The EphA2 receptor drives self-renewal and tumorigenicity in stem-like tumor-propagating cells from human glioblastomas

Abstract

In human glioblastomas (hGBMs), tumor-propagating cells with stem-like characteristics (TPCs) represent a key therapeutic target. We found that the EphA2 receptor tyrosine kinase is overexpressed in hGBM TPCs. Cytofluorimetric sorting into EphA2(High) and EphA2(Low) populations demonstrated that EphA2 expression correlates with the size and tumor-propagating ability of the TPC pool in hGBMs. Both ephrinA1-Fc, which caused EphA2 downregulation in TPCs, and siRNA-mediated knockdown of EPHA2 expression suppressed TPCs self-renewal ex vivo and intracranial tumorigenicity, pointing to EphA2 downregulation as a causal event in the loss of TPCs tumorigenicity. Infusion of ephrinA1-Fc into intracranial xenografts elicited strong tumor-suppressing effects, suggestive of therapeutic applications.

Figure 1. The EphA2 receptor in human glioblastomas tissues and TPCs

(A–B) Example of strong and frequent EphA2 immunoreactivity (brown) in the tumor core (A; arrows) as compared to infrequent, weaker labeling in the periphery (B) of the same hGBM specimen (6 hGBMs yielded similar results). Insets: higher magnification. Bottom: no primary antibody. Blue, hematoxylin counterstain. Bar, 20 μm. See also Figs. S1A–C. (C) Immunolabeling of hGBM tissue shows that cells positive for the putative neural stem cell markers Nestin, Sox2 or Olig2 (red) co-express (arrows) the EphA2 receptor (green; 5 hGBMs yielded similar results). Bar, 20 μm. (D) Co-expression of EphA2 (green) and ephrinA1 (red) on the surface of cells in hGBM tissues (left) and TPC cultures (right). Arrows mark examples of co-expression (yellow). Bars, 10 μm. (E) Higher EphA2 expression as detected by qPCR in cultured TPCs from the core (TPCs; n=14 independent cultures) versus cells from the periphery of the same hGBM (non-TPCs; n=6 independent cultures; Student’s t test, p=0.017). Controls: vhNSCs and human fibroblasts (HF). Error bars: SEM. See also Figs. S1D–J. (F) TPCs differentiated by mitogen starvation show upregulation of astroglial- (GFAP), oligodendroglial (GalC) and neuronal (β Tub, MAP2, tyrosine hydroxylase (TH), glutamate and GABA) markers (left), while loosing EphA2 expression (right).

Figure 2. Enrichment of the stem-like tumorigenic pool based on EphA2 levels

(A) Viable (propidium iodide negative) tumor cells acutely isolated from hGBM specimens (top left) were sorted into EphA2^High and EphA2^Low fractions (bottom left). The EphA2^High fraction displayed higher clonogenic index than the EphA2^Low fraction (right) (n=4 tumors). Error bars: SEM; **p=0.0004 for EphA2^High vs. EphA2^Low by Student’s t test. (B) Intracranial transplantation of 6×10⁴ EphA2^High or EphA2^Low cells confirmed the much higher tumor-propagating capacity of the former (MC test, log-rank p-value <0.0001 for EphA2^High vs. EphA2^Low; n=8). (C) Confocal images show widespread co-localization (arrowheads; yellow) of EphA2 (red) and SSEA-1 (green) in hGBM tissue. Bar, 20 μm. (D) Cells from the same hGBM were sorted and gated according to EphA2 and SSEA-1 levels. (E) Kaplan-Meier survival curves show that mice receiving intracranially 2×10⁴ and 1×10⁴ EphA2^High SSEA-1^High purified TPCs die earlier (median survival: 135 and 164 days, respectively) than mice receiving EphA2^Low SSEA-1^Low cells (56% and 67% survival at 230 days, respectively). MC and GBW tests, log-rank p-value <0.0001 EphA2^High SSEA-1^High vs. EphA2^Low SSEA-1^Low; n=9). Survival was also shorter when implanting 4×10⁴ EphA2^High SSEA-1^High as compared to EphA2^Low SSEA-1^Low TPCs. See also Fig. S2. (F) Limiting dilution intracranial transplant of cultured, luciferase-tagged TPCs sorted into EphA2^High and EphA2^Low pools (top). Light emission imaging analysis (bottom; 5,000, 1,000 and 100 cells per mouse) shows a higher tumor-initiating ability of EphA2^High versus EphA2^Low TPCs. Error bars, SEM; ***p<0.0001, **p=0.002, EphA2^High vs. EphA2^Low. (G) Kaplan-Meier analysis shows that mice receiving EphA2^High TPCs die earlier than mice receiving EphA2^Low cells (MC and GBW tests, log-rank p-value <0.0001 EphA2^High vs. EphA2^Low; n=9).

