The human glioblastoma cell lines T98, U251MG, U251sp, U87, SKMG1, U251nu/nu and AO2 and the human embryonic kidney 293T (HEK293T) cell line were maintained in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% FBS (Equitech-Bio, Kerrville, TX, USA) at 37°C in a humidified atmosphere of 5% CO₂. U251sp and U251nu/nu cells are derived from U251MG cells (18). Anti-E-cadherin, anti-N-cadherin and anti-vimentin antibodies were obtained from BD Biosciences (San Jose, CA, USA) and the anti-β-actin antibody was from Sigma-Aldrich (St. Louis, MO, USA). The anti-HES1 antibody was purchased from Cell Signaling Technology, Inc. (Danvers, MA, USA).

The sequences of the siRNAs specific for PRRX1 consisted of 5′-GCACUAAAGUCUACAUCUA-3′ (siPRRX-1) and 5′-CCACUGUUCUUAUCUCUAU-3′ (siPRRX-2). The sequence of the control siRNA that targeted luciferase was 5′-CUUACGCUGAGUACUUCGATT-3′. siRNAs were obtained from Sigma-Aldrich. Cells were transfected with 20 nM of siRNA using Lipofectamine RNAiMAX (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s instructions.

RNA was extracted from glioma samples and cells using the RNeasy Mini kit (Qiagen, Venlo, The Netherlands), and cDNA was generated using PrimeScript reverse transcriptase (Takara, Tokyo, Japan). Glioma samples were obtained from patients at Nagoya University Hospital with informed consent. Normal human brain RNAs were obtained from Takara and BioChain (Newark, CA, USA). Catalog nos. of the normal samples are 636530 (Takara) and R1234035-50 (BioChain). PCR was performed using the SYBR Premix Ex Taq™ II, and the Thermal Cycler Dice™ Real-time System TP800 (both from Takara) was used for analysis. The relative mRNA expression levels were normalized to GAPDH. The sequences of primers used to amplify each gene were: 5′-AGGTGGAGGAGTGGGTGTCGCTG TT-3′ and 5′-CCGGGA A ACTGTGGCGTGATGG-3′ (GAPDH); 5′-AGTTCCGCAGGAATGAGAGA-3′ and 5′-AT GGCGCTTTTCAGTGTCTT-3′ (PRRX1A); 5′-CATCGTACC TCGTCCTGCTC-3′ and 5′-GCCCCTCGTGTAAACAAC AT-3′ (PRRX1B); and 5′-AGGCGGACATTCTGGAAATG-3′ and 5′-TCGTTCATGCACTCGCTGA-3′ (HES1).

To generate T98 and U251MG cells that constitutively expressed GFP, GFP-PRRX1A and GFP-PRRX1B, 293T cells were transfected with a pQCXIP vector (Clontech Laboratories, Inc., Mountain View, CA, USA) encoding each gene as well as the pVPack-GP and pVPack-Ampho vectors (Stratagene, Tokyo, Japan). The culture supernatant was collected 48 h later and applied to T98 or U251MG cells with 2 μg/ml of Polybrene (Sigma-Aldrich). The cells were cultured for 24 h, and then 1 μg/ml of puromycin (Sigma-Aldrich) was added to select for infected cells. T98 and U251MG cells that constitutively expressed dominant-negative RBPJ (DN-RBPJ) with PRRX1 were produced by infecting PRRX1-expressing cells with recombinant retrovirus that encoded DN-RBPJ. The infected cells were selected with neomycin for 5 days. To generate U87 and U251sp cells that constitutively expressed the shRNAs, oligonucleotides encoding shRNAs specific for human PRRX1 and luciferase were cloned into the pSIREN-RetroQ vector (Clontech Laboratories, Inc.). The sequences of the shRNAs were: 5′-GCTTGAAGCTACAGATTAT-3′ (shPRRX1), 5′-CCACTG TTCTTATCTCTAT-3′ (shPRRX2) and 5′-CTTACGCTGAG TACTTCGA-3′ (control). Recombinant retrovirus was produced, and infected U87 and U251sp cells were selected with 1 μg/ml puromycin for 3 days.

