|Home | About | Journals | Submit | Contact Us | Français|
Glioblastomas often show activation of epidermal growth factor receptor (EGFR) and loss of PTEN (phosphatase and tensin homolog deleted on chromosome 10) tumor suppressor, but it is not known if these two genetic lesions act together to transform cells. To answer this question, we infected PTEN−/− neural precursor cells with a retrovirus encoding EGFRvIII, which is a constitutively activated receptor. EGFRvIII PTEN−/− cells formed highly mitotic tumors with nuclear pleomorphism, necrotic areas, and glioblastoma markers. The transformed cells showed increased cell proliferation, centrosome amplification, colony formation in soft agar, self-renewal, expression of the stem cell marker CD133, and resistance to oxidative stress and ionizing radiation. The RAS/mitogen-activated protein kinase (ERK) and phosphoinositide 3-kinase/protein kinase B (PI3K/ Akt) pathways were activated, and checkpoint kinase 1 (Chk1), the DNA damage regulator, was phosphorylated at S280 by Akt, suppressing Chk1 phosphorylation at S345 in response to ionizing irradiation. The PTEN−/− cells showed low levels of DNA damage in the absence of irradiation, which was increased by EGFRvIII expression. Finally, secondary changes occurred during tumor growth in mice. Cells from these tumors showed decreased tumor latencies and additional chromosomal aberrations. Most of these tumor lines showed translocations of mouse chromosome 15. Intracranial injections of one of these lines led to invasive, glial fibrillary acidic protein–positive, nestin-positive tumors. These results provide a molecular basis for the occurrence of these two genetic lesions in brain tumors and point to a role in induction of genomic instability.
Hyperactivation of the epidermal growth factor receptor (EGFR) in glioblastomas occurs by several mechanisms. The EGFR gene is amplified in many glioblastomas,1 and as a result, EGFR protein is overexpressed. In addition, many tumors with amplified EGFR express a mutated EGFR.2 The most common of these modified receptors is EGFRvIII, which has a 267–amino-acid deletion in the extracellular domain.3 Using an anti-EGFRvIII antibody, EGFRvIII was detected in 53% of glioblastomas.4
EGFRvIII enhances cell proliferation, tumor growth, and invasiveness.5–10 EGFRvIII does not bind EGF, but its kinase domain is constitutively activated, possibly due to receptor dimerization.11 EGFRvIII constitutively activates the c-Jun N-terminal kinase (JNK) pathway and, to a lesser degree, the RAS/mitogen-activated protein kinase (ERK) pathway.12–14 Finally, phosphoinositide 3-kinase (PI3K) is activated to high levels.15
Phosphatase and tensin homolog deleted on chromosome 10 (PTEN) is a major tumor suppressor in the development of glioblastomas.16,17 In a mouse model, inactivation of the retinoblastoma (Rb) and PTEN tumor suppressors leads to astrocytic tumors.18 Deletions or mutations have been detected in other tumor types, including endometrial, melanoma, and prostate, identifying PTEN as an important tumor suppressor.19 The PTEN protein is a phosphatidylinositol phosphate phosphatase specific for the 3-position of the inositol ring.20
Many glioblastomas show EGFR amplification, EGFRvIII expression, and PTEN loss.21–23 Von Deimling et al.22 found that, without exception, glioblastomas with EGFR amplification also had loss of heterozygosity (LOH) of chromosome 10. In a recent study with 220 brain tumors, EGFR amplification was strongly associated with LOH of chromosome 10q (p < 0.0001).23 EGFRvIII expression usually occurs in glioblastomas with EGFR amplification. PTEN loss is more common in primary than in secondary glioblastomas,24 and EGFR amplification and PTEN loss frequently occur in the same tumors.21 EGFR amplification and PTEN loss are positively associated in glioblastomas (p = 0.018) and are negative predictors of patient survival. EGFRvIII expression is more common for tumors with intact PTEN but also occurs for glioblastomas lacking PTEN.25 PTEN expression reverses EGFRvIII-induced cell proliferation and increases susceptibility to anti-EGFR drugs.26–29 In glioblastomas, PTEN loss is closely linked to activation of the PI3K pathway, and EGFRvIII is linked to the RAS/ERK pathway.25 Expression of EGFRvIII or loss of PTEN accelerates development of brain tumors in response to an activated RAS gene.30 These findings led us to hypothesize that EGFR hyperactivation and PTEN loss might act together to transform cells.
