Heterogeneous Expression of the SSX Cancer/Testis Antigens in Human Melanoma Lesions and Cell Lines¹

The SSX genes were initially identified as fusion partners of the SYT gene in human synovial sarcomas carrying a recurrent t(X;18)(p11.2;q11.2) translocation (reviewed in Ref.1). Typically, these tumors express chimeric transcripts in which 3′ sequences of either SSX1 or SSX2,both localized on the X chromosome, are fused to 5′ sequences of SYT, which is localized on chromosome 18(2, 3, 4, 5). Expression studies in normal tissues showed that SSX transcripts are only found in the testis and, at residual levels, the thyroid (4, 6). Preliminary evidence for the existence of additional SSX family members was provided by the detection of genomic fragments from Xp11 that hybridized to an SSX1 probe (7). The screening of a human testis cDNA library led to the isolation of SSX3,which is localized in the region Xp11.1–11.2 (8).

Recently, when searching for melanoma tumor antigens, Türeci et al. (9) found that the SSX2 gene was expressed in melanomas and, in addition, several other malignancies. Further research disclosed that also SSX1 and two newly discovered SSX genes, SSX4 and SSX5, may be expressed ectopically in a variety of cancers(10, 11). In contrast, no SSX3 expression was observed in any of these malignancies. In addition, it was found that at least the SSX2 protein may act as a tumor-associated antigen (also designated as HOM-MEL-40), eliciting humoral immune responses in a subset of melanoma patients (9, 12). Due to the absence of expression in most normal tissues except testis and the ectopic expression in a variety of malignancies, the SSX proteins are considered as new members of the CT⁵family of antigens (10).

The SSX genes exhibit nucleotide homologies ranging from 88 to 95% and encode proteins of 188 amino acids that exhibit homologies ranging from 77 to 91%. Recent studies have shown that the SSX proteins are localized in the nucleus, exhibiting both diffuse and punctated distribution patterns (13, 14, 15). In addition, it was found that the conserved COOH-terminus of SSX is able to repress the transcription of a reporter gene (16). This domain is also responsible for nuclear localization, chromatin association, and colocalization with Polycomb-group nuclear bodies (17).

In the present study, we detected variable levels of SSX RNA expression in several human melanoma and other cancer cell lines. In addition, we developed an anti-SSX monoclonal antibody suitable for SSX protein detection in paraffin-embedded tissues. Using this antibody, we detected heterogeneous SSX expression in 9 of 18 melanoma cell lines,34 of 101 primary and metastatic melanoma cases, and 2 of 24 common nevocellular and atypical nevus cases. Normal SSX expression in the testis was found to be confined to spermatogonia. The implications of our findings for the use of SSX antigens as potential targets for immunotherapy are discussed.

Eighteen human melanoma cell lines were used in this study (listed in Table 1), the origin of which was reported previously (18, 19). Eight nonmelanoma cell lines were also analyzed: UMSCC (head and neck squamous cell carcinoma), HeLa (cervical carcinoma), A431 (cervical epidermoid carcinoma), NT2D1 (testicular germ cell cancer), CaCo2(colon carcinoma), A2243 (synovial sarcoma), U-4SS (sarcoma), and K562(erythroleukemia). All cell lines were cultured and harvested as described previously (18). To induce genome-wide demethylation, cultured cells were grown in DMEM medium containing 1μ m 5-aza-2′-deoxycytidine (Sigma). After 72 h of incubation at 37°C, the cells were processed for RNA extraction or immunofluorescence microscopy. One hundred and twenty-five paraffin-embedded melanocytic lesions were examined by immunohistochemistry (listed in Table 2). The lesions were surgically resected from patients treated at hospital centers in Nijmegen, the Netherlands (88 lesions), Copenhagen,Denmark (39 lesions), Würzburg, Germany (8 lesions), and Brussels, Belgium (6 lesions). Special care was taken that all specimens that were included in this multicenter European study were fixed in 4% formaldehyde in PBS. Two testicular biopsies (9-year-old and 41-year-old donors) and two normal thyroid paraffin-embedded specimens were also examined by immunohistochemistry.

