Neural differentiation of glioblastoma cell lines via a herpes simplex virus thymidine kinase/ganciclovir system driven by a glial fibrillary acidic protein promoter

Glioblastoma is a malignant brain tumor with poor prognosis that rapidly acquires resistance to available clinical treatments. The herpes simplex virus thymidine kinase/ganciclovir (HSVtk/GCV) system produces the selective elimination of HSVtk-positive cells and is a candidate for preclinical testing against glioblastoma via its ability to regulate proliferation and differentiation. Therefore, in this study, we aimed to establish a plasmid encoding the HSVtk/GCV system driven by a glial fibrillary acidic protein (GFAP) promoter and verify its possibility of neural differentiation of glioblastoma cell line under the GCV challenge. Four stable clones—N2A-pCMV-HSVtk, N2A-pGFAP-HSVtk, U251-pCMV-HSVtk, and U251-pGFAP-HSVtk—were established from neuronal N2A and glioblastoma U251 cell lines. In vitro GCV sensitivity was assessed by MTT assay for monitoring time- and dosage-dependent cytotoxicity. The capability for neural differentiation in stable glioblastoma clones during GCV treatment was assessed by performing immunocytochemistry for nestin, GFAP, and βIII-tubulin. Under GFAP promoter control, the U251 stable clone exhibited GCV sensitivity, while the neuronal N2A clones were nonreactive. During GCV treatment, cells underwent apoptosis on day 3 and dying cells were identified after day 5. Nestin was increasingly expressed in surviving cells, indicating that the population of neural stem-like cells was enriched. Lower levels of GFAP expression were detected in surviving cells. Furthermore, βIII-tubulin-positive neuron-like cells were identified after GCV treatment. This study established pGFAP-HSVtk-P2A-EGFP plasmids that successfully ablated GFAP-positive glioblastoma cells, but left neuronal N2A cells intact. These data suggest that the neural differentiation of glioblastoma cells can be promoted by treatment with the HSVtk/GCV system.