To test this hypothesis, we have developed a mouse model by infecting PTEN−/− neural precursor cells with an EGFRvIII retrovirus. Although glioblastomas commonly express both EGFR and EGFRvIII, we chose to analyze EGFRvIII because its mitogenic activity is constant and not dependent on the concentration of EGF in the culture medium. We found that precursor cells bearing both genetic lesions form tumors in immunodeficient mice. Furthermore, secondary changes occurred during tumor growth. These results provide a molecular basis for the occurrence of these two genetic lesions in brain tumors and point to their role in induction of genomic instability.
The EGFRvIII retrovirus was prepared from the green fluorescent protein (GFP)-bearing retrovirus MSCV-XZ066,31 which was also used as a control virus. Pseudotyped VSV-G (vesicular stomatitis virus glyco-protein) viruses were prepared by transfecting the 293GPG packaging cell line32 with the pMSCV-XZ066 or EGFRvIII-pMSCV-XZ066 plasmids. The viral supernatants were concentrated by centrifugation (90 min at 75,000g). As determined by titration of NIH-3T3 cells, the titer was about 107 cfu/ml. Single-cell suspensions of precursor cells in 5 μg/ml Polybrene (Sigma-Aldrich, St. Louis, MO, USA) were infected for 8 h with retrovirus. After 48 h, we selected GFP-positive cells (50%–80% of total) by fluorescence-activated cell sorting (FACS), using a FACS Vantage SE DIVA Cell Sorter (Becton Dickinson, San Jose, CA, USA). At least 95% of the sorted cells were GFP positive.
Neural precursor cells were isolated from brain sub-ventricular zone using the procedure of Weiss and coworkers.33 Striata were removed from newborn mice, triturated to single cells, and placed in serum-free Dulbecco’s modified Eagle’s medium (DMEM)/F12 (1:1) supplemented with glucose (0.3%), transferrin (100 μg/ ml), putrescine (60 μM), progesterone (10 nM), selenium chloride (30 nM), HEPES (5 mM), insulin (23 μg/ml), and EGF (20 ng/ml). Cultures were fed every 48 h with 5 ml medium and passaged every 7 days.
Because the litters included PTEN+/+ and PTEN−/− pups, we could not pool tissue from multiple pups. We were concerned that our yield of precursor cells would be low from a single embryonic day 14 (E14) subventricular zone culture. The cultures from newborn mice grew and behaved like our earlier E14-derived cultures (data not shown).
To produce PTEN−/− neural precursor cultures, we first prepared nestin-Cre+/+ PTEN+/lox mice in a C57BL6/J background, starting with the PTEN lox/lox mice (mouse genomic DNA PTEN exons 4 and 5 were flanked by 2 loxP sites in a direct orientation)34 and nestin-Cre mice (Jackson Laboratories, Bar Harbor, ME, USA). The nestin-Cre mice bear the Cre DNA recombinase under the control of the nestin promoter, which leads to expression of Cre in neural stem and precursor cells. The PTEN lox/lox mice have loxP sites so that, in the presence of the Cre recombinase, exons 4 and 5 of the PTEN gene are deleted. Consequently, PTEN is selectively inactivated in the brain stem/precursor cells expressing nestin. However, the PTEN gene is intact in other tissues that do not express the Cre recombinase. We crossed nestin-Cre+/+ PTEN+/lox mice with themselves. Because the PTEN−/− mice die shortly after birth, we prepared neural precursor cell cultures immediately after birth. The tails were retained for DNA extraction and genotyping. Both the PTEN+/+ and PTEN−/− cultures expressed Cre.
For genotyping, we used two sets of PCR primers. To detect the intact PTEN gene, we used primers P1 (5'-TGT T T T TGACCA AT TA A AGTAGGCT-GTG-3') and P2 (5'-AAAAGTTCCCCTGCTGAT-GATTTGT-3').35 To detect deletion of exons 4 and 5, we used primers P2 and P3 (5'-TTCTCTTGAGCACT-GTTTCACAGGC-3').
