PURPOSE: Vibrational spectroscopy enables the label-free characterization of cells and tissue by probing the biochemical composition. Here, we evaluated these techniques to identify glioblastoma stem cells. MATERIALS AND METHODS: The biochemical fingerprints of glioblastoma cells were established in human cell lines with high and low content of CD133 (cluster of differentiation 133)-positive cells using attenuated total reflection Fourier-transform infrared (ATR FT-IR) on vital cells and FT-IR mapping, which delivers spatially resolved spectroscopic datasets. After data preprocessing, unsupervised cluster analysis was applied. CD133 was addressed with flow cytometry and immunohistochemistry and used as a stemness marker. RESULTS: In all preparations, the algorithm was able to correctly classify the spectra, differentiating CD133-rich and -poor populations. The main spectral differences were found in the region of 1000 cm(- 1) to 1150 cm(- 1) that can be assigned to vibrations of chemical bonds of DNA, RNA, carbohydrates and phospholipids. Interestingly, this spectral region is a key feature to discern glioblastoma from normal brain parenchyma, as FT-IR spectroscopic mapping of experimental brain tumors demonstrated. CONCLUSIONS: We were able to show biochemical differences between glioblastoma cell populations with high and low content of cancer stem cells that are presumably related to changes in the RNA/DNA content.
PURPOSE: Vibrational spectroscopy enables the label-free characterization of cells and tissue by probing the biochemical composition. Here, we evaluated these techniques to identify glioblastoma stem cells. MATERIALS AND METHODS: The biochemical fingerprints of glioblastoma cells were established in human cell lines with high and low content of CD133 (cluster of differentiation 133)-positive cells using attenuated total reflection Fourier-transform infrared (ATR FT-IR) on vital cells and FT-IR mapping, which delivers spatially resolved spectroscopic datasets. After data preprocessing, unsupervised cluster analysis was applied. CD133 was addressed with flow cytometry and immunohistochemistry and used as a stemness marker. RESULTS: In all preparations, the algorithm was able to correctly classify the spectra, differentiating CD133-rich and -poor populations. The main spectral differences were found in the region of 1000 cm(- 1) to 1150 cm(- 1) that can be assigned to vibrations of chemical bonds of DNA, RNA, carbohydrates and phospholipids. Interestingly, this spectral region is a key feature to discern glioblastoma from normal brain parenchyma, as FT-IR spectroscopic mapping of experimental brain tumors demonstrated. CONCLUSIONS: We were able to show biochemical differences between glioblastoma cell populations with high and low content of cancer stem cells that are presumably related to changes in the RNA/DNA content.