Fibronectin matrix assembly suppresses dispersal of glioblastoma cells

Abstract

Glioblastoma (GBM), the most aggressive and most common form of primary brain tumor, has a median survival of 12-15 months. Surgical excision, radiation and chemotherapy are rarely curative since tumor cells broadly disperse within the brain. Preventing dispersal could be of therapeutic benefit. Previous studies have reported that increased cell-cell cohesion can markedly reduce invasion by discouraging cell detachment from the tumor mass. We have previously reported that α5β1 integrin-fibronectin interaction is a powerful mediator of indirect cell-cell cohesion and that the process of fibronectin matrix assembly (FNMA) is crucial to establishing strong bonds between cells in 3D tumor-like spheroids. Here, we explore a potential role for FNMA in preventing dispersal of GBM cells from a tumor-like mass. Using a series of GBM-derived cell lines we developed an in vitro assay to measure the dispersal velocity of aggregates on a solid substrate. Despite their similar pathologic grade, aggregates from these lines spread at markedly different rates. Spreading velocity is inversely proportional to capacity for FNMA and restoring FNMA in GBM cells markedly reduces spreading velocity by keeping cells more connected. Blocking FNMA using the 70 KDa fibronectin fragment in FNMA-restored cells rescues spreading velocity, establishing a functional role for FNMA in mediating dispersal. Collectively, the data support a functional causation between restoration of FNMA and decreased dispersal velocity. This is a first demonstration that FNMA can play a suppressive role in GBM dispersal.

Figure 1. Time-dependent dispersal of aggregates of U87-MG cells.

A) Aggregates of U87-MG cells approximately 50 µm in diameter, plated onto tissue culture plastic and incubated in complete medium under standard tissue culture conditions, spread to a monolayer within 8 hours. This dispersal can be quantified as spreading velocity by calculating the slope of the line representing the change in normalized aggregate diameter over time. A time-lapse movie showing aggregate dispersal can be viewed in Movie S1. B) Linear regression analysis of aggregate spreading. The relationship between normalized aggregate diameter and time was found to be linear for aggregates of all three-cell lines. The red line represents aggregates of U87-MG cells, whereas the blue and green lines represent LN-229 and U118-MG, respectively. Ten to 15 aggregates were used to calculate the mean normalized diameters at each time-point. Standard deviation bars are depicted.

Figure 2. Aggregate dispersal velocity of 3-glioblastoma cell lines.

This bar graph depicts the spreading velocities of aggregates of U87-MG, LN-229, and U118-MG cells. Data sets included sample sizes of 14, 10, and 11 aggregates, respectively. Aggregate spreading was monitored over an 8-hour period. Movies of the spreading were captured and images at various time points were analyzed. Normalized aggregate diameter was plotted as a function of time, generating a slope, which was used to calculate spreading velocity. As can be seen here, the spreading velocity of aggregates of U87-MG cells is considerably faster than that of either LN-229 or U118-MG (ANOVA, p<0.0001 and Tukey's MCT).

Figure 3. Assessment of pan-cadherin expression by glioma cells.

A) Overall cadherin expression by the three cell lines was assessed by immunoblot analysis using a pan-cadherin antibody (clone CH-19, Sigma). Bands, corresponding to the molecular weight of cadherin, are evident. Actin was used as a loading control. B) Quantification of cadherin expression by densitometric analysis. Optical density (OD) of cadherin bands was measured in triplicate as described in the Methods section. These measurements were then normalized by expressing OD of the cadherin signal by that of the corresponding actin signal. Here, the average normalized cadherin OD of the three cell lines is compared. No difference in cadherin expression was detected (ANOVA, p>0.05).

Figure 4. Assessment of fibronectin matrix assembly and α5 integrin expression by glioma cells.

A) Immunofluorescence analysis of FNMA by glioma cells. Cells were plated at equal density and incubated for 24 hours, whereupon they were stained with an anti-fibronectin antibody and an Alexa-568-conjugated secondary IgG. Cell nuclei were counterstained using Syto-16 live-cell nuclear stain. Images from green and red channels were captured, pseudo-colored, and merged in IP lab. Nuclei are depicted in green and fibronectin matrix is depicted in red. B) Biochemical analysis of FNMA by differential solubilization assay and immunoblot analysis. The assembly of high molecular weight fibronectin multimers (HMWFM) by U87-MG, LN-229, and U118-MG cells was assessed using DOC differential solubilization assay. DOC-soluble and insoluble fractions were separated by SDS-PAGE and analyzed by immunoblotting using an anti-fibronectin antibody. Actin in the soluble fraction was used as loading control. C) Analysis of cell surface α5β1 integrin expression by flow cytometry. Cells were trypsinized from near-confluent tissue culture plates, washed, and incubated with a mAb specific for the extracellular domain of α5 integrin, followed by incubation with a FITC-conjugated secondary antibody. The control panel (left) represents cells incubated with secondary antibody only. The right panel represents cell surface expression of α5 integrin. D) Comparison of α5β1 integrin expression by measuring mean fluorescence intensity (MFI). Cells were incubated in either IgG-FITC or in anti-α5β1 integrin antibody and IgG-FITC as described in Materials and Methods. Cells were appropriately gated and statistical analyzed using CellQuest software to calculate MFI values for controls and their matched samples. The MFI values of controls were subtracted from the MFI values of their matched samples, yielding a net MFI value for each cell line. Five separate experiments were performed using exact cytometer settings. The MFI data were averaged and mean MFI were compared by t-test or ANOVA and Tukey's MCT. MFI of U87-MG and U118-MG were statistically identical (ANOVA, P<0.0001, Tukey's MCT, p>0.05). MFI of LN229 was lower than that of either U87-MG or U118-MG (ANOVA, p<0.0001, Tukey's MCT, p<0.001).

