BACKGROUND: Experimental data suggest that glioblastoma cells expressing the stem cell marker CD133 play a major role in radiochemoresistance and tumor aggressiveness. To date, however, there is no clinical evidence that the fraction of CD133-positive cells in glioblastoma that recurs after radiochemotherapy may be relevant for prognosis. METHODS: The authors used immunohistochemistry to assess CD133 expression in 37 paired glioblastoma samples, including 1 primary tumor sample and 1 recurrent tumor sample, after patients received adjuvant radiochemotherapy. To assess the actual composition of the CD133-positive glioblastoma cell population, fluorescence-associated cell sorting (FACS) analysis was used to sort CD133-positive/CD45-negative cells that were assayed for tumor-specific chromosomal aberrations using interphase fluorescence in situ hybridization. To rule out endothelial precursor cells, CD133-positive fractions also were assayed with anti-CD34 by FACS. RESULTS: In recurrent glioblastomas, the percentage of CD133-positive cells was increased by 4.6-fold compared with the percentage in primary glioblastomas, although, in some tumors, it increased up to 10-fold and 20-fold. Unexpectedly, the increase in CD133 expression was associated significantly with longer survival after tumor recurrence. An analysis of tumor-specific chromosomal aberrations and in vivo studies revealed that the CD133-positive cell compartment of recurrent glioblastoma was composed of both cancer stem cells and nontumor neural stem cells. The latter cells represented from 20% to 60% of the CD133-positive cell population, and their relative percentage favorably affected the survival of patients with recurrent glioblastoma. Endothelial CD133-positive/CD34-positive precursors did not contribute to the CD133-positive cell population. CONCLUSIONS: The authors hypothesized that, similar to the phenomenon described in glioblastoma models, neural stem/progenitor cells that are recruited by the tumor from surrounding brain may exert an antitumorigenic effect. Cancer 2011;117:162-74.