Glioblastoma (GBM) is the most aggressive brain malignancy. Although some potential glioblastoma biomarkers have already been identified, there is a lack of cell membrane-bound biomarkers capable of distinguishing brain tissue from glioblastoma and/or glioblastoma stem cells (GSC), which are responsible for the rapid post-operative tumor reoccurrence.
Our aim was to find new candidates for glioblastoma or glioblastoma stem cell markers. Our startig point were the candidates CD9, FTL, S100A9, beta-actin and TRIM28 (starting point candidates – SPC). These candidates have ben recenty proposed as novel GSC biomarkers with elevated expression in GSC. We conducted a literature search in order to find all malignant pathologies in which SPC likewise have elevated expression – as in GBM. In each such pathology, we identified interaction partners for our SPC. For each pair SPC-interaction partner an interaction network was created at the gene and protein levels of expression, using the databases Biomine Explorer (gene level) and String (protein level). In each of these networks we identified common nodes between SPC and its interaction partner, and these nodes represented potential novel GBM/GSC marker candidates. For each candidate represented in the common nodes, a check was made whether it had already been linked to GBM carcinogenesis. Only candidates with no such links were bioinformatically validated by comparing their expression in various GBM/GSC-relevant cell/tissue types with that of established GBM/GSC marker candidates. This validation consisted of meta-analysis of genome-scale mRNA expression data from three data repositories (GEO, ArrayExpress and GLIOMASdb). The search yielded ten appropriate datasets, which were used for validation of the candidates from the common nodes.
16 interaction partners that were linked to our SPC in 10 different malignant pathologies were identified. Analysis of all possible networks SPC-interaction partner yielded 82 common nodes, of which only 23 had not yet been linked to GBM carcinogenesis. Bioinformatic validation of these candidates revealed three potential new GBM/GSC marker candidates with elevated expression in malignant cells (CCT2, RUVBL1, BST1). However, detailed results analysis highlighted that all these three candidates have uneven expression in various GBM/GSC-relevant cell/tissue types (e.g., elevated expression in GBM tissue samples, but not in stem-like cell lines and neurospheres).
Thus, we decided to identify candidates that achieved the best results in the bioinformatic validation tests and had the most consistent expression pattern in all relevant cell/tissue types. In this stage the search was focused on candidates with elevated expression at the protein level on the tumor cells surface. Three datasets (GSE4290/GDS1962, GSE23806/GDS3885, and GLIOMASdb) were used for selection of new GBM/GSC marker candidates, while the other seven (GSE4412/GDS1975, GSE4412/GDS1976, E-GEOD-52009, E-GEOD-68848, E-GEOD-16011, E-GEOD-4536, and E-GEOD-74571) were used for bioinformatic validation of the selected candidates.
The selection identified four new CSP-encoding candidate genes—CD276, FREM2, SPRY1, and SLC47A1—and the bioinformatic validation confirmed their elevated expression in GBM/GSC. A review of the literature revealed that CD276 is not a novel candidate, while SLC47A1 had lower validation test scores than the other new candidates and was therefore not considered for experimental validation.
This validation revealed that the expression of FREM2—but not SPRY1—is higher in glioblastoma cell lines when compared to non-malignant astrocytes. In addition, FREM2 gene and protein expression levels are higher in glioblastoma stem-like cell lines than in conventional glioblastoma cell lines. FREM2 is thus proposed as a novel GBM marker candidate, as well as a putative candidate for GSC biomarker. Both FREM2 and SPRY1 are expressed on the surface of the GBM cells, while SPRY1 alone was found over-expressed in the cytosol of non-malignant astrocytes. SPRY could also be useful as a therapeutic target since it is expressed on the surface of malignant GBM cells but in the cytosol of non-malignant astrocytes.