Preferentially expressed antigen in melanoma (PRAME) was first discovered in melanomas and it was associated with cytotoxic T cell activation [1]. Shortly after its discovery, it was also shown to be expressed in acute leukemia cells [2]. The function of PRAME appears to be extensive though it was first identified as a repressor of the retinoic acid receptor pathway [3]. PRAME also inhibits myeloid differentiation in a retinoic acid-dependent and independent manner as well [4]. PRAME, like other cancer-testis antigens, has been shown be minimally expressed in adult human organs except for gonadal tissues and various human cancers including sarcomas.

Cancer-testis antigens such as MAGE-A and NY-ESO-1 have been widely explored, and these tumor-associated antigens have served as the therapeutic target of various vaccine strategies and adoptive cellular therapies. Objective tumor regressions of cutaneous metastases of melanoma patients have been documented with a MAGE-3.A1 peptide [5]. Similarly, patients with synovial sarcoma treated with genetically engineered autologous T cells with NY-ESO-1 recognition experienced RECIST partial responses as noted in 11 of 18 cases (61%) [6].

Various subtypes of sarcomas have demonstrated expression of cancer-testis antigens including synovial sarcomas, myxoid liposarcomas, chondrosarcomas, and osteosarcomas. Co-expression of PRAME and NY-ESO-1 has been shown to correlate with high-grade histologic features and a worse overall survival in patients with myxoid liposarcomas [7] and synovial sarcomas [8]. High protein expression levels of PRAME have been shown to correlate with a worse overall survival in osteosarcoma, and the expression of PRAME was more common in metastases compared to primary tumors [9]. In chondrosarcoma, a disease with low expression of PRAME at baseline, induction of PRAME with 5-aza-2-deoxycitabine rendered chondrosarcoma cells targetable by PRAME-specific CD8+ T cells [10].

The goal of this study is to evaluate the expression of PRAME across multiple sarcoma subtypes and normal tissues using three large public datasets. We report statistically significant associations to guide PRAME-specific therapies of sarcoma. To our knowledge this is the first comprehensive analysis of PRAME in multiple sarcoma subtypes and clinical subsets. In addition, we evaluated associations of T cell and antigen expression markers with PRAME expression to show how these may affect immunotherapies targeting this antigen.

Sarcomas represent a rare collection of neoplasms of mesenchymal origin that make up less than 1% of all cancer cases diagnosed each year in the United States [14]. While surgery can be curative for low-grade/low-stage disease, unresectable/metastatic disease is treated with systemic therapies. These therapies for soft tissue sarcomas have improved in slow incremental steps over the last 40 years with limited success [15]. Treatment of sarcoma is also difficult because more than one hundred subtypes have been identified. The success of immunotherapies in certain tumors provide a new avenue for investigation for the treatment of sarcoma. In fact, a recent phase II study (SARC028) of pembrolizumab for advanced/metastatic sarcomas demonstrated a response rate of 17% [16]. In the current study, we show that PRAME could be an effective immunotherapy target in specific sarcoma subtypes. Furthermore, we show that PD-1 and PD-L1 are also expressed in a heterogeneous manner, supporting further evaluation of anti-PD-1 and anti-PD-L1 therapies in sarcomas.

We determined that uterine carcinosarcoma, a disease that lacks standard therapies, highly overexpresses the PRAME antigen. Other cancer testis antigens have been identified as potential immunotherapy targets for this malignancy, including MAGE-A4 and NY-ESO-1 [17]. We propose that PRAME might also be an effective and broadly expressed target. Although the other sarcoma tumors included in the sarcoma TCGA showed a lower expression, a subset of samples was clearly characterized by PRAME overexpression. Multiple CCLE cell lines that were derived from sarcoma tumors retained PRAME expression, with the exception of chondrosarcoma, which was not included in the sarcoma TCGA.

Our analysis of PRAME expression associations with subtypes and clinical variables revealed that PRAME is overexpressed in synovial sarcoma and in multifocal leiomyosarcoma. Notably, all other sarcoma subtypes showed a highly heterogeneous PRAME expression, from zero to very high expression. This is in line with heterogeneity of sarcomas, which represents a major challenge. Cancer testis antigens, focusing primarily on NY-ESO-1, are currently being tested in clinical trials and show promise as targets of adoptive immunotherapies and cancer vaccines to treat sarcoma [18]. PRAME co-overexpression with NY-ESO-1 in synovial sarcoma suggests that this antigen may also be targeted efficiently by these immunotherapy approaches.

The negative correlation that we identified between PRAME and expression of antigen presentation-related genes also supports that tumor-associated antigens derived from the PRAME protein can be recognized by the immune system, and tumors lose expression of genes involved in antigen presentation to avoid an effective immune response. Indeed, in the PRAME-overexpressing synovial sarcoma, beta2-microglobulin (B2M) loss appears to be a mechanism by which the tumor avoids immune recognition, evidenced by low expression of the CD8A cytotoxic T cell marker. However, we also found non-zero interferon gamma (IFNG) expression in half of the synovial sarcoma samples that we analyzed, therefore we hypothesize that there are functional and active T cells in those tumors that may be exploited to develop effective immunotherapies. Furthermore, MHC class I recovery approaches [19] might also be needed to ensure the success of T cell-based immunotherapy of sarcoma.

Our analysis of sarcoma subtypes shows that uterine carcinosarcoma, synovial sarcoma, and multifocal leiomyosarcoma samples overexpress this antigen and patients with these malignancies would potentially benefit from PRAME-specific immunotherapies. We also found a negative correlation between PRAME and expression of genes involved in antigen presentation, which may provide a way for tumors to avoid immune recognition.