RNA was isolated using either the RNeasy Mini (Qiagen) or RNAzol B(Campro Scientific) kits. One to 5 μg of total RNA were reverse-transcribed using a random hexamer and SuperScript RNase H⁻ Reverse Transcriptase (Life Technologies)according to the manufacturer’s instructions. One to 2 μl of RT product were PCR-amplified using Taq DNA polymerase (Life Technologies) following standard procedures. The sense primers SSX-Bcl(5′-TCTGATCATGCCCAAGAAGCCAGCAGAG-3′), which carries an artificial BclI restriction site (underlined), and SSX-start(5′-ACGGATCCCGTGCCATGAACGGAGACGAC-3′), which carries an artificial BamHI site (underlined), were both used in combination with the antisense SSXL-rev primer(5′-TTGTCGACAGCCATGCCCATGTTC-GTGA-3′), which carries an artificial SalI restriction site (underlined), to amplify 324-bp and 663-bp SSX fragments, respectively. PCR amplification of PBGD, coding for the enzyme porphobilinogen deaminase, was included as a positive control for the RT products using the primers 5′-GCAGATGGCTCCGATGGTGA-3′ (sense) and 5′-CTGGTAACGGCAATGCGGCT-3′ (antisense) giving rise to a 336-bp product. In some experiments, β-actin RT-PCR was performed as another positive control using primers 5′-GCTACGAGCTGCCTGACGG-3′ and 5′-GAGGCCAGGATGGAGCC-3′. SSX-Bcl/SSXL-rev PCR amplifications were performed for all cell lines, whereas SSX-start/SSXL-rev PCR was carried out in only nine cell lines (A375P, A375 M, BLM, 1F6, 1F6m,MV1, M14, Mel57, and 530). All PCR amplifications were performed using a GeneAmp PCR System 2400 (Perkin-Elmer).

SSX PCR amplification products were purified from agarose gels using the Easy-Pure kit (Biozym) and cloned into pGEM-T vectors using the pGEM-T Vector Systems kit (Promega) for further restriction analysis and sequencing. Confirmation of restriction analysis results was achieved by sequencing of the cloned PCR products using a vector-specific primer, the ABI PRISM Dye Terminator Cycle Sequencing Ready Reaction Kit (Perkin-Elmer), and the ABI 373A automated DNA sequencer (Applied Biosystems).

Mice were immunized with 10 μg of GST-SSX2 protein, produced, and isolated from Escherichia coli, as previously reported(13). During the following weeks, collected serum was tested for reactivity on formaldehyde-fixed and paraffin-embedded 1F6(SSX2-positive) and BLM (SSX2-negative) cells. A seropositive mouse was boosted i.v. for 2 subsequent days with 10 μg of GST-SSX2 protein. Two days after the last booster injection, the spleen was collected,and the spleen cells were isolated. These cells were fused with SP2/0 myeloma cells, using PEG4000, in a ratio of 10 spleen cells:1 myeloma cell. The hybridomas were screened by assessing the reactivity of the various supernatants toward 1F6 and BLM cells by immunofluorescence analyses, immunohistochemistry, and immunoblotting. Hybridomas secreting anti-SSX antibodies were subcloned several times until achieving monoclonality. Finally, the obtained mAbs were assessed for staining of formaldehyde-fixed and paraffin-embedded 1F6 and BLM cells by immunohistochemistry.

Cultured cells were lyzed in RIPA buffer [50 mm Tris (pH 8.0), 150 mm NaCl, 1% NP40, 0.5% sodium deoxycholate,0.1% SDS, 0.1% 2-mercaptoethanol, 1 mmphenylmethylsulfonyl fluoride, 4.8 μg/ml aprotinin] following standard methods (20). Subsequently, equal amounts of each cell extract were electrophoresed in SDS-polyacrylamide gels and immunoblotted using the E3AS mAb (1:1000 dilution) and B39 polyclonal antibody (1:3000 dilution), essentially as previously described(13). As secondary antibodies, horseradish peroxidase-coupled goat antimouse (1:1000 dilution) and horseradish peroxidase-coupled goat antirabbit antibodies (DAKO) were used. Detection was performed by incubating the blotted nitrocellulose filters with enhanced chemiluminescence substrate (Amersham) and subsequent autoradiography.

