Introduction: Glioblastoma is one of the most aggressive cancers known to men. Non-invasive assessment of aggressiveness is crucial for treatment planning, but current MRI protocols lack specificity. Amide proton transfer CEST MRI can grade diffuse gliomas, but not GBM aggression levels. GBM invasiveness arises from a shift from a pro-neural to mesenchymal phenotype. Based on a report that mannose-weighted (MANw) CEST MRI can be used for detection of unlabeled mesenchymal stem cells overexpressing mannose [1], we investigated if mesenchymal cancer stem cells could be detected “label-free” in a similar fashion.

Materials and Methods: Low aggressive GBM1a and highly aggressive M1123 cells were used throughout. Mannose expression was assessed using fluorescein-labeled galanthus nivalis lectin (GNL-FITC, specific for mannose) staining, and the mesenchymal cellular phenotype by anti-CD44 immunostaining. In vitro MANw CEST MRI was conducted using a Bruker 11.7T vertical bore spectrometer. Z-spectral data were collected using a CW RF pulse of B₁=2.4 μT, and T_sat=3 s. For in vivo tumor models, 2×10⁵ M1123 and GBM1a spheres were injected into the right and left striatum of NSG mice brain. In vivo T2-w and MANw CEST MRI was performed 1, 8 and 16 days after injection using an 11.7 T Bruker Biospin horizontal bore scanner. CEST MRI was performed with a saturation pulse B₁=2.4 µT and T_sat=3 s, with the saturation frequency step between ±5 ppm. Tumor and brain ROIs were manually drawn based on T2-w images. For M1123 cells, the mannose-binding lectins LMAN1 and LMAN2 were knocked down using liposomal transfection with LMAN 1/2 siRNA, and LMAN1/2 expression was quantified with qRT-PCR and normalized to GAPDH.

Results and Discussion: GNL staining indicated low mannose expression for both 2D cell cultures, but M1123 3D sphere cultures contained more mannose (MCSCs) compared to GBM1a (Fig. 1A). In vitro MANw CEST MRI showed the highest CEST signal for the M1123 3D spheres (Fig. 1B). After transplantation of 3D tumor spheres in mouse brain. T2-w MRI showed M1123 cells growing much faster than GBM1a invading across the entire hemisphere on day 16, with pronounced hypointense regions due to necrosis and hemorrhage (Fig. 1C). For the MANw CEST MRI on day 1, a distinct signal was observed for M1123, whereas GBM1a signal levels were not distinguishable from normal brain tissue background. Eight and 16-day post-injection follow-up revealed a continuous pronounced MANw CEST signal only for the highly aggressive tumor (Fig. 1D), with an excellent correlation for anti-mannose staining (Fig. 1E). The difference in MANw CEST signal of M1123 was significantly higher (>1.8-fold) than GBM1a and host brain for all three time points (Fig. 1F). Anti-CD44 immunostaining revealed an abundance of mesenchymal cells in the M1123 tumor, but none in the GBM1a. Moreover, the highest density was found in the periphery of the M1223 tumor mass, representing the most malignant, outwards growing mesenchymal cancer stem cells. Silencing LMAN1/2 in M1223 cells resulted in a 4-fold reduction of LMAN1/2 expression and mannose glycosylation, which was accompanied by a ~10% reduction in MANw CEST MRI contrast.

Conclusions: Highly aggressive GBM cancer stem cells overexpress mannose and exhibit a mesenchymal (CD44⁺) phenotype. Knocking down mannose binding lectins resulted in lower mannose expression and a concomitant lower MANw CEST contrast, proving causation of CEST MRI signal and cancer stem cell phenotype instead of a mere correlation. MANw CEST MRI may allow a label-free, non-invasive differentiation of GBM aggressiveness. Once translated, this advancement may decrease the time interval between diagnosis and treatment, hopefully increasing patient survival. Since brain tumor patients already undergo routine MRI, translational approval can be fast, only needing local IRB approval.

Image/Figure:

Click to view full size

Image/Figure Caption:

Figure 1. (A) In vitro mannose N-linked glycan expression (green) of GBM1a and M1123 2D cell and 3D tumor sphere cultures. (B) MANw CEST MR images of 2D cell and 3D tumor sphere phantoms. Color bar in (B) represents data acquired at 1.2 ppm. (C) T2-weighted MRI and (D) MANw CEST MRI of GBM1a (left) and M1123 (right) tumor spheres xenografted in immunodeficient NSG mice obtained at 1, 8 and 16 days post implantation. Color bar in (B) represents data acquired at 1.2 ppm. (E)  Correlating GNL-FITC staining for negative GBM1a (left, circles) and positive M1123 (right) implanted tumor spheres shown in panels (C,D). (F) Summary of MTR_asym values at 1.2 ppm obtained at day 1, 8 and 16. All data are presented as mean±SD, n=3. P-values <0.001 (***), 0.001-0.01 (**) and 0.01-0.05 (*) were considered statistically significant.

Author