Human Research Ethics Committee. Primary enucleations for uveal melanomas without prior treatment were performed for 30/32 specimens (table 1). Patient characteristics and tumour cell type of uveal and conjunctival melanomas examined in this study are provided in tables 1 and 2, respectively.

Table 1

Uveal melanoma characteristics

Table 2

Conjunctival melanoma characteristics

Paraffin sections of uveal (n=32) and conjunctival (n=15) melanomas were dewaxed and rehydrated using alcohols and water, before staining with H&E according to standard protocols. Stained sections were subsequently dehydrated using alcohols and xylenes, mounted with coverslips in DePeX (BDH Pty Ltd, Murarrie, Queensland, Australia) and examined using a Leica light microscope.

Approximately 65% of uveal melanomas were located posterior to the equator with a height of 1–14 mm. Mixed spindle/epithelioid cell morphology was reported in 65% of specimens with spindle cell morphology in 35% of specimens (table 1). Conjunctival melanoma specimens were reported as primary acquired melanosis with atypia or malignant melanoma (table 2).


Monoclonal antibodies (mAbs) produced by the Biological Production Facility at the Ludwig Institute of Cancer Research were used at various concentrations; mAb E978 (NY-ESO-1) at 2.5 μg/ml, mAb A103 (Melan-A) at 2.67 μg/ml, mAb T311 (tyrosinase) at 5.59 μg/ml, mAb MA454 (MAGE-A1) at a dilution of 1:50 and mAb M3H67 (MAGE-A3/6) at a dilution of 1:20 000.21 The mAb CT7-33 (MAGE-C1) was purchased from Abcam, Cambridge, UK and used at a dilution of 1:4000. Anti-MAGE-A4 mAb 57B was kindly supplied by Dr G Spagnoli (Surgical Research Centre, Basel, Switzerland) and used at a dilution of 1:100 for staining.21 Anti-gp100 (HMB-45) was acquired from DakoCytomation (Carpinteria, California, USA) and used at a dilution of 1:100 for staining.


Formalin-fixed paraffin sections were prepared and dried overnight at 37°C. Paraffin sections were dewaxed in xylene, rehydrated using alcohols and rinsed in 0.1 M phosphate buffered saline (pH 7.4). Water bath antigen retrieval was performed for 30 min using EDTA buffer (pH 8.0) (NeoMarkers, Fremont, California, USA) for antibodies to NY-ESO-1, MAGE-A1 and MAGE-A3/6 or citrate buffer (pH 6.0) (NeoMarkers) for antibodies to gp100, Melan A, tyrosinase, MAGE-A4 and MAGE-C1.

Immunohistochemistry was performed using the Dako Envision+ kit (DakoCytomation). Endogenous peroxidase activity was inhibited by immersing sections in 3% H2O2/phosphate buffered saline for 10 min, after which sections were rinsed in phosphate buffered saline. All incubations were performed at room temperature using the Shandon Sequenza immunostainer.

Two groups of antibodies were investigated to label differentiation antigens (Melan-A, gp100 and tyrosinase) and CT antigens (MAGE-A1, MAGE-A3/6, MAGE-A4, MAGE-C1 and NY-ESO-1). Immunostaining was visualised using 3-amino-9-ethylcarbazole chromogen (Sigma-Aldrich, St Louis, Missouri, USA), with Mayer's haemotoxylin (Amber Scientific, Belmont, Western Australia, Australia) counterstaining. Application of Crystal Mount (Biomeda Corp., California, USA) followed dehydration and mounting in DePeX (BDH).

Known antigen-positive tumours were used as positive controls.12 21 22 The positive control used for all the CT antigens was a cutaneous melanoma tumour as well as normal testes tissue. The melanoma sample used was positive for all the CT antigens investigated in the current study using immunohistochemistry and PCR. The Melan-A, gp100 and tyrosinase positive control consisted of a set of three melanoma tumours. Negative substitution controls included replacing the primary antibody with the antibody diluent or an immunoglobulin matched control antibody.


