Increased NY-ESO-1 Expression and Reduced Infiltrating CD3+ T Cells in Cutaneous Melanoma

NY-ESO-1 is a cancer-testis antigen aberrantly expressed in melanomas, which may serve as a robust and specific target in immunotherapy. NY-ESO-1 antigen expression, tumor features, and the immune profile of tumor infiltrating lymphocytes were assessed in primary cutaneous melanoma. NY-ESO-1 protein was detected in 20% of invasive melanomas (16/79), rarely in in situ melanoma (1/10) and not in benign nevi (0/20). Marked intratumoral heterogeneity of NY-ESO-1 protein expression was observed. NY-ESO-1 expression was associated with increased primary tumor thickness (P = 0.007) and inversely correlated with superficial spreading melanoma (P < 0.02). NY-ESO-1 expression was also associated with reduced numbers and density of CD3+ tumor infiltrating lymphocytes (P = 0.017). When NY-ESO-1 protein was expressed, CD3+ T cells were less diffusely infiltrating the tumor and were more often arranged in small clusters (P = 0.010) or as isolated cells (P = 0.002) than in large clusters of more than five lymphocytes. No correlation of NY-ESO-1 expression with gender, age, tumor site, ulceration, lymph node sentinel status, or survival was observed. NY-ESO-1 expression in melanoma was associated with tumor progression, including increased tumor thickness, and with reduced tumor infiltrating lymphocytes.


Introduction
Cutaneous melanoma is a global health problem, with increasing incidence in Caucasians [1][2][3][4]. While more than 95% of patients with early stage melanoma are recurrencefree 5 years after diagnosis, fewer than 50% of patients with metastases survive more than 5 years [5]. Despite the revolutionary developments in immunotherapy, the treatment of patients with metastatic melanoma has been focused more on extending progression free survival and palliation than on effecting a cure [6][7][8][9][10].
Consistent with the recent successes of immunotherapy, there is interest in vaccine development for melanoma. Cancer-testis antigens, such as NY-ESO-1, may serve as robust and specific targets; while they are normally expressed only in the germ cells of the adult testis, they are aberrantly expressed in tumors [11][12][13][14][15][16][17].
NY-ESO-1 is a well-characterized cancer-testis antigen expressed in 20-30% of carcinomas (lung, esophagus, liver, stomach, prostate, ovary, endometrium, and bladder) and some sarcomas [18][19][20][21][22][23][24][25][26][27]. Prior reports showed NY-ESO-1 expression in 20 to 40% of melanomas; metastases more likely express this antigen than primary cutaneous melanoma [28][29][30][31]. NY-ESO-1 is a highly immunogenic antigen in vitro and is presented to T cells via HLA. The presence of circulating NY-ESO-1 antigen-specific T cells in patients with metastatic melanoma correlates with better prognosis [32,33]. To further understand the role of NY-ESO-1 in melanoma progression, we studied the expression of NY-ESO-1 protein in tissue sections of benign and malignant cutaneous melanocytic tumors and analyzed the association if any with melanoma progression and melanoma-specific survival. We also assessed the immune profile of the tumor infiltrating lymphocytes (TIL) in primary invasive melanoma and in the context of NY-ESO-1 antigen expression.

Case Selection.
This cohort was composed of 109 Brazilian patients with melanocytic skin tumors: benign nevi ( = 20), melanoma in situ ( = 10), and invasive primary cutaneous melanoma ( = 79). Data was collected from the Division of Dermatology, University of São Paulo, and this study has been approved by the Institutional Review Boards of University of São Paulo (CAPPesq) and of Massachusetts General Hospital (2013P000172).
Demographic features assessed included gender, age, skin phototype, tumor site, sentinel lymph node status, development of metastasis, and melanoma-associated death. Histopathological evaluation included tumor thickness, histological type, ulceration, and grade of tumor infiltrating lymphocytes (TIL). TIL were graded as brisk when lymphocytes diffusely infiltrated throughout the tumor or were present as a broad band at the advancing tumor margin; nonbrisk when there was focal or multifocal infiltration; and absent when there were no lymphocytes infiltrating the tumor or the lymphocytic infiltrate was not associated with the tumor cells [34].