HeLa cells were grown in DMEM medium containing 10% FCS. Different SSX cDNAs (SSX1 to SSX4 and SSX2 deletion mutants) cloned in eukaryotic expression vectors containing epitope tags (pCATCH and pSG8-VSV) were transiently transfected into HeLa cells using Dosper liposomal reagent according to the manufacturer’s instructions (Boeringher Mannheim). Indirect immunofluorescence assays were performed as described previously (13). As primary and secondary antibodies, we used the E3AS mAb (1:10 dilution) and FITC-conjugated swine antimouse IgG (DAKO; 1:100 dilution).

As a testing system for immunohistochemical staining of paraffin-embedded pathology specimens, we embedded SSX-positive 1F6 and SSX-negative BLM cells in gelatin. These gelatin devices were fixed in 4% formaldehyde/PBS and embedded in paraffin. Four-μm sections were deparaffinized and treated by microwave heating for 20 min at 650 W in 0.1 mm sodium citrate buffer (pH 6.0). The slides were adjusted to room temperature for 30 min and preincubated for 10 min in 20% normal horse serum in PBS. This solution was decanted, and the E3AS mAb was applied as hybridoma culture supernatant for 60 min at room temperature. We used an avidin-biotin complex-peroxidase method for antibody detection as previously described (21). The signal was enhanced with catalyzed reporter deposition, an amplification method developed in our laboratory based on the deposition of BT (22). The BT precipitate was then visualized with peroxidase-labeled avidin and 3-amino-9-ethylcarbazole as a substrate. The normal testis and thyroid specimens were analyzed by immunohistochemistry using a similar protocol but without the use of BT. In addition, these samples were stained with E3AS mAb preincubated with GST-SSX2 or GST-SYT peptide to assess the antibody specificity.

For each section, the percentage of positive melanocytic cells was estimated. Each section was assigned to one of the following percentage categories: 0%, 1–5%, 5–25%, 25–50%, 50–75%, and 75–100%. Positive melanocytic staining was scored when at least 1% of melanocytic cells were stained. The scoring was performed by two observers (N. R. d. S. and T. J. d. V.), but doubtful cases were analyzed and judged by a panel of four observers (N. R. d. S.,T. J. d. V., G. N. P. v. M., and D. J. R.).

The discovery of SSX2 ectopic expression in several human cancers, including melanoma (9), prompted us to evaluate its expression in a panel of established human melanoma cell lines(Table 1). Using primers that recognize and amplify the 3′ ends of the five known SSX genes (SSX1 to SSX5),we observed a specific band of 324 bp in 9 of 18 melanoma cell lines tested (Fig. 1,A; Table 1). Eight nonmelanoma cell lines were also analyzed,and only the U-4SS and K562 cell lines were found to be SSX-positive(Table 1). In all cases, RT-PCR amplification of PBGD mRNA,a housekeeping gene, was included as a control to ascertain the appropriate quality and quantity of the different RNA samples (Fig. 1,B). In nine of the melanoma cell lines, an additional SSX RT-PCR reaction was performed using a primer set that amplifies full-length SSX mRNAs (Fig. 1,C). Accordingly, the cell lines that were previously found to be SSX-positive gave rise to a specific band of 663 bp, whereas negative cell lines remained so, as expected (Fig. 1 B).

Restriction digestion of the short (324 bp) PCR products using different enzymes known to discriminate between SSX1, SSX2, and SSX3 (8; Fig. 1,C) indicated that more than one SSX gene was expressed in four of nine positive melanoma cell lines. To identify the corresponding cDNAs, the PCR fragments were cloned into plasmid vectors and propagated in E. coli. Several of these clones were isolated and subjected to restriction and sequencing analyses (Fig. 1,C). By doing so, we found in three cell lines only SSX1 (Omel2, MZ2-MEL-3.0, and U-4SS) and in three other cell lines only SSX2 (1F6m, 518A2, and SK-mel-28) PCR products(Table 1). Besides, fragments with higher molecular weight (∼450 bp)were also amplified from cell lines 518A2 and SK-mel-28 (Fig. 1,A, arrowhead). These fragments were found to correspond to an alternatively spliced SSX2 variant containing an additional intronic sequence of 148 bp (GenBank accession no. AF190791). By analyzing clones derived from the remaining four positive cell lines (1F6, 530, A375P, and A375 M), we obtained evidence for the expression of the SSX4 and SSX5 genes recently identified by Güre et al. (10)in three of them. SSX4 expression was detected in the 530 and A375 M cell lines, whereas SSX5 expression was detected only in the A375P cell line (Table 1). The SSX1 and SSX2 genes were both expressed in these four cell lines. It should be noted that no SSX3 gene expression was found in any of the melanoma cell lines tested.