Western blotting was carried out as previously described.36 The rabbit antibodies anti-phospho–Akt (ser473, 1:1,000), anti-Akt (1:1,000), anti-phospho–checkpoint kinase 1 (anti-P-Chk1; ser345, 1:1,000), anti-phospho–mitogen-activated protein kinase kinase 1/2 (anti-P-MEK1/2; ser217/221, 1:1,000), anti-MEK1/2 (1:1,000), anti-phospho–stress-activated protein kinase/ JNK (anti-P-SAPK/JNK; thr183/tyr185, 1:1,000), anti–serine-threonine kinase (anti–Aurora A, 1:1,000), and Aurora B (1:1,000) were from Cell Signaling (Danvers, MA, USA). Anti-phospho-Chk1 (ser280, 1:1,000) was a gift from Dr. Ramon Parsons (Dept. of Pathology, Columbia University Medical Center, New York, NY, USA). The mouse anti-Chk1 monoclonal antibody (G4, 1:1,000) was purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Anti-β-tubulin (clone 2-28-33, 1:2,000) was from Sigma-Aldrich (St. Louis, MO, USA). Anti-phosphohistone-H2AX (anti-P-γ-H2AX; ser139, 1:1,000) was from Upstate Biotechnology (Lake Placid, NY, USA).
Cultured cells were spun down on slides with a Cytospin apparatus (Shandon Inc., Pittsburgh, PA, USA) and fixed with methanol (−20°C) for 15 min. After being blocked with 10% bovine serum albumin in phosphate-buffered saline (PBS), cells were incubated with rabbit anti-pericentrin antibody (Steve Doxsey, University of Massachusetts Medical School; 1:100) or mouse monoclonal anti-γ-tubulin antibody (Sigma Chemicals, St. Louis, MO, USA; 1:500) for 12 h at 4°C. The cells were then incubated with fluorescein- or rhodamine-linked anti-immunoglobulin G (IgG) secondary antibody (Vector Laboratories, Burlingame, CA, USA). The coverslips were mounted with Vectashield mounting medium with DAPI (4',6 diamidino-2- phenylindole; Vector Laboratories).
For CD133 immunostaining, cells were fixed with 4% paraformaldehyde in PBS at room temperature for 20 min and washed twice with PBS. Rat anti-mouse prominin-1 (CD133) antibody (Chemicon International, Temecula, CA, USA; final concentration, 5 μg/ml) was added to 5.5 × 105 cells in 25 μl 5% fetal bovine serum (FBS) in PBS. Samples were incubated for 20 min at room temperature, and unbound antibody was removed by washing cells twice with 5% FBS in PBS. The cells were resuspended in 25 μl 5% FBS in PBS and goat anti-IgG fluorescein isothiocyanate (1:200). The samples were incubated for 10 min at room temperature in the dark, and the cells were then washed once with 5% FBS and once with PBS and then resuspended in 200 μl PBS. Cell fluorescence was analyzed with a FACScan flow cytometer (Becton Dickinson) in the University of Massachusetts Medical School Flow Cytometry Core Lab, and the resulting data were analyzed with FlowJo version 7.2.2 software (Tree Star, Inc., Ashland, OR, USA).
Assays for migration, invasion, proliferation, and apoptosis were carried out as previously described.36 To assess colony formation in soft agar, 1,000 cells were cultured in 0.1 ml 0.35% Bacto Agar (BD Bioscience, Franklin Lake, NJ, USA) in a 96-well plate. After 2 weeks, colonies (>100 μm) were scored by light microscopy.
One potential problem was that retroviral insertion might alter cell properties. To meet this concern, we tested multiple independent cultures. In all cases, the results for independently derived cultures were essentially identical. Hence, changes in cell properties are due to EGFRvIII or loss of PTEN and not due to retroviral insertion.
For analysis of Chk1 phosphorylation, cells were irradiated with 5 Gy ionizing radiation from a 137Cs source at 0.87 Gy/min. Cells were harvested 2 h after irradiation for Western blotting. To determine radio-resistance, cells were irradiated with 2, 5, or 10 Gy ionizing radiation. Two days after irradiation, cell survival was assessed with a 3-[4,5-dimethylthiazole-2-yl]-2,5-diphenyltetrazolium bromide (MTT) assay.