Introduction
Glioblastoma is a devastating, uniformly lethal primary brain tumor. The standard treatment for glioblastoma is surgery followed by radiotherapy and chemotherapy, but tumor recurrence is observed in most patients [1]. In fact, the median survival time for patients is only 12-15 months, despite undergoing the maximum treatment [2]. A contributor to the tumor reoccurrence is a cell population with cancer stem cell properties present in glioblastoma.
Previous studies revealed that malignant tumors are initiated by a population of tumor cells sharing similar biological properties with normal adult stem cells [3,4]. These tumor cells, known as cancer stem cells, have self-renewal phenotypes, which may be the cause of resistance to the current standard care of concomitant chemoradiotherapy [5]. Treatment with anticancer drugs and irradiation can cause cancer cells to die by apoptosis, but cancer stem cells can survive and are related to cancer recurrence [6,7]. Therefore, controlling the proliferation and differentiation of cancer stem-like cells is a key factor in controlling malignant cancer [4,8,9]. Inducing differentiation and apoptosis of cancer stem cells offers a promising strategy for the management and eradication of different types of cancers [10]. Similarly, regulating the differentiation of glioma cells has been suggested as a strategy for controlling glioblastoma [11]. Recent research revealed that glioma-initiating cells (GIC) isolated from human glioblastoma differentiated and increased their radiosensitivity when treated with rapamycin [12]. In addition, autophagy was suggested to play an essential role in the regulation of selfrenewal, differentiation, and tumorigenicity of GIC [13].
Ganciclovir (GCV) is an antiviral medication used to treat infections from the herpes family. GCV is first phosphorylated to GCV monophosphate by a viral kinase encoded by the cytomegalovirus (CMV) gene. Afterwards, GCV is metabolized to the triphosphate form by guanylate kinase and phosphoglycerate kinase. GCV-5 0 -triphosphate inhibits the replication of viral DNA. This inhibition includes a selective inhibition of the viral DNA polymerase and leads to disruption of DNA synthesis and inhibition of cell proliferation [14][15][16]. Since the herpes simplex virus thymidine kinase (HSVtk)/GCV system produces selective elimination of HSVtk-positive cells, it has been used in a clinical gene therapy trial as a therapeutic gene to kill cancer cells. Several tumor cell lines, including sarcoma, melanoma, and colon carcinoma, displayed sensitivity to this suicide gene approach [17]. This gene therapy has also shown potential as a treatment for cervical cancer [18] and gastrointestinal cancer [19]. Besides, neural stem/progenitor cells derived from human induced pluripotent stem cells (hiPSCs) with the HSV-TK/GCV suicide gene system showed prolonged survival time and the bystander killing in an orthotopic xenograft mouse model of glioblastoma upon GCV treatment. However, stable constitutive gene expression of HSV-TK was highly cytotoxic and difficult to sustain in hiPSCs and the recurrence occurred 2-3 weeks after GCV treatment [20].
As mentioned, HSVtk-positive cells can be selectively ablated by the HSVtk/GCV system. However, controlling HSVtk expression in mammalian cells is also essential to avoid any influence on other nontarget cell populations. For this purpose, the delivery of the HSVtk/GCV system using embryonic stem cells is a possible approach. Embryonic stem cells have the potential for use in cell replacement therapies, but the transplantation of differentiated cells harbors the risk of teratoma formation. This problem was overcome by establishing a negative selection system that permitted selective removal of undifferentiated stem cells [21]. The HSVtk gene under the control of the Oct4 promoter allowed the destruction of undifferentiated embryonic stem cells by GCV treatment and provided a pure population of completely differentiated cells [22]. The HSVtk/GCV system has been used to ablate undifferentiated cells, but the possibility that it could be used to ablate an unwanted differentiated lineage is still not clear.
Glial fibrillary acidic protein (GFAP), a type III intermediate filament protein, is expressed mainly in astroglia [23,24], and its expression is increased upon maturation of astrocytes [25]. The GFAP promoter is a 2.2-kb promoter region of the GFAP gene that drives astrocyte-specific expression and regulates the development of astrocytes and CNS physiology [26,27]. Previous study showed that the HSV-TK gene driven by human GFAP promoter could selectively kill glioma cell lines [28], but did not investigated the possible ability of neural differentiation under GCV treatment. Therefore, the aims of this study were to establish a recombinant plasmid vector bearing the HSVtk/GCV suicide system driven by a GFAP promoter to ablate GFAP-positive tumor cell lines selectively, and to characterize the possible neural differentiation of glioblastoma under GCV challenge.
To analyze the promoter expression level of pCMV-HSVtk-P2A-EGFP and pGFAP-HSVtk-P2A-EGFP between glial cell lines and neuron/fibroblast cell lines, the images of transfected cells were acquired under an inverted fluorescence microscope (DM IRM HC; Leica, Wetzlar, Germany). Cells that emitted strong green fluorescence were identified as successfully transfected with promoter expression. The promoter expression level was calculated as followed: Promoter expression level = (the number of green fluorescent cells / total cell number under phase image).
After transfection, 400 μg/ml G418 sulfate (Millipore, Cat. # 345810, Darmstadt, Germany) was added to the medium to select successfully transfected U251 cells. G418 sulfate selection lasted for 4 weeks and the cell population was also enriched by fluorescence-activated cell sorting (FACS; BD Bioscience, BD FACSAria™ IIIu, San Jose, CA, USA). Thereby, the U251-pCMV-HSVtk and U251-pGFAP-HSVtk stable clones were established. For N2A stable clone selection, 500 μg/ml G418 sulfate were applied to transfected N2A cells. After 4 weeks of G418 selection, surviving cells with green fluorescence of the N2A-pCMV-HSVtk cells were enriched by FACS and surviving cells of N2A-pGFAP-HSVtk cells were selected. Thus, the N2A-pCMV-HSVtk and N2A-pGFAP-HSVtk stable clones were established for use in further studies. After stable clones were established, G418 was still present for the maintenance of all stable clone cells, but not in the experimental cells.

Western blot analysis
For protein extraction, cells were trypsinized and homogenized with radioimmunoprecipitation assay buffer (RIPA buffer; Cell Signaling Technology, Cat. #9806, Danvers, MA, USA), 1 × phosphatase inhibitor (Roche Applied Science, Cat. #4906837001, Mannheim, Germany), and 1 × protease inhibitor (Roche Applied Science, Cat. # 4693116001). Cell lysates were centrifuged at 14,000 rpm for 30 minutes at 4˚C, and supernatant was collected. Protein samples with 1 × protein sample buffer were boiled at 97˚C for 10 minutes and then placed onto ice immediately. The prepared protein samples were stored at -20˚C for further analysis. Protein samples (40 μg/lane) were resolved in 10% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE), separated at 120 V for 2 hours, and transferred to polyvinylidene fluoride (PVDF) membranes (Millipore, Molsheim, France) at 100 V for 90 minutes. Subsequently, the PVDF membrane with transferred proteins was blocked with a blocking buffer (5% skim milk dissolved in 1 × Tris-buffered saline with Tween-20 [TBST]; 0.1% Tween-20 in 1 × TBS) for 1 hour at room temperature. After blocking, the PVDF membrane was incubated with primary antibodies (see S1 Table) diluted in a blocking buffer overnight at 4˚C. After washing, the PVDF membrane was incubated with horseradish peroxidase (HRP)-conjugated secondary antibodies diluted in 1 × TBST at room temperature for 1 hour. Detection of the protein signals was performed with enhanced chemiluminescence (MaestroGen, Hsinchu City, Taiwan) and the luminescence signal was detected using the UVP AutoChemi Image System. The western blot images were further quantified by ImageJ software (n = 3) and β-actin served for protein normalization in all groups.

