Phenotypic and Expressional Heterogeneity in the Invasive Glioma Cells

Abstract

Background: Tumor cell invasion is a hallmark of glioblastoma (GBM) and a major contributing factor for treatment failure, tumor recurrence, and the poor prognosis of GBM. Despite this, our understanding of the molecular machinery that drives invasion is limited.

Methods: Time-lapse imaging of patient-derived GBM cell invasion in a 3D collagen gel matrix, analysis of both the cellular invasive phenotype and single cell invasion pattern with microarray expression profiling.

Results: GBM invasion was maintained in a simplified 3D-milieue. Invasion was promoted by the presence of the tumorsphere graft. In the absence of this, the directed migration of cells subsided. The strength of the pro-invasive repulsive signaling was specific for a given patient-derived culture. In the highly invasive GBM cultures, the majority of cells had a neural progenitor-like phenotype, while the less invasive cultures had a higher diversity in cellular phenotypes. Microarray expression analysis of the non-invasive cells from the tumor core displayed a higher GFAP expression and a signature of genes containing VEGFA, hypoxia and chemo-repulsive signals. Cells of the invasive front expressed higher levels of CTGF, TNFRSF12A and genes involved in cell survival, migration and cell cycle pathways. A mesenchymal gene signature was associated with increased invasion.

Conclusion: The GBM tumorsphere core promoted invasion, and the invasive front was dominated by a phenotypically defined cell population expressing genes regulating traits found in aggressive cancers. The detected cellular heterogeneity and transcriptional differences between the highly invasive and core cells identifies potential targets for manipulation of GBM invasion.

Figure 1

GBM cellular invasion is maintained in a simplified 3D-milieue by the presence of the core. A) Comparison between migration/invasion on 2D-fibronectin coating, 3D-collagen gel and rodent brain slices. The migration pattern of GBM cells in collagen gel grafted as suspension (B), suspension followed by co-grafting of tumorsphere (C), from tumorsphere (D) followed by the resection of the graft core (E). F) Representative GBM cell migration plots before and after graft core resection. BR – before core resection, AR – after core resection. G) Comparison of migratory directionality and invasive increment in T0965 and T1402 before and after core resection (46 to 87 tracked cells per observation).

Figure 2

Glioma cell invasion is cell culture specific. A) Comparison between grafts from 5 different GBM cultures with plots representing the distribution of invasive cells after image-processing. B) Graphical representation of invasive characteristics. C) Time-lapse frames with cell tracks made immediately after grafting (Day 0), by Day 2 and 5 after grafting. D) Comparison of migrative directionality before and after reaching the “invasive limit” (38 tracked cells). E) Representative GBM cell migration plots before and after reaching the “invasive limit”.

Figure 3

The phenotype of invasive cells. A) Comparison of invasive cell phenotype in moderately (T1008) and highly (T1402) invasive GBM grafts with sector diagrams representing the phenotypic composition of the invasive cell pool (100 cells per observation). B) Single cell tracking of GFAP⁺ (n = 9) and nestin⁺/GFAP⁻ (n = 11) invasive cells and post-time-lapse immunostaining for GFAP, nestin and DAPI C) Migratory velocity of GFAP⁺ and nestin⁺/GFAP⁻ invasive cells. D) GFAP expression in core and invasive cells (T1008). E) GFAP expression in invasive cells in five GBM cultures. F) Relation between GFAP expression in invasive cells and total invasive increment.

Figure 4

Grafted core and invasive cells have significantly different gene expression profiles. A) Unsupervised hierarchical clustering of microarray gene expression in core and invasive cells. B) Comparative qPCR relative expression of selected genes in invasive vs. core cells. C) Confirmation of annexin A1 expression in invasive cells by immunostaining. D) Unsupervised hierarchical clustering of microarray gene expression of genes according to Philips et al. . E) Gene expression of lineage specific markers in core and invasive cells, astrocytes (GLI), oligodendrocytes (OLIG), neural stem cells (NSC), neurons (NEU) and mesenchymal cells (MES).

Figure 5

Core cells express genes related to hypoxia and chemo-repulsive signals, while cells in the invasive front express genes involved in cell survival and malignancy. A) A gene signature of core cells is identified in T1008 and T1456. B) The signature is also present in the highly invasive T1402. Since this culture disperses cells until no core is left, making all cells invasive, this signature characterizes a repulsive signal rather than a marker of stationary cells. C) A gene signature of invasive cells is identified in T1008 and T1456. D) The signature shows relationship between signature and invasiveness in four tumor cell cultures. E) Confirmation of CTGF expression in invasive cells by Western blot and immunostaining. F) Confirmation of TNRS12AF expression in invasive cells by immunostaining.

Sup. Figure 1

The invasion persists even in absence of FBS in medium.

Sup. Figure 2

The invasive pattern is culture specific and maintained over time and passages. Seven different primary established GBM cultures were grafted as single tumorspheres in collagen matrix. The pattern of the invasion was specific and completely reproducible for a given patient, regardless of the number of passage. The images depict unique typical pattern of invasion for given patient derived culture. The last right image in the lower row shows the example of a giant cell (marked with the red arrow) found in a graft of T1018 (not a “giant cell GBM”). Type I cell is marked with the yellow arrow for size comparison.

Sup. Figure 3

Post time-lapse immunostaining of grafts for markers nestin, GFAP and β-III tubulin. Tumors T1008 (A) and T1402 (B) present a similar pattern in marker distribution, with the overwhelming majority of cells being positive for nestin and β-III-tubulin both in the invasive front and in the tumor core. In contrast, GFAP staining was strongest in the core, and only a small proportion of the invasive cells were GFAP positive.

Sup. Figure 4

Comparison of GFAP expression in invasive and core cells. GFAP staining was strongest in the core with a small proportion of the GFAP positive invasive cells. qPCR- expression analysis verified the variation of GFAP between core and invasive cells except one tumor T1402.

Sup. Figure 5

Invasive GBM cells preserve proliferative potential. Invasive GBM cells maintain ability to proliferate that was detected by direct recordings of mitosis (A) and immunostaining for marker of proliferation Ki-67 (B). Interestingly, proliferative marker Ki-67 could be co-localized with marker of neuronal differentiation MAP2 in both migrating and mitotic cells in invasive front (C and D correspondingly).

Sup. Figure 6

Invasive GBM cells are positive for the putative cancer stem/progenitor cell markers CXCR4, CD133, CD166, CD44, CD29 and CD9.