Figure 1.
Deletion of CXCR4 suppresses the growth of intracranial U87 xenografts.
(A) Western blot analysis of CXCR4 expression in U87 cells infected with lentirvirus encoding a scrambled shRNA control (scRNA) or a short hairpin RNA directed against CXCR4 (CXCR4 shRNA). CXCR4 expression declined with increasing viral mulitplicity of infection (MOI). (B) Animals were injected with U87 cells expressing either control (scrambled, sc-U87-Luc) or a CXCR4 specific shRNA (shCXCR4-U87-Luc). Growth curves were derived from serial bioluminesence imaging measurements (six to eight animals per experimental group) over the four-week experimental period post tumor cell implantation. Presented are the mean Bioluminescence Ratios (photon flux week (1–4)/photon flux hr48) ± SEM for each group. * = P<0.005 as determined by two-way ANOVA.
Figure 2.
Primary human brain endothelial cells express CXCL12 and GBM cells express CXCR4.
(A) Human glioblastoma specimen immunostained for CXCL12 (brown) demonstrates expression in vascular endothelial cells. t = tumor cells, e = cross-section through tumor-associated capillary, and Scale bar = 25 µm. (B) HBMECs, in co-culture on Matrigel, express CXCL12 (red). Nuclei are counterstained blue with DAPI. Scale bar equals 25 µm. (C) Human brain micro-vascular endothelial cells cultured on Matrigel (Mat HBMEC) secrete CXCL12 into the media as determined by ELISA. Mat alone indicates results from Matrigel alone-conditioned media. N = 3. ** = p<0.005 as determined by two-tailed t-test. (D) Single cell suspensions from two different adult GBM patients (GBM1, GBM2) and cultured U87 cells express CXCR4 by western blotting. CXCR4 appears red and the actin loading control appears green.
Figure 3.
U87 and primary GBM cells localize to the peri-endothelial domain when plated with HBMECs on Matrigel.
(A) Twenty-four hours after establishing a capillary-like network of mcherry-expressing HBMECs in Matrigel, eGFP-expressing U87 cells were added to the culture. Within 24 hours U87 cells were seen in physical contact with HBMECs. Scale bar = 50 microns. (B) The mean distances between U87 cells (500 to 800 cells) and HBMECs were calculated at different time points after the addition of the tumor cells to the HBMEC networks. There was a significant increase in co-localization (reduction in mean distance) within 24 hrs, which was maintained over a 72 hr period. * = p<0.05 as determined by one way ANOVA for the means of three separate experiments involving 500–800 measurements per experiment. (C) The distance between approximately 1000 eGFP-expressing U87 cells and mCherry fluorescent protein-expressing HBMECs in co-culture (24 hrs) was measured and the distribution was plotted as the percentage of total cells in 20 micron increments (black triangles). More than 50% of the total U87 cells in the culture were within 40 microns of an endothelial cell. A theoretical plot of a random distribution of cells is shown (open circles). (D) The distance between GFAP positive GBM cells and mCherry fluorescent protein-expressing HBMECs in co-culture were measured and the distribution was plotted as the percentage of total cells in 20 micron increments (black triangles). Error bars represent SEM from three independent experiments involving three different GBM isolates. Approximately 500 GBM (GFAP positive) cells were counted. Nearly 80% of the GFAP positive GBM cells in the culture were within 20 microns of an endothelial cell. A theoretical plot of a random distribution of cells is shown (open circles).
Figure 4.
U87 localization to the peri-endothelial space is correlated with CXCL12 expression.
(A) Quantitative PCR reveals that HUVECs express significantly lower levels of CXCL12 than HBMECs. N = 3, ** = p<0.005 as determined by two-tailed t-test. (B) The distance between greater than 1000 eGFP-expressing U87 cells and mCherry fluorescent protein-expressing HUVECs in co-culture was measured and the distribution was plotted as the percentage of total cells in 20 micron increments (filled squares). Approximately 30% of the total U87 cells in the culture were within 40 microns of an endothelial cell (compare to Figure 3B). This distribution does not differ from a theoretical plot of a random distribution of cells (open circles) (C) A statistically significant difference existed between the mean distance of U87 cells to HBMECs versus the mean difference between U87 and HUVECs. * = p<0.05 as determined by two-tailed t-test for the means of four separate experiments involving approximately 330 measurements per endothelial co-culture per experiment. (D) When cultured with HUVECs that are engineered to over-express CXCL12, U87s show a similar co-localization pattern (as seen with HBMECs) consistent with a significant reduction in mean distance when compared to co-cultures with control HUVECs. * = p<0.05 as determined by two-tailed t-test for the means of two separate experiments involving approximately 300 measurements per endothelial co-culture per experiment.
Figure 5.
U87 localization to the peri-endothelial space is blocked by the CXCR4 antagonist AMD 3100.
(A) GFP-expressing U87 cells (green) are localized to mCherry-expressing HBMECs (red). (B) Treatment of parallel co-cultures as in A with AMD 3100 results in failure of U87 cells (green) to make consistent contact with endothelial cells (red). Scale bar equals 25 microns. (C) The CXCR4 antagonist AMD 3100 redistributes U87 cells, decreasing the number of cells within the nearest proximity to HBMECs and increasing the number of cells at greater distance. * = p<0.05 as determined by two-way ANOVA. (D) AMD 3100 increased the mean distance between U87 cells and HBMECs but had no effect on the mean distance between U87 cells and HUVECs. * = p<0.05 as determined by two-tailed t-test.
Figure 6.
The chemotactic and trophic effect of the peri-endothelial space on primary GBMs is blocked by pharmacologic or genetic inhibition of the CXCL12-CXCR4 pathway.
(A) The distance between GFAP positive GBM cells and mCherry fluorescent protein-expressing HBMECs in co-culture was measured as in Figure 3. Error bars represent SEM from three independent experiments involving three different GBM isolates. Approximately 100 GBM (GFAP positive) cells were counted in each condition. Nearly 80% of the GFAP positive GBM cells in the culture were within 20 microns of an endothelial cell. Treatment with AMD3100 resulted in a reduction in the proportion of GBM cells in closest proximity to endothelial cells and an increase in the proportion of cells at greater distances. P<0.005 by two-way ANOVA. (B) Primary GBM cells co-cultured with HBMECs expressing a scrambled shRNA (sh-RNA-sc) show increased growth after 72 hours in culture, as measured by number of GFAP-positive tumor cells, compared to growth on Matrigel alone. The trophic effect of HBMECs on primary GBM cells is abrogated upon depletion of CXCL12 expression. N = 3, * = P<0.005 as determined by one-way ANOVA with Dunnett's post-test for multiple comparisons.