Transdifferentiation of metastatic melanoma into cerebellar angiosarcoma

This is the authors' abstract. We don't add key points for this paper.

To the Editor: A 68-year-old male patient presented with intractable vomiting and headache. Magnetic resonance imaging demonstrated a heterogeneously enhancing 2.8 cm mass of the right cerebellum with edema and mass effect, including compression of the fourth ventricle (not shown). Past medical history included diabetes mellitus type 2, hyperlipidemia, hypertension, and hypothyroidism. Past surgical history included a cutaneous melanoma on his back that, despite immunotherapy, metastasized to the left external iliac node, lungs, and liver. After the metastases, the patient was switched to encorafenib (mitogen-activated protein kinase [MAPK] inhibitor) and binimetinib (MEK inhibitor), and had been on those treatments for approximately 1 year. The patient underwent neurosurgical resection of the right cerebellar mass. Histopathology showed a malignant neoplasm forming vascular channels composed of large, atypical cells with amphophilic cytoplasm and prominent nucleoli (Figure 1A). The tumor cells were immunopositive for the vascular markers CD31, CD34, ERG, and FLI-1, were variably positive for S100, and were negative for Inhibin, GFAP, Olig2, HMB45, Melan A, Sox10, and cytokeratin AE1/AE3 (Figure 1B-J). Histopathology of the cerebellar, cutaneous, and lymph node tumors. (A-J) H&E and key immunostains from the cerebellar lesion. (K-O) The original cutaneous melanoma. (P-T) The iliac lymph node mass. Scale bar in (T) = 50 µm for all panels. Next-generation sequencing (NGS) detected pathologic mutations in BRAF p. V600E and PTPRT p. R936*, along with variants of unknown significance (VUS) in CARD11, EPHA7, FGF23, PIK3C2B, PREX2, PRSS1, SPTA1, TET1, and MAF. Genomic DNA methylation profiling using the sarcoma classifier indicated angiosarcoma (score = 0.87) (Table 1). Summary of histologic, immunophenotypic, and molecular characteristics of all three tumors. Sox10 HMB45 (weak) Melan A Sox10 ERG CD31 CD34 BRAF p. V600E PTPRT p. R936* VUS in CARD11, EPHA7, FGF23, MAF, PIK3C2B, PREX2, PRSS1, SPTA1, and TET1 CK AE1/AE3 CK 8/18 Sox10 HMB45 (weak) Melan A (weak) S100 (weak) CDX2 NKX3.1 PSA BRAF p. V600E PTPRT p. R936* VUS in CARD11, EPHA7, FGF23, MAF, PREX2, PRSS1, SPTA1, TET1, and KEL CD31 CD34 ERG FLI-1 S100 (variable) Inhibin Olig2 HMB45 Melan A Sox10 CK AE1/AE3 GFAP (positive in admixed non-neoplastic brain tissue) BRAF p. V600E PTPRT p. R936* VUS in CARD11, EPHA7, FGF23, MAF, PIK3C2B, PREX2, PRSS1, SPTA1, and TET1 Sox10 HMB45 (weak) Melan A Sox10 ERG CD31 CD34 BRAF p. V600E PTPRT p. R936* VUS in CARD11, EPHA7, FGF23, MAF, PIK3C2B, PREX2, PRSS1, SPTA1, and TET1 CK AE1/AE3 CK 8/18 Sox10 HMB45 (weak) Melan A (weak) S100 (weak) CDX2 NKX3.1 PSA BRAF p. V600E PTPRT p. R936* VUS in CARD11, EPHA7, FGF23, MAF, PREX2, PRSS1, SPTA1, TET1, and KEL CD31 CD34 ERG FLI-1 S100 (variable) Inhibin Olig2 HMB45 Melan A Sox10 CK AE1/AE3 GFAP (positive in admixed non-neoplastic brain tissue) BRAF p. V600E PTPRT p. R936* VUS in CARD11, EPHA7, FGF23, MAF, PIK3C2B, PREX2, PRSS1, SPTA1, and TET1 Abbreviations: IHC = immunohistochemistry; NGS = next-generation sequencing; VUS = variants of unknown significance. Summary of histologic, immunophenotypic, and molecular characteristics of all three tumors. Sox10 HMB45 (weak) Melan A Sox10 ERG CD31 CD34 BRAF p. V600E PTPRT p. R936* VUS in CARD11, EPHA7, FGF23, MAF, PIK3C2B, PREX2, PRSS1, SPTA1, and TET1 CK AE1/AE3 CK 8/18 Sox10 HMB45 (weak) Melan A (weak) S100 (weak) CDX2 NKX3.1 PSA BRAF p. V600E PTPRT p. R936* VUS in CARD11, EPHA7, FGF23, MAF, PREX2, PRSS1, SPTA1, TET1, and KEL CD31 CD34 ERG FLI-1 S100 (variable) Inhibin Olig2 HMB45 Melan A Sox10 CK AE1/AE3 GFAP (positive in admixed non-neoplastic brain tissue) BRAF p. V600E PTPRT p. R936* VUS in CARD11, EPHA7, FGF23, MAF, PIK3C2B, PREX2, PRSS1, SPTA1, and TET1 Sox10 HMB45 (weak) Melan A Sox10 ERG CD31 CD34 BRAF p. V600E PTPRT p. R936* VUS in CARD11, EPHA7, FGF23, MAF, PIK3C2B, PREX2, PRSS1, SPTA1, and TET1 CK AE1/AE3 CK 8/18 Sox10 HMB45 (weak) Melan A (weak) S100 (weak) CDX2 NKX3.1 PSA BRAF p. V600E PTPRT p. R936* VUS in CARD11, EPHA7, FGF23, MAF, PREX2, PRSS1, SPTA1, TET1, and KEL CD31 CD34 ERG FLI-1 S100 (variable) Inhibin Olig2 HMB45 Melan A Sox10 CK AE1/AE3 GFAP (positive in admixed non-neoplastic brain tissue) BRAF p. V600E PTPRT p. R936* VUS in CARD11, EPHA7, FGF23, MAF, PIK3C2B, PREX2, PRSS1, SPTA1, and TET1 Abbreviations: IHC = immunohistochemistry; NGS = next-generation sequencing; VUS = variants of unknown significance. Because the patient had a