PRAME is expressed in sarcoma and shows high overexpression in uterine carcinosarcoma
To determine the relevance of PRAME as a target in sarcoma, we compared all normal (GTEx, n = 30 tissue types, n = 8153 samples) and tumor tissue (TCGA, n = 33 cancers) expressions (Fig. 1). Expression of PRAME in the uterine carcinosarcoma TCGA (n = 57) was significantly higher (p < 0.001) compared to the sarcoma TCGA as a whole, and only skin cutaneous melanoma showed a higher PRAME expression among all normal and tumor tissue types. Furthermore, PRAME expression was significantly higher (p < 0.001) in uterine carcinosarcoma than in normal uterus (n = 83).
PRAME is expressed in sarcoma cell lines
In the CCLE cell line data (n = 46 with microarray, n = 40 with RNA-sequencing data) we found that sarcoma subtypes show diverse PRAME expressions (Fig. 2), however, in the microarray data, all four chondrosarcoma lines had lower expressions than other bone sarcoma cell lines such as Ewing’s sarcoma (p < 0.01) and osteosarcoma (p < 0.1). Analysis of the RNA-sequencing data confirmed chondrosarcoma—Ewing’s sarcoma difference (p < 0.05), and we also observed a trend for overexpression in rhabdomyosarcoma compared to chondrosarcoma (p < 0.1). Notably, with the exception of chondrosarcoma, PRAME over-expressing cell line(s) were found in all CCLE sarcoma types.
PRAME is overexpressed in synovial sarcoma and in multifocal leiomyosarcoma
Analyzing the expression of PRAME in sarcoma subtypes, we found that PRAME was highly expressed in all synovial sarcomas (Fig. 3a). The PRAME expression in these samples was significantly higher (p < 0.001) than in LMS, UPS/MFS, and DDLPS, while LMS expression was significantly lower (p < 0.05) than UPS/MFS, DDLPS, and MPNST PRAME expression (Fig. 3a). Importantly, a few of the LMS, UPS/MFS, DDLPS, and MPNST tumors also showed high PRAME expressions, suggesting that in addition to synovial sarcomas, these subtypes may also be considered for immunotherapies targeting PRAME. Although PRAME median expression was low in LMS, it showed a significantly higher (p < 0.05) expression in multifocal LMS compared to non-multifocal LMS (Fig. 3b), therefore PRAME may be a relevant target in multifocal LMS cases.
PRAME expression negatively correlates with genes involved in antigen presentation
We have determined that PRAME expression was not associated with overall survival in dedifferentiated liposarcoma (Fig. 4a), leiomyosarcoma (Fig. 4b), and UPS/MFS (Fig. 4c), subtypes where a sufficient number of samples were available for Kaplan–Meier analyses.
We also determined whether antigen presentation and other immune-related genes (B2M, CD3E, CD4, CD8A, GZMA, GZMB, HLA-A, HLA-B, HLA-C, IFNG, LCK, PRF1, LMP7, LMP2, TAP1, and TAP2) are associated with PRAME and found that expression of multiple genes involved in antigen presentation (HLA-B, HLA-C, B2M, LMP2, LMP7, TAP2) negatively correlate with PRAME expression in dedifferentiated liposarcoma and leiomyosarcoma (Fig. 4d). PD-L1 and PRAME expression also negatively correlated in dedifferentiated liposarcoma. The other three subtypes did not show significant (p < 0.05) correlations. Furthermore, we found that synovial sarcoma, which overexpresses PRAME, showed a significantly lower B2M and CD8A expressions compared to other subtypes (Fig. 4e). Interestingly, CTAG1B (NY-ESO-1) expression was significantly higher in synovial sarcoma than in the other sarcoma types (Fig. 4e). Furthermore, PD-1 expression was significantly higher (p < 0.05) in synovial sarcoma and UPS/MFS compared to leiomyosarcoma. Interestingly, PD-L1 was expressed in most subtypes except synovial sarcoma, where the expression level was significantly lower than in leiomyosarcoma (p < 0.001), UPS/MFS (p < 0.001), dedifferentiated liposarcoma (p < 0.05), and MPNST (p < 0.05).