Extracellular Vesicles Carrying Tenascin-C are Clinical Biomarkers and Improve Tumor-Derived DNA Analysis in Glioblastoma Patients

Abstract

Extracellular vesicles (EVs) act as carriers of biological information from tumors to the bloodstream, enabling the detection of circulating tumor material and tracking of disease progression. This is particularly crucial in glioblastoma, a highly aggressive and heterogeneous tumor that is challenging to monitor. Using imaging flow cytometry (IFCM), we conducted an immunophenotyping analysis of eight glioma-associated antigens and tetraspanins in plasma EVs from 37 newly diagnosed glioblastoma patients (pre- and post-surgery), 11 matched individuals with recurrent glioblastoma, and 22 healthy donors (HD). Tenascin-C (TNC) positive EVs displayed the strongest differences in newly diagnosed and recurrent glioblastoma patients, when compared to non-tumor subjects. Among dual-positive subpopulations, TNC⁺/CD9⁺ EVs were the most elevated in newly diagnosed (FC = 7.6, p <0.0001, AUC = 81%) and recurrent patients (FC = 16.5, p <0.0001; AUC = 90%) than HD. In comparison with other CNS tumors (n = 25), this subpopulation was also 34.5-fold higher in glioblastoma than in meningioma cases (p <0.01). Additionally, TNC⁺/CD9⁺ EV levels were 3.3-fold elevated in cerebrospinal fluid from glioblastoma patients (n = 6) than controls (p <0.05). Aberrant TNC levels were further observed in glioblastoma EVs from different sources and purified via different methods. Immunohistochemical analysis revealed high levels of TNC in tumor tissues. Spatial transcriptomic analysis indicated a TNC overexpression in malignant cell populations of glioblastoma resections, particularly in cells with mesenchymal-like signatures and chromosomal aberrations. Lastly, we purified TNC⁺ EVs from plasma of 21 glioblastoma patients by magnetic sorting and detected the oncogenic mutation TERT*C228T by droplet digital PCR. The mutant allele frequency was higher in TNC⁺ EVs vs TNC-negative EVs (FC = 32, p <0.001), total EVs (FC = 5.3, p <0.001) or cell-free DNA (FC = 5.3, p <0.01). In conclusion, circulating TNC⁺ EVs may have potential as clinical biomarkers in glioblastoma, and their purification could improve the identification of tumor-specific mutations in liquid biopsies.

Keywords: Tenascin-C; biomarkers; extracellular vesicles; glioblastoma; liquid biopsy.

Figure 1

Plasma cohort and EV phenotyping overview. (A) EV phenotyping workflow showing cohort description and strategy of antigen selection for biomarker investigation by Imaging flow cytometry (IFCM). The candidate antigens TNC, ITGB1, CD44, GPNMB, SPARC, PFN1, HLA-DR/DQ/DP and CD133 (previously highlighted by our group), as well as classical EV markers (the tetraspanins CD81, CD63 and CD9) were selected for investigation, where the levels per milliliter of plasma of double-positive EVs (for each candidate marker and each tetraspanin) were compared among the groups. Figure made in BioRender.com. (B) Heatmap with significant fold changes (FC) when comparing the median levels of double positive EV subpopulations between groups (healthy donors (HD), newly diagnosed glioblastoma patients (GBM), postoperative (post-OP), and GBM relapses). Their main statistical analyses are shown in Figure 2. Nonsignificant results are plotted with FC equivalent to zero. TNC⁺ EVs (independently of the tetraspanins) are the most significantly increased in glioblastoma patients (newly diagnosed and relapses), in comparison to HD and post-OP subjects.

Figure 2

EV phenotyping analyses among the groups. (A–H) Plasma levels of EVs (positive at least for one of the tetraspanins) containing each of the selected candidate markers for investigation in HD (green), glioblastoma (GBM, red), post-OP (blue), and relapses (orange). Medians were compared between groups by Kruskal–Wallis analysis with correction for false discovery rate (FDR). Their respective FC values are shown in Figure 1B. The TNC⁺ EVs (A) were differentially expressed in most of compared groups and had the strongest significances (higher FC and lower p values), followed by ITGB1 (B). (I) Immunogold staining for TNC protein in plasma EVs of a glioblastoma patient (left) and a healthy donor (right). Obvious differences are observed between two immunostainings, where gold particles (black dots) can be visualized in plasma EVs from glioblastoma (white arrows). * = p < 0.05, ** = p < 0.01, *** = p < 0.001, **** = p < 0.0001. Nonsignificant associations are not specifically marked.

Figure 3

Differential levels of TNC⁺/CD9⁺ EVs in glioblastoma. (A) TNC⁺/CD9⁺ EVs are elevated in plasma of newly diagnosed (FC = 7.6) and recurrent (FC = 16.5) glioblastoma patients, in comparison to HD subjects. (B) A 3.9-fold drop is observed in a paired way for TNC⁺/CD9⁺ EVs after tumor removal, which (C) reincreased (FC = 8.4) in these same patients under tumor relapse. (D) ROC graph of TNC⁺/CD9⁺ EV levels, with area under curve (AUC) greater than 80% for discrimination between healthy donors and glioblastoma patients (newly diagnosed and recurrent together). (E) ROC curve of TNC⁺/CD9⁺ EVs, comparing HD and newly diagnosed glioblastoma, also with AUC greater than 80%. (F) ROC curve comparing HD and recurrent GBM, with the highest indicative results for TNC⁺/CD9⁺ EVs as clinical biomarkers, with AUC values greater than 91% (threshold 74,500 EVs for 100% specificity and 63% sensitivity). (G) ROC curve of TNC⁺/CD9⁺ EVs, with AUC of 84% for discrimination of post-OP subjects and recurrent glioblastoma. (H) TNC⁺/CD9⁺ EV levels in glioblastoma patients belonging to different methylation subgroups. Patients classified as RTK-I present 3.4-fold lower levels of TNC⁺/CD9⁺ EVs in plasma, in relation to patients with RTK-II subtype. * = p < 0.05, ** = p < 0.01, *** = p < 0.001, **** = p < 0.0001.