history of widely metastatic melanoma, materials from the original cutaneous melanoma and the left external iliac lymph node were analyzed. The cutaneous melanoma was weakly positive for HMB45 and strongly positive for Melan A, Sox10, and S100 (Figure 1K-O). Compared to the cutaneous tumor, the lymph node tumor was weakly positive for HMB45, retained Sox10, and expressed less Melan A and S100 (Figure 1P-T). The lymph node metastasis also expressed some cytokeratin AE1/3 (Table 1). NGS detected BRAF p. V600E and PTPRT p. R936* in tumor from all 3 sites, as well as the exact same VUS in CARD11, EPHA7, FGF23, MAF, PREX2, PRSS1, SPTA1, and TET1 (Table 1). The only differences were that a VUS in PIK3C2B was detected in the cutaneous and cerebellar lesions but not in the lymph node, and a VUS in KEL was detected in the lymph node but not in the other 2 sites. The primary cutaneous melanoma matched to “melanoma” on the methylation classifier (score = 1.0); not enough tissue from the lymph node remained for methylation profiling (Table 1). Based on sequencing similarities, the final integrated diagnosis for the cerebellar mass was therefore metastatic melanoma with angiosarcomatous transdifferentiation. Seven months post-operatively, there was no evidence of cerebellar disease recurrence on surveillance imaging. This is only the second reported case of complete transdifferentiation of metastatic melanoma to angiosarcoma. In the first case, the angiosarcomatous component was identified in a lymph node metastasis 1 year after the primary resection.1 That angiosarcomatous component developed alongside a conventional component and showed a complete lack of immunoreactivity for the melanoma-lineage markers, MART-1, S100 protein, and Sox10, while expressing the vascular markers ERG and CD31. In contrast, ERG and CD31 were negative in the conventional melanoma component. The common origin of both morphologies was suggested by their having the exact same NRAS p. Q61R mutation. Methylation profiling was not done on that case.1 Our case showed only angiosarcomatous differentiation in the cerebellar metastasis, without morphologic or immunophenotypic evidence of conventional melanoma. There were also no angiosarcomatous features identified in the primary dermal resection or left external iliac lymph node biopsy. A common origin was again supported by shared pathologic mutations, this time involving BRAF and PTPRT, as well as shared VUS. Remarkably, the tumor also completely changed its DNA methylation profile from a melanoma to an angiosarcoma. An angiomatoid variant of cutaneous melanoma has been described and can mimic angiosarcoma. However, angiomatoid melanoma still retains expression of melanocyte markers, including S100 protein, Melan A, HMB-45, and MiTF-1,2 and is negative for vascular markers such as CD31. Melanoma is typically characterized by mutations in the MAPK pathway (BRAF p. V600E being the most common), PI3K-AKT pathway, genes involved in cell cycle regulation (eg, CDKN2A), WNT pathway, MITF pathway, and the NFκB pathway.3 Many angiosarcomas also contain MAPK-activating alterations, including alterations involving BRAF.4 This suggests that a BRAF-driven melanoma may be predisposed toward transdifferentiating into another BRAF-driven tumor. As concerns what may have triggered transdifferentiation in this case, the cerebellar microenvironment may have prompted an epigenomic shift, in turn altering morphology and immunophenotype. There is precedence for such effects by microenvironments on tumor cell epigenetics.5,6 On the other hand, therapies targeting lineage-specific pathways like MAPK/ERK may trigger cancer transdifferentiation as an adaptive mechanism. For example, neuroendocrine differentiation can occur in response to anti-EGFR or anti-androgen therapies in lung and prostate cancer, respectively.7,8 Immunotherapy has also been shown to cause epigenomic shifts and transdifferentiation in hematopoietic neoplasms.9–11 Since targeted therapies and immunotherapies are becoming more widespread, transdifferentiation may also become more commonplace. In sum, transdifferentiation of cancers is currently rare, but may become more common in response to new therapies. The odds of correctly identifying transdifferentiation increases with access to detailed medical history, information on treatments, careful examination of resection specimens, the ability to perform advanced molecular testing, and awareness of this rare and remarkable phenomenon. The presented work did not receive external or internal funding. The authors have no relevant conflicts of interest to disclose.