For further analysis at the protein level, we set out to generate a mouse mAb against SSX proteins. Several hybridomas were produced, and because we were particularly interested in developing an antibody able to detect protein from paraffin-embedded tissues, each hybridoma was assessed for its ability to detect SSX expression in paraffin-embedded 1F6 cells. BLM cells, which were SSX-negative by RT-PCR, were used as a negative control. By doing so, we identified one hybridoma (E3AS) that was able to specifically detect SSX in 1F6 but not in BLM cells (Fig. 4, a-c).

To determine the specificity of the E3AS mAb, HeLa cells were transiently transfected with different epitope-tagged SSX expression constructs (SSX1 to SSX4) and subjected to indirect immunofluorescence analysis. We found that the E3AS mAb was able to detect SSX2, SSX3, and SSX4 but not SSX1 protein in the nucleus of HeLa cells (not shown). The SSX5 protein was not tested. The B39 polyclonal antibody(13) was also found to recognize SSX2, SSX3, and SSX4 but not SSX1 (not shown). Interestingly, SSX1 contains the most dissimilar amino acid sequence in the SSX family of proteins (10),suggesting that the E3AS mAb recognizes a less conserved epitope. To determine the location of the E3AS mAb epitope, we transiently transfected HeLa cells with several SSX2 deletion mutants (Fig. 2) and tested them for reactivity by immunofluorescence analysis. Absence of the most NH₂-terminal 110 amino acids (as in standard SYT-SSX fusions; FLAG-SYT-SSX2) led to loss of E3AS mAb staining. Conversely, COOH-terminal deletions up to amino acid 81 did not abrogate E3AS staining. Because absence of the 24 most NH₂-terminal amino acids did not prevent E3AS recognition, the epitope of this mAb maps between amino acids 25 and 80(Fig. 2).

To test whether the E3AS mAb could also recognize normal SSX expression in paraffin-embedded tissues, we analyzed two testicular and two thyroid biopsies by immunohistochemistry. One biopsy was obtained from an adult male testis, whereas the other was obtained from a prepuberal male testis. SSX expression in mature testis was found to occur only in the nuclei of spermatogenic cells, mainly in spermatogonia and occasionally in primary spermatocytes close to the basement membrane(Fig. 3). The expression was heterogeneous, and only a fraction of spermatogonia was stained. Likewise, in immature testis, SSX expression was found to occur in a portion of spermatogonial cells, the predominant cell type present (data not shown). Neither interstitial cells nor Sertoli cells were stained by the E3AS mAb. Attesting the specificity of the mAb E3AS, we found that preincubation with the immunizing antigen (GST-SSX2) abolished nuclear staining of spermatogenic cells. In contrast, preincubation with an unrelated protein (GST-SYT) had no effect on E3AS staining (data not shown). Surprisingly, none of the thyroid tissue sections examined revealed SSX staining.

Thirteen melanoma cell lines were analyzed by immunoblotting (B39 and E3AS antibodies) and immunofluorescence microscopy (E3AS mAb; Table 1). Immunoblotting using the B39 antibody revealed specific bands of 29 kDa in 1F6, 1F6m, and 518A2 cell lysates, as well as in a positive control(COS-1+SSX2; Fig. 5,A). Less intense bands of similar weight could be discerned in the immunoblotted SK-mel-28, A375P, and A375 M cell lysates. Identical results were obtained using the E3AS mAb (not shown). Immunofluorescence analysis using the E3AS mAb disclosed nuclear stained cells in 9 of 13 analyzed cell lines. However, the percentage of positive cells varied markedly between cell lines. In three of them (1F6, 1F6m, and 518A2), the majority of cells were positive (>80%), whereas five other cell lines (Omel2, 530, A375P,A375 M, and MZ2-MEL-3.0) showed very low numbers of positive cells(<2%). The SK-mel-28 cell line showed an intermediate percentage of positive cells (∼25%; Table 1).