Cytogenetic analysis of cell cultures was performed by spectral karyotyping (SKY),37,38 using cells after three to five passages in culture. SKY is based on the simultaneous hybridization of 21 chromosome-specific painting probes labeled with different fluorochrome combinations (Applied Spectral Imaging, Vista, CA, USA). The measurement of defined emission spectra using interferometer-based spectral imaging permits discernment of all mouse chromosomes in different colors. For cells showing apparent rearrangements, the experiment was repeated using only the SKY probes for the chromosomes included in the putative translocation, conclusively verifying the identities of the translocated chromosomes.
We subcutaneously injected (1 × 107 cells in 0.1 ml) immunodeficient mice (NIH-III [NIH-bg-nu-xidBR], 50-day-old females; Charles River Laboratories, Boston, MA, USA) on the back. Tumors about 1 cm in diameter were removed and divided into three pieces. One piece was used for preparation of paraffin blocks. The resulting sections (5 μm thick) were immunostained with the mouse monoclonals anti-glial fibrillary acidic protein (GFAP) antibody cocktail (BD Pharmingen, San Diego, CA, USA; 1:500), antinestin antibody (BD Pharmingen; 1:500), and antioligodendrocyte marker O4 antibody (Chemicon International; 1:50). In addition, we used rabbit antibodies from Cell Signaling: anti-P-MEK1/2 (ser217/221, 1:100) and anti-P-Akt (ser473, 1:50). The second tumor piece was used to establish PTEN−/− EGFRvIII tumor (PET) cell lines following the procedure for mouse striata. The third piece was frozen, wrapped in aluminum foil, and smashed with a hammer. The fragments were extracted and used for Western blotting.
For intracranial tumors, NIH-III mice (7 weeks old) were anesthetized with i.p. ketamine/xylazine and stereo-tactically injected with 106 cells in 5 μl of DMEM/F12 medium using a Hamilton syringe (Hamilton, Bonaduz Ab, Bonaduz, Switzerland) mounted on a Kopf model 900 stereotaxic apparatus (Kopf Instruments, Tujunga, CA, USA). The injections were in the right corpus striatum (lateral distance relative to the bregma, 2.3 mm; anterior distance, 0.7 mm). The depth of the injection was 3.0 mm from the dura. To avoid backflow, the injection was performed over 5 min. The needle was left in place for an additional 5 min before being slowly withdrawn. Six weeks after injection, the mice were perfused with 4% paraformaldehyde. All animal studies were performed in accordance with the Animal Care and Use Rules at the University of Massachusetts Medical School.
The Tukey’s multiple comparison test and the Student’s t-test were used to analyze data sets. Results with p < 0.05 were considered significant.
We crossed nestin-Cre+/+ PTEN+/lox mice with themselves to produce PTEN−/− neural precursor cells.39 Deletion of the PTEN gene was efficient as judged by PCR of genomic DNA (Fig. 1A) and the absence of anti-PTEN immunostaining (data not shown). These cells were infected with control or EGFRvIII retrovirus. Although both retroviruses bear GFP, we refer to the control as the GFP retrovirus. Infected cells were selected for GFP expression by flow cytometry, and the cultures grew with neurosphere morphology and expressed the stem cell marker nestin (data not shown). By Western blotting, EGFRvIII was expressed at levels comparable to endogenous EGFR (data not shown). In addition, treatment of neural precursor cells with 10% fetal bovine serum converts virtually 100% of wild-type neural precursor cells to GFAP-positive astrocytic cells, and treatment with basic fibroblast growth factor induces neuronal differentiation.40 GFP PTEN+/+, EGFRvIII PTEN+/+, GFP PTEN−/−, and EGFRvIII PTEN−/− cells differentiate along both astrocytic and neuronal lineages (data not shown). Hence, this procedure efficiently produces EGFRvIII PTEN−/− neural precursor cells.
We injected immunodeficient mice subcutaneously with GFP PTEN+/+, EGFRvIII PTEN+/+, GFP PTEN−/−, and EGFRvIII PTEN−/− cells. For the EGFRvIII PTEN−/− cells, tumors were detected 10–14 weeks after injection (seven of seven; Fig. 1B, C). For the other cell groups, no tumors were detected even after 21 weeks. As judged by Western blotting, these tumors strongly expressed EGFRvIII (Fig. 1D). The tumors showed areas of necrosis (Fig. 1E), which are often found in glioblastomas.41 These vascular tumors contained undifferentiated cells at high cell density with numerous mitoses. Many cells also had spindle-shaped cytoplasmic projections (Fig. 1F). All of these tumors (seven of seven) were positive for GFAP and nestin (Fig. 1G, H) and negative for the neuronal markers NeuN and synaptophysin and the oligodendroglial marker O4 (data not shown). Hence, EGFRvIII PTEN−/− cells formed tumors that expressed markers associated with glioblastomas.