To measure cell invasion using Boyden chambers, a filter was pre-coated with Matrigel, and 4.5×10⁴ cells (T98, U251sp, U87) or 1.2×10⁴ cells (U251MG) were seeded into the upper surface of the chamber. Twenty hours after seeding, the cells were fixed with 70% methanol and stained with 0.5% crystal violet. Cells that invaded the lower surface of the filters were counted in five randomly selected fields. Three independent experiments were performed. To evaluate cell invasion in the presence of the Notch inhibitor, DAPT, cells were treated with DMSO or DAPT (10 μM) for 12 h and then subjected to an invasion assay in the presence of DMSO or DAPT.

Cells were cultured in a 6-well, ultra-low attachment plate (Iwaki, Tokyo, Japan) with DMEM/F-12 medium supplemented with 10 ng/ml bFGF and 20 ng/ml EGF (both from PeproTech, Rocky Hill, NJ, USA). Two weeks later, the number of spheres that were >50 μm in diameter was counted in 10 different, randomly selected fields.

Cells were transfected with a reporter construct with pRTK-Luc to normalize to the transfection efficiency. Reporter constructs for Notch (19) and Hedgehog signaling (20) were obtained from RIKEN BioResource Center, and a construct for Wnt signaling (21) was obtained from Addgene (Addgene plasmid 12456). Twenty-four hours after transfection, the activities of firefly and Renilla lucif-erase were measured using the dual-luciferase reporter assay system (Promega, Madison, WI, USA). Luciferase activity was measured in triplicate, and three independent experiments were performed.

Animal experiments were conducted in accordance with the Faculty of Medicine of Nagoya University. U87 cells expressing control or PRRX1 shRNAs (3.5×10⁵ cells/5 μl) were stereotactically injected into the brains of Balb-c nu/nu nude mice (female, 5 weeks old) under anesthesia using a Hamilton syringe (Hamilton, Reno, NV, USA). The coordination was 1.4 mm posterior from the bregma, 3.0 mm to the right and at a 4.0-mm depth from the brain surface.

We first evaluated the mRNA levels of PRRX1 in multiple low-grade and high-grade glioma samples. Two alternative forms of PRRX1, PRRX1A and PRRX1B, exist; both isoforms have a homeobox domain in their central regions, yet they differ at their C-termini. PRRX1B has a so-called otp, aristaless and rax (OAR) domain at its C-terminus, whereas the C-terminal end of PRRX1A lacks this domain. Real-time PCR analysis revealed that the mRNA levels of both isoforms were increased in most of the low-grade and high-grade glioma samples when compared to the levels in the normal brain tissue (Fig. 1A). We next examined expression of PRRX1 in several glioblastoma cell lines. PRRX1A and PRRX1B were highly expressed in the U251sp and U87 cells compared with that in the other glioblastoma cell lines (Fig. 1B).

To examine the role of PRRX1 in glioblastoma cells, we used two different shRNAs that targeted both isoforms of PRRX1. U251sp and U87 cells that constitutively expressed the control (shCtrl) or PRRX1 shRNA (shPRRX-1 and shPRRX-2) were established by retrovirus infection. RT-PCR analysis confirmed a significant reduction in the PRRX1 mRNA in the shPRRX-1 and shPRRX-2 cells (Fig. 2A). Although a previous study reported that PRRX1 is associated with EMT, we did not observe any changes in the expression of EMT markers, such as E-cadherin and vimentin, following PRRX1 knockdown (Fig. 2B). We examined the invasion of PRRX1-knockdown cells using Matrigel-coated Boyden chambers. The invasion of U251sp and U87 cells was significantly reduced by PRRX1 knockdown (Fig. 2C). We performed in vitro neurosphere forming assays to assess the tumorigenicity of the PRRX1-depleted cells. Cells were cultured in suspension with EGF and FGF in the absence of serum for 2 weeks, and then the number of neurospheres was determined. As shown in Fig. 2D, neurosphere formation was suppressed by PRRX1 depletion. We then examined the effect of PRRX1 suppression on the proliferation of glioblastoma cells in vivo. Both shCtrl and shPRRX-1 U87 cells were implanted in the brains of mice by intracranial injection, and the survival of the mice was observed. Mice implanted with the PRRX1-knockdown cells exhibited a significantly longer survival when compared with the survival in the shCtrl cell-injected mice (Fig. 2E). These results indicate that PRRX1 is associated with invasion and proliferative properties of glioblastoma cells.