Using the E3AS mAb, we observed SSX nuclear staining in 36 of 125(29%) melanocytic lesions. Apart from two atypical nevi that displayed weakly positive cells (Fig. 4,k), no SSX staining was found in the remaining benign human melanocytic lesions (Fig. 4,l). Within the malignant lesions, the frequency of E3AS positivity ranged from 24% in locoregional metastases to 40% in primary melanomas(Table 2). Among all lesions studied, SSX expression was very heterogeneous, with the majority of the cases exhibiting low numbers of positive cells (<25%; Table 2). Large SSX-negative areas were observed in several tumors. Four main staining patterns could be distinguished in both primary and metastatic melanomas: (a)widespread, with >75% of positive cells (Fig. 4,d);(b) focal, with positive cells clustered in a restricted area from the tumor (Fig. 4,e); (c) scattered,with a low percentage of positive cells present in several areas of the tumor (Fig. 4,f); and (d) isolated, with few positive cells localized in largely negative areas (Fig. 4,g). Apart from the typically observed nuclear staining,some cases showed aberrant patterns. In one primary tumor, cytoplasmic staining of mitotic cells was observed next to nuclear staining in other cells (Fig. 4,h). In seven melanoma lesions (two primary melanomas and five metastases), tumor cells showed cytoplasmic or membrane staining, either exclusively or in conjunction with the typical nuclear staining (Fig. 4,i). Also, one lymph node metastasis exhibited SSX staining in the cytoplasm of fibroblast-like cells adjacent to tumor cells (Fig. 4 j).

Because 32 patients had both primary and metastatic tumors, we investigated whether the status of SSX expression varied in the different lesions from each patient (Table 3). Fourteen patients developed both primary and metastatic SSX-negative lesions, whereas in six patients, all lesions showed SSX nuclear positivity. Six patients had SSX-negative primary melanomas and at least one SSX-positive metastasis, whereas six other patients had SSX-positive primary tumors and negative metastases. Furthermore, no correlation was found between SSX expression and Clark staging, Breslow tumor thickness, or age of the patients.

To assess whether SSX expression might be regulated by demethylation, we treated one SSX-negative (BLM) and one SSX-positive(K562) cell line with DAC. RNA was extracted from these cell cultures before and after DAC treatment, and SSX expression was evaluated by RT-PCR. Treatment with 1 μm of DAC for 72 h resulted in de novo induction of SSX expression in BLM cells and a clear increase of expression in the K562 cell line (Fig. 5,B). The effect of DAC treatment on SSX expression was also assessed at the protein level for BLM cells. Immunofluorescence microscopy revealed that DAC treatment resulted in SSX nuclear staining in about 9–13% of the cultured cells (Fig. 5 C). In contrast, no cells were stained in the absence of DAC. Furthermore, incubation with higher concentrations of DAC (10μ m) did not increase the percentage of SSX-positive BLM cells (∼7%; not shown).

The SSX gene family is comprised of five members,including the SSX1 and SSX2 genes, that are fused with the SYT gene in t(X;18)-positive synovial sarcomas(4, 5). Recently, it was found that SSX1, SSX2, SSX4, and SSX5, but not SSX3, RNAs may be aberrantly expressed in various tumor types, in particular, human melanoma (9, 11). Besides,normal SSX gene expression, as detected at the RNA level,was found to be restricted to testicular and thyroid tissues (4, 6). In the present study, we developed an anti-SSX-specific mAb and evaluated the levels of SSX protein expression in normal testicular and thyroid tissues, a panel of cancer cell lines, common nevocellular and atypical nevus lesions, and primary and metastatic melanoma lesions.