To better understand how these two mutations transform neural precursor cells, we assessed their effects on cancer-related properties. We characterized proliferation of the precursor cells by reduction of MTT. After 4 days in complete defined medium and medium lacking EGF, there were 2.6- and 5.4-fold more EGFRvIII PTEN−/− cells than GFP PTEN+/+ cells (Fig. 2A). Both EGFRvIII PTEN−/− and GFP PTEN−/− cells grew in medium lacking insulin (Fig. 2B). These results demonstrate that EGFRvIII and PTEN loss make precursor cells less dependent on EGF and insulin.
Cell migration and invasiveness were measured by filter penetration assays. For migration measurements, the filter was coated with gelatin to increase adhesion and prevent loss of cells from the filter. For invasion assays, the filter was coated with Matrigel to clog the pores. EGFRvIII and PTEN loss increased cell migration and invasiveness. Migration was increased 4.8-fold for EGFRvIII PTEN−/− cells and 2.3-fold for GFP PTEN−/− cells compared with GFP PTEN+/+ cells (Fig. 2C). Invasiveness was increased by 5.6-fold for EGFRvIII PTEN−/− cells and 2.9-fold for GFP PTEN−/− cells (Fig. 2D).
Because PTEN haploinsufficiency reduced H2O2-induced apoptosis of neural precursor cells,36 we treated cells with H2O2 and measured apoptosis by TUNEL (terminal deoxynucleotidyl transferase dUTP nick end labeling) staining. PTEN loss, but not EGFRvIII, increased resistance to H 2O2 (Fig. 2E). However, EGFRvIII PTEN−/− cells were substantially more resistant. The H2O2 concentrations for which 50% of cells were apoptotic were, for GFP PTEN+/+, 0.05 mM; EGFRvIII PTEN+/+, 0.06 mM; GFP PTEN−/−, 0.1 mM; and EGFRvIII PTEN−/−, 0.3 mM. The combination of EGFRvIII and PTEN loss synergistically enhanced resistance to oxidative stress–induced apoptosis. We also irradiated these cells and assessed survival. PTEN loss or EGFRvIII expression made the cells more resistant, but EGFRvIII PTEN−/− cells were much more resistant (Fig. 2F). Treatment of EGFRvIII PTEN−/− cells with 10 Gy ionizing radiation decreased the cell number by 16% ± 2% compared with EGFRvIII PTEN+/+ cells (38% ± 1%), GFP PTEN−/− cells (40% ± 1%), and GFP PTEN+/+ cells (55% ± 1%).
We carried out two in vitro assays relevant to self-renewal. Cells were plated at low density (1,000 cells/ ml).42 The number of neurospheres that form is a measure of the number of self-renewing stem cells. EGFRvIII had little effect on sphere formation (GFP PTEN+/+, 11 ± 1 neurospheres/ml; EGFRvIII PTEN+/+, 15 ± 1; Fig. 3A). GFP PTEN−/− cells showed increased sphere formation (50 ± 5 neurospheres/ml), but EGFRvIII PTEN−/− cells showed much greater sphere formation (112 ± 6 neurospheres/ml). We also assayed these cells for colony formation in soft agar. EGFRvIII or PTEN loss only slightly increased colony formation (GFP PTEN+/+, 1 ± 1 colonies; EGFRvIII PTEN+/+, 6 ± 1 colonies; GFP PTEN−/−, 5 ± 1 colonies; Fig. 3B). However, EGFRvIII PTEN−/− cells showed enhanced colony formation (47 ± 5 colonies). As judged by flow cytometry, expression of the stem cell marker CD133 was increased for EGFRvIII PTEN−/− cells (6% CD133+ cells) compared with GFP PTEN+/+ cells (1.2%), EGFRvIII PTEN+/+ cells (1.3%), and GFP PTEN−/− cells (2.0%; Fig. 3C). For neurosphere formation, colony formation in soft agar, and CD133 expression, there was a synergy between EGFRvIII expression and PTEN loss.