To further confirm the effects of PRRX1 expression on the invasion of glioblastoma cells, we expressed both isoforms of PRRX1 in T98 and U251MG cells, which showed lower expression of PRRX1 compared with U87 and U251sp cells. T98 and U251MG cells were infected with recombinant virus that encoded GFP-tagged PRRX1A or PRRX1B and selected with puromycin for 2 days. Most of the selected cells clearly expressed GFP-PRRX1, and both isoforms localized to the nucleus (Fig. 3A). We then examined the invasion of PRRX1-expressing cells using Matrigel-coated Boyden chambers. The expression of either isoform of PRRX1 promoted the invasion of T98 and U251MG cells (Fig. 3B). Neurosphere formation assays were also performed to assess tumorigenicity, yet neither GFP- nor GFP-PRRX1-expressing cells formed neurospheres (data not shown).

We next investigated the molecular mechanisms by which PRRX1 promotes the invasion of glioblastoma cells. A number of pathways, including the Notch, Wnt and Hedgehog pathways, are activated in glioblastoma cells and are known to be associated with the progression of glioblastoma (22–24). We tested whether these pathways were activated by PRRX1 expression using a reporter assay. GFP- or GFP-PRRX1-expressing T98 cells were transfected with a reporter plasmid encoding luciferase that could be activated by either pathway, and the luciferase activity was measured 24 h later. In this analysis, we found specific activation of the Notch pathway in PRRX1-expressing cells (Fig. 4A). To further confirm Notch activation by PRRX1 expression, we examined the level of HES1, which is a target gene of Notch signaling. Consistent with the activation of Notch signaling, increased levels of HES1 mRNA and protein by PRRX1 expression were observed (Fig. 4B and C). Conversely, depletion of PRRX1 expression in U251MG cells by siRNA transfection suppressed HES1 mRNA levels (Fig. 4D). We also tested whether suppression of PRRX1 affected Notch signaling in U87 and U251sp cells, but we did not observe any changes in Notch activation after PRRX1 knockdown (data not shown). It appears that PRRX1-mediated activation of Notch signaling is dependent upon the cellular context. We next examined whether PRRX1 expression was related with HES1 expression in the glioma samples. The expression levels of the PRRX1A and HES1 mRNAs were evaluated in glioma samples using RT-PCR analysis. As shown in Fig. 4E, we observed a correlation between PRRX1A and HES1 expression.

We tested whether Notch activation was responsible for the PRRX1-mediated promotion of cell invasion. To inhibit Notch activation, we used a chemical inhibitor, DAPT. Cells were incubated with DAPT for 12 h and subjected to the invasion assay in the presence of DAPT. Addition of DAPT clearly suppressed the invasion of PRRX1-expressing T98 cells (Fig. 5A). We also used a dominant negative form of the recombination signal binding protein for immunoglobulin kappa J region (RBPJ), which is a transcriptional regulator that is critical for the Notch signaling pathway (25). We established T98 and U251MG cells that constitutively expressed PRRX1 and a dominant-negative form of RBPJ (DN-RBPJ) and examined cell invasion. As shown in Fig. 5B, expression of the DN-RBPJ suppressed the invasion of PRRX1-expressing cells.