Quantification of TIL.
We adapted the morphometric point-counting method previously described for mast cell quantification to assess the volume density of lymphocyte subsets [35,36]. Briefly, slides are viewed in a microscope using a specialized eyepiece graticule that contains a 10 × 10 mm square grid (100 contiguous squares). The intersection of any two lines defined individual points, and cells at an intersection within a 400x magnification field were counted. The grid was superimposed over the TIL stained for CD3, CD8, FoxP3, and CD8FoxP3. Whenever possible, the four fields with highest cell density, either in the tumor or at the periphery, were counted and the mean per each slide was used for statistical analysis. The distribution pattern of CD3+ T cells was reported as isolated cell (only one cell), small clusters (2 to 5), or large clusters (6 or more T cells).

Statistical Analysis.
Association between NY-ESO-1 expression and tumor thickness was performed using a two tailed Student's -test. Associations between NY-ESO-1 expression and gender, age, site, phototype, CD3+, and CD8+FoxP3− cells were analyzed by Pearson's chi-square test. Associations between NY-ESO-1 expression and melanoma type, ulceration, CD8−FoxP3+, and CD8+FoxP3+ cells were analyzed by Fisher's exact test. value of ≤0.05 was considered to be statistically significant. Statistical analysis was performed by SAS software.
Sentinel lymph node (SLN) biopsy was performed in 23% (18/78) of patients, of which 11% (2/18) were positive. The criteria used to indicate sentinel lymph node biopsy at University of São Paulo were clinically localized primary cutaneous melanomas with thickness between 1.0 and 4.0 mm. Regarding the 61 cases without sentinel lymph node biopsy, 57% (45/79) did not meet the criteria and 20% (16/79) were diagnosed with melanoma before sentinel lymph node biopsy becomes the standard of care.
We assessed TIL profile in 57 cases of vertical growth phase melanomas with either brisk or nonbrisk TIL, using immunohistochemical staining for CD3, CD8, and coexpression of CD8 and FoxP3 double stain. The mean number of CD3+ cells in brisk group was significantly higher than nonbrisk group ( = 0.0106). CD3+ T cell density and patterns of tumor infiltration correlated with NY-ESO-1 expression. CD3+ T cells were more numerous in tumors negative for NY-ESO-1 ( = 0.017, Figure 4). Moreover, there was a correlation between large aggregates of CD3+ T cells and absence of NY-ESO-1 ( = 0.0001), whereas NY-ESO-1 positive tumors usually had isolated ( = 0.0019) and small clusters of CD3+ cells within the tumor ( = 0.0125; Figure 5).