Because PTEN haploinsufficiency increased phosphorylation of the downstream effector Akt,36 and PTEN loss correlated with Akt phosphorylation in glioblastomas,25 we measured this parameter. Phosphorylation of Akt was increased by 2.0-fold in GFP PTEN−/− cells and by 3.2-fold in EGFRvIII PTEN−/− cells (Fig. 4A, B). In contrast, EGFRvIII in PTEN+/+ cells induced a modest increase in Akt phosphorylation.
In glioblastomas, RAS/ERK activation correlates with cell proliferation and decreased patient survival.43 Furthermore, glioblastomas frequently show activated JNK.13,44 We assayed activation of these pathways using phospho-specific antibodies (Fig. 4A, B). Phosphorylation of the MEK1/2 kinase in the RAS/ERK pathway was increased by 1.4-fold in EGFRvIII PTEN+/+ cultures and 2.5-fold in GFP PTEN−/− cultures. However, EGFRvIII PTEN−/− cells showed greater activation (5-fold). Phosphorylation of JNK showed modest increases for EGFRvIII PTEN+/+ and EGFRvIII PTEN−/− (1.9-fold; Fig. 4A, B).
Six tumors formed by EGFRvIII PTEN−/− cells in mice were sectioned and stained with anti-P-Akt and anti-P-MEK1/2 antibodies. All six tumors showed activation of the Akt and RAS/ERK pathways (Fig. 1I, J).
As previously reported,45 loss of PTEN can lead to activation of Akt and phosphorylation of Chk1 on ser280. This modification suppresses responses of Chk1 to ionizing radiation. Phosphorylation of Chk1 on ser280 was increased in GFP PTEN−/− (1.8-fold) and EGFRvIII PTEN−/− cells (2.0-fold; Fig. 5A). In the absence of ionizing radiation, phosphorylation of ser345 on Chk1 was not detected. Ionizing radiation induced phosphorylation of Chk1 on ser345 for GFP PTEN+/+ and EGFRvIII PTEN+/+ cells but was greatly reduced for GFP PTEN−/− and EGFRvIII PTEN−/− cells.
We next assessed phosphorylation of γ-H2AX, which is a marker for double-strand DNA breaks. Ionizing radiation induced γ-H2AX expression for all four cell groups (Fig. 5B). However, GFP PTEN−/− and EGFRvIII PTEN−/− cells showed the γ-H2AX band without ionizing radiation, and EGFRvIII enhanced γ-H2AX levels with and without ionizing radiation. Because γ-H2AX is closely associated with double-strand DNA breaks, we conclude that PTEN loss leads to DNA damage, as reported for other cell types,45,46 and that EGFRvIII enhances the effect of PTEN loss.
We next checked for centrosome amplification, which occurs in many tumors.47 The EGFRvIII PTEN−/− cells showed centrosome amplification (12% ± 3% of cells), whereas GFP PTEN+/+, EGFRvIII PTEN−/−, and GFP PTEN−/− cells did not (Fig. 6A–C). The micrographs show cells immunostained with anti-pericentrin. Similar results were obtained using anti-γ-tubulin antibody (data not shown). We also measured levels of Aurora A and B, which when overexpressed can cause centrosome amplification.48 In EGFRvIII PTEN−/− cells, Aurora A and B levels were increased 2.5- and 1.7-fold, respectively (Fig. 6D).
We injected cells from one EGFRvIII PTEN−/− culture into immunodeficient mice, dispersed the cells from six tumors from six different mice, and placed them into culture. These cultures grew with neurosphere morphology (data not shown), and we named them PET1–PET6 (PTEN−/− EGFRvIII tumors 1–6). We subcutaneously injected the parental EGFRvIII PTEN−/− PET1–PET6 cells into immunodeficient mice. Tumors were detected for EGFRvIII PTEN−/− PET1–PET6 cells after 89 ± 18, 48 ± 8, 10 ± 1, 60 ± 10, 50 ± 7, 11 ± 1, and 10 ± 1 days, respectively. All of the PET lines showed shorter tumor latencies than EGFRvIII PTEN−/− cells. PET2, PET5, and PET6 showed shorter latencies than PET1, PET3, and PET4 (Fig. 7A). We also assessed CD133 expression by flow cytometry. The percentages of CD133+ cells for PET1 (6.3%), PET3 (6.1%), and PET4 (6.8%) were similar to that of EGFRvIII PTEN−/− cells (6.0%). However, the percentages of CD133+ cells were increased for PET2 (16%), PET5 (12%), and PET6 (14%). Hence, the lines with the greatest potency for tumor formation also had the highest percentage of CD133+ cells. These results indicate that EGFRvIII PTEN−/− cells underwent tumor progression to more aggressive PET2, PET5, and PET6 cells.