Figure 3. EphA2 downregulation by ephrinA1-Fc inhibits in vitro proliferation and depletes the stem cell-like pool in hGBM TPCs

(A) (Top) Cells were treated with ephrinA1-Fc for 24 hours. EphA2 was heavily downregulated in TPCs but not vhNSCs and non-TPCs. (Bottom) FACS analysis and flow cytometry quantitative data showing EphA2 downregulation in TPCs based on equivalent molecules of phycoerythrin (ME-PE) (n=10 independent cultures); error bars, SEM; **p<0.005 by Student’s t test. See also Fig. S3A. (B) EphrinA1-Fc (5 μg/ml for 24 hours) downregulates EphA2 expression in TPCs acutely isolated from patients or from neurospheres (TPCs #8 shown as an example), but not in vhNSCs. Bar, 20 μm. (C–F) EphrinA1-Fc (5 μg/ml for 48 hours) triggers obvious morphological changes (arrowheads) in acutely isolated TPCs (C, control; D, treated) or serially subcultured TPC neurospheres (E, control; F, treated), promoting cell adhesion. Bar, 50 μm. (G–I) Acutely isolated TPCs cannot establish stable TPC primary lines if exposed to ephrinA1-Fc (G, red line), which also inhibits steady growth in pre-established TPCs (G); p<0.0001 vs. Control-Fc. Negligible inhibition of ephrinA1-Fc was observed in non-TPCs (H) and vhNSCs (I) (p<0.005 vs. Control-Fc; TPCs#8 and non-TPCs#8 are shown as representative examples); error bars, SEM. See also Figs. S3B–C. (J) Co-expression of EphA2 (red) with the putative stem antigens Nestin, Sox2 or Olig2 (green) in TPC spheroids treated for 24 hours with Control-Fc or ephrinA1-Fc. EphrinA1-Fc nearly abolishes expression of both EphA2 and the stem antigens. Insets: higher magnification. Bar, 40 μm. See also Figs. S3D–E. Unless otherwise indicated, the data are representative of three independent experiments giving similar results.

Figure 4. EphrinA1-Fc inhibits TPC self-renewal by inducing a differentiated phenotype

(A) Clonogenic assays show a dose-dependent inhibition of self-renewal by ephrinA1-Fc in TPCs but not in non-TPCs and vhNSCs; error bars, SEM; ***p<0.0001 vs. Control-Fc cells. (B) Up to 5 μg/ml ephrinA1-Fc does not alter TPC cell cycling, as determined by FACS analysis of BrdU incorporation; error bars: SEM. (C) Cytofluorimetric Tunel analysis shows no induction of apoptosis in TPCs or non-TPCs treated with ephrinA1-Fc; error bars: SEM; **p<0.005, *p<0.05 vs. Control-Fc. (D) Quantitative FACS analysis shows a time-dependent increase in equivalent molecules of fluorescein (ME-FITC) for astroglial GFAP but not neuronal βIII tubulin (β–Tub) or oligodendroglial GalC markers in ephrinA1-Fc-treated TPCs; error bars, SEM; *p=0.0095, **p=0.0005 vs. Control-Fc. (Inset) Western blotting confirms a marked, time-dependent increase in GFAP under the same settings. (E) (Top) Tyrosine phosphorylation of EphA2 immunoprecipitated from whole hGBM lysates and from TPCs treated with ephrinA1-Fc. (Bottom) Western blots for EphA2, ERK, Akt and FAK (Tyr397) expression and phosphorylation in TPCs treated with increasing ephrinA1-Fc concentrations over a 24 hours time course. Lower ephrinA1-Fc concentrations preferentially stimulate signaling, higher ones more rapidly downregulate the receptor. (F–J) (F) TPCs spread on Cultrex show an organized actin cytoskeleton and F-actin assembled in stress fibers (arrows). Typical ring-like actin bundles are seen at higher magnification (H, I; arrowheads). (G, J) EphrinA1-Fc-treatment (5 μg/ml for 5 min) causes TPC elongation and actin concentration at cell-cell junctions (arrowhead). Bar, 20 μm. Unless otherwise indicated, the data are representative of three independent experiments giving similar results.