The degree of immunostaining was scored as the percentage of tumour cells that were positive for each antigen and graded independently by one assessor (JB) in consultation with a clinical investigator (JC) as described previously12 21 22: 0% (–), <5% (+), 5–25% (++), 25–50% (+++), 50–75% (++++), >75% (+++++).


Uveal melanomas

Differentiation antigens, Melan-A, gp100 and tyrosinase, were strongly expressed (>75% cells stained) in the majority of uveal melanomas; only one specimen showed <25% of tumour cells positive for these antigens (figures 1A and 2A–C). By contrast, regardless of the tumour cell type, height or location, uveal melanomas uniformly showed low level (<25% cells stained) or no expression for all CT antigens tested (figures 1A and 3A–E). Two or more CT antigens were expressed independently in only 2/32 (∼6%) specimens. Negative controls showed no obvious immunolabelling (figure 4).

Figure 1

Expression of differentiation and cancer-testis antigens in: (A) uveal and (B) conjunctival melanomas. The histograms show the % of cases immunoreactive for differentiation antigens (gp100, tyrosinase and Melan-A) and cancer-testis antigens (MAGE-A1, MAGE-A4, MAGE-C1, NY-ESO-1 and MAGE-A3/6). Immunostaining is scored as % of tumour cells positive for each antigen and graded as described previously12 21 22: 0% (–), <5% (+), 5–25% (++), 25–50% (+++), 50–75% (++++), >75% (+++++).

Figure 2

Immunoreactivity of the differentiation antigens in representative examples of a uveal melanoma and a malignant conjunctival melanoma. (A, D) gp100, (B, E) tyrosinase, (C, F) Melan-A. Both types of melanoma express high levels of immunostaining compared with the surrounding tissue (bar=300 μm (A–F)).

Figure 3

Immunoreactivity of the cancer-testis antigens in representative examples of a uveal melanoma and a malignant conjunctival melanoma. (A, F) MAGE-A1, (B, G) MAGE-A3/6, (C, H) MAGE-A4, (D, I) MAGE-C1 and (E, J) NY-ESO-1. Low level immunoreactivity is expressed in uveal and conjunctival melanomas, respectively (bar=250 μm (A–J)).

Figure 4

Representative uveal (A, B) and conjunctival (C, D) melanoma specimens showing no immunolabelling in immunoglobulin non-specific controls (A, C), compared with intense gp100 immunostaining (B, D) (bar=100 μm (A, B); bar=50 μm (C, D)).

Conjunctival melanomas

Conjunctival melanomas displayed strong cytoplasmic immunostaining for Melan-A and gp100 differentiation antigens (>75% cells stained in 10/15 and 11/15 specimens, respectively). Tyrosinase showed >75% immunopositive tumour cells in 8/15 specimens (figures 1B and 2D–F).

Immunoreactivity for the CT antigens was low or negative in many specimens, and this varied for the antigens and between specimens (figures 1B and 3F–J). Antibodies to MAGE-A1 and MAGE-A4 showed <5% of tumour cells stained in 7% (1/15) of specimens, respectively. MAGE-C1 immunolabelling was seen in 5–25% of tumour cells stained, for 20% (3/15) specimens. MAGE-A3/6 was observed in 33% (5/15) of specimens and expressed in more tumour cells compared with other CT antigens (figure 3F–J). NY-ESO-1 showed very little (<5% cells stained) or no immunoreactivity for all specimens (figure 1B). The level of CT antigen expression did not appear to vary with the type of tumour (primary acquired melanosis with atypia or malignant melanoma) for the conjunctival melanoma specimens examined. Overall, we found that only 20% (3/15) conjunctival melanomas expressed two or more CT antigens independently. Immunolabelling was not seen in negative controls (figure 4).