In vitro GCV sensitivity assay
Stable clone cells (1.5 × 10 4 ) were seeded in 12-well cell culture plates. Cells were incubated for 12 hours for cell adhesion and treated with 0, 1, 5, or 10 μg/ml GCV for 5 days. After treatment, cells were placed in serum-free media containing 5 μg/ml MTT (Thiazolyl Blue Tetrazolium bromide; Sigma-Aldrich1, Cat. #M5655) and incubated at 37˚C for 1 hour. The purple crystals formed by the cells were dissolved in dimethyl sulfoxide (DMSO; Sigma-Aldrich1, Cat. #D8418). The level of cell viability was determined by measuring the absorbance at 550 nm with an ELISA reader (ELx808; BioTek, Winooski, VT, USA).
Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay. The TUNEL assay was performed with a cell death detection kit (Roche Applied Science, Cat. # 11684795910) to detect DNA fragmentation generated during apoptosis. For cell apoptosis analysis, 1.5 × 10 4 U251-pGFAP-HSVtk cells were treated with GCV for 3 days, as described. H 2 O 2 (10 μM) was added to induce apoptosis in the positive control groups. All images were acquired using a DMR fluorescence microscope (Leica) and a LSM880 confocal microscope (Carl Zeiss, Oberkochen, Germany).

Immunocytochemistry for nestin, GFAP, and βIII-tubulin
U251-pCMV-HSVtk or U251-pGFAP-HSVtk cells (1.5 × 10 4 ) were seeded onto poly-L-lysinecoated 18-mm round coverslips in 12-well culture plates. After 12 hours of cell-adhesion, cells were treated with 5 μg/ml of GCV for 1, 3, 5, and 7 days. Cells were fixed with 100% methanol and 4% paraformaldehyde and blocked with 1% of FBS in phosphate buffer saline with 0.1% Triton X-100 (PBS-T) for 1 hour. Thereafter, cells were incubated with primary antibodies (see S1 Table) in blocking buffer at 4˚C overnight. Cells were then incubated in fluorescentlabeled secondary antibodies (see S1 Table) for 1 hour at room temperature. Specimens were mounted with Fluoro-Gel (Electron Microscopy Sciences). Fluorescent images were acquired using a Carl Zeiss LSM880 confocal microscope. Photomicrographs were adjusted both in contract and brightness and merged using Photoshop software (Adobe, Mountain View, CA, USA, RRID: SCR_014199).
Survival rate and differentiation rate of U251-pGFAP-HSVtk and U251-pCMV-HSVtk were determined by quantifying five randomly selected fields of confocal images from each immunostaining. Survival rate were determined by calculating the number of Hoechst-positive cell nuclei on each treatment day compared with those on day 0. The ratio of nestin, GFAP, and βIII-tubulin positive cells in all viable cells (Hoechst-positive cells) were calculated as differentiation rates of U251-pGFAP-HSVtk and U251-pCMV-HSVtk.

Establishment and analysis of recovery of U251-pCMV-HSVtk and U251-pGFAP-HSVtk subclones
After treatment with 5 μg/ml of GCV for 7 days, U251-pCMV-HSVtk and U251-pGFAP-HSVtk cells were supplemented with fresh growth medium for 3 days. The remaining cell populations were regarded as recovery subclones. Subsequently, PI/Hoechst 33342 staining was performed to investigate this cell population. In addition, immunocytochemistry of βIII-tubulin, GFAP, and nestin were conducted to study the expression patterns of neural markers in the recovery cell subclones.

Statistical analysis
GraphPad Prism1 4.0 software (GraphPad Software, La Jolla, CA, USA, RRID: SCR_002798) was used to analyze the data. When comparing two conditions, a t-test was performed, with a p value < 0.05 regarded as statistically significant. All data were expressed as mean ± standard error of the mean (SEM).