Figure 4

Differential levels of TNC⁺ and TNC⁺/CD9⁺ EVs in glioblastoma under different experimental conditions. (A) TNC⁺ EVs in comparison with other CNS malignancies (medulloblastoma, brain metastasis from NSCLC and from melanoma, and meningioma). Glioblastoma remains with the highest levels of TNC⁺ EVs per mL of plasma, also being 27.2- and 11.3-fold elevated than meningioma and brain metastatic melanoma, respectively. TNC⁺ EVs from medulloblastoma cases were also significantly higher than meningioma (FC = 9.7). (B) TNC⁺/CD9⁺ EVs in CNS malignancies. Glioblastoma remains with the highest levels of TNC⁺/CD9⁺ EVs per mL of plasma, also significantly elevated than meningioma cases (FC = 34.5). (C) TNC levels per mL of plasma on EVs isolated by size exclusion chromatography (SEC), which remained significantly increased in glioblastoma patients than in HD subjects (FC = 6.52). (D) TNC⁺/CD9⁺ levels per mL of plasma on EVs isolated by SEC, which remained significantly increased in glioblastoma patients than in HD subjects (FC = 12.05). (E) TNC⁺/CD9⁺ EVs were also significantly higher in cerebrospinal fluid (CSF) from glioblastoma patients, when compared to CSF from controls (suffering from normal pressure hydrocephaly) (FC = 3.3). * = p < 0.05, ** = p < 0.01.

Figure 5

TNC mRNA expression in glioblastoma tissue by spatial transcriptomic analysis. Expression of investigated markers in cell populations of glioblastoma tissue, as classified by spatial transcriptomics analysis. (A) Leveraging the glioblastoma reference single cell data set (GBMap) to explose distinct transcriptional signatures within the tumor tissue, of which the genuine malignant area is represented by glioblastoma subtypes classified as neural progenitor-like (NPC), mesenchymal-like (MES), astrocyte cell-like (AC) and oligodendrocyte progenitor cell-like (OPC). (B) RNA expression of TNC and ITGB1 markers within the tumor tissue. TNC expression mainly corresponds to the glioblastoma malignant signature areas, in contrast to ITGB1. (C) Also in contrast to TNC, and similarly to the other evaluated antigens, the tetraspanins (CD63, CD9 and CD81) are nonspecifically expressed throughout the tumor tissue. (D, E) As visualized by H&E staining, TNC is overexpressed in the malignant area of the tumor tissue. (F) Following the trajectory direction from the malignant to the peritumoral zone, TNC and most of the investigated markers are exclusively overexpressed in glioblastoma cells, which also colocalize with oncogenic CNVs, such as gain in chromosomes 7 and loss in chromosome 10.

Figure 6

Tumor-derived DNA analysis in TNC⁺ and TNC^– EV fractions. (A) Paired analysis of the TERT*C228T tumor mutation frequencies in gDNA and TNC-sorted EV-DNA samples obtained from glioblastoma-tissue resections (n = 7), by droplet digital PCR (ddPCR). The samples are from the cohort of tissue-EVs previously analyzed by IFCM (see Supporting Information Figure S3) and are from tumors carrying the TERT mutation. TNC^– EVs (green) seem to carry lower or similar mutation frequencies than genomic DNA (gray) (nonsignificant differences), in addition to and 1.22-fold lower TERT*C228T frequencies than TNC⁺ EVs (red) (p < 0.05), when analyzed by Student t-test. Tissue-derived TNC⁺ EVs present significant higher frequencies of mutation amplicons (8.3% more abundant, in average) than their paired TNC^– fraction. (B) TERT*C228T frequencies in plasma samples. Left panel: unpaired analysis (Kruskal–Wallis with FDR correction), where circulating TNC⁺ EVs are shown with significantly higher frequencies of tumor mutation than EVs lacking TNC (FC = 32), total EVs (FC = 5.3) and cfDNA (FC = 5.3). Right panel: paired analysis of the same cohort (Wilcoxon test). Differences remained significant in paired comparisons, in addition to a 6-fold decrease in TNC- EVs than total EVs. (C) Examples of ddPCR plots for analysis of TERT*C228T tumor mutation in TNC^– EVs, TNC⁺ EVs, total EVs and bulk cfDNA from plasma samples. Droplets carrying mutation amplicons are shown in blue (upper plots) whereas wild-type copies in green (green droplets, below). The analyzed TERT mutation is mainly observed in DNA of TNC⁺ EVs. On the right side, a similar analysis was performed with HD samples to show the absence of unspecific amplifications. # = p < 0.08, * = p < 0.05, ** = p < 0.01, *** = p < 0.001, **** = p < 0.0001.