We investigated the distribution and expression of differentiation and CT antigens in primary uveal and conjunctival melanomas. Differentiation antigens (tyrosinase, gp100 and Melan-A) are specific to normal and malignant melanocyte lineage cells and their expression is routinely assessed in the diagnosis of malignant melanoma.12 We found that both uveal and conjunctival melanomas consistently expressed high levels of differentiation antigens, similar to earlier observations in uveal melanoma and metastases, and in other melanocytic lesions and types of melanoma (eg, cutaneous melanoma).12 23 24 Despite the strong expression of differentiation antigens commonly seen in melanomas, their use as targets for immunotherapy has been limited in the treatment of metastatic melanoma of the viscera, lymph nodes and skin.25

Peptides for some differentiation antigens have been considered for use in vaccines for uveal melanoma, although in the only study to date, no detectable CTL response was found to the peptide gp100457–466.26 However, vaccines that use combined tumour-antigen peptides may be more effective in eliciting T-cell responses, since tumours often display heterogeneous antigen expression (both within and between tumours). Valmori et al26 used multiple tumour antigen-derived peptides (differentiation antigens including gp100457–466 and a CT antigen) to successfully stimulate CD8 T cell responses, suggesting that multi-targeted approaches may be appropriate for ocular melanoma. This approach could be useful in targeting primary and metastatic ocular melanoma cells for example, limiting the potential for positive-selection of antigen-negative tumour cells that may ‘escape’ CTL killing.

When we examined for CT antigens, low or no MAGE-A1, MAGE-A3/6, MAGE-A4, MAGE-C1 and NY-ESO-1 immunoreactivity was found in uveal melanomas, and this appeared homogeneous within tumours and across specimens. We found expression of two or more CT antigens in only ∼6% of specimens. Mulcahy et al17 reported low or no gene expression for MAGE-1, MAGE-2, MAGE-3 and MAGE-4 in uveal melanoma; however, they could not determine if all tumour cells had low level MAGE gene expression, or whether small populations of tumour cells with higher levels of mRNA may also have been present. In another study, fresh and cultured primary and metastatic melanoma cells from one ocular melanoma patient were examined for MAGE-1, MAGE-2 and MAGE-3 gene expression at different stages of tumour progression.18 MAGE expression was found in all tumour-derived clones and from metastatic cells (although this expression was variable).18 They concluded that there was a homogeneous distribution of MAGE-positive cells within the primary tumour and that MAGE genes were expressed by a high percentage of tumour cells.18 This contrasts with our findings and Mulcahy et al,17 where little or no MAGE expression was detected in primary uveal melanoma. As discussed by Chen et al18 only one primary ocular melanoma removed 15 years after first detection, following plaque radiotherapy was examined, and this does limit generalisation of the findings. Plaque radiotherapy was used for only 2/32 primary uveal melanomas in our study, although this is not clear for tumours examined by Mulcahy et al17 MAGE-1, MAGE-2 and MAGE-3 gene expression has also been reported in several ocular melanoma cell lines,19 and in cells grown from primary uveal melanomas in ∼40–50% of patients examined (n=17).20 Normal choroidal melanocytes and retinal pigment epithelial cells did not express MAGE genes.19 20 In the current study, CT antigen immunoreactivity was not observed in retina, areas of choroid without tumour or surrounding tissue. The variable CT expression reported in uveal melanoma to date most likely reflects differences in the stage of the specimens examined, including plaque therapy as discussed above and the comparison of gene versus protein expression (current study).