Plasmid design and the specificity of the CMV promoter and the GFAP promoter in various cell lines
HSVtk and P2A were subcloned into pEGFP-N3, and named as pCMV-HSVtk-P2A-EGFP, to serve as a positive control vector in which the HSVtk and the reporter gene EGFP were separately expressed (S1A Fig). P2A, a self-cleavage peptide, played a role similar to an IRES, but was shorter with higher cleavage efficiency. Therefore, the HSVtk and EGFP sequences could be cleaved after translation and would become individual proteins with the assistance of P2A. Based on the design of pCMV-HSVtk-P2A-EGFP, the experimental plasmid, pGFAP-HSVtk-P2A-EGFP, was cloned (S1B Fig). The CMV promoter was replaced by the GFAP promoter, the DNA for which was amplified from a pAAV-GFAP-hChR2(H134R)-mCherry plasmid and cloned into AseI-XhoI restriction sites using an In-Fusion cloning kit.
To examine further the promoter specificity of pCMV-HSVtk-P2A-EGFP and pGFAP-HSVtk-P2A-EGFP in non-glial and glial cell lines, transient transfection into three cell lines-NIH-3T3 (mouse fibroblast cell line), N2A (mouse neuroblastoma cell line), and U251 (human glioblastoma cell line)-was conducted, and green fluorescent cells were analyzed on day 2 (Fig 1). A high proportion of cells containing green fluorescence were detected in all three cell lines transfected with pCMV-HSVtk-P2A-EGFP (Fig 1A, 1C and 1E).
However, non-glial (NIH-3T3 and N2A) and glial (U251) cell lines displayed different expression levels after transfection with pGFAP-HSVtk-P2A-EGFP. Very few EGFP-positive cells were detected when NIH-3T3 and N2A cells were transfected with pGFAP-HSVtk-P2A-EGFP (Fig 1B and 1D). By contrast, many U251 cells expressed green fluorescence when transfected with pGFAP-HSVtk-P2A-EGFP ( Fig 1F). All images were quantified by counting EGFP-positive cell numbers (Fig 1G). There was no significant difference in the number of EGFP-positive cells when pCMV-HSVtk-P2A-EGFP was transfected into the NIH-3T3, N2A, and U251 cell lines. However, after transfection with pGFAP-HSVtk-P2A-EGFP, the glioblastoma cell line, U251, had a higher proportion of green fluorescent cells (62.80±3.84%) than did NIH-3T3 and N2A cells (0.36±0.31% and 2.31±1.01% of EGFP-positive cells, respectively). These data on promoter specificity suggest that a large proportion of glia cells, but not neurons and fibroblasts, could be specifically targeted by a vector under the control of the GFAP promoter.

Establishment of the stable clones and characterization of their HSVtk expression levels
To produce stable clones that constitutively expressed HSVtk, 500 μg/ml and 400 μg/ml of G418 were applied to N2A and U251, respectively. During 14 days of selection, cells with a strong EGFP signal formed colonies, which were isolated and amplified for further FACS enrichment. Cells that expressed strong green fluorescence were sorted and further enriched by cell sorting and regarded as candidates for stable clones. There was no significant difference in the morphology or cell doubling time between parent cells and the transfected cells used in this selection. We established four stable clones-N2A-pCMV-HSVtk, N2A-pGFAP-HSVtk, U251-pCMV-HSVtk, and U251-pGFAP-HSVtk-for use in subsequent experiments, as described below.
To verify the protein expression of these stable clones, the levels of HSVtk protein expression were determined by western blot (

In vitro GCV sensitivity assays for stable clones treated for 5 days
To investigate the sensitivity of the stable clones to GCV and determine the optimal working concentration for all cell lines, the cell viabilities in different GCV concentrations were tested. All stable clones were treated with 0, 1, 5, or 10 μg/ml GCV for 5 days, and then an MTT assay was performed to evaluate cell viabilities. Live-cell phase images and immunocytochemistry were used to examine cell morphology and EGFP expression. Phase images collected using an inverted microscope revealed that N2A parent cells (Fig 2A-2D) and transfected N2A-pGFAP-HSVtk cells (Fig 2E-2H) remained intact and maintained more than 90% cell viability. N2A-pCMV-HSVtk was the only group in which the cell number decreased in a dosedependent manner (Fig 2I-2L). Furthermore, green fluorescence was still detectable in the surviving cells, even after treatment with 10 μg/ml GCV. Similar results were obtained from the MTT assay analysis, and N2A-pCMV-HSVtk was the only stable clone affected by GCV, with only 9.90±1.04% of the cells remaining after 10 μg/ml GCV treatment (Fig 2M). These results indicate that only N2A-pCMV-HSVtk cells were sensitive to GCV, since N2A parent cells and transfected N2A-pGFAP-HSVtk cells do not express HSVtk.