Of 18 melanoma cell lines studied by RT-PCR, 9 were found to express at least one SSX gene, SSX1 and SSX2being most frequently expressed (Table 1). The levels of SSXexpression were variable, as also found by Güre et al.(10) in their panel of melanoma cell lines. In three pairs of melanoma cell lines derived from the same patient (A375P/A375 M,1F6/1F6m, and MV1/MV3), the overall presence or absence of SSX expression was concordant. However, the SSXfamily members were different in each pair, suggesting that the expression of certain SSX genes might have been switched on or off after cell line establishment. Using a newly developed mAb(E3AS) that recognizes the NH₂ terminus (amino acids 25–80) of SSX2, SSX3, and SSX4 (but not SSX1) proteins, we searched for expression in 13 of 18 melanoma cell lines by immunoblotting and immunofluorescence. In contrast to the former method, which revealed SSX bands in only six cell lines,immunofluorescence analysis allowed the detection of SSX protein in all cell lines that were found to be SSX-positive by RT-PCR. However, the percentage of SSX-positive cells varied considerably among these cell lines (Table 1). A low percentage of SSX-positive cells was observed in two cell lines (530 and MZ2-MEL-3.0) that showed strong SSXexpression by RT-PCR. This discrepancy is likely due to the fact that the E3AS mAb does not detect SSX1 expression. Altogether, these results indicate that the different levels of SSX mRNA detected by RT-PCR are not due to differential intensities of transcriptional activity but rather to cell-to-cell heterogeneity.

The E3AS mAb was designed and found to be suitable for the detection of SSX protein in paraffin-embedded tissues. Using this antibody, we analyzed 125 cases of benign and malignant melanocytic lesions by immunohistochemistry. Heterogeneous SSX expression was found in 40% of primary melanomas analyzed. This percentage is similar to what has been observed by other investigators using RT-PCR (43% [11]), but it may be underestimated because SSX1 expression is not detected by this antibody. Yet, this underestimation may not be of major significance because others found that only 1 of 37 (3%) melanomas tested expressed SSX1 alone (11). The heterogeneity of SSX expression in the melanocytic lesions was striking, ranging from widespread to scarce. This finding parallels our observations in the melanoma cell lines and, therefore, indicates that SSX-expressing cells do not undergo positive or negative selection during cell line establishment.

Thirty-two patients from our series developed both primary and metastatic melanomas (Table 3), allowing the assessment of whether SSX expression could be related to tumor progression. Because six patients had SSX-positive primary melanomas and SSX-negative metastases, we conclude that expression of these proteins does not confer a clonal selective advantage to melanoma cells in vivo. This is in contrast with the reported association between MAGEexpression and tumor progression and increased aggressiveness(19, 23).

The SSX proteins are members of the still growing family of CT antigens. All CT antigens (e.g., MAGE, BAGE, GAGE,NY-ESO-1/LAGE-1, SCP-1, CT7, and SSX) share common characteristics:(a) normal expression in the testis only, (b)ectopic expression in malignancies of various histological origins,(c) existence of several family members, and (d)localization on the X chromosome (10, 24, 25). However,not all CT antigens fully comply with these rules. For example, several MAGE genes are also expressed in the placenta(26), SSX1 and SSX2 RNAs are detected in the thyroid (4, 6), and the SCP-1gene is localized on chromosome region 1p12-p13 (27). SSX expression at the protein level was detected in the testis but not in two normal thyroid specimens examined by immunohistochemistry (this report). Additional RT-PCR analyses were not performed on these samples due to lack of fresh and/or frozen material. However, our results combined with those of others that did find SSX expression in the thyroid by RT-PCR (4, 6) suggest that this expression may be too low to be detected at the protein level. Alternatively, the discrepancy may be explained by a bias in sample biopsies.

Immunohistochemical detection of SSX proteins in the testes of two donors revealed nuclear expression in spermatogonia, both from mature and immature testes. Several studies have shown that spermatogenic cells undergo profound changes in methylation patterns(28, 29, 30). In several cases, it was observed that hypomethylation of certain promoter sequences occurs at the early stages of spermatogenesis (i.e., in spermatogonia), thereby being linked to the activation of transcription (29, 31, 32). SSX expression in spermatogonia suggests that also these genes may be activated after demethylation of their promoter sequences. Our finding that DAC incubation induces SSXexpression in otherwise non- or low-expressing cells further supports this hypothesis. A similar phenomenon has been observed for MAGE-1, a gene that is also primarily expressed in spermatogonia (33) and is activated by DAC treatment(19, 34).