Figure 5. Reastoration of FNMA promotes compaction of glioma cells.

A) Restoration of FNMA by Dexamethasone, MEK inhibitor, and Geldanamycin. Cells were plated at equal densities and incubated for 24 hours in the presence of Dexamethasone, MEK inhibitor, Geldanamycin, or carrier-only controls. FNMA by U87-MG and U118-MG was assessed by immunofluorescence as previously described. Fibronectin matrix was detected using anti-fibronectin antibody (ab6584) and Alexa-568-conjugated secondary IgG and is depicted in red. Nuclei were stained with Syto-16 live-cell nuclear stain and are depicted in green. Dex and GA appear to be most effective in restoring FNMA in both lines. B) Restoration of FNMA results in aggregate compaction. Untreated and drug-treated U87-MG cells were placed in hanging drop culture and incubated for 24 hours, whereupon aggregate size was measured as previously described. Depicted are mean aggregate size in pixels, with corresponding standard error. Note the marked compaction of aggregates treated with Dex, MEKi, or GA relative to carrier controls (water and DMSO). C) Immunoblot analysis of α5 integrin and pan-cadherin expression and of FN secretion in response to drug treatment. Antibodies specific for α5 integrin, secreted fibronectin, or pan-cadherin were used to determine whether drug treatment altered expression or secretion levels. Immunoblot analysis reveals that drug treatment does not appear to upregulate the expression of either α5 integrin nor of any cadherin recognized by the CH-19 pan-cadherin antibody. FN secretion also appears to be unaffected by drug treatment. Depicted, are protein levels relative to actin, which was used as loading control.

Figure 6. Effects of drug treatment on aggregate spreading velocity.

A) Dex treatment significantly reduces the slope of the regression line. Aggregate spreading assays were performed using untreated and Dex-treated aggregates of U87-MG cells. Slopes were found to be significantly different (F-test, P<0.0001), with untreated aggregates having a steeper slope than that of their Dex-treated counterparts. B) Dex treatment promotes cell-cell contact during aggregate spreading. Depicted are phase-contrast images of control and Dex-treated aggregates of U87-MG glioma cells after 8 hours of incubation. Dex treatment markedly reduces spreading and appears to promote cell-cell cohesion at the advancing cell front. C) Spreading velocity of control and drug-treated aggregates. Spreading velocity of aggregates of U87-MG (solid bars) and U118-MG (hatched bars) was compared in response to treatment with Dex, MEKi or GA. Treatment by Dex or GA consistently reduced aggregate dispersal velocity in both cell lines.

Figure 7. Blocking Dex-mediated FNMA rescues spreading velocity.

A) Blocking of FNMA by the 70 KDa fragment of fibronectin. Aggregates of U87-MG cells generated and incubated in the absence and/or presence of the 70 KDa fragment were assessed for FNMA by differential solubilization assay. DOC-soluble and insoluble fractions were separated by SDS-PAGE and analyzed by immunoblotting using an anti-fibronectin antibody. Actin in the soluble fraction was used as loading control. As expected, Dex-treatment significantly increased the amount of soluble fibronectin and high molecular weight fibronectin multimers (HMWFM). Addition of the 70 KDa fragment had no effect on soluble fibronectin but significantly reduced the amount of insoluble FN matrix, represented here by a specific band of molecular weight >220 KDa. Knock-down of the FN matrix was more pronounced in aggregates generated and incubated in 50 µg/ml of the 70 KDa fragment. B) Blocking Dex-induced matrix assembly influences aggregate compaction. U87-MG cells were plated either in conventional 2D culture (IF panel) or in 3D hanging drop culture (compaction panel) in the absence and/or presence of Dex and 70 KDa fragment. Cells in 2D culture were incubated in anti-fibronectin antibody followed by an Alexa-488-conjugated secondary antibody. Fibronectin matrix is depicted in green. DAPI was used to label nuclei, here depicted in blue. Inclusion of the 70 KDa fragment in both untreated and Dex-treated aggregates alters the appearance of the matrix from one in which long fibers extend between cells to a more punctate pattern. A corresponding change in compaction was also noted; Dex-treated cells tending to form less compact aggregates when incubated in the presence of the 70 KDa fragment. C) Blocking Dex-induced fibronectin matrix assembly rescues aggregate spreading velocity. The spreading velocity of aggregates prepared in the presence of a combination of Dex and the 70 KDa fragment was greater than that of aggregates generated in Dex-containing medium alone and approached the spreading velocity of untreated aggregates.