Figure 5. EPHA2 siRNA knockdown in hGBM TPCs inhibits self-renewal and increases differentiation, concomitant with ERK and Akt activation

(A) TPCs were treated with an EPHA2 siRNA pool, non-targeting control (NTC) or GAPDH control pool siRNAs. Three days post-transfection, only the EPHA2 siRNAs caused a substantial decrease in EphA2 mRNA, as detected by qPCR. Between 12 and 16 days post-transfection (DIV), EphA2 levels began to normalize (arrow); error bars, SEM. (B) siRNA-mediated knockdown of EPHA2 expression causes loss of TPC clonogenicity; error bars, SEM; **p<0.005 EphA2 vs. NTC siRNAs. (C) TPC growth decrease is concomitant with siRNA-mediated EphA2 downregulation and normalizes when EphA2 levels begin recover (arrow); error bars, SEM; p<0.0001 EPHA2 siRNAs vs. NTC. (D) siRNA-mediated EphA2 downregulation decreased vhNSCs growth; error bars, SEM; ***p<0.0001 EPHA2 siRNAs vs. NTC. Inset: Western blot for EphA2 upon EPHA2 siRNA treatment. (E) Confocal analysis shows that siRNA-mediated EphA2 downregulation in TPC spheroids is associated with depletion of putative stem markers, as compared to NTC-treated spheroids (NTC siRNAs). Insets: magnification. Bar, 40 μm. (F–G) Western blot analysis show increased GFAP but not β–Tub or GalC levels in TPCs treated with EPHA2 siRNAs versus NTC or GAPDH siRNAs (F). ERK phosphorylation is also strongly increased concomitant with EphA2 downregulation, with more prominent effects at 72 hours than at 5 days, when EphA2 levels begin to recover. (G) Representative Western blot showing that only the siRNA sequences that effectively reduce EphA2 expression increase GFAP expression and activate ERK. An EPHA2 construct lacking the 3’UTR sequence targeted by EPHA2 siRNA #3 was used to transfect TPCs in a control rescue experiments. (H) The reduced TPC clonogenicity caused by knockdown of EPHA2 expression was partially restored by 10 μM UO126, which inhibits ERK; error bars, SEM; ***p<0.0001, **p<0.005 vs. untreated TPCs. Unless otherwise stated, data are representative of three independent experiments giving similar results.

Figure 6

EphA2 does not appear to be activated by ephrins in hGBM TPCs is phosphorylated on Ser897 in hGBM TPCs and treatment with the soluble EphA2 extracellular domain to inhibit ephrin binding to endogenous EphA2 does not affect TPC signaling pathways. (A) Top: EphA2 is constitutively phosphorylated on Ser897 in the core of hGBMs (left) but not in the periphery of hGBMs (centre) or normal brain (right). Bottom: EphA2 is highly phosphorylated on Ser897 in hGBM TPCs but not non-TPCs or vhNSCs. EphA2 Ser897 phosphorylation suggests ephrin-independent oncogenic activities. (B) Representative Western blots showing that treatment of TPCs with 10 μg/ml EphA2 extracellular domain to inhibit possible interactions with endogenous ephrin ligands does not affect GFAP levels or ERK phosphorylation as compared to Fc protein used as a control. Data are representative of three independent experiments giving similar results.