PLOS ONE
were sensitive to GCV. Furthermore, U251-pCMV-HSVtk cells, as a control for all the stable clones, reached LD 50 at 5 μg/ml GCV treatment. Therefore, the optimal dosage of GCV chosen for ablating stable clones in subsequent experiments was 5 μg/ml.
In summary, the experimental stable clone, U251-pGFAP-HSVtk, and the positive-control stable clone, U251-pCMV-HSVtk, displayed a dose-dependent response to GCV, while only N2A-pCMV-HSVtk was sensitive to GCV. N2A-pGFAP-HSVtk cells remained intact because the GFAP promoter would not be expressed in N2A neuronal cells. These data demonstrate that the GFAP promoter can specifically drive the HSVtk/GCV system and selectively ablate glioblastoma cell lines, while keeping neuronal cell lines intact.

Cell apoptosis assays and cell viability assay after GCV treatment
A previous study demonstrated that the HSVtk/GCV system triggered cells to undergo apoptosis rather than necrosis [29]. Therefore, we investigated whether the apoptotic pathway was activated in the cell lines treated with GCV and performed the TUNEL assay which can identify DNA nicks resulting from apoptotic signaling cascades (S3 Fig). Only cells that were both TUNEL-and Hoechst-positive, displayed condensed nuclei, and were magenta-colored in merged confocal images were considered apoptotic cells. In the GCV-treated group, apoptotic cell nuclei with TUNEL-and Hoechst-positive signals were identified (S3A- S3C Fig)

Pattern of nestin, GFAP, and βIII-tubulin expression in U251 stable clones after GCV treatment
Enlargement and process extension of surviving (or survival) cells, which are morphological features of cell differentiation, were observed in U251-pGFAP-HSVtk and U251-pCMV-HSVtk cells during GCV treatment. Therefore, immunocytochemical staining for nestin, GFAP, and βIII-tubulin was performed to characterize the identity of the survival cells in the U251-pGFAP-HSVtk (Fig 4) and U251-pCMV-HSVtk (Fig 5) groups. In addition, cells in recovery subclones were also investigated to mimic recurrence after drug withdrawal. Nestin, a stem cell marker, was used to identify neural stem cells. Only a trace of nestin signal was detected on day 1 (Fig 4B) of GCV treatment, but cells with filamentous nestin expression were detected from day 3 to day 7 (Fig 4C-4E) in U251-pGFAP-HSVtk cells. After 3 days of recovery, recovered U251-pGFAP-HSVtk cells had strong nestin expression (Fig 4F). These results indicate that cells with stem cell properties had more resistance to GCV and this population of neural stem-like cells was enriched after GCV withdrawn. The expression of GFAP in U251-pCMV-HSVtk cells gradually decreased during the 7 days of GCV treatment (Fig 4G-4K), suggesting that cells with strong GFAP expression were eliminated by GCV treatment. However, cells with weak GFAP expression might survive and resume proliferation after GCV withdrawal; as a result, more cells with strong GFAP expression were observed in the recovered cell population (Fig 4L). These results also confirmed that GFAP-positive glioblastoma cells could be ablated by the HSVtk/GCV system controlled by the GFAP promoter.
For identification of neuronal differentiation, the expression of βIII-tubulin, a neuronal marker, was investigated. Weak βIII-tubulin expression was detected by day 3 of GCV treatment (Fig 4O), and survival cells with strong βIII-tubulin expression in their processes were detected on days 5 and 7 and in the recovery cell population (Fig 4P-4R). According to DIC images, these βIII-tubulin-positive cells had elongated processes, rounded cell nuclei, and bulging cell bodies, while other cells were smooth (Fig 4P 0 -4R 0 ). This βIII-tubulin expression indicated that neuron-like cells could be identified after 5 days of GCV treatment, and these cells remained after recovery. In summary, U251-pGFAP-HSVtk cells had the ability to undergo neural differentiation under the stress of GCV, and the recovered cells were enriched with neural stem-like cells.
The survival rates and the expression ratio of the nestin-, GFAP-, and βIII-tubulin-positive cells in the U251-pGFAP-HSVtk ( Fig 6A) and U251-pCMV-HSVtk (Fig 6B) cell groups were also determined by quantifying confocal images of the respective immunocytochemistry. Survival rates were determined by calculating the number of Hoechst-positive cell nuclei on each day of treatment compared with day 0 (without GCV treatment). The ratio of the nestin-, GFAP-, and βIII-tubulin-positive cells to all viable cells (Hoechst-positive cells) were calculated in U251-pGFAP-HSVtk and U251-pCMV-HSVtk cell groups during the 7 days of GCV treatment ( Table 1). The survival rate of both stable clones gradually decreased during GCV treatment, but increased after GCV was withdrawn. Nestin-positive U251-pGFAP-HSVtk and U251-pCMV-HSVtk cells increased during the GCV treatment and even after GCV withdrawal. GFAP-positive U251-pGFAP-HSVtk and U251-pCMV-HSVtk cells decreased during GCV treatment and increased after 3 days of recovery. The βIII-tubulin-positive cells increased during GCV treatment and decreased after recovery. In summary, based on these results with U251-pGFAP-HSVtk and U251-pCMV-HSVtk cells, we suggest that neural stem-like cells with nestin expression are enriched after GCV treatment. Furthermore, a rapid decrease in the GFAP-positive cell number was observed during the 5 days of treatment. Finally, βIII-tubulinpositive neuron-like cells were detected after 3 days of GCV treatment.