CT antigen expression has not been previously reported in conjunctival melanoma. We found low but variable expression for the CT antigens studied. In particular, MAGE-C1 and MAGE-A3/6 antigens were more often expressed in conjunctival compared with uveal melanomas. As mentioned, the growth patterns, gene expression and microenvironment of cutaneous and ocular melanomas differ.15 16 Several studies further suggest that conjunctival melanoma is more similar to cutaneous than uveal melanoma, clinically, histologically, and in terms of gene expression and immunophenotype.16 27–30 Cutaneous melanomas express a range of CT antigens with a higher frequency of expression than conjunctival melanoma. However, only a low proportion of cutaneous melanomas showed simultaneous expression of MAGE-A1, MAGE-A4 and NY-ESO-1.12 The frequency of expression reported by Barrow et al for each of these antigens was 20%, 9% and 45%, respectively, and the expression of MAGE-A1 and MAGE-A4 antigens increased with tumour progression and metastases, while NY-ESO-1 expression did not vary.12 The presence of infiltrating leucocytes (including lymphocytes and macrophages) has been reported in both uveal and conjunctival melanomas,31–33 although the role of these in conjunctival melanomas in particular, requires further investigation. The growth of both uveal and conjunctival melanomas despite the presence of obvious immune cell infiltration is challenging, and whether this in any way relates to the low level and heterogeneous CT antigen expression in these tumours remains to be clarified.

The use of immunotherapy/vaccines for conjunctival melanoma remains to be explored and is of interest since these can be applied topically for these patients. For example, despite its low expression in conjunctival (and uveal) melanoma, NY-ESO-1 may still be a useful target for ocular melanoma immunotherapy. NY-ESO-1 is recognised to be highly immunogenic and can frequently induce spontaneous immune responses.34 Recent clinical trials of NY-ES0-1 ISCOMATRIX vaccine indicate effectiveness in cutaneous melanoma patients with minimal residual disease, although not in advanced disease.14 The potential of this vaccine for ocular melanoma patients, including its immunogenicity, remains to be established.

The CT antigen family consists of many other antigens, apart from those considered in the present study and these may be expressed in conjunctival and uveal melanoma. We have observed strong expression of PRAME (preferentially expressed antigen of melanoma, a CT antigen family member) in uveal melanoma cell lines (mRNA and protein) (M Madigan, personal communication, 2010). Conjunctival melanomas and lymph node metastases were also recently reported to express high levels of PRAME with implications for retinoic acid-based therapies.35 The potential for epigenetic modulation of CT antigen expression has also been reported, with de novo induction of CT antigens following treatment of tumour cells with hypomethylating agents such as 5-aza-2′-deoxycytidine (5AC).36 Interestingly, this effect in multiple myeloma was further enhanced when 5AC was combined with a histone deacetylase inhibitor (MGCD0103).32 We recently found that histone deacetylase inhibitors can induce in vitro killing of conjunctival melanoma cells.37 This combined approach may thus provide heterogeneous targets for therapy in ocular melanoma (and other tumours), the outcome being to limit the potential for antigen-negative tumour cells to escape therapy.

Uveal and conjunctival melanomas consistently expressed high levels of the differentiation antigens (gp100, Melan-A/MART1 and tyrosinase). The low levels of expression of CT antigens observed in uveal and conjunctival melanomas, compared with cutaneous melanoma, suggest that immunotherapy directly targeting these antigens may not be effective for ocular melanoma. However, recent reports in other tumours showing the potential for induction of CT antigens using epigenetic modifying agents such as histone deacetylase inhibitors may be worth further consideration.


  • Funding The authors acknowledge support from The Claffy Foundation, Sydney Hospital/Sydney Eye Hospital and the Sydney Foundation for Medical Research (RMC, NW-C, MCM); the National Health & Medical Research Council (NHMRC), Australia, Melanoma Research Alliance, the Cancer Vaccine Collaborative, New York, the Victorian Cancer Agency and the Ludwig Institute for Cancer Research as well as Operational Infrastructure Support Program funding of the Victorian Government and NHMRC Independent Research Institutes Infrastructure Support Scheme.

  • Competing interests None.

  • Ethics approval Ethics approval was provided by the South Eastern Sydney and Illawarra Health Human Research Ethics Committee and Austin Health Human Research Ethics Committee.

  • Provenance and peer review Not commissioned; externally peer reviewed.


Register for free content

Free sample
This recent issue is free to all users to allow everyone the opportunity to see the full scope and typical content of BJO.
View free sample issue >>

Don't forget to sign up for content alerts so you keep up to date with all the articles as they are published.