Figure 7. EphA2 downregulation by ephrinA1-Fc or by siRNA-mediated knockdown inhibit TPC tumorigenicity in immunodeficient mice

(A) Treating TPCs with ephrinA1-Fc (5 μg/ml for 48 hours) prior to subcutaneous implantation (PRE-treatment) lessens their tumor-initiating capacity (left). Similar results were obtained by co-injecting TPCs and ephrinA1-Fc (CO-treatment) or injecting ephrinA1-Fc (10 μg/day) around the tumor starting 11 days after cell transplant (POST-treatment). (Right) Volumes of subcutaneous tumors 35 days after TPCs injection. Histograms, mean volume ± SEM; **p=0.0002 vs. Control-Fc mice, n=6. (B) Imaging of luciferase-tagged TPCs (luc-TPCs) injected into the brain of Scid/bg mice. After 42 days, untreated TPCs established larger tumors (vehicle, top left) than ephrinA1-Fc PRE-treated TPCs (bottom left). Luc-TPC tumors established for 7 days (7 DPT, top center) grew quickly when a mini-pump delivered Control-Fc for 14 days starting at 11 days post-transplant (27 DPT, bottom center). In contrast, tumor growth was markedly inhibited by ephrinA1-Fc (top and bottom right panels). (C) (Left) Quantitative analysis of luc-TPC signals for the PRE-treatment intracranial transplants. Histograms, mean ± SEM; ***p<0.0001 vs. Control-Fc mice; n=6. (Right) Kaplan-Meier survival curves showing that mice receiving ephrinA1-Fc treated TPCs have a significant longer life span than mice injected with Control-Fc cells (MC and GBW tests, log-rank p-value=0.0005 and 0.0013 respectively; n=9). (D) (Left) Quantitative time course analysis of the luc-TPC signal for POST-treatment paradigm (arrows mark the time of mini-pumps implantation). Histograms, mean ± SEM; *p<0.05 vs. Fc-treated mice; n=8. (Right) Kaplan-Meier survival curves are shown (MC and GBW tests, log-rank p-value <0.0001 and p=0.0002 vs. Control-Fc mice; n=9). (E–H) Mouse brain sections immunolabeled for luciferase show that tumors established from luc-TPCs PRE-treated with Control-Fc (E) spread through the brain parenchyma more than those established from cells PRE-treated with ephrinA1-Fc (F) at 42 DPT. Similarly, ephrinA1-Fc infused into the brain for 2 weeks starting 11 days after tumor establishment (H) inhibits the growth of luc-TPC tumors more than Control-Fc (G). Arrowheads mark the edges of the tumors. CC: corpus callosum; LV: lateral ventricle; St: Striatum. See also Fig. S4. Bar, 1 mm. (I) Loss of intracranial tumorigenicity in luc-TPCs treated with EPHA2 siRNAs for 72 hours prior to transplantation as compared to NTC siRNA-treated or untreated TPCs. Tumor growth was monitored by quantitative imaging analysis (top). Histograms, mean ± SEM; ***p<0.0001 EPHA2 siRNAs vs. NTC; n=6. (Bottom) Kaplan-Meier survival analysis. Mice receiving EPHA2 siRNA-transfected TPCs die significantly later than those receiving untreated TPCs or TPCs transfected with non-targeting control siRNAs (MC and GBW tests, log-rank p-value <0.0001 vs. NTC siRNAs treatment; n=9).

Figure 8. EphA2 is differentially expressed in GBM subtypes and its abundance correlates with patient survival

(A) Relative EphA2 mRNA expression in the four GBM sutbypes of the TCGA dataset. (***p<0.0001 by ANOVA; n=495). (B) Kaplan-Meier survival curves for the Classical, Mesenchymal, Proneural and Neural subtypes. High and low expression are defined as above and below the median expression value for each subtype, log rank p-values were determined by MC and GBW tests.