The testis is an immune privileged organ because spermatogenic cells do not express HLA class I and II antigens at the cell surface (35, 36). Concomitantly, the testis has a so-called blood-testis barrier in the seminiferous tubuli generated by the Sertoli cells. Because ectopic expression of CT antigens may thus lead to an autologous cellular and/or humoral immune response (12, 37), these antigens are presently being used as targets for the development of cancer immunotherapy protocols. Several clinical trials have been set up with relative success seeking immunization against CT or melanocyte-specific antigens (e.g., gp100, tyrosinase,and MART-1/Melan-A) in metastatic melanoma patients(38, 39, 40, 41). As CT antigens, the SSX proteins may also be immunogenic when ectopically expressed and, therefore, be suitable as targets for immunotherapy. Previous studies have reported the presence of anti-SSX antibodies in the sera of melanoma patients (9, 12). In the present study, we found that the proportion of SSX-expressing cells varied considerably among the different samples tested, thereby raising the question as to whether all SSX-positive tumors would be equally suitable for immunotherapy. Heterogeneous expression in melanomas has also been reported for the MAGE-1 protein (42, 43), but the MAGE proteins tend to be uniformly expressed in multiple melanoma lesions from the same patient(44). In contrast, we found that 12 of our patients developed both SSX-negative and -positive lesions (Table 3), suggesting that, in such cases, anti-SSX immunotherapeutic approaches would not be equally effective for all these lesions. Also, melanoma cases with a low percentage of SSX-expressing cells may be less suitable for immunotherapy unless a so-called bystander effect occurs, i.e., if the SSX-targeted action of CTLs would create a cytokine environment, triggering an inflammatory response that would kill neighboring tumor cells (45).

Because DAC induces SSX expression in vitro, it would be of interest to study the effectiveness of DAC as a coadjuvant to SSX vaccination in cases with low levels of SSX expression. Previously,DAC has been shown to inhibit tumor growth and induce differentiation(46, 47), and it has been successfully administered to leukemia patients as a therapeutic agent (48). Furthermore, the recent observation that DAC also up-regulates the expression of HLA class I antigens and costimulatory molecules(e.g., ICAM-1 and LFA-3) in cultured melanoma cells(49) suggests that this agent may enhance the immunogenicity of tumor cells and, therefore, may serve as an important coadjuvant in immunotherapeutic protocols targeting SSX and/or other CT antigens for the treatment of melanoma.

In conclusion, the generation of a mAb (E3AS) specifically directed against SSX proteins allowed the detection of these proteins in a series of primary and metastatic melanoma lesions. The use of this antibody revealed a high level of heterogeneity among the different cases studied. This finding stresses the importance of assessing SSX expression at the protein level in melanoma cases that may be selected for immunotherapeutic trials. However, because the E3AS mAb does not recognize SSX1 protein, RT-PCR amplification remains useful to complement immunohistochemistry to evaluate more accurately the status of SSX expression in tumors. Despite the relatively low levels of SSX expression in melanomas when compared with other antigens(e.g., MAGE, gp100, and tyrosinase), we maintain that melanoma immunotherapy targeted to SSX should be contemplated,particularly in those cases with high levels of SSX protein expression.

We thank Drs. S. A. Aaronson, J. Pontén, and C. van Roozendaal for kindly providing the A2243, U-4SS, and NT2D1 cell lines,respectively. We are also grateful to Bert Janssen, Aïcha Fourkour, and Miriam Smeets for invaluable technical assistance. Members of the pathology subgroup of the European Organization for Research and Treatment of Cancer-melanoma cooperative group are acknowledged for providing primary and metastatic melanoma specimens. Klaus Hou-Jensen (Copenhagen, Denmark), Eva B. Bröcker(Würzburg, Germany), and Nathalie Renard (Brussels, Belgium)generously provided paraffin-embedded specimens.