Discussion
In this study, the GFAP promoter, an astrocyte-specific promoter, was used to restrict HSVtk gene expression in the glioblastoma cell line, U251. The specificity of the GFAP promoter in pGFAP-HSVtk-P2A-EGFP was confirmed by comparison with the control plasmid, pCMV-HSVtk-P2A-EGFP. An extremely low number of EGFP-positive cells was detected when NIH-3T3 and N2A cells were transfected with pGFAP-HSVtk-P2A-EGFP. A previous study demonstrated that a different GFAP promoter could be expressed in various kinds of cell lines, and the expression level of the GFAP promoter was 1.3% and 3.5% in N2A and NIH-3T3 cell lines, respectively [30]. In addition, gene regulation studies have indicated that the promoter and its distal elements could engage in looping interactions [31]. Long-range interactions between promoters and distal sites, including elements resembling enhancers, promoters, and CCCTC-binding factor (CTCF)-bound sites interacted with elements located~120 kb upstream of the transcription start sites [32]. Furthermore, mammalian cells possess a high activity of illegitimate recombination. When DNA is transfected into mammalian cells, illegitimate recombination is a predominant pathway of integration of transfected DNA into the host genome [33,34]. Thus, some other gene regulators that are located more than several hundred bp upstream from the GFAP promoter may exist, and may exhibit long-range interaction with the HSVtk gene, which was integrated into the host genome randomly. These issues may be the causes of the leaking problem found in NIH-3T3 and N2A cells.
In this research, the GFAP promoter was applied to control the expression of the HSVtk gene in glioblastoma cell lines, while the cytotoxicity of GCV was limited in neurons and fibroblasts. Our result could correspond to the previous study which showed that the HSV-TK gene driven by human GFAP promoter could selectively kill C6 and U251 glioma cell lines, with no  The survival rate and expression ratio of nestin, GFAP, and βIIItubulin in U251-pGFAP-HSVtk (A) and U251-pCMV-HSVtk (B) cells were determined by quantifying confocal images of immunocytochemistry (n = 3). The survival rate of both stable clones was decreased when cells were treated with GCV, but increased after recovery. The ratio of nestin-positive cells in both clones increased during GCV treatment and even after recovery. The ratio of GFAP-positive cells in both clones decreased during GCV treatment and increased after recovery. The ratio of βIII-tubulin-positive cells in both clones increased during GCV treatment and decreased after recovery. (C) The HSVtk/GCV system driven by the GFAP promotor could successfully ablate GFAP-positive glioblastoma cells while leaving non-glial cells intact. In addition, neural differentiation was promoted after challenge with this HSVtk/GCV system. effect on the ovarian cells [28]. Besides, the efficacy of HSV-TK gene-mediated cell killing was also reported in the primary cultures of human glioblastomas, which the majority of the tumor cells had detached or showed cytopathic effects after four-day GCV treatment [19]. These results suggested that HSVtk/GCV system could be applied to glioblastoma research. In addition, the HSVtk/GCV system was also driven by other promoters in previous studies. The CMV promoter is a constitutive mammalian promoter [35]. Bone marrow mesenchymal stem cells (MSCs) transduced with a recombinant baculovirus vector containing the HSVtk gene driven by the CMV promoter were reported to act as a systemic cancer therapy. The transduced MSCs were able to deliver the HSVtk suicide gene to glioblastoma and inhibited tumor growth on introduction of GCV [36]. Transcriptional targeting promoters were also used to mediate HSVtk expression in order to selectively target tumors [37]. Prostate-specific antigen (PSA) promoter was another tissue-specific promoter used in targeting prostate cancer [38]. Because telomerase is highly active in human cancer cells, the promoter regions of the telomerase RNA and catalytic component, human telomerase reverse transcriptase, have been used in many targeted cancer gene therapies to drive therapeutic gene expression [39]. Glucose-regulated protein 78 (Grp78) plays an essential role in cancer cell survival, so the Grp78 promoter along with HSVtk expression was applied to infect tumor cells for cancer suicide therapy [40]. In the present study, we confirmed that the GFAP promoter could specifically control HSVtk expression in glioblastoma cell lines while leaving neuronal cell lines intact. Moreover, we demonstrated that a glioblastoma cell line with a high level of GFAP expression could be eliminated on day 7 of GCV treatment.
Our results also demonstrated that the HSVtk/GCV system promoted neural differentiation of U251 cells and enriched the neural stem-like cell population after 7 days of GCV treatment. We demonstrated that cells with a high GFAP expression level were killed, while βIII-tubulinpositive neuron-like cells were identified after day 5 of GCV treatment. Previous studies have also revealed that some drugs can induce neural differentiation in other glioblastoma cell lines. The novel small molecule, CG500354, was reported to induce growth arrest of GBM-derived cells and neural differentiation via the cAMP/CREB signaling pathway. In contrast to our results, CG500354 treatment decreased the number of nestin-positive neural progenitor cells and increased the number of GFAP-and Tuj1-(βIII-tubulin)-positive neural cell types [41]. Furthermore, rapamycin, a therapeutic agent with both immunosuppressant and antitumor properties [42], reduced the proliferation of glioblastoma cells and increased the expression of neural differentiation markers (GFAP and βIII-tubulin) [12]. Another natural product, curcumin, has a potent anticancer effect on a variety of cancer cell types. Curcumin activated autophagy and triggered the differentiation cascade of human glioblastoma by producing higher levels of βIII-tubulin and GFAP expression [13]. Therefore, our finding that the HSVtk/GCV system driven by the GFAP promotor promoted the neural differentiation of U251 cells displayed a similar effect to other reports.
