1. ¹Department of Neurosurgery, Samsung Medical Center and Samsung Biomedical Research Institute, Sungkyunkwan University School of Medicine, Seoul, South Korea
2. ²Cancer Stem Cell Research Center, Samsung Medical Center and Samsung Biomedical Research Institute, Sungkyunkwan University School of Medicine, Seoul, South Korea
3. ³School of Life Sciences and Biotechnology, Korea University, Seoul, South Korea
4. ⁴Department of Pathology, Samsung Medical Center and Samsung Biomedical Research Institute, Sungkyunkwan University School of Medicine, Seoul, South Korea
5. ⁵Department of Biochemistry and Molecular Biology, Seoul National University College of Medicine, Seoul, South Korea
6. ⁶Department of Anatomy, Seoul National University College of Medicine, Seoul, South Korea

Correspondence: Dr Do-H Nam, MD, PhD, Department of Neurosurgery, Samsung Medical Center and Biomedical Research Institute, Sungkyunkwan University School of Medicine #50, Irwon-dong, Gangnam-gu, Seoul 135-710, South Korea. E-mail: nsnam@skku.edu; Dr H Kim, PhD, Cell Growth Regulation Laboratory, School of Life Sciences and Biotechnology, Korea University, Anam-dong, Seongbuk-gu, Seoul 136-713, South Korea. E-mail: hg-kim@korea.ac.kr

⁷These authors contributed equally to this work.

Received 9 March 2008; Revised 3 April 2008; Accepted 3 April 2008; Published online 16 June 2008.

Abstract

A number of recent reports have demonstrated that only CD133-positive cancer cells of glioblastoma multiforme (GBM) have tumor-initiating potential. These findings raise an attractive hypothesis that GBMs can be cured by eradicating CD133-positive cancer stem cells (CSCs), which are a small portion of GBM cells. However, as GBMs are known to possess various genetic alterations, GBMs might harbor heterogeneous CSCs with different genetic alterations. Here, we compared the clinical characteristics of two GBM patient groups divided according to CD133-positive cell ratios. The CD133-low GBMs showed more invasive growth and gene expression profiles characteristic of mesenchymal or proliferative subtypes, whereas the CD133-high GBMs showed features of cortical and well-demarcated tumors and gene expressions typical of proneuronal subtype. Both CD133-positive and CD133-negative cells purified from four out of six GBM patients produced typical GBM tumor masses in NOD-SCID brains, whereas brain mass from CD133-negative cells showed more proliferative and angiogenic features compared to that from CD133-positive cells. Our results suggest, in contrast to previous reports that only CD133-positive cells of GBMs can initiate tumor formation in vivo CD133-negative cells also possess tumor-initiating potential, which is indicative of complexity in the identification of cancer cells for therapeutic targeting.