Abstract C80: Characterizing models for personalized medicine: Endosialin in sarcoma from cultured cells to mice to humans

A survey of endosialin by IHC was previously conducted in 250 clinical specimens of cancer, revealing a key difference between sarcomas and carcinomas: in sarcomas, endosialin was expressed in tumor cells, stromal fibroblasts and perivascular cells, while in carcinomas tumor cells were negative. Endosialin expression was frequent among sarcoma patients and reached high levels: 70 of 86 sarcomas (81%) were positive for endosialin, with 44 (51%) displaying at least 50% coverage of all 3 immunoreactive cell types as a whole. Staining intensity was scored on the scale 0, 1+, 2+, 3+: all 9 sarcoma subtypes surveyed included specimens reaching 2+ and 3+. Endosialin emerged from that study as a potential molecular target for sarcomas. However, despite frequent expression and high levels observed in patients, some sarcomas are endosialin-negative and some express low levels. Therefore, targeting endosialin would mandate a personalized medicine approach where patients are tested for endosialin expression in order to determine eligibility for endosialin-directed therapy. We characterized the expression of endosialin in human tumor cell lines in vitro and in vivo, from mRNA and protein in cell culture to protein in xenograft tumors. Thirty-seven human sarcoma cell lines and one neuroblastoma cell line were assayed for endosialin by real-time PCR and flow cytometry. Eighteen cell lines were positive for the protein at various levels. There was a positive correlation between transcript and protein expression, with a few exceptions: 3 cell lines expressed very low levels of the transcript yet clearly detectable protein and one cell line expressed detectable transcript yet no protein. Fifteen of the 18 cell lines positive for the protein in vitro were implanted in mice subcutaneously. Eight of the 15 cell lines implanted failed to form tumors. Seven formed tumor nodules that were excised and tested for endosialin by IHC using the same assay as that used for clinical specimens. Protein expression was not always as robust in vivo as the in vitro expression might have suggested. Staining in the xenograft tumors was scored on the same scale as the clinical specimens. Some models showed robust and homogeneous expression throughout the tumor tissue (SK-N-AS neuroblastoma), while others showed highly heterogeneous staining. In one model representing extreme heterogeneity (SJSA-1 osteosarcoma), endosialin was highly regional, with areas staining positive at the 3+ intensity level and adjacent areas being negative. Both homogeneous and heterogeneous staining was seen in human clinical specimens. Our work demonstrates the importance of characterizing disease models to ascertain target expression at all stages of testing from in vitro to mice and to humans, since expression in vivo cannot be assumed from expression in cell culture and expression patterns in mice may or may not be similar to expression patterns in humans. Citation Information: Mol Cancer Ther 2009;8(12 Suppl):C80.