Discussion
A recent study showed that 37% of 348 primary cutaneous melanomas were NY-ESO-1 positive by immunohistochemical analysis (IHC) [28]. Other authors assessing 586 melanoma samples (IHC) observed positivity of 46% for NY-ESO-1, including 251 primary melanomas and 335 metastases [29]. As an alternative, PCR testing done in 52 patients with melanoma showed that 33% of tumors were positive for NY-ESO-1 [30]. In another research, metastatic melanomas had a higher NY-ESO-1 expression of 32% compared to 13% in primary cutaneous melanoma (IHC) [31]. The variability  between these studies may be due to the stage of melanoma and the use of different laboratory techniques.
In the present study, NY-ESO-1 expression was associated with melanocytic tumor progression. Of 20 cases of benign nevi, 10 melanomas in situ, and 11 radial growth phase melanomas, only one melanoma in situ had detectable NY-ESO-1. On the other hand, 24% of vertical growth phase melanomas had detectable NY-ESO-1 protein (16/68).
We showed that NY-ESO-1 protein expression is associated with tumor thickness. Other researchers had similar findings. One study evaluating 251 cutaneous melanomas showed increased NY-ESO-1 in the intermediate group, between 1.1 and 4.0 mm of thickness, when compared with melanoma <1.1 mm [29]. In another one, which analyzed 61 cutaneous melanomas, the NY-ESO-1 positive tumors had a median thickness of 4.7 mm versus 1.53 mm in the NY-ESO-1 negative group [31]. Recently, 321 patients with cutaneous melanoma also presented higher NY-ESO-1 antigen expression with increasing tumor thickness [28]. Our results and those previously published support the association between NY-ESO-1 protein expression and advanced melanomas, more invasive and thicker. However, a causal relationship cannot be established, as these findings come from retrospective cohort studies with biopsy done only once.
The expression of NY-ESO-1 protein in primary cutaneous melanoma is markedly heterogeneous. In our study, 33% of melanomas presented complete expression, regional expression in part of the tumor was observed in 47%, and, less commonly, a diffusely scattered expression was seen in 18%. These findings are somewhat different from those previously reported, wherein a diffuse pattern was the most common and focal (regional) distribution was the least common [29]. Another series reported yet different results, with 37% 6 Journal of Immunology Research complete, 13% regional, and 50% scattered expression [31]. Therefore, the studies have consistently showed significant intertumoral heterogeneity in the expression of NY-ESO-1 protein. We also observed significant variability in the percentage of tumor cells expressing NY-ESO-1 protein: 41% of the NY-ESO-1 (+) melanomas ( = 7) expressed the protein in 2-20% of tumor cells, 18% ( = 3) expressed the protein in 21-60% of tumor cells, and 41% ( = 7) showed NY-ESO-1 protein in more than 60% of the tumor cells.
Despite the scientific evidence of an association between NY-ESO-1 expression and tumor thickness in primary cutaneous melanoma, an association with prognosis has not been confirmed. Cancer-testis antigen expression has been reported as an adverse prognostic indicator in a wide range of cancers, including neuroblastoma, ovarian cancer, breast cancer, and multiple myeloma [38][39][40][41]. NY-ESO-1 expression associated with poor prognosis in melanoma has been reported by some [29][30][31]. Nevertheless, our results and another report showed no association with overall survival [28]. We found no association of NY-ESO-1 expression with gender, age, race, skin phototype, tumor anatomic location, or sentinel lymph node status, findings that are consistent with a prior study [28].
As a potential immunogenic protein, it is important to know how NY-ESO-1 is related to the immune response and not just whether it is present or not. The tumor immunological environment is critical to the impact, if any, of NY-ESO-1 protein expression on prognosis. It is well established that the presence of TIL is associated with better prognosis in melanoma [42][43][44][45]. The present study included all grades of TIL (absent, nonbrisk, and brisk) but showed no statistically significant association between NY-ESO-1 and TIL grade. Another study has also reported a heterogeneous association of NY-ESO-1 expression and grade of TIL. Of 77 primary melanomas expressing NY-ESO-1, 57% had absent TIL, 38% nonbrisk TIL, and 5% brisk TIL, but the difference was not significant [28].
In order to further explore the immune environment, we assessed specific T cell subsets. Similar to the protective effect reported for TIL in general, CD3+ cells in particular have been associated with improved survival in cancer [46]. It is likely that the composition of the lymphocytic infiltrate is critical in this setting. We showed fewer CD3+ T cells in NY-ESO-1 (+) melanomas, and these cells were isolated or in small clusters, while, in the negative NY-ESO-1 tumors, the number of CD3+ cells was higher and the cells were arranged in large aggregates. Although NY-ESO-1 expression was not associated with the grade of TIL, it was associated with fewer infiltrating CD3+ T cells, a puzzling finding not described before. It would be expected that NY-ESO-1 expression could induce and enhance an antitumor immune response.
Clearly, the interplay between NY-ESO-1 and the immune response is very complex and large prospective cohorts investigating this interaction are need. Nevertheless, the association of NY-ESO-1 expression and the composition of the tumor lymphocytic infiltrate provide insights into the physiopathogenesis of immune response in primary melanoma and may drive future immunotherapeutic approaches [49].

Conclusions
NY-ESO-1 is a melanocytic tumor progression marker that is not present in benign nevi but is observed in approximately one-fifth of primary cutaneous melanomas, more commonly in thicker tumors and those with vertical growth phase. NY-ESO-1 protein expression displays marked heterogeneity, from 2% to 100% of the tumor cells, with varying expression patterns.
The present study is original in showing that NY-ESO-1 positive melanomas had fewer CD3+ TIL, and the T cells were most often dispersed as isolated cells or small clusters.