To further analyze tumor progression, we determined the karyotypes of our original four cultures and six PET lines. The GFP PTEN+/+, EGFRvIII PTEN+/+, and GFP PTEN−/− cells had essentially normal karyotypes (data not shown). EGFRvIII PTEN−/− cells had multiple chromosomal aberrations (Fig. 8). Additional chromosomal aberrations were noted for the PET cells. The PET1 and PET5 lines were hypertriploid and triploid, respectively. PET2 was near diploid, with numerous chromosomal aberrations. Chromosomes 15 and 17 were derived from three different chromosomes. A striking observation is that the EGFRvIII PTEN−/− line and five of six PET lines had chromosome 15 translocations. These results led us to test two additional, independently derived EGFRvIII PTEN−/− lines. Each of these lines had a chromosome 15 translocation [der(11)t(11;15) and der(15)t(15;11); data not shown].
We pooled karyotype data for three EGFRvIII PTEN−/− lines and six PET lines. Chromosome 15 was rearranged in far more cell lines than any other chromosome (Fig. 7B). Currently, we are mapping the chromosome 15 break points so that we can elucidate the functional consequences of these rearrangements.
We next assessed whether PET cells can form tumors in mouse brain and whether these tumors are diffuse. We made intracranial injections of PET2 cells into 10 immunodeficient mice. After 4 weeks, two of these mice became less active and began to lose weight. These mice were sacrificed after 6 weeks and found to have brain tumors (Fig. 9A). The tumors were highly cellular and showed focal invasion of normal parenchyma, including white matter tracts (Fig. 9B, C). There also was apparent infiltration along blood vessels (data not shown). These tumors were positive for both GFAP and nestin (data not shown). Both of these mice also had intraventricular tumors (Fig. 9A, D). We did not observe areas of necrosis in these tumors.
Because some glioblastomas show both EGFRvIII expression and PTEN loss,21–23,29 we tested whether these mutations would transform neural precursor cells. We found that these two lesions, but not the single lesions, transformed the cells and induced tumors. Furthermore, during tumor formation, the cells became more aggressive, as reflected by more rapid formation of tumors, a greater percentage of CD133+ cells, and additional chromosomal aberrations. The vascular hypercellular tumors formed in this model contained undifferentiated spindle-shaped cells expressing GFAP and nestin but not neuronal (NeuN and synaptophysin) or oligodendroglial (O4) markers. These cells showed nuclear pleomorphism and had high mitotic activity. In this orthotopic transplantation model, there was infiltration of ventricles, normal parenchyma, and white matter tracts. There also appeared to be infiltration along blood vessels. Subcutaneous, but not intracranial, tumors included areas of coagulative necrosis. This result may be due to the smaller size of the intracranial tumors, as suggested for tumors derived from human cancer stem cells.49 Hence, EGFR activation and PTEN loss act together to transform cells to form tumors that resemble human glioblastomas.