To test the HSVtk/GCV system on another genomic background, another human primary glioblastoma cell line, U87, was examined in our laboratory (unpublished data). The HSVtk/ GCV system driven by the GFAP promoter could ablate GFAP-positive glioblastoma cell lines and promote neural differentiation within 5 days of GCV treatment. A similar trend found in U87 cells indicated that this system represents a promising preclinical test for glioblastoma.
In this study, we showed that cell number of GFAP-positive cells would decrease during the treatment of GCV, and increase after 3 days of GCV withdraw. We demonstrated that GFAP promoter could specifically drive the HSVtk/GCV system and selectively ablate GFAP-positive cells. One reason for the increase of GFAP-positive cells after 3 days recovery may be due to the differentiation of the nestin-positive neural stem cell. Previous studies reported that cancer stem cells could be identified in U251 glioma cell line [43,44] and cancer stem cells represented a population of drug-resistant cells that could survive treatment and repopulate the tumor. Besides, side populations of glioblastoma cells (U251 and primary glioblastoma sample) were less sensitive to HSV-TK/GCV system than the non-side population [45], indicating that GFAP-positive cells may not be fully ablate under GCV treatment in our study. Therefore, the second reason for the increase of GFAP-positive cells after 3 days recovery may come from the proliferation of surviving GFAP-positive cells.
We observed that the nestin-positive neural stem cell increased after the GCV treatment and βIII-tubulin-positive neuron-like cells gradually increased after 5 days of GCV treatment and decreased after 3 days recovery. We suggested that the increase of βIII-tubulin-positive neuron-like cells during GCV treatment may be further differentiated from nestin-positive neural stem cell which increased from day 3 of GCV treatment. Because previous study suggested that cellular origin may affect lineage differentiation propensity of human induced pluripotent stem cells [46], nestin-positive neural stem cell may tend to differentiate into GFAPpositive cells instead of neurons after GCV was withdrawn. As a result, βIII-tubulin-positive neuron-like cells decreased after 3 days recovery. In this study, we did not investigate the maturation of βIII-tubulin-positive neuron-like cells, so further neuronal markers could be applied to identify the cell property in the future.
The HSVtk/GCV system can cause the death not only of HSVtk-positive cells but also of the surrounding HSVtk-negative tumor cells, which is called bystander effect and is another important feature of this system [47]. Bystander effect is critical for the clinical application because gene transfer currently available typically result in fewer than 10% of the cells expressing the transgene [48]. Previous study suggested that gap junction communication was the mediator of the bystander effect in primary cultures of human glioblastoma cells by transferring phosphorylated GCV from HSVtk-transfected cells to untransfected ones [49]. U251 glioblastoma cell line was shown to exhibit high gap junctional intercellular communication between neighboring cells (>80%) [50] and U251 cells expressing HSVtk gene could induce cytotoxicity in cocultured U251 bystander cells that lack the viral kinase when incubated with GCV [51]. In this study, the bystander effect has not been evaluated and further study could investigate the difference of the bystander effect between U251-pGFAP-HSVtk and U251-pCMV-HSVtk cells and the differentiation ability of the bystander tumor cells induced by bystander effect.
While the precise mechanisms underlying the neural differentiation are not known, some possible mechanisms are suggested by other studies. Autophagy is involved in differentiation, survival, and cell death, and the PI3K/AKT/mTOR pathway was reported to promote growth and proliferation over differentiation of adult stem cells, and neural stem cells, specifically [52]. Inactivation of the mTOR pathway was sufficient to induce autophagy following the administration of rapamycin [12]. Furthermore, curcumin promoted differentiation by inducing autophagy [13]. Therefore, it could be speculated that the HSVtk/GCV system promotes neural differentiation of glioblastoma, and this is mediated by the PI3K/AKT/mTOR-related pathway. However, evidence for the involvement of a specific signaling pathway should be further investigated in future studies.
In conclusion, this study established pGFAP-HSVtk-P2A-EGFP plasmids that could successfully ablate GFAP-positive glioblastoma cell lines, while neuronal N2A cells remained intact. The data also suggested that the neural differentiation of glioblastoma cells could be promoted after challenge with the HSVtk/GCV system.

