Figure 1.
Sample preparation and computational deconvolution of the mixed culture RNA signal.
(A) Flow chart of sample preparation from culture to gathering RNA and computational analysis. (B) The percentage of each monoculture to use for the control dataset was determined by creating one thousand test datasets, each with a different amount of U87 or HBMEC signal, and then calculating the Pearson correlation coefficient between each of these datasets and the measured co-culture dataset.
Table 1.
Genes differentially expressed by co-culture of U87 and HBMECs.
Figure 2.
Specification of the cell of origin for expression changes by qRT-PCR validation of differentially expressed genes using conditioned media swap.
Conditioned medias were Serum Free DMEM (SF DMEM), SF DMEM conditioned by HBMECs (HBMEC), SF DMEM conditioned by U87 cells (U87 MG), or SF DMEM conditioned by a co-culture of HBMECs and U87s. HBMECs (left column) or U87s (right column) were grown in one of the respective conditioned media types for 48 hours, RNA was isolated, and qRT-PCR was performed for CXCL6, CXCL1 and THBS1. Expression of CXCL6 and CXCL1 specifically changed in the HBMECs in response to U87 or co-culture CM, while changes in THBS1 expression were limited to the U87 cells and occurred most significantly in response to co-culture CM. Expression is in arbitrary quantification units calculated by performing standard delta-delta C(t) while setting GAPDH expression to an arbitrary level of 10,000. N = 3 for all sets, significance by Student’s t-test.
Figure 3.
Changes in PDE7B Expression require physical contact between endothelial cells and U87 cells.
(A) PDE7B expression was measured by qRT-PCR in RNA isolated from U87 cells grown alone or on methanol fixed HBMECs. *p-value = 0.04 by Student’s t-test. (B) PDE7B expression similarly measured in RNA isolated from a primary GBM line, B18, grown alone or on fixed HBMECs. #p-value = 0.02 by Student’s t-test.
Figure 4.
PDE7B is expressed in human GBM.
(A) Oncomine Murat et al [28] dataset showing overexpression of PDE7B in GBM samples compared to normal brain. (B) Oncomine Sun et al [29] dataset also showing overexpression of PDE7B in GBM samples compared to normal brain. (C) cDNA was prepared from twenty-two primary human GBM RNA samples. PDE7B expression was quantified by qRT-PCR and compared to expression in normal human astrocytes. (D) Relationship between PDE7B expression and molecular subtypes of GBM, Classical (C), Mesenchymal (M), Proneural (P) and Neural (N). Shown are box and whisker plots of data accessed through the TCGA as described in Materials and Methods. P-values for differences in subtype specific expression are shown.
Figure 5.
PDE7B expression is correlated with Survival.
(A) Data from NCI’s Rembrandt database indicates that expression of PDE7B at levels two-fold lower than median is strongly correlated with improved survival compared to median levels of expression in all glioma samples. p-value = 2.55E-5 (B) Oncomine’s Murat et al [28] dataset showing that in GBM patients those who were alive at 5 years had very much lower levels of PDE7B than those who had succumb to disease at that point (p-value = 2.56E-9).
Figure 6.
Overexpression of PDE7B increases stem cell fraction and accelerates in vivo tumor growth and decreases survival.
(A) Frequency of clonogenic stem-like cells as determined by Extreme Limiting Dilution Assays (ELDA) with U87 cells expressing PDE7B (WT), catalytically inactive PDE7B (H217Q) or no vector control (U87GL). PDE7B-WT expression increased the clonogenic cell frequency (*indicates p = 0.037, two tailed t-test) compared to the catalytically inactive form. (B) Plot of the log fraction of wells without spheres as a function of plated U87 cell number. The more vertical the line, the higher the percentage of clonogenic, sphere-forming cells. (C) Frequency of clonogenic stem-like cells as determined by Extreme Limiting Dilution Assays (ELDA) with G144 cells expressing PDE7B (WT), catalytically inactive PDE7B (H217Q) or empty vector control (TdTomato). PDE7B-WT expression increased the clonogenic cell frequency compared to TdTomato (**indicates p = 0.001, two tailed t-test) and compared to catalytically inactive PDE7B (H217Q) (**indicates p = 0.003, two tailed t-test). (D) Plot of the log fraction of wells without spheres as a function of plated G144 cell number. The more vertical the line, the higher the percentage of clonogenic, sphere-forming cells. (E) Intracranial growth of luciferase-expressing U87 cells was monitored by weekly bioluminescence imaging (BLI). Data for each individual mouse is normalized to its own week one value and shown are the means and SEM for each group (N = 14 (Wild Type), N = 13 (H217Q)). P-value = 0.0002 as determined by two-way ANOVA. (F) Kaplan-Meier plot of survival. Logrank test indicated a significant decrease in survival (p-value = 0.02) in mice xenografted with U87-GL cells overexpressing wild type PDE7B compared to cells overexpressing catalytically inactive PDE7B (N = 14 (Wild Type), N = 13 (H217Q).
Figure 7.
Expression of PDE7B increases tumor vascularity and invasiveness.
(A) H&E staining of representative U87-GL+PDE7B(WT) tumor, with tumor tissue outlined in black. Scale bar = 1 mm (B) H&E staining of representative U87-GL+PDE7B(H217Q) tumor, with tumor tissue outlined in black. Scale bar = 1 mm (C) Endoglin staining of a representative U87-GL+PDE7B(H217Q) tumor illustrates usual pattern of U87 growth. Black arrows demarcate tumor-surrounding brain boundary. Tumor cells are not detected in the surrounding brain and there is no angiogenic response in the surrounding brain. Scale bar = 300 µm. (D) Immunolabelling for Endoglin of a U87-GL+PDE7B(WT) xenograft revealed a highly invasive tumor leading edge with tumor cells evident in the surrounding brain accompanied by a robust angiogenic response (asterisks). Scale bar = 300 µm.