How do these two genetic lesions alter cell signaling and thereby regulate cell proliferation, apoptosis, and chromosomal stability? Although EGFRvIII activates Akt kinase in fibroblasts,15 we observed only a slight activation in the neural precursor cells (Fig. 4A). Another study reported that EGFRvIII activation of PI3K varied depending on cell type.50 In our study, the combination of EGFRvIII expression and PTEN loss robustly activated Akt (Fig. 4A, B). We also observed activation of the RAS/ERK pathway in EGFRvIII PTEN−/− cells (Fig. 4A, B). By DNA sequencing, we did not detect activating mutations of RAS genes (Candace Gilbert, unpublished results). The synergistic activation of the RAS/ERK pathway by EGFRvIII expression and PTEN loss may be due to phosphorylation of MEK by 3-phosphoinositide-dependent kinase 1 (PDK1).51
Do EGFRvIII expression and PTEN loss affect properties commonly associated with cell transformation? EGFRvIII PTEN−/− cells had increased proliferation, even with suboptimum levels of growth factors, and showed enhanced migration and invasiveness. Activation of the RAS/ERK and PI3K/Akt pathways enhances cell proliferation and invasiveness.36,43,52–55 EGFRvIII PTEN−/− cells were resistant to H2O2 and ionizing radiation. EGFRvIII expression and PTEN loss acted synergistically to increase the percentage of CD133+ cells, neurosphere formation, and colony formation. Aggressive brain tumors in the clinic also show increased CD133 expression and self-renewal assessed by neurosphere formation.56 Activation of the PI3K/Akt pathway enhances self-renewal of fetal neural precursor cells57 and embryonic stem (ES) cells.58 Activation of the RAS/ ERK pathway enhances self-renewal of ES cells.59
Even in the absence of environmental insults, EGFRvIII PTEN−/− cells showed DNA damage, which likely occurs by several mechanisms. Consistent with an earlier study,45 we found that loss of PTEN led to phosphorylation of Chk1 on ser280, a site phosphorylated by Akt kinase. Phosphorylation on ser280 prevents phosphorylation on ser345 of Chk1 in response to ionizing irradiation. EGFRvIII did not affect these processes. For GFP PTEN−/− and EGFRvIII PTEN−/− cells, we detected γ-H2AX even in the absence of ionizing radiation. Some human tumors also show γ-H2AX staining in the absence of ionizing radiation.60,61 The simplest model to explain our observation is that PTEN loss leads to phosphorylation and inhibition of Chk1.45 Loss of functional Chk1 can lead to induction of γ-H2AX and chromosomal instability in the absence of ionizing radiation,62–65 likely reflecting the role of Chk1 in normal cell division.66,67 In addition, loss of PTEN from the centromeres may contribute to these rearrangements.46 We also observed centrosome amplification for EGFRvIII PTEN−/− cells. EGFRvIII expression or PTEN loss did not lead to centrosome amplification. We can offer several possible explanations. First, levels of Aurora A and B were elevated and may contribute to centrosome amplification.48 Constitutive activation of the RAS/ERK pathway may lead to upregulation of Aurora A.68,69 Second, inhibition of Chk1 function has been suggested as a cause of aneuploidy in PTEN-negative tumors.45,70
Additional changes occurred during tumor formation in mice. All of the PET lines have decreased latent periods for tumor formation. The PET2, PET5, and PET6 cells had especially short latent periods for tumor induction as well as increased percentages of CD133+ cells. In addition, each of the PET lines has a unique karyotype. We propose a model to explain this genomic instability. The genome of EGFRvIII PTEN−/− cells is inherently unstable. The double-strand breaks indicated by γ-H2AX staining for EGFRvIII PTEN−/− cells may facilitate chromosomal translocations.71 In addition, the propensity of these cells to undergo centrosome amplification may lead to aneuploidy. Because these cells are resistant to apoptosis, an increased number of cells undergoing DNA damage survive, allowing new clones with novel karyotypes to emerge.
The variety of karyotypes for the EGFRvIII PTEN−/− cells and the PET lines suggests that multiple chromosomal rearrangements contribute to tumor evolution. However, one change was common to many of these lines: eight of nine EGFRvIII PTEN−/− lines had a translocation involving chromosome 15. The recurrent chromosome 15 translocations indicate that there is an important oncogene or tumor suppressor near these translocations. Identification of this gene is an important topic for future studies.
We thank Jonas Ekstrand for EGFRvIII coding sequence, Marie-Claire Daou and Taichang Jang for technical help, Yu Liu and S.Y. Wang for preparation of tumor sections, Ruibao Ren for retroviral vector and Dan Ory for packaging cell line, Candace Gilbert for RAS gene analysis, Ramon Parsons for anti-phospho-Chk1 antibody, David Weaver for help with intracranial injections, and Steve Doxsey, Jack Rosa, Lucio Castillo, Steve Jones, Tim Kowalik, and Tom Smith for helpful discussions. The University of Massachusetts Flow Cytometry Center sorted GFP-labeled cells. This work was supported in part by NIH NS21716, the Intramural Program of the Human Genome Research Institute, the Goldhirsh Foundation, and the Worcester Foundation.