Conclusions
This study constructed a pGFAP-HSVtk-P2A-EGFP plasmid bearing the HSVtk/GCV system controlled by the GFAP promoter to establish a strategy for cancer management in vitro. We demonstrated that this system could selectively ablate GFAP-positive glioblastoma cell lines, while it left the neuronal N2A cells intact and promoted neural differentiation after challenge with the HSVtk/GCV system ( Fig 6C). Therefore, we propose that this system has potential for use in the study of the neural differentiation of glioblastoma stem cells in vitro and in vivo and offers a possible preclinical test for glioblastoma. A 66 bps self-cleavage peptide P2A was also subcloned into pEGFP-N3 at EcoRI and SalI restriction enzyme cutting sites to obtain a construct named pCMV-HSVtk-P2A-EGFP. This construct could express HSVtk and EGFP as a reporter gene separately. Kanamycin/neomycin resistance gene was used for both kanamycin selection of E.coli colony and G418 sulfate selection of cell stable clone. (B) CMV promoter of pCMV-HSVtk-P2A-EGFP was substituted by 2207 bps GFAP promoter from another plasmid, pAAV-GFAP-hChR2(H134R)-mCherry, at restriction enzyme cutting sites of AseI and XhoI. HSVtk and EGFP in this clone were controlled by GFAP promoter, which is a specific promoter of glia cell. Kanamycin/neomycin resistance gene was used for both kanamycin selection of E.coli colony and G418 sulfate selection of stable clone in the further experiments. The survivals of U251-pGFAP-HSVtk (A-F) and U251-pCMV-HSVtk (G-L) cells were determined by PI/Hoechst co-staining for cell nuclei after 5 μg/ml GCV treatment for 7 days. PI/Hoechst colocalized and condensed cell nuclei were identified at the 5th day and the 7th day of GCV treatment in both U251-pGFAP-HSVtk (D and E, arrows) and U251-pCMV-HSVtk (J and K, arrows) groups. No PI signal was detected after 3 days of recovery (F and L). Scale bar = 50 μm. (TIF) S1 Table. List of antibodies applied in this research. (DOCX) S1 Raw images. (PDF)