Whatever marker you use to isolate cancer stem cells, "it should be tumorigenic in mice". Check literature for studies showing FACS sorting using specific markers and shows tumorigenecity assays.
and to add to Goodwin's comment...theoretically whatever is left behind from your sorting (cells negative for those markers) should not be tumourigenic in mice.
Also CSC (cancer stem cells) and TIC (tumour initiating cells) are not necessarily the same thing.
Dilution assay (serial transplantation) to determine glioma stem cells is the useful method. Theoretically, a single CD44+/nestin+/CD133+ tumor cell has a potential to form glioma tissue after orthotopical transplantation. The syngenic transplantation is much better than xenograft model using human glioma cell line.