SOX10-Internal Tandem Duplications and PLAG1 or HMGA2 Fusions Segregate Eccrine-Type and Apocrine-Type Cutaneous Mixed Tumors.

Cutaneous mixed tumors exhibit a wide morphologic diversity and are currently classified into apocrine and eccrine types based on their morphologic differentiation. Some cases of apocrine-type cutaneous mixed tumors (ACMT), namely, hyaline cell-rich apocrine cutaneous mixed tumors (HCR-ACMT) show a prominent or exclusive plasmacytoid myoepithelial component. Although recurrent fusions of PLAG1 have been observed in ACMT, the oncogenic driver of eccrine-type cutaneous mixed tumors (ECMT) is still unknown. The aim of the study was to provide a comprehensive morphologic, immunohistochemical, and molecular characterization of these tumors. Forty-one cases were included in this study: 28 cases of ACMT/HCR-ACMT and 13 cases of ECMT. After morphologic and immunohistochemical characterization, all specimens were analyzed by RNA sequencing. By immunohistochemistry, all cases showed expression of SOX10, but only ACMT/HCR-ACMT showed expression of PLAG1 and HMGA2. RNA sequencing confirmed the presence of recurrent fusion of PLAG1 or HMGA2 in all cases of ACMT/HCR-ACMT, with a perfect correlation with PLAG1/HMGA2 immunohistochemical status, and revealed internal tandem duplications of SOX10 (SOX10-ITD) in all cases of ECMT. Although TRPS1::PLAG1 was the most frequent fusion, HMGA2::WIF1 and HMGA2::NFIB were detected in ACMT cases. Clustering analysis based on gene expression profiling of 110 tumors, including numerous histotypes, showed that ECMT formed a distinct group compared with all other tumors. ACMT, HCR-ACMT, and salivary gland pleomorphic adenoma clustered together, whereas myoepithelioma with fusions of EWSR1, FUS, PBX1, PBX3, POU5F1, and KLF17 formed another cluster. Follow-up showed no evidence of disease in 23 cases across all 3 tumor types. In conclusion, our study demonstrated for the first time SOX10-ITD in ECMT and HMGA2 fusions in ACMT and further refined the prevalence of PLAG1 fusions in ACMT. Clustering analyses revealed the transcriptomic distance between these different tumors, especially in the heterogenous group of myoepitheliomas.